technical committee 202 transportation · pdf filetechnical committee 202 transportation...

Technical Committee 202Transportation Geotechnics

Comité technique 202Géotechnique des transports

1213

General Report TC202 Transportation Geotechnics

Rapport général du TC202 Géotechnique pour les infrastructures de transport

Indraratna B. Centre for Geomechanics and Railway Engineering, Program Leader, ARC Centre of Excellence for Geotechnical Science and Engineering University of Wollongong, Wollongong City, NSW 2522, Australia

Correia A. School of Engineering, Univ. of Minho, DEC Campus de Azurém, 4800-058 Guimarães, Portugal

ABSTRACT: Today’s needs of urban transportation including roads, railways, airports and harbours demand significant resources forinfrastructure development in view of rapid and efficient public and commercial (freight) services. In most cases, authorities have haddifficulties in meeting these service demands due to the rapidly growing public, industrial, mining and agricultural sectors in manyparts of the world. In order to maximise efficiency and to reduce the costs of maintenance, sound technical knowledge is required.This general report presents major technical advancements around the globe encompassing 33 articles from 19 countries and it isclassified into 6 key categories, namely: compaction and subgrade improvement, laboratory testing, theoretical advancements andcontributions to design, applications of geosynthetics, numerical modelling and field performance evaluation.

KEYWORDS: urban transportation, compaction, design, geosynthetics.

RÉSUMÉ : De nos jours, les besoins en transports urbains (routes, chemins de fer, aéroports aériens et maritimes) nécessitent d’importantes ressources pour le développement des infrastructures en vue d’assurer des services commerciaux rapides et efficaces.Dans la plupart des cas, en raison de la croissance rapide des secteurs public, industriel, minier et agricole, les autorités se trouventconfrontées à des difficultés pour atteindre les services escomptés. Un savoir technique est alors nécessaire en vue de maximiserl’efficacité et de réduire le coût d’entretien. Le présent rapport général expose les avancées techniques majeures à travers le monde synthétisant 33 articles émanant de 19 pays ; six thèmes clés sont classés : compactage et amélioration des assises, expérimentation en laboratoire, développements théoriques et contributions au dimensionnement, applications des géosynthétiques, modélisationsnumériques et évaluation des performances sur le terrain.

Mots clés : Transports urbains, compactage, dimensionnement, géosynthétiques.

1 INTRODUCTION

Modes of transportation including roads, railways, airports and harbours demand the most essential infrastructure in industrialised countries. While almost 90% of the population depends on public transport every day in many developing nations, in large developed countries such as Australia, the mining and agricultural sectors almost entirely depend on the efficient roads and rail services. Geotechnical aspects of infrastructure design and construction play an important role and face key challenges in optimising the performance of road pavements, rail tracks, runways and ports, as well as their maintenance throughout their operational life cycle. Maximising efficiency with acceptable longevity and ensuring the minimum cost of maintenance requires sound technical knowhow, implementation of new and appropriate technologies and effective administration of strategic public policies and investments. In essence, the key phases of transport geotechnics can be primarily categorised to general soil and rock mechanics applications in design, development of theoretical concepts and analysis, construction innovations, challenges and in-situ quality control, ground improvement schemes including compaction and problematic soil remediation, field performance monitoring and data interpretation, laboratory techniques and physical modelling, numerical simulation for design and performance verification, constitutive modelling of pavement materials, conventional in-situ testing and non-destructive techniques for foundations, risk assessment and failure prediction, among other topics.

TC202 (Transportation) of ISSMGE fosters the relevant challenges through an active cohort of international membership, and through an array of meetings, workshops and conferences disseminates the new ideas, technical concepts and innovative technologies to the worldwide geotechnical community. The 18th ICSMGE (Paris) has attracted about 33 articles from 19 countries in the area of Transportation. These papers cover the entire domain of transportation, but for the purpose of this General Report, they can be predominantly classified under the following 6 categories: a) Compaction and subgrade improvement b) Laboratory testing c) Theoretical advancements and contributions to design d) Applications of geosynthetics e) Numerical modelling f) Field performance evaluation

2 COMPACTION AND SUBGRADE IMPROVEMENT FOR TRANSPORT INFRASTRUCTURE

There are 5 articles included in this section. Two papers discuss the salient aspects related to the application of novel techniques in field compaction (Adam et al. 2013, Kuo et al. 2013) while one article focuses on the site evaluation of the compaction quality using an array of different instruments (Conde et al. 2013). Useful practical information related to the application of ground improvement methods in a brown coal landfill site (Kirstein et al. 2013) and laboratory characterization of cement

1214

Proceedings of the 18th International Conference on Soil Mechanics and Geotechnical Engineering, Paris 2013

aggregates (Fonseca et al. 2013) applied to transport infrastructure are also presented and discussed.

Adam et al. (2013) introduced a finite element modeling framework for analyzing the performance and efficiency of an impact compactor in relation to the surface velocity, weight of impact compactor and number of passes. Field observations indicate that the impact compactor is suitable for treating a wide variety of loose soils and fills, but the effective treatment depth is dictated by the grain size, typically ranging from 4.5m to 10m depth. Experience of two case studies suggests that Dynamic Probing Tests (Figure 1) are adequate for evaluating the efficiency of compaction.

depth [m]

4.5 m 12.0 m

4.0

mgravel core(loess / loam)

compacted zone

N10

BEFORE RIC

AFTER RIC

depth [m]

4.5 m 12.0 m

4.0


compacted zone

N10

BEFORE RIC

AFTER RIC

4.5 m 12.0 m

4.0


compacted zone

4.5 m 12.0 m

4.0


compacted zone

N10

BEFORE RIC

AFTER RIC

Figure 1. Test dike and correspondent dynamic probing test results. (Source: Fig 8, Adam et al. 2013).

Kuo et al. (2013) have described the effectiveness of Rolling Dynamic Compaction (RDC) by the combination of field studies with numerical modeling (Figure 2). At the ground surface, there are noticeable large deformations, and RDC proves to be most effective between the depths of 0.8m and 3.0m. The preliminary parametric study showed that most significant factors were soil cohesion, Poisson’s ratio and shear modulus, as well as the width and mass of the RDC module.

An interesting study on the feasibility of a stiffness-based specification for embankment soil compaction quality control is discussed by Conde et al. (2013). An array of instruments are adopted for compaction control, which measures soil stiffness and then discussed on the basis of an earth dam construction. Among the different equipment used, the DCP (dynamic cone penetrometer) equipment showed greater promise as a compaction control tool, partly attributed to the strong negative correlation with water content values.

Figure 2. FEM model (Source: Fig 3, Kuo et al. 2013).

Kirstein et al. (2013) have described the application of a combination of ground improvement techniques to stabilize a recently placed brown coal landfill embankment for supporting a new road. Owing to significant stability problems and the small settlement tolerance of the structure (15 m deep), “floating” stone columns were also installed. The design and the associated settlements were significantly influenced by the combination of different soil improvement techniques. The settlement predictions were obtained using a finite element model (Figure 3) and successfully verified against the results of pressuremeter tests.

Figure 3. Representation of the predicted total settlements obtained with Plaxis (Source Fig 10, Kirstein et al. 2013).

Fonseca et al. (2013) presented some intriguing results obtained through laboratory studies performed on compacted mixtures of cement and limestone aggregates. The results indicated that the differences observed in dynamic and static stiffness properties and shear strength parameters were directly associated with the variation of porosity/cement ratio. As expected, a higher stiffness and strength were obtained by increasing the cement content and the degree of compaction. While a hardening soil model could be employed to adequately describe the observed stress-strain behaviour, the volumetric predictions and the post-peak strain softening response could not be reproduced satisfactorily.

3 LABORATORY TESTING

This section includes 6 articles. Two papers demonstrate the results of California Bearing Ratio (CBR) tests performed on the subbase (Ishikawa et al. 2013) and the subgrade (Moayed et al. 2013). Some studies focus on cyclic loading tests on ballast (Kumara and Hayano 2013) and subgrade (Mohanty and Chandra 2013), while the others investigate the overall performance of railway track (Calon et al. 2013, Hayano et al. 2013).

Ishikawa et al. (2013) examined the effects of freeze-thaw and water content on the deformation-strength properties of granular base materials. Two types of tests are conducted on these materials under various water contents. One test is based on the newly developed CBR equipment (Figure 4), and the other using medium-size triaxial apparatus. The freeze-thaw of granular base showed a strong influence on the fatigue life of pavement structures. When number of freeze-thaw process cycles increased, CBR values decreased regardless of the water content. Resilient modulus showed a decreasing tendency with the increasing water content.

150

Surcharge

Water supply/ drainage

Coolantcirculating line

Insulation

Acrylic cell

Base cooling plate

Temperaturesensor (pt100)

Porous metalwith filter paper

O-ring



Top cooling plate




Figure 4. Freeze-thawing CBR test apparatus. (Source: Fig 1, Ishikawa et al. 2013).

1215

Technical Committee 202 / Comité technique 202

Calon et al. (2013) have studied the potential benefits from the ground reinforcement by vertical soil-cement columns. Laboratory tests are performed to study the influence of the column location and the efficiency of geosynthetics on the reduction of stiff zones effects. These tests together with subsequent numerical modelling determined the optimum column layout (depth, spacing and positioning) and the effects of geosynthetics on the reduction of ballast damage.

Hayano et al. (2013) analysed the influence of ballast thickness and tie-tamper repair on the settlement of tracks by conducting a series of cyclic loading tests. Figure 5 shows the shear strain distribution generated before the tie-tamper implementation. This shear strain distribution is obtained using the method of particle image velocimetry. They found that the 250 mm ballast thickness currently adopted as the standard design is ineffective for minimizing settlement that occurs when the nonlinearity of roadbed compressibility is relatively moderate. Moreover, characteristics of the initial settlement process are altered significantly after the tie-tamper implementation, although the degree of gradual subsidence undergoes minimal change regardless of ballast thickness and roadbed type.

Figure 5. Distribution of maximum shear strain generated before tie-tamper implementation . (Source: Fig 5, Hayano et al. 2013).

Mohanty and Chandra (2013) have reported a series of cyclic load triaxial tests on reconstituted pond ash specimens at different moisture content and stress levels simulating environmental and traffic conditions. They concluded that both traffic and environmental conditions play an important role in the permanent axial strain behavior of the material. Furthermore, within the design context, they also highlighted the existence of a shakedown limit describing a critical stress level between stable and unstable conditions.

A series of CBR tests was conducted by Moayed et al. (2013) to investigate effects of lime-microsilica additive as a modern additive stabilizer on a silty soil to use it as a subgrade. They also evaluated the effects of the wetting-drying cycles. The CBR values were found to increase significantly as the soil was stabilized with lime-microsilica additive. An increase in the CBR values of the stabilized soil owing to wetting-drying cycles was also observed. Results showed that lime-microsilica additive can successfully be considered as a suitable option to stabilize silty soils.

Kumara and Hayano (2013) presented a series of cyclic loading models to investigate the effects of sand intrusion into ballast (i.e. fouling) and tie-tamping application on settlements of ballasted rail track. They found that the initial settlement process and the rate of residual settlement increases when the ballast is mixed with more than 30% fine materials. Therefore, tie-tamping application was found effective for fouled ballast with less than 30% fines.

4 THEORETICAL ADVANCEMENTS AND CONTRIBUTIONS TO DESIGN

A total of 7 articles are categorized in the area of theoretical advancements and contributions to design. There are 6 papers investigating the behavior of road embankments (Simic 2013, Ohta et al. 2013, Shin et al. 2013, Vogt et al. 2013, Eekelen and Bezuijen 2013, Brown and Thom 2013), while one article reports the development of a non-linear ballasted track model using the finite element technique (Fernandes et al. 2013).

Brown and Thom (2013) proposed a Precision Unbound Materials Analyser (simplified version of the repeated load triaxial test) to quantify both resilient and plastic strain characteristics (Figure 6). Unlike CBR testing, this technique can be very useful in allowing the designer to evaluate alternative foundation material combinations in order to achieve the desired bearing capacity.

Figure 6. The Precision Unbound Material Analyser (PUMA) (Source: Fig 1, Brown and Thom 2013).

Simic (2013) adopted the average suction compression index of the plate loading tests and the routine soil parameters to carry out a comparison between the methods of estimating swelling. It is found that the potential vertical rise method is overly dependent on the active moisture depth, which should be adopted in the design based on the local experience.

Ohta et al. (2013) proposed the structure of seismic retrofit technique for asphalt concrete pavements using the Confined-Reinforced Earth (CRE) principle. Construction method and the results of full scale in-situ tests are well-described where the crushed stones and the associated design procedures are clearly introduced. Full-scale in-situ tests show the acceptable performance of CRE after the forced settlement to simulate severe earthquake-induced damage.

Shin et al. (2013) determined the frost penetration depth of paved road using field measurements. They found that the subbase and base courses were influenced by the temperature below 0 regardless of the anti-frost layer. The frost penetration depth, estimated by the empirical equation proposed by Korea Institute of Construction Technology, shows a similar trend at lower frost index. This design concept is proposed for road design as an acceptable and reasonable approach.

Vogt et al. (2013) presented project-specific conditions during the dumping process, and the properties of the dumped soils along the future A-44 route. A simple model for the description of the time-dependent deformation of the dump and the effectiveness of soil compaction methods is discussed and evaluated. The simulation results and geodetic measurements reveal that by allowing a rest period of at least 6 months between the end of the dumping process and the start of the construction work, the settlements of structures and/or pavements can be reduced significantly.

Eekelen and Bezuijen (2013) compared three equilibrium models describing the phenomenon of arching in basal geosynthetically reinforced (GR) piled embankments, namely the models of Hewlett and Randolph (1988), Zaeske (2001) and the model of concentric arches by Van Eekelen (2013b). The load distributions predicted by Hewlett and Randolph (1988) and Zaeske (2001) show a uniform load distribution on the GR between the piles. The concentric arches model provides a load concentration on the GR strips, with an inverse triangular load

1216


distribution on those GR strips. This is in agreement with the observations in scaled-down model tests, numerical analysis and field measurements.

Fernandes et al. (2013) carried out a 2D finite element model with a modified width (plane strain) where viscous boundaries are implemented using a Kelvin-Voigt viscoelastic mechanical model to reduce the wave reflexion on the boundaries. The importance of initial state evolution of track materials on the context of non-linear mechanical behaviour is discussed to assure the correct combination of laboratory tests based on the current track conditions, especially for ballast (Figure 7).

Figure 7. Finite element mesh discretisation of the railway structure (Source: Fig 3, Fernandes et al. 2013).

5 APPLICATIONS OF GEOSYNTHETICS

In this section, 4 articles are described. The majority of articles discuss general issues of geosynthetic reinforcement (Indraratna et al. 2013, Wayne et al. 2013, Huckert et al. 2013), while other article examines the stiffness of the soil-geosynthetic intereaction under small displacement conditions (Zonberg et al. 2013).

Huckert et al. (2013) presented full-scale experiments to study load transfers of geosynthetics-reinforced embankments prone to sinkholes which are related to the complexity of various mechanisms. Numerical model coupling both finite and discrete element methods were performed and the results compared with the experimental data. These simulations provide a better understanding of load transfers towards the edges of the cavity.

Wayne et al. (2013) presented results of two field studies and model tests to evaluate performance of a geogrid-stabilized unpaved aggregate base overlaying relatively weak and non-uniform subgrade soils. Piezoelectric earth pressure cells (EPC) were used to measure horizontal stress below and above the geogrid location versus the passage of construction and truck traffic. The variation of dynamic horizontal stresses within the subgrade against the passage of truck traffic is presented in Figure 8. This result indicated an enhanced fully confined zone above the geogrid, resulting in an uniform vertical stress across the subgrade that leads to reduced lateral stresses.

25

20

15

10

5

050 100 150

Unstabilised section

Stabilised section

Roller/Test cumulative pass count

Stress (kPa)

Figure 8. Horizontal stress within the subgrade layer after roller compaction and test vehicle passes (Source: Fig 3, Wayne et al. 2013).

Figure 9 presents the horizontal stress in the base layer after roller compaction and trafficking. It is clearly seen that the geogrid confines the unbound aggregate leading to an increased

lateral stress within the aggregate. The results demonstrate that the inclusion of geogrid at the interface of soft subgrade and aggregate layers affects the development of the “locked-in” horizontal stress upon loading. A higher horizontal stress within the sttabilized aggregate layer gives a direct indication of the lateral restraint mechanism. The result of increased aggregate stresses leads to an increase in the resilient modulus of aggregate adjacent to the geogrid.

25

20

15

10

5

050 100 150


Stabilised section


Stress (kPa)

Figure 9. Horizontal stress within the base layer after roller compaction and test vehicle passes (Source: Fig 4, Wayne et al. 2013).

Indraratna et al. (2013) presented the results of full-scale field tests conducted on rail track sections in the towns of Bulli and Singleton (NSW, Australia) to measure track deformations associated with cyclic stresses and impact loads. The vertical and horizontal stresses induced in the track bed were recorded by pressure cells. Vertical deformations of the track were measured by settlement pegs, and lateral deformations were measured by electronic displacement transducers. The settlement pegs and displacement transducers were installed at the sleeper-ballast and ballast-subballast interfaces, respectively, as shown in Figure 10.

Figure 10. Installation of settlement pegs and displacement transducers at Bulli site (Source: Fig 3, Indraratna et al. 2013).

The average lateral deformations of ballast at various number of load cycles (N) are illustrated in Figure 11. It is shown that the geocomposite decreased the lateral deformation of fresh ballast by about 49% and that of recycled ballast by 11%. The capacity of the ballast to distribute loads was improved by the placement of the geocomposite, which substantially reduced settlement under high repeated loading. Indraratna et al. also discuss the results of large scale drop-weight impact testing equipment to evaluate the effect of using shock mats in mitigating ballast breakage. The ballast breakage was measured using the ballast breakage index (BBI) as shown in Table 1. Installing layers of synthetic materials such geogrids and rubber pads (shock mats) in rail tracks was found to significantly reduce ballast degradation.

Table 1. Ballast breakage under impact loading (Source: Table 1, Indraratna et al. 2013)

Base Test Details BBI

Stiff Without shock mat 0.170

Stiff Shock mat at top and bottom 0.091

Weak Without shock mat 0.080

Weak Shock mat at top and bottom 0.028

1217


Figure 11. In-situ response of the ballast layer: lateral deformations (Source: Fig 4, Indraratna et al. 2013).

Zornberg et al. (2013) introduced a mathematical model to investigate the soil-geosynthetic interaction behavior under small displacements. A new parameter, defined as ‘Stiffness of Soil-Geosynthetic Interaction’ (KSGI) is proposed to evaluate soil-geosynthetic interaction. This parameter is capable of quantifying the performance of geosynthetic reinforcement under small displacement conditions. KSGI was proposed on the assumption of a linear relationship between unit tension and strain in geosynthetic reinforcement and uniform soil-geosynthetic interface shear over the active length of the geosynthetic. Zornberg et al. (2013) conducted several geosynthetic pullout tests of biaxial geosynthetic with dimensions of 300 600 mm in clean poorly graded sand to validate the proposed model. Figure 12 shows a good agreement between the experimental data and the results obtained with the proposed model.

Figure 12. Results for the pullout test for LVDTs 2, 3, and 4 in (T2-u) space (Source: Fig 7, Zornberg et al. 2013).

6 NUMERICAL MODELLING

1.1 Finite element modelling (FEM)

There are 5 articles described in this section. The majority of the papers discusses Finite Element Modelling (FEM) on the stability analysis of soft clay subgrade and embankments (Carvajal and Romana 2013, Mansikkamaki and Lansivaara 2013, Islam et al. 2013 and Chirica et al. 2013), while one paper examines the application of a stochastic subsoil model on the deformation of bridge piers considering the soil heterogeneity (Jacobse et al. 2013).

Carvajal and Romana (2013) developed a FEM model of a multilayered system to investigate the influence of soft soil depth on pavement response during static and cyclic loading (Figure 13). They concluded that deep ground treatments should be applied to achieve an allowable capacity of soft soils up to a depth of 6 m to reduce the maintenance costs.

0 1x105 2x105 3x105 4x105 5x105 6x105 7x105-14

-12

-10

-8

-6

-4

-2

-0

Ave

rage

late

ral d

efor

mat

ion

of b

alla

st, (

S h) avg (m

m)

Fresh ballast (uniformly graded) Recycled ballast (broadly graded) Fresh ballast with geocomposite Recycled ballast with geocomposite

Number of load cycles, N Figure 13. Geometry of the finite element model (Source: Fig.1c, Carvajal and Romana 2013).

Islam et al. (2013) investigated the long-term performance of the instrumented preloaded Nerang-Broadbeach Roadway (NBR) embankment founded on a soft sensitive estuarine clay. Fully coupled nonlinear Finite Element Analyses (FEA) were carried out adopting an elasto-viscoplastic (EVP) and an elasto-plastic Modified Cam Clay (MCC) constitutive model. It was concluded that the MCC model under-predicted the ultimate settlement whereas the creep-based EVP model captured settlement quite well, albeit over-predicting the pore pressure response (Figure 14).

0 50 100 150 200 250Days

0

20

40

60

80

Exce

ss P

ore W

ater

Pre

ssur

e (kP

a)

PP1MCCEVP

Figure 14. Comparison of measured and predicted settlements and pore water pressures (Source: Figs. 4 and 5, Islam et al. 2013).

Mansikkamaki and Lansivaara (2013) introduced a 2D and 3D FEM analysis to evaluate the embankment stability of slopes reinforced with a wooden pile structure and a sheet pile wall. Wooden piles and sheet pile walls can be used to improve embankment stability if the supporting forces are reasonable. FEM can provide valuable additional information to evaluate how sensitive the structural forces can be for the soil strength variation and also to determine what would be the real nature of the failure. Figure 15 shows the failure surfaces observed with and without reinforcement.

(a) (b)

Figure 15. Failure surfaces from the safety analysis. (a) without the reinforcement (b) with the sheet pile wall (Source: Fig 8, Mansikkamaki and Lansivaara 2013).

1218


Chirica et al. (2013) presented the analysis of a road embankment with variable height located at Iassy (Romania). The FEM model had taken into account various hypotheses: (1) modeling the soil in natural state, (2) modeling the foundation in flooded state, and (3) modelling the foundation soil in a flooded state and with different imposed consolidation conditions. They reported that out of all the test hypotheses, the flooded state exhibited the highest strain and lowest bearing capacity.

Jacobse et al. (2013) developed a 3D FEM model to capture the deformation of new lifting bridge constructed across the river Oude Maas in the Rotterdam Harbour. They applied a simplified stochastic subsoil model to quantify the risk in order to deal with the uncertainty. They highlighted that the distribution of expected rotation is more or less equal to zero (Figure 16) and was in agreement with the deterministic settlement calculations.

Figure 16. Results Monte Carlo analysis pier 40, residual rotations (source: Fig 4, Jacobse et al. 2013).

1.2 Discrete element modelling (DEM)

The use of Distinct Element Method (DEM) in transportgeotechnics is gaining popularity, but regrettably there is no significant contribution made in this theme at this Conference. Therefore, for completeness of this General Report, a succinct description is provided herewith. Ballast layer is often subjected to large dynamic stresses (Yang et al. 2009), which contribute to track settlement caused by particle breakage and densification, leading to frequent maintenance (e.g. McDowell and Harireche 2002, Lobo-Guerrero and Vallejo 2006, Indraratna et al. 2010, Indraratna et al. 2012).

McDowell and Harireche (2002), and Indraratna et al. (2010) considered each particle as an agglomerate of several bonded particles. Disintegration of this agglomerate during loading is considered as breakage (Figure 17). Lobo-Guerrero and Vallejo (2006) simulated particle breakage by replacing the original particles with an equivalent set of smaller particles, when the original particle satisfies a predefined failure criterion.

Figure 17. Final fracture of a typical 0.5 mm diameter agglomerate showing intact contact bonds (after McDowell and Harireche 2002).

Indraratna et al. (2010) developed a DEM (PFC2D) model to capture the influence of frequency on the deformation and degradation of ballast during cyclic loading. DEM simulations were performed at frequencies of 10 Hz, 20 Hz, 30 Hz, and 40 Hz and for low values of loading cycles (N < 1000). The

cumulative bond breakage (Br), defined as a percentage of bonds broken compared to the total number of bonds is shown at different f and N (Fig. 18). It is observed that Br increases with the increase in f and N. Most of the bond breakages occurred during the initial cycles of loading, causing rapid permanent deformation at the start of loading, as this is exactly what is observed on new tracks upon the passage of initial trains.

Figure 18. Effects of frequency (f) on bond breakage (Br) with number of cycles (N) (after Indraratna et al. 2010).

Huang and Tutumluer (2011) assessed the behavior of fouled ballast using a “half-track” 2D DEM model. They studied the effects of different percentages of fouling and the corresponding and locations on track settlement. Recently, Indraratna et al. (2012) employed a 3D DEM model to study the shear behaviour of fresh and coal fouled ballast in direct shear testing. Fouled ballast with void contaminant index (VCI) ranging from 20% to 70% was modeled by injecting a specified number of miniature spherical particles into the ballast voids. The micro-mechanical observations obtained through DEM studies imply that fouling decreases particle breakage due to diminished stress concentrations or contact forces between ballast grains, but considerably impedes drainage when the VCI > 40%.

2 FIELD PERFORMANCE EVALUATION

There are 6 papers that have been included in this section. Two papers discuss the results of monitoring of full scale embankments used for ground improvement (Boutonnier et al. 2013, Buggy 2013) while one paper focuses upon the stability and settlement analysis of the road embankment (Murjanto et al. 2013). Effects of moisture, mechanical indices and asphalt reinforcement on the performance of concrete pavements are presented (Teltayev 2013, Touole and Thesseling 2013). Laboratory studies as well as field studies are conducted to evaluate the performance of shale as fill and embankment material (Solomon et al. 2013).

Boutonnier et al. (2013) describe the monitoring of six full-scale embankments to measure settlements and the time of consolidation. They estimate the preconsolidation pressure using undrained cohesion Cu and consider the coefficient of consolidation Cv as ten times the laboratory measured Cv value. They further conclude that the calculated settlements and time of consolidation are in good agreement with the measurements.Buggy (2013) describes the observational approach used to control embankment stability primarily by means of monitoring filling rates, pore pressures and deformation ratio (ratio of lateral toe displacement to vertical crest settlement). Embankments up to 10 m height are constructed in multiple stages with continuous monitoring of performance by means of piezometers, inclinometers, settlement plates and survey monuments. A combination of prefabricated vertical drain

1219


(PVD), geosynthetic basal reinforcement and 2 to 2.5 m of surcharge is adopted for ground improvement. A typical filling rate and deformation ratio history for one of the instrumented locations is shown in Figure 19. Buggy (2013) concludes that deformation ratios offer a reliable method for controlling stability of multi-stage embankments when used in conjunction with pore pressure instrumentation.

Figure 19. A typical filling rate and deformation ratio history for the instrumented location at Ch 4+185 m (Source: Fig 2, Buggy 2013).

Murjanto et al. (2013) presented a comprehensive stability and settlement analysis of the road embankment using a detailed site investigation. A 7.3 km long embankment with flexible pavements is built over North Jakarta-soft alluvial deposit. The pavement level was raised several times in order to compensate for the settlement. The results of the stability analysis indicated that the road is relatively in critical condition and some proposed trial designs were analyzed to fulfill minimum FS by strengthening of the road embankment using: (i) corrugated prestressed concrete sheet piles; (ii) corrugated prestressed concrete sheet piles and horizontal bars; (iii) concrete sheet piles and ground anchor; and (iv) secant pile walls.

The variation of moisture and mechanical indices on the soil basement is often neglected while designing concrete pavements. An interesting study addressing these issues is reported by Teltayev (2013). He has shown that the sagging, tensile stress and vertical deformation of the surface of soil basement are very sensitive to seasonal climate changes. He has mentioned that design of cement concrete slabs often incorporate sagging of pavement in spring, however sagging also increases in summer and autumn seasons and this can be the cause of various forms of cracks (Figure 20).

Figure 20. Transversal crack in cement concrete pavement (Source: Fig. 8, Teltayev 2013).

In another study by Touole and Thesseling (2013), tensile strengths of two different asphalt reinforcement products with different raw materials (polyester and fiberglass) are analysed considering the influence of installation damage. Results of full-scale tests after loading from truck passes and asphalt compaction revealed that the polyester grid undergo a loss of 30% of its tensile strength while the fiberglass grid showed a loss of strength up to about 90%. The fiberglass grid was damaged significantly more than the polyester grid reinforcement (Figure 21).

Figure 21. Results of installation damage test (Source: Fig 5, Touole and Thesseling 2013).

Solomon et al. (2013) described the performance of shale as fill and embankment material through laboratory studies and field trials. In order to reduce costs involved in the hauling of suitable material over longer distances, possible use of shale is evaluated. Laboratory tests including index properties, compaction, California Bearing Ratio (CBR) and triaxial tests are conducted at six different laboratories and the results, particularly the CBR values indicated that the shale was of marginal quality for its intended purposes. However, in a field trial road section constructed using the shale that was monitored for a period of two months, the results indicated high CBR and bearing resistance values with insignificant settlement. The field performance based characteristics of the shale merited its selection for use.

3 CONCLUSIONS

The Discussion Session TC202 on Transportation of the 18th ICSMGE consists of 33 papers (135 pages) describing numerous efforts on experimental research, field monitoring and data interpretation, design approaches, analytical methods and numerical modelling in six distinct categories: a) Compaction and subgrade improvement b) Laboratory testing c) Theoretical advancements and contributions to design d) Applications of geosynthetics e) Numerical modelling f) Field performance evaluation

In this General Report, an attempt has been made to offer a critical review of the majority of papers that have made a significant contribution in the area of Transportation, and the salient aspects of all papers have been summarised in the Annexure (Tables 2-7). Considering the extensive worldwide efforts put in by practitioners, academics, research associates and research students (125 contributors from 19 countries), there is no doubt that this Technical Session has offered one of the most comprehensive compilations in Transport Geotechnics, representing its current state-of-the-art. However, it is noted that only a limited number of evolving techniques have been presented to any significant extent, and these include load transfer analyses including probabilistic approaches, seismic retrofitting, intelligent compaction control, micro-mechanics of granular media through DEM modeling, analysis of soil-geosynthetic interfaces, stabilization of rail and road sub-base and sub-ballast using geocells, and other ground improvement methods addressing problematic subgrade, among others. While some additional papers are cited in this General Report especially in DEM modeling of granular media, further details of evolving techniques in Transport Geotechnics have been

1220


reported by Correia et al. (2012), Correia et al. (2007) and Indraratna et al. (2011).

4 ACKNOWLEDGEMENT

The assistance of Dr Sanjay Nimbalkar, Dr Nayoma Tennakoon, Dr Ana Heitor, Dr Cholachat Rujikiatkamjorn, Dr Jayan Vinod and Dr Ngoc Trung Ngo of the Centre for Geomechanics and Railway Engineering, and School of Civil, Mining and Environmental Engineering, University of Wollongong is gratefully appreciated.

5 REFERENCES

Papers in proceedings of discussion session TC 202 Transportation: ISSMGE, Paris 2013.

(Please see Tables 2-7)

5.1 Additional references

Correia, A.G., Momoya, Y. and Tatsuoka, F. 2007. Design and construction of pavements and rail tracks : geotechnical aspects and processed materials Taylor and Francis, London.

Correia, A.G., Quibel, A. and Winter, M.G. 2012. The work of ISSMGE TC3 (geotechnics of pavements) and how it links to earthworks. Geological Society Engineering, Engineering Geology Special Publications (EGSP)(26), 67-77.

Esveld, C. 2001. Modern railway track MRT Press, The Netherlands. Huang, H. and Tutumluer, E. 2011. Discrete Element Modeling for

fouled railroad ballast. Construction and Building Materials 25(8), 3306-3312.

Indraratna B., Salim, W. and Rujikiatkamjorn, C. 2011. Advanced Rail Geotechnology – Ballasted Track CRC Press/Balkema.

Indraratna B., Thakur, P.K. and Vinod, J.S. 2010. Experimental and Numerical Study of Railway Ballast Behaviour under Cyclic Loading. International Journal of Geomechanics ASCE 10(4), 136-144.

Indraratna, B., Ngo, N.T., Rujikiatkamjorn, C. and Vinod, J.S. 2012. Behaviour of fresh and fouled railway ballast subjected to direct shear testing - A discrete element simulation. International Journal of Geomechanics ASCE (accepted, in press).

Lobo-Guerrero, S. and Vallejo, L.E. 2006. Discrete element method analysis of railtrack ballast degradation during cyclic loading. Granular Matter, 8, 195-204.

Lu, M. and McDowell, G.R. 2006. Discrete element modelling of ballast abrasion. Géotechnique 56(9), 651-655.

McDowell, G.R and Harireche, O. 2002. Discrete Element Modelling of Soil particle fracture. Géotechnique 52(2), 131-135.

Yang L.A., Powrie W. and Priest, J.A. 2009. Dynamic stress analysis of a ballasted railway track bed during train passage. Journal of Geotechnical and Geoenvironmental Engineering ASCE 135(5), 680-689.

1221


6 ANNEXURE - TABLE 2-7: SUMMARY OF ALL CONTRIBUTIONS IN DISCUSSION SESSION

Table 2. Compaction and subgrade improvement for transport infrastructure

Title of Paper Authors Country Summary of Main Contribution

Five years of Impact Compaction in Europe – successful implementation of an innovative compaction technique based on fundamental research and field experiments

Adam D., Paulmichl I., Adam, C. and Falkner F.J.

Austria The impact compaction methods are more efficient if greater depths of densification are of interest.

Key aspects related to the application of impact compaction such as the surface velocity, weight and number passes are analyzed both in numerical simulations and field cases studies. The results are compared and compaction efficiency is also linked to the type of soil.

Assessing the Effectiveness of Rolling Dynamic Compaction

Kuo Y.L., Jaksa M.B., Scott B.T., Bradley A.C., Power C.N., Crisp A.C. and Jiang J.H.

Australia The efficiency of rolling dynamic compaction (RDC) is examined by means of a combination of field studies and numerical modeling

RDC is more effective for depths between 0.8 m to 3.0 m and the most significant factors governing its efficiency are soil cohesion, Poisson’s ratio and shear modulus, as well as the width and mass of the RDC module.

Applicability of the Geogauge, P-FWD and DCP for compaction control

Conde M. C., Lopes M. G., Caldeira L. and Bilé Serra J.

Portugal The feasibility of a stiffness-based specification for embankment soil compaction quality control is discussed.

DCP equipment showed greater suitability as a compaction control tool, due to the strong negative correlation with water content values.

Ground improvement methods for the construction of the federal road B 176 on a new elevated dump in the brown coal region of MIBRAG

Kirstein J. F., Ahner C., Uhlemann S., Uhlich P. and Röder K.

Germany The design and the settlements are significantly optimized by the combination of different soil improvement techniques, particularly in cases where significant stability problems are expected.

The settlements predictions by Finite element modeling agreed well the results obtained with in situ pressuremeter tests.

Laboratory characterization and model calibration of a cemented aggregate for application in transportation infrastructures

Viana da Fonseca A., Rios S., Domingues A.M., Silva A. and Fortunato E.

Portugal The differences observed in dynamic and static stiffness properties and shear strength parameters of compacted mixtures of cement and limestone aggregate are directly associated to the variation of porosity/cement ratio.

Hardening soil models may be employed to describe the stress-strain behavior, but do not provide satisfactory predictions of the volumetric behavior and post-peak strain softening.

Table 3. Laboratory Testing


Railways platforms reinforced by soil-mixing columns without track removing

Calon N., Robinet A., Costa S., D’Aguiar L., Cojean, B.C. and Mosser J.F.

France Investigated the potential benefits from the ground reinforcement by vertical soil-cement columns

Performed laboratory tests to investigate the influence of the column location and the efficiency of geosynthetics on the reduction of stiff zones

With subsequent numerical model, they determined the optimum columns mesh

Effects of ballast thickness and tie-tamper repair on settlement characteristics of railway ballasted tracks

Hayano K., Ishii K. and Muramoto K.

Japan Conducted a series of cyclic loading tests on model grounds to investigate the effects of ballast thickness and tie-tamper repair on the settlement characteristics of ballasted tracks

Maximum shear strain distributions generated in the model grounds were analyzed with particle image velocimetry

Effect Evaluation of Freeze-Thaw on Deformation-Strength Properties of Granular Base Course Material in Pavement

Ishikawa T., Zhang Y., Kawabata S., Kameyama S., Tokoro T. and Ono T.

Japan CBR tests of freeze-thawed subbase course materials under various water contents, and the resilient modulus tests in unsaturated condition were conducted using two newly developed test apparatus

The test results were compared with long-term field measurement at a model pavement structure, including FWD tests

On the Permanent Deformation Behavior of Rail Road Pond Ash Subgrade

Mohanty B. and Chandra S.

India Repeated load triaxial tests were conducted on reconstituted pond ash specimens

Permanent deformation calculations take into account the stress history

1222


and number of passes of vehicular traffic loadingTest results were analyzed to study the effects of confining pressure, deviatoric stresses, and degree of saturation on the permanent deformation response of pond ash

Effect of wetting- drying cycles on CBR values of silty subgrade soil of Karaj railway

Moayed R.Z., Lahiji B. P. and Daghigh Y.

Iran Conducted series of CBR tests to investigate the effect of lime-microsilica additive as a modern additive stabilizer on a silty soil

Lime and microsilica were mixed with the soil in different percentages and specimens were prepared at optimum moisture content.

Model tests on settlement behaviour of ballasts subjected to sand intrusion and tie tamping application

Kumara J. and Hayano K.

Japan Model tests were conducted on 1/5th scale of the actual size of railway track

The relationship between the number of loading cycles and settlement was obtained and results were discussed with degree of ballast fouling

Table 4. Theoretical Advancements and Contributions to Design


Evaluation of roadbed potential damage induced by swelling/shrinkage of the subgrade

Simic D. Spain The average suction compression index of the plate load tests and the routine soil parameters were adopted to carry out a comparison between the methods of estimating swelling deformation.

The potential vertical rise method is very dependent on the active moisture depth, which should be adopted based on the local experience.

Development of a non-linear ballasted railway track model

Fernandes V.A., D’Aguiar S. C., and Lopez-Caballero F.

France A 2D finite element model with a modified width plane strain condition is used where viscous boundaries are implemented using a Kelvin-Voigt viscoelastic mechanical model to reduce wave reflexion on boundaries.

The importance of initial state evolution of track materials on the context of non-linear mechanical behavior is discussed to assure the correct combination of laboratory tests based on current track conditions, especially ballast.

Seismic Retrofit Technique for Asphalt Concrete Pavements

Ohta H., Ishigaki T. and Tatta N.

Japan The structure of the seismic retrofit technique for asphalt concrete pavements using Confined-Reinforced Earth (CRE), construction method and the results of full scale in-situ tests were described where crushed stones and new design procedures were introduced.

Full scale in-situ tests show the acceptable performance of CRE after the forced settlement to simulate the severe earthquake-induced damage.

Influence of Anti-freezing on the Frost Penetration Depth for Paved Road Design

Shin E.C., Cho G.T. and Lee J.S.

Korea The frost penetration depth of paved road was determined by the field measurement. The subbase and base courses are influenced by the temperature below 0o regardless of anti-frost layer.

The frost penetration depth estimated by the empirical equation proposed by Korea Institute of Construction Technology shows a similar trend in lower frost index.

The reasonable design concept is proposed for road design.

Special Aspects for Building a Motorway on a 185 m Deep Dump

Vogt N., Heyer D., Birle E., Vogt S., Dahmen D., Karcher C., Vinzelberg G. and Eidam F.

Germany The paper presents the project-specific conditions during the dumping process and the properties of the dumped soils along the future A 44 routing.

A simple model for the description of the time-dependent deformation of the dump and the effectiveness of soil compaction methods is discussed and evaluated.

The simulation results and geodetic measurements have shown that by allowing the resting period of at least 6 months between the end of the dumping process and the start of the construction work, the settlements of structures or pavements can be reduced significantly.

Equilibrium models for arching in basal reinforced piled embankments

Eekelen S.J.M. and Bezuijen A.

Netherlands The paper compares three equilibrium models describing arching in geosynthetic basal reinforced (GR) piled embankments, namely the models of Hewlett and Randolph (1988), Zaeske (2001) and the concentric arches model of Van Eekelen (2013b).

The load distributions predicted by Hewlett and Randolph (1988) and Zaeske (2001) show a uniform load distribution on the GR between the piles. The concentric arches model (Van Eekelen et al. 2013b) provides

1223


a load concentration on the GR strips with an inverse triangular load distribution on those GR strips. This is in agreement with observations in scaled model tests, numerical analysis and field measurements.

Recent developments in pavement foundation design

Brown S.F. and Thom N.H.

United Kingdom A Precision Unbound Materials Analyzer (simplified version of the repeated load triaxial test) has been developed to quantify both resilient and plastic strain characteristics.

Unlike CBR testing, this technique can be very useful in allowing a designer to evaluate alternative foundation material combinations in order to achieve a desired foundation.

Table 5. Applications of Geosynthetics


Load transfer mechanisms in geotextile-reinforcedembankments overlying voids: experimental and numerical approaches

Huckert A., Garcin P., Villard P., Briançon L. and Auray G.

France Full-scale experiments were conducted on non-cohesive granular embankments to study load transfers of geosynthetics-reinforced embankments prone to sinkholes.

Numerical model was performed to provide a better understanding of the load transfers towards the edges of the cavity

Performance verification of a geogrid mechanically stabilised layer

Wayne M., Fraser I., Reall B. and Kwon J.

USA Series of full-scale field tests and model tests were conducted to evaluate performance of a geogrid- stabilized unpaved aggregate base overlaying relatively weak and non-uniform subgrade soils.

The results confirm that the geogrid promotes improved aggregate confinement and interaction, leading to enhanced structural performance of the unpaved aggregate base.

Performance Assessment of Synthetic Shock Mats and Grids in the Improvement of Ballasted Tracks

Indraratna B., Nimbalkar S., Rujikiatkamjorn C., Neville T. and Christie D.

Australia Full-scale field tests were conducted on rail track sections in the towns of Bulli and Singleton (Australia) to measure track deformations associated with cyclic stresses and impact loads.

The results indicated that the use of geocomposites as reinforcing elements for ballast proved to be a feasible and economically attractive alternative.

Characterization of Soil-Geosynthetic Interaction under Small Displacements Conditions

Zornberg J.G., Roodi G.H., Gupta R. and Ferreira J.

USA A mathematical model with a new parameter, defined as “Stiffness of Soil-Geosynthetic Interaction” or KSGI, was introduced to address soil-geosynthetic interaction behavior under small displacements.

Pull out tests with geosynthetics embedded in poorly graded sand were conducted to evaluate the proposed model.

Table 6. Numerical Modeling


Analysis of the influence of soft soil depth on the subgrade capacity for flexible pavements.

Carvajal E. and Romana M.

Spain Presented the analyses of a flexible pavement structure founded on soft soil subgrade, through the FEM of a multilayered system.

Deep ground treatments should be applied to achieve an allowable capacity of soft soils up to minimum depth of about 6 m, otherwise maintenance cost of pavements might be excessive.

Long-term performance of preloaded road embankment

Islam M.N., Gnanendran C.T.,

Sivakumar S.T.,

Karim M.R.

Australia Investigated the long-term performance of the preloaded Nerang-Broadbeach Roadway (NBR) embankment near the Gold Coast in Queensland (Australia).

The MCC model under-predicted the ultimate settlement while the creep-based EVP model captured it well but over-predicted the pore pressure response.

The modified calculation of the Asaoka method predicted almost identical magnitudes of ultimate settlement as the Hyperbolic method and FEAs.

Stability improvement methods for soft clays in a railway environment

Mansikkamäki J. and Länsivaara T.

Finland This paper introduces an evaluation of alternative methods to improve embankment stability with wooden pile structures or with sheet pile walls.

Based on the 2D and 3D finite element analysis and to the soil behavior calibrated in the failure test and existing, they highlighted that analysed railway embankments are under poor stability conditions.

Probabilistic Settlement Analysis For The Botlek Lifting Bridge

Jacobse J.A., Nehal R.S., Rijneveld B.

Netherlands Presented the deformation analysis, deterministic 3D FEM calculations, on a lifting bridge constructed across the river Oude Maas in the

1224


Design and

Bouwmeester D.

Rotterdam harbour area

Application of a simplified stochastic subsoil model captured quantitative risk analysis in order to deal with the uncertainties.

The geotechnical analysis corresponding to the high road embankments close to a bridge

Dragos A.C., Tenea V. D.

Romania Presented a complex case study corresponding to a road embankments upto10m height placed at Iassy (Romania).

The analysis shows high strain and low bearing capacity for soils in flooded state.

Table 7. Field Performance Evaluation


LGV EST section 41 : measured and calculated settlements under embankments

Boutonnier L., Hajouai F., Bacar Fadhuli N. and Gandille D.

France Six embankments are monitored to measure the settlements and the time of consolidation.

The preconsolidation pressure of the soil is estimated from the undrained cohesion Cu and the coefficient of consolidation Cv used in the design was ten times the Cv measured with laboratory test.

Influence of installation damage on the tensile strength of asphalt reinforcement products

Touole L.S. and Thesseling B.

Germany The effective tensile strength of asphalt reinforcement products, considering installation damage was analysed.

A considerable difference in loss of tensile strength, due to the effects of installation damage was observed.

Influence of Mechanical Indices for Soil Basement on Strength of Road Structure

Teltayev B. Kazakhstan Moisture value and its phase content in soil basement of the highway vary substantially in annual cycle and according to the depth of basement.

The maximum values of sagging, tensile stress and vertical deformation of cement concrete pavement occur in summer and autumn seasons.

The performance of shale as fill and embankment material for a trunk road in Ghana

Solomon K.M., Oddei J.K. and Gawu S.K.

Ghana The CBR values indicated variations between 8% and 12%, which are below the contract special specification minimum value of 15%. This was therefore considered as having marginal quality for its intended purposes.

Results obtained from the field evaluation indicated high CBR and bearing resistance values including insignificant settlement.

Evaluation of the Performance of Road Embankments over North Jakarta-Soft Soils

Murjanto D., Rahadian H., Hendarto and Taufik R.

Indonesia To fulfill stability and settlement analysis, the road at Zone 1, and 3 should be strengthened by secant pile walls combined with raising of 0.7 m.

The road at Zone 2 should be strengthened by concrete sheet piles and ground anchor.

Deformation Performance and Stability Control of Multi-stage Embankments in Ireland

Buggy F.J. Ireland Deformation ratios offer a reliable method for controlling stability of multi-stage embankments when used in conjunction with pore pressure instrumentation. For the specific conditions at Limerick Tunnel, a limit deformation ratio of 0.6 was shown to give satisfactory performance and acceptable stability.

Excessive lateral deformation related to local failure occurred at several locations in the vicinity of creeks, ditches and historical excavations located within 10 m of the embankment toe.

1225

Five years of Impact Compaction in Europe – successful implementation of an innovative compaction technique based on fundamental research and field experiments

Cinq ans de compactage par impact en Europe – mise en œuvre avec succès d'une technique de compactage novatrice basée sur la recherche fondamentale et expériences sur le terrain

Adam D., Paulmichl I. Vienna University of Technology, Austria

Adam C., Falkner F.-J. University of Innsbruck, Austria

ABSTRACT: In the year 2007, the innovative Impact Compactor was widely introduced in Central Europe on the initiative of anAustrian company to compact and improve the ground. At the beginning, the application of the novel impact-like compaction technique was based on empirical data and experience gained on several construction sites. Soon after, a funded research project was initiated including both fundamental research and field experiments. The outcomes of the research work provided the basis for theoptimized and economic application of this novel compaction method on-site. Since 2007, in numerous applications the ImpactCompactor has been successfully employed for ground improvement for industrial, administrative and apartment buildings, bridges,bridge abutments, embankments, dams and dikes, and other civil engineering structures.

RÉSUMÉ : Le compacteur par impact pour l'amélioration des sols a été introduit en Europe Centrale en 2007, sur l'initiative d'unesociété Autrichienne. Au début de son utilisation cette technique novatrice de compactage et fondée sur des données empiriques etl'expérience acquise sur plusieurs chantiers. Peu de temps après, un projet de recherche a été lancé en se focalisant sur la recherchefondamentale et les expériences sur le terrain. Les résultats de ces travaux de recherche ont fournit la base d'une application optimisée et économique de ce procédé novateur de compactage sur site. Dans de nombreux projets, le compacteur à impact a été mis en œuvreavec succès pour l'amélioration des sols pour des projets des bâtiments industriels, administratifs et appartements, ponts, culées de ponts, remblais, barrages et digues et autres travaux de génie civil.

KEYWORDS: Impact Compactor, dynamic compaction, soil dynamics, ground improvement, earth works.

1 INTRODUCTION

1.1 Background and history of the Impact Compactor

The Impact Compactor was developed for the British military forces to compact and improve the ground. In the year 2007 the Austrian company TERRA-MIX introduced this device in Central Europe.

In the early days after implementation the application of the novel impact-like compaction technique was based on empirical data and experience gained on several construction sites. Later, a basic research project funded by the Austrian Research Promotion Agency (FFG) was initiated to quantify the effect of this innovative device, and to optimize its application. At the same time, a GPS-based data recording system for the documentation of the compaction process including stop codes as indication for maximum possible compaction was developed.

Since implementation, the Impact Compactor has proven to improve efficiently the ground for industrial, administrative and apartment buildings, bridges, bridge abutments, embankments, dams and dikes, and other civil engineering structures.

1.2 Basic principle and setup of the Impact Compactor

The Impact Compactor is a dynamic compaction device based on the piling hammer technology that is used to increase the load-bearing capacity of soils through controlled impacts. The general idea of this method is to drop a falling weight from a relatively low height onto a special foot assembly at a fast rate while the foot remains permanently in contact with the ground. The lately introduced compaction equipment aims at closing the gap between the surface compaction methods and the deep

compaction methods, and permitting a middle-deep improvement of the ground up to a depth of 4.5 to 7.5 m (10 m) (Adam and Paulmichl 2007).

The Impact Compactor consists mainly of three impact components: the impact foot, the driving cap, and the hammer with the falling weight. The impact foot made of steel has a diameter of 1.5 m. Since the driving cap is connected loosely to the foot, only compression forces load the subsoil, which allows an efficient energy transfer. Impact foot, driving cap, and falling weight are connected to the so-called hammer rig. Falling weights of mass 5,000, 7,000, 9,000 or 12,000 kg are dropped from a falling height up to 1.2 m at rate 40 to 60 repetitions per minute. For further details see (Falkner et al. 2010).

Gravels, sands, silts, industrial byproducts, tailings material, and landfills can be successfully compacted by the Impact Compactor to increase the load-bearing capacity of foundations, to improve the ground bedding conditions for slabs, to reduce the liquefaction potential of soils, and to stabilize waste materials.

2 FUNDAMENTAL RESEARCH

2.1 Numerical simulations

Theoretical investigations comprised numerical computer simulations of the impact-type compaction effect, energy transfer into the soil, and wave propagation.

A theoretical study of the dynamic impact of the Impact Compactor on its environment was performed employing a simple mechanical model of the Impact Compactor-subsoil interaction system. Thereby, the falling weight was modeled as

1226


lumped mass, which hits the impact foot after a free fall. The initial velocity of the impact foot, which excites the underground, was derived assuming an idealized elastic impact between falling weight and the mass of the impact foot. The soil medium was modeled as homogenous, isotropic, and rate-independent elastoplastic halfspace based on Mohr-Coulomb theory with isotropic hardening. The axially symmetric impact foot made of steel rests on the surface of the halfspace. A sliding interface between the foot and the soil was adopted, i.e. only normal stresses are transferred between the foot and the soil. The numerical model takes advantage of the rotational symmetry of this subsystem, which is divided into a near-field and a far-field. The near-field was discretized by means of Finite Elements. Infinite Elements model the far-field in order to avoid wave reflections at the boundary between the near- and far-field, and to allow for energy propagation into the semi-infinite halfspace. The model and its parameters are described in more detail in Adam et al. (2010).

As an example, Figure 1 shows the peak velocity magnitude vR,max with respect to the distance of the compaction point for the subsoil condition silty fine sand after the first, third, fifth, and tenth compaction pass. The outcomes of this figure prove field observations that the pronounced increase of vR,max after each compaction impact leads to a parallel shift of the regression line, and thus, the arbitrary assumed limit value of 10 mm/s is shifted to a larger distance from the compaction point.

Figure 1. Magnitude of maximum resultant surface velocity as function of the distance from the impact foot after a specified number of compaction impacts applied to an elastoplastic silty fine sand.

Figure 2. Distribution of the velocity magnitude at two specified instants after the first compaction impact. Elastoplastic silty fine sand.

Figure 2 shows the propagation of the velocity magnitude at two instants after the first impact is applied to the subsoil condition silty fine sand. Spherical propagation of the waves can be observed. Comparison of Figure 15(a) and Figure 15(b) prove that geometric damping leads to a rapid decay of the response amplitudes. According to Figure 15(b) the maximum peak velocities develop at the soil surface, because Rayleigh waves have the largest energy content. Furthermore, the faster propagating P-waves can be distinguished from the slower S-waves. According to the characteristics of P-waves in zones between compression and dilatation the velocities are zero.

The effect of compaction and the compaction depth have been investigated, because these properties serve to define the application fields of the Impact Compactor with respect to soil

type and soil stratification. In numerical studies it was assumed that the equivalent plastic strain is the characteristic parameter for evaluation of the compaction depth. A threshold of 0.02 separates the compacted space from the non-compacted subsoil. After each impact in the compaction zone the soil properties were modified. Here, an isotropic hardening constitutive model was used for an engineering-like approximation of soil compaction.

Figure 3 shows the expansion of the equivalent plastic strains in a cross-section of homogeneous silty fine sand below the impact point after the first and tenth compaction pass. The colored area within the outer contour is considered as compaction zone. The largest equivalent plastic strains occur below the boundary of the compaction foot. The domains of equal plastic strains, i.e. the domains of equal degree of compaction, show the shape of a “stress bubble”. It can be seen that in this example the soil is compacted laterally and downwards with approximately the same magnitude. A thin surface layer shows as well distinct equivalent plastic strains, which are induced by Rayleigh waves. After the tenth impact the compaction depth is about 4.3 m.

Figure 3. Spread of the equivalent plastic strain after the first (left) and after tenth compaction impact (right). Elastoplastic silty fine sand.

2.2 Field experiments

Field tests on different soil conditions were performed to verify theoretically derived outcomes. Moreover, they provide the basis for the optimized and economic application of this compaction method in the field.

Experimental results and field investigations confirm the trends of the presented numerical outcomes (see chapter 3).

3 DEVELOPMENT AND APPLICATION

3.1 GPS-based recording system

The Impact Compactors are provided with a monitoring system. The compaction monitor is a kit of parts, which can be coupled to the compaction device in order to record the performance of the hammer and the rate of ground improvement. The following parameters are automatically recorded during the compaction process and monitored from the cab with an on-board data acquisition system (see Figure 4): number of blows final settlement at the last blow total settlement (depth of the compaction crater) compaction energy average number of blows In addition to these parameters a more novel device monitors electronically the coordinates of the compaction points, date, and time for each compaction point during the compaction process, and all data are documented via GPS controlled data acquisition (see Figure 4).

GPS-based data recording during the compaction process and the online display in the operator’s cab facilitates compaction control, an economic application of the compaction

1227


tool, and a work integrated quality control. Thus, local heterogeneities of the subsoil can be identified, and compaction with the Impact Compactor can be adjusted systematically. If necessary, additional compaction passes are conducted.

Figure 4. GPS-based recording system of the Impact Compactor.

3.2 Parameter setting and quality control

Optimization and control of compaction with the Impact Compactor is ensured by meeting the stop code criteria, GPS based compaction including work integrated documentation of the performance parameters for each compaction spot, and conduction of cone penetration tests and/or dynamic probing before and after compaction. During the compaction process the following stop codes are applied: stop code 1: total settlement (depth of the compaction crater) stop code 2: number of blows per compaction point stop code 3: final settlement of the last blow

Figure 5: Compaction process (left) and compaction control (right).

The stop codes have to be verified and optimized on a test field that is located within the site (see Figure 5). In dependence of the subsoil conditions and the complexity of the project the calibration field can comprise up to three different compaction patterns and point grids. The compaction process at the test field is usually carried out by applying stop codes defined by a geotechnical expert based on the results and experiences from comparable sites. After the test compaction the treatment depth is determined and compared with the required compaction depth in order to find the suitable compaction point grid. The compaction pattern and point grids, the number of compaction passes and the stop codes are finally defined by the geotechnical expert.

The compaction depth is determined conducting cone penetration tests (CPT) and/or dynamic probing light, medium, or heavy (DPL, DPM, or DPH).

In Figure 6 the number of blows N10 determined by dynamic probing heavy and light before and after compaction is plotted against the depth. The dynamic probing heavy was performed in non-cohesive primarily sandy gravelly soil; the dynamic

probing light was carried out in cohesive soil consisting of silts and sands. It can be seen that the depth effect of the Impact Compactor depends on the soil condition, and it varies from about 4 m (silts and sands) to 7 (8) m (sandy gravelly soils).

In cohesive soils of soft to stiff consistency dynamic probing heavy allows only a low number of blows independent of the degree of compaction. Consequently, for checking the compaction effect it is recommended to use dynamic probing light (DPL) or cone penetration tests (CPT) (Adam et al. 2010).

Typical depths of influence (treatment depth) are summarized in Table 1 in dependence of the soil type based on the results of numerous experimental investigations.

For quality control recorded compaction parameters are evaluated graphically. As an example, in Figure 5 (right) the “final set” (stop code 3) is used as control criteria, and the compaction points are hatched in blue, green, yellow or red color in dependence on the numerical value of the recorded “final set”. It can be seen that another compaction pass had to be carried out on the red colored points. Consequently, this plot gives information on the compaction quality (whether the stop codes are met all over the site or not), and allows conclusions to be drawn about the subsoil quality before compaction.

0 10 20 30

-8

-6

-4

-2

0

before compaction

after compaction

number of blows N10(DPH)

(a)

dept

h [m

]

0 20 40 60 80

-8

-6

-4

-2

0

number of blows N10(DPL-5)

before compaction

after compaction

(b)

dept

h [m

]

Figure 6: Dynamic Probing Heavy (DPH) in non-cohesive soil (left) and Dynamic Probing Light (DPL-5) in cohesive soil (right).

Table 1. Characteristic compaction depth for the Impact Compactor ith a falling weight of 9,000 kg mass. w

Type of soil Type of dynamic probing

Number of blows

Treatmentdepth

Sa/Gr DPH N10 > 20 6 – 7.5 (10) m si Sa DPH N10 > 15 5 – 6 m sa Si DPL N10 > 20 4.5 – 5 m Miscellaneous graded soils DPL/DPH N10 > 15 / 20 4.5 – 7 m

3.3 Vibration emission and immission

On numerous test sites the maximum surface velocity induced by the Impact Compactor as function of the distance were determined. The data acquisition tool MR2002DIN-CE (RED BOX) of the company SYSCOM was applied to monitor and record the vibrations. The velocities were measured in situ with tri-axial velocity transducers according to the German Standard DIN 45669 and saved with a data recorder. The velocity was measured in three orthogonal directions in the frequency domain of 1 to 315 Hz. The subsequent data processing was done with the software package VIEW 2002 (Ziegler Consultants). Subsequently, regression analyses were performed to obtain the magnitude of the maximum resulting velocity vR,max as function of the distance from the impact foot.

Figure 7 shows selected linear regression lines for different homogeneous ground conditions determined through free-field velocity measurements during impact compaction with a falling weight of 9,000 kg mass. It is seen that smallest peak velocity magnitudes develop during compaction of homogeneous loose sandy gravels. For this subsoil condition a coefficient of decay of about 1.8 is determined. Note that only one compaction pass was performed. Largest peak velocity magnitudes were measured during compaction of dense gravels. Compaction of sandy silts and gravelly silty sands led to peak velocity magnitudes in-between. The coefficient of decay of about 1.3 is

1228


practically identical for dense gravels, sandy silts, and gravelly silty sands. The results show that the peak velocity magnitude falls below the value of max vR,max = 10 mm/s, i.e. the limit value for buildings of the class no. III according to the Austrian Standard ÖN S 9020, at a distance of 11 to 34 m from the impact foot, depending on the subsoil condition and soil type. Based of hitherto experience the required minimum distance to buildings of class no. III is about 20 m. In comparison compaction of heavy tamping techniques induces resulting velocities of more than 10 mm/s at a distance of 30 m.

Figure 7. Magnitude of maximum resulting velocity as function of the distance from the impact foot. Measured values for different soil types.

4 SELECTED CASE HISTORIES

4.1 Ground improvement for embankments and foundations

In the last five years the standard application for the Impact Compactor was the ground improvement for embankments and foundations. Typical fields of application are: improvement of the ground in the embankment base compaction to increase the bearing capacity of foundations

and/or reduce the liquefaction potential of soils improvement of the ground bedding conditions for slabs combined application with other compaction methods such

as heavy tamping or deep vibro-compaction when large compaction depth is required, or lime stabilization of soft cohesive soils on top of the ground (Adam et al. 2010)

4.2 Rehabilitation of flood protection dikes

The efficiency of the Impact Compactor to improve existing flood protection dikes alternatively to e.g. the mixed-in-place method (MIP) was investigated by compaction of the core of a test dike (Adam et al. 2010).

The test dike was constructed on a gravelly ground, which is covered with a loess layer of about 0.75 m thickness. The core of the embankment was built layer-wise with a layer thickness of about 1 m. Each layer was only “pre-compacted” with a vibratory roller in order to simulate the weak compactness of existing old flood protection dikes. For one half of the embankment core sandy silt (loess) was used as filling material, for the other half silt (loam). The shoulders and slopes were constructed with sandy gravel (see Figure 8).

Optimization and control of compaction was realized by the following tasks and criteria: meeting the stop code criteria GPS-based documentation of the compaction parameters performance of dynamic probing heavy (DPH) before and

after compaction performance of dynamic load plate test using the LFWD

before and after compaction in-situ permeability tests

In the following selected results of dynamic probing tests are presented exemplary, which were carried out to determine the compaction depth. Figure 8 (right) illustrates the number of blows N10 over depth determined with dynamic probing heavy

in the test section consisting of loess. It is obvious that the depth effect of the Impact Compactor is about 4.5 m. Figure 8 reveals that the upper zone of the gravelly ground beneath the embankment was compacted as well.

depth [m]

4.5 m 12.0 m

4.0


compacted zone

N10

BEFORE RIC

AFTER RIC

depth [m]

4.5 m 12.0 m

4.0


compacted zone

N10

BEFORE RIC

AFTER RIC

4.5 m 12.0 m

4.0


compacted zone

4.5 m 12.0 m

4.0


compacted zone

N10

BEFORE RIC

AFTER RIC

Figure 8. Section of the test dike (left) and Dynamic Probing Heavy in the loess (right).

5 CONCLUSION

In Central Europe the Impact Compactor was introduced in 2007. The novel compaction equipment provides a technically sound and economic method of improving the capacity of a wide variety of loose soils (silts, sands, gravels, cobbles, boulders) and fills. The effective treatment depth in soils is dictated by grain sizes and is typically in the range of 4.5 m (silt and sand) up to 7.5 m (10 m) depth (sand and gravel). Due to the numerous benefits, e.g. monitoring of the compaction process through a GPS-based recording system (on-board computer), reliability and safety in operation, quality assurance, versatility and working speed, the Impact Compactor is now a well established dynamic compaction method throughout Europe.

6 ACKNOWLEDGEMENTS

The Austrian Research Promotion Agency (FFG) has funded this research project. This support is gratefully acknowledged.

7 REFERENCES (TNR 8)

Adam D., and Paulmichl I. 2007. Impact compactor – an innovative dynamic compaction device for soil improvement. In: Proc. 8th International Geotechnical Conference (June 4-5, 2007, Slovak University of Technology, Bratislava, Slovakia), pp. 183-192.

Falkner F.-J., Adam C., Paulmichl I., Adam D., and Fürpass J. 2010. Rapid impact compaction for middle-deep improvement of the ground – numerical and experimental investigation. In: 14th Danube-European Conference on Geotechnical Engineering "From Research to Design in European Practice", June 2-4, 2010, Bratislava, Slovakia, CD-ROM paper, 10 pp.

Adam C., Falkner F.-J., Adam D., Paulmichl I., and Fürpass J. 2010. Dynamische Bodenverdichtung mit dem Impulsverdichter (Dynamic soil compaction by the Rapid Impact Compactor, in German).Project No. 815441/13026 – SCK/KUG, Final report for the Austrian Research Promotion Agency (FFG), 184 pp.

Adam C., Adam D., Falkner F.-J., and Paulmichl I. 2011. Vibration emission induced by Rapid Impact Compaction. In: Proc. of the 8th

International Conference on Structural Dynamics, EURODYN 2011, p. 914-921, 4 – 6 July 2011, Leuven, Belgium.

Fürpass J., and Bißmann, M. 2012. 5 Jahre Impuls-Verdichtung in Europa. Rückblick auf ein Erfolgsmodell (in German). In: 2. Symposium Baugrundverbesserung in der Geotechnik, p. 149-163, 13 – 14 September 2012, Vienna, Austria.

1229

Développement d’un modèle non linéaire de la voie ferrée ballastée

Development of a non-linear ballasted railway track model

Alves Fernandes V., Costa d’Aguiar S. Innovation & Recherche SNCF, Paris, France

Lopez-Caballero F. LMSSMat – Ecole Centrale Paris, Châtenay-Malabry, France

RÉSUMÉ : L’objectif de ce travail est d’étudier la réponse mécanique de la voie ferrée dans le contexte de comportement mécaniquenon-linéaire des matériaux ferroviaires. Un modèle éléments finis 2D avec épaisseur en déformation plane modifiée est utilisé ; des frontières absorbantes sont implémentées avec un modèle de comportement viscoélastique de type Kelvin-Voigt afin de réduire lesréflexions d’onde aux bords. Le modèle de comportement de sols ECP est considéré pour la couche intermédiaire et la plateforme,dont les paramètres sont calibrés à partir des essais triaxiaux disponibles dans la littérature. Afin de montrer l’effet de l’utilisationd’un modèle de comportement non-linéaire, les déplacements verticaux normalisés obtenus sont comparés à un modèle élastique.L’importance de l’état initial et de son évolution est évoquée afin d’assurer une transposition correcte des essais laboratoires aux conditions d’utilisation des voies ferrées, principalement pour le ballast.

ABSTRACT: The aim of this work is to study track mechanical response in the context of non-linear mechanical behavior of tracklayers. A 2D finite element model with a modified width plane strain condition is used in this work; viscous boundaries are implemented using a Kelvin-Voigt viscoelastic mechanical model as to reduce wave reflexion on boundaries. The ECP constitutive model is considered to simulate the behavior of soils of both the intermediate layer and platform. The model parameters werecalibrated from triaxial test results available in the literature. In order to show the effect to use a non-linear soil behavior, the obtainednormalized vertical track displacements were compared to those obtained with an elastic model. The importance of initial state evolution of track materials on the context on non-linear mechanical behavior is discussed as to assure the correct transposition oflaboratory tests to track current conditions, especially to the ballast material.

MOTS-CLÉS : voie ferrée, éléments finis, mécanique non-linéaire, rotation de contraintes principales

KEYWORDS: railway track, finite-element model, non-linear mechanical behaviour, principal stress axis rotation

1 INTRODUCTION

Les géomatériaux ferroviaires possèdent différentes échelles de complexité. La taille des grains, la géométrie et la nature des matériaux peuvent varier selon les couches ferroviaires et à l’intérieur de chaque couche. La réponse mécanique des géomatériaux est très non-linéaire et dépend du chemin de contraintes appliqué, du taux de déformation plastique ; et selon la taille des grains et/ou nature du sol, sa réponse mécanique peut être aussi dépendante de la présence de l’eau. Cet ensemble de caractéristiques contribue à l’évolution de la réponse mécanique de la voie ferrée ; cependant, seulement un modèle élastoplastique non-linéaire est capable d’en prendre compte.

L’utilisation des modèles rhéologiques avancés pose deux questions majeures : comment obtenir les paramètres nécessaires au modèle ? et combien représentatif est leur transposition directe vers la modélisation de la réponse mécanique d’une structure sous charge ? La première question est fréquemment répondue à partir des essais en laboratoire ou in-situ. La deuxième, par contre, est spécifique des modèles non-linéaires, pour lesquels le chemin de contraintes et l’évolution de l’état initial doivent être pris en compte afin de modéliser correctement la réponse de la structure.

L’objectif de ce papier est d’étudier la réponse mécanique de la voie ferrée en utilisant des modèles de comportement élastique et élastoplastique pour les différents composants. Le modèle de comportement avancé ECP (Aubry et al., 1982, Hujeux, 1985) est choisi puisque il permet de simuler un large éventail de possibles réponses mécaniques. Les deux aspects auparavant évoqués – obtention de paramètres et leur transposition vers la structure – sont discutés dans le contexte des géomatériaux ferroviaires.

2 MODÈLE DE COMPORTEMENT ECP

Le modèle élastoplastique multimécanisme développé à l’Ecole Centrale Paris, appelé modèle ECP (Aubry et al., 1982, Hujeux, 1985) est écrit selon le concept de contraintes effectives de Terzaghi. Le modèle est basé dans le critère de rupture de type Coulomb est dans le concept d’état critique. Tout phénomène irréversible est modélisé par trois déformations planes déviatoriques plastiques dans trois plans orthogonaux et une dans le plan isotrope. L’évolution de l’écrouissage est contrôlée par la déformation plastique selon le mécanisme : déformation déviatorique et volumique pour les mécanismes déviatoriques et déformation volumique pour le mécanisme isotrope. Le comportement cyclique utilise l’écrouissage cinématique basé dans les variables d’état au dernier changement de direction de chargement.

Le modèle est basé dans la théorie de la plasticité incrémentale, laquelle considère une décomposition complète des déformations dans une partie élastique et plastique. Le modèle considère l’élasticité non-linéaire pour la réponse élastique. Le module isostatique (K) et de cisaillement (G) sont fonctions de la pression moyenne effective (p’) :

(1)

(2)et Gref sont le module Kref

isostatique et de cisaillement mesurés à la pression moyenne de référence pref, et ne est le dégrée de non-linéarité. La surface de réponse déviatorique dans le plan k est la suivante :

1230


(3)

avec:

(4)

(5)

pp est l’angle de frottement à l’état critique, b contrôle la forme de la surface de réponse, est le module de compressibilité plastique et pco représente la contrainte à l’état critique correspondant à l’indice de vides initial. rk est appelé le dégrée de frottement mobilisé et il est associé à la déformation déviatorique plastique. Sa loi d’évolution est la suivante :

(6)

est le multiplicateur plastique du mécanisme k et a :

(7)

(8)

a1, a2 et m sont des paramètres du modèle. Cette écriture permet la décomposition de l’écrouissage déviatorique dans les domaines pseudo-élastique, hystérétique et mobilisé. Chaque domaine est défini par les rayons respectifs relast, rhyst et rmob. La surface de réponse du mécanisme isotrope est définie par :

(9)

avec:

(10)

d est un paramètre du modèle et cmon contrôle l’écrouissage volumique. Le modèle considère une fonction de charge associée dans le plan déviatorique (k). La fonction de dilatance proposée par Roscoe (Roscoe et al., 1958) est considérée afin d’obtenir l’incrément de déformation déviatorique plastique de chaque mécanisme :

(11)

est l’angle caractéristique et un paramètre constant. Lopez-Caballero et Modaressi-Farahmand-Ravazi (2008) ont

proposé une classification des paramètres du modèle selon leur méthode d’estimation : directement et non directement mesuré.

3 ESSAIS EN LABORATOIRE

Les paramètres du modèle ont été calibrés pour les différents géomatériaux ferroviaires à partir des essais triaxiaux disponibles dans la littérature.

Tableau 1. Classification des paramètres du modèle élastoplastique ECP

Directementmesurés

Non directement mesurés

Elasticité Kref, Gref,ne, pref

Etat critique et Plasticité ’pp, pco, d B

Fonction de charge et écrouissage Isotrope a1, a2, ,

m, cmon

Taille de domaines rela, rhys,rmob, rela

iso

3.1 Matériau ballast

Des essais mécaniques sur le ballast ont été effectués par différents auteurs (Suiker 2002, Fortunato 2005 and 2010, Indraratna et al. 2011). Un comportement très dilatant à faible confinement a été mis en évidence. Lors des opérations de rénovation ou de bourrage des voies ferrées, une procédure de stabilisation du ballast est effectuée, puisque le ballast présente grand tassement et réarrangement des grains dans les premières centaines de cycles (Jeffs and Maritch, 87).

Du point de vue de la mécanique des sols, l’état initial du ballast évolue lors de la procédure de stabilisation. Du point de vue de modélisation numérique, cela revient à développer une procédure d’initialisation des contraintes qui tienne compte de cette évolution et qui considère le ballast déjà dans son état stabilisé. Même si la calibration des paramètres du modèle ECP a été effectuée selon les essais triaxiaux, le comportement de ce matériau a été choisi comme linéaire élastique dans une première approche, puisque le développement d’une procédure d’évolution de l’état initial est encore en cours.

3.2 Matériau de la couche intermédiaire

La couche intermédiaire est composée d’un mélange des matériaux provenant de la couche de ballast pollué et de la plateforme. Elle existe seulement dans les voies classiques anciennes, dans lesquels aucune couche de forme n’a été construite et dont la vitesse est actuellement limitée à 220 km/h. Trinh (2010) a étudié les propriétés mécaniques et hydrologiques de ce matériau à partir d’une série d’essais triaxiaux, à différentes teneurs en eau, et à partir d’essais d’infiltration. Même si la réponse mécanique a été très influencée par la teneur en eau, seulement les essais effectués à la saturation ont été considérés.

Les Figures 1 et 2 montrent la réponse déviatorique et volumique des paramètres calibrés pour le modèle ECP lors du chemin de contrainte de l’essai triaxial. Sur les mêmes figures, le résultat expérimental est aussi exposé. Les paramètres numériques obtenus sont cohérents et suivent le résultat expérimental de Trinh (2010).

Figure 1. Contrainte déviatorique q vs. déformation axiale 1

1231


3.3 Matériau de la plateforme

La plateforme est supposée raide. Afin de prendre en compte ce comportement, les paramètres proposés par Costa d’Aguiar (2008), obtenus à partir des essais mécaniques dans le sable de Toyoura (Fukushima and Tatsuoka, 1984), sont utilisés.

Figure 2. Déformation volumétrique v vs. déformation axiale 1

4 MODELE ELEMENT FINIS DE LA VOIE FERREE

Un modèle éléments finis 2D en déformation plane avec épaisseur (Saez, 2009) est considéré dans ce travail. La superstructure est composée du rail, des semelles, des traverses et du ballast non-confiné entre traverses. Le rail est modélisé par des éléments poutres et les autres composants par des éléments volumiques. Le modèle considère une structure de voie ferrée classique appartenant au réseau français, composée de ballast sain, ballast pollué, une couche intermédiaire et la plateforme. La Figure 3 montre la cellule périodique représentant la structure ferroviaire avec les respectives épaisseurs de couches. Ces épaisseurs sont issues de valeurs moyennes obtenues à partir des mesures de caractérisation in-situ déjà utilisées par les auteurs dans des travaux précédents (Alves Fernandes et al., 2012). Afin de réduire la réflexion des ondes P et S dans les bords du maillage, des couches absorbantes avec un modèle de comportement de type Kelvin-Voigt ont été utilisées dans les deux bords latéraux.

Figure 3. Cellule périodique représentative de la structure ferroviaire

Les propriétés mécaniques des matériaux de la superstructure sont définies dans le Tableau 2.

Tableau 2. Propriétés élastiques des composants de la superstructure

E [MPa]

Semelles 40 0.3

Traverses 34.103 0.3

Ballast non confiné 50 0.2

Ballast sain 150 0.2

Ballast pollué 250 0.3

La force appliquée par le train sur la voie est considérée ponctuelle et appliquée directement aux nœuds des éléments poutre. Le déplacement longitudinal de la force est effectué nœud à nœud à la vitesse voulue. Le chargement d’un boggie est modélisé à partir de deux charges de 85 kN représentant chaque roue, F1 et F2, écartées de 3 m. Les déplacements verticaux obtenus par le modèle élastique et élastoplastique sont comparés afin de monter l’effet du modèle de comportement.

5 RESULTATS

Le modèle élastique est analysé en premier afin de définir le cas de référence. La Figure 4 montre le déplacement vertical uz de différentes couches normalisé par le déplacement absolu maximum au niveau du rail. L’abscisse est recentrée sur la deuxième force F2, qui passe sur un point après la force F1. Une réduction rapide des déplacements entre le rail et la traverse est observée. Les semelles atténuent 60% du déplacement vertical total dans ce cas. La couche intermédiaire présente encore 28% du déplacement vertical total et la plateforme 20%.

F2 F1

Figure 4. Déplacement vertical normalisé pour les différentes couches de la structure ferroviaire à l’instant de force maximum appliqué sur une traverse (cas élastique ; vitesse de 150km/h)

La Figure 5 montre le déplacement vertical uz normalisé par le déplacement absolu maximum au niveau du rail pour le cas élastoplastique. Contrairement au cas élastique, un tassement au niveau du rail de l’ordre de 20% du déplacement vertical maximum est obtenu. Le déplacement vertical maximum de chaque essieu n’atteint plus la même valeur, le deuxième essieu (F2) présente une valeur maximale plus grande que le premier essieu (F1). Ce fait peut être expliqué par une augmentation du confinement sous F1 dû à F2 et pourtant une augmentation de la résistance du sol sous F1. Dans les voies réelles en utilisation courante, le tassement est observé à plus petite échelle. Comme discuté auparavant, une procédure numérique d’évolution de l’état initial des matériaux non-linéaires doit être développée afin d’assurer qu’ils soient au même état qu’en voie courante.

F2 F1

Figure 5. Déplacement vertical normalisé pour les différentes couches de la structure ferroviaire à l’instant de force maximum appliqué sur une traverse (cas élastoplastique ; vitesse de 160km/h)

1232


Une différence importante entre le chemin de contraintes de l’essai triaxial et des charges mobiles est le cisaillement. Différemment des points directement sous la charge appliquée, les points en amont et en aval expérimentent aussi une légère réduction du confinement et une augmentation de la contrainte déviatorique due au cisaillement. La présence de cisaillement mène aussi à la rotation de l’axe de contraintes. Différents auteurs ont montré (Grabe and Clayton, 2009, Ishikawa et al., 2011, parmi d’autres) que le ballast et les matériaux des plateformes ferroviaires répondent différemment à ces deux chemins de contraintes. Une augmentation de la déformation verticale permanente est observée dans le cas de rotation de l’axe des contraintes principales.

L’évolution de l’angle de rotation de l’axe des contraintes principales 1 en fonction du temps est étudiée selon la position dans la cellule périodique (Figure 3). La position entre traverses est montrée sur la Figure 6. On observe que dans ce cas la couche de ballast subi une rotation maximale de plus de 60° (30° positive et négative), valeur qui réduit avec la profondeur. La plateforme subit une rotation maximale de 20%. Par contre, les points directement sous la traverse sont soumis à des angles de rotation plus petits (Figure 7). Comme la voie ferrée possède un support discret, les points de la cellule périodique suivent des chemins de contrainte différents. La discontinuité du support mène à une concentration de la charge sous traverse, ce qui augmente la part de la contrainte verticale vis-à-vis du cisaillement et donc limite la valeur maximale de 1. Les points entre traverses subissent des contraintes normales moins importantes et de contraintes de cisaillement plus élevées, ce qui augmente l’angle de rotation. Une répartition progressive des contraintes normales et de cisaillement est obtenue en profondeur et cette non-homogénéité est moins observée dans la couche intermédiaire et la plateforme.

Figure 6. Rotation de l’axe des contraintes principales 1 pour différentes couches entre traverses (cas élastique; vitesse de 150km/h)

Figure 7. Rotation de l’axe des contraintes principales 1 pour différentes couches sous traverse (cas élastique; vitesse de 150km/h)

6 CONCLUSION ET TRAVAUX À VENIR

Les géomatériaux ferroviaires possèdent différents dégrées de complexité, leur réponse mécanique dépend de leur état initial, du chemin de contraintes imposé et pour les matériaux fins aussi

du dégrée de saturation. Seulement des modèles rhéologiques complexes peuvent prendre ces différents aspects en compte.

Le modèle élastoplastique ECP a été choisi afin de modéliser la réponse mécanique de la couche intermédiaire et de la plateforme. Les paramètres du modèle ont été calibrés à partir de résultats d’essais triaxiaux disponibles dans la littérature. Même si des paramètres ont été aussi calibrés pour le ballast, une attention particulière doit être portée à la modélisation de ce matériau. Il présente grande dilatance à faible confinement, condition rencontrée en voie avant stabilisation. Dans ce sens, une procédure pour faire évoluer l’état initial doit être développée, basée sur les procédures effectuées en voie.

Les résultats des simulations élastique et élastoplastique ont montré les atouts des modèles complexes : prise en compte des phénomènes irréversibles et évolution de l’état initial. L’influence du support discret sur les voies ballastées dans l’angle de rotation de l’axe de contraintes principales a été aussi étudiée. Le support contribue à des chemins de contraintes différents selon la position dans la cellule périodique. L’angle de rotation est plus petit sous traverse vis-à-vis entre traverse pour les couches proches du support. Une homogénéisation progressive des contraintes est obtenue en profondeur.

7 REFERENCES

Alves Fernandes, V., Costa d’Aguiar, S., Lopez-Caballero, F. 2012. Influence of soil properties variability on the railway track response under moving load. Proceedings of the 2nd Int. Conf. on Tansportation Geotechnics (en anglais)

Aubry, D., Hujeux, J.-C., Lassoudière, F., and Meimon, Y. 1982. A double memory model with multiple mechanisms for cyclic soil behaviour. Proceedings of the Int. Symp. Num. Mod. Geomech.,Balkema, 3-13. (en anglais)

Costa d’Aguiar, S. 2008. Numerical modelling of soil-pile axial load transfer mechanisms in granular soils. PhD, Ecole Centrale Paris, France. (en anglais)

Fortunato, E. 2005. Renovação de plataformas ferroviarias – Estudos relativos à capacidade de carga. PhD, Universidade do Porto, Portugal (en portugais).

Fortunato, E., Pinelo, A., Fernandes, M. M. 2010. Characterization of the fouled ballast layer in the substructure of a 19th century railway track under renewal. Soils and Foundation, 50(1):55-62. (en anglais)

Fukushima, S. and Tatsuoka, F. 1984. Strength and deformation characteristics of saturated sand at extremely low pressures. Soils and Foundations, 24(4):30–48. (en anglais)

Gräbe, P. J., Clayton, C. R. I. 2009. Effects of Principal Stress Rotation on Permanent Deformation in Rail Track Foundations. Journal of Geotechnical nad Geoenvironmental Engineering 135(4):555-565. (en anglais)

Hujeux, J.-C. 1985. Une loi de comportement pour le chargement cyclique des sols. In V. Davidovici, editor, Génie Parasismique, Presses ENPC, France, pp. 278-302. (en anglais)

Indraratna, B., Salim, W., Rujikiatkamjorn, C. 2011. Advanced Rail Technology – Ballasted Track. CRC Press/Balkema. (en anglais)

Ishikawa, T., Sekine, E., Miura, S. 2011. Cyclic deformation of granular material subjected to moving-wheel loads. Canadian Geotechnical Journal 48(5):691-703. (en anglais)

Jeffs, T., Marich, S. 1987. Ballast characteristics in the laboratory.Conference on Railway Engineering, Perth, pp. 141-147. (en anglais)

Lopez-Caballero, F., Modaressi-Farahmand-Ravazi, A. 2008. Numerical simulation of liquefaction effects on seismic SSI. Soil Dynamics and Earthquack Engineering 28(2): 85-98. (en anglais)

Roscoe, K. H., Schofield,A; N., Wroth, C. P. 1958. On the yielding of soils. Géotechnique 8(1), 22–52. (en anglais)

Suiker, A. 2002. The mechanical behaviour of ballasted railway tracks.PhD, Delft University of Technology, Netherlands. (en anglais)

Saez, E. P. R. 2009. Dynamic non-linear soil-structure interaction.PhD, Ecole Centrale Paris, France. (en anglais)

Trinh, V. 2010. Comportement hydromécanique des matériaux constitutifs de plateformes ferroviaires anciennes. PhD, Ecole Nationale des Ponts et Chaussées, France.

1233

LGV EST lot 41 : tassements calculés puis mesurés sous remblais

LGV EST section 41 : measured and calculated settlements under embankments

Boutonnier L., Hajouai F. Egis Géotechnique, Seyssins, France

Bacar Fadhuli N., Gandille D. Guintoli, Saint-Etienne du Grès, France,

RÉSUMÉ : Dans le cadre du projet de la LGV EST lot 41, six remblais présentant une problématique de consolidation ont étéinstrumentés. Les tassements et temps de consolidation mesurés sont comparés à ceux calculés dans les études d’exécution. Il apparaît que la méthodologie mise en œuvre pour les études d’exécution donne de très bons résultats en réalisant deux correctionssystématiques des procès-verbaux des essais oedométriques : (i) estimation de la contrainte de préconsolidation à partir de Cu ; (ii) correction de Cv mesuré en laboratoire en le multipliant par 10.

ABSTRACT: Six embankments from the “LGV EST lot 41” project have been monitored to measure the settlements and the time ofconsolidation. The comparison of the calculated and measured settlements and time of consolidation give good results. The oedometric tests were corrected according to the following methodology: (i) the preconsolidation pressure of the soil is estimatedfrom the undrained cohesion Cu; (ii) the coefficient of consolidation Cv used in the design is ten times the Cv measured withlaboratory test.

MOTS-CLÉS : tassements, remblais, coefficient B, instrumentation, pressions interstitielles, cohésion non drainée, surconsolidation

KEYWORDS: settlement, embankment, B coefficient, instrumentation, pore pressure, undrained cohesion, overconsolidation

1 INTRODUCTION

Les lignes ferroviaires à grande vitesse admettent très peu de tassements après la mise en service. Sur le réseau ferré français, les critères de dimensionnement sur les opérations de lignes nouvelles à grande vitesse en cours en 2012 (LGV EST 2ème phase, LGV SEA, LGV BPL) admettent 10 cm de tassement résiduel sur 25 ans avec une vitesse inférieure à 1 cm par an dès réception du génie civil. L’estimation des tassements et des temps de consolidation dans les sols fins en assise de remblai dépend de nombreux paramètres : effet de la non saturation et du coefficient B de Skempton (Skempton, 1954 ; Tavenas et Leroueil, 1980 ; Boutonnier, 2009), anisotropie de perméabilité et présence de fissures dans le sol (Mitchell, 1992 ; Skempton et Northey, 1952), difficulté pour prélever des échantillons intacts et effet sur la contrainte de préconsolidation ’p (Bat et Blivet, 2000). Dans le cadre de la LGV EST 2ème phase (lot 41 attribué au groupement d’entreprises piloté par Guintoli), huit remblais présentent une problématique de tassements et de temps de consolidation. Egis Géotechnique a pris un soin particulier pour évaluer la contrainte de préconsolidation de manière réaliste, en se basant sur les essais oedométriques, mais également sur des corrélations entre la cohésion non drainée et la contrainte de préconsolidation (Leroueil et al., 1985). Néanmoins, pour les ouvrages les plus sensibles (cohésion non drainée Cu entre 5 et 30 kPa avec présence de tourbes et risque de fluage), les études ont conduit à préconiser et réaliser des purges totales des sols, les critères dimensionnants étant la stabilité au glissement et le fluage associé aux problèmes de stabilité à court terme. Pour les six remblais sur des sols intermédiaires qui n’ont pas été purgés (Cu = 30 à 100 kPa), les tassements et les temps de

consolidations calculés et mesurés sont comparés dans la présente communication afin d’évaluer la méthodologie prise en compte dans les études.

2 LES ETUDES GÉOTECHNIQUE D’EXÉCUTION

2.1 Contexte géologique du lot 41

Le lot 41 se trouve en bordure Est du bassin Parisien. Les formations anciennes recoupées sont des roches sédimentaires du Trias marno-calcaires ou franchement argileuses ou marneuses (Marnes bariolées du Keuper, Grès à roseaux, Marnes irisées inférieures, Marnes bariolées de la Lettenkohle, Couches blanches et grises). La frange altérée de ces roches présente une compressibilité non négligeable. Par ailleurs, ces formations sont souvent recouvertes en fond de vallon par des sols fins alluvionnaires compressibles.

2.2 Méthodologie des études d’exécution

Dans le cadre de la LGV EST lot 41, Egis Géotechnique réalisait les études géotechniques d’exécution des ouvrages pour le compte du groupement d’entreprises piloté par Guintoli. Au droit de chaque remblai, l’ensemble des données géotechniques disponibles ont été synthétisées et corrélées afin de définir un modèle au droit de chaque profil de calcul représentatif (voir tableau 1). Dans un premier temps, un soin particulier a été pris pour estimer correctement la cohésion non drainée Cu dans les sols fins en corrélant les données pressiométriques, pénétro-métriques et scissométriques. La cohésion Su mesurée au

1234


scissomètre a été corrigée (Pilot, 1972 ; Bjerrum, 1972) pour estimer Cu. Dans un deuxième temps, les contraintes de préconsolidation mesurées à l’oedomètre ont été comparées à celles estimées d’après les corrélations présentées par Leroueil et al. (1985) et qui conduisent à retenir, de manière prudente :

’p = Cu/0,35 (1)

Ce type de relation peut aussi être retrouvé de manière théorique (Boutonnier et Virollet, 2001) en utilisant des lois de comportement élasto-plastique avec écrouissage qui prennent en compte la surconsolidation du sol dans des conditions anisotropes (correspondant aux dépôts sédimentaires naturels : Ylight, Mélanie,…). La plupart du temps, les essais oedométriques sont apparus remaniés et / ou les contraintes de préconsolidation mesurées sous-estimées. Elles ont donc été corrigées en fonction de Cu comme illustré dans le tableau 1.

Enfin, les coefficients de consolidation verticale Cv et radiale Cr(cas de drains verticaux) ont été estimés à partir des essais de laboratoire mais en appliquant les corrections suivantes :

Cv_in_situ = 10.Cv_labo (2)

Cv_labo : coefficient de consolidation verticale mesuré sous la charge finale du remblai en laboratoire Cv_in_situ : coefficient de consolidation verticale estimé in-situ sous la charge finale du remblai

Cr_in_situ = Cv_in_situ (3)

Cr_in_situ : coefficient de consolidation radiale estimé in-situ sous la charge finale du remblai

Ces corrections sont justifiées par les retours d’expérience observés dans les sols compressibles (Leroueil et al., 1985) et dans les marnes altérées sur les projets Egis sur la LGV EST lot 32 (Boutonnier et Guerpillon, 2005) et sur la LGV Rhin Rhône lot B3 (résultats non publiés au moment de la rédaction de cette communication). D’un point de vue plus fondamental, plusieurs pistes peuvent expliquer ces observations : fissuration du sol (Mitchell, 1992 ; Skempton et Northey, 1952), présence de passées plus perméables non détectables, même avec un pénétromètre, etc. Pour les sols restant dans le domaine surconsolidé sous la charge finale du remblai, il a été considéré que le tassement se dissipait au fur et à mesure de la construction du remblai. Cette hypothèse est justifiée compte tenu : (i) du coefficient de consolidation Cv qui augmente avec le module oedométrique dans le domaine surconsolidé, (ii) du coefficient B de Skempton (1954) qui diminue lorsque le module oedométrique augmente dans le domaine surconsolidé, (iii) des retours d’expérience chantier qui montrent peu ou pas de montée des pressions interstitielles dans les sols surconsolidés.

3 LES TASSEMENTS ET TEMPS DE CONSOLIDATION MESURÉS

3.1 Principe des mesures effectuées

Les mesures ont été effectuées par le groupement d’entreprises pilotées par Guintoli. Les profils en travers instrumentés sont équipés au minimum de 3 tassomètres ou d’un profilomètre comme illustré sur la figure 1. Parfois, des cellules de pressions interstitielles à différentes profondeurs sont mises en œuvre.

Dans ce dernier cas, un piézomètre de référence est mis en œuvre à coté du remblai pour suivre les variations naturelles des niveaux d’eau en relation avec la nappe.

CPI tassomètre

profilomètre

Figure 1 : principe d’instrumentation des assises de remblais pour contrôler l’amplitude des tassements, les pressions interstitielles et les

temps de consolidation

3.2 Tassements et temps de consolidation mesurés

Pour chaque profil instrumenté, les tassements et temps de consolidation mesurés et calculés sont comparés dans le tableau 2. Lorsque le profil de calcul présente une hauteur de remblai différente du profil instrumenté, le calcul a été repris avec la hauteur réelle de remblai pour faciliter la comparaison entre mesures et calculs.

1235


Tableau 1. Synthèse des modèles géotechniques au droit de chaque profil de calcul. Les valeurs de la contrainte de préconsolidationissue des essais oedométriques et corrigée en fonction de Cu sont présentées dans le tableau

remblai km formation z(m)

e0 Cc Cs ceCu

(kPa)

'pmesuré(kPa)

'precalé(kPa)

Cv_labo

(m²/s)Cr_retenu

(m²/s)

limons 0 à 2 0.76 0.17 0.019 55 71 157 3.10-8 3.10-7

Argiles de Chanville altérée 2 à 7 0.87 0.18 0.022 100 100 286 2.10-8 2.10-7

purge 0 à 3

grès à roseaux (altération) 3.5-6.5 0.76 0.188 0.023 55 65 157 1.10-8 1.10-7

limons argileux 0 à 1 0.71 0.156 0.012 50 102 140 4.10-7

grès à roseaux (altération) 1 à 6.5 0.71 0.156 0.012 67 102 191 5,8.10-8 4.10-7

grès à roseaux (altération) 6.5 à 12 0.81 0.208 0.013 96 81 274 3,8.10-8 4.10-7

alluvions récentes 0 à,3 0.95 0.226 0.01 0.009 30 30 86 1.3.10-7 4.10-7

marnes irisées inf. altérées 3 à 8 0.948 0.226 0.01 0.009 72 205 205 4.10-7

purge 0 à 1.5

marnes irisées inf. altérées 1.5 à 8 0.84 0.182 0.02 63 59(29 à 101) 180 1.3.10-7 1.3.10-6

limons argileux 0 à 1 0.527 0.126 0.014 80 95 229 2.7.10-7 2.7.10-6

marnes irisées inf. altérées 1 à 6.5 0.527 0.126 0.014 80 95 229 2.7.10-7 2.7.10-6

purge 0 à 2

marnes irisées inf. altérées 2 à 6.5 0.527 0.125 0.014 80 95 229 2.7.10-7 2.7.10-6

limons argileux 0 à 2 0.739 0.186 0.047 40 90 114 2.0.10-8 2.0.10-7

Marnes bariolées altérées 2 à 6 0.739 0.186 0.047 100 285 2.0.10-7

purge 0 à 1

couches bl. et grises alt 1 à 6 0.646 0.193 0.016 80 98 229 1.57.10-7 1.57.10-6

couches bl. et grises alt 6 à 10

couches bl. et grises alt 0 à 6 0.646 0.193 0.016 80 98 229 1.57.10-7 1.57.10-6

couches bl. et grises alt 6 à 10

limons 0 à 1.3

couches bl. et grises alt 1.3 à 5 0.5 0.116 0.009 65 79 185 1.57.10-7 1.57.10-6

couches bl. et grises alt. 5 à 12 0.5 0.116 0.009 82 235 1.57.10-7 1.57.10-6

calcul tassement par méthode pressiométrique W = ..h/Em, = 0.5, Em = 10000kPa352+680

pas de tassement calculé

353+460

RBT 546 347,600

calcul tassement par méthode pressiométrique W = ..h/Em, = 0.5, Em = 10000kPa

352,510352,910


RBT 555

RBT 527 341,200


RBT 527 - PRA 41230 341,362


RBT 505 secteur 3 331,880

RBT 525 340,270


331,520331,780

RBT 505 secteur 2

RBT 503modèle 1 328,253

RBT 505secteur 1c 330,700

Tableau 2. Synthèse des modèles géotechniques au droit de chaque profil de calcul. Les valeurs de la contrainte de préconsolidationssue des essais oedométriques et corrigée en fonction de Cu sont présentées dans le tableau i

remblai km Hremblai(m)

tassements mesurés (cm)

tassements calculés avec correction 'p

(cm)

tassement calculés sans correction 'p

(cm)

(durée consolidation prévue) / (durée

consolidation mesurée)dispositions constructives

RBT 503- modèle 1 328,265 8.0 10 7 28 3.0 drains 1.5 x 1.5m

RBT 505 secteur 1c 330,710 8.0 9 6 17 3.0 drains 1.2 x 1.2m

331,036 7.5 6 6 43 1.0 drains 2.5 x 2.5m

331,828 10.5 10 10 46 drains 1.2 x 1.2m + purge 3mRBT 505 secteur 2

PRA 331,540 9.0 4 9 27 instantané drains 2.5 x 2.5m

RBT 505 secteur 3 331,900 10.5 22 23 38 montée trop lente drains 1.2 x 1.2m

340,240 8.5 7 9 37 1.5 drains 2.5 x 2.5 m + purge 1,5m

340,300 8.0 4 8 35 2.0 drains 1.5 x 1.5m + purge 1,5 m

RBT 527 341,200 7.5 6 5 19 0.60 drains 2.5 x 2.5m + purge 0.7m

341,285 (C0) 10.0 5 7 19 pas comparable drains 2.5 x 2.5m + purge 2m

341,340 (C3) 10.5 11 8 20 2.60 drains 2.5 x 2.5m + purge 2m

RBT 546 347,610 4.8 5 4 pas comparable drains 2.5 x 2.5m + purge 3m

352,418 11.0 7 11 31 pas comparable drains 1.5 x 1.5m + purge 1m

352,520 20.0 24 28 48 pas comparable drains 1.5 x 1.5m + purge 1m

352,885 9.0 3 7 25 instantané drains 1.5 x 1.5m

352,932 9.0 4 7 25 instantané drains1.5 x 1.5m

353,447 12.0 22 16 1.00 drains 1.5 x 1.5m + purge 1m

RBT 555

RBT 527-PRA 41230

RBT 525

RBT 505 secteur 2

1236


4 ANALYSE DES RÉSULTATS

Les amplitudes des tassements mesurés et calculés sont très bien corrélées comme illustré sur la figure 2. Si la contrainte de préconsolidation des essais oedométriques avait été prise en compte sans la correction de l’équation (1), l’amplitude des tassements aurait été surestimée de 100 à 500%. Ce résultat valide la méthodologie de détermination de la contrainte préconsolidation ’p.

y = 1.0042xR2 = 0.7364

0.00

5.00

10.00

15.00

20.00

25.00

30.00

35.00

40.00

45.00

50.00

0.00 10.00 20.00 30.00 40.00 50.00

tassements mesurés (cm)

tass

emen

ts c

alcu

lés

(cm

)

contrainte préconsolidationcorrigée

contrainte préconsolidationnon corrigée

Linéaire (contraintepréconsolidation corrigée)

Figure 2 : comparaison des tassements mesurés et calculés. Chaque point correspond à un profil instrumenté différent. La contrainte de

préconsolidation corrigée ’p = Cu/0,35 permet d’obtenir une estimation réaliste des tassements

Les paliers de stabilisation des tassements sont en général nets et aucune évolution n’est observée sur la période de quelques semaines à quelques mois qui suit la fin de la construction des remblais. En ce qui concerne la consolidation, les paramètres Cv_in_situ et Cr_in_situ retenus dans les études d’exécution permettent d’estimer de façon suffisamment sécuritaire les temps de consolidation. Néanmoins, la construction de beaucoup de remblai a été lente et / ou par étapes ce qui fausse un peu l’analyse par rapport aux hypothèses d’une construction « instantanée » des études d’exécution. Par ailleurs, pour les phases de montée rapide des remblais, les montées de pression interstitielles sont très faibles et le tassement « instantané » est important.

5 CONCLUSION

La démarche de comparaison des tassements mesurés et calculés valide, sur ce chantier, la méthode de correction de la contrainte de préconsolidation in situ ’p en fonction de la cohésion non drainée Cu. Dans la détermination de ’p, les incertitudes liées à l’utilisation de corrélations semblent plus faibles que celles introduites par le remaniement des échantillons dans les opérations de prélèvement.

L’aspect cinétique de la consolidation reste à analyser plus en détail pour déterminer la part de déformation non drainée (effet du coefficient B de Skempton) et drainée.

Le modèle développé par Boutonnier (2007, 2009) permet de modéliser ces phénomènes en calculant le coefficient B de Skempton à partir d’hypothèses sur la non saturation parfaite des sols, même sous la nappe. Cette analyse sera effectuée et

présentée dans une publication ultérieure. Elle permettra aussi de faire un recalage de Cr_in_situ, l’utilisation de la méthode d’Asaoka (1978) étant délicate pour les cas de montée lente des remblais.

6 REFERENCES

Asaoka A., 1978. Observational procedure of settlement prediction. Soils and foundations, vol. 18 (4), 87-101.

Bat. A., Blivet JC (2000). Incidence de la procédure de prélèvement des sols fins sur les caractéristiques géotechniques mesurées en laboratoire. Revue Française de Géotechnique, n°91, 3-12.

Bjerrum L. (1972). Embankments on soft ground. Proc. ASCE Specialty Conference on Earth and Earth-supported Structures, Purdue University, vol. II, 1-54.

Boutonnier, L. (2007). Comportement hydromécanique des sols fins proches de la saturation. Cas des ouvrages en terre : coefficient B, déformations instantanées et différées, retrait / gonflement. Thèse INPG Grenoble soutenue le 23 octobre 2007. http://geotec-luc.blogspot.com/

Boutonnier L. (2009). Prise en compte de la non saturation dans l'interprétation de l'essai oedométrique. Proceedings XVIIème congrès international de Mécanique des Sols et de la Géotechnique,Alexandrie, 5-9 octobre 2009

Boutonnier L., Guerpillon Y. (2005). Reconnaissances géotechniques et critères en déformation dans la conception des ouvrages en terre des lignes ferroviaires à grande vitesse. Géoline 2005, Lyon, 23-25 mai 2005. Editeurs M. Arnould et P. Ledru, BRGM, Orléans, 1 volume, 298 pages.

Boutonnier L., Virollet M. (2001) – Nouvelles tables de poussée et de butée dans les sols surconsolidés. Comptes rendus du XVème congrès international de Mécanique des Sols et de la Géotechnique, Istanbul, Balkema, p. 1095-1098.

Leroueil S., Magnan J.P., Tavenas F. (1985). Remblais sur argiles molles. Technique et documentation Lavoisier.

Mitchell J.K. (1992). Fundamentals of soil behaviour. Second Edition. John Wiley & Sons.

Pilot G. (1972). Rupture d’un remblai sur sols compressibles. Bulletin de liaison des L.P.C n°61, 1972.

Skempton A.W., Northey R.D. (1952). The sensitivity of clays. Geotechnique, vol. 3, n°1, 30-53.

Skempton, A.W. (1954). The pore-pressure coefficients A and B. Géotechnique , vol. 4, p. 143-147

Tavenas F., Leroueil S. (1980). The behaviour of embankments on clay fondations. Canadian Geotechnical Journal, vol. 17 (2), 236-260.

1237

Recent developments in pavement foundation design

Développements récents dans la conception des fondations des chaussées

Brown S.F., Thom N.H. University of Nottingham, UK.

ABSTRACT: For many years the design of flexible pavements was heavily dependent on use of the California Bearing Ratio (CBR) of the soil subgrade within an empirical method that dated back to the 1920’s. Recent changes to design practice in the UK have at last recognised that the key parameters are the resilient stiffness modulus (Er) and the resistance to permanent (plastic) deformation under repeated loading of the subgrade and foundation materials. The foundation platform on which the asphalt layers are placed ischaracterised by a ‘Surface Modulus’, which is measurable on site using a dynamic plate test. Resistance to rutting under construction traffic is determined from proof loading with a truck. The surface modulus is a function of Er for the individual foundation layers. To aid the selection of materials, a simplified version of the repeated load triaxial test has been developed to quantify both resilient and plastic strain characteristics. This apparatus is known as the PUMA and is similar to the Springbox device that has been successfullyused in recent years. It can accommodate both soils and lightly stabilised materials for use in sub-base construction.

RÉSUMÉ : Pendant de nombreuses années la conception des chaussées souples a été fortement tributaire de l'utilisation de l'indice portant californien (CBR) de la plate-forme du sol par le biais d'une méthode empirique qui remontait à 1920. Les changementsrécents dans la pratique du dimensionnement au Royaume-Uni ont enfin reconnu que les paramètres clés sont le module de rigiditéélastique (Er) et la résistance à la permanente déformation (plastique) sous charge répétée des matériaux du sol de fondation et de fondation. La plate-forme de fondation sur laquelle les couches d'asphalte sont placées se caractérise par un «Module Surfacique», qui est mesurable sur le site en utilisant un essai à la plaque dynamique. La résistance à l'orniérage sous trafic est déterminée à partir d’un essai de chargement référence avec un camion. Le module de surface est fonction de Er pour les couches de fondation individuelles. Pour faciliter le choix des matériaux, une version simplifiée du test triaxial à charge répétée a été développée pour quantifier à la fois les caractéristiques de déformation résilientes et plastiques. Cet appareil est connu comme le PUMA et est une version améliorée de l'appareil Springbox qui a été utilisé avec succès ces dernières années. Il peut être utilisé à la fois pour des sols et des matériaux stabilisés utilisés comme couche de fondation.

KEYWORDS: Pavement foundations, laboratory testing, repeated loading, resilient modulus, plastic strain, design.

1 INTRODUCTION.

The California Bearing Ratio (CBR) continues to be used to characterise subgrades in modern flexible pavement design methods in most parts of the World (eg. Highways Agency, 2006, AASHTO, 2008) despite the fact that it has long been recognised as, at best, a simple index of undrained shear strength. The mechanical properties of subgrade soils and of capping and sub-base materials that are relevent to pavement design are, however, the resilient stiffness modulus (Er) and the resistance to plastic strain under repeated loading, concepts that are gradually being incorporated into pavement design methods. Developments both in laboratory and field testing have assisted this process. This has allowed a wider choice of materials to be used in construction within a framework of performance related specifications. While field testing using some form of dynamic plate load represents the end-product performance test for resilient characteristics on the as-built foundation, engineers also require element tests that can be used on candidate materials for the individual layers and the subgrade as part of the design process. Combining this information within an elastic analysis of the foundation can deliver a target value for what is known as the ‘Surface Modulus’, which is the parameter determined from the dynamic plate loading test.

The repeated load version of the triaxial test was developed in the 1960s to determine resilient modulus but it has proved time consuming to use and has essentially remained a research tool. By contrast, simplified methods of testing asphalt were developed for use in engineering practice in the 1990s against a

sound research background (Brown, 1995). These have been adopted within European Standards. A similar philosophy has since been used to provide an appropriate test method for soils, granular materials and lightly stabilised materials.

2 PAVEMENT FOUNDATION DESIGN

The concept of designing a pavement in two stages was proposed by Brown and Dawson (1992). The first stage involves all layers up to and including the sub-base, known as the pavement foundation, and requires that a relatively small number of heavy wheel loads should be accommodated from construction traffic. The second stage involves design of the completed pavement for the long term incorporating bound layers over a foundation of known, measured effective stiffness; the ‘Surface Modulus’.

The UK Highways Agency (2009) have introduced an interim design guide for pavement foundations following these principles (Highways Agency, 2009). Four classes of foundation are defined in terms of minimum values for the Surface Modulus; Class 1 ≥ 50MPa, Class 2 ≥ 100MPa, Class 3 ≥ 200MPa and Class 4 ≥ 400MPA. The top two classes generally require the use of stabilized aggregates of some type. A similar approach is used in France (LCPC and SETRA, 1994).

The definitive test method for measuring the Surface Modulus is the Falling Weight Deflectometer (FWD) which applies load pulses that are simulative of moving heavy wheel

1238


loads. Given the non-linear resilient characteristics of soils and granular materials, it is important to use an appropriate stress level when testing. If a hand-held Lightweight Deflectometer (LWD) is used, the stress level will be lower and the zone of influence smaller so a correlation exercise with the FWD has to be carried out and this is site specific.

In order to assess resistance to rutting, a standard truck of known axle load is operated on the completed foundation to deliver the equivalent of 1000 standard 80kN axle loads. The accumulated permanent deformation in the wheel tracks is measured and assessed against allowable values which depend on the layer thicknesses and material types.

During the design phase, it is convenient to use a simple laboratory test to determine the resilient characteristics of potential materials for use in the foundation. This provides input data to the design analysis which can generate a combination of layers of particular thicknesses with an overall effective surface modulus to satisfy the design specification for one of the foundation classes. A site trial involving one or more potential solutions can then provide confirmation that the target surface modulus has been achieved.

3 LABORATORY TESTING

3.1 Background

The testing philosophy is an extension of that adopted for asphalt, which resulted in the Nottingham Asphalt Tester (NAT), (Cooper and Brown, 1989) being developed as a simple low cost facility for use in engineering practice. It consisted of a pneumatically driven loading frame into which different test modules could be placed in order to measure various mechanical properties of an asphalt test specimen under repeated loading, including stiffness modulus, resistance to fatigue cracking and to permanent strain accumulation.

Using similar principles to those applied by Semmelink and de Beer (1995) for their K-Mould, Edwards et al (2005) developed a test module for unbound and weakly stabilized materials (known as a ‘Springbox’) that fitted into the NAT test frame. This involved a 170mm cubical specimen contained in a box with one opposite pair of vertical faces rigid and the other pair spring loaded to simulate the confining situation insitu for an element of compacted aggregate. Vertical loading was applied through a square loading platen at a frequency of 1Hz by the pneumatic actuator and measurements of both resilient and plastic deformations were taken under repeated loading.

3.2 The Precision Unbound Materials Analyzer (PUMA)

One of the perceived disadvantages of the Springbox is its square cross-section, raising the possibility of non-uniform compaction and non-homogeneous stress conditions. The PUMA, shown in Figure 1, therefore, adopts a similar 150mm diameter circular shape to that of the K-Mould but with a slightly increased height of 150mm. Like the K-Mould it is confined within eight curved wall segments. The specimen, which is compacted using standard equipment (e.g. a vibrating hammer) at a desired water content, is then loaded on its top surface by a circular platen. Side walls are confined within a rubber-lined steel band, the rubber providing the possibility of wall movement under load, simulating the elasticity of surrounding material in-situ. Under repeated vertical loading a residual horizontal stress will accumulate, typically between 10kPa and 50kPa, again simulating the in-situ condition.

Measurements are taken as follows: a) vertical load from a load cell; b) vertical displacement of the top surface from LVDTs, optionally inserted through holes in the top platen; c) horizontal strain in the steel confining band from a strain gauge. This last measure is directly proportional to the stress in the steel band and, therefore, to the horizontal stress in the

specimen. It is also proportional to the horizontal strain in the specimen, via the known compressibility of the rubber lining. Thus, while only vertical stress is controlled, vertical and horizontal stress and strain are all monitored during the test.

Figure 1. The Precision Unbound Material Analyzer (PUMA)

3.3 Analysis of Test Conditions

In order to maximize the information that could be derived from a single test during the development of the equipment, specimens were loaded in four stages each involving 1000 load applications. Vertical stress levels up to about 250kPa were used, above an initial preload of typically 5kPa. Since horizontal stress is not controlled, values varied according to the material tested. Figure 2 illustrates stresses measured during a typical test on a natural gravel.

0

50

100

150

200

250

0 1000 2000 3000 4000

Stress during test (kPa

)

Number of load applications

Vertical stress ‐minVertical stress ‐maxHorizontal stress ‐minHorizontal stress ‐max

Figure 2. Stresses measured during a typical PUMA test

As a first approximation, friction between the walls and the

specimen could be neglected and the measured stresses and strains converted directly into a stiffness modulus and Poisson’s ratio, treating the material as a linear elastic solid. This is the method specified in EN 13286-7 (CEN, 2004) in relation to interpretation of triaxial data. Nevertheless, it is self-evidently inaccurate to ignore friction between the walls and the specimen as well as friction against the upper and lower platens. Direct measurements taken during development of the PUMA equipment suggested that a coefficient of friction of around 0.5 could be expected. The effect of this would be to transfer vertical load to the walls, reducing the stress at the lower platen. Similarly, platen friction would mean that not all the internal horizontal stress within the specimen would reach the walls and be measured. An approach to take account of this has been outlined by Thom et al (2012) who developed the following correction equations.

v(corrected) = v(measured) – 0.5μh (hmax+hmin)/r (1) h(corrected) = h(measured) + 2μr (vmax+vmin)/15h (2)

1239


where: vmax = vertical stress at full load vmin = vertical stress on unloading (slightly > 0) hmax = horizontal stress at full load

hmax = horizontal stress on unloading v = vmax – vmin h = hmax – hmin

μ = coefficient of friction (0.5 assumed) h = specimen height (approximately 150mm) r = specimen radius (75mm)

With these corrected stresses, the stiffness modulus and Poisson’s ratio can be calculated with greater accuracy. Figure 3 shows a typical set of results from a PUMA test on a gravel aggregate. On the figure, these are compared with a set of predictions based on equations for the resilient non-linear stress-strain behaviour of a gravel aggregate contained in Thom (1988), providing a degree of added confidence that the measurements and their interpretation are approximately correct.

0

50

100

150

200

250

300

0 1000 2000 3000 4000

Mod

ulus (MPa

) / v x 10

0


Siffness modulus ‐ testStiffness modulus ‐ predictionPoisson's ratio (v) ‐ testPoisson's ratio (v) ‐ prediction

Figure 3. Stiffness modulus and Poisson’s ratio – corrected for friction

Figure 4 shows some typical results for the accumulation of

permanent strain under the four stages of repeated loading. Use of data of this type is not presently catered for in the Highways Agency’s design method but it can be useful at the material selection stage for use on a comparative basis.

0

0.5

1

1.5

2

2.5

0 1000 2000 3000 4000

Vertical strain (%

)


Vertical stress: 5‐23kPaVertical stress: 5‐43kPaVertical stress: 5‐85kPaVertical stress: 7‐175kPa

Figure 4. Accumulation of permanent strain in a typical PUMA test

4 USE OF DATA

4.1 Selection of Appropriate Stress Level

A key requirement for a realistic stiffness modulus test is that the stress conditions should be representative of those in the pavement. For a completed pavement, an estimate of such conditions can be derived from multi-layer linear elastic analysis and this was carried out for two cases, one with 140mm of asphalt (Case 1) and the other with 240mm (Case 2), assuming a temperature of around 20C. At mid-depth in a

200mm base layer below the asphalt, the computed stresses due to a 100kN axle (50kN wheel) load were found to be as shown in Figure 5.

0

50

100

150

200

250

300

0 500 1000 1500 2000

Compressive

stress (kPa

)

Base Stiffness Modulus (MPa)

Case 1: vertical stressCase 1: horizontal stressCase 2: vertical stressCase 2: horizontal stress

Figure 5. Predicted traffic induced stresses at mid-depth in the base layer

Although the stresses in Figure 5 only represent a limited

range of examples, they suggest the sort of stress levels that should be applied to achieve a realistic stiffness modulus for pavement design. For example, using the data in Figures 2 and 5 and taking the case of a 300MPa base layer, the stress conditions would be similar to Stage 3 of the test routine in the case of a 140mm asphalt pavement and Stage 4 with 240mm of asphalt.

It is also necessary to consider the case of insitu testing using an LWD, which typically applies a vertical stress of about 100kPa, and for which it is difficult to predict the appropriate horizontal stress due to the non-linear nature of granular materials. Nevertheless, adopting an earth pressure coefficient approach the situation is akin to an active rather than passive state, in which case the ratio of vertical to horizontal stress is likely to be of the order of 4 to 5. This gives a likely horizontal stress of 20kPa to 25kPa near to the surface under an LWD load. Since it is known that the horizontal stress state has a controlling influence on measured stiffness modulus, similar to the effect of confining stress in a triaxial test, it is logical to ensure that this is correctly simulated. This suggests that either Stage 1 or 2 of the proposed test routine is likely to give a stiffness modulus suitable for inclusion in a foundation surface modulus prediction. Stage 1 is likely to be most appropriate for the uppermost layer, while Stage 2 may represent conditions in an underlying foundation layer.

4.2 Design example

By way of example, the data shown in Figure 3 have been used to generate a design for a UK Highways Agency Class 2 foundation (Highways Agency, 2009) which requires an equivalent surface modulus of 100MPa under a 240mm thick asphalt layer. The designation ‘Class 2’ represents the condition in the finished pavement and it is, therefore, appropriate to use Stage 4 of the PUMA test, which gives a material stiffness modulus of 150MPa.

It is also necessary to evaluate the stiffness modulus of the subgrade soil. This can also be carried out in the PUMA, again taking Stage 4 conditions for the completed pavement and it is assumed here that this gave a stiffness modulus of 60MPa.

It is now possible to use multi-layer linear elastic analysis to determine the equivalent foundation modulus under the completed pavement. It is suggested here that the most appropriate design methodology is to compare computed asphalt tensile strains (the asphalt fatigue cracking design criterion) under a given load, first with the intended 2-layer foundation, then with a single layer only, representing the equivalent foundation with a single stiffness value. On this basis, 260mm of the gravel material in Figure 3 is required.

1240


The surface modulus obtained from testing with an LWD will however be significantly less than 100MPa due to material non-linearity. Stage 1 of the PUMA test generated a stiffness of just 50MPa for the gravel. For the subgrade, had it been tested, Stage 2 would be appropriate due to the increased level of confinement at depth which, if Stage 4 had given 60MPa, would be likely to have given a value of around 40MPa. Turning to multi-layer linear elastic analysis once more, with 260mm of material of 50MPa stiffness overlying a subgrade of 40MPa stiffness, the equivalent surface modulus is calculated to be 46MPa. This value would therefore represent the direct LWD test equivalent of a Class 2 foundation.

4.3 Discussion

The design approach described in Highways Agency (2009) represents a considerable step forward in that it puts pavement analysis onto a sound basis. Nevertheless, implementation in the UK has given rise to questions as to the appropriate in-situ measured surface modulus requirement for a given foundation class. For example, the limits currently included in Highways Agency (2009) for a Class 2 foundation are an absolute minimum LWD modulus of 50MPa and a mean value (from groups of 5 tests) of 80MPa which, from the foregoing example, would appear to be an unnecessarily conservative requirement. It is understood that a similar point has been made on several occasions by contractors who have found the 80MPa mean value difficult to achieve with standard unbound crushed rock aggregates. Clearly this is an issue that would benefit from further comparative study, including tests carried out on completed pavements using a Falling Weight Deflectometer.

Nevertheless, judgments have to be made in order to move forward and testing in a PUMA or similar confined compression test (Springbox, K-Mould) can provide suitable information on which to base such judgments. The evidence put forward in this paper suggests that the LWD-measured stiffness modulus on a granular pavement foundation should typically be less than half the value that would apply in the finished pavement, although it is clearly important to ensure that limits placed in specifications are suitably robust.

Once appropriate limits are agreed, including suitable factors of safety, the PUMA or similar tests can then be very useful in allowing a designer to evaluate alternative foundation material combinations in order to achieve a desired foundation class. The challenge for any highway authority is to allow road construction and design organizations to innovate and, thereby, save resources and costs while maintaining an appropriate degree of conservatism.

5 CONCLUSIONS

This paper represents a snapshot of developments in the UK that are still ongoing. It is believed that the generic test type, one version of which has been described, could be widely used in pavement foundation design. This is, however, reliant on there being a specification that commands the trust of the industry and that ensures the designed long-term in-service foundation stiffness modulus is achieved subject to matters that are outside the remit of direct testing, notably the provision of appropriate drainage.

6 ACKNOWLEDGEMENTS

The authors would like to acknowledge the support of Cooper Research Technology in obtaining data from the PUMA equipment.

7 REFERENCES

AASHTO. 2008. Mechanistic-Empirical pavement design guide; a manual of practice.

Brown, S.F. 1995. Practical test procedures for mechanical properties of bituminous materials, Proc. Inst. of Civil Engineers Transport, 111 (4), 1995, 289-297.

CEN. 2004. Cyclic load triaxial test for unbound mixtures, EN13286-7 Unbound and hydraulically bound mixtures: Part 7, Comité European de Normalisation.

Cooper, K.E. and Brown, S.F. 1989. Development of a simple apparatus for the measurement of the mechanical properties of asphalt mixes, Proceedings of the Eurobitume Symposium, Madrid, 494-498.

Brown, S. F. and Dawson, A. R. 1992. Two-stage approach to asphalt pavement design, Proceedings of the 7th International Conference on Asphalt Pavements, 1, Nottingham, 16-34.

Edwards, J. P., Thom, N. H., Fleming, P. R. and Williams, J. 2005. Accelerated laboratory based mechanistic testing of unbound materials within the newly developed NAT Springbox, Transportaion Research Record 1913, Transportation Research Board, Washington, DC, 32-40.

Highways Agency. 2006. Design manual for roads and bridges, 7. The Stationery Office, London.

Highways Agency. 2009. Design guidance for road pavement foundations (draft HD25). Interim Advice Note 73/06 Revision 1,The Stationery Office, London.

LCPC and SETRA. 1994. Conception et dimensionnement des structures de chaussée, Le Laboratoire Central des Ponts et Chaussées and Le Service d’Études Techniques des Routes et Autoroutes, Paris.

Semmelink, C. J. and de Beer, M. 1995. Rapid determination of elastic and shear properties of road-building materials with the K-Mould,Unbound aggregates in roads, ed A. R. Dawson and R. H. Jones, University of Nottingham, 151-161.

Thom, N.H. 1988. Design of road foundations, PhD Thesis, University of Nottingham.

Thom, N.H., Cooper, A., Grafton, P., Walker, C., Wen, H. and Sha, R. 2012. A New Test for Base Material Characterisation, Proceedings of the International Symposium on Heavy Duty Asphalt Pavements and Bridge Deck Pavements, International Society for Asphalt Pavements, Nanjing, China.

1241

Deformation Performance and Stability Control of Multi-stage Embankments in Ireland

Performance en déformation et contrôle de stabilité de remblais construits par étapes en Irlande

Buggy F.J. Roughan & O’Donovan, Dublin, Ireland

ABSTRACT: The Limerick Tunnel project includes approximately 10 km of approach roads, most of which are constructed on embankments up to 10m high within the River Shannon flood plain. The ground conditions consist of very soft organic silts up to 13mdeep. A combination of vertical drains and basal reinforcement plus 2 to 2.5m of surcharge was generally adopted as ground improvement. The embankments were carefully built at controlled rates in multiple stages with continuous monitoring of performanceby means of piezometers, inclinometers, settlement plates and survey monuments. The paper describes the observational approachused to control embankment stability primarily by means of monitoring filling rates, pore pressures and deformation ratio of lateral toe displacement to vertical crest settlement. Performance data relating to a failure of a 3m high embankment during the first stage ofloading is included. Comparison of critical filling rates prior to minimum stability as indicated by maximum deformation ratio ispresented for the Limerick Tunnel project as well as other case histories of multi-stage embankments in Ireland and some critical conclusions are drawn.

RÉSUMÉ : Le projet du Tunnel de Limerick comprend environ 10 km de voies d’accès, la plupart desquelles sont construites sur desremblais jusqu'à 10 m de hauteur dans la plaine d’inondation de la Rivière Shannon. Les conditions de sols consistent en une vaseorganique très molle qui va jusqu'à 13 m de profondeur. Une solution combinant des drains verticaux avec un renforcement de la base du remblai, plus une surcharge de 2 à 2.5 m de hauteur a généralement été utilisée pour améliorer le sol. Les remblais étaientsoigneusement construits en plusieurs étapes avec une surveillance continuelle du comportement par des piézomètres, des inclinomètres, des plaques de règlements et des repères topographiques. Le document décrit la méthode observationnelle utilisée pour contrôler la stabilité du remblai, principalement par la surveillance des vitesses de remblaiement, des pressions interstitielles, et duratio de déformation latérale en pied de remblai rapporté au tassement en crête. Les données du comportement d’une rupture d’unremblai de 3m de hauteur lors d’une première étape de chargement sont incluses. La comparaison des vitesses de remblaiementcritiques juste avant d’atteindre une stabilité minimum telle qu’indiquée par le ratio de déformation maximale est présentée pour le Tunnel de Limerick, ainsi que pour d’autres études de cas de remblais construits par étapes en Irlande, et quelques conclusionsimportantes sont données.

KEYWORDS: Ground Improvement; Surcharge; PVD; Multi-stage Embankments; Performance Monitoring; Observational Design.

1 INTRODUCTION.

1.1 Project Description

The Limerick Tunnel project is located to the south and west of Limerick City in SW Ireland and provides a dual carriageway bypass of the city via an immersed tube tunnel beneath the River Shannon. The route passes through flat, low lying alluvial flood plains of the River Shannon and its tributaries. Much of the 10 km long roadway is on embankment to maintain the road above potential flood levels and for crossings of existing creeks, roads and railways. The flood plain of the River Shannon is underlain by extensive deposits of very soft to soft alluvium comprising mainly organic silt to depths typically from 3 to 13m deep.

The embankments were carefully built at controlled rates in multiple stages with continuous monitoring of performance by means of piezometers, inclinometers, settlement plates and survey monuments. This paper describes the observational approach used to control embankment stability primarily by means of monitoring filling rates, pore pressures and deformation ratio of lateral toe displacement to vertical crest settlement. Principle methods adopted for earthworks along the project include one of more of the following ground improvement solutions: Full or partial excavation and replacement; Prefabricated Vertical Drainage (PVD); Geosynthetic Basal Reinforcement; Multi-Stage Construction Techniques; and Surcharging.

A more detailed description of the design, construction and performance of embankments along the project is contained in Buggy & Curran (2011).

1.2 Site Characterization and Alluvium Properties

A brief summary of the engineering properties of the soft alluvium follows but a much more extensive description is given in Buggy & Peters (2007). The uppermost 1m, approximately, of alluvial material is a firm to stiff desiccated “crust” overlying very soft to soft, grey, silt with organic material or lightly overconsolidated grey silt with abundant organic material. The stratum occasionally contains bands of more sandy material or shell fragments but is generally free of distinct laminations and partings. Classification test data for the alluvium are summarized as follows: Natural moisture content - 40 to 120 % in organic silt and

150 to 300 % in peaty layers; Liquid Limit - 40 to 150 % in organic silt and 150 to 300 %

in peaty layers; Plasticity Index – 30 to 75 %; Organic content (loss on ignition) typically 2 to 10 % but up

to 34% in peaty layers; Undrained strength ratio cu / po’ varies 0.36 CAUC triaxial

tests; 0.3 DSS test; 0.2 CAUE triaxial tests; 0.3 average assumed in design; and

Coefficient of consolidation Cv = 0.5 to 4 m2/yr (lab tests) and 0.8 to 1.5 m2/yr (field back calculated).

The alluvium sediments are underlain by deposits of predominantly fine grained glacial tills with occasional coarse grained layers and limestone bedrock.

1242


2 MULTI-STAGE EMBANKMENT STABILITY

At Limerick Tunnel the designers adopted the undrained strength analysis approach as developed by Ladd (1991). This employs a normalised undrained strength ratio cu / p0’ to predict the operational shear strengths that would apply at some future time after initial loading based on the estimated (or measured) partial consolidation and pore pressure conditions of the layer of soil in question. Stability at any stage of construction was evaluated by limit equilibrium methods using Bishops Modified Method for circular and Janbu’s Method for block shaped failure planes, the most critical of either being adopted in design. A minimum operating Factor of Safety of 1.25 was adopted for short term loading conditions assuming that the embankment was fully instrumented. In the long term fully drained condition a Factor of Safety of 1.3 was selected. A cautious design value of Cvr = 1 m2/year was adopted for radial drainage and the contribution of vertical drainage was ignored. Vertical drains consisting of Mebradrain MD7007 were typically installed at 1.3 m c/c triangular spacing but in one 200m long high fill area drains were installed at 1.0 m c/c spacings. If the total filling duration to achieve maximum height was deemed excessive, typically in excess of 6 to 9 months depending on the Contractor’s programme, then the use of basal geosynthetic reinforcement was considered to increase the temporary stability and thereby reduce the total time required for initial filling to full height. Basal reinforcement was required for approximately 1.7 km of the 6 km total embankment length requiring PVD and surcharge, typically where the total temporary embankment height (including surcharge fill) exceeded 6m.

Multi-stage embankment construction designs were summarized in tabular format for each design profile and the earthwork drawings also reflected the reinforcement and stage hold durations. Earthworks construction was controlled in the field by careful review of instrumentation data by the designer’s site staff and the filling schedules and hold periods were altered to reflect the true soil behavior as monitored by field instrumentation.

Jardine (2002) gives an excellent summary of the behaviour of multi-stage embankments constructed on soft foundation soils. Based on a number of fully instrumented and well documented case histories he notes the following key principles of their behaviour which can be of use in monitoring performance and assessing stability: Large ground movements due to volume changes can occur

as instability is approached; Instability is primarily related to lateral spreading of the

foundation and this can be monitored by assessing deformation ratios of lateral movement at the toe Y to maximum settlement at the crest S (see Figure 1). The limit criteria for such ratios will be different for single stage compared to multi-stage embankments and indeed will vary with each site due to soil material properties, embankment geometry, soil profile and loading rate;

Similarly as instability is approached the ratio of pore pressure change in the foundation soils to increased total loading approaches and exceeds unity; and

An observational approach is only valid if adequate instrumentation and a degree of redundancy due to loss is provided. The time necessary to acquire, process, evaluate and provide a control response must also be sufficiently short to avert a failure.

Deformation ratio limits reported in CIRIA C185 (1999) typically range from 0.3 to 0.4 for embankments on soft ground. Data from Japan published by Wakita & Matsuo (1994) has suggested that the deformation ratio for a given degree of stability reduces as total settlement increases, failures being expected for deformation ratios in excess of 0.4 for observed settlements greater than 2m.

Figure 1 Definition of Embankment Deformation Ratio (Y/S) (Jardine, 2002)

Constitutive models used for soft alluvium included standard isotropic soft clay model in the PLAXIS suite plus anisotropic models S-CLAY1 and ACM which was performed by University of Strathclyde. Further details of the anisotropic model parameters and results are given by Kamrat-Pietraszewska et al. (2008). The FEM results suggested that the maximum deformation ratio to be expected for the proposed stage loading schedules at adequate Factors of Safety might range up to 0.6. The following threshold limits for monitoring data were adopted as indicative of developing failure based on an average filling rate of 0.5 m/week with an absolute prohibition on any single incremental fill rate exceeding 1m/week: Incremental pore pressure ratios u/v > 1.0 ; Global Deformation Ratios (Y/S) > 0.5; Deformation Ratios > 0.3 represented warning conditions

where fill rates and performance data had to be more closely monitored; and

Incremental change in settlement or toe movement > 0.1m between consecutive readings.

3 INSTRUMENTATION MONITORING

A total of 13 fully instrumented cross sections were selected at representative locations and near structures where temporary fill heights were greatest. A standard instrumentation cross section included a pair of settlement plates 5m inset from the embankment crest, survey monuments 1m offset from each toe, VW piezometers arranged at the centre point of the triangular PVD layout under the embankment centreline typically at 3m depth increments plus a single piezometer under both mid slopes at 2 metres depth. Inclinometers extended to stiff glacial till soils or bedrock were installed at the embankment toes.

Settlement plates and toe survey monuments were generally arranged in pairs at 50 m c/c spacing along the mainline. Active areas of filling with settlement rates > 20 mm / week required twice weekly monitoring but daily monitoring was triggered when monitoring threshold limit values were exceeded.

4 EMBANKMENT PERFORMANCE

4.1 Deformation Ratio & Stability

A typical filling rate and deformation ratio history for the instrumented location at Ch 4+185 m is shown on Figure 2. During initial filling to heights of 4 m the deformation ratio rapidly rose to local maximum values of 0.4. The ratio then reduced to below 0.2 as settlement continues under constant load before increasing again during the next filling stage but

1243


remaining near or below 0.3. This behaviour was commonly observed elsewhere along the project..

Figure 2 Ch 4+185m Deformation Ratio & Fill Height v Time

The distribution of measured peak deformation ratio throughout areas of the project improved by PVD and surcharge is shown in Figures 3 (a), (b) & (c) for the mainline embankment south and north of the River Shannon plus Clonmacken Link respectively.

In general peak deformation ratios south of the Shannon were kept below 0.6. Exceptions occurred at Ch 1+900 & 2+850 m where the embankment was constructed adjacent to and within the original course of Ballinacurra Creek and temporary sheet piling had to be employed to maintain stability.

Within the mainline north of the Shannon peak deformation ratios were typically below 0.5. Exceptional values in excess of 1.0 occurred near Ch 7+550 m & 7+650 where local failure of an existing back drain behind the Shannon flood protection bund and instability in an area where the surface crust had been historically removed for brick manufacture. In both cases local instability occurred at relatively low embankment heights and was mitigated by culverting the drain and incorporation of geogrid reinforcement over short sections of the outer embankment slope. Other high ratios occurred near the locations of cross ditches.

Along Clonmacken Link several values in excess of 1 occurred associated with failures which are described in more detail below. Interestingly the occurrence of high deformation ratios is not particularly correlated to maximum embankment height nor to the inclusion of basal geosynthetic reinforcement within the embankment. Around 90% of all survey monitored sections demonstrated a deformation ratio of under 0.6.

4.2 Embankment Failures at Clonmacken Link

During the period from June to September 2007 two significant failures occurred over 50m lengths of embankment (from Ch 140 to 190 and from Ch 780 to 830m ) as filling extended their height to around 3 - 3.5m. Both failures occurred on the left (northern) side of the embankment and displaced a relatively shallow block of alluvium up to 2.5m deep at the toe outwards and closing a ditch constructed from 5 to 15m distant from the toe. In both cases the failure plane was restricted to the outermost section of embankment and did not pass through the basal reinforcement or the drainage blanket which appeared to function normally based on the nearest piezometer records.

Forensic investigations of both failures revealed the following factors which had contributed to the instability: Accidental over steepening of side slopes to around 1:1.3 in

lieu of the design slope of 1:2 (V:H); Presence of nearby ditches, especially drains cut skew or

transverse to the embankment;

Figures 3 (a), (b) & (c). Peak Deformation Ratio v Chainage, Mainline South, North & Clonmacken Link respectively.

Poor quality fill (Moisture Condition Value 5 to 8 with significant organics) compounded by wet weather conditions (Ch 140 – 190 m only); and

High filling rates around 1.5 to 3 m/week. Typical deformation ratio history for instrumented location

at Ch 0+150 m is shown on Figure 4. The large incremental settlements > 0.1m observed in early to mid September 2007 plus deformation ratio approaching 1.0 immediately prior to failure in mid September both validate the selection of design threshold values and were both reliable indicators of future instability. Regrettably the data was not passed quickly enough to engineers who could have acted to avert the failure by preventing further filling after 17 September 2007. The final filling rate of around 3m / week (1.3 metres in 3 days) could never be sustained at this site and resulted in failure.

1244


Figure 4 Ch 150m Deformation Ratio & Fill Height v Time

5 PEAK DEFORMATION RATIO & FILLING RATES

The variation of peak deformation ratio associated with the maximum filling rate has been investigated for the Limerick Tunnel project as well as other published case histories of multi-stage embankments in Ireland. The maximum filling rate is defined herein as the local filling rate typically measured over a period of a few weeks immediately prior to the peak deformation ratio being observed. This is not the same as the slower average filling rate for a discrete stage in construction.

The data is presented in Figure 5 and includes 18 data points from Limerick Tunnel representative of 6 km length of embankment on PVD improved ground and a range of depths of soft ground from 3 to 11m. Also included are 9 data points from 4 sites in Cork, Bunratty Co. Clare, Derry and Athlone with a diverse range of embankment geometry from 2:1 to 3:1 side slopes (Derry & Athlone with toe stability berms); depth of soft ground from 6 to 13m; range of Cv typically from 0.5 to 3 m2/yr and PVD spacing 0.6 to 1.4 m (Dunkettle Cork did not contain PVD and exhibited a range of Cv from 7 to 16 m2/yr).

Figure 5 Peak Deformation Ratio v Max Filling Rate

While there is considerable scatter in the data, there is a discernable trend of increasing deformation ratio and thus decreasing stability as the local filling rate increases. Below maximum filling rates of 0.5 and 1m / week, deformation ratios generally are below 0.25 and 0.5 respectively, suggesting stable conditions. An exception to this occurs at A2 Maydown site in Derry where significant deformation ratio of 0.65 occurred during construction at a modest maximum filling rate of only 0.4 m / week. This was found to be caused by a pre existing failure plane and remedied by addition of toe stability berms. At Limerick Tunnel filling rates in excess of 2 m / week were consistently associated with large observed deformations or failures. Local failures were strongly influenced by the presence of creeks, ditches or excavations and occurred at filling rates above 1m / week. Peak ratios were nearly always observed during first stage filling at embankment heights of 4m or less. Only 4 peak ratios (15% of the total for all sites) occured at

embankment heights over 5m during second stage or near final embankment height.

6 CONCLUSION

Deformation ratios offer a reliable method for controlling stability of multi-stage embankments when used in conjunction with pore pressure instrumentation. An assessment of critical deformation ratio values during loading should be based on modelling of the specific embankment geometry and soft ground properties, trial embankments and case history precedence in similar conditions. For the specific conditions at Limerick Tunnel a limit deformation ratio of 0.6 was shown to give satisfactory performance and acceptable stability.

Excessive lateral deformation related to local failure occurred at several locations in the vicinity of creeks, ditches and hisotrical excavations located within 10m of the embankment toe. More general failure of embankments occurred at 2 locations during construction, in both cases related in part to excessive filling rates. Local filling rates in excess of 1 m / week have a significantly higher risk of failure and rates below 0.5m / week are advisable to sustain a well controlled, stable stage filling for typical Irish soils.

7 ACKNOWLEDGEMENTS

The author would like to acknowledge the National Roads Authority and DirectRoute (Limerick) Limited for their kind permission to publish the data contained within this Paper. The views expressed in this paper are solely those of the author.

8 REFERENCES

Buggy, F. J. & Peters, M., 2007. Site Investigation and Characterisation of Soft Alluvium for Limerick Southern Ring Road - Phase II, EI Proc. Conf. Soft Ground, Port Laoise, paper 1.6.

Buggy, F. J. and Curran, E., 2011. Limerick Tunnel Approach Roads – Design Construction & Performance. Geotechnical Society of Ireland Meeting, Engineers Ireland, Dublin 8th December, 2011.

CIRIA Report C185, 1999. Observational Method in Ground Engineering. CIRIA London.

Connolly, C., Davitt, S. & Farrell, E. 1990. The Design and Construction of Bunratty Bypass, Engineers Ireland MeetingLimerick.15th October 1990.

Creed, M. J. 1996. Dunkettle Embankments, Glanmire Co. Cork – Pore Water Pressures Reviewed. Engineers Ireland Conference Road Embankments on Soft ground in Ireland. Dublin.

Dauncey, P.C., O’Riordan N.J. & Higgins J. 1987. Controlled failure and back analysis of a trial embankment at Athlone. Proc. IX Europ. Conf. Soil Mechnanics. Dublin pp 21 – 24.

Jardine, R.J. 2002. Stability and Instability: soft clay embankment foundations and offshore continental slopes.” Keynote Paper. International Symposium on Coastal Geotechnical Engineering In Practice. Volume 2, Balkema, Rotterdam. pp 99-118.

Kamrat-Pietraszewska, D., Buggy, F., Leoni, M. & Karstunnen, M. 2008. Numerical Modelling of an Embankment on PVD-Improved soft Alluvium – Limerick Southern Ring Road Phase II. BCRI 08 Symposium NUI Galway pp 443 – 452.

Ladd, C.C. 1991. Stability Evaluation During Staged Construction. ASCE Journal Geotech. Engng Divn. Vol 117. No. 4. pp 540 -615.

Long, M. & O’Riordan N., 2001. Field behavior of very soft clays at Athlone embankments. Géotechnique Vol. 51 (4) pp 293 – 309.

Raven, K. & Anketell-Jones, J. 2012.A2 Maydown to City of Derry Airport – Construction over Soft Ground. Proc. Conf. Geotechnics on Irish Roads 2000 - 2010, Portlaoise. 11 October 2012.

Wakita, E. & Matsuo, M. 1994. Observational design method for earth structures constructed on soft ground. Géotechnique Vol. 44 (4) pp 747 – 755.

1245

Renforcement de plates-formes ferroviaires par colonnes de soil mixing réaliséessans enlever la voie.

Railways platforms reinforced by soil-mixing columns without track removing

Calon N., Robinet A.,Costa D’Aguiar S.SNCF, Paris, France

Briançon L., Cojean C.Cnam, Paris, France

Mosser J.-F.Soletanche Bachy, Rueil-Malmaison, France

RÉSUMÉ : Afin de réduire le coût global de l’infrastructure ferroviaire, la Société Nationale des Chemin de fer Français (SNCF) et SOLETANCHE BACHY développent un procédé de renforcement des structures d’assise à l’aide de colonnes de sol ciment. D’un point de vue mécanique, les colonnes sont susceptibles de créer des points durs provoquant une dégradation de l’armement (ballast et rail) et donc un effort de maintenance plus important. Des essais de laboratoire ont été réalisés pour vérifier l’influence du positionnement des colonnes et l’efficacité des géosynthétiques pour gommer les points durs. En parallèle, des modélisations numériques pour connaitre l’influence des colonnes de soil-mixing sur la raideur globale de la voie sont réalisées.

ABSTRACT: In order to reduce the Life Cycle Cost of railway infrastructure, the French National Railway company (SNCF) andSOLETANCHE BACHY have investigated the potential benefits from the ground reinforcement by vertical soil-cement columns.From a mechanical point of view, stiff zones could be created at the column location and damaged the ballast, which would involvemaintenance additional works. Laboratory tests have been performed to verify the influence of the column location and the efficiencyof geosynthetics on the reduction of stiff zones effects. In parallel, numerical simulations were carried out to analyze the impact of thecolumns on the behavior of the railway track structure in terms of deformability.

MOTS CLÉS: Plate-forme; ferroviaire; Soil-mixing; modélisation.

KEYWORDS ; platform ; railway ; soil mixing ; modelisation

1 INTRODUCTION.

Le réseau ferroviaire exploité en France compte environ29 000 km de lignes anciennes dites « lignes classiques »circulées jusqu’à 220 km/h et 2 000 km de Lignes à GrandeVitesse (LGV). Les structures d’assise des premières citées se sont constituées au cours du temps en fonction de leur histoire(Trinh 2011, 2012) : chargement et nature de matériaux mis enœuvre. Quelques tronçons de ces lignes classiques sontaujourd’hui affectés par des pathologies liées à la plate-forme.Afin de trouver une alternative aux travaux actuels, longs etcouteux, la Société Nationale des Chemins de fer Français,l’IFSTTAR et SOLETANCHE BACHY développent une méthode de renforcement des structures ferroviaires, parcolonnes de soil-mixing, sans nécessiter une dépose préalablede la voie. L’intérêt de cette technique est qu’elle s’inscrit pleinement dans les préceptes du développement durable, eneffet, elle permet de renforcer la voie en réutilisant lesmatériaux du site, tout en diminuant les interceptions descirculations. C’est dans ces perspectives (technique etenvironnementale) que le projet de recherche Renforcement etréUtilisation des plates-formes et Fondations Existantes(RUFEX) a été initié permettant ainsi le développement etl’optimisation de l’outil Springsol.

D’un point de vue mécanique, la mise en œuvre de colonne de soil-mixing est susceptible de créer des points dursprovoquant une dégradation de l’armement (ballast, rail) et d’engendrer des efforts dynamiques supplémentairesoccasionnant une maintenance accrue.

L’influence de ces colonnes, sur la voie, a donc été testée aulaboratoire sur une structure échelle une. Divers complexes degéotextiles ont, également, été mis en œuvre pour établir leureffet sur le gommage des points durs. En parallèle, desmodélisations numériques ont été réalisées pour définir lameilleure adéquation entre l’espacement inter colonne et leur

profondeur. L’étude de l’interaction véhicule-voie est égalementprévue dans le cadre de cette recherche.

2 ESSAIS EN LABORATOIRE

Plusieurs recherches ont été menées, en laboratoire, pourvérifier l’influence des géosynthétiques dans l’amélioration ducomportement du ballast. Idraratna et al. (2006) ont évalué lescaractéristiques du ballast à l’aide de cellules triaxiales permettant de modéliser une portion de voie. Brown et al.(2007) ont mis au point un appareil permettant de tester desgéogrilles et ainsi déterminer les paramètres essentiels qui ontun rôle dans le renforcement du ballast. Kenedy et al. (2009) ontdéveloppé une machine « Geopavement & Railway AcceleratedFatigue Testing » permettant, en laboratoire, de reproduire desconditions de sollicitations proches de la réalité. A la lecture deces documents, un dispositif d’essai existant (« RailwayAccelerated Fatigue Testing » - RAFT), au laboratoire « voie »de la SNCF, a été d’adapté pour tester l’influence des pointsdurs sur la voie et l’effet des géosynthétiques pour lisser ces défauts.

1.1 Dispositif et protocole d’essai

1.1.1 Dispositif d’essaiLes dimensions (3x1.5 m) de la RAFT (Figure 1a ;b) permettentde positionner à l’intérieur un châssis de voie comprenant deuxtraverses et deux rails. L’épaisseur de ballast mise en œuvre est conforme à celle présente en voie à savoir 30cm. Lescirculations sont simulées à l’aide d’un vérin hydraulique de 200kN, approchant d’un trafic de 22,5 t/essieux, ayant unefréquence de sollicitation de 5Hz, simulant un convoi circulant à100 km/h (Trinh, 2011). Pour simuler la zone de remontéeboueuse, un tapis anti vibratile (2 cm d’épaisseur ; raideur0,03 N/mm3) a été utilisé. Les colonnes de soil-mixing ont été

1246


simulées par des disques de bétons de 40 cm d’épaisseurpositionnés dans des réservations réalisées dans le tapis.

a)

b)

Figure 1. Vue du banc d'essai (RAFT) : a) vue globale du système ; b)coupe type

Des capteurs de pression () et de tassement (S) ont étédisposés à la base du ballast (Figure 2) pour étudier l’effet de chaque renforcement sur les contraintes appliquées au niveau dusol support. Des mesures au niveau du rail ont également étéréalisées afin de vérifier l’effet du renforcement sur la géométrie du rail.

1

2

3

4 s1s4 s3 s2

s5s8 s7 s6

5

a)

3

6

1 7s1

s4 s3 s2

s5s8 s7 s6

54

b)Figure 2. Localisation des capteurs de contrainte ( et de tassement(Si), en fonction du positionnement des points durs

Trois types de configuration ont été testés : absence decolonne ; une colonne dans l’axe de la voie ; et deux colonnessous chaque file de rails. Pour chaque configuration, différentsrenforcements ont été expérimentés : absence degéosynthétique ; une géogrille à la base ou dans le ballast ; ungéotextile à la base du ballast ; un géotextile associé à unegéogrille le tout placé à la base du ballast.

1.1.2 Protocole d’essaiPour chaque essai, la procédure suivante a été appliquée : cinqchargements statiques de 200kN permettant de stabiliser la voie(limiter les tassements différentiels), un chargement cyclique de100 000 périodes, encadré de trois chargements statiques de200 kN. Un essai de répétabilité consistant en la mesure de lacontrainte après les sollicitations cycliques a permis de validercet essai (Tableau 1)

Tableau 1. Essais de répétabilité.Un point dur Deux points durs

Essai 1 Essai 2 Essai 1 Essai 21 (kPa) 53 58 85 922 (kPa) 58 573 (kPa) 64 57 52 654 (kPa) 66 445 (kPa) 50 49 66 -6 (kPa) 66 617 (kPa) 84 96

1.2 Résultats et analyses

Dans un premier temps, ces essais ont mis en évidence l’effet des chargements cycliques sur la modification du mécanisme detransfert de charge. En effet, à la suite de la sollicitation, et pourla configuration sans géosynthétique, la contrainte moyenne surles colonnes augmente de façon non négligeable (1,5x), alorsque celle relevée en présence de géosynthétique est peu ou prouéquivalente (Tableau 2). Notons que dans le cas où la géogrilleest posée sous le ballast, nous observons une augmentation de lacontrainte moyenne après sollicitation. Cette augmentation peutêtre due à une mauvaise mise en œuvre du produit expliquant ainsi qu’il ne fonctionne pas de façon optimum.

Tableau 2. Influence du chargement cyclique sur la contrainte à la basedu ballast

moy Sol (kPa) moy point dur (kPa)essais Avant Après Avant AprèsSans GSY 71 67GTX 66 62Sans

point durGTX + GGR 74 60

Absence depoint dur

Sans GSY 66 60 89 130GTX 64 57 Absence de mesureUn point

durGTX + GGR 62 56 73 87Sans GSY 61 68 68 105GTX 69 62 69 82GGR* 60 60 65 85GGR* 61 62 74 94

Deuxpointsdurs

GGR 62 59 80 104Note: GGR* = GGR dans le ballast

Pour mettre en évidence l’effet des géosynthétiques sur legommage des points durs, il est nécessaire de faire un focus surl’ensemble des capteurs de pression utilisés pour ces essais.Dans le cas d’un renforcement sous chaque file de rails (deuxcolonnes) et en présence de géosynthétique (hors géogrille à labase du ballast), la contrainte sur les têtes de colonne estdiminuée (Tableau 3). Par ailleurs, une symétrie de chargementest observée lorsque pour le cas avec deux colonnes sous la voiece qui d’un point de vue de la maintenance permettra d’éviter d’avoir des défauts de géométrie du rail.

Tableau 3. Pression mesurée après une sollicitation cyclique pour uneaugmentation de charge de 200 kN (i = F = 200 kN - F = 0 kN).

Contrainte (kPa)1 3 4 5 6 7

Sans GSY 106 63 - 62 74 104GTX sous ballast 84 64 64 71 58 80GGR sous ballast 85 52 56 66 67 84GGR sous ballast 101 56 62 56 63 107

Le Tableau 4 met en évidence une augmentation de la contrainteentre les colonnes. En effet, la contrainte obtenue dans l’axe dela voie est plus élevée pour un chargement de 200kN, dans lecas où la géogrille est placée dans le ballast ou dans celui dugéotextile. Ce qui n’est pas le cas pour l’essai sans géosynthétique.

Ballast

Géotextile

Point dur

1247


Tableau 4. Pression mesurée après un chargement cyclique pourune charge statique de 200 kN (i = F = 200 kN).

Contrainte (kPa)1 3 4 5 6 7

Sans GSY 101 63 - 69 85 102GTX sous ballast 76 63 64 125 85 82GGR sous ballast 79 51 56 113 94 80GGR sous ballast 102 63 62 90 88 96

A l’inverse, le Tableau 3 montre que l’augmentation de contrainte dans les cas précités est la même quelque soit lecapteur. Ainsi, en l’absence de chargement, le sol est confiné entre les colonnes par le biais d’une tension développée dans lesgéosynthétiques mettant en évidence un effet membrane.

1.3 Conclusion

Les tests réalisés ont montré l’efficacité des géotextiles pour gommer les points durs engendrés par les colonnes de soil-mixing. Cette efficacité dépend bien entendu du type derenforcement (et association) mis en œuvre et sa localisationdans la structure.

2 MODÉLISATION NUMÉRIQUE

Dans ces travaux, la modélisation numérique a été réalisée àl’aide du logiciel GEFDyn (Aubry, Chouvet, Modaressi and Modaressi, 1986). Ce modèle, aux éléments finis, prend encompte les caractéristiques du sol dans une large gamme dedéformation, permet de représenter les déformationsirréversibles et permet de gérer jusqu’à quatre lois de comportement élastoplastiques avec quatre mécanismesdifférents (3 déviatoriques et un volumétrique).

Le logiciel GEFDyn permet de modéliser la voie enreprésentation 2,5D avec un chargement sous différents cheminsde contraintes (Saez, 2009). La superstructure est composée derails, semelle sous rail, traverses ; la structure est constituée duballast, de la couche intermédiaire et du sol support. Le rail estmodélisé par un élément poutre, tandis que les autres élémentssont constitués par des éléments quadrilatéraux. Dans unpremier temps, le comportement des matériaux est modélisé parune loi élastique linéaire. Les paramètres utilisés pour lesdifférents éléments de structure sont repris dans le Tableau 5.

Tableau 5. Paramètres de modélisation (e, épaisseur de la couche)E

MPa

ρkN/m3

emm

Rail 210.103 0.3 7850 -Patin 40 0.25 900 90Traverse 30.10 0.25 2400 210Ballast 130 0.2 1700 250Couche Intermédiaire 180 0.3 2135 400Sol 12.5 0.4 1800 8140Colonne * 0.2 1800 2160

Le tronçon de voie modélisé est long de 26 m, pour uneprofondeur de 9 m (Figure 3). La charge roulante est modéliséepar un chargement quasi statique (F=100 kN à 15 km/h)appliqué sur le rail. Ce convoi simule le passage d’un engin Suisse EMW permettant de mesurer la raideur de la voie.

Ainsi dans le but d’étudier numériquement l’influence des colonnes sur la voie, trois paramètres ont été étudiés :L’espacement entre colonnes, leur profondeur et les propriétésmécaniques de la colonne de soil-mixing.

Figure 3. Maillage du tronçon de voie par élément fini, une colonnetoutes les deux traverses. (Colonne : Ø400, profondeur 2.16m)

Pour représenter l’impact des colonnes sur la voie, il a été décidé de tracer la déflexion du rail sous charge roulante enfonction de sa position (x).

2.1 Résultat et analyse

La Figure 4 présente la déflexion du rail sous circulation pourdifférents espacements entre colonnes. Dans le cas présenté, larigidité de la colonne est cent fois supérieure à celle du sol(Ecol/Esol=100).

Figure 4. Déflexion du rail sous circulation, pour différents espacementsentre colonnes (x, position de la charge)

La comparaison entre le cas avec une colonne toutes les deuxou trois traverses et le cas sans colonne montre que ces types derenforcement n’apporte pas de point dur sous la voie. Le signalest quasiment identique par contre, la déflexion du rail se trouvelargement diminuée pour un tronçon renforcé d’une colonnetoutes les deux ou trois traverses. Les variations présentes sur cegraphique, par exemple sur la courbe 0col-A-ACI sont dues à laprésence des traverses. Pour les cas avec une colonne toutes lesquatre ou six traverses, des pics apparaissent. Ceux-ci sont dus àla présence des colonnes de soil-mixing sous la voie. Donclorsque les colonnes sont trop éloignées, l’effet de groupe disparaît et des points durs apparaissent occasionnant desvariations de déflexion importantes.

Dans le cadre de cette étude, les simulations numériquesréalisées ont également fait apparaître l’influence de la prise du ciment sur le comportement des colonnes et donc sur la réponsede la voie. De précédents articles mettent en évidence la relationentre le module de Young et la résistance à la compressionsimple (Tan, Goh and Yong 2002 ; Ajorloo 2010). La résistancedu mélange sol/ciment dépendant principalement de laproportion de ciment et de ses caractéristiques chimiques, nousavons réalisé différents tests afin de déterminer le ratio optimumentre le module de sol et celui de la colonne et de déterminerl’impact sur la réponse de la voie. Ainsi, les cinq configurationsde renforcement ont été testées avec des ratios de modulecompris entre 1 et 1 000.

1248


Les Figure 5 (a) et (b) présentent les déflexions et amplitudesdu rail pour un renforcement, en fonction du Ecol/Esol.

Figure 5. Impact du module de Young de la colonne de soil-mixing surla réponse de la voie, en fonction de l’espacement entre colonnes

Ce graphique met en évidence que les déflexions du rail seréduisent à mesure que le module de la colonne augmente. Cetteréduction est très importante jusqu’à ce que la colonne soit 100 fois plus raide que le sol, au-delà l’augmentation du module de la colonne n’est pas significative sur le comportement de lavoie. Cette information est très importante, car elle nous permetainsi d’optimiser la teneur en ciment du mélange.

Si nous regardons l’amplitude de variation du rail (Figure5b) nous constatons également l’influence de ce ratio Ecol/Esolsur l’apparition de points durs. Ce graphique met en évidence un point d’inflexion, identifiant l’amplitude maximum du rail, dont le pic varie entre 25 et 100 et qui est fonction del’espacement entre colonnes. Au-delà de ces valeursmaximums, l’amplification de raideur tend à se stabiliser voiredécroitre légèrement.

L’effet de la longueur de la colonne sur la géométrie de la voie est mis en évidence par la Figure 6 (a) et (b) (1 colonne/4traverses). Les résultats montrent que la déflexion du raildiminue à mesure que la longueur de la colonne augmente, maisparallèlement les points durs sont plus marqués.

Figure 6. Impact de la longueur de la colonne sur la géométrie du rail

2.2 Conclusion

Ces modélisations ont mis en évidence l’influence des colonnes de soil-mixing sur la réponse globale de la voie et parconséquent sur la création de points durs. En parallèle, lescalculs menés sur le ratio module sol/colonne, ainsi que ceuxréalisés sur l’espacement inter colonne et la profondeurd’ancrage de celles-ci doivent permettre d’optimiser le maillage du renforcement et ainsi que la qualité du coulis.

3 CONCLUSION

L’intérêt de développer une technique de renforcement de plate-forme par des colonnes de soil-mixing, et ce, sans enlever lavoie, était devenu une nécessité pour la SNCF. Si peu de douteconcernait le procédé de mise en œuvre de ces colonnes, il étaitnécessaire de vérifier leur impact sur la tenue de la géométrie durail, et donc sur l’absence de création de points durs. Les essaismenés en laboratoire et par simulations numériques ont permisd’une part de valider l’apport des géosynthétiques pour homogénéiser les contraintes appliquées au sol et donc de

gommer ces points et d’autre part de vérifier l’influence des colonnes sur le renforcement d’une portion de voie.Par la suite, les travaux de modélisation 3D permettront dedéfinir le maillage optimum des colonnes (profondeur,espacement, positionnement) et également intégrer lesgéosynthétiques dans le modèle pour avoir une modélisationplus représentative. Pour ce qui concerne le développement durenforcement par géosynthétique il est nécessaire de définir unproduit standard qui répondra aux différentes exigences defonctionnement (séparation / filtration / renforcement) maiségalement de mise en œuvre (légèreté, maniabilité).

4 REMERCIEMENTS

Les auteurs tiennent à remercier la Direction Générale de laCompétitivité et des Services (DGCIS) et le Conseil Général du93 qui cofinancent cette recherche. Nous tenons également àremercier tous les acteurs du projet RUFEX qui contribuent àson bon déroulement : Soletanche Bachy, Terrasol, IFSTTAR,INSA de Lyon et l’Ecole des Ponts ParisTech et la SNCF.

5 REFERENCES

Ajorloo A. M. (2010). Characterization of the mechanical behavior ofimproved loose sand for application in soil cement deep mixing.Thèse de docteur d’université, université Lille 1 science ettechnologie, University of Illinois,USA.

Aubry D., Chouvet D., Modaressi A., and Modaressi H. (1986).GEFDYN: Logiciel d’Analyse de Comportement M´ecanique desSols par Eléments Finis avec Prise en Compte du Couplage Sol-Eau-Air. Manuel scientifique, Ecole Centrale Paris, LMSS-Mat

Brown, S.F., Kwan, J. & Thom, N.H. 2007. Identifying the keyparameters that influence geogrid reinforcement of railway ballast,Geotextiles and Geomembranes, 25 (6): 326–335.

Indraratna, B., Khabbaz, H., Salim, W. and Christie, D. 2006.Geotechnical Properties of Ballast and the Role of Geosynthetics,Journal of Ground Improvement, 10(3): 91-102.

Kennedy, J.H., Woodward, P.K. Medero G. & McKinney J. 2009. Full-scale cyclic geopavement & railway accelerated fatigue testing,Proc in the 10th International Conference on Railway Engineering,June 24-25, 2009, London, UK.

Saez E. (2009). Dynamic nonlinear Soil-Structure interaction. Mémoirede Thèse, Ecole Centrale de Paris.

Tan T. S., Goh T. L., and Yong K. Y. (2002). Properties of Singaporemarine clays improved by cement mixing. Geotechnical TestingJournal 25(4).

Trinh V.N. 2011. Comportement hydromécanique des matériauxconstitutifs de plateformes ferroviaires anciennes. Mémoire deThèse, Ecole Nationales des Ponts et Chaussées - Université Paris –Est

Trinh V.N., Tang A.M., Cui Y.J., Dupla J.C., Canou J., Calon N.,Lambert L., Robinet A., Schoen O. (2012) Mechanicalcharacterisation of the fouled ballast in ancient railwaytracksubstructure by large-scale triaxial tests. Soils and Foundations52(3), 511–523

1249

Analysis of the influence of soft soil depth on the subgrade capacity for flexible pavements.

Analyse de l'influence de la profondeur d’un sol mou sur la capacité portante pour les chaussées souples.

Carvajal E. Kellerterra S.L., Madrid, Spain

Romana M. Universidad Politécnica de Madrid, Spain

ABSTRACT: It is presented the analysis of a flexible pavement structure founded on soft soil subgrade, through the finite elementmodelling of a multilayered system, with the objective to evaluate the influence of soft soil depth on pavement response. The analysis also comprises an iterative procedure to take into account the influence of small strains on soil stiffness. A simple static load of aheavy truck has been used to evaluate the pavement response; furthermore a cyclic loading has been considered in the form of haversine function in order to simulate the traffick effects on cumulative permanent deformation. The results of permanent verticaldeformation from the followed procedure is compared to an empirical equation, so that rutting failure intensity is estimated. It is concluded that deep ground treatments should be applied to achieve an allowable capacity of soft soils up to miminum depth of about6 m, otherwise maintenance cost of pavements might be excessive.

RÉSUMÉ: L’analyse d'une structure de pavement flexible sur un terrain du sol mou est réalisée grâce à une modélisation éléments finis appliquée à un système multicouches. L’objectif est d'évalué l'influence de la profondeur du sol mou sur la réponse de la structure. L'analyse comprend un procédé itératif pour tenir compte de l'influence de petites déformations sur la rigidité du sol. Une charge statique correspondante à un camion poids lourd a été utilisée pour évaluer la réponse du pavement. Une fonction de charge cyclique de type Haversine a été considérée pour simuler les effets de trafic dans la déformation permanente accumulée. Les résultats de la déformation permanente verticale obtenue par cette méthode sont comparés à une équation empirique, pour évaluer l’orniérage. Le constat est la nécessité de traiter sur une profondeur minimum de 6 m le terrain pour atteindre une capacité de charge admissible dans les sols mous. Dans le cas contraire les prix de maintenance et d’entretien des pavements pourraient être excessifs.

KEYWORDS: Pavement, subgrade, cyclic load, permanent deformation, small strain, stiffness, damping ratio, finite element model

MOTS-CLES : pavement, fondation, chargement cyclique, déformation, raideur, amortissement, éléments finis 1 INTRODUCTION.

Flexible pavements over shallow soft soils could be built through the application of a wide range of techniques to improve the low capacity of subgrade, such as lime and cement stabilisation, geogrids and geosynthetic treatments. Nevertheless when the soil layers which compose the subgrade reach certain depths, the intensity of shallow treatments become inefficient and it is necessary to evaluate the depth that is influenced by the load, and the effects on the capacity and long term behaviour of such deep soft subgrade. Here below is presented a theoretical procedure to analyze the response of a flexible pavement on soft soil under static and cyclic loading. It is also presented an estimation of the rut depth failure based on permanent deformation of subgrade layers affected by determined number of load repetitions.

2 PAVEMENTS ANALYSIS

The pavement analysis through mechanistic approaches is increasingly adopted, with the development of numerical modelling tools, considering complex behavior of pavement structure. Actually, because of the amount of variables that have to be dealt, the design of flexible pavements could be divided in two parts, one mechanical and the other empirical, thus nowadays most frequently used methods for the design are often called Mechanstic-Empirical Methods.

The mechanistic part consists on determining the elastic response of the pavement structure in terms of stress, strain and displacement when a heavy truck equivalent load (P) is applied on the surface.The parameters that determine the properties and

thickness of surface layer are radial strain and tensile strain (t ; t) at the bottom of the surface layer. Whereas, the capacity of foundation soil is governed by the vertical compressive stress and deformation in the top of the subgrade (v ; v). Figure 1.

On the other side, the empirical approach is related to relationships between each components of the elastic structural response (, , y) and a fatigue law. Thus, the damage accumulation for a given number of the load application could be estimated. For this purpose, the fatigue law has to take into account other essential factors that are difficult to assess from the mechanical point of view, e.g. rainfalls, temperature changes, drainage conditions, etc. Particurarly, the evaluation of pavement foundation (subgrade layers) is usually estimated through an allowable number of load repetitions (Nd) that produces unacceptable permanent deformation, which is commonly known as rutting failure mechanism. Otherwise, after construction stage settlements produced due to weight of fills and pavement structure has to be considered, although, even more important matter will be estimation of the time of consolidation process and determination of the soils stress state after consolidation and at the begining of traffic operation.

Figure 1. Parameters for pavement design

1250


Load = 80 kN

Rutting = 0.075 m

Rutting = 0.15 m

(a) (b) (c)

h (m)

Cu (kPa)

Figure 2. (a) Base layer thickness as a function of subgrade undrained shear strength, number of load application (80 kN) and rut depth (Giroud and Noiray 1981); (b) cyclic load used for finite element modelling; (c) geometry of finite element model.

Figure 2a shows a chart proposed by Giroud and Noiray (1981) to select the required thickness of un-reinforced granular layer, which provides an allowable rut depth on unpaved roads. This chart shows that even for low trafficked road it is required a ganular thickness larger than 1 m when subgrade soil presents undrained shear strenght less than Cu = 20 kPa, and number of load application N is greater than 10000. With more greater N and less values of Cu, base layers tend to be too large. In these cases the settlements after construction stage and the consolidation periods may be unacceptable. 3 MODELLING OF SOFT SUBGRADE

3.1 Introduction

A theoretical procedure to estimate the structural response of a pavement founded over a deep soft subgrade is presented, focusing on vertical stresses and strains within the subgrade. For this purpose a mechanistic finite element model has been performed with the program Plaxis v8.2, considering the pavement structure depicted in Figure 2c. The analysis is separately presented according to static and cyclic load conditions, and basically is focused on the influence of subgrade behavior. Accordingly only stresses and strains regarding subgrade layers will be analyzed.

3.2 Geometry and general inputs

The pavement modelling is carried out with axisymmetric conditions for a multi-layered system composed by an asphalt surface layer, followed by two unbound granular layers as base and subbase. A subgrade composed by natural soil is adopted below pavement structure considering the water table at 1 m depth. Vertical boundaries are restrained for lateral displacements, while the bottom horizontal boundary is full restrained for both lateral and vertical displacements. It was adopted an extra-fine mesh of 15-noded elements close to upper part of the model axis, where loading is applied. All of materials constituting the pavement superstructure are modeled as linear elastic, so that the required parameters are only the young modulus E and Poisson’s ratio . In Table 1 are outlined the parameters of pavement superstructure. Table 1. Parameters of asphalt layer and unbound granular layers

Thickness E Layer m kPa

Asphalt 0.1 1.3·106 0.2 Base 0.25 1.5·105 0.33

Subbase 0.5 1·105 0.33

On the other hand, isotropic hardening soil behavior is considered for the subgrade, which is governed by a stress-dependent stiffness that is different for both virgin loading and unloading-reloading process (Schanz 1998). The hardening soil model may be considered as an extension of the well known hyperbolic model (Duncan and Chang 1970), owing to the use

of plasticity theory and yield cap surface to account for the hardening effect produced by isotropic compression strains. The parameters adopted for the soft subgrade soil are presented in Table 2., and its characterization is in accordance with the stiffness increase due to the typical small strains that affects pavement performance.

Table 2. Parameters of subgrade soil Parameter Symbol Values Unit Unit weight 16.5 kN/m3

Small Strain stiffness G0ref 45000 kN/m2

Shear strain at 0.7G0 0.7ref 1.75·10-4 -

Poisson's ratio ur 0.3 - Triaxial compression stiffness E50

ref 10000 kN/m2

Primary oedometer stiffness Eoedref 10000 kN/m2

Unloading - reloading stiffness Eurref 20000 kN/m2

Rate of stress-dependency m 0.85 Cohesion c’ 5 kN/m2

Friction angle ' 10 o

Failure ratio (qf/qasymptote) Rf 0.9 - Stress ratio in primary compression K0

nc 0.83 - The parameter G0 and 0.7 are used in order to consider the variation of moduli at small strains, as describe below. The load condition has been evaluated by means of a static surcharge, which represents the application of a single axle load of 13 t, with a contact area of 0.15 m radius, so that maximum pressure on the surface reaches up to 900 kPa. The cyclic loading conditions have been performed considering a waveform pattern in order to simulate the accumulation of permanent deformation. Thereby, it is assumed a stress pulse over the pavement surface in the form of a haversine function, which may be defined by the use of an equivalent pulse time of 0.1 second, associated to a vehicle speed of 30 km/h (Barksdale 1971). Figure 2b shows the cyclic load pulse adopted.

3.3 Modelling procedure

The calculation consists of one stage that is related to the pavement construction, where unbound granular and asphalt layers are laid out over the subgrade soil. Following stage consist of cyclic load application, in order to simulate the action of traffic. Initially, no drained condition is assumed for two stage considered. After the load application a consolidation phase is included to take account of the final stress state and to evaluate the time of the pore-pressure dissipation.

The effects of cyclic loading on the pavement behavior are analyzed in several steps in order to consider the influence of strain level and the soil damping on the pavement response. Thus, it is adopted an iterative procedure to consider the updating of the soil stiffness according to the strain level during the cyclic loading. This procedure was performed through the stoppage of the loading process once 10 cyclic of load repetition completed, determining the average values of shear strains related to the subgrade layers. In order to update the actual

1251


stress state after load repetitions, overconsolidation ratio OCR of natural soft soil is systematically recalculated due to the modification of the stress history. In total 10 iterations of cyclic loading have been adopted with 10 load repetition each one, so that the analysis reaches 100 load applications.

The decrease of clay stiffness due to increase of the load repetitions is well known issue. To take into account this effect Idriss et al. (1978) proposed that the decrease in modulus could be accounted for by a degradation index according to (1).

= (ES)N / (ES)1 = N-t (1) Where: (ES)N is the secant young’s modulus for Nth cycle;

(ES)1 is the secant young’s modulus for the first cycle; and t is a degradation parameter, which represent the slope of the curve log ES – logN.

On the other hand, in order to take into account the influence of the typical small strains produced below the pavement structures, it was assumed the soil stiffness degradation due to strain level. Hardin and Drnevich (1972) proposed a simple hyperbolic law to describe how the shear modulus of soil decays with the increase of shear strains. Afterward, Santos and Gomes Correia (2001) modified this relationship to determine de variation of secant shear modulus GS as a function of the initial (and maximum) shear modulus at small strains G0, and of the shear strain 0.7 related to the 70% of the maximum shear modulus 0.7G0 as a threshold. Equation 2 describes this relation in the domain of a certain range of shear strain, commonly within values from 1·10-6 to 1·10-2. The tangent shear modulus Gt can be determined taking the derivative of GS.

7.0

0S 0.3851

GG

;

Figure 3. Hyperbolic stress-strain relation and moduli E0, E50 and Eur adopted in the HS-small model.

SD Eπ4Eξ (3) Where ED is the dissipated energy in a load cycle comprised

from the minimum to maximum shear strain (Equation 4), while ES is the energy stored at maximum shear strain c (Equation 5).

0.7

c0.7

c0.7

cc

00.7D γ

aγ1ln

a2γ

aγγ1γ

2γa

G4γE (4)

7.02

202

S222

1E

c

ccS

aGG

(5)

Where a= 0.385

27.0

0t

0.3851GG

(2)

This approach is based on the research carried out by

Vucetic and Dobry (1991) and Ishihara (1996), which demonstrated that beyond a volumetric threshold strain v the soil starts to change irreversibly. At this strain level, in drained conditions permanent volume change will take place, whereas in undrained conditions pore water pressure will build up (Santos and Gomes Correia 2001). Furthermore, it is well known that the degradation of G/G0 with shear strains depends on many factors e.g. plasticity index, stress history, effective confine pressure, frequency and number of load cycle, etc. Indeed, with the purpose of consider the influence of the most importance factors on the degradation of shear moduli, Santos and Gomes Correia (2001) used the average value of v related to the stiffness degradation curves (G/G0 = f()) presented by Vucetic and Dobry (1991), in order to define a unique curve of G/G0 as a function of normalized strain /v. In this way they concluded that when the ratio /v = 1 the best fit tends to correspond to a ratio G/G0 = 0.7.

This approach aids to develop the small-strain stiffness model (HSsmall) proposed by Benz (2006) that is already included in the latest version of Plaxis code. Unlike the standard Hardening Soil Model (Schanz et al. 1999) where a linear stress strain relationship controlled by the stiffness Eur is assumed during unloading-reloading process, the HS-small model takes into account hysteresis loops during loading and unloading cycles with moduli variation among initial E0, secant E50 and unloading-reloading Eur stiffness (Figure 3). In fact, such model also presents a typical hysteretic damping when the soil is under cyclic loading due to energy dissipation caused by the strains; Brinkgreve et al (2007) proposed an analytical formulation to estimate the local hysteretic damping ratio according to Equation 3:

Once assumed the calculation procedure described above, the hysteretic damping ratio according to Equation 3 is estimated from the strain updating after each iterative calculation (10 iterations each consisting of ten load repetitions), considering the calibration of soil stiffness due to strain level (), current stress state (OCR) and number of load repetitions (N). Moreover, viscous damping effects may be added by means of the rayleigh damping features of the used Plaxis version (v8.2). Rayleigh damping consists in a frequency-dependent damping that is directly proportional to the mass and the stiffness matrix through the coefficients and respectively (C = ·M + ·K). In fact, the damping process subjected to a cyclic loading should be analyzed from the combination of two approaches: mechanical hysteretic damping depending on the strains level, and viscous damping depending on the time, which has to fit well with both, natural material and load application frequency. For this purpose the values adopted for Rayleigh coefficients are = = 0.01. As small-strain stiffness model is not included in the version 8.2 of Plaxis code used here, the analytical solution of Brinkgreve et al (2007) is adopted to verify and compare the damping ratios between the results from the finite element modeling and the results from such analytical solution. Finally, the behavior of subgrade soil is analyzed regarding the accumulated vertical strains due to the cyclic load repetitions.

3.4 Modelling results

3.4.1 Subgrade response under static loading The settlement after construction stage rise up to 25 mm, which should last less than few months according to the typical projects requirements. Nevertheless, the consolidation process may lasts between 1.5 to 2.5 years considering a soil permeability of 10-8 to 10-9 m/s. Regarding the operation stage, the maximum value of excess pore pressure due to axle load rises up to 5.5 kPa. The response of subgrade soil when the axle load is applied under static conditions can be appreciated through the values of deviator stresses as well as vertical resilient and permanent strains depending on the depth (Figure 4a).

1252


(a) (b) (c)

Figure 4. Subgrade response under static loading; (a) deviator stress, permanent and resilient strains, and excess pore pressure; (b) stress state at different depths and failure criteria under drained and undrained conditions;(c) horizontal distribution of stresses and strains at 1.5 m depth.

The influence of axle load below pavement surface reaches 6 m depth, although the most of strains take place in the upper layers of subgrade.

It is also noted that within upper 5 m the stresses reaches cap yield surface, leading to an isotropic hardening of soil. Besides, Figure 4b shows the stress states at different depths, where can be seen that the current effective stress state remains far of Mohr Coulomb failure condition, although it is near the undrained failure e.g. according to the empirical equation Su = 0.35·v’·OCRm proposed by Ladd (1991), with OCR = 1 and m = 0.85. The ratio between current deviator stress and deviator stress at failure R=q/qf could be used to express the extent to which permanent deformation might develops; usually it is assumed that permanent deformation will start to rise for R > 0.70 – 0.75 (Korkiala-Tanttu 2008). In any case, isotropic compression produces plastic volume strains once the excess pore pressures are completely dissipated.

The horizontal distribution of shear and vertical strains at 1.5 m depth are shown in Figure 4c; it can be seen that in the vertical axis, there is no shear strains and the vertical strain reaches its maximum value, which indicates a purely triaxial compression state just below the load. On the other hand, the largest shear strain is located at a horizontal distance of 1.10 m from the load, where soil is under a general stress regime with shear and axial stresses. In reality, when cyclic load of moving wheel over the pavement surface is applied, these two stress state are successively changed. Also it is noted that at 1.5 m depth the maximum deviator stress reaches values near 6 kPa, and spreads horizontally up to 2 m from the load axis. Whereas the influence of excess pore pressure spreads horizontally until 4.5 m, approximately.

3.4.2 Subgrade response under cyclic loading

The effect of one cyclic loading stage composed by 10 load

repetitions is shown in Figure 5; the deviator stress, as well as recoverable and permanent displacement at different depths are also depicted in Figure 5. In Figure 6a are shown the curves of shear moduli degradation GS/G0 and Gt/G0 determined by the Equation (2) as a function of strain level and the parameters G0 and 0.7 outlined in the table 1. Also in Figure 6a are shown the results of finite element modelling for the maximum shear strains produced after each iterative calculus under cyclic loading, for a reference depth of 1.5 m. The interception of these maximum shear strains with the curves GS/G0 and Gt/G0 gives the proportion at which soil modulus is changed.

In a next step, stiffness are reduced due to number of load repetitions by means of Equation 1, considering a parameter t = 0.045 (Dobry and Vucetic 1987).

It was observed that maximum shear strain obtained in the first 10 load repetitions ( = 2.9·10-4) is larger than the reference shear strains 0.7(1.75·10-4), which coincides with the proportion of permanent deformation at this loading stage. After 20 load repetitions the shear strain was lower than the 0.7, and after the subsequent load repetitions the strains were reduced more

slowly until reach values leading to ratios GS/G0 = 0.82 and Gt/G0 = 0.67. The OCR was gradually increased in between each iterative calculation up to a maximum value of 1.50. In the Table 3 are shown the results of final subgrade soil stiffness at a reference depth of 1.5 m. The damping ratio obtained from the adopted procedure is verified according to the Equations (3) to (5), taking the maximum shear strain after each iteration, in order to reproduce a representative hysteresis loop. Thus, in total 10 damping ratios have been estimated. In Figure 6b are shown the results of damping ratio compared to the analytical results for hysteretic damping considering variation of initial shear modulus G0 from values of 2Gur to 6Gur, and variation of 0.7 from 1·10-4 to 2·10-4.

It is observed that results obtained from finite element modelling fit well with hysteretic damping related to 0.7 = 1.75·10-4. In the two first iterations was obtained damping ratios close to 0.1, which were reduced gradually until the final iterations reached values close 0.045. This tendency agrees with the typical reduction in the amount of permanent deformation with the load repetitions, due to the reduction in the amount of dissipated energy.

Figure 5. Subgrade response under cyclic loading.

3.4.3 Subgrade performance in the long term

In order to estimate the long term behavior of the soft

subgrade analyzed, one could adopt an empirical power equation for calculating the cumulative plastic strain. Chai and Miura (2000) proposed an enhanced formula of a former empirical model proposed by Li and Selig (1996) for estimation of cumulative plastic strains with the number of repeated load applications. This new model is defined in the Equation (6), which has demonstrated that its prediction agrees with actual measurements taken from low height embankments on soft soil.

bnfis

mfdp Nqq1qqaε (6)

Where p = cumulative plastic strain (%); qis = initial static deviator stress; qd = dynamic load induced deviator stress;

1253


(a) (b) (c)

Figure 6. (a) Shear moduli variation with strain level; (b) damping ratio according to cyclic shear strain; (c) cumulative strain with load repetitions.

Figure 7. Cumulative vertical displacement (rutting depth).

qf = static failure deviator stress; N = number of repeated load applications; and a, m, n and b are constant largely dependant on soil properties, which could be assumed for the soft soil analyzed here, according to those values suggested by Li and Selig (1996): a = 1.2; b = 0.18; m = 2.4; n = 1.

Considering the results of qis, qd and qf depicted in Figures 4b and 5, the cumulative plastic strain is calculated at different depths according to the Equation 6. In Figure 6c is observed that results from finite element model fit reasonably well with the empirical equation proposed by Chai and Miura (2000) considering N = 100. Otherwise, Figure 7 shows a cumulative permanent deflection of about 45 mm (rut depth), given from Equation 6 for a number of repeated load applications up to N = 5·105, and considering the thickness of soft soil influenced by the cyclic load. Such results suppose an unallowable level of rutting failure, even for a low trafficked road as its typical thresholds range from 10 to 15 mm for N greater than 106. Possible solutions for pavements based over such soft soil might be achieved by means of deep ground improvement, which could overcome the detrimental effects of the induced deviator stress and excess pore pressure throughout the depth of load influence (Elias et al. 2004, Sonderman and Wehr 2004).

Table 3. Results of subgrade soil stiffness at 1.5 m depth.

Number of load repetitions Parameter Unit 10 30 50 100 OCR - 1.2 1.4 1.47 1.5 - 2.90E-04 1.16E-04 1.20E-04 1.07E-04

GS/G0 - 0.61 0.80 0.79 0.81 Gt/G0 - 0.37 0.63 0.63 0.66

GS kPa 27473 35851 35601 29485 - 0.891 0.844 0.822 0.794

Eur* kPa 63661 78635 73191 58551 E50* kPa 25464 31454 29276 23421

*final stiffness 4 CONCLUSION

The theoretical procedure presented by means of finite element modelling has shown that deep soft soils might be decisive to long term behavior of flexible pavements, especially in the cases when shallow treatments of subgrade would be uneconomic or inefficient. Deep soil treatments should be applied to achieve an allowable capacity of soft soils up to minimum depth of about 6 m, otherwise maintenance cost of

pavements might be excessive. The analysis presented has included the response of soft subgrade layers under static and cyclic loading taking into account the influence of small-strain levels on the soil stiffness; the results fit reasonably well with the analytical solution of hysteretic damping ratio presented by Brinkgreve et al. (2007).

5 REFERENCES

Barksdale R. G. 1971. Compressive stress pulse times in flexible pavements for use in dynamic testing. Highway Research Record 345. pp 32-44. Highway Research Board.

Benz T. 2006. Small-strain stiffness of soils and its numerical consequences. Ph.d. thesis. Universität Stuttgart.

Brikgreve R. B. J. Kappert M.H. and Bonnier P.G. 2007. Hysteretic damping in a small-strain stiffness model. NUMOG X. 737-742.

Chai J. C. and Miura N. 2000. Traffic load induced permanent deformation of low road embankment on soft subsoil. Proceedings of International Conference on Geotechnical and Geological Engineering. CD Rom, Paper No. DE0239

Dobry R. and Vucetic M. 1987. Dynamic properties and seismic response of soft clay deposits. Proc. International Symposium on Geotechnical Engineering of soft soils. Mexico city. pp 51-87.

Duncan J.M. and Chang C.Y. 1970. Nonlinear analysis of stress and strain in soil. J. Soil Mech. Found. Div. ASCE 96. 1629-1653.

Elias V. Welsh J. Warren J. Lukas R. Collin J.G. and Berg R.R. 2004. Ground Improvement Methods. Participant Notebook. NHI Course 132034. FHWA NHI-04-001. Washington. D.C. 1022 pp.

Giroud J.P. and Noiray L. 1981. Geotextiles-reinforced unpaved road design. ASCE. Journal of Geotech. Engg. 107(9). 1233-1253

Hardin B. O. and Drnevich V. P. 1972. Shear modulus and damping in soils: Design equations and curves. Proc. ASCE: Journal of the Soil Mechanics and Foundations Division. 98(SM7). 667-692.

Idriss I. M. Dobry R. and Singh R.D. 1978. Nonlinear behavior of soft clays during cyclic loading. Journal of Geotech. Engg. ASCE. vol. 104. No. GT12. Dec. pp. 1427-1447.

Ishihara K. 1996. Soil Behaviour in Earthquake Geotechnics. Oxford Engineering Science Series. Oxford University Press.

Korkiala-Tanttu L. and Laaksonen R. 2004. Modelling of the stress state and deformations of APT tests. In Proc. of the 2nd Int. Conf. on Accelerated Pavement Testing. Minnesota. Worel, B. 22 p.

Ladd C. C. 1991. Stability evaluation during stage construction. Journal of Geotechnical Engineering. ASCE. Vol. 117. No. 4. pp. 541-615.

Li D. and Selig E.T. 1996. Cumulative plastic deformation for fine-grained subgrade soils. Journal of Geotechnical Engineering. ASCE. Vol. 122. No. 12. pp. 1006-1013.

Santos J. A. and Correia A.G. 2001. Reference threshold shear strain of soil. Its application to obtain a unique strain-dependent shear modulus curve for soil. Proc. 15th Int. Conf. on Soil Mechanics and Geotechnical Engg. Istanbul. Vol 1. 267-270.

Schanz T. 1998. Zur Modellierung des Mechanischen Verhaltens von Reibungsmaterialen. Habilitation. Stuttgart Universität.

Schanz T. Vermeer P.A. and Bonnier P.G. 1999. The hardening-soil model: Formulation and verification. In R.B.J Brinkgreve. Beyond 2000 in Computational Geotechnics. Balkema. Rotterdam. 281-290.

Sonderman W. and Wehr W. 2004. Deep vibro techniques. In Moseley M.P. and Kirsch K. eds. Ground Improvement 2nd edition. Spon Press. London and New York.

Vucetic M. and Dobry R. 1991. Effect of soil plasticity on cyclic response. Journal of Geotech. Engg. ASCE. vol. 117. pp. 89-107.

1255

The Use of Jet Grouting to Enhance Stability of Bermed Excavation

L'utilisation de Jet Grouting pour améliorer la stabilité d’une excavation avec risbermes

Cheuk J.C.Y., Lai A.W.L., Cheung C.K.W. AECOM Asia Co. Ltd., HKSAR, China

Man V.K.W., So A.K.O. MTR Corporation Ltd., HKSAR, China

ABSTRACT: Jet grouting has been widely used as a ground treatment method to improve the mechanical behaviour of soft soils inmany different types of constructions. This technique has been used to facilitate the construction of the West Kowloon Terminus(WKT) of the Hong Kong section of the Guangzhou-Shenzhen-Hong Kong Express Rail Link (XRL) in Hong Kong. The constructionsequence of the central portion of the deep excavation involves the formation of temporary cut slopes which serve to partially support the diaphragm wall until the core station structure is completed. The stability of the temporary cut slopes, hence the excavation, isaffected by the presence of soft marine deposits. Jet grout columns are therefore constructed to enhance the overall stability of the temporary cut slopes before excavation. The paper discusses the design philosophy of the deep excavation supported by temporary cutslopes which have been pre-treated by jet grouting. The performance of the jet grout columns have been verified by post-construction coring together with in-situ and laboratory testing. Results of these verification measures and field monitoring data which demonstrate the overall performance of the excavation supported by slopes treated by jet grouting are also presented.

RÉSUMÉ : Le jet grouting a été largement utilisé comme une méthode de traitement du sol pour améliorer le comportementmécanique des sols mous dans de nombreux types de constructions. Cette technique a été utilisée pour faciliter la construction duTerminus West Kowloon (WKT) de la section située à Hong Kong de la liaison ferroviaire express Guangzhou-Shenzhen-Hong Kong (XRL). La séquence de la construction de la partie centrale de l'excavation profonde implique la création de talus temporaires quiservent à supporter partiellement la paroi moulée jusqu'à ce que la structure de la station de base soit terminée. La stabilité des talus temporaires, et donc de l'excavation, est affectée par la présence de dépôts marins mous. Des colonnes de jet grouting sont doncréalisées de manière à améliorer la stabilité globale des talus temporaires avant excavation. Cet article traite du concept de dimensionnement de l'excavation profonde soutenue par des talus temporaires pré-traités par jet grouting. La performance des colonnes de jet grouting a été vérifiée par un carottage post-construction associé à des essais in-situ et en laboratoire. Les résultats de ces mesures de vérification et de surveillance des données en place qui démontrent la performance globale de l'excavation soutenue par des pentes traitées par jet grouting sont également présentés.

KEYWORDS: Deep excavation, slope stability, soil berm, soft clay, diaphragm wall, jet grouting

1 INTRODUCTION

The West Kowloon Terminus (WKT) is the underground terminus of the Hong Kong section of the Guangzhou-Shenzhen-Hong Kong Express Rail Link (XRL). The multi-storey 10-hectare terminus, located at West Kowloon to the north of the West Kowloon Cultural District (WKCD), will be linked to the Kowloon Station and Austin Station and is expected to be commissioned in 2015. The entire station will be an underground structure with an iconic roof erected above it.

The construction of the underground terminus involves deep excavation of about 30m supported by a 1.5m thick reinforced concrete diaphragm wall. To meet the tight programme, a construction sequence involving open cut excavation supported by the diaphragm wall and temporary cut slopes formed in front of the wall – sometimes referred to as bermed excavation – was adopted at the central portion of the deep excavation as shown in Figure 1. The concept of using soil berm as lateral support for deep excavation can be dated back to 1960s (Peck, 1969). Various analysis methods have been discussed in Simpson and Powrie (2001). These methods have also been used to back analyse centrifuge model tests (e.g. Powrie and Daly, 2002).

The temporary cut slopes serve to provide lateral support to the diaphragm wall against earth and water pressures. The stability of the temporary cut slopes, hence the excavation, is adversely affected by the presence of soft marine deposits which is overlain by reclamation fill. To enhance overall

stability of the temporary cut slopes and to reduce wall deflection during excavation, jet grout columns (JGCs) were constructed before excavation. The construction of jet grout columns involved the use of a high energy jet of fluid to break up and loosen the ground, and subsequently replaced the slurry by cement grout.

This paper discusses the design philosophy of the deep excavation supported by temporary cut slopes which have been pre-treated by jet grouting. The performance of the jet grout columns was verified by post-construction coring together with in-situ and laboratory testing. Results of these verification measures are also presented.

N

Figure 1. Deep excavation at West Kowloon Terminus (WKT).

1256


2 DESIGN OF THE DEEP EXCAVATION

The 30 m deep excavation at WKT has a plan area of approximately 550 m by 220 m, and is surrounded by high-rise residental and commerical buildings in West Kowloon – an urban area of Hong Kong. Construction of large diameter bored piles and rock-socketted steel H-piles for supporitng the station structure and the perimeter diaphragm wall was carried out before excavation commenced. The diaphragm wall serves as the temporary retaining structure for the deep excavation, the permenent wall of the station structure, as well as a load bearing wall to support the vertical load from the superstructure. This Section describes the ground conditions of the site, the construction seqenence and the design considerations of the deep excavation.

2.1 Ground conditions

The solid geology within the site area comprises Kowloon Granite from the Cretaceous Period of the Mesozoic Era which is a monzogranite pluton centered on Kowloon and Hong Kong Island. The superficial deposits include fill and transported materials such as alluvium, colluvium, marine deposits, estuarine deposits and the like.

Reclamation fill has been placed on the site following the West Kowloon reclamation works carried out in the 1990s. The fill comprises the reclamation deposits and the remnants of the old seawalls, old breakwaters, revetments, old ferry pier and existing layers of building debris and rock fill. The in-situ deposits include weathered rock and the soil derived from the weathered rock such as saprolite and residual soil (i.e. the Grade VI material according to GEO (1988)). Variably jointed rock and soil masses with different proportions of rock and soil are present within the in-situ deposits.

The site topography prior to the bulk excavation is generally characterised as flat and the average ground level is at about +5.5 mPD. The available groundwater level records from standpipes and vibrating wire piezometer indicate that the highest recorded groundwater level ranged from +1.56 mPD to +3.60 mPD and the lowest groundwater level recorded ranged from -1.29 mPD to +1.94 mPD.

2.2 Construction sequence

Excavation and construction of the station structure is carried out by two separate contractors who adopt different construction methods. The contractor on the northern side adopts open cut excavation (see Figure 1) which is then followed by bottom-up construction. On the southern side, the contractor adopts top-down construction for the top two levels of slabs and then changes to bottom-up construction for lower levels.

The focus of this paper is the open cut excavation carried out near the northern part of the site. Upon completion of the foundation works and diaphragm wall construction, jet grouting was carried out at locations where soft marine clayey deposits are present which adversely affect the stability of the bermed excavation. After sufficient strength had been gained in the jet grouted material, temporary cut slopes at a gradient of 1 on 2 were formed in front of the diaphragm wall. This was followed by bottom-up construction of the central core of the station structure. Construction of the entire structure would then be completed by top-down method between the diaphragm wall and the core station structure where the temporary slopes are gradually removed and replaced by reinforced concrete slabs connecting the diaphragm wall to the core station structure.

2.3 Design considerations

Figure 2 shows a simplified design scenario which illustrates the design concept. The diaphragm wall and the soil berm are to support an excavation with a maximum depth of approximately 30 m. Since the diaphragm wall is designed as a foundation

element for vertical loading, the wall is founded on the bedrock with a nominal embedment of 300 mm. The sufficiency of the embedment depth of the wall has been checked by trial wedge methods. For areas with a shallow rockhead, mini piles, denoted as shear pins, were constructed beneath the base of the diaphragm wall to increase the resistance against overturning.

Figure 2. Simplified design scenario of bermed excavation design with ground treatment by jet grouting.

The trial wedge method considers the soil berm as a passive support as far as stability of the embedded wall is concerned. The stability of the soil berm has to be considered separately. The marine deposits sandwiched between the reclamation fill and alluvium consist of interbeded cohesive and granular materials. The cohesive portion typically comprises of clay and in places silt with various proportion of sands and gravels. The undrained shear strength ranged from a few kilopascals to >200 kPa. A general design value of 30 kPa was adopted, except for marine clay at shallow depths where a lower value of 20 kPa was used.

(a)

(b)

Figure 3. Predictions of failure mode from finite element analysis (a) total deformation, (b) incremental shear strain distributions

Finite element analyses have been conducted using Plaxis to predict the behaviour of the bermed excavation for the

1257


simplified design scenario shown in Figure 2 if ground treatment is not carried out. In the absence of ground treatment, equilibrium solutions could not be obtained and the likely failure mechanisms are shown in Figure 3. The predicted failure mechanism involves concentrated shear strains mobilised along the base of the marine clay layer due to the constant strength assumption. The overall stability is directly dictated by the elevation of the soft marine clay layer which controls the overburden pressure exerted on the potential failure soil mass.

To enhance the overall stability, jet grouting works was proposed. The potential working principle of the jet grout columns has been studied by finite element analyses. The technique of strength reduction has been used in the analysis to quantify the margin of safety and to identify the most critical failure mechanism of the design scheme. In the analyses, several rows of 2 m diameter discrete jet grout columns which pass through the marine clay and alluvium layers are assumed. The uniaxial compressive strength (UCS) and the Young’s modulus of the jet grouted material are assumed to be 2 MPa and 300 MPa respectively. Due adjustment of these parameters have been made in the plane strain models.

Figure 4. Finite element prediction for bermed excavation with jet grouting - incremental shear strain distributions at failure.

Figure 4 shows the incremental shear strain distributions at failure when soil strengths have been reduced by a factor of 1.45 through strength reduction calculations. It can be seen that the most critical failure mechanism no longer involves a sliding plan along the base of the marine clay layer due to the presence of the jet grout columns. Instead, a local failure in the marine clay and alluvium is observed.

In the analysis, the jet grout columns are modeled as a non-porous elastic perfectly plastic material, with the maximum shear strength governed by the Mohr-Coulomb failure criterion. Jet grouted material is brittle and therefore the mobilised shear strains in the jet grout columns have been calculated to ensure sufficient strength can be mobilised at small deformation. A limiting criterion of maximum shear strain of 0.5% has been adopted.

The design scenario depicted in Figure 2 is a gross simplification of the actual site conditions. It conservatively considers the adverse effects of the presence of soft marine clay on the overall stability. The actual characteristics of the marine clay, including its strength, thickness and elevation, may vary across the site. The finite element analyses merely confirm the feasibility of the design scheme under an extreme condition. The actual amount of jet grout columns to be installed is determined by considering the local ground conditions, excavation profiles and the characteristics of the marine clay. The key design criteria are the overall safety margin of the excavation including the slope in front of the wall, the mobilised deformation in the jet grouted material as well as the structural forces induced in the diaphragm wall. Verification of the assumed material parameters for the jet grout columns is described in the next section.

3 PERFORMANCE VERIFICATION

Although jet grouting has been used widely in many parts of the world, it has not been common in Hong Kong. Therefore, not much field data was available. As such, a site trial was carried out before commencement of the actual jet grouting works. The purpose of the site trial is to determine the control parameters of the grouting operation, for example, the grout pressure, nozzle size and lifting rate, etc. In addition, the site trial serves to confirm that the assumed design strength and stiffness of the jet grouted materials can be obtained. A total of twelve trial jet grout columns were constructed using different combinations of operation parameters. A photo showing the jet grouting works is shown in Figure 5. The entire process is fully automated with all the operation parameters shown on a digital display panel.

Figure 5. Plant used for jet grouting works with fully automated control system.

A comprehensive post-grouting investigation was carried out. This included multiple full-depth coring in the trial jet grout columns, laboratory testing of the core samples and in-situ pressuremeter tests in the core holes. Core samples were obtained at differnet depths and at different locations in order to confirm that an effective diameter of 2 m could be acheived. The core samples were tested to obtain the Young’s modulus and the UCS of the specimens. In-situ pressuremeter tests were conducted in the core holes to measure the stiffness of the grouted zone. The final operation parameters of the jet grouting works were determined according to the results of site trial.

For working jet grout columns, full-depth corings were obtained at a particular sampling frequncy as a quality control measure. Typical cores are shown in Figure 6.

Figure 6. Typical core sample obtained from jet grout column

1258


Figure 6. Measured uniaxial compressive strengths of the jet grouted materials.

Figure 7. Measured Young’s modulus of the jet grouted materials.

The UCS and Young’s modulus measured from the core samples of the working jet grout columns are plotted in Figures 7 and 8. It can be observed that the mean values of UCS and Young’s modulus are well above the assumed design values. For the cases where the measured UCS or Young’s modulus of a particular jet grout column was below the design value, a design review was carried out. The design review assessed the likely behaviour of the excavation using finite element technqiues assuming that the strength or stiffness of the jet grout columns was reduced to the measured value. If the design criteria set out in Section 2.3 could still be met, the revised design was considered acceptable. On the contrary, additional jet grout columns were constructed to ensure that the excavation is of suffcient safety margin.

The purpose of coring and testing conducted on the core samples of the working jet grout columns was to verify the design assumptions made in the design analyses. It is important

to monitor the actual performance of the construction through field instrumentation. A comprehesive instrucmentation plan has been adopted for such purposes. Inclinometers have been installed in the diaphragm wall as well as in some of the jet grout columns. The measured deflection is used to closely monitor the actual performance of the construction using the predictions made by finite element analyses as a reference.

Design value

4 CONCLUSIONS

The use of soil berm to support deep excavation is not uncommon and has been used for decades. However most of the previous cases involved excavation in stiff clay. The stability of the soil berm is therefore not a concern – at least for short term conditions.

This paper describes a case where bermed excavation is adopted in a geologically complex site. The overall stability is adversely affected by the presence of soft marine clay. Without ground treament, finite element analyses predicted a potental failure mechanism initiated from sliding along the base of the marine clay layer. In order to enhance the overal stability, jet grouting as a ground treatment method was introduced. Finite element analyses also showed that the jet grout columns could avoid the formation of a global sliding failure along the marine clay layer if sufficent strength of the jet grouted material can be mobilised at small deformation. Verification of the design parameters was done through a site trial which consisted of labortoary and in-situ tests. Quality control and verification of the acutal performance were achieved by obtaining core samples from the working jet grout columns together with a comprehensive monitoring scheme.

Design value

5 ACKNOWLEDGEMENTS

The permission to publish this paper by the Mass Transit Railway (MTR) Corporation Limited is gratefully acknowledged.

6 REFERENCES

Daly, M. P. and Powrie, W. 2001. Undrained analysis of earth berms as temporary supports for embedded retaining walls. Proceedings of the Institution of Civil Engineers, Geotechnical Engineering, 149 (4), 237-248.

Geotechnical Engineering Office (GEO) 1988. Guide to Rock and Soil Descriptions, Geoguide 3, Hong Kong Government.

Peck, R.B. 1969. Deep excavations and tunneling in soft ground. State-of-the-art Report, 7th International Conference on Soil Mechanics & Foundation Engineering, Mexico City, State-of-the-art Volume, 225-290.

Powrie, W. & Daly, M.P. 2002. Centrifuge model tests on embedded retaining walls supported by earth berms. Géotechnique 52 (2), 89-106.

Simpson B. and Powrie. W. 2001. Embedded retaining walls: theory, practice and understanding. Proceedings of the 15th International Conference on Soil Mechanics and Geotechnical Engineering, Istanbul.

1259

The geotechnical analysis corresponding to the high road embankments close to a bridge

L'analyse géotechnique correspondant aux remblais routiers de grande hauteur à proximité d'un pont

Chirica A. Technical University of Civil Engineering, Faculty of railways, roads and bridges, Bucharest, 020396 Romania

Vintila D., Tenea D. Ovidius University Constanta, Faculty of civil engineering, Constantza, 900524 Romania

ABSTRACT: The paper presents a complex study case corresponding to a road embankment until 10m height placed at Iassy, Romania. It is about the access embankments at a bridge with total length of about 300m. After one year of service the studiedembankments presented the following geotechnical problems: longitudinal cracks parallel and along with road axis, lateral knobs corresponding to each compacted layer, infiltrations through the backfill from the top (faulty pluvial system) and from the bottom(flooded foundation soil without drainage system). Foundation soil is metastable clay. Triaxial tests type CKoD on stress loading paths has shown that this soil is sensitive to moistening at shear stresses. The embankment is made also from a clay unusually used forsuch type of structures and several geosynthetic layers. At the top there is an elastic pavement. The footwalks over the embankment are from reinforced concrete and are bracket assembled. The finite element model has taken into account various hypothesis: 1. Modelwith the soils in natural state, 2. Model with foundation soil in flooded state, 3. Model with foundation soil in flooded state and different artificial consolidation on embankment width.

RÉSUMÉ : Cet article présente une étude de cas complexe correspondant à un remblai routier de 10m de hauteur placé à Iassy, Roumanie. L’étude porte sur les remblais d'accès à un pont d'une longueur totale de 300 mètres. Après une année de service lesremblais étudiés ont présenté les problèmes géotechniques suivants: fissures longitudinales parallèles et le long de l’axe de la route, bourrelets latéraux correspondant à chaque couche compactée, infiltrations à travers le remblai depuis le haut (système pluvial défectueux) et par le bas (sol de fondation inondé sans système de drainage). Le sol de fondation est constitué d'argile métastable. Des essais triaxiaux de type CK0D en chemin de contraintes ont montré que ce sol est sensible à l'humidification lorsqu’il est soumis à des contraintes de cisaillement. Le remblai est également constitué d'une argile inhabituelle pour ce type de structures et est renforcé de plusieurs couches de matériaux géosynthétiques. Au sommet se trouve une couche de roulement souple. Les trottoirs en crête deremblai sont en béton armé et sont assemblés sur place. Le modèle aux éléments finis a pris en compte différentes hypothèses: 1. Modèle avec les sols à l'état naturel, 2. Modèle avec les sols de fondation inondés, 3. Modèle avec les sols de fondation à l’état inondé et différentes consolidations artificielles sur la largeur du remblai.

KEYWORDS: high road embankments, seismic loads, geotechnical investigations.

1 INTRODUCTION. FIRST LEVEL HEADING

The object of this paper is to present the analysis of a road embakment with variable height (between 4-10m) placed at Iassy, Romania (Fig. 1, 2, 3, 4). This embakment presents from the first year of service longitudinal cracks parallel with road axle, lateral knobs corresponding to each compacted layer, infiltrations through the backfill from the top and from the bottom. The owner employed an investigation team to identify, analyse and propose consolidation work for this embakment. Investigation team lead by the first author paied a visit to establish the “to do” list. First of all we identify the problems named before. After this we have made an geotehnical study to identify corectly the soil parameters, dimensions of foundations and water level. Based on this parameters we were able to do an analysis of this embakments using Plaxis software. The analysis contains the folowing models: 1. Model with the soils in natural state, 2. Model with foundation soil in flooded state, 3. Model with foundation soil in flooded state and diferent artificial consolidation on embankment width.

Figure 1 General view of the embankment

Figure 2 Lateral view of the embankment.

1260


Figure. 3 Longitudinal view of the embankment.

Figure 4 Zone between embankment - bridge.

2 GEOTECHNICAL INVESTIGATIONS

Geotechnical studies show the followings: - the lithology of soil is: vegetable soil 0,5m, black/yellow

plastic clay for up to 5,00m (Bahlui clay), saturated sand, saturated sand with gravel (5-7m) and marl clay from 12m;

- underground water from 2-4m from terrain level, this level can be ascensional with 0,8m;

- peak ground acceleration ag=0,2g, Tc=1sec (P100-2006); - Bahlui clay is very active, with high compressibility and big

variations of volume (shrinkage-belly); - plasticity index Ip = (30÷45)%; - saturation degree Sr = 0.80÷0.90; - oedometric modulus M2-3 = 4.000÷10.000 kPa; - modulus of linear deformation E 50.00kPa; - dry volumic weight γd = 14.8÷15.5kN/m3; - natural volumic weight γ = 18.75÷19kN/m3; - porosity n = (40÷45)%; - void ratio e = 0.838; - angle of internal friction Ø = 12⁰÷16⁰; - cohesion c = (15÷25)kPa.

For construction supervision of soil works have been made

the following tests: (a) tests in open system (CK0D), for which the specimens during shearing until breaking have been in contact with water from the beginning, soil being free to change his humidity with the raising the intensity of shearing force, (b) tests in closed system (CK0D-A), for which the specimens during shearing untill breaking have been in natural state humidity without any contact with a free source of water.

For both type of tests the specimens are consolidated under stress states coresponding to “K0 line”, after which they are sheared as presented above.

We can observe that on both loading systems, in the zone of normal stresses σ' < 0,8 daN/cm2 intrinsec curve has big values for angle of internal friction and low values for cohesion and in the zone of normal stresses σ' > 0,8 daN/cm2 situation is reversed. Also it can be seen that for closed system of testing intrinsec curve near the origin of axis Bahlui clay has values 4 times bigger for apparent cohesion c', and in the zone of normal stresses σ' > 0,8 daN/cm2 presents values a little bigger for apparent angle of internal friction '.

From tests we have seen that, invariable, the specific volume deformation of specimens tested in open system, correspond to

a reduction of void ratio through shearing and for specimens tested in closed system specific volume deformation correspond to a mechanical growing – dilatancy who appears in a specified point in load path function of latteral pressure σ3. Also it is important to note that dilatancy appears when deviatoric stresses q is in direct raport with spherical stress p and volume variation depends of q. This experimental observation has a great practical importance because it shows the zones in wich dilatancy occures funtion of the report between deviatoric stress and spherical stress.

In conclusion, material properties for analysis are: 1. Bahlui clay:

a. Dry state: γ=17kN/m3, Ø=23°, Cd=20kPa, E=15.000kPa, ν=0,30

b. Floded state: γ=21kN/m3, Ø=25°, Cd=5kPa, E=5.000kPa, ν=0,35

2. Backfill for embakment: γ=20kN/m3, Ø=20°, c=50kPa, E=18.000kPa, ν=0,30

3. Loose backfill: γ=20kN/m3, Ø=20°, c=50kPa, E=10.000kPa, ν=0,30

4. Stone layer: γ=20kN/m3, Ø=25°, c=1kPa, E=30.000kPa, ν=0,30

5. Asphalt: γ=22kN/m3, E=20.000kPa, ν=0,20 Loads are: ‐ self weight, ‐ on road – 100kN/m2, ‐ on sidewalk – 10kN/m2.

Figure 5 General section with materials. 3 FINITE ELEMENT ANALYSIS

The analysis was made using PLAXIS software. Model was

plane strain with 15 node elements. The analysis was made to predict future behavior of the

embankment. Different models were taken into consideration taken into consideration the following:

‐ Foundation soil of embankment is almost every time of the year flooded. Bahlui river is not flood controlled in that area.

‐ Backfill was loose on the edges of the embankment due to the lack of technology used in civil works (Fig. 4).

The 3 models taken into analysis are: a) MODEL 1. Model with soils in natural state b) MODEL 2. Model with foundation soil in flooded state c) MODEL 3. Seismic response due to earthquake with

foundation soil in flooded state a) MODEL 1. Model with soils in natural state.

This model is the simplest model taken into consideration. This means that the properties of materials are in natural state.

1261


Figure 6 Deformed mesh.

Figure 7 Total displacements.

Figure 8 Horizontal displacements.

Figure 9 Vertical displacements.

The conclusion of this calculus is that the maximum total displacement is 280mm (at the edge the sidewalks - Fig. 9). As it can be seen in Fig. 10 the plastic points appear at the edges of the embankments and in the central part of soil foundation at a approx. depth of 5m. Also, a very important notice is that the plastic points also appear at the edge of the embankments at maximum 1m around the body of backfill.

Figure 10 Plastic points.

a) MODEL 2. Model with foundation soil in flooded state. This 2nd model is taken into consideration the flooded state

of materials.

Figure 11 Deformed mesh.

Figure 12 Total displacements.

The conclusion of this calculus is that the maximum total displacement is 33,5mm (at the edge the sidewalks - Fig. 12). As it can be seen in Fig. 15 the plastic points appear at the edges of the embankments and in the central part of soil foundation at a approx. depth of 10 m. Also, a very important notice is that the plastic points also appear at the edge of the embankments at around 5m around the body of backfill. From Stability analysis we can see that the structure is almost permanent at limit having a factor of safety 1,105. Fig. 17 show us that the embankment structure has a very small reserve in strength for seismic action.

1262


X-Ac

c. (g

)

Time

-0.1

-0.2

-0.3

-0.4

0.0

0.1

0.2

0.3

0.4

0 5 10 15

Figure 17 Time history (x-acceleration) - output.

Figure 13 Horizontal displacements. The conclusion of this calculus is that the maximum

horizontal acceleration at the top of the embankment is 0,35g. From stability analysis we can see that the structure is unstable having a factor of safety 0,995.

4 CONCLUSIONS

a) Model 1: the maximum total displacement is 280mm (at the edge the sidewalks - Fig. 6). As it can be seen in Fig. 10 the plastic points appear at the edges of the embankments and in the central part of soil foundation at a approx. depth of 5m. Also, a very important notice is that the plastic points also appear at the edge of the embankment at maximum 1m around the body of backfill.

Figure 14 Vertical displacements. b) Model 2: maximum total displacement is 33,5mm (at the edge the sidewalks - Fig. 11). As it can be seen in Fig. 15 the plastic points appear at the edges of the embankments and in the central part of soil foundation at a approx. depth of 10 m. Also, a very important notice is that the plastic points also appear at the edge of the embankments at around 5m around the body of backfill. From Stability analysis we can see that the structure is almost permanent at limit having a factor of safety 1,105. Fig. 16 show us that the embankment structure has a very small reserve in strength for seismic action.

c) Model 3: the maximum horizontal acceleration at the top of the embankment is 0,35g. From stability analysis we can see that the structure is unstable having a factor of safety 0,995. Fig. 17 shows that the embankment structure has no reserve in strength.

Figure 15 Plastic points.

Shear Strength

Shear Mob.

Shea

r Res

istanc

e

Slice #

0

50

100

150

0 10 20 30 40 50

d) All tests and calculations made underline high strain and low bearing capacity of flooded state soil foundation.

e) Soil foundation is high compressibility terrain with great sensibility at moistening under stresses according to specific macrostructure.

f) To realize this works uncohesive soils are recommended; all tests on Bahlui clay show that this material is not proper to be used safely for embankments.

5 REFERENCES

Bowles, J. E. [1997] “Foundation analysis and design”, Ed. McGraw-Hill international editions, Civil Engineering Series, Fifth Editions.

Figure 16 Shear resistance versus slice. b) MODEL 3. Seismic response due to earthquake with

foundation soil in flooded state. Chirica, A. [1995]. “Tasarea şi cedarea pământurilor

macrostructurate”, Editura UTCB For this model GEOSLOPE is used for analysis. Here a

dynamic analysis was performed according to the romanian seismic code P100-2006. A scaled accelerogram was used with peak ground acceleration of 0,2g and 15s. Time increment was 0,02s and results were saved at every 10 steps.

Ieremia, M. [1998]. “Elasticitate, plasticitate, neliniaritate” Ed. Printech, Bucuresti.

Tenea, D. [2007]. “Contributii privind metodele de tratare si ranforsare a pamanturilor cu structura metastabila in cazul cailor de comunicatie”, Ed. UTCB.

1263


1

Applicability of the Geogauge, P-FWD and DCP for compaction control

Étude des conditions d’application du Geogauge, DP et PDL dans le contrôle du compactage

Conde M.C., Lopes M.G.ISEL, Lisboa, Portugal

Caldeira L., Bilé Serra J.LNEC, Lisboa, Portugal

ABSTRACT: Soil compaction is a critical issue in the construction of highway, airport and dam embankments and foundations. Thecurrent specifications address embankment compaction in terms of dry density and moisture content. However, achieving a certaindry density and moisture content does not guarantee performance adequacy. So, a comprehensive experimental testing program isunder development, on compacted layers to investigate the feasibility of developing a stiffness-based specification for embankmentsoil compaction quality control. The field test program includes 11 test points on the upstream shell and 6 points on the downstreamshell on an earth dam during construction. In each point, 3 geogauge, 10 portable falling weight deflectometer (P-FWD) and 4dynamic cone penetrometer (DCP) tests were performed. This paper aim is to analyze the experimental data and to show thefeasibility of employing these devices for earth work evaluation.

RÉSUMÉ : Le compactage des sols est un point critique dans la construction des fondations et remblais d’autoroutes, aéroports etbarrages. Le contrôle classique du compactage se fait par la mesure du poids volumique sec et de la teneur en eau. Cependant lesmesures de ces paramètres ne garantissent pas une bonne capacité portante des couches compactées. Ainsi, un programme d’essais esten cours pour étudier la faisabilité de remplacer le contrôle du compactage par des paramètres liés à la portance du sol. Le programmed'essais comprend 11 points de mesure sur la recharge amont et 6 points de mesures sur la recharge en aval d’un barrage en terre encours de construction. À chaque point ont été pris 3 mesures avec le geogauge, 10 mesures avec le déflectomètre portable (DP) et 4mesures avec le pénétromètre dynamique léger (PDL). L’objectif principal de cet article est d'analyser et de comparer les résultatsobtenus par le geogauge, le DP et le PDL et de montrer la faisabilité d'employer ces appareils pour le contrôle in situ du compactage.

KEYWORDS: soil compaction, compaction control, Geogauge, P-FWD and DCP.

1 INTRODUCTION

Soil compaction is essential in the construction of highways,airports, bridges and embankment dams. Usually, compaction iscontrolled by measuring the dry density and the water content ofthe compacted soil. These physical properties are comparedwith reference values so that adequate mechanical and hydraulicproperties may be ensured. An alternative approach based onsoil stiffness modulus has been emerging, particularly intransportation infrastructures construction. This approach issupported by the concept that the performance requirements(e.g. soil compressibility) may not correspond to the maximumsoil dry density at its optimum water content.

This paper focuses on a comprehensive experimental testingprogram aiming to correlate the data from three devices, i.e.geogauge, portable falling weight deflectometer (P-FWD) anddynamic cone penetrometer (DCP), for stiffness or penetrationmeasurement to dry density and water content of the compactedsoil. The latter experimental data was gathered by traditionalmethods (sand cone density and microwave oven heating tests,respectively).

The ultimate goal of this study is the assessment of theapplicability of earth work evaluation and control by stiffnessperformance data, just described.

2 EXPERIMENTAL WORK

With the objective to evaluate the applicability conditions ofgeogauge, portable falling weight deflectometer (P-FWD) anddynamic cone penetrometer (DCP) in compaction control ofembankment layers, a series of tests was performed oncompacted layers of a zoned earth dam during construction atAlentejo in the southern Portugal. A unique soil was used in

both upstream and downstream shells. For this purpose, a totalof 11 test points at the upstream shell and 6 test points at thedownstream shell were considered.

Soil samples from each test point were collected close to thefield test locations and stored for laboratory characterization.

2.1 Laboratory experimental program

Laboratory tests included index tests and Proctor compactiontests, as summarized in Table 1. An example of the grain sizedistribution and Proctor curve of upstream and downstreammaterial are shown in Figure 1 and Figure 2, respectively.

Figure 1. An example of a grain size distribution of shell material.

Figure 2. An example of a Proctor curve of shell material.Table 1. Index properties and compaction tests results.

Location d max

(kN/m3)wopt

(%)wL

(%)PI

(%)AASHTO

Classif.USCS

Classif.Shell material 18.32 14.5 N/P N/P A-1-b (0) SM

1264



2

2.2 Experimental layout

In Figure 3 the field tests performed at the different locationswithin 35 m wide bands, from A to G, are summarized. Thelayout of the field tests carried out at each position is illustratedin Figure 4. A cross-shaped configuration was selected with thetraditional tests at the center and the performance base methodsat the cross ends. The sand cone density test was used tomeasure in-place unit weight, according to ASTM D 1556standard. Soil samples were also collected from each testposition for water content (w) determination in the laboratoryby the microwave oven heating procedure, following ASTM D4643 standard. Measurements with geogauge and P-FWD weretaken at surface. The DCP readings were taken continuouslyalong the compaction layer depth, i.e. 40 cm, and recordedevery 10 cm.

Figure 3. Plan view of the test positions.

Figure 4. Tests arrangement at each test position.

In order to limit the disturbance caused by each type of teston the results, the following sequence of tests was selected:geogauge, P-FWD and DCP.

2.3 Compaction control

The required properties of the compacted fill layers wereestablished during the construction of a trial embankment.Accordingly, the layers of the shells were compacted to 40 cmthickness by eight passes of a smooth steel drum vibratoryroller, model CAT 583. A minimum relative compaction of95% to the reference standard Proctor was required forcompaction approval. Further, regarding maximum watercontent deviation to optimum water content (owc) up to + 2% atthe upstream shell and between – 2% to + 1% at thedownstream shell was required.

The soils used in the dam construction came from borrowareas; therefore some degree of heterogeneity in their physicaland mechanical properties was anticipated. The compactioncontrol in heterogeneous materials, based on dry density andwater content determination at each controlled point wouldinvolve a volume of work and delay in the results presentation,with potential interference with the construction schedule. Thus,the Hilf method was selected for control of compaction work(ASTM D 5080). It allows the determination of the relativecompaction, RC, and the water content deviation from owc, w,based uniquely on the soil density value, thus without watercontent measurement and previous knowledge of the Proctorreference curve. Table 2 shows the range of results of thecompaction control by the Hilf method in upstream and

downstream shells. As summarized in Table 2 the relativecompaction and water content deviation values lie within theexpected ranges.

Table 2. Range of compaction results control by the Hilf method.

Location (kN/m3)

d max.(kN/m3)

RC(%)

w(%)

Upstream 20.2 to 21.3 17.5 to 18.8 98.5 to 99.7 +0.1 to +1.7Downstream 19.9 to 20.8 17.6 to 18.3 97.6 to 99.7 -0.7 to +0.7

2.4 Geogauge testing

The geogauge device testing procedure is based on the responseof a linear elastic medium to a dynamic force applied at thesurface. It allows the determination of the elastic Youngmodulus of the near surface material. The geogauge is acylinder with a height of 270 mm and a diameter of 280 mm, asshown in Figure 5. The equipment weighs approximately100 N. The device rests on a circular ring placed and seatedfirmly at the soil surface. The base cylinder has an outsidediameter of 114 mm and an inside diameter of 89 mm. Thegeogauge shaker scans the frequency domain between 100 and196 Hz with 4 Hz increments, totalizing 25 individualfrequencies (Alshibli et al., 2005). During the test sequence, thesmall amplitude deflection and the applied force F arerecorded, thus enabling the determination of the soil verticalspecific stiffness, the so-called geogauge stiffness (KGG). Theaverage of the 25 stiffness values is taken as the representativevalue of KGG. The elastic Young’s modulus (EGG) of the soil isthen computed by the equation:

R.ν-kE GGGG 771

1 2 (1)

where is the Poisson’s ratio and R is the radius of thegeogauge base (57.15 mm), being EGG expressed in MPa andKGG in MN/m.

Figure 5. Geogauge device.

2.5 P-FWD testing

The P-FWD device used was a Prima 100. It consists of fourmajor parts: the sensor body, load plate, buffer system andsliding weight (Figure 6).

The sensor body encloses a load cell and a geophone. Thelatter is spring mounted at the center of the load plate andmeasures the deflection of the surface caused by the impactload. The Prima 100 allows the user to vary the drop height,weight, plate diameter and the number of rubber buffers.

The adjustment of the weight and drop height allows one toadjust the impact energy. Additional drop weight increases thestress exerted by the plate. By changing the size of the loadingplate diameter the stress imparted onto the sub-grade soil mayalso be adjusted. The number of rubber buffers can be selectedto alter the duration of the load impact impulse.

1265



3

Figure 6. Prima 100 P-FWD device.

The device measures both force and deflection. The softwareenables the selection of test setup and to visualize and save thetest results. Time histories and peak values of load anddeflection are displayed in a hand-held computer (PDA). Thepeak values of load and deflection allow determining the elasticstiffness modulus, EP-FWD. The equation used to determineEP-FWD is based on the Boussinesq’s equation. It corresponds to calculating the surface modulus of a layered material under auniform circular load of radius R, assuming an uniformPoisson’s ratio:

c

FWDPRfE

21

(2)

where f is the stress distribution factor, assumed 2.0 (flexibleplate), ν is the Poisson’s ratio, assumed 0.35, is the (peak)impact stress under the loading plate (kPa), R is the P-FWDloading plate radius (150 mm) and δc is the (1st peak) P-FWDdeflection (m).

Initially, a 300 mm diameter loading plate, four buffers, a 15kg falling mass and a 0.8 m drop height were adopted. Thisconfiguration proved to be inadequate due to the excessiveenergy involved which caused the apparatus to detach from theground after impact. It was observed that this affected theaccuracy of the deflection measurement. The experimental setupwas then changed to a 10 kg falling mass and a 0.4 m dropheight, which proved to be adequate in terms of contact andreading accuracy (Conde et al., 2009). With this setup theequipment applies a contact pressure between 85 and 100 kPa.

2.6 DCP testing

Dynamic cone penetrometer (DCP) allows a simple, fast, andeconomical usage and provides continuous measurement of thepenetration resistance of embankment or pavement layers. TheDCP consists of a steel rod with a cone tip at the end. In thisstudy a light weight configuration was used, i.e. with a 10 kghammer, with a falling height of 50 cm (Figure 7).

Figure 7. DCP device.

In this study DCP testing was performed according toEN ISO 22476-2 standard. The readings were takencontinuously through the compaction layer depth, i.e. along 40cm, and recorded every 10 cm. Based on the total number ofblows required to drive the penetrometer through the layer, theaverage penetration rate at each 10 cm penetration PN10(mm/blow) or the cumulative number of blows N10 were wascalculated.

3 RESULTS ANALYSIS

To evaluate the performance of these devices as tools for thecompaction control of embankment layer, the sensitivity of theresults to variations in water content and in dry density ofgeomaterials determined by traditional methods was assessed.

3.1 KGG results

Figure 8 shows the correlation between water content (w) andKGG. The chart shows some scattering about the adjustednegative exponential trend which limits the quality of theadjustement. A significant increase of the geogauge stiffnesswith decreasing water content may be inferred from the databoth in the upstream and in the downstream shells.

Figure 8. Relation between soil stiffness, kGG, and water content, w.

Regarding the dependence of KGG on the in situ dry density,the results in Table 2 present minor variations of the relativecompaction, i.e. of d SC, thus making the correlation analysisdificult. Conde et al. (2010) analyzed these results concludingthat small and erratic sensitivity of stiffness values determinedby geogauge occurred with only relatively small variations indry density (Figure 9).

Figure 9. Soil stiffness, kGG, and dry density, d SC, results.

The joint consideration of both results seems to indicate thatsoil stiffness is only a reliable predictor of the water contentvariation. This sensitivity of stiffness to changes in watercontent were also observed by Abu-Farsakh et al. (2004) in astudy conducted on fine soils (silt, sandy clay and clay).

3.2 EP-FWD results

Figure 10 shows the relationship between the elastic stiffnessmodulus (EP-FWD) and the in situ water content (w). While theadjusted trend is again of the negative exponential type, asmaller scatter is now observed in comparison with that of thegeogauge results.

Figure 1. Prima 100 P-FWD device.

The device measures both force and deflection. The software enables the selection of test setup and to visualize and save the test results. Time histories and peak values of load and deflection are displayed in a hand-held computer (PDA). The peak values of load and deflection allow determining the elastic stiffness modulus, EP-FWD. The equation used to determine EP-FWDis based on the Boussinesq’s equation. It corresponds to calculating the surface modulus of a layered material under a uniform circular load of radius R, assuming an uniform Poisson’s ratio:

c

FWDPRfE

21

(1)

where f is the stress distribution factor, assumed 2.0 (flexible plate), ν is the Poisson’s ratio, assumed 0.35, is the (peak) impact stress under the loading plate (kPa), R is the P-FWD loading plate radius (150 mm) and δc is the (1st peak) P-FWD deflection (m).

Initially, a 300 mm diameter loading plate, four buffers, a 15 kg falling mass and a 0.8 m drop height were adopted. This configuration proved to be inadequate due to the excessive energy involved which caused the apparatus to detach from the ground after impact. It was observed that this affected the accuracy of the deflection measurement. The experimental setup was then changed to a 10 kg falling mass and a 0.4 m drop height, which proved to be adequate in terms of contact and reading accuracy (Conde et al., 2009). With this setup the equipment applies a contact pressure between 85 and 100 kPa.

1.1 DCP testing

Dynamic cone penetrometer (DCP) allows a simple, fast, and economical usage and provides continuous measurement of the penetration resistance of embankment or pavement layers. The DCP consists of a steel rod with a cone tip at the end. In this study a light weight configuration was used, i.e. with a 10 kg hammer, with a falling height of 50 cm (Figure 2).

Figure 2. DCP device. In this study DCP testing was performed according to EN ISO 22476-2 standard. The readings were taken continuously

through the compaction layer depth, i.e. along 40 cm, and recorded every 10 cm. Based on the total number of blows required to drive the penetrometer through the layer, the average penetration rate at each 10 cm penetration PN10 (mm/blow) or the cumulative number of blows N10 were was calculated.

2 RESULTS ANALYSIS

To evaluate the performance of these devices as tools for the compaction control of embankment layer, the sensitivity of the results to variations in water content and in dry density of geomaterials determined by traditional methods was assessed.

2.1 KGG results

Erreur ! Source du renvoi introuvable. shows the correlation between water content (w) and KGG. The chart shows some scattering about the adjusted negative exponential trend which limits the quality of the adjustement. A significant increase of

1266



4

The correlation between P-FWD results and dry densityvalues showed significant scatter. However, the same tendencywas verified for EP-FWD, which increases with dry densityincrease, as happened with the values obtained with thegeogauge equipment.

Similarly to the case of the KGG, the variation of EP-FWD withdry density was negligible, thus restraining the conclusionsabout the sensitivity of the elastic stiffness modulus to watercontent variation.

Figure 10. Relation between elastic stiffness modulus, EP-FWD, and watercontent, w.

3.3 DCP results

DCP tests were carried out with penetration through the layerthickness, i.e. 40 cm. The cumulative number of blows wascalculated by adding N10 for each 10 cm of penetrationsuccessively. Given the reduction of N10 in the last 10 cmpenetration, lower compaction efficiency at the lower of thelayer was identified in some test points. As the vibrating rollerwith smoth drum transmits energy to the layer from the surfaceto the base, deficiency in compaction energy is prone to occur atthe base of the compaction layer. The decrease in N10 between30 and 40 cm depth occurred mainly in the downstream shellcompacted at the dry side of optimum water content.

Whenever a relatively homogeneous condition wasidentified, equivalent conclusions were obtained based on thecommulative number of blows or on average penetration rate(Conde et al. 2012). Otherwise, it was decided to select the mostrepresentative deph for data processing. In these cases the bestquality correlation between water content and cumulativenumber of blows were obtained at 30 cm depth, as ilustrated atFigure 11. The determination coefficient is here much higherthan those obtained with the previous equipments, showing isadequacy for compaction control.

Figure 11. Relationship between cumulative number of blows at 30 cmdepth and water content, w.

Alike to the results of the other two equipments, therelationship between DCP cumulative number of blows and drydensity had a significant dispersion, and it wasn’t possible to establish a correlation. Nevertheless, the downstream shoulderpenetration was observed to be higer than that of the upstreamone.

4 CONCLUSION

In order to assess the applicability of geogauge, portable fallingweight deflectometer and dynamic cone penetrometer devicesas compaction control tools, they were used during the

construction of an earth dam in southern Portugal to control thecompaction of the upstream and downstream shells. Thefollowing conclusions and remarks may be drawn from thecurrent research : P-FWD results can be affected by an inadequate

configuration choice. Stiffness values by geogauge tests and stiffness modulus by

P-FWD tests, despite some dispersion, showed anexponential negative correlation with water content. Highercorrelation to water content was apparent on downstreamshell, i.e. at dry compaction conditions.

A good quality linear correlation between DCP results andwater content was found. As a remark, in the presence ofheterogeneous conditions within the compaction layercarefull choice of the reference testing depth is needed.

In all testing points, only a small variation in dry densitywas observed (RC between 98 and 100%), thus putting thisexperimental program off as a data base provider for theassessment of the applicability of geogauge, P-FWD andDCP to relative compaction control. Further research isnecessary with significant variation of dry density betweentests in order to clarify the correlation of the readings tocompaction.

Among the equipments used in this study the DCPequipment showed greater suitability as a compactioncontrol tool, due to the strong negative correlation withwater content values.

5 ACKNOWLEDGMENTS

The authors gratefully acknowledge the dam owner EDIA forthe permission for testing and the dam contractor MONTEADRIANO for the in situ assistance. Also thanks are due toLNEC technicians Mr. Joaquim Timóteo da Silva, Mr. RuiCoelho and Mr. António Cardoso.

6 REFERENCES

Abu-Farsakh, M. Y.; Alshibli, K.; Nazzal, M. and Seyman, E. 2004.Assessment of In-Situ Test Technology For Construction Control ofBase Courses and Embankments. Technical ReportnºFHWA/LA.04/385, Louisiana Transportation Research Center,Baton Rouge, LA. USA. 126p.

Alshibli, K. A.; Abu-Farsakh, M. and Seyman, E. 2005. LaboratoryEvaluation of the Geogauge and Light Falling WeightDeflectometer as Construction Control Tools. Journal of Materialsin Civil Engineering, 17 (5), 560-569.

American Society for Testing and Material. ASTM D4643 - 2000.“Standard Test Method for Determination of Water (Moisture) Content of Soil by the Microwave Oven Method”, ASTM International, USA.

American Society for Testing and Material. ASTM D1556 - 2007.“Standard Test Method for Density and Unit Weight of Soil inPlace by the Sand-Cone Method”, ASTM International, USA.

American Society for Testing and Material. ASTM D5080 - 2008.“Standard Test Method for Rapid Determination of PercentCompaction”, ASTM International, USA.

Conde, M. C.; Caldeira, L. and Lopes, M. G. 2010. Study of applicationconditions of the geogauge and the portable falling weightdeflectometer in compaction control. in Portuguese Proceedings ofthe XII Congresso Nacional de Geotecnia, Guimarães, Portugal.

Conde, M. C.; Lopes, M. G. and Caldeira, L. 2009. Stiffness methodsfor compaction control: the P-FWD device. Proceedings of the 17th

International Conference on Soil Mechanics and GeotechnicalEngineering, Cairo, Egypt.

Conde, M. C; Caldeira, L.; Bilé Serra, J. and Lopes, M. G. 2012. Studyof dynamic cone penetrometer performance for soil compactioncontrol. in Portuguese Proceedings of the XIII Congresso Nacionalde Geotecnia, Lisbon, Portugal.

EN ISO 22476-2. 2005. Geotechnical investigation and testing – Fieldtesting – Part 2: Dynamic probing.

1267

Equilibrium models for arching in basal reinforced piled embankments

Modèles d’équilibre par effet voute pour l'amélioration des sols de fondation par inclusions rigides

Eekelen van S.J.M. Deltares, Unit Geo-Engineering and Delft University of Technology, Netherlands

Bezuijen A. Ghent University, Belgium and Deltares, Netherlands

ABSTRACT: Several analytical models are available for describing arching in basal reinforced piled embankments using geosynthetics. Some of them are limit state equilibrium models. Two of them are frequently applied in Europe: the model of Zaeske (2001), the model of Hewlett and Randolph (1988), but both models have only been described very briefly in the English language.This paper considers these two models along with another, new one: the Concentric Arches Model (Van Eekelen et al. 2013b). The paper gives a graphical presentation of the models and summarizes and discusses them.

RÉSUMÉ : Plusieurs modèles analytiques sont disponibles pour décrire la distribution en arcs des forces dans une l'amélioration des sols de fondation par inclusions rigides et géosynthétique. Parmi eux, il y a des modèles d'équilibre aux états-limites. Deux d'entre eux sont fréquemment appliquées en Europe : le modèle de Zaeske (2001), le modèle de Hewlett et Randolph (1988, mais les deux modèles ont seulement été décrits très brièvement dans la langue anglaise. Le présent article examine ces deux modèles et les compare avec notre nouveau modèle: le Modèle Arches Concentriques (Van Eekelen et al. 2013b). L’article donne une représentation graphique des modèles qui sont résumés et discutés.

KEYWORDS: arching, piled embankments, geosynthetic reinforcement, basal reinforced load transfer platforms

MOTS-CLES: effet voutes, inclusion rigide, renforcement géosynthétique, plateforme de transfert de charge

1 DESIGN OF BASAL REINFORCED PILED

EMBANKMENTS

Many analytical design models for the design of piled embankments distinguish two calculation steps. Step 1 is the arching behaviour in the fill. This “arching step” divides the total vertical load into two parts: load part A, and the ‘rest load’ (B+C in Figure 1). Load part A, also called the ‘arching’, is the part of the load that is transferred to the piles directly.

Calculation step 2 describes the load-deflection behaviour of the geosynthetic reinforcement (GR) (see Figure 1). In this calculation step, the ‘rest load’ is applied to the GR strip between each two adjacent piles, and the GR strain is calculated. An implicit result of step 2 is that the ‘rest load’ is divided into a load part B, which goes through the GR to the piles, and a part C, resting on the subsoil, as indicated in Figure 1.

geometryproperties

loadstrain εstep 1

“arching”

load part A

load part B+C step 2“membrane”

B

AA

C C soft subsoil

B+C

support from subsoil (C)

z

GR strip

Figure 1. Calculating the geosynthetic reinforcement (GR) strain comprises two calculation steps.

This paper focuses on calculation step 1 only and thus on the determination of the load distribution in the load transfer platform. The two most interesting results of the arching step are:

1. The calculated value for the arching A (kN/pile) 2. The load distribution of B+C (kN/pile)

Van Eekelen et al. (2012a, b and 2013a) showed with experiments, numerical calculations and field measurements that load B+C is concentrated on the GR strips between each two adjacent piles, and that the load distribution on these strips approaches the inversed triangular shape, as shown in Figure 1 (right hand side of the figure). The two most applied models in Europe (Zaeske 2001 and Hewlett and Randolph 1988) are summarized, analysed and discussed in this paper.

Zaeske (2001), between several other researchers, showed the great influence of the application of a sufficient stiff GR in a piled embankment. The concentration of load on GR strips is only found for GR basal reinforced piled embankments, not for piled embankments without GR. Therefore, it is necessary to make a distinction between arching models for piled embankment with and without GR. This paper focuses on GR reinforced piled embankments only.

2 EQUILIBRIUM MODELS DESCRIBING ARCHING

In equilibrium models, an imaginary limit-state stress-arch is assumed to appear above the void between stiff elements. In the 3D situation these stiff elements are piles, in the 2D situation they are walls. The pressure on the void (GR) is calculated by considering the equilibrium of the arch. In most models, the arch has a thickness.

The model of Hewlett and Randolph (1988, see Figure 2) is adopted in the French ASIRI guideline (2012) and suggested in BS8006 (2010) as an alternative for the originally first empirical model in BS8006. The other frequently applied equilibrium model is the model of Zaeske (2001, also described in Kempfert, 2004). See Figure 3. This model is adopted in the German EBGEO (2010) and the Dutch CUR226 (2010), and is hereafter called EBGEO. These two models are of great importance.

1268


a

sx

Figure 2. Hewlett & Randolph (1988) consider the ‘crown’ element of the diagonal arch and the ‘toe’ element (just above the pile cap) of the plane strain arch as indicated in this figure.

sd

d

d/2z

Figure 3. Zaeske (2001) considers the equilibrium of the crown elements of the diagonal arches

sx

a

Hg3

DH

g2D

Figure 4. Van Eekelen et al. (2013b), the Concentric Arches Model. The load is transferred along the 3D hemispheres (right hand side) towards the GR strips and then via the 2D arches (left hand side) towards the pile caps

A third model is the concentric model presented by Van

Eekelen et al. (2013b). Figures 2 to 4 present these three models and are presented in the following sections.

3 HEWLETT AND RANDOLPH (1988)

Hewlett & Randolph (1988) based their model on 3D door trap tests, without geosynthetic reinforcement. Their analytical model consists of a series of thick-walled 3D-shells, or arches, in the embankment. They consider two arch elements separately: a ‘crown element’ and an element just above the pile cap, the ‘toe’ element, as shown in Figure 2 and Figure 5. For the toe element, the pile load (A) is calculated by assuming

radial equilibrium of the crown element in the plane strain arch (left in Figure 2) and assuming that the principal stresses follow the arches with the major principal stress and r the minor principal stress and that the arches are in a nearly-plastic situation:

1 sin1 sinp rK

r

(1)

Where (kPa) is the tangential stress, r (kPa) is the radial stress, Kp (-) is the Rankine passive earth pressure coefficient and (o) is the friction angle. The pile load (A) is obtained by integrating over the pile area, indicated in Figure 5. For the crown element, the vertical stress r;i below the crown is calculated using equilibrium of the crown element in the 3D

1269


(diagonal) arch right in Figure 2 and the limit state of equation (1). Note that this is different from figure 78 in BS8006, where the crown element is the crown of the plane strain arch (but the equations in BS8006 are correct and thus derived for the diagonal arch). The soil weight below the arch is added to calculate the pressure on the subsurface v:GR:

; 2v GR i

s a

(2)

It is supposed that the entire area between the piles is loaded by this load v;GR. This load is thus assumed to be equally distributed. The remaining load is assumed to be the pile load (A). This pile load is calculated for both the crown and the foot element. The minimum pile load of Acrown and Atoe is considered to be normative.

ri=(s-a)/2

ro=s/2

s a

Figure 5. Hewlett & Randolph (1988) detail of the ‘toe’ element

4 ZAESKE (1988)

De model of Zaeske (2001) exists of a set of scales. The crown of each scale is thicker than the feet of the scales as indicated in Figure 3. Only the diagonal of the arches is considered, and this diagonal rests on the pile caps. Zaeske derived the vertical (radial) stress σz in the central line between 4 piles by considering the vertical (radial) equilibrium of the set of crown elements of the arches, as indicated in Figure 3. He assumed that: The stress situation in the feet of the arches, thus just on top

of the pile cap, is in a nearly-plastic situation. Thus the earth pressure coefficient is maximal: K = Kp (equation (1)).

The total tangential force just above the top of the pile cap equals the total tangential force along the vertical line of crown elements. Thus the passive earth pressure coefficient K at the top of the arch is less than the K at the toe of the arch, according to the ratio of the scale width at the toe and the crown (d/sd).

This way, Zaeske determined the radial stress in the crown element of each scale and extended this downwards for the entire vertical line through the crown elements. For z=0, he finally obtained the value for the vertical pressure on the GR in the mid of 4 piles (σz0 = σz(z=o)). Zaeske assumes that this pressure σzo is constant for the entire GR area between the piles.

5 CONCENTRIC ARCHES (VAN EEKELEN ET AL.

2013B)

In the concentric arches model, 3D concentric arches (hemispheres) occur above the square between each four piles (Figure 4). These hemispheres exert part of the load to their subsurface, the square between the four piles. The rest of load is transported laterally in the direction of the GR strips. The load is then further transported along the 2D arches, in the direction of the pile caps. The 2D arches also exert part of the load to the subsurface (the GR). Thus, both the 3D hemispheres and the 2D arches exert a load on its GR subsurface, and this exerted force increases towards the exterior. The part of the load not resting on the GR is the load on the piles (arching A).

Following Hewlett and Randolph (1988), the radial stress r and tangential stress in the 2D and 3D arches is calculated by assuming radial equilibrium of the crown element and assuming that: - The principal stresses follow the arches with the major

principal stress and r the minor principal stress. These concentric hemispherical stress paths were observed in several numerical studies, like Vermeer (2010) and Nadukuru and Michalowski (2012).

- The arches are in a nearly-plastic situation (equation (1)). The forces exerted on the subsurface (the GR) are calculated

by integrating the tangential stress over the GR area. This is fully elaborated and presented in Van Eekelen et al. (2013b). Figure 6 presents the resulting load distribution on the GR subsurface. The figure shows that the load is indeed concentrated on the GR strips, and the load distribution on the GR strips indeed approaches the inversed triangular load distribution found earlier in model tests, numerical analysis and field measurements (Van Eekelen et al., 2012a, b and 2013a).

0.27

5

0.23

1

0.19

1

0.15

2

0.11

2

0.07

3

0.03

3

02004006008001000

1200

1400

1600

0.2750.224

0.1780.132

0.0860.040

1400-16001200-14001000-1200800-1000600-800400-600200-4000-200

Figure 6. Pressure exerted on the GR subsurface by the arches and hemispheres of the concentric arches model.

6 DISCUSSION

Both Hewlett and Randolph (1988) and Zaeske (2001) determine the pressure exerted on the GR at the central point between four piles only. They continue with assuming that the entire GR area is loaded with this pressure, thus resulting in an equally distributed load on the GR. The concentric arches model, however, gives a load distribution that resembles the observed load distribution: a concentration on the GR strips between adjacent piles, and approximately an inversed triangular load distribution on the GR strips.

All three considered models obtained the load distribution (Efficacy E, which is the ratio pile load (A in kN/pile) / total load (A+B+C in kN/pile), thus E = A/(A+B+C)) while assuming that the surcharge load p = 0. Afterwards, the resulting Ap=0 and (B+C)p=0 are multiplied with (H+p)/(H) to obtain the A and B+C for surcharge load p>0. This results in robust calculation models.

Hewlett and Randolph (1988) as well as Zaeske (2001) compared their analytical model with measurements in scaled model tests without GR. As discussed before, it would be better to compare with measurements in situations with GR.

Van Eekelen et al. (2013b) give many comparisons between the three models and results of scaled model tests, field measurements and numerical calculations. All with GR. Figure 7 shows a comparison with numerical calculations of Le Hello et al. (2009), Figure 8 with field measurements in a high way exit in Woerden (Van Eekelen et al. 2012c). In these figures is H (m) the embankment height, a (m) the (equivalent) width square pile cap, d (m) the (equivalent) diameter of circular pile

1270


cap, sx and sy (m) the centre-to-centre distance piles along and across the road, sd (m) the diagonal centre-to-centre distance piles, (kN/m3) the unit fill weight, p (kPa) the surcharge load and (o) the friction angle. Figure 7 also gives the minimum embankment height as required in respectively EBGEO (2010) and CUR226 (2010).

The figures, as well as most other comparisons in Van Eekelen et al. (2013b), show that the concentric arch model agrees best with the numerical calculations, and most measurements in the scaled model tests. For the considered field test, the model of Zaeske and the concentric arches model give comparable good results.

0%

10%

20%

30%

40%

50%

60%

70%

80%

0.0 0.5 1.0 1.5 2.0 2.5

load

par

t A (p

erce

ntag

e of

tota

l loa

d, %

)

H/(sd-d) (m)

concentric archesLe Hello et al. 2009Hewlett and RandolphEBGEOEBGEO minimum HCUR minimum H

a=0.6m, sx=sy=1.5m, gamma=19 kN/m3, p=0kPa, phi=29deg

Figure 7. Variation of embankment height H, comparison analytical models with numerical calculations of Le Hello et al. (2009).

0

40

80

120

160

arch

ing

A (k

N/p

ile)

A pile 692A pile 693EBGEO/CUR A (phi=43)BS8006 A (phi=43)conc arches var 2 (phi=43)

=43o, average values geometry: sx=sy=2.25 m, H=1.86 m, 17 m soft soil: k=0 kN/m3

Figure 8. Comparison measured and calculated arching A in highway exit Woerden, the Netherlands, described in Van Eekelen et al., 2012c

7 CONCLUSIONS

It is important to make a distinction between models for piled embankments with or without geosynthetic basal reinforced (GR). In the case with GR, the load is concentrated on the GR strips between the piles (and the piles), and the load on the GR strips is inversed triangular distributed. This paper deals with the situation with GR.

The paper summarizes three equilibrium models describing arching in GR basal reinforced piled embankments, namely the models of Hewlett and Randolph (1988), Zaeske (2001) and the concentric arches model of Van Eekelen (2013b).

It is shown how the three models obtain their load distribution. Hewlett and Randolph (1988) as well as Zaeske (2001) find an equally distribution load on the GR between the piles. The concentric arches model (Van Eekelen et al. 2013b) finds a load concentration on the GR strips, and approximately an inversed triangular load distribution on those GR strips. This is more in accordance with observations in scaled model tests, numerical analysis and field measurements. The considered numerical calculations agree best with the concentric arches model. Measurements in the field agree equally well with the concentric arches model and the model of Zaeske (2001).

8 ACKNOWLEDGEMENTS

The financial support of Deltares and the financial support and fruitful discussions with manufacturers Naue, TenCate and Huesker for the research is greatly appreciated.

9 REFERENCES

ASIRI, 2012. Recommandations pour la conception, le dimensionnement, l'exécution et le contrôle de l'amélioration des sols de fondation par inclusions rigides, ISBN: 978-2-85978-462-1 (in French with in the appendix a digital version in English).

BS8006-1:2010. Code of practice for strengthened/reinforced soils and other fills, BSI 2010, ISBN 978-0-580-53842-1.

CUR 226, 2010. Ontwerprichtlijn paalmatrassystemen (Design Guideline Piled Embankments), ISBN 978-90-376-0518-1 (in Dutch).

EBGEO, 2010 Empfehlungen für den Entwurf und die Berechnung von Erdkörpern mit Bewehrungen aus Geokunststoffen e EBGEO, vol. 2. German Geotechnical Society, Auflage, ISBN 978-3-433-02950-3. (in German, also available in English): Recommendations for Design and Analysis of Earth Structures using Geosynthetic Reinforcements EBGEO, 2011. ISBN 978-3-433-02983-1 and digital in English ISBN 978-3-433-60093-1.

Hewlet, W.J., Randolph, M.F. Aust, M.I.E., 1988. Analysis of piled embankments. Ground Engineering, April 1988, Volume 22, Number 3, 12-18.

Kempfert, H.-G., Göbel, C., Alexiew, D., Heitz, C., 2004. German recommendations for reinforced embankments on pile-similar elements. In: Proceedings of EuroGeo 3, Munich, pp. 279-284.

Le Hello, B., Villard, P., 2009. Embankments reinforced by piles and geosynthetics – Numerical and experimental studies with the transfer of load on the soil embankment. Engineering Geology 106 (2009) pp. 78 – 91.

Nadukuru, S.S., Michalowski, R.L., 2012. Arching in Distribution of active Load on Retaining Walls. Journal of geotechnical and geoenvironmental engineering, May 2012. 575-584.

Van Eekelen, S.J.M., Bezuijen, A., Lodder, H.J., van Tol, A.F., 2012a. Model experiments on piled embankments Part I. Geotextiles and Geomembranes 32: 69-81.

Van Eekelen, S.J.M., Bezuijen, A., Lodder, H.J., van Tol, A.F., 2012b. Model experiments on piled embankments. Part II. Geotextiles and Geomembranes 32: 82-94 including its corrigendum: Van Eekelen, S.J.M., Bezuijen, A., Lodder, H.J., van Tol, A.F., 2012b2. Corrigendum to ‘Model experiments on piled embankments. Part II’ [Geotextiles and Geomembranes volume 32 (2012) pp. 82e94]. Geotextiles and Geomembranes 35: 119.

Van Eekelen, S.J.M., Bezuijen, A., 2012c. Does a piled embankment ‘feel’ the passage of a heavy truck? High frequency field measurements. In: proceedings of the 5th European Geosynthetics Congress. Valencia. Vol 5. Pp. 162-166.

Van Eekelen, S.J.M. and Bezuijen, A., 2013a, Dutch research on piled embankments, Proceedings of Geo-Congres, California, March 2013.

Van Eekelen, S.J.M., Bezuijen, A., Lodder, H.J., van Tol, A.F., 2013b. Analytical model for arching in piled embankments. To be published in Geotextiles and Geomembranes.

Vermeer, P.A., Punlor, A., Ruse, N., 2001. Arching effects behind a soldier pile wall. Computers and Geotechnics 28 (2001) 379–396.

Zaeske, D., 2001. Zur Wirkungsweise von unbewehrten und bewehrten mineralischen Tragschichten über pfahlartigen Gründungselementen. Schriftenreihe Geotechnik, Uni Kassel, Heft 10, February 2001 (in German).

1271

Prise en compte des effets de la surconsolidation dans la stabilité des talus

Consideration of Overconsolidation in slopes stability

Guerpillon Y., Virollet M. Egis Structures et Environnement, Seyssins, France

RÉSUMÉ : Les études de stabilité des talus de déblai dans des sols très surconsolidés posent des problèmes à savoir : le choix descaractéristiques mécaniques de cisaillement et les méthodes de calcul de stabilité. Les essais triaxiaux basse pression montrent que les courbes contrainte-déformation présentent des pics très prononcés, le pic correspondant à la surface d’état critique. Les sols surconsolidés gardent en mémoire des contraintes horizontales importantes qui ne sont pas prises en compte dans les calculs de stabilité. Dans cet article nous développons le comportement des sols surconsolidés tant au niveau des contraintes horizontales, que ducomportement géomécanique. Enfin à l’aide d’un modèle de calcul simple nous montrons l’influence des contraintes horizontales surla stabilité.

ABSTRACT: Stability studies of highly overconsolidated cutting slopes shows such as: the choice of shearing mechanic characteristics, stability calculation methods. Low-pressure triaxial tests show that stress-strain curves contain very pronounced peaks, the peak corresponding to the critical state. Overconsolidated soils retain significant horizontal stresses that are not taken into accountby stability calculations. In this article, we develop the behaviour of overconsolidated soils by consideration of both horizontalstresses and geomechanical behaviour. Finally, using a simple calculation model we show the influence of horizontal stresses on thestability.

MOTS-CLÉS : surface d’état limite, surface d’état critique, stabilité KEYWORDS : horizontal stresses, failure criterion, stability 1 INTRODUCTION

Nous allons tout d’abord analyser l’influence du déchargement généralisé d’un sol en montrant tour à tour les effets sur les contraintes horizontales et sur les caractéristiques mécaniques. Enfin, nous montrerons, à l’aide d’un modèle de calcul, l’effet des contraintes horizontales sur le coefficient de sécurité général.

2 CONTRAINTES HORIZONTALES DANS LES SOLS SURCONSOLIDÉS

Ce sont des sols qui ont subi au cours de leur histoire des contraintes beaucoup plus importantes que celles qui existent actuellement.

Fig. 1 Représentation d’un déchargement par érosion

Le terrain naturel a subi une érosion d’épaisseur , à la

profondeur

he

z la pression de consolidation est p, zhe

2.1 Effet du déchargement sur les contraintes horizontales

Nous prendrons un trièdre de référence avec 1 , 2 , 3 1

contrainte verticale et les contraintes horizontales. 2,3L’érosion se produisant sur une surface semi infinie, les déformations horizontales sont nulles. Comme de plus

, nous obtenons la relation : 1 3

1

1 3 (1)

2.2 Représentation des chemins suivis dans l’espace s, t

Nous rappelons que dans cet espace nous avons :

s 1 32

(2)

t 1 3

2 (3)

Dans l’article (ref. 1), JP. Magnan et JF. Serratrice montrent que dans le domaine des essais d’extension obtenus au triaxial sur des sols très surconsolidés, la surface d’état limite est analogue à celle qui est utilisée pour les argiles dans ce même domaine. L’équation de la droite de décharge s’écrit : t (12)s t0 (12)s0 (4) s0 et étant les coordonnées du point I qui représente l’intersection de l’ellipse avec la droite (pression des terres au repos).

t0k0

z

he

P= (He+z)

Dans cet espace les droites de Mohr –Coulomb sont symétriques par rapport à l’axe Os. Quand C touche la surface d’état limite (ellipse), il y a rupture et le point G vient sur la surface de rupture avec . cst1Comme sur les falaises de marne très surconsoldée de la région toulousaine, il n’est pas observé de surface de rupture, il est alors possible de déterminer le coefficient de Poisson, en confondant G avec l’origine. Nous avons trouvé 0,275 .

Fig. 2 Représentation du trajet de déchargement IC en surface du sol

-0,15

-0,05

0,05

0,15

0,25

0,35

0,00 0,20 0,40 0,60 0,80 1,00 1,20

s'/sp

I

'pG

C

O

t/sp

1272


Les coordonnées du point C sont obtenues en calculant l’intersection de la droite de décharge avec la surface d’état limite qui est une ellipse d’équation :

A2 s , cos tsin sp

,

AC

2

B2 tcos sp, sin 2

sp2

C2 0

avec

A 2(sin cos )

cos2

sincos

2

22 C

ACAB

6,0C

Les notations suivantes sont utilisées :

- angle entre l’axe Os et l’axe principal de l’ellipse , désignant le coefficient de pression des terres au repos ;

tg (1k0 ) /(1k0 ) k0

- pression de préconsolidation (contrainte moyenne dans un essai œdométrique)sp

,

La contrainte régnant dans le massif peut être obtenue en combinant les équations (1) à (4). Nous obtenons alors :

3

1

)1(3 0001003

kkks p(5)

p0 pression de consolidation au niveau de la surface du sol

pp 11

L’expression (5) montre qu’il subsiste des contraintes horizontales importantes dans les sols.

3 CARACTERISTIQUES MECANIQUES

Dans les sols surconsolidés se pose toujours le choix des caractéristiques de cisaillement à prendre en compte.

3.1 Aspect théorique

Lors de la réalisation des déblais, le chemin de contrainte suivi, fait qu’en premier lieu la résistance au cisaillement mobilisable est supérieure à la résistance correspondant à l’état critique (droite de Mohr - Coulomb), mais il est clair que dès que la surface d’état limite est atteinte, la résistance au cisaillement chute brutalement en revenant sur la surface d’état critique. Il faut donc caractériser la courbe d’état limite pour des sols très surconsolidésJF Serratrice (ref. 2) a étudié ces surfaces dans les domaines des très fortes surconsolidations.

3.1.1 Rappel de la méthode

La surface d’état limite à basse pression est tirée du modèle Camclay avec une représentation par une spirale logarithmique d’équation :

,0

,, )/ln( pbpppaq (6) avec :

32, rap

(7)

q a raet sont des coefficients positifs bLa relation (6) n’est valable que pour des basses pressions.

La courbe d’état critique (droite de Mohr-Coulomb) est représentée par la droite : q Mc p

25)(deg

sin3sin6 ,

,

, résM c

(8)

Le point d’intersection N a pour coordonnées :(8) p0

, 1kPa (9)

Fig. 3 Représentation de la courbe d’état limite. Document JF Serratrice

3.1.2 Détermination des paramètres a et b

Dans son article JF Serratrice fournit un tableau donnant les principaux paramètres des marnes étudiées tableau 1.

Tableau 1 : Données Serratrice

Marne W % e Mc ’ a b a/Mc pnMPa3 10,8 0,39 1,18 29,5 0,53 5,68 0,45 5,07 4 24,8 0,68 1,15 28,9 0,43 4,30 0,37 1,51 5 18,0 0,48 1,28 31,8 0,53 5,45 0,41 2,73 6 11,3 0,35 1,30 32,3 0,60 7,25 0,46 19,53 7 14,7 0,40 1,15 28,9 0,45 5,32 0,39 11,06 8 12,0 0,34 1,00 25,4 0,51 5,30 0,51 4,92 9 17,4 0,47 0,86 22,1 0,45 4,43 0,52 2,87

En utilisant les équations (6) à (9) nous obtenons :

ab

MaM

c

anc

23

ln,

(10)

En effectuant une régression linéaire entre b/a et logan, nousavons établi la relation suivante.

b / a 2,17 logan 2,207 (11)

9,00

10,00

11,00

12,00

13,00

1000 10000 100000

�an

b/a

Fig 4 Droite de régression

En reportant l’équation (11) dans l’équation (10), il est alors possible de calculer a puis b.

1273


Nous avons appliqué cette méthode à un sol ayant subi 300m d’érosionLes caractéristiques géotechniques sont reportées dans le tableau 2

Tableau 2 : Données du sol étudié

he (m) 300,00 300,00 z (m) 0,00 40,00

an (kPa) 6300 7140pn (kPa) 2405 2725

lnpn 8,1906 8,3158 b/a 10,45 10,57 Mc 1,12 1,12

Mc/a 2,26 2,26 a 0,49 0,50 b 5,17 5,25

A partir des données du tableau 2 nous avons déterminé la courbe d’état limite en surface et à une profondeur de 40m dans l’espace p;q

Fig 5 Courbe d’état limite

Du fait de la forte surconsolidation la courbe d’état limite varie peu avec la profondeur.

3.2 Détermination des caractéristiques mécaniques

Les modèles décrits dans les paragraphes précédents montrent que la stabilité d’un talus dépend : de la surface d’état limite en premier lieu ; de la courbe de rupture de Mohr- Coulomb étant, bien entendu, que dès que l’état de contrainte touche la surface, il y a rupture. Pour ce qui concerne les caractéristiques mécaniques de Mohr- Coulomb, elles sont caractérisées par un angle de frottement et une cohésion qui est telle que .

,

c , 0

Nous allons étudier plus avant la surface d’état limite pour la modéliser simplement.

3.2.1 Représentation de la SEL dans l’espace s, t

A partir de l’espace p,q, nous avons calculé la surface d’état limite dans l’espace s, t, qui représente le lieu des sommets des cercles de Mohr.

Nous constatons que dans le domaine des basses pressions la surface d’état limite Fig 6 peut être représentée par deux droites.

Fig 6 : Courbe d’état limite espace s, t

Les notations utilisées sont les suivantes : l’angle de frottement est noté ,

la cohésion t0Ces droites ont été obtenues par régression linéaire. La première droite passant par l’origine a pour caractéristiques :

tg , 0,86 kPat 68,3,0

Pour la seconde nous obtenons : tg , 0,70 kPat 65,

0

0

1000

2000

3000

4000

0 1000 2000 3000

p kPa

3.2.2 Caractéristiques mécaniques de cisaillement

q kP

a

o État critique D’après le Tableau 2, , il en résulte , la cohésion quant à elle, est quasiment nulle

Mc 1,12 , 28c , 0

o Surface d’état limite Nous avons vu qu’elle pouvait être décomposée en deux droites. Le passage des caractéristiques déterminées dans l’espace (s,t) à l’espace (,) dans lequel la droite de Mohr- Coulomb est définie par ’ et c’, s’effectue avec les relations de passage suivantes :

presson consolidation=6300kPapression consolidation=7140kPa

sin, tg

,

,0,

costc

Nous obtenons donc : Première droite passant par l’origine

, 59

, kPac 8 Deuxième droite

, 44

Ces deux droites se coupent pour une contrainte : c , 91kPa

122kPa soit une hauteur de sol mh 80,5

4 MÉTHODE DE CALCUL

4.1 Choix de la méthode

Pour montrer l’influence des contraintes, nous avons choisi une méthode de calcul par bloc pour pouvoir prendre en compte les efforts horizontaux dus à la surconsolidation, ce qui n’est pas possible dans les autres méthodes.

Fig 7 : schéma de calcul de stabilité F

0

AEH

1274


La surface de rupture considérée est OAB. avec k 1tg2

(13) Les triangles OAE et EAF représentent respectivement les contraintes horizontales appliquées respectivement sur OA et AB.

5 APPLICATION4.2 Hypothèses de calcul

Nous avons admis : que sur la partie verticale OA les contraintes sont peu

modifiées,

Nous allons étudier un talus de hauteur H 10m avec une

pente tg 23

.

Les caractéristiques mécaniques retenues résultent du § 3, 2, 1 à savoir : ce qui signifie que les contraintes liées à la surconsolidation

restent horizontales. que sur la surface AB les contraintes décroissent de k Z

(contrainte en A) à 0 (contrainte en B) en restant horizontales.

, 59

D’après l’équation (13) il en résulte c , 8kPa

k 2,25

4.3 Détermination du coefficient de sécurité

Le principe consiste à déterminer la projection des efforts le long de la surface AB. Nous prendrons l’angle comme variable. Si OA=Z il vient :

Z H (tg tg )

tg (12)

avec : tg pente du talus.

Coefficient de poussée 4.3.1 Efforts horizontaux dus à la surconsolidation

OEA FH1 12

k Z2 Fig. 8 : Variation du coefficient de sécurité avec la poussée horizontale

EAF FH2 12

k Z(H Z ) La Figure 8 montre que pour des sols très surconsolidés, l’état de contrainte peut atteindre la surface d’état limite .et donc que dans ce cas il faudrait prendre les caractéristiques de l’état critique.

FH H2

2k tg tg

tg

4.3.2 Poids du sol glissé 6 CONCLUSION

W H2

2tg tg

tg2

Lors d'un déchargement généralisé, phénomène d'érosion par exemple, nous avons montré que les sols conservaient des contraintes horizontales élevées. Tant que la surface d'état limite n'est pas atteinte c'est cette dernière qui conditionne la stabilité. A partir des essais réalisés par JF Serratrice sur des marnes, nous avons développé une méthode pour déterminer cette surface à partir de la seule pression de consolidation mesurée ou estimée dans ces matériaux. Cette surface peut être décomposée avec des segments de droites. Enfin à partir d'une méthode de calcul simple nous avons montré l'influence des contraintes horizontales qui peuvent conduire à atteindre la surface d'état limite et donc à une rupture. Si les contraintes en place sont mal connues il est donc dangereux de dimensionner avec les caractéristiques de pic.

4.3.3 Calcul des forces normales et tangentielles

Forces normales :

FN H2

2tg tg

tg2cos k H2

2tg tg

tgsin

Le calcul implique que FN 0 :

soit k 1tg2

Dans la réalité, il se produit des déformations qui modifient l’état de contrainte. Forces tangentielles motrices :

FT H2

2tg tg

tg2sin k H2

2tg tg

tgcos

7 REFERENCES Force tangentielle résistante Il s’agit de la résistance qu’engendre la cohésion : 1. Détermination de la courbe d'état limite d'une marne-

JP Magnan, JF Serratrice Séminaire; de la géologie au calcul des ouvrages-Reconnaissance des propriétés mécaniques des terrains Grenoble 6-10 novembre 1995.

FC cH

tg cos

4.3.4 Détermination du coefficient de sécurité 2. Essais de laboratoire à haute pressions sur des marnes-

JF Serratrice- Craies et schistes Bruxelles, 20-22 mars 1995.

cos2

sin2

cossin

2cos

22

2

2

2

2

2

tgtgtgHk

tgtgtgH

tgHctg

tgtgtgHk

tgtgtgH

3. Lois de comportement des géomatériaux et modélisation par la méthode des éléments finis- P. Mestat-études et recherches des laboratoires des ponts et chaussées Série géotechnique GT 52.

1275

Effects of ballast thickness and tie-tamper repair on settlement characteristics of railway ballasted tracks

Les effets de l'épaisseur de ballast et de la réparation de lien-bourreur sur le tassement des voies chemin de fer

Hayano K., Ishii K. Yokohama National University

Muramoto K. Railway Technical Research Institute

ABSTRACT: The effects of ballast thickness and tie-tamper repair on the settlement characteristics of ballasted tracks areinvestigated by conducting a series of cyclic loading tests on model grounds. A model sleeper at a one-fifth scale was used, and tie-tamper implementation was physically simulated in the model tests in which relationships between the number of loading cycles andsleeper settlement were obtained. In addition, maximum shear strain distributions generated in the model grounds were analyzed withparticle image velocimetry. Results suggest that the 250 mm ballast thickness currently adopted as the standard design is ineffectivefor minimizing settlement that occurs when the nonlinearity of roadbed compressibility is relatively moderate. Moreover,characteristics of the initial settlement process are altered significantly after tie-tamper implementation, although the degree of gradual subsidence undergoes minimal change regardless of ballast thickness and roadbed type.

RÉSUMÉ : Les effets de l'épaisseur de ballast et de la réparation de lien-bourreur sur les caractéristiques de nivellement des voieslestées sont étudiés. Une série d'essais cycliques de chargement sur un modèle à une échelle d'un cinquième a été effectuée. L’exécution de lien-bourreur a été physiquement simulée dans les essais sur maquette. Les relations entre le nombre de cycles dechargement et le déplacement sont respectées. De plus, les distributions des contraintes de cisaillement maximales du modèle sont analysées par analyse d’image. Les résultats montrent que l’épaisseur de ballast de 250 mm adoptée actuellement comme standard estinefficace pour minimiser le tassement qui se produit pour une compressibilité de terre-plein non linéaire relativement modérée. De plus, le processus de tassement initial change considérablement après mise en oeuvre du lien-bourreur, malgré les effets minimesdurant l’implantation.

KEYWORDS: Railway ballasted track, Maintenance, Residual settlement, Model test

1 INTRODUCTION

Railway ballasted tracks, which are composed of crushed stones, rails, and sleepers, usually undergo residual settlement due to railway traffic. In order to perform appropriate maintenance on these tracks, it is important to clarify such settlement characteristics. However, optimum relationships between ballast thickness and roadbed rigidity have not been well understood1), particularly with the 250 mm thick ballast currently used as the standard design. Ballasted tracks that show a substantial amount of settlement is often restored to the original positions by tie-tamper implementation. However, the manner in which this type of implementation alters the settlement characteristics of the ballasted tracks is poorly understood.

In this study, therefore, the effects of ballast thickness and tie-tamper repair on the settlement characteristics of ballasted tracks are investigated. A series of cyclic loading tests are conducted on a model sleeper at a one-fifth scale, as shown in Fig. 1. In the loading tests, tie-tamper repair was physically simulated by inserting a small tool into the ballasts. In addition, particle image velocimetry (PIV) analysis was performed to interpret deformation of the ballasts and roadbeds.

2 MODEL GROUNDS AND CYCLIC LOADING

Figure 1 shows the model test apparatus used in this research. Model grounds at a scale of one-fifth were constructed in a sand box with interior dimensions of 800 mm wide, 304 mm deep, and 300 mm high. A duralumin footing with a width of 48 mm was used to model the sleeper. Crushed stones approximately one-fifth the size of actual ballasts were selected to model the

ballasts. The maximum particle diameter Dmax was 19 mm, and the mean diameter D50 was 8.0 mm.

Cyclic loading tests were conducted on 12 model grounds under various conditions. Crushed stones with 20, 50, and 80 mm thicknesses were constructed on four types of roadbeds (Table 1). Crushed stones were compacted to achieve a dry density of 1.60 g/cm3 in each test.

Figure 1. Model test apparatus for cyclic loading test in the case of a steel roadbed

As shown in Table 1, the roadbed in Case 1 was represented by the bottom steel plate of the sand box. The roadbed in Case 2 was constructed with Toyoura sand (Dr = 90%) with a thickness

1276


H of 60 mm, mean diameter D50 of 0.21 mm, and uniform coefficient Uc of 1.70.

Reinforced roadbeds were introduced in Cases 3 and 4. A Toyoura sand roadbed (Dr = 90%, H = 60 mm) was overlain by an asphalt mixture layer in Case 3, and a Natom sand roadbed (Dr = 95%, H = 60 mm) was overlain by the same asphalt mixture layer in Case 4. The D50 of the Natom sand was 0.70 mm, and its Uc was 3.09. The 10 mm thick asphalt mixture was composed of straight asphalt 80–100 and sands.

Table 1. Model ground conditions

Cyclic loadings were applied to the model grounds with footing at a constant displacement rate of 0.05 mm/s. The amplitude of the cyclic stress applied in Case 1 was 110 kN/m2;that applied in Cases 2, 3, and 4 was 80 kN/m2. During the cyclic loadings, consecutive images of the model grounds were captured by a digital camera. In each test, 100 cyclic loadings were first applied. Tie-tamper repair modeling was performed in the following manner. As shown in Fig. 2, the footing was reset to the initial position after 100 cyclic loadings were applied. A small spoon was next inserted into the model ground near lateral sides of the footing. After the spoon reached a fixed ground depth, it was tilted several times to permit the crushed stones to move laterally. This procedure was followed at several locations until the voids between the footing and the ground surface were completely filled by the crushed stones. Finally, additional crushed stones were introduced to the ground surface near the footing sides to produce a flat ground surface. After this tie-tamper modeling was implemented, 100 of cyclic loadings were applied again.

Figure 2. Tool and procedure used for simulating tie-tamper repair

3 RESIDUAL DEFORMATION CHARACTERISTICS

3.1 Effects of ballast thickness

The relationships between the number of cyclic loadings N and footing settlement were obtained before and after tie-tamper repair, as shown in Fig. 3. Each relationship obtained could be fitted by the following equation2):

NeC N 1 (1)

where C and are parameters representing the initial settlement process, and represents the process of gradual subsidence.

0 20 40 60 80 100

10

8

6

4

2

0

Case1-2 Case2-2 Case3-2 Case4-2

Hb=50mm

Number of cyclic loading , N

Settl

emen

t of f

ootin

g,

(mm

)

Figure 3. Relationships between number of cyclic loading cycles and footing settlement before tie-tamper implementation. Ballast thickness, Hb, = 50 mm

Figure 4. Relationships between gradual subsidence parameter and ballast thickness Hb before tie-tamper implementation

Figure 4 shows the relationships between the gradual subsidence parameter and ballast thickness Hb before tie-tamper implementation. It should be noted that 50 mm Hb was used to represent the 250 mm ballast thickness adopted for the standard design because the model size was at a scale of one-fifth. Interestingly, it is seen in the figure that was highest when Hb = 50 mm in Cases 2 and 3. High values indicate a substantial amount of gradual settlement; thus, these results suggest that the standard ballast thickness of 250 mm is ineffective for minimal settling. The residual settlement characteristics were investigated in detail with PIV analysis. First, the displacement magnitude and direction of crushed stones and roadbeds induced by 100 cyclic loadings were estimated by analyzing consecutive digital images. The distributions of maximum shear strain maxgenerated in the crushed stones and roadbeds were next calculated. Figures 5 to 7 show the results obtained from Cases 1, 2, and 3. Results could not be obtained from Case 4 because the monotonic color of dark gray Natom sand resulted in ineffective pattern matching of PIV. As shown in Fig. 5, a high value of max was noted in Cases 1-1 and 1-2 until the ground depth reached the bottom steel plate. However, the concentration of max could not be observed in Case 1-3 near the bottom steel plate. Similarly, the concentration of max could not be observed in the roadbeds for Cases 2-3 and 3-3, as shown in Figs. 6 and 7. These results indicate that when Hb = 80 mm, max can be sustained in

Case No. Roadbed Ballast thickness Hb (mm)

1-1 201-2 5011-3

Steel (Bottom plate of a sand box)

802-1 202-2 5022-3

Toyoura sand (Dr = 90%, H =60 mm)

803-1 203-2 5033-3

Toyoura sand (Dr = 90%, H =60 mm) + Asphalt mixture (layer thickness = 10 mm) 80

4-1 204-2 5044-3

Natom sand (Dr = 95%, H = 60 mm) + Asphalt mixture (layer thickness = 10 mm) 80

20 30 40 50 60 70 800.000

0.005

0.010

0.015

0.020

0.025

0.030

Case1 Case2 Case3 Case4

Ballast thickness, Hb (mm)

1277


roadbeds, which can be explained by the limited distribution of stress applied by the footing with a width of 48 mm.

Except for the cases in which Hb = 80 mm, residual settling of the footing was attributed to total compression of crushed stones and roadbed materials. In general, stress concentration in roadbeds should be higher in the Hb = 20 mm cases than those in the Hb = 50 mm cases. Therefore, owing to the plastic deformation of roadbeds, the highest value was observed in Case 4, in which Hb = 20 mm, as shown in Fig. 4.

Figure 5. Distribution of maximum shear strain generated before tie-tamper implementation (Case 1)

Figure 6. Distribution of maximum shear strain generated before tie-tamper implementation (Case 2)

Case3-1

Case3-2

Case1-1

Case3-3Case1-2

Case1-3 Figure 7. Distribution of maximum shear strain generated before tie-tamper implementation (Case 3)

Conversely, the compression of crushed stones was higher in the Hb = 50 mm cases than those in the Hb = 20 mm cases. If nonlinear compression of roadbeds is relatively moderate, the deformation modulus of the roadbeds changes slightly through the change in stress levels. In this situation, can be higher in the Hb = 50 mm cases compared to that when Hb = 20 mm.

3.2 Effects of tie-tamper implementation Case2-1

Figure 8 shows typical relationships between footing settlement and applied stress, represented by convex curves, in Case 1-1 when the 1st, 10th, and 100th cyclic loadings were applied before tie-tamper implementation. In this research, the curves were fitted by bilinear lines, and the slopes of the two lines were estimated as k1 and k2. Displacement u2 was estimated by dividing the applied stress by k2. The parameter u2 decreased and tended to show a constant value in each case with an increase in the number of cyclic loadings (Fig. 9). Therefore, these constant values will be used in the following discussion.

Case2-2

Case2-3

Figure 8. Relationships between footing settlement and applied stress in Case 1-1 before tie-tamper implementation

1278


0 20 40 60 80 1000.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

Case1-1(before) Case1-1(after)

u 2

Number of cyclic , N0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14

0.000

0.005

0.010

0.015

0.020

0.025

0.030

Before repair After repair

u2(mm)Figure 9. Relationships between displacement u2 and number of cyclic loadings in Case 1-1 before and after tie-tamper implementation

Parameters u2, C, , and in Eq. 1 were evaluated from 100 cyclic loadings conducted before and after tie-tamper implementation. The relationships between u2 and the remaining three parameters before and after tie-tamper implementation are shown in Figs. 10, 11, and 12.

Figure 12. Relationships between displacement u2 and degree of gradual settlement

Figure 10 shows that were in the range 0.9–1.5 before tie-tamper implementation regardless of roadbed type and ballast thickness. Here, represents the duration periods of the initial settlement process (Eq. 1). The figure also shows that decreased more after tie-tamper implementation than that before. This tendency can be clearly observed when u2 is higher, which indicates that the duration periods of the initial settlement process increased after tie-tamper implementation.

Parameters C and proportionally increased with an increase in u2, as shown in Figs. 11 and 12. Here, C represents the amount of initial settlement, and represents the degree of the gradual settlement. Figure 11 shows a higher decrease in Cafter tie-tamper implementation than that before. The same tendency was also clearly observed at higher u2 because roadbeds became denser as a result of cyclic loadings; therefore, the amounts of initial settlement decreased after tie-tamper implementation. Conversely, Fig. 12 shows that was nearly the same after tie-tamper implementation as that before. These results suggest that although the characteristics of the initial settlement process are significantly altered after tie-tamper repair, the degree of gradual subsidence is minimal regardless of ballast thickness and roadbed type.

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.140.0

0.5

1.0

1.5

2.0

2.5

3.0


u2(mm)

4 CONCLUSION

The effects of ballast thickness and tie-tamper repair on the settlement characteristics of ballasted tracks were investigated by conducting a series of cyclic loading tests on model grounds. The following conclusions were derived from this research: (1) The standard 250 mm ballast thickness is ineffective for minimizing settlement, particularly when the nonlinearity of roadbed compressibility is relatively moderate. (2) The characteristics of the initial settlement process are altered considerably after tie-tamper implementation; however, the degree of gradual subsidence is minimal regardless of ballast thickness and roadbed type.

Figure 10. Relationships between displacement u2 and duration periods of the initial settlement process

5 ACKNOWLEDGEMENTS

The authors would like to thank Mr. Kazunori Ito of the Railway Technical Research Institute for his assistance in conducting experiments.

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.140

2

4

6

8

10


C

u2(mm)

6 REFERENCES

Ishikawa, T. and Namura, A. 1995. Cyclic deformation characteristics of the railroad ballast in full scale tests, Journal of JSCE, No.512, 47-59 (in Japanese).

Sekine, E., Ishikawa, T and Kouno, A. 2005. Effect of ballast thickness on cyclic plastic deformation of ballasted track, RTRI Report, Vol. 19, No.2, 17-22 (in Japanese).

Figure 11. Relationships between displacement u2 and amount of initial settlement C

1279

Mécanismes de transfert de charges dans les remblais sur cavités renforcés par géotextiles : approches expérimentales et numériques

Load transfer mechanisms in geotextile-reinforced embankments overlying voids: experimental and numerical approaches

Huckert A., Garcin P. Egis Géotechnique, Grenoble, France

Villard P., Briançon L. Laboratoire 3SR, Grenoble, France

Auray G. Texinov, La Tour-du-Pin, France

RÉSUMÉ : Le dimensionnement des remblais renforcés par géosynthétiques sujets à cavités potentielles reste problématique dans la mesure où il fait appel à des mécanismes encore peu compris ou quantifiés, tels que les transferts de charge au sein de l’ouvrage. Desexpérimentations en vraie grandeur ont ainsi été menées sur des remblais granulaires non cohésifs afin de compléter les connaissances actuelles sur le sujet. Les grandeurs mesurées sont ici confrontées aux résultats de simulations numériques obtenus à partir d’un modèle couplant les méthodes finies et discrètes. Ainsi, les mesures expérimentales de déflexion et de déformation du géosynthétique valident le modèle numérique. Ce modèle numérique permet alors d’appréhender avec pertinence les mécanismes de report de charge vers les bords de la cavité.

ABSTRACT : The design of geosynthetic-reinforced embankments prone to sinkholes raises questions linked to the complexity ofvarious mechanisms that remain still not well understood or quantified, like load transfers inside the structure. Hence full-scale experiments were led on non-cohesive granular embankments in order to complete the current knowledge on the subject. Measurements are then confronted to the results of numerical simulations obtained thanks to a numerical model coupling both finiteand discrete elements methods. Experimental deflexions and strains of the geosynthetic thus validate the numerical model, which isthen used to get a better understanding of the load transfers towards the edges of the cavity.

MOTS-CLÉS: cavités, renforcements géosynthétiques, remblai, simulations numériques, expérimentations en vraie grandeur

KEYWORDS: sinkholes, geosynthetic reinforcement, embankment, numerical simulations, full scale experiment

1 INTRODUCTION

L’aménagement de nouvelles infrastructures de transport routier et ferroviaire est contraint lors de la traversée de régions où les terrains présentent de faibles caractéristiques mécaniques ou sont sujets à la formation de cavités remontant vers la surface. C’est particulièrement le cas des régions karstiques, ou sur les emplacements d’anciennes exploitations minières. Parmi les solutions de renforcement possibles, l’utilisation de géosynthétiques est une technique économique, à la mise en œuvre relativement simple. Cette solution permet de limiter les tassements en surface du remblai à des valeurs acceptables lors de la formation d’une cavité à sa base. Le dimensionnement de tels ouvrages reste néanmoins problématique car leur comportement est régi par des mécanismes complexes combinés tels que : comportement en membrane, effet voûte et transferts de charge au sein du remblai.

Les premiers travaux français effectués sur le sujet font partie intégrante du projet de recherche R.A.F.A.E.L (Renforcement des Assises Ferroviaires et Autoroutières contre les Effondrements Localisés) (Gourc et al. 1999). Ils ont donné lieu à une méthode de dimensionnement largement répandue (Blivet at al. 2001), reformulée et complétée (Villard et al. 2002, Briançon et al. 2006, Villard et al. 2008). Malgré tout, différents mécanismes tels que la répartition des charges sur la nappe géosynthétique, le foisonnement du matériau de remblai ou l’influence d’un renforcement au comportement non linéaire restent peu compris.

Afin de compléter les connaissances sur le sujet, des études expérimentales et numériques sur des remblais granulaires ou des couches de sols traitées à la chaux ont été menées dans le cadre d’un projet de recherche FUI GéoInov. Ce projet, intitulé « Conception de géosynthétiques hautes performances sous

contraintes environnementales » regroupe des industriels, des géotechniciens et des chercheurs dans l’optique d’optimiser le dimensionnement et les propriétés mécaniques des renforts géosynthétiques sur cavités. Il est co-labellisé par les pôles de compétitivité Techtera (Technical Textiles Rhône Alpes) et Fibres dans le cadre du 10e appel à projets de la DGIS (Direction Générale de la Compétitivité, de l’Industrie et des Services). Seule une synthèse des principaux résultats se référant aux matériaux de remblais granulaires est présentée ici.

2 EXPÉRIMENTATION EN VRAIE GRANDEUR

2.1 Principe et présentation des expérimentations

Afin de prendre en considération le caractère évolutif des remontées de fontis en surface (INERIS 2007) nous avons réalisé des expérimentations en vraie grandeur simulant l’ouverture progressive et concentrique d’une cavité circulaire sous un remblai renforcé à sa base par un géosynthétique. Pour ce faire, un dispositif expérimental composé d’une buse, d’une trappe, de chambres à air de grandes dimensions et de billes d’argiles, est implanté sous le remblai renforcé (Figure 1). La vidange progressive des billes d’argile et des chambres à air depuis un regard de visite permet l’ouverture de la cavité à des diamètres croissants. Une fois le dispositif mis en place, la plateforme de travail est nivelée. Le géosynthétique est alors posé et un remblai de grave roulée lavée 20/40 mm est élevé à une hauteur de 1,2 m environ.

La cavité est alors ouverte de manière concentrique en trois étapes : 0,5 m, 1 m puis 2,2 m de diamètre. Seul le diamètre

1280


final de 2,2 m est bien maîtrisé, les diamètres d’ouvertures intermédiaires ne sont pas parfaitement contrôlés.

Après la vidange complète de la cavité un délai de 20 jours a été respecté pour permettre une stabilisation des mécanismes de transferts de charge. Par la suite, la déflection de surface a été comblée et des essais de chargement réalisés. Ceux-ci ont consisté aux passages répétés d’une pelle mécanique permettant d’appliquer une surcharge de 1t au droit de la cavité. Au total, 10 allers-retours ont été réalisés.

Figure 1. Installation du dispositif expérimental permettant l’ouverture progressive et concentrique d’une cavité circulaire sous le niveau du remblai renforcé.

2.2 Plots expérimentaux et matériaux mis en place

Deux plots expérimentaux ont été réalisés, chacun mettant en œuvre des renforcements géosynthétiques de nature différente : un géosynthétique au comportement linéaire sur le plot GR1 et un géosynthétique de raideur plus faible et au comportement non linéaire sur le plot GR2. Ces renforcements sont volontairement sous-dimensionnés afin que des mesures significatives de tassements de surface, de déformation, ou de déflexion de la nappe géosynthétique, puissent être observées.

Un soin particulier a été apporté à la caractérisation en laboratoire des matériaux utilisés (Tableau 1). Les sols ont été caractérisés par une campagne d’essais à la grande boîte de cisaillement (0,3 m x 0,3 m) selon la norme NF P94-071, et les géosynthétiques ont été testés par des essais de traction selon la norme NF EN ISO 10319.

Tableau 1. Caractéristiques des matériaux sur les plots expérimentaux

Remblai Géosynthétique Plot

(kN/m3) C

(kPa) φsol (°) Type J=3%

(kN/m) remblai/gsy

(°) GR1 15,5 0 36 GSY1 2988 23

GR2 15,5 0 36 GSY2 Non

linéaire 30*

* Valeur estimée

2.3 Instrumentation

Une instrumentation spécifique est mise en place sur les plots expérimentaux. Une campagne de mesures topographiques permet d’évaluer les tassements de surface au cours des différentes étapes expérimentales. Lorsque la cavité est totalement ouverte, ces mesures sont complétées par des mesures manuelles de tassements de surface à la règle graduée.

La déflexion du géosynthétique est évaluée en cours de l’ouverture des cavités par des investigations au radar géologique. Un suivi par fibres optiques des déformations des renforcements géosynthétiques a également été réalisé.

Ces mesures sont complétées par l’installation de capteurs de pression totale sous les nappes géosynthétiques, en bord de cavité, qui permettent d’appréhender les reports de charge au sein du remblai. On note que les résultats de mesure obtenus sont soumis à la précision des différents dispositifs de mesure,

eux-mêmes contraints par les conditions de chantier et la taille relativement importante des particules du remblai granulaire.

3 SIMULATIONS NUMÉRIQUES: COUPLAGE ENTRE ÉLÉMENTS DISCRETS ET ÉLÉMENTS FINIS

Le modèle numérique (Villard et al. 2009) est basé sur un couplage entre les méthodes éléments discrets et éléments finis, qui sont utilisées respectivement pour décrire le comportement du sol granulaire et du renforcement géosynthétique.

Le remblai granulaire est simulé par un ensemble de 15000 clusters d’élancement 1.5 (constitués de deux sphères enchevêtrées) qui interagissent entre eux en leurs points de contact. Les particules du milieu granulaire sont mises en place à une porosité donnée par une procédure d’expansion qui garantit un contrôle très précis de leur agencement (Salot et al. 2009) et de leurs propriétés mécaniques. Les paramètres du modèle influant sur le comportement du matériau granulaire sont les raideurs normales et tangentielles de la loi de contact, l’angle de frottement microscopique, l’agencement et la forme des particules. Les paramètres micro mécaniques sont déterminés sur la base d’essais triaxiaux numériques de sorte que soient restituées les propriétés mécaniques de la grave roulée utilisée pour les expérimentations (notamment l’angle de frottement interne). L’utilisation d’éléments discrets permet de prendre en considération des mécanismes complexes tels que les grands déplacements, les rotations, le foisonnement et les transferts de charge.

Les éléments finis utilisés pour décrire le comportement du géosynthétique sont des éléments triangles à 3 nœuds de faible épaisseur (3200 éléments) qui permettent de décrire la nature fibreuse et les directions de renforcement de la nappe (Villard et Giraud 1998), tout en reproduisant son comportement en membrane et en tension. De fait, aucun effort de flexion ou de compression n’est considéré dans les fibres. L’interaction entre les éléments de sol et de la nappe est gérée par des lois de contact similaires à celles utilisées entre les particules de sol et permettent de restituer parfaitement le comportement d’interface.

Au final, le modèle numérique illustré en Figure 2 tient compte des caractéristiques géométriques et mécaniques des expérimentations en vraie grandeur. Il comprend un matelas granulaire à la base duquel est interposée une nappe géosynthétique. Des sphères de petit diamètre positionnées sous la nappe géosynthétique simulent l’action d’un sol support élastique. C’est le contrôle de la position des sphères support au niveau de la cavité qui permet de simuler son ouverture progressive. Pour des raisons de symétrie un quart du problème est considéré.

Figure 2. Aperçu du modèle numérique.

On notera enfin que ce modèle permet une large exploitation

des données : les déplacements, les forces de contact et les contraintes au sein du matelas granulaire ; les tensions, les déformations et les déplacements du renfort géosynthétique, ainsi que les forces d’interaction entre le sol et le géosynthétique, et ce à chaque étape d’ouverture de la cavité. La Figure 3 permet de mettre en évidence les mécanismes de transfert de charge et notamment les rotations des contraintes principales au bord de la cavité.

1281


C

ote

par r

appo

rt à

la b

ase

du

Figure 3. Aperçu en coupe des contraintes principales au sein d’un remblai de 1 m de haut renforcé par un géosynthétique de raideur 3000 kN/m, sur une cavité de 1,1 m de rayon.

4 CONFRONTATION ENTRE RÉSULTATS EXPÉRIMENTAUX, ANALYTIQUES ET NUMÉRIQUES

4.1 Outils d’analyse analytique

La méthode analytique utilisée à titre de comparaison avec les résultats expérimentaux et numériques a été établie par Villard et Briançon (2008) pour les remblais granulaires non cohésifs. Le problème est considéré en deux dimensions (déformations planes). Le géosynthétique est renforcé dans la direction longitudinale (le sens du trafic) et son comportement en tension est décrit par une loi linéaire élastique : T = J., où T, J et sont respectivement la tension, la raideur et la déformation de la nappe. Le comportement d’interface entre le renforcement et le sol est régi par une loi de frottement élastique-plastique de Coulomb. La non linéarité du comportement du géosynthétique peut être grossièrement approchée en utilisant une raideur sécante calculée dans la section de nappe la plus sollicitée.

La distribution des contraintes sur le géosynthétique au droit de la cavité et en zone d’ancrage est supposée uniforme. Trois mécanismes sont pris en compte pour le dimensionnement : la mobilisation du frottement en zone d’ancrage, l’effet membrane au droit de la cavité, et le changement d’orientation de la nappe sur le bord de la cavité.

4.2 Flèche du renforcement géosynthétique

Les Figures 4 et 5 illustrent les déplacements verticaux des géosynthétiques sur les plots GR1 et GR2 après ouverture de la cavité à 2,2 m de diamètre. Ces données sont estimées numériquement ou analytiquement, ou calculées à partir des mesures expérimentales de positionnement du géosynthétique au radar géologique. On note que les mesures expérimentales au radar géologique ont donné lieu à une interprétation minutieuse des données dont la précision est toutefois estimée à ± 2 cm. A partir de la déformée de la nappe géosynthétique et des déflexions de surfaces on peut estimer le coefficient de foisonnement du sol à environ 1.03 sur les deux plots.

En règle générale (Figures 4 et 5) pour une cavité de 2,2 m de diamètre, les ordres de grandeur des flèches des géosynthétiques sont assez semblables que l’on considère les résultats expérimentaux, numériques ou analytiques. Les résultats obtenus sont une flèche numérique de 20,3 cm pour une flèche analytique de 21,7 cm et une mesure expérimentale de 20 cm environ sur le plot GR1. Sur le plot GR2, l’estimation numérique de la flèche est de 38,2 cm pour un calcul analytique de 30,9 cm et une mesure expérimentale de 35 cm environ.

On constate (Figures 4 et 5) que les courbes expérimentales sont légèrement décalées par rapport aux courbes numériques et analytiques. Les mesures expérimentales ayant été effectuées en conditions de chantier, les profils de mesure peuvent être légèrement désaxés, ce qui expliquerait les différences entre flèches numériques, analytiques et expérimentales observées sur les plots GR1 et GR2.

Figure 4. Plot GR1 – Déplacements verticaux du géosynthétique après ouverture de la cavité. Figure 5. Plot GR2 – Déplacements verticaux du géosynthétique après ouverture de la cavité.

En comparant les résultats analytiques et numériques sur le plot GR1, on constate que la méthode analytique utilisée permet une bonne estimation des valeurs de déflexions de la nappe. Les différences constatées avec les résultats numériques peuvent s’expliquer par une mauvaise prise en compte des mécanismes de reports de charge dans la méthode analytique (hypothèse d’une charge uniformément repartie) qui peuvent être plus complexes dans la réalité. Sur le plot GR2 les écarts entre les résultats analytiques et numériques sont plus prononcés. Une explication est que la méthode analytique suppose un comportement linéaire du géosynthétique (ou l’utilisation d’un module sécant approchant) alors que le modèle numérique tient compte de manière très réaliste du comportement non linéaire.

4.3 Déformation du renforcement géosynthétique

Une comparaison des mesures expérimentales de déformation de la nappe géosynthétique par fibre optique avec les résultats de déformation des simulations numériques et des calculs analytiques est présentée sur la Figure 6. On constate pour le plot GR1 (Figure 6) que les mesures expérimentales effectuées sur une cavité de 2,2 m de diamètre sont bien corrélées aux prédictions numériques. Les résultats du calcul analytique se corrèlent également assez bien aux résultats expérimentaux en zone d’ancrage. En revanche, au droit de la cavité, les déformations analytiques sont supérieures aux valeurs numériques et expérimentales, ce qui reste logique puisque la flèche analytique est plus importante sur le plot GR1 (Figure 4).

max = 30 kPa

1

-0,05

0,00

0,05

0,10

0,15

0,20

0,25

-1,5 -1,0 -0,5 0,0 0,5 1,0 1,5Abscisse linéique par rapport au centre de la cavité (m)

Flèc

he d

u gé

osyn

thét

ique

sel

on c

alcu

l nu

mér

ique

ou

rada

r de

sol (

m)

Numérique - D=2,2m Analytique - D=2,2mExpérimental - D=2,2m

0,8 0,6

0,4

0,2 rem

blai

(m)

0

-0,2

Abscisse linéique par rapport au centre de la cavité (m) -2,5 -2 -1,5 -1 -0,5 0

-0,05

0,00

0,05

0,10

0,15

0,20

0,25

0,30

0,35

0,40

-1,5 -1,0 -0,5 0,0 0,5 1,0 1,5Abscisse linéique par rapport au centre de la cavité (m)

Flèc

he d

u gé

osyn

thét

ique

sel

on c

alcu

l nu

mér

ique

ou

rada

r de

sol (

m)

Numérique - D=2,2m Expérimental - D=2,2mAnalytique - D=2,2m

1282


Au final, on retient les valeurs de déformations suivantes au droit de la cavité de 2,2 m de diamètre : 1,02 % pour les résultats numériques, 1,07 % sur les mesures expérimentales, et 1,4 % par la méthode analytique. Là encore on peut attribuer les différences observées par une prise en compte des reports de charge analytiques différant de la réalité.

Sur le plot GR2 (Figure 6), les valeurs expérimentales des déformations n’ont pu être correctement enregistrées en raison des forts déplacements observés lors de l’ouverture de la cavité à un diamètre de 2,2 m. Les valeurs de déformation à retenir sont, pour une cavité de 2,2 m de diamètre : 3,06 % pour les résultats numériques, et 3,23 % par la méthode analytique. Ces valeurs sont plus élevées que sur le plot GR1, la raideur du renforcement géosynthétique du plot GR2 étant plus faible.

Les résultats numériques et expérimentaux ont été analysés avec la méthode analytique de dimensionnement la plus récente (Villard et Briançon, 2008). Il en ressort que si les résultats analytiques sont comparables aux données expérimentales et numériques, cette méthode peut néanmoins être optimisée. En effet l’hypothèse analytique des transferts de charge du remblai sur le géosynthétique semble perfectible. De même des développements complémentaires pour une meilleure prise en considération d’un comportement non linéaire du géosynthétique semblent nécessaires. Des travaux numériques sont engagés dans ce sens afin de préciser les mécanismes de report de charge au sein du remblai.

Enfin, les différences de comportement en zone d’ancrage constatées sur le plot GR2 entre les méthodes numériques et analytiques peuvent s’expliquer de deux manières différentes : la prise en compte du caractère non linéaire du renforcement dans la méthode analytique par des modules sécants identiques dans chaque section de la nappe (d’où des raideurs surestimées dans les zones d’ancrages) ; ou une mauvaise approximation analytique des mécanismes de transfert de charge (répartition réelle des contraintes non uniforme, ou mal évaluée). Les différences entre mécanismes de transfert de charges peuvent encore s’expliquer par la non considération du caractère progressif de la formation de la cavité par la méthode analytique, alors que la simulation numérique en tient compte.

6 REMERCIEMENTS

Les auteurs souhaitent remercier les pôles de compétitivité Techtera et Fibres, les enseignants-chercheurs Fayçal Rejiba et Albane Saintenoy des universités Paris 6 et Paris 11 pour les mesures et l’analyse des données du radar géologique, la société de terrassement Carrey TP pour le prêt du terrain, l’IUT1 de Grenoble pour le prêt du matériel topographique, ainsi que la société Texinov, pilote du projet GéoInov, pour la conception et la caractérisation des géosynthétiques testés.

7 RÉFÉRENCES

Blivet J.C., Khay M., Gourc J.P., Giraud H. 2001. Design considerations of geosynthetic for reinforced embankments subjected to localized subsidence. Proceedings of the Geosynthetics’2001 Conference, February 12-14, 2001, Portland, Oregon, USA, 741-754.

0

0,5

1

1,5

2

2,5

3

3,5

4

-4 -2 0 2 4Abscisse linéique par rapport au centre de la cavité

(m)

Déf

orm

atio

n de

la n

appe

géo

synt

hétiq

ue (%

)

GR1 - Numérique GR1 - ExpérimentalGR1 - Analytique GR2 - NumériqueGR2 - Analytique

Briançon L., Villard P. 2006. Dimensionnement des renforcements géosynthétiques de plates- formes sur cavités. Revue Française de Géotechnique, n° 117, 4° trimestre 2006, pp 51-62.

Briançon L., Villard P. 2008. Design of geosynthetic reinforcements of platforms subjected to localised sinkholes. Geotextiles and Geomembranes, Volume 26, 5: 416-428.

Gourc J.P., Villard P., Giraud H., Blivet J.C., Khay M., Imbert B., Morbois A., Delmas P. 1999. Sinkholes beneath a reinforced earthfill – A large scale motorway and railway experiment. In proceedings of Geosynthetics’ 99, Boston, Massachusetts, USA, 28-30 April 1999, 2: 833-846.

INERIS 2007. Mise en sécurité des cavités souterraines d’origine anthropique : Surveillance – traitement. Guide technique / Rapport d’étude INERIS-DRS-07-86042-02484A.

Figure 6. Plot GR1 – Déformations des géosynthétiques pour une cavité de 2,2 m de diamètre.

Salot C., Gotteland Ph. , Villard P. 2009. Influence of relative density on granular materials behavior: DEM simulations of triaxial tests. Granular Matter Vol. 11, N° 4, pp. 221-236.

Villard P., Giraud H. 1998. Three-Dimensional modelling of the behaviour of geotextile sheets as membrane. Textile Resarch Journal, Vol. 68, N° 11, November 1998, pp. 797-806

5 CONCLUSION

Des expérimentations en vraie grandeur permettant de reproduire l’ouverture concentrique d’un fontis sous un remblai renforcé à sa base par un géosynthétique ont été menées. L’instrumentation mise en place a pu être testée en conditions de chantier. Si les capteurs de pression totale ou la mesure de fibre optique n’ont pas pleinement joué leur rôle, respectivement pour des soucis d’implantation ou de capacité de mesure, les relevés topographiques et mesures de tassement ont été performants. Après un calage topographique minutieux, le radar géologique donne lui aussi des résultats très satisfaisants.

Villard P., Gourc J.P., Blivet J.C. 2002. Prévention des risques d’effondrement de surface liés à la présence de cavités souterraines: une solution de renforcement par géosynthétique des remblais routiers et ferroviaires. Revue Française de Géotechnique, 99: 23-34.

Villard P., Briançon L. 2008. Design of geosynthetic reinforcements of platforms subjected to localized sinkholes. Canadian Geotechnical Journal, volume 45, 2: 196-209.

Les mesures ont été confrontées aux résultats des simulations numériques effectuées selon un modèle couplant éléments finis et éléments discrets, avec une correspondance relativement bonne. Après ouverture de la cavité et équilibre des plots, les valeurs expérimentales et numériques de déflexion et de déformation de la nappe géosynthétique sont relativement bien corrélées, ce qui montre l’intérêt du modèle numérique pour ce type d’application.

Villard P., Chevalier B., Le Hello B., Combe G. 2009. Coupling between finite and discrete element methods for the modeling of earth structures reinforced by geosynthetic. Computers and Geotechnics (2009), doi:10.1016/j.compgeo.2008.11.005

1283

Performance Assessment of Synthetic Shock Mats and Grids in the Improvement of Ballasted Tracks

Évaluation de la performance des nappes synthétiques à effet d’amortissement et des géogrilles dans l'amélioration des plates-formes ferroviaires ballastées

Indraratna B., Nimbalkar S., Rujikiatkamjorn C. Centre for Geomechnics and Railway Engineering, University of Wollongong, Wollongong City, NSW Australia; ARC Centre of excellence in Geotechnical Science and Engineering, Australia

Neville T. Australian Rail Track Corporation Ltd. Broadmeadow NSW Australia

Christie D. Geotechnical Consultant, Hazelbrook, NSW Australia

ABSTRACT: In Australia, railways offer the most prominent transportation mode in terms of traffic tonnage serving the needs of bulkfreight and passenger movement. Ballast is an essential constituent of conventional rail infrastructure governing track stability andperformance. However, in recent times, higher traffic induced stresses due to dramatically increased train speeds and heavier axle loads have caused excessive plastic deformations and degradation of ballast. This seriously hampers safety and efficiency of express tracks, forinstance, enforcing speed restrictions and effecting more frequent track maintenance. Installing layers of synthetic materials such geogrids and rubber pads (shock mats) in rail tracks can significantly reduce ballast degradation. Field trials were conducted on rail track sectionsin the towns of Bulli (near Wollongong City) and Singleton (near Newcastle) to measure track deformations associated with cyclicstresses and impact loads. This paper describes the results of large-scale laboratory testing as well as the observations from full-scale instrumented field trials characterising the behaviour of rail ballast improved by shock mats and synthetic grids.

RÉSUMÉ : En Australie, les chemins de fer offrent le mode de transport plus important en terme de tonnage de trafic apte à répondre aux besoins de transport de passagers et de fret en vrac. Le ballast est un constituant essentiel de l'infrastructure ferroviaire conventionnellerégissant les performances et la stabilité de la voie. Toutefois, dans les temps récents, les contraintes plus fortes induites par un trafic se faisant à vitesse de plus en plus élevée et avec des charges à l’essieu plus importantes provoquent des déformations plastiques excessiveset la dégradation du ballast. Cela entrave sérieusement la sécurité et l'efficacité des voies expresses en nécessitant, par exemple, des restrictions de vitesse et un entretien des voies plus fréquent. L’installation de couches de matériaux géosynthétiques tels que les géogrilles et les nappes de caoutchouc dans les plates-formes ferroviaires peuvent réduire de façon significative la dégradation du ballast.Des essais en place ont donc été réalisés sur des sections de plates-formes ferroviaires dans les villes de Bulli (près de Wollongong) etSingleton (près de Newcastle) afin de mesurer les déformations de la voie associées à des charges cycliques et d’impacts. Cettecommunication présente les résultats des essais en laboratoire à grande échelle ainsi que des observations résultant des essais en placegrandeur nature instrumentés, caractérisant le comportement du ballast ferroviaire amélioré par les renforcements en grilles géosynthétiques.

KEYWORDS: ballast, degradation, field trial, geosynthetics, impact loads, shock mats.

1 INTRODUCTION

The rail track structure consists of rail, sleeper (crossties), ballast, sub-ballast (capping and structural-fill) and subgrade. Ballast is one of important track components and is used as the primary means of distributing of the wheel loads to underlying layers, and for holding the track in proper alignment, cross level and grade. The ballast assembly undergoes irrecoverable deformations due to particle breakage and cyclic densification. The breakage of ballast particles due to wheel loading can occur due to: (a) the particle splitting, (b) breakage of angular projections and (c) grinding of small-scale asperities (Raymond and Diyaljee 1979). In Australia, most breakage of latite ballast is primarily attributed to the presence of highly angular corners of quarried aggregates (Lackenby et al. 2007).

Several previous studies focused on the laboratory testing of the soil-geogrid interfaces (Tang et al. 2008, Liu et al. 2009) and the ballast-geogrid interfaces (Raymond 2002, Indraratna and Salim 2003, Brown et al. 2007, Indraratna et al. 2010a,b). In order to reduce ballast degradation, the use of geosynthetic grids has been recommended (Selig and Waters 1994, Indraratna et al. 2006, 2007, Indraratna and Nimbalkar 2012). The geosynthetic grids hinder the lateral movement of ballast due to frictional interlock among aggregates. The grid-particle interlock in turn increases the track stability and prolongs the maintenance period. Wheel-rail irregularities such as wheel flats produce high levels of impact loading (Indraratna et al. 2010).

This impact load induces high frequency vibration of the track components (Jenkins et al. 1974, Indraratna et al. 2011a,b,c). It has been proven that excessive impact loads aggravate ballast degradation (Indraratna et al. 2012a,b, Nimbalkar et al. 2012). A field trial was conducted on sections of an instrumented rail track in the town of Bulli (near Wollongong) and Singleton (near Newcastle) to study the effectiveness of geosynthetic grids and shock mats. This paper describes the large-scale laboratory studies and full-scale field trials.

2 USE OF SHOCK MATS IN MITIGATING BREAKAGE

In order to evaluate the effectiveness of shock mats, a large scale drop-weight impact testing equipment was used.

0

50

100

150

200

250

300

350

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20Elapsed Time, t (sec)

Impact Force Peaks P1

Impact Force Peak P2Im

pact

For

ce, F I (

kN)

Without Shock matShock mat placed at top and bottom

AB

BB

A

A

Figure 1. Typical impact force responses for stiff subgrade (data sourced from Nimbalkar et al., 2012).

1284


2.1 Test setup and procedure

A steel plate of 300 mm diameter and 50 mm thickness was used to represent a hard base such as the deck of a bridge, or hard rock. A thick sand layer of 100mm thickness was used to simulate a typical ‘weak’ subgrade. The drop hammer was raised mechanically to the required height and then released by an electronic quick release system. After 10 blows, an attenuation of strains in the ballast layer was reached.

2.2 Single impact loading

The impact load-time history under a single impact load is shown in Figure 1. Two distinct types of peak forces were seen during impact loading: (a) an instantaneous sharp peak with very high frequency P1, and (b) a gradual peak of smaller magnitude with a relatively smaller frequency P2 (Jenkins et al. 1974). It was also evident that multiple P1 type peaks followed by the distinct P2 type peak often occurred. The multiple P1peaks occurred when the drop hammer was not restrained vertically, so consequently it rebounded after the first impact and impacted the specimen again.

Figure 2. Variation of impact force with number of blows (data sourced from Nimbalkar et al. 2012).

2.3

agnitude of impact force compared to a stiffer

, 2005, Lackenby et al. 2007, Indraratna and Nimbalkar 20

fo ak sub s that the weak subgrade itself acts flexibl

Table 1. Ballas ge under impact loading (Indraratna et al., 2011a).

pe

Multiple impact loading

Figure 2 shows the variation of P2 force peak with repeated hammer blows (N). The P2 force showed a gradual increase with the increased number of blows due to the densification of ballast. A dense aggregate matrix offers a higher inertial resistance which leads to an increased value of P2. Even without a shock mat, a ballast bed on a weak subgrade leads to a decreased msubgrade.

2.4 Particle breakage

After each test, the ballast sample was sieved to obtain the ballast breakage index (BBI) as shown in Table 1. The particle breakage encountered under 10 impact blows was significantly higher than that under both static and cyclic loads (Indraratna et al. 1998

11).The higher breakage of ballast particles can be attributed to

the considerable non-uniform stress concentrations occurring at the corners of the sharp angular particles of fresh ballast under high impact stresses. When a shock mat was placed above and below the ballast bed, particle breakage was reduced by approximately 47% for a stiff subgrade, and approximately 65%

r a we as a

grade. This impliee cushion.

t breaka

Base ty Test Details BBI

Stiff Without shock mat 0.170

Stiff Shock mat at top and bottom of 0.091 ballast

Weak Without shock mat 0.080

Weak Shock mat at top and bottom of ballast

0.028

3 USE OF GEOSYNTHETICS FOR STABILISING A BALLASTED TRACK: BULLI CASE STUDY

In order to investigate deformations of a caused by train traffic, and the associat

multi-layer rail track ed benefits of using

3.1

mposite layer were the other two sections, fresh and

th a layer of geocomposite at the

The vertical and horizontal stresses induced in the track bed were measured by pressure cells. Vertical deformations of the track were measured by settlement pegs, and lateral deformations were measured by electronic displacement transducers. The settlement pegs and displacement transducers were installed at the sleeper-ballast and ballast-subballast interfaces, respectively, as shown in Figure 3.

geosynthetics in fresh and recycled ballast, a field trial was carried out on a fully instrumented track in the town of Bulli north of Wollongong City [Indraratna et al. 2009, 2010]. The proposed site was located between two turnouts.

Site geology and track construction

A site investigation comprising 8 test pits and 8 Cone Penetrometer tests was carried out to assess the condition of the sub-surface soil profiles. The subgrade consisted of a stiff over consolidated silty clay that showed high values of cone resistance (qc) and friction ratio (Rf)(Robertson 1990, Choudhury 2006).

The instrumented section of track was 60 m long and it was divided into four equal sections. The layers of ballast and subballast (capping) were 300 mm and 150 mm, respectively. Fresh and recycled ballast without a geocoused in two sections, while inrecycled ballast was used wiballast-subballast interface. The physical and technical specifications of the fresh ballast, recycled ballast and geosynthetic material used at this site have been reported elsewhere (Indraratna et al. 2011a, 2012a).

3.2 Track instrumentation

Figure 3. Installation of settlement pegs and displacement transducers at Bulli site (data sourced from Indraratna et al. 2012b)

1285


Figure 4. In-situ response of the ballast layer: lateral deformations (data sourced from Indraratna et al. 2010).

3.3 Lateral ballast deformations

Average lateral deformations of ballast are plotted against

y high indicate

o investigate the performance of different types of

study was also undertaken on an instrumented track se Singleton, Newcastle.

4.1 Site geology and nstruction

Nine experimental se ere included in the trial track while it was under con , on three diffe of sub-grades, including (i) th ely soft gener vial silt clay deposit (S 1-4 and Sec the intermediate cut siltsto tions 5 and C), he stiff reinforced concrete bri k supported by piled abutment (Section B), as shown in Table 2. Further details of track construction and material specifications c und in Indraratna et al. (2012c).

the number of load cycles (N) in Figure 4. The recycled ballast with moderately graded particle size distribution (Cu = 1.8) showed less lateral deformations compared to the very uniform fresh ballast (Cu = 1.5). Recycled ballast often shows less breakage because the individual particles are more rounded which prevents high angular corner breakage caused b

in Figure 4stress concentrations. The results presentedthat the geocomposite reduced lateral deformation of fresh ballast by about 49 % and that of recycled ballast by 11 %. The apertures of the geogrid offered strong mechanical interlocking with the ballast. The capacity of the ballast to distribute loads was improved by the placement of the geocomposite, which substantially reduced settlement under high repeated loading.

4 USE OF GEOSYNTHETICS FOR STABILISING A BALLASTED TRACK: SINGLETON CASE STUDY

Tgeosynthetics for improving the overall track stability under in situ conditions, an extensive

ctions in near the City of

track co

ctions wstruction rent types e relativections

al fill and allution A), (ii) y

ne (Sec and (iii) tdge dec a

an be fo

Table2. Reinforcement at experimental sections using geogrids, geocompolistes, and shock mats.

Section Location Reinforcement

A 234.75 -

1 234.66 Geogrid 1

2 234.40 Geogrid 2

3 234.22 Geogrid 3

4 234.12 Geocomposite

B 232.01

4.2 Track instrumentation

Shock mat

C 228.50 -

5 228.44 Geogrid 3

The strain gauges were installed in groups, 200 mm apart, and ds in both longitudinal and

5. ls o k i me n str uge

4.3

ared, gradencrete

on the top and bottom sides of the gritransverse directions (Figure 5). The strain gauges were of a post-yield type suitable to measure strains in the range of 0.1 to 15%. Two pressure cells were installed at Sections 1, 5, A and C. At these locations, one pressure cell was installed at the sleeper-ballast and another at the ballast-sub-ballast interface. At Section B, three pressure cells were installed at the synthetic mat-deck interface. Settlement pegs were also installed at the sleeper-ballast and ballast-sub-ballast interfaces to measure the vertical deformations of the ballast layer.

Figure Detai f trac nstru ntatio using ain ga s.

Vertical ballast deformations

The settlements (sv) and vertical strains (v) of the ballast layer after 2.3 105 load cycles are reported in Table 3. The vertical settlements of sections with reinforcement are generally smaller than those without reinforcement. This observation is mainly attributed to the effective interlocking between the ballast particles and grids, thus inducing increased track confinement as explained earlier. When sections a, b, and c are compthe results indicate that sv and v are larger when the substiffness becomes smaller, i.e. S is smallest on the covbridge deck and largest at the alluvial deposit.

Table 3. Vertical deformation and strain of ballast after 2.3105 load cycles.

Instrumented section details

1 2 3 4 5 A B C

Sv

(mm) 16.3 21.2 14.8 16.0 16.3 23.8 8.8 17.8

v (%) 5.4 7.1 4.9 5.3 5.4 7.9 2.9 5.9

It is also observed that the geogrid is more effective in terms

ported by Ashmawy and Bourdeau (1995) thorough geogrid at Section 3 performed better, although the tensile strength did not differ much with the other types. is is a

um aperture size (40 mm) which would enable better cking between the b parti d the rid.

4.4 Strains accumulation in geocomposites

ins after

of reducing track settlement for relatively weak subgrades. Similar observations have been re

full scale testing. The

Th ttributed to theoptiminterlo allast cles an geog

geogrids &

Accumulated longitudinal (l) and transverse (t) stra2.3 105 load cycles are given in Table 4. The transverse strains were generally larger than the longitudinal strains, and this is attributed to the ease of lateral spreading of the ballast layer upon loading. It was also observed that l and t were mainly influenced by the subgrade deformations. The strains of geogrid at Section 4 were relatively large although its higher stiffness could have resulted in smaller strains. This is because, the thicker general fill underwent large lateral deformations shortly after the track was commissioned. Induced transient strains in both longitudinal and transverse directions due to the passage of

0 1x105 2x105 3x105 4x105 5x105 6x105 7x105-14

-12

-10

-8

-6

-4

-2

-0

Ave

rage

late

ral d

efor

mat

ion

of b

alla

st, (

S h) avg (m

m)

Fresh ballast (uniformly graded) Recycled ballast (broadly graded) Fresh ballast with geocomposite Recycled ballast with geocomposite

Number of load cycles, N

1286


trains (axial load of 30 tons) travelling at 40 km/h were of magnitude in the order of 0.14-0.17 %. Table 4. Accumulated longitudinal and transverse strain in geogrid and geocomposite after 2.3105 load cycles.

Instrumented section details

1 2 3 4 5

l (%) 0.80 0.78 0.61 0.60 0.62

t (%) 0.85 1.50 0.80 1.80 0.85

5 CONCLUSIONS AND RECOMMENDATIONS

forcement and shock mats on

ere examined. The results of the Bulli field the use of geocomposites as reinforcing

ed ballast proved to be a feasible and

studrela lated strains in the

the

mofutuloa

The

(ARTC), and QR National for

Iansign

and

Ash

Cho

Indr

Indr

Indr B., Nimbalkar S. and Christie D. 2009. The performance of

Roads,

Indr

2378-2387.

Indference

Indr

nternational Conference of International Association for

nd Improvement, 10(3), 91-101.

l Engineering, M. A. Shahin & H. Nikraz

Indr rack

Journal of Railway Technology 1 (1), 195-219.

nthetic Grids in the

do 2012, 10-

Indrspeeds with heavier freight.

Sixteenth annual Symposium of Australian Geomechanics Society,Sydney Chapter, 10 October 2012, Sydney, Australia, 1-24.

Jenkins H. M., Stephenson J. E., Clayton G. A., Morland J. W. and Lyon D. 1974. The effect of track and vehicle parameters on wheel/rail vertical dynamic forces. Railway Engineering Journal 3, 2-16.

Lackenby J., Indraratna B., McDowel G. and Christie D. 2007 Effect of confining pressure on ballast degradation and deformation under cyclic triaxial loading. Géotechnique 57(6), 527-536.

Liu C. N., Ho, Y-H, and Huang J. W. 2009. Large scale direct shear tests of soil/PET-yarn geogrid interfaces. Geotextiles and Geomembranes, 27, 19-30.

Nimbalkar S., Indraratna B., Dash S. K. and Christie D. 2012 Improved performance of railway ballast under impact loads using shock mats. Journal of Geotechnical and Geoenvironmental EngineeringASCE 138(3), 281-294.

Raymond G. P. 2002. Reinforced ballast behaviour subjected to repeated load. Geotextiles and Geomembranes 20(1), 39-61.

Raymond G. P. and Diyaljee V. A. 1979. Railroad ballast load ranking classification. Journal of Geotechnical Engineering ASCE 105(10), 1133-1153.

Robertson P. K. 1990. Soil classification using the cone penetration test, Canadian Geotechnical Journal, 27, 151-158.

Selig E. T. and Waters J. M. 1994. Track Geotechnology and Substructure Management. Thomas Telford, London.

Tang X., Chehab G. R. and Palomino A. 2008. Evaluation of geogrids in stabilizing weak pavement subgrade. International Journal of Pavement Engineering, 9(6), 413-429.

The effects of geosynthetic reinthe performance of ballasted rail tracks were discussed in this paper. The use of shock mats was beneficial in terms of reduced ballast breakage and attenuated impact forces. A few impact blows were observed to have caused considerable ballast breakage (BBI = 17%). Due to the placement of shock mats, BBI could be reduced by approximately 47% over a stiff subgrade and by approximately 65% over a weak subgrade.

The performance of instrumented ballasted tracks at Bulli and Singleton was evaluated, in which different types of geosynthetics wstudy indicated that elements for recycleconomically attractive alternative. The results of the Singleton

y revealed that the effectiveness of geogrids is greater for tively weak subgrades. The accumu

geogrids were influenced by the subgrade deformation, while induced transient strains were mainly affected by the

geogrid stiffness. An in-depth understanding of the geogrid and shock mat stabilised performance would allow for safer and

re effective ballasted track design and construction in the re, especially for increased trains speeds where high cyclic

ding together with impact is almost inevitable.

6 ACKNOWLEDGEMENTS

authors are grateful to the CRC for Rail Innovation for funding a significant part of this research. The authors express their sincere thanks to RailCorp (Sydney), Australian Rail Track Corporation

their continuous support. The assistance provided by senior technical officers, Mr Alan Grant, Mr

Bridge, and Mr Cameron Neilson is also appreciated. A ificant part of the contents of this paper are described in a number cholarly joof s urnals including Géotechnique, and ASCE Journal of

Geotechnical and Geoenvironmental Engineering, as cited in the text listed below.

7 REFERENCES

mawy A.K. and Bourdeau P.L. 1995. Geosynthetic-reinforced soils under repeated loading: a review and comparative design study. Geosynthetics International 2(4), 643-678.

Brown S.F., Kwan J. and Thom N.H. 2007. Identifying the key parameters that influence geogrid reinforcement of railway ballast. Geotextiles and Geomembranes 25(6), 326-335. udhury J. 2006. Geotechnical investigation report for proposed bulli track upgrading between 311 & 312 turnouts: dn track 71.660~71.810km, up track 71.700~71.780km, Memorandum, Engineering Standards & Services Division, Geotechnical Services, NSW, Australia

Indraratna B. and Nimbalkar S. 2012. Stress-strain-degradation response of railway ballast stabilised with geosynthetics. Journal of Geotechnical and Geoenvironmental Engineering ASCE (accepted, in press). aratna B. and Salim W. 2003. Deformation and degradation mechanics of recycled ballast- stabilised with geosynthetics. Soils and Foundations 43(4), 35-46.

Indraratna B., Ionescu D. and Christie D. 1998. Shear behaviour of railway ballast based on large-scale triaxial tests. Journal of Geotechnical and Geoenvironmental Engineering ASCE 124(5), 439-439.

aratna B., Lackenby J. and Christie D. 2005. Effect of confining pressure on the degradation of ballast under cyclic loading. Géotechnique 55(4), 325-328. aratna rail track incorporating the effects of ballast breakage, confining pressure and geosynthetic reinforcement. Proceedings of 8th International Conference on the Bearing Capacity of Railways, and Airfields, London: Taylor and Francis Group, 5-24. aratna B., Nimbalkar S. and Tennakoon N. 2010b. The behaviour of ballasted track foundations: track drainage and geosynthetic reinforcement. GeoFlorida 2010, ASCE Annual GI Conference,February 20-24, 2010, West Palm Beach, Florida, USA,

Indraratna B., Nimbalkar S., Christie D., Rujikiatkamjorn C. and Vinod J.S. 2010a. Field assessment of the performance of a ballasted rail track with and without geosynthetics. Journal of Geotechnical and Geoenvironmental Engineering ASCE 136(7), 907-917.

Indraratna B., Nimbalkar S., Rujikiatkamjorn C. and Christie D. 2011b. State-of-the-art design aspects of ballasted rail tracks incorporating particle breakage, role of confining pressure and geosynthetic reinforcement. Proceedings of 9th World Congress on Railway Research WCRR 2011, Lille, France, 1-13.

Indraratna B., Salim W. and Rujikiatkamjorn, C. 2011a Advanced Rail Geotechnology – Ballasted Track CRC Press/Balkema.

raratna B., Shahin M.A. and Salim W. 2007. Stabilising granular media and formation soil using geosynthetics with special reto Railway engineering, Ground Improvement, 11(1), 27-44. aratna, B. and Nimbalkar, S. 2011. Implications of Ballast Breakage on Ballasted Railway Track based on Numerical Modelling, Proc.13th IComputer Methods and Advances in Computational Mechanics,IACMAG 2011, Melbourne Australia, May 09-11, 2011, 1085-1092.

Indraratna, B., Khabbaz, H., Salim, W. and Christie, D. 2006. Geotechnical properties of ballast and the role of geosynthetics in rail track stabilization. Grou

Indraratna, B., Nimbalkar, S. and Rujikiatkamjorn, C. 2011c. Stabilisation of Ballast and Subgrade with Geosynthetic Grids and Drains for Rail Infrastructure, International Conference on Advances in Geotechnica(Eds.), November, 7-8, 2011, Perth, Australia, 99-112. aratna, B., Nimbalkar, S. and Rujikiatkamjorn, C. 2012a. TStabilisation with Geosynthetics and Geodrains, and Performance Verification through Field Monitoring and Numerical Modelling, International

Indraratna, B., Nimbalkar, S. and Rujikiatkamjorn, C. 2012b. Performance Evaluation of Shock Mats and SyImprovement of Rail Ballast. Proc. Second International Conference on Transportation Geotechnics, IS-Hokkai12 September 2012, Sapporo, Japan, 47-62. aratna, B., Nimbalkar, S. and Rujikiatkamjorn, C. 2012c. Future of Australian rail tracks capturing higher

1287

Effect Evaluation of Freeze-Thaw on Deformation-Strength Properties of Granular Base Course Material in Pavement

Évaluation des effets de gel-dégel sur les propriétés de résistance à la déformation des matériaux granulaires de couche de base des chaussées

Ishikawa T., Zhang Y. Hokkaido University, Sapporo, Japan Kawabata S., Kameyama S.Hokkaido Institute of Technology, Sapporo, Japan

Tokoro T. Tomakomai National College of Technology, Tomakomai, Japan Ono T.Hokkai-Gakuen University, Sapporo, Japan

ABSTRACT: This paper examines the effects of freeze-thaw and water content on the deformation-strength properties of subbase course materials to evaluate the mechanical behavior of granular base in cold regions. CBR tests of freeze-thawed subbase coursematerials under various water contents, and the resilient modulus tests in unsaturated condition were conducted using two newly developed test apparatuses. Moreover, these results were compared with long-term field measurement at a model pavement structure,including FWD tests. As the results, it was revealed that the deformation-strength characteristics of unbound granular base course materials deteriorate due to freeze-thaw and increment of the water content in thawing season. This indicates that the freeze-thaw ofgranular base has a strong influence on the fatigue life of pavement structures.

RÉSUMÉ: Cet article examine les effets de gel-dégel et de la teneur en eau sur les propriétés de résistance à la déformation desmatériaux de couche de fondation de chaussée pour évaluer le comportement mécanique des bases granulaires dans les régionsfroides. Des essais CBR de gel-dégel des matériaux de couche de fondation de chaussée avec diverses teneurs en eau, et des essais demodule résilient, dans des conditions non saturées, ont été réalisés à l'aide de deux appareils récemment mis au point. De plus, cesrésultats ont été comparés, avec les mesures de terrain à long terme, à un modèle de structure de chaussée, y compris des essais FWD.Les résultats ont mis en évidence que les caractéristiques de résistance à la déformation, des matériaux granulaires de couche de basenon liés, se détérioraient en raison du gel-dégel et de l'accroissement de la teneur en eau durant la saison de dégel. Ceci indique que legel-dégel des bases granulaires a une forte influence sur la durée de vie en fatigue des structures des chaussées.

KEYWORDS: unbound granular base course matrials, freeze-thaw action, unsaturated soil, CBR test, triaxial test

1 INTRORDUCTION

In snowy cold regions such as Hokkaido, the 0 °C isotherm may penetrate deep into pavement, thereby causing frost heave and swelling of pavement surface, or cracking in asphalt-mixture layer. Such phenomena specific to cold regions are thought to accelerate deterioration of pavement structures and losing of the functions. Recently, a theoretical design method that can predict the long-term performance of transportation infrastructures has come to be used as a structural design method of asphalt pavement in cold regions. The theoretical design method can take the above-mentioned degradation of pavement structures into consideration. However, the frost-heave phenomenon and the temporary degradation in the bearing capacity during the thawing season have not been sufficiently elucidated as well as the modelling of these phenomena. To develop an optimal design method against fatigue failure of asphalt pavement in Japan, it is necessary to understand the mechanical behaviour of subgrade and base course during freeze-thaw in detail.

This paper examines the effects of freeze-thaw action and water content on the deformation-strength characteristics of subbase course materials to evaluate the change in mechanical behaviour of granular base caused by freeze-thaw and concurrent seasonal fluctuations in water content, and the influences on fatigue life of pavement structures. For that reason, we developed a freeze-thawing CBR test apparatus and a medium-size triaxial apparatus for unsaturated soils. CBR tests of freeze-thawed subbase course materials under various water

contents, and the suction-controlled resilient modulus (MR) tests in unsaturated condition were carried out. Moreover, this paper compares results of the above-mentioned laboratory element tests with those of long-term field measurement at a model pavement structure, including FWD tests.

2 TEST APPARATUS

2.1 Freeze-thawing CBR test apparatus

A schematic diagram of a freeze-thawing CBR (California Bearing Ratio) test apparatus is shown in Figure 1. This test apparatus is based on a general CBR test apparatus that has been improved to reproduce the freeze-thaw history expected to be applied to subbase course materials at the in-situ pavement structures, in a laboratory environment. It has following features: – The apparatus, which allows free water absorption or

drainage (open-system freezing) or suppresses it (closed-system freezing) during the freeze-thaw process, can perform a frost-heave test compliant with “Test Method for Frost Heave Prediction of Soils (JGS 0172-2003)” on a CBR test specimen (=150mm, H=125mm).

– Since the temperatures of both ends of the specimen are controlled independently, the apparatus can subject a CBR test specimen to a desired one-dimensional freeze-thaw history at a constant freezing rate (moving speed of frost line).

1288


– Since the apparatus can conduct a CBR test immediately after the freeze-thaw process without moving the sample, the effects of the freeze-thaw action on the bearing-capacity characteristics of unbound granular base course materials can be examined under clear boundary condition, as well as the initial conditions.

2.2 Medium-size triaxial apparatus for unsaturated soils

A schematic diagram of a medium-size triaxial apparatus for unsaturated soils is shown in Figure 2. One key feature of the apparatus is the structural design of the cap and pedestal as shown in Figure 2. Here, the versapor membrane filter is a kind of microporous membrane filters made from hydrophilic acrylic copolymer, and polyflon filter is a hydrophobic filter made from polytetrafluoroethylene. The other key features are as follows: – Since the apparatus can use a medium-size cylindrical

specimen (�=150mm, H=300mm), a triaxial compression test can be performed in accordance with the “Standard Method of Test for Determining the Resilient Modulus of Soils and Aggregate Materials (AASHTO Designation: T307-99)” (AASHTO, 2003).

– The apparatus can apply the matric suction from both ends of the specimen (Figure 2). Besides, pore water is allowed to drain from both cap and pedestal. Accordingly, the apparatus can reduce the testing time by shortening the length of drainage path to half of the specimen height, in addition to the effect of versapor membrane filter.

– The apparatus can apply axial load to a specimen with high accuracy by both strain control method and stress control method with only one hybrid actuator. Moreover, the apparatus can perform both monotonic loading tests with very slow loading rate, and cyclic loading tests in which the maximum frequency of cyclic loading is up to about 10 Hz.

3 METHODOLOGY

3.1 Method of freeze-thaw CBR test

CBR tests on the specimens exposed to different patterns of freeze-thaw history under three different water contents were conducted by using the newly developed freeze-thawing CBR test apparatus. As a test sample of the CBR test, a natural crusher-run (C-40, Figure 3) made from angular, crush, hard andesite, which is employed at the subbase course of pavement structures in Japan, was used. The specimen was prepared by compacting the air-dried samples (water content, w=1.8%) with a vibrator at a degree of compaction (Dc) of 95% (“air-dried condition”). Then, air-dried specimens were saturated with permeating water for 1 hour (“saturation condition”), and after the saturation process saturated specimens were allowed to

drain by gravity for 3 hours (“wet condition”). Therefore, there were three types of specimens defined by the difference in initial water content.

A freeze-thaw CBR test of C-40 was conducted as follows. Freeze-thaw of the specimen was performed according to JGS 0172-2003, though this research adopted closed-system freezing so that the initial water content of the specimen could be maintained. The freeze-thaw process was repeated, and the number of freeze-thaw process cycles (Nf) was given in three patterns of Nf=0 (no freezing), 1, and 2 cycles. After subjection to the freeze-thaw history, CBR test was carried out as per Japanese Industrial Standards “Test Method for the California Bearing Ratio (CBR) of Soils in Laboratory (JIS A 1211: 2009)”.

3.2 Resilient modulus test

Cyclic loading triaxial compression tests on C-40 were performed under three different water contents in conformance with the AASHTO Designation: T307-99 by using the newly developed medium-size triaxial apparatus for unsaturated soils as follows. In the air-dried condition (w=1.2%), an air-dried specimen after compaction (Dc=95%) was isotropically consolidated under an effective confining pressure (σc') of 49.0

150

Surcharge



Insulation

Acrylic cell

Base cooling plate

Temperature sensor (pt100)


O-ring



Top cooling plate


Coolant circulating line

Coolant circulating line

Figure 1. Freeze-thawing CBR test apparatus.

Hybrid Actuator

Triaxial Cell

Pedestal

Cap Gap Sensors

LVDT

Dial Gauge

Specimen150mm x 300mm

LLVDTLoad Cell

(a)

Water Plumbing path

(mm)

Screw

Air supply path(12mm)

Polyflon filterScrew

Versapor membrane filter

Porous metalO-ring

Figure 2. Medium-size triaxial apparatus for unsaturated soils. (a) Schematic diagram of test apparatus. (b) Structural design of cap and pedestal.

0.01 0.1 1 10 1000

10

2030

4050

60

7080

90100

C-40D

50 = 9.1 mm

Uc = 37.1Fc = 5.20PI = NPs=2.74 g/cm3

dmax=2.07 g/cm3

wopt=8.2%Soundness = 4.6%

Perc

enta

ge p

assin

g (%

)

Grain size (mm)

Figure 3. Physical properties of subbase course material.

(b)

1289


kPa. In the saturated condition (w=14.3%), a saturated specimen after compaction (Dc=95%) and permeation was isotropically consolidated under σc' of 49.0 kPa. In the unsaturated condition (w=5.3%), first, after compaction (Dc=95%) and permeation, a capillary-saturated specimen was isotropically consolidated under a net normal stress (σnet) of 49.0 kPa by applying confining pressure (σc) of 249 kPa, pore air pressure (ua) of 200 kPa and pore water pressure (uw) of 200 kPa. Here, σnet is defined as σnet=σcua. Next, an unsaturated specimen under a matric suction (s) of 10 kPa was produced by decreasing uwwhile keeping both σc and ua constant. Here, s is defined as s=uauw. Upon attaining an equilibrium condition in the consolidation process, MR tests were performed under fully drained condition (CD test) as follows. For repeated loading, a haversine-shaped load pulse with a load duration of 0.1 sec followed by a rest period of 0.9 sec was applied. A MR test requires both conditioning process with 1000 loading cycles (Nc) followed by actual testing process with 100 loading cycles under 15 successive paths with varying combinations of confining pressure and deviator stress.

4 RESULTS AND DISCUSSIONS

4.1 Results of freeze-thaw CBR tests

The frost heave rate (Uh), which is used as a frost-susceptibility index, was Uh=0.1mm/h or lower for all test conditions, and thus frost-susceptibility of C-40 is considered to be low

regardless of the freeze-thaw history and the soil water content. Whereas, Figure 4 shows the relationships between CBR and initial volumetric water content (θ) under different Nf. The overall tendency shows a decrease in CBR caused by the increase in water content. Comparing test results of specimens without freezing (Nf=0) to examine differences due only to water content, CBR is found to decrease to nearly 50% when the condition changes from air-dried to saturated, indicating that the water content has an extremely major influence on CBR. On the other hand, a drop in CBR accompanied by an increase in the number of Nf is observed regardless of the water content. In particular, the ratio of decreasing CBR tends to become larger with the decrease in the water content. The volumetric water content at the subbase course in an actual pavement structure is lower than that of the specimen in wet condition (Ishikawa et al. 2012). Thus, it is expected that the influence of the freeze-thaw action on the bearing-capacity of granular base course materials is more pronounced in in-situ condition.

4.2 Results of resilient modulus tests

Figure 5 shows the relationships between the resilient modulus (Mr) and the effective mean principal stress (p’) or the deviator stress (q), respectively, obtained from MR tests on C-40 under different water contents. Here, Mr is defined as qcyclic/εr (qcyclic : amplitude of repeated axial stress, εr : amplitude of resultant recoverable axial strain due to qcyclic). Note that the test data in unsaturated condition was arranged by using σnet instead of σc'in air-dried and saturated conditions. Besides, the regression analysis results of Eq. 1, which is utilized as a resilient modulus constitutive equation in the MEPDG (AASHTO 2008), are also shown in the figure.

2 3

1 1k k

ii octr a

a a

M k pp p

(Yan and Quintus 2002) (1)

Where, k1, k2, k3 are regression constants, σii is bulk stress, pa is normalizing stress, and τoct is octahedral shear stress. For plots with the same σc', Mr decreases with the increase in p’ and q,while for plots with the same p’ and q, Mr increases with the increase in σc'. A dominant effect for the deformation behavior of C-40 is an increase in Mr with increasing confining pressure, regardless of water content. On the other hand, when comparing plots with the same p’ and q under the same σc', the remarkable decreasing tendency of Mr followed by the increase in the water content is recognized irrespective of σc'. The stress-dependency of Mr obtained from this research qualitatively agrees well with the tendency of past researches like the regression analysis by Eq. 1, regardless of the water content.

4.3 Effects of freeze-thaw and water content on Mr

Under different water contents, Figure 6 compares the resilient modulus (Mr(CBR)) estimated by the following empirical formula (Eq. 2) based on the correlation between CBR and Mr, with the resilient modulus (Mr(MR)) derived from the regression analysis results as shown in Figure 5. Note that Mr(MR) are estimated by assuming the stress state, calculated using multi-layered elastic

0 5 10 15 20 25 300

10

20

30

40

50

60C-40

: Nf = 0 cycle : Nf = 1 cycle : Nf = 2 cycles

CBR

(%)

Volumetric water content, Figure 4. Results of freeze-thaw CBR tests.

0 20 40 60 80 10

100

200

300

400

500

0 5 10 15 20 25 300

100

200

300

400 MR test ('1/'

3=4): CBR test:

: Mr(MR)

at 'c=20.7kPa : M

r(CBR)

: Mr(MR)

at 'c=10.0kPa FWD test:

: Mr(MR)

at 'c= 5.0kPa : E

2(FWD)

Res

ilien

t mod

ulus

, Mr (M

Pa)

Volumetric water content, Figure 6. Influence of water content on resilient modulus.

00

MR-1 to MR-6 : Approximation

curve by Eq. 1

'c=20.7kPa: Air-dried: Unsaturated: Saturated

'c=34.5kPa

: Air-dried: Unsaturated: SaturatedR

esili

ent m

odul

us, M

r (MPa

)

Effective mean principal stress, p' (kPa)

(a)

0 40 80 120 1600

100

200

300

400

500

MR-1 to MR-6 : Approximation

curve by Eq. 1Res

ilien

t mod

ulus

, Mr (M

Pa)

Deviator stress, q (kPa)

(b)

'c=20.7kPa: Air-dried: Unsaturated: Saturated

'c=34.5kPa

: Air-dried: Unsaturated: Saturated

Figure 5. Results of resilient modulus tests.

1290


5 CONCLUSIONS

analysis (Ishikawa et al. 2012), at an actual Japanese pavement structure. The elastic moduli (E2(FWD)) for subbase course layer calculated from FWD test results using the static back-analysis program BALM (Matsui et al. 1998) are also plotted against the volumetric water content (θ) measured at the long-term field measurement (Figure 7).

2010/09/012010/12/01

2011/03/012011/06/01

2011/09/010

5

10

15

thawing

regular

freezing

regular

Am

ount

of p

reci

pita

tion

(mm

)

Vol

umet

ric w

ater

con

tent

,

Date (day)

r0=9.13% Subbase courseDc = 94.2%

0

30

60

90

120

150

Figure 7. Results of long-term field measurement

0 5 10 15 20 25 300

50

100

150

200

250

300

opt

CBR test resultswith AASHTO (2008):

: Nf=0 cycle: N

f=1 cycle

: Nf=2 cycle

Res

ilien

t mod

ulus

, Mr(C

BR) (M

Pa)

Volumetric water content,

r

Figure 8. Resilient modulus estimated from CBR test results.

The following findings can be mainly obtained: – Two new test apparatuses have high applicability in the

evaluation of the deformation-strength characteristics of granular base course materials exposed to repeated freeze-thaw and concurrent seasonal fluctuations in water content.

– A dominant effect for mechanical behavior of base course materials in cold regions is a decrease in the CBR and the resilient modulus with increasing water content and freeze-thaw action, and with decreasing confining pressure.

– Empirical formulas adopted in AASHTO standards have sufficient applicability in evaluation of resilient modulus of subbase course layer in Japanese pavement structures.

– Decreasing tendencies of resilient modulus against water content derived from MR tests, CBR tests, and FWD tests qualitatively and quantitatively agree well with each other.

These indicate that when developing a theoretical model for predicting the mechanical behavior of pavement structures in cold regions, it is important to give a special consideration to the degradation in the bearing capacity and resilient modulus caused by cyclic freeze-thaw actions even in non-frost susceptible granular base, in addition to the effects of an increase in water content during the thawing season. However, further examination of the validity, limitation of application and so forth needs to be conducted in the future in order for the outcomes of this research to be practically applicable.

6 ACKNOWLEDGEMENTS

This research was supported in part by Grant-in-Aid for Scientific Research (B) (20360206 & 23360201) from Japan Society for the Promotion of Science (JSPS) KAKENHI.

7 REFERENCES

Mr=17.6·CBR0.64 (AASHTO 2008) (2) The decreasing tendencies of all types of Mr with increasing water content are in fair agreement with each other, irrespective of the calculation method. Though Mr(MR) noticeably depends on σc' in case of the same water content, Mr(MR) estimated at σc' of 10 kPa closest to the in-situ overburden pressure is almost equal to the upper limit of E2(FWD). Besides, Mr(CBR) approximately coincides with Mr(MR) when the principal stress ratio (σ'1/σ'3) is 4 under the σ'c of 10.0 kPa, irrespective of θ. Accordingly, it seems reasonable to conclude that the suction-controlled MR test results in this research quantitatively match those in previous laboratory element tests and field measurement, and that Eq. 1 adopted in the AASHTO standard has high applicability in the evaluation of the resilient modulus of subbase course layer in Japanese pavement structures.

AASTHO. 2003. Standard Method of Test for Determining the Resilient Modulus of Soils and Aggregate Materials. AASTHO Designation T 307-99. Standard Specifications for Transportation Materials and Methods of Sampling and Testing, T307-1-T-307-41.

AASHTO. 2008. Mechanistic-Empirical Pavement Design Guide: A Manual of Practice. Washington.

Ishikawa, T., Kawabata, S., Kameyama, S., Abe, R., and Ono, T. 2012. Effects of freeze-thawing on mechanical behavior of granular base in cold regions. In Miura, Ishikawa, Yoshida, Hisari & Abe (eds), Advances in Transportation Geotechnics II, Proc. intern. conf., Sapporo, 10-12 September 2012, 118-124.

Ishikawa, T., Zhang, Y., Segawa, H., and Miura, S. 2012. Development of medium-size triaxial apparatus for unsaturated granular base course materials. In Miura, Ishikawa, Yoshida, Hisari & Abe (eds), Advances in Transportation Geotechnics II, Proc. intern. conf., Sapporo, 10-12 September 2012, 185-191.

Figure 8 shows the relationships between Mr(CBR) and initial volumetric water content (θ) under different Nf. Note that the range from residual volumetric water content (θr) to θoptcorrespond to optimum water content obtained from a water retentivity test and compaction tests on C-40 (Ishikawa et al. 2012) is indicated in the figure. The overall tendency is identical to that observed in Figure 4. When being focused on the range, a decrease in Mr(CBR) due to the increase in the number of freeze-thaw process cycles is severe as compared with a decrease in Mr(CBR) due to the increase in water content. Besides, according to results of long-term field measurement (Figure 7), it is expected that the resilient modulus of subbase layer at the actual pavement structure deteriorates along the path shown by the arrows in Figure 8 when it is exposed to repeated freeze-thaw and the concurrent seasonal fluctuations in water content. Therefore, freeze-thaw action seriously influences the resilient deformation characteristics of granular base course materials and hence it also affects the fatigue life of pavement structures in cold regions.

Japanese Geotechnical Society. 2003. Test method for frost susceptibility of soils (JGS 0172-2003). Standards of Japanese Geotechnical Society for Laboratory Soil Testing Methods, 45-50.

Japanese Standards Association. 2009. Test Method for the California Bearing Ratio (CBR) of Soils in Laboratory (JIS A 1211: 2009). Japanese Industrial Standards. (in Japanese).

Matsui, K., Kurobayashi, I. and Nishiyama, T. 1998. Effort for improving accuracy of pavement layer moduli estimated from FWD test data. Journal of Pavement Engineering 3, 39-47. (in Japanese).

Yan, A., and Quintus, H. L. V. 2002. Study of LTPP Laboratory resilient modulus test data and response characteristics: Final Report. Publication No. FHWA-RD-02-051, U.S. Dept. of Transportation, Federal Highway Administration, McLean, VA, 1-161.

1291


Comportement à long terme d’un remblai routier préchargé

Islam M.N., Gnanendran C.T.School of Engineering and Information Technology, UNSW Canberra, Australia

Sivakumar S.T.Queensland Department of Transport and Main Roads, AustraliaKarim M.R.School of Engineering and Information Technology, UNSW Canberra, Australia

ABSTRACT: The results from an investigation into the long-term performance of the preloaded Nerang-Broadbeach Roadway (NBR)embankment near the Gold Coast in Queensland, Australia, are presented in this paper. The soil profile along this roadway consists ofdeep Cainozoic estuarine alluvial soft clay deposit. To predict the performance of the preloaded embankment, two fully couplednonlinear Finite Element Analyses (FEA) were conducted adopting an elasto-viscoplastic (EVP) and an elasto-plastic Modified CamClay (MCC) model to represent the soft clay using the UNSW Canberra modified version of the nonlinear stress analysis programAFENA. It was found that the MCC model under-predicted the ultimate settlement while the creep-based EVP model captured it wellbut over-predicted the pore pressure response. Observational approaches using the Asaoka and Hyperbolic methods were also appliedfrom which it was observed that, when the soft soil exhibited creep, after a certain cutoff time increment (∆t), the Asaoka plot becameparallel to the 45 line and the settlement prediction was unrealistic compared with the field measurement. After a modification wasintroduced into the Asaoka method for creep-susceptible soil, the predicted settlement was found to be in good agreement with thatobtained from Hyperbolic method and the details are presented in the paper.

RÉSUMÉ : Les résultats d'une enquête sur le comportement à long terme de la route Nerang-Broadbeach (NBR), en remblai ayant faitl’objet d’un préchargement près de la Gold Coast dans le Queensland (Australie) sont présentés dans ce papier. Le profil de sol le longde cette chaussée se compose de dépôts profonds d’argile molle marine de l’horizon alluvial « Cainozoic ». Afin de prédire laperformance du remblai préchargé, deux analyses couplées aux élément finis non linéaires (FEA) ont été réalisées en adoptant unmodèle élasto-viscoplastique (EVP) et un modèle élasto-plastique modifié du type Cam Clay (MCC) ; ces modèles ont été choisispour représenter l’argile molle en utilisant la version modifiée de l’analyse en contrainte non linéaire du programme AFENA. Il a étéconstaté que le modèle MCC a sous estimé les tassements ultimes, tandis que le modèle EVP basé sur le fluage a bien évalué cestassements mais a surestimé la réponse en terme de pression interstitielle. Les approches observationnelles utilisant la méthoded’Asaoka ou la méthode hyperbolique ont été aussi utilisées ; il a été constaté que lorsque le sol exhibe du fluage, après un certainseuil dans l’incrément de temps (∆t), la courbe d’Asaoka devient parallèle à la droite à 45° et la prédiction de tassement devient irréaliste par rapport aux mesures sur le terrain. Après qu’une modification ait été introduite dans cette méthode d’Asaoka pour les sols sujets au fluage, le tassement prédit a été trouvé en bon accord avec celui tiré de la méthode hyperbolique.

KEYWORDS: Preload, soft clay, Modified Cam Clay model, Elasto-viscoplastic model, creep, Asaoka method, Hyperbolic method.

1 INTRODUCTION

The Nerang-Broadbeach Roadway (NBR) was constructed bythe Queensland Department of Transport and Main Roads(QDTMR) and completed in 2001. It is located closer to theGold Coast Highway in the South part of Surfers Paradise, GoldCoast, Queensland, Australia. It was constructed toaccommodate the region’s transport network and enhance roadsafety. The roadway embankment was founded on deepCainozoic estuarine alluvial, soft sensitive deposits ofthicknesses from 5 to 21 m overlaying greywackes and argillitebedrock. This estuarine deposit is highly compressible, exhibitslow bearing capacity and undergoes extensive time-dependentsettlement when subjected to extrinsic loads. Although there areseveral techniques for accelerating the ongoing settlement ofestuarine clay and to mitigate post-construction damage,preloading in conjunction with surcharging has been proven tobe one of the most efficient ground improvement techniques forestuarine clay in the Queensland region (Islam et al. 2012,2013). The NBR was divided into five distinct preloadingembankment sections: North of Main Drain; Main Drain toMeadow Drive; Meadow Drive to Witt Ave Drain; South of WittAvenue Drain; and Gin House Creek (Fig. 1). Performance ofthe embankment section located in between Gin House Creekand Witt Avenue Drain and nearer to settlement plate SP18 thathad a preloading height of 3 m is examined in this paper.

Field monitoring data (measured settlement and excess porewater pressure) of this embankment was compared with thecorresponding predicted responses obtained from Finite

Element Analysis (FEA). In particular, nonlinear fully coupledFEAs were carried out adopting a creep-based elastic-viscoplastic (EVP) model and Modified Cam Clay (MCC)elasto-plastic model for the foundation soil from which it wasfound that the creep-based EVP model captured the fieldsettlement of the embankment better than the MCC model butover-estimated the excess pore water pressure. The ultimatesettlement was estimated using Asaoka’s and Hyperbolicobservational methods in this study. Since the foundation softsoil exhibited creep, after a certain cutoff time increment (∆t),Asaoka plot became parallel to the 45 line and the predictedsettlement was unrealistic compared to those obtained fromFEA using the creep-based EVP model as well as Hyperbolicmethod. Therefore, some modification was necessary for theAsaoka method for capturing the ultimate settlement of creep-susceptible foundation soil and it is the focus of this paper.

2 SUBSURFACE CONDITIONS OF SITE

To delineate the subsurface conditions of the NBR, two subsoilinvestigations were carried out by the QDTMR, in 1991 (MainRoads 1991) and 1999 (Main Roads 1999), from which a poorsubsoil strata was identified. This led to further investigations ofQDTMR in 2000 (Main Roads 2000) and 2001 (Main Roads2001) which included six borehole tests, twenty electric conepenetrometer tests (CPT) and four piezocone dissipation tests(CPT-u). The reasons behind the boreholes were to obtainundisturbed soil samples for laboratory testing and to conductfurther in-situ field testing. The CPT and CPT-u tests were


1292


conducted to profile the soil layer and to determine the potentialsand lenses as well as to correlate the parameters. Fieldresponses were monitored using settlement plates andpiezometers. In this paper, the data from a settlement plate(SP18 – See Fig. 1a) and a piezometer are used for assessing thepredictability of the performance. The site plan, locations offield tests and instrumentation, along with the depths of the claydeposits, are shown in Fig. 1.

The subsoil consisted of moisture contents from 24 to101%, liquid limits of 35 to 68% and plasticity index of 15 to33%. The saturated unit weight of the estuarine soft clay variedfrom 14.43 to 20 kN/m3. The sensitivity of the soil rangedbetween 3.75 to 7.00 and the undrained shear strengths of theclay deposits obtained from field vane shear tests from 30 to 92kN/m3. The soil properties of the NBR site are shown in Fig. 2.

Figure 1. (a) Site plan and (b) Soil profile along NBR

Figure 2. Foundation soil properties of NBR

The soil profile along the NBR area comprised alluviumoverlying bedrock in two distinct strata: upper alluvium and

lower alluvium. The upper alluvium consisted of a 2m thicktopsoil or silty clay overlying 2 m of loose sand. Depending onthe physical properties and compressibility characteristics; thelower alluvium was divided into three distinct layers: Clay-1 (7m); Clay-2 (5 m); and Clay-3 (5 m). It was observed that, in theClay-1 layers, the organic component was 8.4 %.

3 FINITE ELEMENT ANALYSIS

Fully coupled, elasto-plastic (Roscoe and Burland, 1968) andelasto-viscoplastic (Karim et al., 2010) nonlinear FEA of theNBR embankment were carried out considering plane strainanalyses using a UNSW Canberra, modified version of the FEAprogram AFENA (Carter and Ballam 1995). Due to thesymmetry of the embankment section to reduce computationaltime, only half part of the embankment was considered foranalysis.

The soft soil was initially modelled as an elasto-plasticMCC material and the results were compared with thosesubsequently obtained adopting the creep-based EVP model.The sand layer in the foundation soil, argillite bed rock andembankment fill materials were modelled as elastic perfectlyplastic materials using the Mohr-Coulomb failure criterion.

Consolidation parameters (λ and κ) were calculated from a 1-D consolidation test data and the strength parameter (φ or M) estimated from the correlation of the CPT and CPT-u tests. Theflow parameter (co-efficient of permeability) was back-calculated from the CPT-u test data using the relationshipproposed in Teh and Houlsby 1991 and Karim et al. 2010. Forthe CPT and CPT-u test data interpretations CPeT-IT 2012 wereused. The void ratio ( Ne ) of the in situ soil at the unit mean-normal effective stress on the normal consolidation line, thepreconsolidation pressure ( 0cp ) and conventional secondaryconsolidation co-efficient (Cα) were calculated from a 1-Dconsolidation of the test data. The model parameters used inMCC and EVP models are tabulated in Tab. 1.

Figure 3. FE geometry used for 2D plane strain analysis (X-X section)

The length and width of the embankment section were 1.3km and 40 m respectively, with the height of fill materials anddepth of its foundation 3 m and 21 m respectively. Theconstruction period of the embankment was 15 days. Thesettlement plate was placed at RL = + 1.5 m on the centre lineof the embankment to monitor the ongoing field settlement andthe piezometer at RL = - 4.6 m to monitor the field’s excesspore water pressure. The side slope of the embankment is 1V:2H.

The finite element mesh consisted of 11,267 nodes and5,520 elements with six noded nonlinear triangular elementsused for finite element discretisation. It was observed from theFEAs that the predicted settlements from the MCC and EVPmodels were 425.55 mm and 498.00 mm respectively for 360days. On the other hand, the measured settlement for the sametime duration was 478.00 mm. It is evident that the MCC modelunder-predicted the settlement which may have been due toongoing creep settlement.

RL=+4.5 mRL=+1.5 mRL=-0.5 mRL=-2.5 m

RL=-9.5 m

RL=-14.5 m

RL=-19.5 m

RL=-32.5 m2.5 m

20.0 m7.5 m5.0 m7.5 m17.5 m

(a)

(b)

Piezometer

SP18

1293


Table 1. Material parameters used in analyses

Notes: Poisson’s ratio considered 0.3 From 1D consolidation tests* At top of soil layer ‡ Gradient of 0cp after -14.5 m 5.5 kPa/m of depth

The excess pore water pressure was monitored for 217 days andobserved to be better predicted by the MCC than EVP model.Up to 73 days, the MCC model captured the measured excesspore water pressure well but then started to over-predict it.

Figure 4. Comparison of measured and predicted settlements

Figure 5. Comparison of measured and predicted excess pore waterpressures

4 OBSERVATIONAL APPROACH

Observational approaches, such as the Asaoka (1978) andHyperbolic (Tan 1995) methods, allow predictions of theultimate settlement of estuarine clay. In Asaoka (1978) method,settlement ( t ) at any time (t) can be expressed as a linear plotdefined by Eqn. 1 and the ultimate settlement by Eqn 2.

0 1 1t t (1)

0

11ult

(2)

where 0 1, are the co-efficients representing the intercept andslope of the fitted straight line proposed by Asaoka (1978)respectively, and the intercept point of the fitted line and 45lines stands for the ultimate settlement. Applicability of Asaokamethod for predicting the creep-included settlement of softclays has been questioned previously (Islam et al. 2012,Lansivaara 2003). Moreover, effectiveness of the Asaokamethod is biased by the selection of the time interval ( t ). Forthese reasons, in the present study, the prediction of the ultimatesettlement obtained from the Asaoka method was comparedwith the ultimate settlement prediction from ‘Hyperbolic’ method.

For Asaoka Plot, the settlement data obtained for thesettlement plate (SP18) was extracted for a particular constanttime interval value t (e.g. 7 days) and the maximummonitored settlement over the field monitoring period wasconsidered as the peak settlement value. By trial and error,consideration of the settlement-time data range after 60%consolidation was found to be appropriate for predicting theultimate settlement of the NBR embankment using Asaokamethod. Similar approaches have been reported by Tan (1996)which was supported by Bergado et al. (1991).

Different values of t ( = 7, 14 and 21 days) were attemptedfor predicting ult. It was observed from the application ofAsaoka method for this field case that, with increases in thetime interval ( t ), the predicted ultimate settlement decreasedbut, after a certain cutoff time interval ( t ), their magnitudesbecame identical which is in agreement with the findings ofArulrajah (2005). The regression value for the correspondingAsaoka plot was found to be about 0.99.

For the NBR embankment, the ultimate settlementpredicted using the Asaoka and Hyperbolic methods werealmost identical (517.00 mm and 517.25 mm respectively). Inboth cases, data beyond 60% of the consolidation (Tan 1996)were considered, as supported by Bergado et al. (1991). It istherefore concluded that when the soft soil exhibits significantcreep, the ultimate settlement prediction by the Asaoka methodonly provided good agreement with the Hyperbolic method aftera certain cutoff time interval ( t ) and the data range after 60 %consolidation state. Therefore, the ultimate settlement predictionby the Asaoka method for creep-susceptible soft estuarine clayrequires scrutiny.

Vertical permeability coefficients

RL (m) M Ne 0K 0cp *(kPa)

iK (m/day) 0e kCFill Materials --- --- --- --- 0.42 E 3000 kPa, =35, c = 5.0 kPa+1.5 to -0.5 1.51 0.43 0.043 4.10 0.40 50.00 2.50×10-5 1.70 1.00-0.5 to -2.5 --- --- --- --- 0.50 E 5000 kPa, = 32, c = 4.0 kPa-2.5 to -9.5 1.33 0.39 0.062 3.85 0.46 80.00 2.5×10-5 1.70 1.00

-9.5 to -14.5 1.20 0.23 0.030 2.70 0.50 112.00 2.5×10-5 1.70 1.00-14.5 to -19.5 1.07 0.13 0.013 2.51 0.55 114.00‡ 2.5×10-5 1.70 1.00-19.5 to -32.5 --- --- --- --- 0.42 E 15000 kPa, = 35, c = 50.0 kPa

0 100 200 300 400

Time (Days)

-600

-500

-400

-300

-200

-100

0

Settl

emen

t(m

m)

MeasuredMCCEVP

0 50 100 150 200 250Days

0

20

40

60

80

Exce

ssPo

reW

ater

Pres

sure

(kPa

)

PP1MCCEVP

1294


In the Hyperbolic method, the relationship betweensettlement ( t ) at any time (t) is given by Eqn. 3 and theultimate settlement by Eqn. 4.

t

t mt

(3)

This is a linear straight line in a plot oft

t

against t.1

ult m (4)

where , , m are the co-efficients representing thetheoretical slope factor, intercept and slope of the straight linerespectively

Figure 6. Observational approaches for predicting settlement: (a)Asaoka; and (b) Hyperbolic

In the NBR embankment field study, the ultimate settlementpredicted using the Asaoka and Hyperbolic methods werealmost identical (517.00 mm and 517.25 mm respectively). Onthe other hand, it was observed that the measured fieldsettlement at 360 days was about 480 mm, whereas thepredicted settlements obtained from the MCC and EVP analysesat 360 days were 425.55 mm and 498.00 mm respectively.

5 CONCLUDING REMARKS

The long-term performance of the instrumented preloaded NBRembankment founded on a soft sensitive estuarine clay wasnumerically modelled using the MCC and creep-based EVPmodels. It was observed from the field monitoring data througha settlement plate and piezometer that creep-based settlementwas ongoing and, after 360 days, ultimate settlement was notattained. Although the MCC model ignored creep, whichresulted in it under-predicting ultimate settlement. Measuredsettlement was well captured by the creep-based EVP model.On the other hand, the MCC model captured the excess porewater pressure prediction better than the creep-based EVPmodel, particularly up to 73 days. As there may have been rainafter 73 days, which would have raised the ground water level,

further study is being undertaken to ascertain the excess porewater pressure. To predict ultimate settlement through anobservational approach, two methods were used: the Asaokaand Hyperbolic, with their predictions compared with thosefrom FEA-based MCC and EVP analyses. It was observed thatthe modified calculation of the Asaoka method predicted almostidentical magnitudes of ultimate settlement as the Hyperbolicmethod and FEAs.

6 ACKNOWLEDGEMENTS

The first author was supported by the Tuition Fee Scholarship(TFS) while conducting his Doctoral research at UNSWCanberra. The support provided by the National ComputationalInfrastructure (NCI) facility, Australia and QDTMR, Australiaare gratefully acknowledged.

7 REFERENCES

Arulrajah A. 2005. Field measurements and back-analysis of Marineclay geotechnical characteristics under reclamation fills. DoctoralThesis, Department of Civil Engineering, Curtin University ofTechnology, Australia.

Asaoka A. 1978. Observational procedure of settlement predictions.Soils and Foundations 18(4), 87-101.

Bergado D., Asakami H., Alfaro, M.C. and Balasubramaniam A.S.1991. Smear effects of vertical drains on soft Bangkok clay.Journal of Geotechnical Engineering, ASCE, 117(10), 1509-1530.

Carter J.P. and Ballam N.P.1995. AFENA User’s Manual. Version5.0[computer program]. Center for Geotechnical Research, Universityof Sydney, Sydney-2006, Australia.

Geologismiki Geotechnical Software, CPeT-IT v 1.7,CPT interpretationsoftware, (2012), http://www.geologismiki.gr/Products/CPeT-IT.html

Islam, M. N., Gnanendran C. T., Sivakumar S. T. 2012. Effectiveness ofPreloading on the Time Dependent Settlement Behaviour of anEmbankment. GeoCongress 2012, ASCE, 2253-2262.

Islam, M. N., Gnanendran C. T., Sivakumar S. T. 2013. Time dependentsettlement behaviour of embankment on soft sensitive clay. 18thSouth Asian Geotechnical & Inaugural AGSSEA Conference,Singapore (Submitted and under review).

Karim M. R.,Gnanendran C. T.,Lo S. C. R., Mak J. 2010. Predictingthe long-term performance of a wide embankment on soft soil usingan elastic–viscoplastic model. Canadian Geotechnical Journal47(2), 244-257.

Länsivaara T. 2003. Observational approach for settlement predictions.Deformation Characteristics of Geomaterials / Comportement DesSols Et Des Roches Tendres, Taylor & Francis, 1277-1285.

Main Roads 1991. Gooding’s Corner Geotechnical Investigation. MainRoads of Queensland, Materials and Geotechnical Services,Report: R1748, Australia.

Main Roads 1999. Additional Geotechnical Investigation, Nerang-Broadbeach Road, Gooding’s Corner. Main Roads of Queensland,Transport Technology, Geotechnical and Geological Services,Report: R3161, Australia.

Main Roads 2000. Preload Monitoring: Nerang- Broadbeach road,Goodings corner deviation and Neilsens road intersection. MainRoads of Queensland, Transport Technology, Geotechnical andGeological Services, Report: MR1822, Australia.

Main Roads 2001. Additional Geotechnical Investigation for theProposed Western RSS Wall Area, Nerang-Broadbeach Deviation,Gooding’s Corner. Main Roads of Queensland, Report: R3233,Australia.

Roscoe K. H. and J. B. Burland 1968. On the generalized stress-strainbehavior of wet clay. Cambridge University Press, Cambridge, UK,535-609.

Tan S.A. 1995. Validation of hyperbolic method for settlements inclays with vertical drains. Soils and Foundations 35(1), 101-113.

Tan S.A. and Chew S.H. 1996. Comparison of the Hyperbolic andAsaoka observational method of monitoring consolidation withvertical drains. Soils and Foundations 36(3), 31-42.

Teh C.I. and Houlsby G.T., “An Analytical Study of the Cone Penetration Test in Clay”, Geotechnique, 41(1), (1991), pp. 17-34.

(a)

(b)

t ult R2

Days mm7 519.00 0.99114 517.50 0.99621 517.00 0.992

2

0.90261 1000

1.7450.994517.25ult

mR

mm

1295

Probabilistic Settlement Analysis For The Botlek Lifting Bridge Design

Analyse probabiliste de tassement pour la conception du pont levant Botlek

Jacobse J.A., Nehal R.S. GEO2 Engineering B.V.

Rijneveld B. Fugro GeoServices B.V.

Bouwmeester D. Ballast Nedam Engineering B.V.

ABSTRACT: A new lifting bridge is being constructed crossing the river Oude Maas in the Rotterdam harbour area in theNetherlands. For the deformation analysis deterministic 3D FEM calculations were performed. In order to take the effect of soil heterogeneity on the deformation behaviour of the bridge piers into account, a probabilistic model has been developed. This modeland the applications are described in this paper. The application of a simplified stochastic subsoil model enables a quantitative risk analysis in order to deal with this uncertainty. Furthermore the model is used to determine design values of the deformations ofseveral components of the bridge.

RÉSUMÉ : Au port de Rotterdam aux Pays-Bas on construit un nouveau pont levant qui traverse la rivière Oude Maas. Des calculs déterministes 3D FEM sont effectués pour analyser la déformation. On a développé un modèle probabiliste pour tirer l'effet del'hétérogénéité du sol sur la déformation des piles du pont. Cet article décrit ce modèle et ses applications. L'application d'un sous-sol simplifié et stochastique permet une analyse de risque quantitative qui sait régler l'incertitude des paramètres du sous-sol. En outre le modèle est utilisé pour déterminer la valeur de calcul des déformations des différentes pièces du pont.

KEYWORDS: Foundation design, shallow foundation, soil heterogeneity, probabilistic deformation analysis, quantitative risk analysis

MOTS-CLES: Calcul de fondations, foundation superficielle, hétérogénéité, analyse probabiliste, analyse quantitative, analyse de risques

1 INTRODUCTION

The Dutch highway A15 in the Rotterdam harbour area is being widened due to an increase in traffic load. One of the main challenges in this project is the construction of a new lifting bridge over the river Oude Maas. Consisting of two lifting spans of approximately 100 m and pylons reaching over 60 m above water level, this new bridge will be one of the largest lifting bridges in Europe (see Figure 1).

Figure 1. Typical cross section of new Botlek Lifting Bridge

The three main bridge piers (from left to right in Figure 1:

Pier 30, Pier 40 and Pier 50) are founded on rigid concrete blocks with footing dimensions of 15 x 60 m, at 8 m below river bed at the top of the first dense (Pleistocene) sand layer. For the geotechnical design the foundation was essentially treated as a shallow foundation.

At a depth of approximately 16 m below the foundation footing a relatively soft clay layer is present with varying thicknesses between 0 and 4 m. This stratum complicated the design, especially with respect to the settlement behaviour which has a major impact on the performance of the total bridge and influences the different design disciplines (e.g. mechanical,

electrical and structural). This article considers the risk analysis with respect to the deformation behaviour of the subsoil which was undertaken as part of the foundation design.

2 SOIL INVESTIGATION AND PARAMETER DETERMINATION

For the determination of the soil parameters an extensive soil investigation has been performed. A relatively dense grid of Cone Penetration Tests (CPT’s) with a mutual distance of about 15 m was executed to a depth of about 3 times the foundation width. In addition, a number of boreholes were drilled and undisturbed samples were taken at regular intervals for geotechnical laboratory tests by means of light percussion drilling in combination with thin-walled samplers. From the CPT’s and borehole logs the soil stratigraphy is determined, see Table 1. Table 1. General soil stratigraphy

Top of layer [m NAP]

Soil description Soil layer

-7 à -14 SAND, clayey cover layer -14 à -20 SAND, (medium) dense 1st sand layer -33 à -39 CLAY, stiff deep clay

layer -34 à -42 SAND, (medium) dense 2nd sand layer -60 Max. investigation depth

The thickness of the deep clay layer varies strongly. At some

locations the thickness is about 4 m, whereas this layer was not encountered at other locations.

Classification tests, such as particle size distribution

(granular layers) and volumetric weight and water content

1296


(cohesive layers), were performed on samples from the different soil layers. In order to determine the strength properties of the second sand layer isotropically consolidated drained triaxial tests were performed. The samples were prepared in the laboratory at relative densities of 40%, 60% and 80%. The in situ relative density was determined from the CPT’s from the correlation deduced by Baldi (Lunne, 1997), and turned out to be approximately 70% for this sand stratum. The characteristic strength properties were determined from statistical analyses on the results from the triaxial tests. For the effective angle of shearing resistance of the sand below foundation level a representative value of 33° was determined at higher axial strain levels, which corresponded well with the cone resistance as can be found in literature, e.g. in the Dutch Code (NEN-EN 1997-1, 2005).

Since deformations of the deep clay layer were expected to

have a relatively large influence on the superstructure, additional oedometer tests were performed on samples from the deep clay layer. From experience in the area, it is known that this layer is overconsolidated, which was confirmed by the CPT results. However, the overconsolidation ratio (OCR) could not be accurately determined from the oedometer tests, most likely due to relaxation of the samples. Therefore the OCR is determined from the following correlation with the cone resistance (Lunne, 1997):

v

vc

ncu

ocu qcc

OCR'3.0

17/

;

;

(1)

In which: OCR = overconsolidation ratio [-] cu;oc = in situ (overconsolidated) undrained shear

strength [kPa] cu;nc = normally consolidated shear strength [kPa] qc = cone resistance [kPa] σv = vertical total stress [kPa] σ’v = vertical effective stress [kPa]

The calculated OCR corresponded well with experience

from other projects in the area and geological information. The virgin stiffness and unloading/reloading stiffness was

determined from the oedometer tests, which included an unloading/reloading step. The determination of these stiffness parameters from the laboratory tests was expected to be reliable, since these were determined beyond the preconsolidation stress, so relaxation effects are expected to be minimal.

Based on the soil investigation and laboratory tests,

representative values for the soil stiffness’s were determined. A representative elasticity modulus (Eoed;ref) of approximately 40 MPa and 3.5 MPa was determined for respectively the 1st sand and deep clay layer. This is the oedometer stiffness at a reference vertical effective stress of 100 kPa. For the stress-stiffness relationship a power law was adopted (Brinkgreve, 2011), with a power 1.0 for sand and 0.8 for the stiff clay (based on oedemeter tests). An unloading/reloading oedometer stiffness ratio of 4 is applied.

3 DETERMINISTIC DEFORMATION ANALYSIS

During the design process it was recognised that deformations of the foundation have a large influence on the design and construction of the superstructure, especially for the mechanical and structural design. Due to the large ratio between the height of the pylons and the width of the foundation, a small rotation of the foundation base results in a large deflection of the pylon heads. This effect has a significant influence on the design of the superstructure. In order to determine safe

tolerances which have to be taken into account by the other design disciplines, a thorough deformation analysis was performed.

First step in the deformation analysis was to perform ‘best estimate’ deformation calculations. In the early design stages analytical 2D settlement calculations were performed. In the detailed design phase additional 3D FEM calculations were performed. The software program Plaxis was used for these calculations.

In the calculations the Hardening Soil (HS) model is used for

the deep clay layer. Aspects of this model include: - Stress dependent stiffness of the soil - Plastic straining due to primary deviatoric loading - Plastic straining due to primary compression - Elastic unloading/reloading - Failure according to the MC criterion The Hardening Soil model does not take creep effects into

account. However, from the laboratory tests it turned out that about 80% of the settlements are primary and only 20% of the settlements are related to secondary compression. Therefore the choice was made to consider the creep effect separately, instead of applying a soft soil creep model. Soft soil creep models are especially useful if the influence of creep is more pronounced.

For the sand layers underneath the foundation surface the

Hardening Soil Small-Strain Stiffness (HSSmall) model is used. This model is similar to the HS model, but additionally takes the higher stiffness of the soil at small strain levels into account. For the clayey sand layer above the foundation layer, also the Hardening Soil (HS) model is used.

The serviceability limit state (SLS) foundation pressures for

the different main piers are in the range between 500 to 700 kPa. The calculated ‘best estimate’ final settlements of the foundation footings range from 0.10 to 0.25 m. For the rotations maximum values in the order of 1/1000 were calculated. These calculated rotations are mainly the result of the bending moments loads, rather than soil heterogeneity.

In general, soil deformations are difficult to predict

accurately, since various uncertainties can be present. See for instance Figure 2 where the thickness of the deep (Kedichem) clay layer is plotted over the footprint of the main bridge piers. The variation is based on factual data from CPT’s and boreholes with interpolation between these data.

Figure 2. Thickness Kedichem clay layer [m] To get a better understanding with regard to subsoil

uncertainties, a sensitivity analysis with the 3D FEM model was performed. The influence of variations in OCR, (virgin) stiffness and the thickness of the deep Kedichem clay layer between the soil investigation points were considered.

1297


From the sensitivity analysis it was concluded that the variation in calculated average settlement was about 25% for the different subsoil scenarios. The variation in the calculated rotations was small. The calculated absolute rotations of the piers were still in the range of 1/1000.

4 PROBABILISTIC DEFORMATION ANALYSIS

The sensitivity analysis with the 3D FEM results in a better understanding in the range of settlements which could be expected. However, the variation of soil properties within one homogeneous soil layer is hardly taken into account in standard 3D FEM calculations. Since this effect can have a large influence on the rotations of the foundation, a practical stochastic subsoil model was set up to take this effect into account.

Figure 3. Description of probabilistic model

With this probabilistic model it was possible to determine

the probability of exceedance of a certain design rotation. With this model it was also possible to perform a quantitative risk analysis, regarding the effects of foundation rotations. The model is described section 3.1 to 3.6 and schematically presented in Figure 3.

4.1 Step 1 - Soil-structure interaction

The foundation is modelled as an infinitely stiff foundation block, supported by linear elastic (stochastic) springs at a spacing of about 3 m. Since the foundation consists of a massive concrete block with a thickness of about 20 m the assumption of a stiff foundation is considered reasonable.

The linear elastic soil springs are stochastic, representing the uncertainty in soil behaviour. The stochastic (correlated) stiffness ki;j of the springs Si;j under the foundation is determined according to:

JW

IL

zqk

jiji

,. (2)

In which: q = uniform distributed foundation load, Pz /(L·W)

[kN/m2] zi;j = settlement at location (xi;j;yi;j) [m]

L = length of the foundation [m]

W = width of the foundation [m] I = number of equally spaced springs along the length

of the foundation [-] J = number of equally spaced springs along the width of

the foundation [-] Pz= Vertical foundation load [kN]

All parameters in eq. (2) are deterministic, except for the

settlements. A linear transformation between the probability density function (PDF) of the settlements (see section 4.3) and the soil stiffness is applied.

4.2 Step 2 - Settlements

For the determination of the expected value of the (residual) settlements the results of the 3D FEM model are used. Since the soil stiffness is primarily a soil property, the influence of the stiff foundation should not be taken into account in the determination of the settlements. Therefore the stiffness of the foundation block is neglected for these settlement calculations, by using a flexible footing in the 3D FEM model. The settlements are calculated with a uniform load on the foundation surface.

4.3 Step 3 – Parameters probability density function

4.3.1 Model parameters The settlements are modelled as random variables with a

lognormal distribution. The lognormal distribution is often used to model non-negative random variables, such as thickness of layers and soil properties. The calculation results from 3D FEM model are interpreted as the expected value z of the PDF of the settlements.

The parameters of the lognormal distribution of the settlements has the following parameters (i.e. Fenton and Griffiths, 2008):

)1ln( 2)ln( zz V (3)

2

)ln()ln( 21)ln( zzz (4)

The coefficient of variation of the settlements is based on the assumption of 30% inaccuracy in the settlement calculations. That means that there is a probability of about 5% that the settlements will be 30% larger than the calculated average settlements. This is a generally applied rule of thumb in the Netherlands. So:

zz 3.1%95 (5)

Based on the lognormal distribution the z95% can estimated

by:

)65.1(%95

)ln()ln( zzez (6) Equating eq. (5) and (6) in combination with eq. (3) and (4)

the coefficient of variation (Vz) of the PDF of the settlements can be estimated by:

3.1)1ln(65.1)1ln(21 22

zz VV ee (7) Eq. (7) results in a coefficient of variation Vz of

approximately 0.17.

4.3.2 Correlation parameters Due to its natural fabric, the soil properties, can be

considered as spatially correlated. Different autocorrelation

Step 4: Create n correlated realisations of a set of spring values

Step 5: Determine for every set of spring values the rotation of the foundation block

Step 6: Estimate the probability of exceedance of a certain rotation

Monte Carlo simulation

Step 1: Create a model for the soil-structure interaction. In this case: infinitely stiff foundation block, supported by linear elastic (stochastic) springs

Soil-structure interactionmodel

Step 2: Determine expected value of the settlements under the foundation surface with an

advanced 3D FEM model

Step 3: Determine the parameters for the PDF of the settlements and corresponding spring values for the soil-structure model

Geotechnical input

1298


functions are known from literature, from which the functions that adopt an exponential shape are commonly used. In this case the following expression is used (Breysse, 2004 and DeGroot, 1993):

cLd

zz e /)2ln();1ln(

(8) In which:

d = horizontal distance between two springs [m] Lc = autocorrelation length of ln(z) [m] The covariance can be determined according to (CUR190,

1997):

)))ln(),(ln( 2ln()1ln()2ln();1ln(21 zzzzzzCov (9) From this the covariance matrix C can be constructed.

4.3.3 Autocorrelation length The autocorrelation length Lc can be interpreted as the

distance over which a certain parameter is significantly correlated. In literature several indicative values for the horizontal and vertical correlation length for soil parameters are given. In this case especially the horizontal correlation length is relevant. Typical values for the horizontal autocorrelation length for soil properties are in the range Lc ≈ 20 to 100 m (DeGroot, 1993; TAW, 2001 and Gruijters, 2009).

Before determining the autocorrelation length the influence

of this parameter is checked. If Lc → 0, the logarithm of the settlement at two locations is independent. Because of the averaging effect of the stiff foundation, the rotation is expected to approach the value as found in a deterministic approach, e.g. zero rotation if a homogeneous soil is modelled. If Lc → ∞, the logarithm of the settlement at two locations is fully correlated. In this case the rotation is also expected to approach the value as found in an deterministic approach, e.g. zero rotation if a homogeneous soil is modelled. The maximum rotation is found for an intermediate value of Lc, typically half the foundation size.

For the deformation analysis especially the spatial variation

of the compressibility of the different soil layers and the thickness of the clay layer are important. From the soil investigation it turned out that the horizontal correlation length with respect to the thickness of the clay layer is typically in the order of 10 to 20 m. However, in general the horizontal correlation length with respect to soil properties is typically in the order of 50 to 100 m. Therefore the most critical value for the horizontal correlation length within the range between 10 to 100 m was selected. In this case a correlation length of 20 m has been used.

4.4 Step 4 - Realisations of spring values

For the probabilistic analysis a Monte Carlo (MC) procedure is used (CUR190, 1997 and Haugh, 2004). To generate correlated values for the spring values an algorithm in a spreadsheet program was set up. The following procedure is applied for each realisation 1 to n: 1) Generate a vector with realization of the standard normal

distribution X N(0,I). In which I is the identity matrix and the size of the vector is equal to the number of springs s = I.J

2) Decompose the covariance matrix Cln(z) = A.AT (Cholesky decomposition (Haugh, 2004))

3) Determine the correlated vector Z’ = A.X + μln(z) N(μln(z),Cln(z))

4) Determine the vector with correlated settlement values Zi;j = exp(Zi;j’)

5) Determine the vector with the correlated spring values from eq. (1)

4.5 Step 5 – Determining settlement and rotation foundation

Since an infinitely stiff foundation is assumed in step 1, the rotation of the foundation can be determined from the vertical force equilibrium and the moment equilibrium in 2 directions:

I

i

J

j jiz qLWRF1 1 . (10)

ayI

i

J

j jijiy MqLWxRM _2

1 1 .. 21 (11)

axI

i

J

j jijix MWqLyRM _1 12

.. 21 (12)

In which:

Ri;j = force in spring Si;j [kN] xi;j = x coordinate of spring Si;j [m] yi;j = y coordinate of spring Si;j [m] My_a = acting bending moment around the y axis [kNm] Mx_a = acting bending moment around the x axis [kNm]

The force in every spring can be determined according to:

jijiji ukR .;; (13)

In which: ui;j = deformation in spring Si;j [m]

The deformation in every spring can be expressed as:

jiyjixji yxuu ;;0;0; (14)

In which: u0;0 = deformation in the point x = 0, y = 0 [m] θx = rotation around the y axis, long axis [-] θy = rotation around the x axis, short axis [-] These are exactly the variables of interest, which can be

filled in into the equilibrium equations. This leads to a system of linear equations, which can be presented in matrix notation:

854

973762321 0;0

AAAu

AAAAAAAAA

y

x

(15)

Wherein the parameters A1 to A9 can be derived from eq.

(10), (11) , (12), (13) and (14). The matrix equation can be solved by Cramer’s rule (Lay,

2003), which states that: DDU u /0;00;0 (16)

DD xx / (17) DD yy / (18)

In which:

D = determinant of the coefficient matrix Du0;0 = determinant of the matrix formed by replacing the

u0;0 column of the coefficient matrix by the answer matrix

Dθx = determinant of the matrix formed by replacing the θx column of the coefficient matrix by the answer matrix

Dθy = determinant of the matrix formed by replacing the θy column of the coefficient matrix by the answer matrix

1299


4.6 Step 6 – Determination probability of exceedance

For every simulation a set of spring values was generated. With the soil-structure interaction model the bending moment and vertical force equilibrium the rotation and (average) settlement was calculated for every set of springs. The probability of exceedance for a certain rotation can be estimated by:

nnP fr /)θ( (19)

In which: P(θ>θr) = exceedance probability of rotation θr [-] nf = number of simulations for which the calculated

rotation is larger than the reference rotation [-] n = total number of simulations [-]

The accuracy of this estimate strongly depends on the

number of simulations in relation to the probability of exceedance; for smaller probabilities, a higher number of simulations is necessary to reach the same reliability of the estimation. The relative error is given by (CUR190, 1997):

))θ(/)θ(/( rrf PPnn (20)

For a certain value of the relative error E with an accuracy of

95% can be estimated by: )/)1/(4( nnnE f (21)

For this study a relative error E of maximum 20% is

assumed to be acceptable. In order to be able to determine probabilities of exceedance of 1·10-4 sufficiently accurate, therefore at least 1·106 simulations are necessary.

5 RESULTS AND APPLICATION

5.1 Results

Figure 4 shows the results of realisations for the residual rotations of pier 40. In this figure the combined realisations of rotation around the long axis (θx) and the rotations around the short axis (θy) are shown.

Figure 4. Results Monte Carlo analysis pier 40, residual rotations

Figure 4 shows that the distribution of realisation is located around the origin what means that the expected rotation is more or less equal to zero. This is in line with the deterministic settlement calculations. It is also shown that rotation around the long axis has a higher probability than rotation around the short axis. The shortest side (width) of the foundation block is more sensitive for rotation.

The calculated probabilities of exceedance for different rotations are presented in Figure 5 for the rotation around the long axis (θx). The results for pier 30 and 50 are almost equal

because the calculated deformations with the FEM model are also almost equal for these piers. From Figure 5 related to the average settlement of the piers it can be concluded that larger average settlements (pier 40) result in a higher probability of larger rotations, which is reasonable.

Important for the bridge deck is the combined rotation of two

piers. Based on the results of the individual piers also the probability of a combined rotation of two piers could be determined. For the combined rotation it is assumed that the deformation behaviour of the piers is uncorrelated.

Figure 5. Results probabilistic deformation analysis (θx). Note that the results for P30 and P50 are almost identical

5.2 Application of results

The results of the model are used for the design of the different components of the bridge which are influenced by the settlement and rotation of the pier.

Based on the calculated probability of exceedance of a certain rotation, safe boundary conditions for the other design disciplines could be determined. Relevant components are the towers with the guiding system, the deck, the expansion joints for the deck and the supports of the deck. A design value of the deformation is derived for these components based on the acceptable probability of exceedance.

During construction the deformations will be monitored and

control measures can be applied if necessary.

6 CONCLUSION

For the design of the new Botlek Lifting Bridge soil deformations can potentially have a major affect on one of the most critical design requirements, which is a limited rotation of the large foundation footing.

Alongside a well designed site investigation campaign,

laboratory tests and the application of appropriate constitutive models, a quantification of the probability of exceedance of soil deformations was desired. Application of a simplified stochastic subsoil model enabled a quantitative risk analysis in order to deal with the uncertainties described in this paper.

Based on the calculated probability of exceedance of a

certain rotation, safe boundary conditions for the other design disciplines could be determined.

1300


7 REFERENCES

Breysse, D., Niandou, H. Elachachi, S. & Houy, L. 2004. A generic approach to soil-structure interaction considering the effects of soil heterogeneity, Geotechnique 54, No. 2, p. 143-150.

Brinkgreve R.B.J., Engin E. and Swolfs W.M. 2011. Plaxis 3D 2011, Manual, Plaxis BV, Delft.

CUR 190 1997. Kansen in de civiele techniek – deel 1: Probabilistisch ontwerp in theorie, CUR/Ministerie van Verkeer en Waterstaat.

DeGroot, D.J. and Baecher, G.B. 1993. Estimating autocovariance of in-situ soil properties, ASCE J. Geotech. Eng., 119(1), p. 147-166.

Fenton G.A. and Griffiths D.V. 2002. Probabilistic Foundation Settlement on a Spatially Random Soil, ASCE J. Geotech. & Geoenv. Engrg., 128 (5), p. 381-390.

Fenton G.A. and Griffiths D.V. 2008. Risk Assessment in Geotechnical Engineering, Hoboken, New Jersey.

Gruijters S.H.L.L. 2009. Blijvend Vlakke Wegen, kenmerk 0910-0235, Delft Cluster, Delft

Haugh M. 2004. The Monte Carlo Framework, Examples from Finance and generating Correlated Random variables, Course Notes IEOR E4703: Monte Carlo Simulation, Columbia University.

Lay D.C. 2003. Linear algebra and its applications, third edition, University of Maryland – College Park, Addison Wesley.

Lunne T., Robertson P.K. and Powell J.J.M., 1997. Cone Penetration Testing in Geotechnical Practice, Blackpool Typesetting Services Limited, UK

NEN-EN 1997-1:2005 2005. Eurocode 7: Geotechnisch ontwerp Deel 1: Algemene regels, Nederlands Normalisatie-instituut, Delft.

TAW 2001. Technisch Rapport Waterkerende Grondconstructies, Technische Adviescommissie voor de Waterkeringen, Den Haag.

1301

Ground improvement methods for the construction of the federal road B 176 on a new elevated dump in the brown coal region of MIBRAG

Méthodes d'amélioration de sols pour la construction de la route nationale B 176 traversant un remblai récent d’une mine de lignite de MIBRAG

Kirstein J. F. BVT DYNIV GmbH; Germany

Ahner C. Landesamt für Straßenbau und Verkehr; Germany

Uhlemann S. MIBRAG; Germany

Uhlich P., Röder K. CDM Smith Consult GmbH; Germany

ABSTRACT: The MIBRAG Company operates two surface mines in the region south of Leipzig in Germany. This is the reason whythe existing B176 road needed to be relocated up to 1 km on a length of more than 5 km. This move will place the new road on almost 60 m recently placed landfill area. Different ground improvement techniques such as Controlled Modulus Columns (CMC),Dynamic compaction and Dynamic Replacement were used for the foundation of the bridge and the road depending on the soilconditions and settlement tolerance of the structure. Because of significant stability problems, 15 m deep “floating” stone columns areinstalled in the landfill. The design and the settlements were significantly optimized by the combination of the different soilimprovement techniques. The settlement forecast based on DIN EN 4094-5 Ménard pressuremeter results and finite elements calculations are validated with the results of inclinometer measurements under the largest 15-m-high and 70-m-wide embankment.

RÉSUMÉ: La société minière MIBRAG exploite deux bassins de lignite dans le sud de Leipzig en Allemagne. Pour continuerl’extraction de lignite, la route nationale B 176 doit être déplacée sur une longueur de 5 km à une distance de 1 km sur un remblairécent d’environ 60 m d’épaisseur. Le renforcement du sol de fondation de la route et des ouvrages d’art a nécessité la mise en œuvredes techniques d’amélioration de sol CMC, consolidation dynamique et colonnes ballastées. Grâce à la combinaison de ces méthodes,des solutions d’exécution sûres et économiques ont été proposées et réalisées aussi bien pour la fondation principalement flottante dela route et des ouvrages d’art que pour la zone de transition au terrain naturel. Les déformations sous un remblai de 15 m de haut et 70 m de large ont été estimées grâce au pressiomètre Ménard et aux calculs aux éléments finis ; elles sont validées par les mesuresinclinométriques.

KEYWORDS: Embankment, brown coal mining, Controlled Modulus Columns (CMC), stone columns, Ménard pressuremeter. MOTS-CLÉS: remblai, mine de lignite, Colonne Module Contrôlé (CMC), colonnes ballastées, Pressiomètre Ménard 1 INTRODUCTION

The area south of Leipzig is characterized by brown coal mining, and with an area of 500 square kilometer it is one of the largest landscape construction sites in Europe. The existing federal road B 176 between Pödelwitz and Neukieritzsch will be relocated by the MIBRAG for brown coal mining. The first construction section has an overall length of 8.3 km. The federal road B 176 will be rebuilt by MIBRAG 5.5 km on the young elevated dump of the Vereinigtes Schleenhain mine and will be handed over to the State Agency for Road and Transport. The young mixed dump fill areas are deep, from at least 60 m and partially up to 105 m depth down to the natural soil. Because of the large thickness of this fill deposit, these areas are usually founded with shallow techniques to remain cost-effective. The connection to the existing B176 takes place on the so-called mainland area. The future road will vary in elevations from existing ground level to about 15m above the current elevation resulting in varying settlements profiles across the length of the road. Without the use of ground improvement methods, differential settlements of up to one meter across the road would have to be expected. For areas with more than 3 m of additional fill, ground improvement was performed using mainly the intensive dynamic compaction (Dyniv), while for the areas with embankment of up to 15 m in height and 70 m in width, stone columns were used. In the transition zones between Dyniv and

stone columns, Dyniv columns were executed as dynamic replacement. At the bridge abutment, Controlled Modulus Columns (CMC), full-displacement-columns were used. in which the soil will be replaced by non-reinforced concrete.

Figure 1. Development of the new B176 (G.U.B. Ingenieur AG 2010).

With the help of various ground improvement methods, the embankment stability verifications were performed by using the results of Ménard pressuremeter with enhanced factor of safety. By adjusting the ground improvement methods to the various sections and ground conditions, a technical and economical optimization could be achieved. The improvement of the soil characteristics as well as the latest results of settlement measurements confirmed the success of the method.

Méthodes d’amélioration de sols pour la construction de la route nationale B 176 traversant un remblai récent d’une mine de lignite de MIBRAG

1302


2 SITE INVESTIGATION, GEOLOGICAL FEATURES

In central Germany, there are mainly landfill sites with mixed soils with varying silt contents. Due to the technique used for filling, dumped from a great height, the fill is deposited in poorly compacted heterogeneous horizontal thin layers (varved). These mixed soil man-made deposits are much more compressible than natural soils. Particularly in the upper soil layers down to a depth of 10 m, the soils are often arranged in a very loose state. A reliable assessment of the interaction between the structure and the subsoil is very difficult on mixed dump soil.

Figure 2. The dump of Schleenhain

As a result of loose deposition or rather the heterogeneity of the material and its density, the soil behavior is different from natural deposits. Experiences and methods used with the development of soft natural soil are not transferable to these types of man-made deposits (Lausitzer und Mitteldeutscher Bergbau – Verwaltungsgesellschaft 1999). The process of conveying, transporting and dumping is the reason that the deposit is locally marked by extreme material and density heterogeneity in a very confined space. Due to the mixture of cohesive fines and loose grain, these deposits consistently characterized by a very strong sensitivity and the risk of loss of strength due to plastification by water ingress.

The soil investigation results from the fill show a large variation in grain size.

Figure 3. Spread of grading curves of the dump soil (G.U.B. Ingenieur AG 2010)

The following fractions have been used for the design of the ground improvement for the new road, according to the geotechnical report (G.U.B. Ingenieur AG 2010):

- 20% boulder clay - 20% sands and gravels - 25% tertiary clay - 30% fine and medium sand - 5% brown coal

This mixed-grain fill material represents almost the entire typical soil in the mining area. Details on the natural soil with boulder-clay and silt layers of the quaternary will not be presented here.

Currently, the groundwater level has been lowered to at least 35 – 40 m below ground-level by the ongoing dewatering operations of the daylight mining, whereupon the relatively steeply running saturation lines of the depression cone have all developed in tilted, grown border slopes. By the year 2100, a static groundwater level should be reached after turning off the pumps at an altitude of 125 m of about 15 m significantly below the gradient and at least slightly below the lowest level of the foundation of new embankments.

With wn = 10%...14%, the calculated current water content under the influence of dewatering is in the normal range of natural humidity. Laboratory tests showed a proctor density about ρPr = 1.846 g/cm3, and a proctor water content about wPr = 9.9% and thus good compression options.

Cone penetration tests from the landfill at up to ca. 20 m depth vary mainly in the range of qc ~ 1.0 MN/m2...3.5 MN/m2, which lead to the estimation of a stiffness modules Es ~ 2 MN/m²...9 MN/m², with average = 5.5 MN/m², as described in the geotechnical report.

It is a mixed-grain fill material (fine particle fraction > 20%...25%) in the earth-moist state. With water saturation a plastification is possible, because of the fine particle fraction and the storage of loose grains and mixed pseudo-cohesive soil. In the event of plastification, the shear strength drops to about 50% of the baseline values.

3 PLANING OF THE SOIL IMPROVEMENT METHODS

The connection to the new B176 is located on the so-called mainland area. This is characterized by the transition from natural soil towards the central part of the road to the landfill area. From this point of view it is a very challenging geotechnical transition in the route. Other settlement issues are arising from the different embankment height as a result of the area conditions. Without the use of specially adapted ground improvement methods, settlements of up to one meter could be expected. In order to reduce settlements and to avoid critical differential settlements, a special sequence and quality control of soil improvement techniques was chosen at the transition with the main land and the three structures.

Figure 4. Design section along the road centerline in the transition from the main land (natural ground) to the landfill.

The following figure 5 shows the floating foundation of two building structures on the 50 to 70 m thick fill deposit.

Figure 5. Design section of the tunnel structures crossing the up to 15 m high embankment over the landfill

1303


The tunnel structure 2 above the belt conveyor is founded by the vibration-free CMC method. For the construction of the 15 m tall embankment 70 m wide at its base, the stone columns technology were used.

Figure 6 gives an overview of the different improvement techniques carried out in the area of the belt conveyors structure 2. The longitudinal section elevations showed a height difference of respectively 1.75 m. The shallow stone columns reach 10 m to 15 m deep.

Figure 6. 15m high difference including ripping holes for the two devices RSV, CMC directly on BW2 and pressumeter PMT in front.

4 RESULTS OF THE SOIL IMPROVEMENT

Hereafter, examples of results for the stone columns treatment area are shown. Test areas and boring before and after treatment were performed near the highest embankment section to be able to derive the soil parameters of the calculation model.

The Menard pressuremeter tests were performed within the stone columns and in the center of the grid of installation.

Figure 7. Execution of the Ménard pressuremeter between and centrally in the columns

Due to the compaction at optimum proctor water content corresponding to figure 8 most often an improvement between the columns with the factor of 2 was measured. With the mean stiffness modulus of Ec = 100 MN/m2 in the columns with at least 70 cm diameter, this results in a 3 times higher design relevant modulus of Es =30 MN/m2 for the improved ground.

The Stiffness modulus was doubled after treatment in the center of the grid of installation, in between columns. This fairly remarkable result was made possible by the water content close to the optimum Proctor of the deposits of the mining ground, and also by the powerful V23 vibrator. A transfer of these high values to other constructions projects without these optimal conditions is not possible and it is highly recommended to use a project-specific calibration with test fields and the Ménard pressuremeter for other projects.

The results of the cone penetration tests also showed an improvement factor of 2. It should be noted that the initial values may have been too low. The improvement as shown by CPT`s is similar to the Ménard pressuremeter. The stiffness modulus in cohesive soils can only be measured by pressuremeter and oedometer tests on undisturbed soil samples.

Preliminary investigation between columns centrally in columns

Figure 8. Results from in-between and at the center of the columns

5 NUMERICAL ESTIMATION OF SETTLEMENT

Based on the design soil parameters, the settlement calculation was created in the course of a Diploma thesis (Vogel 2011) with the PLAXIS software.

Figure 9. CAD model and PLAXIS model

The layers of the fill deposit from ground level up are divided per figure 9 into different layers. The first 20 meters from the edge of the model to the toe of the embankment are modelled using the thin layered structure of the graded soil charactistics of the dump.

The improvement depth by stone columns below the embankment was 15 m. This improved ground beneath the dam is modelled using a composite layer with a composite average modulus Es = 30 MN/m2.

The last two layers of fill are assumed to be each 20 m thick at the lower part of the model domain.

As part of the thesis (Vogel 2011), different behaviour laws for the modelling of the stress-strain relation of the fill material were compared and each corresponding calculations resulted in very variable estimation of the settlements. Between the linear-elastic, perfectly plastic Mohr Coulomb model with one meter expected settlements to the more realistic elastoplastic Hardening Soil model, a difference of half a meter in the estimated long term settlement were calculated using PLAXIS software.

In this paper, we present the results of the Hardening Soil model, where the stiffness modulus for the settlement calculation for the next load level are sequentially recalculated and increased after each load level according to oedometer and pressuremeter test results.

Each of the following six stages corresponds to an embankment height increase of nearly 3 m with corresponding

1304


settlements. On the one hand, these settlements decrease with each load step due to a smaller width of loading and with it decrease the loads at depth. On the other hand, additional settlements of the dam itself are calculated for each step.

7 SUMMARY AND CONCLUSION

The construction of roads on mining areas requires close cooperation between mining engineers, geotechnical engineers and contractors, as the mining technical characteristics, requirements in road construction and geotechnical characteristics must always be brought into line and finished with high-end quality.

A very comprehensive site investigation is essential in dealing with man-made fill deposits. The Ménard pressuremeter brought significant findings throughout the additional preliminary soil investigation.

A technical and economic optimization was achieved by adapting the ground improvement methods to the respective sections. With the help of various ground improvement methods, settlement reduction and improved stability safety factors were successfully obtained.

The high quality of the construction with stone columns, dynamic replacement and CMC was documented not only in foundation with the usual protocols of the manufacturer, but also tested regularly during the construction phase by the Ménard pressuremeter. The settlement calculations and subsequent field measurements confirmed a significant increase of quality.

Station 1+780 HS 30 Steps of the Deposit

Ground Settlement [cm]

1 10 2 10 3 10 4 10 5 5 6 5

Traffic 6 Total

Ground Settlement

56

8 REFERENCES

Figure 10. Representation of the predicted total settlements of 56 cm with Plaxis calculation section of 15 m embankment height

Lausitzer und Mitteldeutsche Bergbau – Verwaltungsgesellschaft mbH, 1999, Schlussbericht – Bauen auf Mischbodenkippen des Braunkohletagebaus im Mitteldeutschen Revier. Senftenberg

Buja,H.-O.,2009, Handbuch der Baugrunderkennung – Geräte und Verfahren. Wiesbaden

6 RESULTS OF THE MONITORING

Seven horizontal inclinometers were installed below the embankment across its section and three vertical inclinometers were also installed at the landfill along the construction road at the location of the highest embankment. The measurements are performed during the embankment construction phase to reach a total height of 15m according to the following sequence:

G.U.B Ingenieur AG, 2010, Ersatz der Bundesstraße B176 zwischen Pödelwitz und Neukieritzsch, Baugrundgutachte,. Dresden

DIN Deutsches Institut für Normung, 2001, DIN EN 4094-5: Bohrlochaufweitungsversuche.

Kirstein, J.F.; Chaumeny J.L., 2005, Ein neues Verfahren zur Bodenverbesserung: CMC (Controlled Modulus Columns) aus Frankreich. Veröffentlichungen des Instituts für Geotechnik TU Bergakademie Freiberg, Heft 2005-2, Freiberg

- every 3 m of fill placement - immediately after reaching the final height and then

every 3 month. The following figure presents the cross-section measurement

at the location of the design cross section.

Meyer, N.; Emersleben, A.; Kirstein J.F., 2007, Probebelastungen von CMC-Säulengruppen – Einfluss der Lastverteilungsschichtauf die Beanspruchung des Untergrundes und der Säulen. Pfahl – Symposium 2007, Institut für Grundbau und Bodenmechanik, TU Braunschweig

Vogel, S., 2011, Setzungsprognosen und Monitoring beim Neubau einer mit unterschiedlichen Bodenverbesserungstechniken auf einer jungen Kippe gegründeten Bundesstraße. Diplomarbeit, Dresden

ISO/FDIS 22476-4:2009, 2009, Geotechnical investigation and testing – Field testing – Part 4: Ménard pressuremeter

Chaumeny, J.L.; Hecht, T.; Kirstein, J.F.; Krings,M.; Lutz, B., 2008, Dynamische Intensivverdichtung (DYNIV®) für die Kreuzung eines aktiven Erdfallgebietes im Zuge der Bundesautobahn BAB A 71. VERÖFFENTLICHUNGEN des Grundbauinstitutes der Technischen Universität Berlin, Heft 42, Berlin

EIBS GmbH, 2010, Planfeststellungsunterlage zum Ersatz der Bundesstraße B 176 zwischen Pödelwitz und Neukieritzsch, Bauabschnitt 1 (unpublished)

Ahner, C., Kirstein, J., Uhlemann, S., Röder, K., Uhlig, P.: Baugrundverbesserungsverfahren zur Gründung der Bundesstraße B 176 auf einer jungen Hochkippe im Braunkohlenrevier der MIBRAG. Baugrundtagung 2012

Figure 11. measurement results: the settlement curves fit the predictions in the different earthwork steps of the embankment construction

The measured settlement values in all cross sections were similar to the predicted settlement and underline the accuracy of the Ménard pressuremeter and the soil parameters derived from this test which were used for the finite element using the hardening soil model.

Ahner, C., Kirstein, J., Uhlemann, S. Uhlig, P.: Ground improvement methods for establishment of the federal road B 176 on a new elevated dumb in the brown coal area of MIBRAG, ISSMGE - TC 211 International Symposium on Ground Improvement IS-GI Brussels 2012

1305

Model tests on settlement behaviour of ballasts subjected to sand intrusion and tie tamping application

Tests de modélisation sur le comportement en tassement des ballasts sujets à l’intrusion de sable et au bourrage

Kumara J., Hayano K. Yokohama National University

ABSTRACT: Effects of sand intrusion into ballast (i.e., ballast fouling) and tie tamping application on settlement characteristics ofballasted trail track were investigated by series of model tests with cyclic loading. Model tests were conducted on 1/5th scale of the actual size of railway track. Ballast fouling was simulated by sand-gravel mixtures (i.e., gap graded particle size distribution). Tie tamping application was physically simulated in the model tests using a simple tool. The relationship of number of loading cycles and settlement was obtained and results were discussed with degree of ballast fouling (i.e., amount of sand in sand-gravel mixtures). The results indicated that initial settlement process and rate of residual settlement alter after 30% sand. Initial settlement period is higherfor fouled ballast with 30% or more sand after tie tamping application. Rate of residual settlement is higher for fouled ballast withmore than 30% sand after tie tamping application. That’s to say, tie tamping application is effective for fouled ballast up to 30% fines.

RÉSUMÉ : Les effets de l’intrusion de sable dans le ballast et du compactage par bourrage du ballast sur le comportement entassement des voies ferrées ballastées ont été étudiés en effectuant une série d'essais cycliques de chargement sur des échantillonstests. Les tests ont été réalisés sur modèle à une échelle d'un cinquième, et l'exécution du bourrage a été physiquement simulée dansles essais à l’aide d’un outil simple. La relation entre le nombre de cycles de chargement et le tassement a été obtenue et les résultatsen fonction du degré d’intrusion de sable et du compactage ont été étudiés. Les résultats indiquent que le processus de tassementinitial et le taux de tassement résiduel sont modifiés au delà de 30% d’intrusion de sable. Le taux de tassement résiduel est plus élevépour le ballast bourré et compacté comportant plus de 30% de sable, ce qui revient à dire que le compactage pour le ballast bourré esteffectif jusqu’à 30% de fines.

KEYWORDS: Ballast fouling, ballasted railway track, model test, residual settlement, tie tamping application.

1 INTRODUCTION

In railway tracks, ballast fouling occurs when fine materials mix with ballast due to heavy repeated train loads. Generally, fine materials come mainly from underneath layers, and to a lesser extent, due to particle crushing too (Indraratna et al., 2004). Sand intrusion alters the original particle size distribution (PSD) of ballast, resulting different settlement characteristics than that of fresh ballast. Once the settlement reaches the allowable limit, a maintenance method should be implemented to bring the railway track into the original position. Usually, tie tamping application is used worldwide as the main maintenance method. However, effects of ballast fouling on settlement characteristics and tie tamping application itself haven’t been well understood in the past.

In this study, effects of degree of ballast fouling and tie tamping application on settlement characteristics of ballasted tracks were investigated. A series of cyclic loading tests were conducted on a model sleeper of 1/5th scale of the actual rail track as shown in Figure 1. In the model tests, tie tamping application was physically simulated by inserting a small tool into the ballasts.

2 MODEL GROUNDS AND CYCLIC LOADING

Figure 1 shows the model test apparatus used in this research. Model grounds at a scale of 1/5th were constructed in a sand box with interior dimensions of 800mm wide, 304mm deep, and 300mm high. A duralumin footing with a width of 48mm was used to model the sleeper. Gravel approximately 1/5th of the size of actual ballasts were selected to model the ballasts. Medium size sand (M sand) was used as sand. PSDs of gravel and M sand are shown in Figure 2.

The model tests were conducted on fouled ballast (i.e., 5 cases) and fresh ballast as given in Table 1. Ballast thickness was made as 50mm (i.e., 1/5 of 250mm of actual ballast layer) in each case and 100 loading cycles were applied before tie tamping application. Cyclic loadings were applied to the model grounds through the sleeper at a constant displacement rate of 0.05mm/s. The amplitude of the cyclic stress applied was 120kN/m2 (i.e., approximately 70% of maximum stress M sand can withstand). All the specimens were prepared with 80% of relative density, Dr. How void ratios, emax and emin of fresh ballast (i.e., gravel) change with amount of sand mixed can be seen in Figure 3. Tie tamping application was simulated with the tool shown in Figure 3. First, the sleeper was lifted to the initial position after 100 loading cycles were applied. Next, a small spoon was inserted (e.g., about 8-10mm) into the model ground by sides of the sleeper. After the spoon reached the fixed ground depth (i.e., 8-10mm), it was tilted several times to permit the particles to move laterally. This procedure was followed at several locations until the voids between the sleeper and the ground surface were completely filled by the particles. Finally, additional gravels (except in case 6 where M sand was used) were introduced to the ground surface near the sleeper to produce a flat ground surface. After this tie tamping application, 100 loading cycles were applied again. Axial displacement was measured using two displacement transducers, placed at the front and back of the sleeper.

1306


Figure 1. Model test apparatus for model test

0 20 40 60 80 1003.2

2.8

2.4

2.0

1.6

1.2

0.8

0.4

0.0Before Tamping

100% sand

70% sand50% sand30% sand

15% sandFresh ballast

Settl

emen

t, S

(mm

)

No of loading cycles, N

1003.2

2.8

2.4

2.0

1.6

1.2

0.8

0.4

0.0

0 20 40 60 80

After Tamping

70% sand

15% sand

50% sand30% sand

100% sand

Fresh ballast

Settl

emen

t, S

(mm

)


Figure 2. PSDs of gravel and M sand

Table 1. Model test conditions CaseNo % sand Dry density,

(kg/m3)Relative density,

Dr (%) 1 0 1519 802 15 1698 803 30 1829 804 50 1929 805 70 1788 806 100 1684 80

Figure 3. Max and min void ratios vs. % sand (Kumara et al., 2012)

Figure 4. The tool and procedure used for simulating tie tamping application

3 RESULTS

3.1 Effects of PSDs on settlement characteristics

Figures 5 and 6 show the results of settlement, S with no. of loading cycles, N for fresh ballast (i.e., gravel) and fouled ballast (i.e., 15, 30, 50, 70% and 100% sand) cases before and after tie tamping application respectively. It clearly shows that fouled ballast alters the settlement characteristics of fresh ballast significantly, both before and after tie tamping application. As shown in Figures 5 and 6, the smallest settlement was observed in cases of 30 and 50% sand specimens. The smallest settlement observed for 30 and 50% sand cases can be understood from the results of void ratios, emax and emin with % sand as shown in Figure 3 where it shows minimum values of void ratios were observed for the mixtures with 30-50% sand.

1 2 3 4 5 6 7 8 90

20

40

60

80

100

10 12

Perc

ent p

assin

g (%

)

Grain size, D (mm)

Gravel M sand

Figure 5. Relationships between no. of loading cycles and settlement before tie tamping application

0 20 40 60 800.2

0.4

0.6

0.8

1.0

1.2

100

emax

emin

Max

and

min

voi

d ra

tio, e

max

and

e min

Amount of M sand (%)Figure 6. Relationships between no. of loading cycles and settlement after tie tamping application

Figure 7 shows the results of settlement at 100th loading cycle, S100 vs. %sand for all the tests and clearly indicates how PSDs affect settlement both before and after tie tamping application. The relationship is quite similar to emax and emin relationships with % sand (Figure 3). The relationships between no. of loading cycles, N and sleeper settlement, S were obtained using Eq. 1 (Sekine et al., 2005),

100 loading cycles Lift the sleeper by Insertingtie tamper and filling the gapby moving it laterally

100 loading cycles

N=100 N=10Tie tamperTie tamper

100 loading cycles

N=100

Ballast

Sleeper

p=120kPap=120kPa

NecS N 1 (1)

where c and represent the initial settlement process, and represents the process of residual settlement. The relationships for fresh ballast (i.e., gravel) and 30% sand cases are shown in Figure 8.

1307


0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.80

20

40

60

80

100

120

140

2.0

k2

k2

u2

1st cycle

App

lied

pres

sure

, p (k

N/m

2 )

Settlement of sleeper, S (mm)

0 20 40 60 800.0

0.1

0.2

0.3

0.4

0.5

100

Before tamping After tamping

M sand

u 2 (mm

)


30%

1000 20 40 60 803.0

2.5

2.0

1.5

1.0

0.5

0.0

Settl

emen

t afte

r 100

load

ing

cycl

es, S

100 (m

m)

Amount of sand (%)

Before Tamping After Tamping

Figure 7. The settlement at 100 loading cycle vs. % sand

Figure 8. Relationships between S and N for fresh ballast and 30% sand cases (before tie tamping application)

3.2 Effects of roadbed stiffness on settlement characteristics

Figure 9 shows typical relationships between sleeper settlement and applied stress, p (30% sand case is shown here). It can clearly be seen that settlement reduces in the 2nd phase (i.e., after tie tamping application), due to densification of the specimen in the 1st phase (i.e., before tie tamping application with 100 loading cycles). In this research, the loading curves were fitted by bilinear lines, and the slopes of the two lines were estimated as k1 and k2 as shown in Figure 10. Displacement u2 was estimated by dividing the applied stress by k2 as shown in Figure 10. The parameter u2decreases and tends to show a constant value with N as shown in Figure 11. Therefore, these constant values were used in the following discussion.

Figure 9. Applied stress vs. settlement for 30% sand case

Figure 10. Calculation of u2 (Gravel specimen)

0 20 40 60 802.8

2.4

2.0

1.6

1.2

0.8

0.4

0.0

100

S30% = 1.49(1 - e-1.44N) + 0.0047N

S0% = 2.06(1 - e-1.67N) + 0.0043N

Before Tamping

30% sand

Fresh ballast

Settl

emen

t, S

(mm

)

No of loading cycles, NFigure 11. Relationships of u2 and N (30% sand case)

The relationships between u2 with %sand, c, and before and after tie tamping application are shown in Figures 12-15 respectively. Figure 12 shows that 50% sand specimen is the densest specimen (i.e., showing the smallest u2). Figure 13 shows that c increases with u2 (though not very clearly), indicating that loose specimen will result in higher initial settlement, in both before and after tie tamping application.Figure 14 shows that initial settlement period alter by tie tamping application with a wider gap between the highest and smallest compared to those of before tie tamping application.

0 20 40 60 80

0.036

0.040

0.044

0.048

0.052

0.056

100

u 2 (mm

)

Amount of sand (%)


0.0 0.4 0.8 1.2 1.6 2.00

20

40

60

80

100

120

140After tamping Before tamping

App

lied

stres

s, p

(kPa

)

Settlement, S (mm)

Figure 12. Relationships of u2 and% sand

0.036 0.040 0.044 0.048 0.052 0.0560.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

c

u2 (mm)


Figure 13. Relationships of u2 and c

1308


0 20 40 60 80 1000.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

Amount of sand (%)


0 20 40 60 800.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

2.2

100

c

Amount of sand (%)


Figure 14. Relationships of u2 and

0.036 0.040 0.044 0.048 0.052 0.0560.2

0.3

0.4

0.5

0.6

0.7

0.8

u2 (mm)


0.036 0.040 0.044 0.048 0.052 0.0560.002

0.004

0.006

0.008

0.010

0.012

u2 (mm)


Figure 17. Relationships of and % sand

0 20 40 60 800.000

0.002

0.004

0.006

0.008

0.010

0.012

100

Amount of sand (%)


Figure 15. Relationships of u2 and

3.3 Effects of tie tamping application on settlement characteristics Figure 18. Relationships of and % sand

Figures 16-18 show how c, and change with %sand. Figure 16 shows that c, parameter indicating initial settlement amount, reduces until 30-50% sand and then increases, almost same as how emax and emin change with %sand (Figure 3). While parameter indicating period of initial settlement process, reduces with %sand, , parameter indicating rate of residual settlement, increase with %sand. The results of can be interpreted as initial settlement period increases with degree of fouled ballast (i.e., increasing of %sand). The results of can be interpreted as rate of residual settlement increases with degree of fouled ballast (i.e., increasing of %sand).

4 CONCLUSIONS

The effects of sand intrusion into ballast and tie tamping application on settlement characteristics were investigated using series of cyclic loading model tests. The following conclusions were derived from this research: (1) The characteristics of the initial settlement process are altered considerably after tie tamping application; especially if ballast is mixed by more than 30% fine materials.

As shown in Figure 16, c reduces with %sand up to 30-50% and then increases. This tendency is same for both before and after tie tamping application. However, change of with %sand is more after tie tamping application for the specimens with more than 15% sand. That’s, initial settlement period increases significantly with %sand after tie tamping application (Figure 17). The results also showed that rate of residual settlement is higher after tie tamping application for the specimens with more than 30% sand (Figure 18). That’s, tie tamping application seems effective for fouled ballast mixed with up to 30% fines.

(2) Rate of residual settlement increases after tie tamping application if ballast is mixed by more than 30% fine materials. Therefore, tie tamping application seems effective for fouled ballast with less than 30% fines.

5 ACKNOWLEDGEMENTS

Japanese Government is acknowledged for providing financial assistance to the first author to study in Yokohama National University, Japan through a Monbukagakusho scholarship.

6 REFERENCES

Indraratna B., Shahin M., Rujikiatkamjorn C. and Christie D. 2004. Stabilisation of ballasted rail tracks and underlying soft formation soils with geosynthetics grids and drains. GeoShanghai International Conference, Shanghai, China, June 2-4, 2006.

Kumara G.H.A.J.J., Hayano K., Ogiwara K. and Takeuchi M. 2012. Fundamental study on the simple evaluation methods for particle size distribution and maximum/minimum void ratio of sand-gravel mixtures, 2nd International Conference on Transportation Geotechnics, Hokkaido, Japan, 572-577, September 10-12, 2012.

Sekine E., Ishikawa T. and Kouno A. 2005. Effect of ballast thickness on cyclic plastic deformation of ballasted track, RTRI Report, 19 (2), 17-22 (in Japanese).

Figure 16. Relationships of c and % sand

1309

Assessing the Effectiveness of Rolling Dynamic Compaction

Évaluation de l'efficacité du compactage dynamique roulant

Kuo Y.L., Jaksa M.B., Scott B.T., Bradley A.C., Power C.N., Crisp A.C., Jiang J.H. The University of Adelaide, Adelaide, Australia

ABSTRACT: Rolling Dynamic Compaction (RDC) is a soil improvement technique, which involves a heavy (6– to 12–tonne) non-circular module (impact roller) that rotates about a corner as it is towed, causing the module to fall to the ground and compact itdynamically. This paper focuses on the 4-sided module and aims to quantify the effectiveness of RDC by means of a combination offield studies and numerical modeling. The field studies involved embedding earth pressure cells beneath the ground at varying depthsand measuring the in situ stress over a range of module passes. In addition, a variety of in situ tests were performed includingpenetrometer, field density and geophysical testing to measure density improvement, again as a function of the number of modulepasses. The field measurements indicated that the depth of improvement exceeded 2 meters below the ground surface. Numericalmodeling was undertaken using the dynamic finite element analysis software, LS-DYNA; the results align well with those obtainedfrom the field studies. Parametric studies were also undertaken to determine the influence of varying soil parameters on theeffectiveness of RDC.

RÉSUMÉ: Le Compactage Dynamique Roulant (CDR) est une technique d'amélioration du sol, qui implique un lourd module de forme non circulaire (6 à 12tonnes), rouleau à impact, qui tourne autour d'un coin lorsqu’il est tiré, ce qui provoque la chute du module sur le sol et le compacte dynamique. Cet article se concentre sur le module à 4 faces et vise à quantifier l'efficacité du CDRpar le biais d'une combinaison d'études sur le terrain et de modélisation numérique. Les études de terrain ont comporté l’installationde cellules de contraintes dans le sol à différentes profondeurs et à mesurer ainsi la contrainte lors des passages du module. En outre, de nombreux essais in situ ont été réalisés, comprenant des pénétromètres, des essais de densité en place et des tests géophysiques afin de mesurer l’amélioration de la densité en fonction du nombre de passes de modules. Les mesures sur le terrain ont indiqué que laprofondeur de l'amélioration a dépassé les 2 mètres sous la surface du sol. La modélisation numérique a été réalisée en utilisant lelogiciel d’analyse par éléments finis en dynamique, LS-DYNA ; les résultats concordent bien avec ceux obtenus dans les études sur le terrain. Des études paramétriques ont également été entreprises pour déterminer l'influence de divers paramètres du sol sur l'efficacitédu CDR.

KEYWORDS: Rolling dynamic compaction, impact roller, LS-DYNA

1 INTRODUCTION

Rolling dynamic compaction (RDC) is a generic term used to describe the densification of the ground using a heavy non-circular module (of three, four or five sides), that rotates about a corner as it is towed, causing the module to fall to the ground and compact it dynamically. An example of RDC is illustrated in Figure 1. RDC is able to compact the ground more efficiently because of its greater operating speed – 12 km/h compared with 4 km/h of conventional rollers. Due to the combination of kinetic and potential energies, RDC has demonstrated improvement to more than one meter below the ground surface (and greater than three meters in some soils); far deeper than conventional static or vibratory rolling, which is generally limited to depths of less than 0.5 m. As a result, RDC has been used on applications such as land reclamation projects, compaction of sites with non-engineered fill, in the agricultural sector to reduce water loss, and in the mining sector to improve haul roads and construct tailings dams.

Quantifying the effectiveness of RDC via field based trials has been the focus of different researchers over the years, including Avalle and Carter (2005), Avalle (2007), Avalle et al. (2009) and Jaksa et al. (2012). Mentha et al. (2011) conducted a trial that involved three main focus areas: (a) the use of earth pressure cells (EPCs) for direct measurements of stress change to determine the extent of depth of influence and the stress distribution induced by the RDC; (b) undertaking field tests, including dynamic cone penetration tests (DCPs) and field

density measurements and the spectral analysis of surface waves (SASW) geophysical technique to measure and infer changes in density as a function of the number of module passes; and (c) conducting a series of laboratory tests (e.g. particle size distribution, hydrometer test, Atterberg’s limits, standard and modified Proctor tests) on the samples collected from the site to characterize the soil. Field-based research typically involves a team of professional operators and technicians spending days diligently preparing a test pad, placing and burying EPCs at the required depth(s) and spacing, undertaking field tests before and after a number of rolling passes, collecting data from EPCs, and collecting soil samples for further laboratory testing.

Figure 1. An example of RDC – Broons BH-1300 4-sided impact roller.

Results from field-based research are typically site specific; supporting the notion that the effectiveness of RDC is highly dependent on the soil type and site conditions. The influence depth is typically a measure of the depth to which the imposed load from the module quantitatively affects the soil; this can

1310


vary considerably, due to inherent differences between sites and how the magnitude of improvement is both defined and quantified. For example, Avalle and Carter (2005) reported a depth of improvement to approximately 1.4 m in botany sands, whereas Avalle (2007) reported a depth of 7 m in calcareous sands. Additionally, time and cost constraints typically limit the number of field tests that can be undertaken. In the case of Mentha et al. (2011), there were requirements on the minimum depth from the surface that cells could be placed to avoid damage to the EPCs, as well as the minimum spacing between adjacent EPCs to reduce stress shadowing effects. Such arrangements provide physical limitations on the spatial resolution of data that can be collected. As a result, contour plots of vertical and lateral stress produced lack of resolution.

There is currently no employable theoretical model or robust predictive model to accurately predict the depth of influence of RDC, the energy imparted per blow or the effectiveness of RDC on different soil types and site conditions. Moreover, there is also limited published information from case studies to indicate the optimal number of passes needed to attain the targeted soil density for a given site or ground condition. A consequence is that costly and time consuming field trials are inevitably required before using RDC. Due to cost and time constraints this can limit the usage of RDC in some projects.

2 NUMERICAL MODELLING

This research aims to fill the knowledge gap discussed previously by evaluating the effectiveness of RDC using the dynamic finite element modeling (FEM) software LS-DYNA (Hallquist 2006). A 3D numerical model was developed that allowed the rolling dynamics of the 4-sided impact roller to be simulated. The model was then validated against field data collected by Mentha et al. (2011). The adopted final FEM model is illustrated in Figure 2.

Figure 2. FEM model.

The FEM model consisted of two major parts: the 4-sided impact roller itself, which is a simplified model of the Broons BH-1300 roller (Figure 1), and the soil mass. The module is a steel encased concrete block. As it rolls, any deformation caused by the impact on the roller is very small and negligible. Therefore, it is reasonable to assume that the roller acts as a rigid body. The model utilized shell elements on the roller, whilst 8-node quadrilateral solid elements were used on the soil mass. To simulate the confinement and far field conditions, LS-DYNA boundary conditions *BOUNDARY_SPC_BIRTH _DEATH and *BOUNDARY_NON_REFLECTING were prescribed to the sides and base of the soil mass. Two of LS-DYNA’s soil constitutive models were examined, namely the MAT_005_Soil_and_Foam and the MAT_193_Drucker_and_ Prager models. It was found that the MAT_005 underestimated the soil settlement caused by the roller and was therefore excluded from further modeling. During the initialization stage of the modeling, the effects of gravity loading were added using *LOAD_BODY_Z and *LOAD_RIGID_BODY. The contact definitions between the roller and the soil mass is described in LS-DYNA’s *CONTACT_AUTOMATIC_ SURFACE_TO_

SURFACE_ID. Finally, the *BOUNDARY_PRESCRIBED_ MOTION_RIGID option was used to define the rolling motion (both horizontal and rotational speeds) of the roller. A detailed description of the FEM is given by Bradley et al. (2012).

3 FIELD WORK AND LABORATORY TESTING

The field work undertaken by Mentha et al. (2011) took place at the Project Magnet Tailings Storage Facility at the Iron Duke Mine, South Australia. The fill material typically consisted of coarse iron magnetite tailings that are a by-product of a consistent treatment process. The results from sieve analyses and plasticity tests indicated that the soil is a well graded sand (SW) with small quantities of gravel-sized material (14%) and clay fines (6%) of low plasticity (LL = ~22%, PL = ~11%). The average field moisture content was ~5% and the water table was located well below the influence depth of RDC.

The test pad consisted of three lanes; three lift heights of 1200 mm were achieved. The EPCs were strategically placed at various vertical and lateral levels. The EPCs were connected to a data acquisition system and a laptop to continuously record the pressures induced by the 8-tonne BH-1300 impact roller. EPC data for the roller at rest (static case) and in motion (dynamic case) were analyzed and reported.

Triaxial and direct shear testing was carried out by the authors to complement the results from Mentha et al. (2011) to characterize further the engineering properties of the tailings material. The results for key soil parameters, which are essential for MAT_193, are summarized in Table 1. The Poisson’s ratio was assumed to be 0.3. These values were used in the subsequent numerical modeling.

Table 1. Summary of laboratory test results for key soil parameters.

Soil parameters Results

Cohesion (kPa) 7

Angle of internal friction (°) 37

Elastic shear modulus (MPa) 5.7

Mass density (t/m3) 2.55

4 VALIDATION OF FEM

Validations of the FEM focused on both static and dynamic (single pass at 10.5 km/hr rolling speed) cases. In the static case, the variations of influence stress with respect to depth from the FEM model were compared with solutions derived from classic Boussinesq theory and Fadum’s chart. Comparisons are summarized in Figure 3. Note that, the influence stress plotted is due to the impact roller only; excluding the overburden pressure due to the soil’s self weight. Moreover, the FEM predicted an immediate settlement of 4.4 mm, which is very close to the solution given by theoretical elastic theory of 5.1 mm.

In the dynamic case, the FEM was validated against field data collected by Mentha et al. (2011), and the results are shown in Figure 4. The comparison showed that the FEM accurately predicted the influence stress at various depths and exhibited a smooth trend. The FEM also predicted an immediate settlement of 17.8 mm after a single pass. Mentha et al. (2011) reported settlement data after the 8th and 16th passes only. In order to directly compare the results, approximations using linear and quadratic trend fitting to the field data yielded 17.0 and 18.5 mm respectively for a single pass. The settlement predictions from the FEM lay between these two values. In summary, the results showed that the FEM is able to predict accurately the influence stress and soil settlement in the static and dynamic cases.

1311


Figure 3. Validation of FEM model – static case.

5 RESULTS OF NUMERICAL MODELING

Some of the key results of a single pass are summarized in Table 2. In order to quantify the effectiveness of the impact roller on certain soils and specific site conditions, there is a need to distinguish between the depth of influence zone (or influence depth in short) and improvement depth. The traditional definition of depth of influence zone refers to the depth of soil affected by the load imposed at the ground surface; generally using 10% of the peak stress as a limit. On the other hand, the improvement depth is the depth over which the soil undergoes significant improvement in density and shear strength due to RDC, as illustrated in Figure 5. Improvement depth is, in the authors’ opinion, a more appropriate measure of the effectiveness of the impact roller, as it is a function of soil type, site characteristics and the weight and operating speed of the RDC module. The results indicate that the influence depth is not equal to the improvement depth, as the low influence stress at greater depths may only cause soil to deform elastically, resulting in no change (or improvement) in soil density upon load removal.

Table 2. Summary of a single pass of the impact roller.

Parameters Single pass

Peak stress (kPa) 720

Settlement (mm) 18

Influence depth (mm) 2,640

Maximum density change (%) 0.2

Improvement depth (mm) 2,350

Figure 5 shows the change in soil density varying with depth for both single and multiple passes of the impact roller. The positive change implies that density of the soil increased and the volume decreased. On the other hand, a negative change indicates decreased density and a volume increase. A few curious trends are observed in Figure 5. Firstly, the density of the soil is found to decrease within 200 to 250 mm of the ground surface. Kim (2011) found similar results, where the near surface soils actually become looser due to RDC. This is further confirmed by visual inspection of the surficial soil which is disturbed and loosened as a result of RDC where the soil is displaced laterally by the module rather than compacted. Additionally, the depth where the maximum density change is observed (~900 to 1,150 mm) does not correspond to the depth where peak influence stress occurs (~200 mm), as shown in Figure 5. This indicates that the compaction for the top layer of soil is inefficient; a higher influence stress does not necessarily result in an increased density. Furthermore, the depths and the magnitude of the peaks increase with the number of passes.

Figure 4. Validation of FEM model – dynamic case.

Figure 6 shows the relationship between influence stress and depth with the number of RDC passes, and Figure 7 shows the in situ stress measured in the field using the EPCs. It is evident from these figures that there is an upward trend of peak influence stress as the number of passes increases. This upward trend is expected, as the force imparted by the roller causes the void ratio of the soil to decrease, resulting in increased soil density and surface settlement. With increased density, the pressure wave can more readily propagate to deeper layers, resulting in increased pressures.

Inconclusive

Effective Depth

Figure 5. FEM predicted change in soil density versus depth after single and multiple passes.

Figure 6. Influence stress versus depth for single and multiple passes.

Mentha et al. (2011) used SASW, in conjunction with dynamic cone penetration tests, to assess the same location at intervals of eight passes of the impact roller. Typical SASW results are shown in Figure 8, where it can be observed that increased number of passes results in a noticeable increase in shear modulus between depths of 0.5 to 2.1 m. This is an indication of increased soil density. Similar behavior is observed in the FEM model (Figure 5) between depths of 0.8 to 3.0 m. In Figure 8, above a depth of 0.8 m (same 0.8 m in Figure 5) the results were inconclusive, which is consistent with conclusions drawn from penetrometer and FEM results. Below

1312


a depth of 2.5 m (3.0 m in Figure 5), the varying numbers of passes begin to converge, suggesting that this is the depth of influence of the roller for which there is quantifiable improvement. Below a depth of 2.5 m, results from field study were inconclusive due to insufficient data points.

8 ACKNOWLEDGEMENTS

The authors wish to acknowledge Mr. Stuart Bowes and the staff at Broons for their technical assistance with undertaking site work and for supplying module drawings used in this study. The authors also wish to acknowledge the research and technical staff of the School of Civil, Environmental and Mining Engineering at the University of Adelaide for their valuable assistance.

Figure 7. Peak influence stress recorded by EPCs (Mentha et al. 2011).

(a)

Effective

Depth

Inconclusive

Figure 8. SASW test results for varying passes (Mentha et al. 2011).

(b) 6 RESULTS OF PARAMETRIC STUDIES

A limited series of parametric studies was undertaken. The parameter that were examined were cohesion, shear modulus, soil’s density, internal angle of friction, Poisson’s ratio, mass and width of the roller and its application speed. It was found that the soil parameters that were the most significant in terms of the effectiveness of RDC were shear modulus, Poisson’s ratio and cohesion (to a lesser extent). The variation of the module mass and roller width were also found to significantly affect the magnitude and depth of influence. Some of the results are presented in Figure 9.

Figure 9. Results of parametric studies, varying: (a) cohesion; (b) shear modulus.

9 REFERENCES

Avalle D.L. 2007. Trials and validation of deep compaction using the “square” impact roller. Australian Geomechanics Society, Chapter Mini Symposium: Advances in Earthworks, Sydney, Australia, 17 October, 1-7.

Avalle D.L. and Carter J. P. 2005. Evaluating the improvement from impact rolling on sand. 6th International Conference on Ground Improvement Techniques, Coimbra, Portugal, 18-19 July, 153-160.

7 CONCLUSIONS

This paper discussed and compared results obtained from a field study by Jaksa et al. (2012) with those obtained from finite element analysis modeling (FEM), to assess the effectiveness of rolling dynamic compaction (RDC). The FEM was validated against both theoretical solutions and field data obtained by Mentha et al. (2011). The numerical model was found to predict the soil settlement and soil stresses reasonably accurate for both the static and dynamic cases. It was observed that large surface deformations were noticeable within the first 0.8 m below the ground, with RDC proving to be most effective between depths of 0.8 m to 3.0 m. The soil within this effective depth range demonstrated an increase in soil density with increasing number of passes. A preliminary parametric study found that the most significant factors were soil cohesion, Poisson’s ratio and shear modulus, as well as the width and mass of the RDC module.

Avalle D.L., Scott B.T. and Jaksa M.B. 2009. Ground energy and impact of rolling dynamic compaction – results from research test site. 17th Int. Conf. on Soil Mechanics and Geotech. Engrg.,Alexandria, Egypt, 5–9 October, Vol. 3, 2228–2231.

Bradley A., Crisp A.J, Jiang J. and Power C., 2012. Assessing the effectiveness of RDC using LS-DYNA. Adelaide, Australia: B.Eng.(Hons), The University of Adelaide.

Hallquist, J.O. (2006), LS-DYNA Theory Manual, Livermore Software Technology Corp., March 2006.

Jaksa M.B., Scott B.T., Mentha N.L., Symons A.T., Pointon S.M., Wrightson P.T. and Syamsuddin E. 2012. Quantifying the zone of influence of the impact roller. Int. Symposium on Recent Research, Advances and Execution Aspects of Ground Improvement Works,Brussels, Belgium, 30 May – 1 June, Vol. 2, pp. 41–52.

Kim K. 2011. Impact rollers (soil compaction) numerical simulation of impact rollers for estimating the influence depth of soil compaction,1st edition, Saarbrücken: LAP Lambert Academic Publishing GmbH & Co.

Mentha N., Pointon S., Symons A. and Wrightson P. 2011. The Effectiveness of the Impact Roller. Adelaide, Australia: B.Eng.(Hons), The University of Adelaide

1313

Determination of distribution of modulus of subgrade reaction

Détermination de la distribution du module de réaction d’un sol de fondation

Larkela A., Mengelt M., Stapelfeldt T. Ramboll Finland Oy

ABSTRACT: In this investigation, three dimensional finite element soil modeling was used to model the behavior of a rectangularfoundation block for on-grade cases in which the foundation experiences uniform loading. Modeling was completed using the commercially available three-dimensional finite element software Plaxis 3D Foundation, Version 2.2. Based on analyzing results,distribution of modulus of subgrade reaction was determined.

RÉSUMÉ : Lors de cette étude à éléments finis tridimensionnels, on a analysé le comportement d'une fondation rectangulaire de taille variable qui a été soumise à différentes sollicitations uniformes. La modélisation a été effectuée à l'aide du logiciel commercial auxéléments finis Plaxis 3D Foundation, version 2.2. A l'aide des résultats obtenus, le module de réaction a été terminé.

KEYWORDS: Modulus of subgrade reaction, soil structure interaction, foundation design, three dimensional finite element modeling.

1 GENERAL

The modulus of subgrade reaction is a parameter expressing the pseudo-elastic behavior of subgrade soil beneath the foundation, and is used for structural analysis of soil-structure interaction. The subgrade modulus concept models the behavior of soil as a series of single springs, and is commonly called Winkler’s spring model. The method allows estimation of settlement beneath a loaded shallow foundation member, and thereby facilitates calculation of shear stresses and bending moment magnitudes within the foundation. These calculated values are used for example to dimension the reinforcements for a concrete footing.

The basis of the extensive usage of Winkler’s soil model is in its simplicity to the differential equations which have been used to develop the method. However, it is commonly acknowledged that the assumption of soil units acting as separate, elastic springs of uniform stiffness below a footing does not model realistic foundation behavior. Results are unconservative for example in cases when the foundation is loaded with uniform loading.

In this investigation, three dimensional finite element soil modeling was used to model the behavior of a foundation block on grade cases in which the foundation experiences uniform loading. Modeling was completed using the commercially available three-dimensional finite element software Plaxis Foundation 3D, Version 2.2.

The modeling program was completed for Wärtsilä Finland, a major northern European supplier of integrated power generation solutions. The foundations considered in this program were foundation blocks for large diesel engines. A variety of engine foundations were considered for the study, representing the foundation blocks of a variety of diesel engine sizes and weights.

2 NATURE OF MODULUS OF SUBGRADE REACTION

Modulus of subgrade reaction is a spring constant describing the relationship between applied pressure and resulting deflection (settlement) below a structural element founded on grade. In the structural analysis the soil is modeled as an elastic

half space, and local supporting pressure is assumed to be directly proportional to settlements. The subgrade modulus is not a fundamental soil property and its magnitude depends on many factors, among them shape of the foundation, stiffness of foundation slab, shape of loading on the foundation, depth of the loaded area below the ground surface, and time. As such, it is not constant for a given type of soil, which makes the estimation of a single general value for design a challenging task.

Modeling work performed in this study demonstrates that the theoretical springs of the subgrade reaction modulus also vary below a foundation. The variation of the subgrade reaction modulus arises especially when a foundation slab is loaded with uniform loading, because vertical pressure near the edges and corners of the foundation are significantly higher while the settlements at the same locations are smallest due to bending of the foundation slab, producing high spring constant. Conversely, at the center of the foundation the pressure is smaller and the settlement higher; thus, the spring constant is smaller at that location. The variation in the calculated subgrade reaction modulus causes the foundation to bend even under uniformly distributed loading.

Further, plastification of soil plays an important role in the determination of modulus of subgrade reaction. If the soil is modeled as purely elastic, pressure concentrations are observed at foundation edges and corners, leading to over conservative reinforcement design of the concrete foundation.

3 METHOD OF CALCULATION

3.1 General

The Finite Element Program PLAXIS 3D Foundation 2.2 was used for the analyses. PLAXIS 3D is programmed and built specifically for analysis of interaction between soil and structures.

3.2 Foundation Model

The foundation type used to support the engines is a rectangular block of reinforced concrete. The analyses

1314


considered the behavior of a variety of different engine types/sizes, and the various engine foundation blocks associated with each engine. Depending on the engine type and size, the thickness of an individual engine foundation varies between 0.6m and 1.2m. The length of a foundation block varies between 10.4m and 20.9m and width between 3.3m and 4.8m. The analysis also considered groups of engines, ranging from one to six engines in a row placed along their short axes. During the usage period, the engine is founded on a spring packet on top of the foundation slab. The purpose of the packets is to damp vibration induced by the engine to the foundation. Each engine is founded on a packet of 20 springs; modeling work by others indicates that the load transmitted through the foundation slab and into the subgrade is evenly distributed. Pressure below each engine type varies between 24kPa and 50kPa. Due to relatively thick foundation slab and heavy reinforcement within the slab, the engine foundations were modeled as linearly elastic; this design assumption was discussed and agreed with the structural engineer to be a valid assumption.

Plaxis 3D Foundation 2.2 allows modeling of structural plate units as two dimensional floor elements. This simplification from three dimensional elements was employed, because using this method the program provides bending moments, shear stresses and settlements of plate units in an easily-usable format.

For the purposes of brevity a single engine type is discussed in this report. The results of this engine analysis are reflective of the results obtained for other engine foundation types. Foundation dimensions of this example engine model are 11.9m by 4.1m by 0.6m, with uniformly distributed pressure on a foundation slab of 28.5kPa. Group affects of engines placed closely in rows was found to influence the results obtained; as such, the results provided in this article are representative for the two middle engines in 6 engine foundation group. This arrangement produces the largest estimated foundation settlements, bending of foundations and therefore the largest bending moments within the foundation blocks. Free distance between each engine foundation was 1.1m, reflecting the distance between the engines after installation in a typical facility.

The analysis results reveal that with these rectangular foundation dimensions, the variation of the subgrade modulus is more significant over the length of the foundation as compared with the width of the foundation. Thus the distribution of modulus of subgrade reaction was determined only in longitudinal direction of foundation. This assumption was confirmed by investigating the bending moments and settlements in the diagonal direction across the foundation.

3.3 Soil Model

Plaxis 3D Foundation offers various soil models for different purposes and applications. In this particular case, application of the hardening soil model was considered to be the most suitable model because it is formulated in an elasto-plastic framework.

The soil model considers hardening of the soil by shear hardening and isotropic compression hardening. The isotropic compression hardening can be simplified as hardening of soil, when the soil is placed under isotropic pressure and the pore pressure within the soil is allowed to dissipate. The shear hardening of soil is the increased shear strength of soil, as the pore pressure between the soil particles decreases.

Yielding of the soil occurs if the shear strength of the soil is exceeded in any element node point, because the soil is modeled as elasto-plastic. Yielding of soil below the edges and corners of foundation slab is considered to be very important in determination of naturalistic bending of the foundation and especially in determination of bending moments and shear stresses within the foundation slab. If the yielding of soil would be neglected from the analyses, unrealistically high bending

moments and shear stresses would occur at the edges and corners of foundation slab.

The three dimensional model of the soil space below and around the foundation slab allowed more realistic distributions of stresses. Modeling of the soil space as three dimensional around and below the foundation slab, was the key element in resolving the bending of the foundation slab even with evenly distributed loading in top of the foundation slab. As the pressure is distributed on a wider area around the corners of the foundation slab, the corners and edges of the foundation settle less than the center portion of the foundation even when the loading on the foundation slab is evenly distributed.

These considerations allowed more realistic analyses of the soil especially when compared to conventional Winkler’s spring model, even though simplifications were made.

The dimensions of the model were chosen to be 200m x 200m; depth was chosen to be 50m. Given these dimensions, no boundary effects were observed due to induced stresses during calculation stages.

3.4 Procedure and Order of Analysis

Due to the reason that modulus of subgrade reaction is specifically used for structural analyses the modeling work with PLAXIS was performed in close co-operation with structural engineers. Therefore the following work flow net of work was used in order to determine distribution of modulus of subgrade reaction:1. Dimensions and elastic modulus of foundation were determined by the structural engineer. 2. Load acting on the foundation was determined by structural engineer.3. Soil parameters were defined by geotechnical engineer. 4. Soil – structure interaction was determined by geotechnical engineer using PLAXIS 3D finite element modeling program. 5. Results from soil-structure interaction modeling, including bending of slab, settlement of slab, and bending moments with in the slab were provided for the structural engineer. Example settlement and bending moment distribution maps are presented in Figure 1 of this article. 6. Structural engineer evaluated whether assumption of elastic behavior of foundation was valid. 7. The structural engineer used the modeled values and a finite element model of a beam on Winkler’s springs to determine subgrade reaction modulus (spring constant) distribution below the foundation, to obtain a match between results from the Plaxis model and the model of the beam on springs. Match between both foundation settlements and bending moments within the foundation slab were required to approve the spring model.8. The varying spring constants calculated in Step 7 were used for design of the foundation reinforcements. Using this method, no unique value of subgrade modulus was found, and the actual soil-structural interaction was reflected by the non-uniform values calculated from the modeling and analysis program.

1315


Figure 1: Longitudinal bending moment and settlement map of example engine foundation.

4 RESULTS OF ANALYSES

The investigations and models performed for this study agree with the observation that subgrade reaction modulus is not a fundamental soil property and, for a single soil, varies not only with the foundation dimensions, but also beneath a given foundation. Significant variability in calculated subgrade modulus was observed beneath a single foundation.

A close match was found between computed foundation slab settlements and bending moments within the foundation slab using the Plaxis model and the model using Winkler’s Spring Model. The subgrade modulus is observed to be lowest in the middle and highest at the ends of the foundation slab. Between these two points the subgrade modulus variation correlates well with a second-degree parabolic distribution.

In the example engine foundation case presented in this article, from which the settlement and bending moment distributions are presented in Figure 1, the best correlation between the modeled foundation settlements and bending moments were found with subgrade modulus being 1190kN/m3

at the center and 3160kN/m3 at the ends of the foundation and being parabolically distributed between these peak values.

Due to good correlation between results from parabolic subgrade modulus distribution used in Winkler’s soil model and 3D soil-structure model, parabolic distribution of subgrade modulus was considered to provide sufficiently accurate results. Therefore, more detailed analysis on which the peak subgrade modulus would be located somewhere near the edge of the foundation, due to yielding of soil, were not considered to be required.

5 DISCUSSION OF COMPUTATION METHODS

The method which was used to determine distribution of modulus of subgrade reaction below a foundation was time consuming, but the results were considered positive and logical.

The method allowed definition of subgrade modulus in such a way that reinforcement quantity in the engine blocks could be reduced significantly; resulting in large cost savings for the customer. The designs have since been employed in the field and the engine foundations exhibit acceptable performance in operation.

In common structural design programs Winkler’s soil spring method is often used to model the behavior of soil below a foundation, even though this method is widely found to be

inaccurate and not reflective of reality. The usage of soil spring method of Winkler was justified in the past, when only structural computation methods based on differential equations have been available. However the structural design programs based on finite element methods commonly use Winkler’s soil model for analysis as well in modern engineering practice, even though modern computation methods (e.g. FE modeling) is capable of modeling the soil in much greater detail and more accurately. Use of more sophisticated soil models can be expected to increase the accuracy of design significantly.

As improvements to the current situation, the authors propose the following:

Clause 1: Three dimensional finite element soil space shall be modeled below and around footings.

Clause 2: Soil shall be modeled as elasto-plastic.

Clause 1 would lead to more correct stress distribution within soil. Using this method a foundation slab would bend even when being loaded with uniformly distributed loading. Clause 1 would also take into consideration settlements due to closely spaced foundations, as was the case with the modeled 6 engine foundations, presented in this article.

The plasticization requirement set in Cause 2 would result in more realistic pressure distribution below foundation and problems of extremely high peak pressures occurring at the corners of foundation would not be observed. This will result in more realistic bending moments and shear stresses within foundation slabs.

When the yielding stress of soil is being determined, it should be noted that if the yielding pressure is set to be too low, the region of yielded soil will become too large and the analyzed foundation may not bend as much as it should. This will result into too small bending moments within the analyzed foundation slab, resulting into under reinforcement. Conversely, if the yield strength of soil is set too high, the total displacements may become too small.

In the conducted analysis, it is found to be very time consuming for structural engineer to determine distribution of subgrade reaction below foundation manually. This is due to the reason that distribution of spring stiffness varies significantly below footing and different spring variation shall be determined separately for each load condition. As such it is recommended that in structural design programs using finite element methods, the soil would be modeled using more sophisticated soil models than Winkler’s soil model.

6 REFERENCES

Abdullah W. S. 2008. New elastoplastic method for calculating the contact pressure distribution under rigid foundations. Jordan journal of civil engineering, Volume 2, No. 1, 2008.

Bergdahl U., E. Ottosson and B. S. Malmborg. 1993. Platt grundläggning. SIG Statens geotekniska institut. ISBN 91-7332-662-3

Brinkgreve R. B. J. and Swolfs W. M. 2007. Plaxis 3D foundation. Version 2. ISBN-13: 978-90-76016-04-7

Coduto D. P. 1994. Foundation design Principles and practices. ISBN 0-13-335381-8

Das B. M. 2006. Principles of foundation engineering 6th edition. Cengage learning. ISBN 978-81-315-0202-0

Horvath J. S. and Colasanti R. J. January 2011. Soil-structure interaction research project – a practical subgrade model for improvement soil-structure interaction analysis: parameter assessment. <http://jshce.com/files/ceenge-2011-1.pdf>

Lambe T. W. and Whitman R. V. 1969. Soil mechanics. Wiley. ISBN 978-81-265-1779-4

Robobat ROBOT Millenium version 20.0 – User’s manual <www.robobat.com>

1317

Stability improvement methods for soft clays in a railway environment

Méthodes d’amélioration de la stabilité des argiles moles sous remblai de chemin de fer

Mansikkamäki J., Länsivaara T. Tampere University of Technology

ABSTRACT: In October of 2009 a full-scale railway embankment failure experiment was conducted in Finland. The data gathered from the test established a good verification base for the soil models used in this study. In Finland, the most commonly usedimprovement method for railway embankments with low stability is a counter weight berm, which is designed based on the undrainedshear strength of clay. Undrained shear strength is often underestimated and this inaccuracy is constantly leading to overdesignedcounter weight berms, which can be tens of meters wide. This paper introduces an evaluation of alternative methods to improveembankment stability with wooden pile structures or with sheet pile walls. The study contains a comparison of different pile elements and an evaluation of piles capability for stability improvements. The evaluation is based on 2D and 3D finite element analysis and tothe soil behavior calibrated in the failure test and existing, well investigated Finnish railway embankments with poor stabilityconditions.

RÉSUMÉ : Une expérimentation grandeur réelle d’une défaillance de remblai de chemin de fer été menée en Finlande en octobre2009. Les données recueillies à partir de ce test ont fourni une base pour le modèle de géométrie et de simulation du comportementdes sols exploité pour cette étude. En Finlande, la méthode d'amélioration la plus fréquente est une berme contre-poids conçue sur la base de la résistance au cisaillement de l'argile. La résistance au cisaillement non drainé est souvent sous-estimée et cette imprécision conduit à des bermes contre-poids surdimensionnées, qui peuvent avoir des dizaines de mètres de largeur. Cet article présente uneévaluation de méthodes alternatives pour améliorer la stabilité du remblai à l’aide de pieux en bois ou de murs de palplanches. L’étude présente une comparaison des différents éléments de piliers et une évaluation de la capacité des piliers à améliorer la stabilité.L’analyse s’effectue par éléments finis 2D et 3D, avec un comportement du sol calibré dans le test de défaillance et l’existence bien documentée de remblais finlandais dans des conditions médiocres de stabilité.

KEYWORDS: FEM, 3D analysis, soft clay, embankment, wooden piles, sheet pile wall, stability improvement, railway.

1 INTRODUCTION

Stability of railway embankments on soft clays is commonly calculated with limit equilibrium method (LEM) using undrained strength parameters. In Finland the undrained strength is defined with the Field Vane Test. However, calculations with undrained strength might in some cases underestimate the factor of safety. In some of the soft soil areas the calculated total factor of safety is less than F=1.0 for existing embankments. On the other hand, LEM calculations with effective strength parameters φ’ and c’ tend to overestimate the safety factor for undrained conditions, when the excess pore pressure is not accurately taken into account.

A major problem in effective stress analyses is the assumptions for stress and pore pressure distribution and the difficulty in accounting for yield induced pore pressure. These can be taken into account with finite element method (FEM), if the analyses are conducted with advanced material models and correctly defined parameters (Mansikkamäki et. al., 2011).

To clarify the real stability conditions of Finnish railway embankments, a full scale failure test was conducted in October of 2009 on a soft marine clay deposit in southern part of Finland. Embankment was loaded to failure in 2 days as shown in figure 1. The goals for the test were to gather data for the purpose of improving stability calculation methods and testing the suitability of different instruments for monitoring embankment stability.

The extensive instrumentation is well documented in the work by Lehtonen (2011). Data considering displacements and excess pore pressure development has given good basis for the evaluation of FE analysis and the material models.

Figure 1 Test site after the failure. Instrumented area is between the

containers and the ditch. Loading structure has overturned and the slip surface is protruding from the ditch.

So far FEM stability analyses have been mostly done using plane strain 2D analyses. Recent development of FE programs and increase of the computational capacity have enabled an increasing use of 3D analysis (e.g. Nian et.al., 2012). A stopped freight train on embankment is relatively close to a plane strain stability problem, even though 3D modeling provides possibility to analyze effect of axles or concentrated bogie loads. What comes to stability improvement methods, modeling of three-dimensional structures, for example piles, can be much more precise with a 3D analysis compared to a plane strain approximation.

1318


2 THREE-DIMENSIONAL ANALYSES

The earlier 2D stability analyses with Plaxis 2D 2010 contained evaluation of different material models for soft clays. It was shown that anisotropic S-CLAY1 based (Wheeler et.al. 2003) material models can well express most of the important features of soft clay, such as failure induced pore pressure. The Soft Soil model was also found to be suitable with adjusted soil parameters. It was also found that the counter weight berms can be significantly smaller if design is conducted with the effective strength parameters and a suitable material model compared to the traditional undrained analysis.

The scope of the 3D analyses was to compare them with the 2D analyses and to model stability improvement objects, which would be indefinite to model as plane strain. The 3D FEM analyses were conducted with the Plaxis 3D program (version 2010.2.0.7044). At the first phase the whole test site was modeled to compare 2D and 3D analyses. The geometry model is shown in figure 2.

Figure 2. Full 3D geometry model containing 240 000 nodes and the

2 bogie section (12 m) with a pile row.

However, for the needs of modeling reinforcements, geometry was reduced to two different options. The larger model contained a section of two bogies (12 m) and the smaller geometry was only a 1.0 m thick section from the middle of the site. Larger model was used to evaluate different pile row installations and the smaller model to observe an influence of a single pile in more detail. Observations from the latter analyses are shown in this study.

The Plaxis Soft Soil model was used for the soft clay, while the Hardening Soil model was used for the coarse layers. Parameters and soil behavior is calibrated with the displacement and pore pressure data gathered from the conducted failure test. The basic parameters of each soil layer are shown in table 1. Table 1. Basic material parameters of the soil layers. Corresponding layers are shown in figure 3.

γ

[kN/m3] E50

[MPa] λ* ‐

φ' [°]

c' [kPa]

POP [kPa]

1 Ballast 20 50 38 0.2

2 Sandy fill 19 15 35 0.2

3 Dry crust 16 12 0 30

4 Clay 15 0.166 25 0.2 13

5 Clayey silt 17 0.08 27 0.2 20

3 MODELING WOODEN PILES

Wooden piles can be a cost-efficient method to improve stability in a railway environment. There is also a lot of research data available about the laterally loaded piles (Cai and Ugai 2000, Thompson et. al. 2005).

There are several options available to model laterally loaded piles in a FEM program. The most convenient way is to use Embedded pile elements, which are special beam (line) elements creating a elastic region around them imitating real structural element with a volume. The elastic region around the pile is equal to the pile diameter. The element does not create new geometry points to the model and therefore the analysis can be conducted with coarser mesh compared to the volume pile.

Embedded pile elements cannot take into account a soil-pile interaction. There is no interface between pile and soil and therefore pile always moves with soil without sliding (Plaxis 3D 2010, Dao 2011).

Other options to model piles in a 3D program are a volume pile and a plate element. In practice, the volume pile is a solid soil element, which material model is linear elastic and diameter equal with the pile diameter. To be able to inspect forces affecting the pile, a beam element with very low elastic modulus was inserted to the center of the pile. A plate element is also applicable when the lateral forces are studied. In that case width and stiffness of the plate should be equal to the wooden pile. One should notice that the skin surface area of the plate is not equal to a cylinder shaped pile, which should be accounted in interface strength between soil and pile.

In this case the strength of the soil was fully accounted for the pile skin, even though with a volume pile and a plate element it is possible to use reduced interface strength. The geometry model was a 1 m thick cross section, where one vertical d200 mm wooden pile was inserted 5 m from the center line of the track, equal to 2 m from the embankment toe. The pile was installed through the clayey silt layer to the surface of the sand layer, where the approximated tip resistance of the pile head would be 24 kN.

Figure 3. Vertical pile and displacements in 1 m thick cross section.

Location of the pile is 5.0 m from the center line. Displacement contours are in 5 mm steps from 10 to 55 mm. Soil layers are sketched and numbered.

Train load was set to 70.0 kN/m3. With this load the overall safety factor of the embankment is F=1.23 without a pile and maximum displacement of the embankment is 60 mm, as shown in figure 3. Number of nodes was 19700 in the original geometry without a pile. Volume of the elements was 0.02…0.03 m3 which is very dense mesh for 3D analysis. The embedded pile was modeled using 2 different meshing options. First calculation conducted with the original mesh and then with a refined mesh, where a 200 mm diameter tube was created around the embedded pile. The tube had equal properties with the surrounding soil but it induced a mesh refinement around the embedded pile similar with the mesh, which was automatically created around the volume pile. Otherwise the meshing options were similar for soil layers in the parallel analysis.

In figure 4 a lateral displacement of different pile types from parallel analysis at the end of the loading is shown. From left to right the piles are embedded pile, embedded pile with refined mesh, volume pile and plate element. Maximum displacement was very similar at every case; 29, 31, 32 and 33 mm respectively. Maximum value was slightly smaller for the original embedded pile which could be due to coarser element mesh. On the other hand it also indicates slightly smaller bending moments.

1319


Figure 4. Lateral displacement of a wooden pile. Modeled with

embedded piles, volume pile and plate element.

As the displacements and structural stiffnesses of the piles are equal, the bending moments should also be similar. However, notable difference could be found in bending moments as shown in figure 5. The moment distribution of the embedded pile is very irregular, indicating inexact values. The embedded pile with refined mesh whereas produced practically identical bending moments with the volume pile.

Figure 5. Bending moment of laterally loaded wooden pile. Modeled

with embedded pile, embedded pile with refined mesh, volume pile and plate element.

The outcome of the analysis is that the element mesh should be refined around the embedded pile, if accurate structural forces are important to find out. Inaccuracy of the embedded pile element will probably be emphasized in actual design projects, where coarser element mesh is used. Other outcome was that the different pile elements produced very similar displacements and bending moments, if the element mesh around the piles was similar.

In figure 6 the safety analysis conducted for the different pile element types is shown. Initial settlement of 60 mm is caused by 70.0 kN/m3 train load. One should notice that none of the pile elements have a failure criterion as they are purely elastic. Therefore the safety analysis is not reliable for large displacements as the bending moment of the pile increases beyond the structural capacity of the pile. As the bending moment capacity of a d200mm wooden pile is known to be approximately 15 kNm (Ranta-Maunus 2000), it was further analyzed at which displacement level structural failure may occur. Accordingly the bending moment capacity is reached when the settlement of the embankment is approximately 0.15 m.

It is shown in figure 6 that the safety factor without reinforcements is F=1.23. The volume pile and the embedded pile with refined mesh produces similar safety factors, F=1.29

and F=1.28 respectively for the displacement level of 0.15 m. Safety factor with the plate element is slightly smaller, equal to F=1.26. The embedded pile with the original mesh gives higher safety factor than the other. The factor was found to be F=1.33 indicating that the element can overestimate the stability conditions if the analysis is made without mesh refinement around the pile.

0

0,05

0,1

0,15

0,2

0,25

1 1,1 1,2 1,3 1,4 1,5

Settl

emen

t of e

mba

nkm

ent [

m]

Overall safety factor ΣMsf

no pile plate element

embedded pile (refined mesh) volulme pile

embedded pile Figure 6. Safety analysis of different piles as a function of

embankment settlement.

In general it can be said that the different structural elements produced similar results under operational loading conditions. Embedded pile was influenced by the coarser mesh even thought the magnitudes of forces were correct as an average. In all cases the mobilized forces are clearly smaller than the structural capacity of the piles. The value of maximum mobilized bending moment was 3.78 kNm, when corresponding lateral displacement was 33.4 mm.

A reason for this kind of behavior is a failure mechanism, where the piles are tilting with the soil mass. The foot of the pile has a hinged joint with soil, which causes smaller forces compared to a rigid connection that would be plausible if piles are driven deeper into the dense soil layers.

The installation effects or the effect of interface elements were not taken into account in this study. Obviously these effects should be considered if the piles are used near the railway track. One should also notice that even if the soil behavior is well known due to failure test, the study considering piles is theoretical as no piles were installed for the conducted failure test.

4 SHEET PILE WALLS

Permanent sheet pile walls are used occasionally for the stability improvements. The reason for using this method is usually the lack of space around the embankment and therefore a counter weight berm is not possible.

In the following, a case study from western Finland near Seinäjoki town is presented. A double track was supported with sheet pile walls anchored through the embankment as shown in figure 7. Sheet piles are installed through the soft clay layer (+27…+38) to the hard soil layer. There are no triaxial test results available from this site and therefore the FEM analysis are conducted using typical effective strength parameters of soft Finnish coastal clays. Thus the real stability conditions of this specific site can differ from the factors presented here.

1320


5 CONCLUSION

3D FE analysis can provide useful and valuable information in geotechnical projects even though robustness and mesh independency are not yet at the same level than in the 2D programs.

The embedded pile element seems to give imprecise results when it is used with standard element mesh. Performance is clearly improved when the mesh is refined around the pile element. In that case the results are similar with the volume pile. This feature slightly reduces calculation performance and handiness of the element.

Figure 7. Embankment supported with the sheet pile walls.

Overall safety factor of the cross section is traditionally calculated in LEM so that the slip surface goes under the foot of the sheet pile wall. Often the adequate safety level is not reached until the wall is extended deep to the hard soil layers.

Wooden piles can be used to improve embankment stability if the demanded supporting forces are reasonable. Still, several piles per track meter should be used, as the mobilized lateral forces are quite small.

In figure 8 the results from the FEM stability analysis with the strength reduction method is shown. The initial overall safety factor is F=1.15. With the sheet pile wall, stability is improved so that the safety factor is F=1.76. However, the failure surface is not passing under the wall but through the wall. In this case the wall is modeled as an elasto-plastic plate element which bending moment capacity is 426 kNm, which corresponds a section modulus of w=1200cm3. In this case the failure mechanism includes a structural failure of the sheet pile wall. It was further observed that also the tensile stress of the anchors was very close to failure at this safety level.

If sheet pile walls are used to improve embankment stability, FEA can provide valuable additional information on how sensitive the structural forces are for the soil strength variation and what is the real nature of the failure. FEA was found to be a useful tool for these evaluations as the structural behavior is also accounted for the analysis. It was shown that the bending moment and the anchor force can be so sensitive for soil strength variation that the safety margin can be lower than expected.

6 ACKNOWLEDGEMENTS

The research work presented in this paper is a part of a Life Cycle Cost Efficient Track research programme (TERA), which is conducted by Tampere University of Technology. Research programme is financially supported by the Finnish Transport Agency. Their support is gratefully acknowledged.

Figure 8. Failure surfaces from the safety analysis. Initial FOS=1.15

without the reinforcement and FOS=1.76 with the sheet pile wall.

7 REFERENCES

In the present design codes the design values of maximum bending moment and anchor force is defined by applying partial safety factors for the permanent and variable loads. Factor is lower for permanent, and higher for variable load. In this case the characteristic train load was 40.4 kPa and design load 50.9 kPa. This design load was used to calculate the bending moment Mk and anchor force Fk. The design values for bending moment and anchor force are calculated as follows; Md=1.15Mk=114.3 kNm and Fd=1.15Fk =96.3 kN/m.

Cai F. & Ugai K. 2000. Shear Strength Reduction FEM Evaluating Stability of Slopes with Piles or Anchors. Proceedings of an International Conference on Geotechnical & Geological Engineering Melbourne

Dao T.P.T. (2011).Validation of PLAXIS Embedded Piles For Lateral Loading. Delft University of Technology, Plaxis bv.

Dawson, E.M., Roth, W.H. and Drescher, A. 1999. Slope stability analysis by strength reduction, Geotechnique, vol. 49, no. 6, pp. 835-840.

Farias M.M., Naylor D.J. 1998. Safety analysis using finite elements. Computers and Geotechnics, Vol.22, No. 2, pp.165-181.

Next, a parallel analysis was conducted, as it can be argued that the loads are quite well known compared to the strength parameters of the soil. The strength parameters of the soil layers were reduced using a partial factor of γφ=1.20. Calculation was conducted with the characteristic train load 40.4 kPa. In this analysis maximum bending moment was M=157.5 kNm and the anchor load T=105.0 kN/m.

Lehtonen, V (2011). Instrumentation and analysis of a railway embankment failure experiment. 29/2011 Research report of the Finnish Transport Agency.

Mansikkamäki J., Länsivaara T. 2012. 3D Stability Analysis of a Full-scale Embankment Failure Experiment. Conference proceedings, NGM2012–Nordic Geotechnical Meeting. Copenhagen, Denmark.

Mansikkamäki, J.; Lehtonen, V; Länsivaara, T. (2011). Advanced stability analysis of a failure test on an old railway embankment. Conference proceedings, GeoRail 2011, Paris.

Hence, a relatively small decrease in soil strength caused higher bending moment and anchor force with the characteristic loads than the design values are. The overall safety margin for the bending moment by the means of soil strength was F<1.20. When the stability of the embankment is poor, a small change in soil strength parameters builds up a significant amount of excess pore pressure, which significantly increases the stress in the supporting structure.

Nian T.-K., Huang R.-Q., Wan S.-S. and Chen G.-Q. May 2012. Three- dimensional strength-reduction finite element analysis of slopes: geometric effects. Canadian Geotechnical Journal. Volume 49, Number 5. NRC Research Press Journal.

Plaxis 3D. (2010). Reference and Material Models Manual. Ranta-Maunus A. 2000.Bending and compression properties of small

diameter round timber. Proceedings of World Conference on Timber Engineering. Whistler Resort, British Columbia, Canada July 31 - August 3.

Sensitivity analysis with FE shows that the structural forces are in this case sensitive for soil strength variation. This kind of sensitivity analysis would also be beneficial in practical design cases to ensure a sufficient safety margin.

Thompson M., White D. J. and Schaefer V. R. Dec. 2005. Innovative Solutions for Slope Stability Reinforcement and Characterization: Vol. III. Final Report. Iowa State University, Center for Transportation Research and Education.

Wheeler S.j., Näätänen A., Karstunen M., Lojander M. An anisotropic elastoplastic model for soft clays. Canadian Geotechnical Journal 40: 403–418. 2003.

1321

Effect of wetting- drying cycles on CBR values of silty subgrade soil of Karaj railway

Effet des cycles d’humidification et séchage sur les valeurs CBR des sols de limoneux de fondation de la voie ferrée Karaj

Moayed R.Z., Lahiji B.P. Imam Khomeini International University, Qazvin

Daghigh Y. Azad University of Karaj, Karaj

ABSTRACT: In this research we have investigated the effect of lime-microsilica additive as a modern additive stabilizer on a siltysoil and have evaluated the wetting- drying cycles on it. Thus, for this purpose and also to observe their usage on a practical project,we have taken some samples from bed soil of a region of Karaj railway in Iran, to improve its strength and use it as a railwaysubgrade. Lime and microsilica in different percentage of dry soil weight were mixed with the soil at the soil optimum moisture. Thenafter 28 days curing time, to create saturated condition, they were put in water for 96 hours under a surcharge load of 10 pound (4.5kilogram). Then California Bearing Ratio (CBR) tests were conducted in order to find the best additive that have the maximum effecton soil strength. In the next step, to observe the effect of wetting- drying cycles on the stabilized soil, several specimens which shows the desired CBR value (from an economic and resistance viewpoint) were rebuilt and were exposed to wetting- drying cycles. Results showed that the CBR values were greatly increased as the soil was stabilized with lime- microsilica additive. In addition, an increaseon the CBR values of the stabilized soil by wetting- drying cycles was observed. Results showed that lime- microsilica additive cansuccessfully be considered as a suitable option to stabilize silty soils.

RÉSUMÉ : Dans cette recherche, nous avons étudié l'effet d’un additif de chaux et microsilice en tant que stabilisateur moderne surun sol limoneux et avons évalué l’effet des cycles de humidification-séchage. Ainsi, dans ce but, et afin d'observer aussi leurutilisation sur un projet concret, nous avons pris des échantillons de sol de la région du chemin de fer de Karaj en Iran, pour en améliorer la résistance et pouvoir l'utiliser comme une plate-forme ferroviaire. La chaux avec microsilice a été mélangée avec le sol àsa teneur en eau sol optimale à différentes teneurs en pourcentage du poids du sols sec. Puis, après 28 jours de temps de prise, les échantillons ont été mis dans de l'eau pendant 96 heures sous une surcharge supplémentaire de 10 livres (4,5 kg), afin de créer des conditions saturées. Des tests CBR tests ont été ensuite effectués afin de trouver le meilleur additif vis-à-vis de la résistance du sol. Dans l'étape suivante, afin d’observer l'effet des cycles d’humidification séchage sur le sol stabilisé, plusieurs spécimens ayant lavaleur souhaitée de CBR (d'un point de vue économique et mécanique) ont été reconstitués et exposés à des cycles d’humification etséchage. Les résultats ont montré que les valeurs de CBR ont été considérablement augmentées pour les sols stabilisés avec l’additifde chaux et microsilice. En outre, une augmentation des valeurs de CBR du sol stabilisé par les cycles d’humidification séchage a été observée. Ces résultats ont donc montré que l’additif de chaux et microsilice peut avec succès être considéré comme une optionappropriée pour stabiliser les sols limoneux.

KEYWORDS: Stabilization, Lime, Microsilica, CBR, Wetting - Drying Cycles.

1 INTRODUCTION

Increasing the bearing capacity of weak soils is always one of the most important issues in civil engineering projects especially in road construction. Silts are one of the problematic soils which are needed to be replaced with suitable material or improved by various improvement methods like compaction and stabilization. Silt is a kind of sedimentary geomaterial consisting primarily of very fine particles, including fine sand particles, silt particles, and some clay particles which are often less than 10% by weight. Silt is a type of transitional soil between sand and clay. A soil is defined as silt if its plasticity index is no greater than 10 and the amount of particles greater than 0.075 mm is no greater than 50% of the total.

Silty soils aren’t considered as suitable materials in civil engineering projects due to their low cohesion and friction angel. Using the soils as a road or railway subgrade is generally not possible without stabilization as their characteristics fall below the minimum required. Consequently, stabilization is needed for this kind of soil. Application of stabilizing agents on soils has a long history. Cement was first used as stabilizing agent at the beginning of the twentieth century to mix with soils and form road materials in the United States. Since then, many other kinds of materials, such as lime (Bell 1996) and special additives such as Pozzolanic materials like Fly Ash (Dermatas and Meng 2003), Microsilica (Abd El Aziz 2003), and Rice

Husk Ash (Choobbasti et al 2010), which are as waste material, may be used for soil improvement. Most of the existing stabilizers like lime and cement are not much useful for silts, so the stabilized silts with such kind of stabilizing agents usually cannot satisfy the requirements of road construction. The encountered problems mainly are lower early strength, greater shrinkage, easy cracking, and bad water stability (Bell 1996), (Sheng and Ma 2001).

Indeed, a successful stabilization method depends on many factors such as:

(1) Soil type and properties; (2) stabilizing agent; (3) Stabilizer content; (4) Potential use of the stabilized soil; (5) Field mixing method; and (6) Economical considerations (Mohamedzein et al, 2003).

Therefore, new methods are still being researched to increase the strength properties of silty soils. In this study we evaluate the feasibility of using stabilized silt with microsilica and lime for Karaj railway subgrade in Iran.

Microsilica (or silica fume) is one of the by- product materials which is obtained from silicon material or silicon alloy metal factories. It was discharged into the atmosphere by the factories smoke before the mid-1970s. Nowadays each year nearly 100,000 tons of microsilica is produced on purpose word wide (Karimi et al, 2011). Iran also has a large amount of microsilica production. Although the microsilica is a waste

1322


material of industrial applications, it has become the most valuable by-product among the pozzolanic materials due to its very active, high pozzolanic property and very fine particles. These particles are approximately 100 times smaller than the average cement particle (Karimi et al, 2011).

In previous studies, there have been many researchers investigating the effects of microsilica on the strength and swelling characteristics of clayey soils were investigated. It was seen that microsilica improved the properties of clayey soils (Kalkan, 2009, Kalkan, 2011, Abd El-Aziz et al, 2004, McKennon et al, 1994). Likewise, recently, the effects of microsilica and lime have been investigated on CBR values of sand (Karimi et al, 2011), (Kalkan, 2009, Yarbasi et al, 2007). So their effects on cohesionless soils especially silts aren’t investigated enough yet. Therefore our aim in this study is to evaluate the feasibility of using stabilized silt with microsilica and lime for a railway subgrade and then evaluate the effect of wetting - drying cycles on the soil resistant.

2 MATERIALS

2.1 Soil

The silt used in this research was obtained from an area in Karaj railway project in Iran. Atterberg limits tests were carried out according to ASTM D 4318. The soil Plasticity Index (PI) was obtained 2. The soil was classified as a low plasticity soil according to the unified soil classification system ASTM D 422 - 87. The soil name is ML according to USCS (silty soil with low plasticity). The soil classification is shown in Figure 1.

2.2 Lime

Quick lime which was used in this experiment was obtained from the industrial group Qom-Iran limestone and its chemical composition is shown in Table 1.

2.3 Microsilica

Microsilica has been obtained from Ferroalloy Industrial Co (I.F.I) in Azna. The composition of microsilica mineral is shown in Table 2.

Figure 1. Grain size distribution curve of the silty soil

able 1. Chemical properties of lime

Chimical names Percentage

T

K2OSO3

MgO

40.8 2.65

CaO 51.64

Fe2O3 0.13

Al2O3 0.24

SiO2 1.36

3 EXPERIMENTAL PROGRAM

3.1 Tests procedure

e- microsilica on CBR values of

Ta 2. Chemical properties of microsilica

Chimical names Percentage

To evaluate the effects of limstabilized silty soils, first the optimum moisture of soil was calculated from compaction test. Then the soil was mixed with various contents of lime and microsilica at the soil optimum moisture. Then the oven- dried soil was sieved from sieve #4 and lime and microsilica were added into them in 1, 3 and 5% for lime and 2, 5, 8 and 12 percent of dried soil weight for microsilica. Required amount of water was added to the mixture to obtain soil optimum moisture, beyond. Time and attention were paid to provide homogenous soil additive mixture samples. The CBR tests were carried out on samples which were cured for 28 days after 96 hours immersing according to ASTM D 1883 - 99. And at the end, several wetting- drying cycles were conducted to the optimum mixture of samples which was economic and had proper CBR values to evaluate the effect of the cycles on them.

ble

MgO 0.5~2

3

2O3

8

re 0.4

CaO 0.5~1.5

Fe2O 0.3~1.3

Al2O3 0.6~1.2

SiO2 90~95C 0.2~0.4Na 0.3~0.5 SiO2 0.04~0.0MO 0.02~0.07 P2O5 0.04 MoistuPH

0.01~6.6~8.8

3.2 Compaction tests

um water content and the soil To determine the soil optimmaximum dry unit weight, the modified compaction tests were carried out according to ASTM D 1557 – 91. For this purpose, the oven- dried soil passing sieve #4 was compacted in five layers by 56 blows with 4.5 Kg hammer from 45 cm height in 6- inch mold according to procedure C from respective standard test method.

1323


3.3 California bearing ratio (CBR) Tests

The California Bearing Ratio (CBR) test is one of the most widespread tests to determine strength and bearing capacity of base, subbase and subgrades for use in road, railway and airfields pavements. To demonstrate the influence of lime- microsilica additive on the bearing ratio of the silty soil, a series of bearing ratio tests were carried out on stabilized and unstabilized specimens. The tests were conducted according to ASTM D 1883 – 99. The soil with different mixtures of lime and microsilica were compacted in 6" modified proctor mold in five layers by 56 blows in per layer at the soil optimum moisture obtained from compaction tests. For curing the samples, they were placed in constant moisture and temperature for 28 days. To conduct the tests in soaked condition, they were immersed in water for 96 hours under the 4.5 Kg (10 pound) overload according to standard test method. The CBR tests were carried out after 20 minutes to drain the samples. Meanwhile swelling potential changes were measured during the soaking time.

3.4 Wetting - drying tests

After performing the CBR tests, one mixture was chosen as a desired sample from an economic and resistance viewpoint. To evaluate the effect of wetting-drying cycles on strength of selected sample, CBR tests were taken. The desired sample was rebuilt three more times in the same previous condition on 6-inch CBR molds. The samples were subjected to wetting- drying cycles after 28-day curing time and required 96 hours for soaking. The samples were placed in room to air-dry after soaking for 24 hours. Then they were again submerged in water for next 24 hours and thus to expose to one wetting-drying cycle. This process was repeated 3 and 5 times for samples; Then CBR tests were carried out on them.

4 RESULTS AND DISCUSSION

4.1 Compaction tests

Compaction tests were carried out on the silty soil. The soil optimum moisture and the soil maximum unit weight were found to be 14.2% and 17.2 KN/m3 respectively. Compaction tests results are drawn in Figure 2.

Figure 2. Compaction test curve

4.2 Effect of additives on the CBR

To compare the soil resistant with different amount of additive, a series of samples were prepared in modified proctor mold. The CBR tests were conducted in both stabilized and unstabilized silty soils at the soil optimum moisture with

different amount of lime and microsilica. The CBR value of the unstabilized soil was 4.8%. The effect of various amount of additive on CBR values of samples are shown in Figure 3.

From Figure 3, it can be observed that in low amount of lime (1 percent of dry soil) increase in microsilica amount up to 8% causes increase in CBR values and then decrease but for 3% and 5% lime increase in microsilica amount causes increase in CBR values. The maximum CBR value of the samples was occurred in 5% lime and 12% microsilica. CBR value in this composition was increased from 4.8% for unstablized soil to 470.8% for the stabilized soil. So it is seen that up to 466% increase in CBR value of stabilized soil in compare of unstabilized silty soil.

In addition, it is observed that the dry unit weights were increased by adding the lime-microsilica additive to samples and samples moistures were decreased by adding the lime-microsilica additive to them in overall.

0

50

100

150

200

250

300

350

400

450

500

0 2 4 6 8 10 12 14

1% Lime

3% Lime

5% Lime

Microsilica Content (%)

CBR(%

)

Figure 3. The effect of various amount of lime- microsilica additive on CBR values of stabilized soil

4.3 Effect of additives on samples swelling

The samples swelling were measured during the 96-hour of CBR samples soaking. There were seen swelling potential rate were decreased reverse of strength. Unstabilized soil swelling was 0.55mm and stabilized swelling samples were decreased up to 0 mm.

4.4 Effect of wetting - drying cycles on samples CBR values

The sample stabilized with 3% lime and 2% microsilica was chosen as the most desirable sample in terms of economy and resistance and alternate wetting-drying cycles were conducted on it. The result of wetting-drying cycles on CBR values of the sample are given in Figure 4. It is observed that the CBR value was increased after first wetting- drying cycle. Thereafter the sample CBR starts to decrease gradually. The reason for increasing CBR at first is assessed by decreasing in permeation due to lime- silica fume stabilizer that 96 hours submerging was not enough for required moisture for the reaction between lime, silica fume and soil that noticed in introduction section. It is noteable that the CBR rate after fifth cycle is still more than initial CBR rate. Therefore wetting- drying cycle not only had no negative effect on specimen strengths but also help to gain the soil strength stabilized with lime- silica fume additive.

1324


120

130

140

150

160

170

180

190

0 1 2 3 4 5 6

Wetting‐Drying Cycles

CBR(%

)

6 REFERENCES

Bell, F.G. 1996. "Lime stabilization of clay minerals and soils", Engineering Geology 42, pp. 223–237.

Dermatas, D., Meng, X.G. 2003. "Utilization of fly ash for stabilization/ solidification of heavy metal contaminated soils", Engineering Geology, pp. 377–394.

Abd El Aziz, 2003. "The Effect of Silica Fume Substitution on the Characteristics of Ordinery Portland cements Pastes and Mortars", Civil Engineering Magazine, Vol. 24 No. 2, p. 715-725.

Choobbasti, A.J., Ghodrat, H., Vahdatirad, M.J., Firouzian, S., Barari, A., Torabi, M., Bagherian A., 2010 "Influence of using rice husk ash in soil stabilization method with lime", Earth Sci. China, pp. 471–480

Bell, F.G., 1995. "Cement stabilization and clay soils, with examples", Environmental and Engineering Geoscience, pp.139–151.

Sheng, A.Q., Ma, M., 2001. "Experimental study on stabilization of subbase of bearing sand silt with low liquid limit", East China Highway 5, pp.42–46.

Figure 4. The effect of wetting-drying cycles on CBR values of the 28-day sample stabilized with 3% lime and 2% microsilica

Mohamedzein, Y.E., Al-Rawas, A.A. and Al-Aghbari, M.Y. 2003. "Assessment of sand– clay mixtures for use in landfill liners", Proceedings of the International Conference on Geo- environmental Engineering, Singapore, pp. 211–218.

5 CONCLUSIONS

The influences of lime- microsilica additive on silts and also wetting-drying cycles on them and its utilization in Karaj railway subgrade were investigated in this study and following conclusions were drawn:

Karimi, M., Ghorbani, A., Daghigh, Y., Kia Alhosseini, S., Rabbani, P., 2011. "Stabilization of silty sand soils with lime and microsilica admixture in presence of sulfates" Pan- Am CGS Geotechnical conference.

Lime- microsilica additive played an important role in the development of the CBR values of the soil. The CBR values increased in response to adding the stabilizer. The CBR value of the unstabilized soil was increased from 4.8% to 470.8% by adding 5% lime and 12% microsilica.

Kalkan E, 2009. "Influence of silica fume on the desiccation cracks of compacted clayey soils", Applied Clay Science, pp. 296–302.

Kalkan, E., 2011. “Impact of wetting–drying cycles on swelling behavior of clayey soils modified by silica fume", Applied Clay Science, pp. 345–352.

Results show lime-microsilica additive increase the samples dry unit weight and sample's moistures are decreased after soaking by adding lime-microsilica additive.

Abd El-Aziz, M.A., Abo-Hashema, M.A., El-Shourbagy, M., 2004. “The Effect of Lime-Silica Fume Stabilizer on Engineering Properties of Clayey Subgrade", Engineering Conference, Faculty of Engineering, Mansoura University, Paper No. 96.

It is an important result that samples swellings had a large reduction by increase of the lime- microsilica additive content.

It is an important result that wetting- drying cycles not only had no negative effect on CBR values of the sample but also help to gain the strength of the silt which is stabilized with lime- silica fume additive due to increasing required moisture for lime, silica fume and soil reaction.

McKennon, J.T., Hains, N.L., Hoffman, D.C., 1994. "Method for stabilizing clay bearing soils by addition of silica and lime", Patent Cooperation Treaty (PCT), International Application Published Under the Patent Cooperation Treaty (PCT), Patent Classification: C09K 17/00, Publication Number: WO 94/06884.

In conclusion, because of considerable strength of stabilized silty soil with lime- microsilica additive in comparison of unstabilized soil, application of lime- microsilica additive is recommended for subgrade and even base of civil projects.

1325

On the Permanent Deformation Behavior of Rail Road Pond Ash Subgrade

Sur le comportement en déformation permanente d’une assise ferroviaire en cendres volantes de bassin

Mohanty B. Dept. of Civil Engg., IISc Bangalore, India

Chandra S. IIT Kanpur, Indian

ABSTRACT: In this study repeated load triaxial tests were conducted on reconstituted pond ash specimens and permanent deformation calculations have been made taking the stress history and number of passes of vehicular traffic loading into consideration. Tests were performed at different moisture content levels with varying dry unit weights, and at different stress levels simulating the environmental and traffic conditions. Specimens were prepared using moist tamping technique so as to obtain densitycloser to the field density. Test results were analyzed to study the effects of confining pressure, deviatoric stresses, and degree of saturation on the permanent deformation response of pond ash. Results show that both traffic and environmental condition play an important role in the permanent axial strain behavior of the material. Furthermore, the shakedown limit describing a critical stress level that exists between stable and unstable condition is also examined for the design purpose.

RÉSUMÉ : Dans cette étude des essais triaxiaux à charges répétées ont été effectués sur des échantillons reconstitués de cendre volante de bassin et des calculs de déformation permanente ont été effectués, prenant en compte l’historique de contraintes et lenombre de passages du chargement de circulation des véhicules. Des essais ont été réalisés à différents niveaux de teneur en eau et avec des densités sèches variables, et à différents niveaux de contrainte simulant les conditions environnementales et de trafic. Des échantillons ont été préparés utilisant la technique du compactage humide afin d'obtenir la densité la plus proche de la densité en place. Les résultats d'essai ont été analysés afin d’étudier les effets de la pression de confinement, des contraintes déviatoriques et du degré de saturation, sur la réponse en déformation permanente de la cendre. Les résultats prouvent que le trafic et l'état environnemental jouent tous deux un rôle important dans le comportement axial en déformation permanente du matériau. De plus, la limite de shakedown caractérisant un niveau de contrainte critique séparant l'état stable de l’état instable est également examinée vis-à-vis du dimensionnement.

KEYWORDS: Pond ash; Train loading; Triaxial tests; Permanent deformation.

1 INTRODUCTION

Pond ash is a by-product of coal-fired electric power plants found abundantly in India. In order to avoid environmental problem, it can be used in the construction of transportation facilities in bulk quanties. The use of pond ash containing a large fraction of bottom ash in rail road pavements or subgrade will experience repeated rail traffic loading while in-service. This material should be assessed for its suitable use by considering resistance to permanent deformation measured from repeated load triaxial tests. It has not been adequately researched in the past and is investigated in this study simulating the environmental and traffic conditions. In laboratory, one-way cyclic triaxial tests are generally conducted to obtain the deformation characteristics of subsoils under repeated traffic loading simulating the in-service loading conditions induced by passing vehicles. Hence, one-way cyclic triaxial tests on remolded pond ash specimens were performed in this study under undrained conditions with a constant confining pressure and different cyclic applied compressive (non-reversal) deviatoric stresses for each test. In this case the axial deviatoric stress remains the major principal stress and shear reversal does not occur during the test. The permanent axial strain accumulated with respect to number of applied loading cycles were recorded for each test and analyzed to study the influence of different controlling parameters on the one-way cyclic behavior of pond ash.

2 TEST MATERIALS AND SAMPLE PREAPRATION

The pond ash used for preparation of remolded samples were sampled near the discharge point, near the margins of the wet disposal ash pond of a thermal power plant producing fly ash and bottom ash with a typical production ratio of approximately 80:20 by weight. The disturbed, completely saturated pond ash

samples were oven-dried, and then thoroughly mixed to obtain the representative homogeneous samples. The specific gravity (Gs), optimum moisture content (wopt), and maximum dry unit weight (d max) of the pond ash were found to be 2.36, 33.6%, and 11.2 kN/m3, respectively. From grain size distribution, it is observed that the dominant particle size is in the sand size range. It contains 77.81% sand, and 20.56% silt size particles. The coefficient of uniformity, Cu is obtained as 7.39, while the coefficient of curvature, Cc is 2.07. The ash was classified based on the classification system proposed by Prakash and Sridharan (2006). It is found to be Non-plastic sand-silt size fractions and is designated as SMN. The samples were reconstituted at different initial dry unit weights [Relative Compaction: RC = 90%, 95%, 97%, and at standard proctor maximum dry unit weight (at MDD)] and at different water contents giving different degrees of saturation (Sr). The relative compaction, RC is defined as the percentage of desired dry unit weight (d) to the maximum dry unit weight (d max) that obtained from the standard proctor compaction curve. The specimen at 97% RC was reconstituted using water content value on the wet side of the standard proctor curve. The samples were prepared in accordance with the conventional moist-tamping technique, as it is a simple and easy method to provide good control over obtaining the wide range in target density (Ladd 1978). First, an appropriate quantity of oven-dried representative pond ash sample was calculated with respect to the desired dry unit weight. Then, de-aired water corresponding to desired moisture content was measured and mixed to form a mixture. Cylindrical split mold of 50 mm in diameter and 100 mm in height was selected for sample preparation. The prepared mix material (wet or moist ash mixture) was carefully placed and compacted inside the specimen mold in five identical layers, subdividing the total mass into five equal parts approximately. The specimen was prepared on a trial for at least three times to check the desired

1326


dry unit weight. Following compaction, the cylindrical specimens were subsequently removed from the split mold sampler, placed, isotropically consolidated and sheared in the cyclic triaxial apparatus.

2.1 Testing Apparatus

The one-way compressive cyclic triaxial device supplied by M/s. Geotechnical Instruments International Limited, Germany, was used in this research. The apparatus is computer controlled and has a provision for testing cylindrical soil specimens under both drained and un-drained conditions, with programmed deviatoric loading sequences and data acquisition rates at eight readings per applied loading or stress cycles. The system consists of a pneumatic stress-controlled actuator which is capable of generating reasonable representation of multiple cycles of compressive axial deviatoric stresses at multiple applied loading frequencies between 0.1 Hz and 10 Hz (cycles per second), with three types of built-in semi-sine, triangular, and square waveforms defined by means of external input. The vertical cyclic compressive deviatoric stresses could be applied to the specimen via the top specimen cap connected to the vertically movable frictionless shaft or loading piston going through the plexi-glass triaxial pressure cell. The loading ram or piston is directly connected to the actuator for application of one-way cyclic compressive loading. A load transducer with a capacity of 5 kN located below the bottom end platen, inside the plexi-glass triaxial pressure cell was used to monitor and measure the applied deviatoric stresses during testing. It is a constant confining pressure triaxial set-up applying the confining pressure with the use of pressurized air, which remains the same during consolidation and shearing. A sensitive Linear Variable Displacement Transducer (LVDT) of capacity 50 mm (resolution 0.01 mm) located outside of the triaxial pressure cell was used to monitor and measure the low-amplitude axial/vertical deformations of the specimen with high accuracy during testing. The applied initial effective confining pressure, back pressure, one-way compressive cyclic deviatoric/axial load, development of axial deformations etc. could be monitored using a built-in data acquisition system and recorded in a notepad file during testing with a computer connected to the device. The apparatus is supported by software which enables the user to perform stress-controlled testing only. A plexi-glass triaxial tank with full of de-aired water at the bottom of the one-way cyclic triaxial test set-up was used to fill the triaxial pressure cell when necessary and has the provision of draining the water from the triaxial pressure cell by gravitation after each testing

2.2 Testing Procedure

It was clear from the literature that, compositional and environmental factors primarily influence the permanent deformation characteristics of subgrade soil under one-way induced traffic loading. In the field, presence of moisture plays a vital role in either a road or railway pavement system and is one of the most important environmental considerations for strength and deformation behavior of material under cyclic loading. The moisture content may vary during the life time of the structure from the construction moisture content to full saturation with the ingress of moisture with seasonal changes or capillary action. Hence specimens were reconstituted to different moisture contents giving different initial degree of saturation. Three compaction moisture contents and dry density conditions were selected for the study. The applied level of confining pressure and deviatoric stresses also affect the deformation characteristics of the material under traffic loading. Hence, tests were conducted under a range of initial effective confining pressure (3c) of 15, 25, and 35 kPa, which is the range of stresses for embankment

of small height. All the remolded specimens were isotropically consolidated under an initial effective confining pressure. Following, samples were sheared cyclically under undrained condition. Tests were performed with different deviatoric stress levels. Fig. 1 shows the typical sinusoidal semi-sine wave cyclic load applied during the cyclic triaxial compression tests, with corresponding response recorded using data acquisition system during testing.

Figure 1. Typical sinusoidal semi-sine wave cyclic load form applied on the specimen and the response received using data acquisition system during the one-way cyclic triaxial compression tests

Each test was of constant-amplitude, consisted of cycling the stress pulse at only one level of cyclic deviatoric stress varying between zero and a preset value at a frequency of 1 Hz. During the tests, only a deviatoric stress (σd) is applied cyclically while the confining pressure (3c) remains constant. Tests were conducted on unsaturated or partially saturated specimens, i.e. the degree of saturation (Sr) employed during reconstitution of the sample was maintained same during the testing, without a back pressure saturation. Few samples were reconstituted at relative compaction dry unit weight equal to 95% giving degree of saturation of 52.70% and the samples were partially saturated by applying back pressure to obtain degree of saturation ranging between 65 and 95% before shearing, to study the effect of degree of saturation (post compaction) on the deformation response of the material. Since during the application of cyclic shear stress, the samples were not fully saturated, pore water pressure was not measured during shearing. During the test, the software presents the results in the form of a table in a note pad file. The raw data was then transferred to an excel sheet and plots of the desired quantities were obtained for the study.

The performance of road and railway pavements resting on compacted material primarily depends upon the stiffness or load-deformation characteristics of the material. Hence, in the present study, during each one-way cyclic triaxial test, the total and permanent deformations of the specimens were monitored and recorded to calculate the plastic or permanent (p) and resilient axial strains (a). The accumulation of permanent axial strain with load cycles is presented in this paper. As the development of permanent deformation in the specimen under repeated loading is a gradual process during which each load cycle contributes a small increment to the accumulation of strain, all the tests were conducted up to the development of sufficient permanent strain in each of the specimens tested. During the test, as the stiffness of the material gradually increases, causing a reduction in the development of permanent deformation under subsequent repetitive loading, tests were stopped after 10,000 applied load cycles.

3 TEST RESULTS AND DISCUSSION

Permanent axial strain mainly depends on the intensity of applied cyclic axial deviatoric stress and number of loading cycles and generally used to study the deformation characteristics of the compacted material. In this study, the effects of various factors such as applied cyclic deviatoric

1327


stress, initial effective confining pressure, and degree of saturation on the permanent axial strain response of pond ash are studied. In order to study the effect of applied cyclic deviatoric stress on the permanent axial strain response, results of tests performed at same initial effective confining pressure but different deviatoric stress for the specimen compaction dry unit weight, d = 10.64 kN/m3 (RC = 95%) are plotted in Fig. 2. It is observed that at a constant initial effective confining pressure the applied cyclic deviatoric stress showed a considerable influence on the permanent axial strain. Higher the cyclic deviatoric stress higher is the permanent axial strain at the same initial effective confining pressure. With increase in applied cyclic deviatoric stress the variation in permanent axial strain is observed to be less for high initial effective confining pressure. For example, with increase in applied cyclic deviatoric stress from 71.27 to 203.64 kPa, the corresponding increase in permanent axial strain is approximately 85%, at initial effective confining pressure, 3c = 35 kPa.

Figure 2. Relationships between permanent axial strains versus applied cyclic deviatoric stress

Fig. 3 shows the effect of initial effective confining pressure on the permanent axial strain response, at same applied cyclic deviatoric stress but different initial effective confining pressures for same reconstituted dry unit weight/density.

Figure 3. Relationships between permanent axial strains versus initial effective confining pressure

It may be observed from the figure that the applied cyclic deviatoric stress has a significant effect on permanent axial strain at low initial effective confining pressure. High value of

permanent axial strain is observed at low initial effective confining pressure. The permanent axial strain decreases as the initial effective confining pressure increases and the decrease is more pronounced at high applied cyclic axial deviatoric stress.For example, with increase in initial effective confining pressure from 15 kPa to 35 kPa, the corresponding decrease in permanent axial strain value is approximately 89%, at an applied cyclic deviatoric stress, d = 152.73 kPa.

The variation of permanent axial strain is plotted in Fig. 4 for three degrees of saturation, for initial effective confining pressure, 3c = 25 kPa and applied cyclic deviatoric stress, d = 127.27 kPa. The degree of saturation of the specimen during the test was kept same as the degree of saturation during compaction. It may be observed that with increase in degree of saturation, a significant increase in permanent axial strain values is obtained. The increase in permanent axial strain is gradual up to the degree of saturation corresponding to the MDD and is more rapid beyond this value. It is observed that the permanent axial strain increases by 57.80% as the degree of saturation increased from Sr = 52.7% to Sr = 77.71%.

Figure 4. Relationships between permanent axial strains versus degree of saturation at compaction stage (without a back pressure saturation)

The effect of degree of saturation during shearing on permanent axial strain response of pond ash is shown in Fig. 5.

Figure 5. Relationships between permanent axial strains versus degree of saturation (with back pressure saturation)

Tests were conducted at the degree of saturation maintained during compaction stage (Sr = 52.70%) and increased to Sr = 65%, 75%, 85%, 90%, and 95% respectively, before shearing.

1328


All the tests were conducted at initial effective confining pressure of 25 kPa and applied cyclic deviatoric stress of 50.91 kPa. It is observed from the figure that the degree of saturation during shearing has a significant effect on permanent axial strain response of pond ash specimen. With increase in degree of saturation the permanent axial strain values increase. A 110% increase in permanent axial strain is observed when the degree of saturation increases from 52.70% to 95%.

Fig. 6 shows the relationship between accumulation of permanent axial strain and number of applied loading cycles at a constant effective initial confining pressure of 15 kPa for a range of applied deviatoric stress levels. It can be clearly seen that with increasing deviator stress levels the magnitude of accumulated permanent strains increases with loading cycles. Depending on the level of applied stress, at small stress levels specimens experienced some value of permanent strain but at high stress levels test specimens have achieved failure after a finite number of applied load cycles.

Furthermore, as the stress level exceeds a specific value (critical stress), the permanent axial strain accumulates rapidly with the number of applied load cycles which exhibits the unstable conditions in terms of excessive permanent deformation in the test specimen. The test results reported here suggest that at stress levels greater than or equal to 91.64 kPa, permanent strain accumulates rapidly. Hence applied cyclic deviatoric stress should not exceed this value, so as to avoid the excessive plastic strain in the subgrade.

Figure 7. Relationship between permanent axial strain rates (log scale) versus permanent axial strain 4 CONCLUSIONS

This study aimed in understaing and characterising the developement of traffic load induced cumulative permanent axial strain in the compacted ash specimens in repated loading triaxial (RLT) tests. The investigation aimed specifically at evaluation of the magnitude of the permanent axial strain, with combination of various applied deviatoric stress and confining stress level, and the factors affecting it, as it has not been done before. The following conclusions are drawn from the investigation.

The occourance of permanent strain under traffic loading is controlled by several factors. It incresaes with increase in number of load cycles, applied cyclic deviatoric stress, and degree of saturation, and decreases with increase in initial effective confining pressures.

Figure 6. Relationship between permanent axial strain versus number of applied load cycles in undrained conditions

If the ash specimen is subjected to deveiatoric stress smaller than the critical stress, peramanent strain increases at the beginning of test, and reaches a peak value after a finite number of applied load cycles, and then remains constant till the end of test or practically unchanged, attributing to stable state. If cyclic deviatoric stress is higher than the critical stress then the strain will change permanently with number of applied loading cycles attributing to unstable state. Hence the amplitude of permanent strain represents a boundary between two fundamentally different kinds of one-way cyclic behavior in the compacted pond ash specimen under induced repeated traffic loading.

Fig. 7 shows the associated permanent strain rate during the one-way cyclic triaxial tests in undrained conditions. A total of seven tests with combination of various deviatoric stress and confining stress values have been presented. Two different cases viz. stable and unstable states are considered and labeled on the figure for the illustration of permanent deformation behaviour under repeated loading in this study.

5 ACKNOWLEDGEMENTS

The experimental work presented in this paper was conducted at the Geotechnical Engineering Laboratory at the Indian Institute of Technology Kanpur, India. The first author gratefully acknowledges Indian Institute of Technology Kanpur for providing the one-way cyclic triaxial testing device for conducting the experiments successfully.

As it is seen in the figure, in stable state, during the applied load cycles the permanent strain rate decreases gradually and reaches a constant value depending on the cyclic stress level applied to the specimen. Where ash material is in stable equilibrium and can be said to be in shake down range and would be permitted in the subgrade. In this case total accumulated strain is sufficiently small. In contrast, in unstable state the permanent strain rate decreases very slowly depending on the applied stress level than that observed in the stable state. It would result the failure in the subgrade and should be prevented.

6 REFERENCES

Ladd R.S. 1978. Preparing test specimens using undercompaction. Geotech. Testing J. 1 (1), 16-23.

Prakash K. and Sridharan A. 2006. A geotechnical classification system for coal ashes. Proc. Inst. Civ. Eng. Geotech. Eng. (UK) 159 (GE2), 91-98.

1329

Evaluation of the Performance of Road Embankments over North Jakarta-Soft Soils

Évaluation de la performance de remblais routiers sur les sols mous du Nord de Djakarta.

Murjanto D., Rahadian H., Hendarto Directorate General of Highways, Ministry of Public Works, Indonesia

Taufik R. Institute of Road Engineering, Ministry of Public Works, Indonesia

ABSTRACT: R.E. Martadinata road located in North Jakarta is a 7.3 km long - arterial road connecting the Tanjung Priok Port and the Western part of Jakarta. It was initially built 30 years ago crossing over the reclamation of Ancol area. Its structure consisted ofthe embankment with flexible pavements over North Jakarta-soft alluvial deposit which is classified as CH-MH soil based on USCSclassification system. To compensate settlement, the pavement level was raised several times. The first embankment failure is 100 mlong, occurred at KM 2+250 in September 16, 2010 at 03.16 AM. This collapsed segment was already repaired. However, anotherindication of an embankment failure appeared at Km. 4+100 m in March 2011. To avoid another possible failure, a morecomprehensive stability and settlement analysis of the road embankment using a more detailed site investigation was conducted. Thispaper presents geotechnical data collection, geotechnical characterisation, geotechnical analyses at 4 zones which probable prone to stability and settlement problems, and some proposed design to strengthen the road embankment.

RÉSUMÉ : La route R.E. Martadinata, située au nord de Djakarta est une artère longue de 7,3 km reliant le port de Tanjung Priok et la partie occidentale de Djakarta. Elle a été construite il y a 30 ans dans la zone Ancol gagnée sur la mer. Sa structure se composait de remblai surmonté de chaussées souples, sur les dépôts alluviaux mous du nord de Djakarta, classés comme sols de type CH-MH selonla classification USCS. Pour compenser le tassement, le niveau de la chaussée a été rehaussé à plusieurs reprises. La rupture d’unpremier remblai de 100 m de longueur s'est produite au PK 2+250 le 16 septembre 2010 à 3h16. Ce tronçon effondré a déjà été réparé. Cependant, d’autres signes de rupture de remblai sont apparus au PK 4+100 en mars 2011. Pour éviter une nouvelle rupture potentielle, une étude plus approfondie de stabilité et de tassement a été entreprise, basée sur une reconnaissance géotechnique plusdétaillée. Cet article présente la synthèse des données géotechniques, la caractérisation géotechnique, les analyses géotechniques de 4 zones sujettes à des risque de rupture et de tassements, et quelques unes des mesures préconisées pour renforcer le remblai.

KEYWORDS: slope stability, settlement, raising, lightweigth material, ground anchore

1 INTRODUCTION

R.E. Martadinata road which was originally constructed over 30 years ago is an arterial road linking the Tanjung Priok port in northen part of Jakarta and the western part of Jakarta. Along 7.5 km of this road, from Simpang Lodan to Gate 3 of the port, consists of embankment which lies over soft deposit aluvium soil classified as CH soil based on USCS classification system and flexible pavement.

The first collapse of a road embankment of R.E. Martadinata Road occurred at Km 2+250 on September 16, 2010. Investigation indicates that the road embankment failure was significantly triggered by the riverbed and slope embankment scour coupled with the decline in sea water level which was suspected to be at the lowest level at the failure (Rahadian et al, 2011). The raise of pavement thickness also contributes to reduce the factor of safety (FS) value. The rehabilitation of this collapsed road was successfully done by Ministry of Public Works.

On March 2011, another potential collapse of a road embankment occurred at KM 4+100. In order to investigate potential problems that will lead to failure, an extensive site investigation was carried out.

This paper presents stability and settlement analysis, and geotechnical data collection for detailed engineering design to prevent potential loss due to the collapse of Martadinata Road.

Numerical analyses of four zones were performed by using Plaxis software.

2 SITE CONDITIONS AND SOIL PROPERTIES

Soil parameter determined from both field and laboratory testing. The field investigation consists of traffic volume survey, topography and bathymetry survey, tidal measurements, 14 bored holes and 17 Cone Penetration Tests (CPT on shore) on road embankment, 35 CPT tests in Japat River (CPT off shore) parallel to the road (Institute of Road Engineering, Ministry of Public Works. 2011). Soil borings were carried out on the shoulder of the road towards Tanjung Priok while tests on the road lanes towards West Jakarta were not done due to insufficient space. CPT tests were done both on the road and the Japat River.

Based on the drilled bore logs and CPT, the soil beneath the road up to 30m deep is classified into 4 layers (see Fig. 1). On top is a quite thick layer of alluvium clay deposit to a deep of about 11 m-16 m with the cone tip resistance qc values are around 784 kPa. Beneath this layer, a layer of sandy silty clay with average 4 m thick. The third layer is a layer of dense sand with the average thickness about 8 m with the cone tip resistance qc values are around 4923 kPa. The last layer is a layer of sandy silty clay founded between 25 m and 30 m deep. A 1m- lens of dense sand also founded between 28 m – 29 m deep at BH 10.

1330


Figure 1. Soil stratigraphy of RE. Martadinata Road.

2.1 Laboratory testing

Based on plot of Atterberg limits value (liquid limit, LL, and plastic limit, PL), moisture content, the liquidity index and consistency index versus depth; the soil has a very soft to soft consistency. The water content of the soil is also close to its liquid limit.

Atterberg limits values can be used to determine the classification of cohesive soil by plotting the values of LL and PI on the plasticity chart based on the study of Casagrande (1932). By plotting a value of PI and LL on a USCS plasticity chart, the majority of dots being around A line that can be classified as inorganic clays or inorganic silt with high plasticity CH-MH (See Fig. 2).

CH or OH

MH or OH

"A" Line

"U" Line

ML or OL0

10

20

30

40

50

0 10 20 30 40 50 60 70 80 90 100 110 120

Plas

ticity

Inde

x,

Liquid Limit, LL (%)

60

70

80

90

100

Ip (%

)

BH 2 BH 3 BH 4 BH 5 BH 6 BH 8

BH 9 BH 10 BH 11 BH 12 BH 13 BH 14

Figure 2. USCS plasticity chart

3 STABILITY AND SETTLEMENT ANALYSIS

Stability and settlement analyses were performed at 4 zones whereas Zone 1: Sta. 2+050 – Sta. 2+376, Zone 2: Sta. 3+275, Zone 3: Sta. 3+400 – Sta. 3+860, and Zone 4: Sta. 3+750 and Sta. 6+497 – Sta. 7+333. Traffic loading for long term stability is taken as 15 kN/m2. .

3.1 Zone 1: STA. 2+050 - STA. 2+376 3.1 Zone 1: STA. 2+050 - STA. 2+376

Raising and stone works had been done at this zone. Potential problems in thiz zone are stability and settlement. Analysis of proposed design in this zone is based on the road cross-section at approximately Sta. 2+275.

Raising and stone works had been done at this zone. Potential problems in thiz zone are stability and settlement. Analysis of proposed design in this zone is based on the road cross-section at approximately Sta. 2+275.

Design parameters such as the friction angle, modulus of elasticity obtained by correlating the CPT #4 off shore and evaluation of laboratory tests on undisturbed soil samples in BH2 at Sta. 2+275.

Design parameters such as the friction angle, modulus of elasticity obtained by correlating the CPT #4 off shore and evaluation of laboratory tests on undisturbed soil samples in BH2 at Sta. 2+275.

The results of stability analysis of the existing road condition indicating the road is relatively in critical condition due to FS = 1.05. Predicted magnitude of settlement was 0.36 m at the center of the road, 0.41 m at the middle and of the road towards to Tanjung Priok, and 0.47 m at the edge of the road towards Tanjung Priok. The time required to reach the amount of settlement is 6800 days.

The results of stability analysis of the existing road condition indicating the road is relatively in critical condition due to FS = 1.05. Predicted magnitude of settlement was 0.36 m at the center of the road, 0.41 m at the middle and of the road towards to Tanjung Priok, and 0.47 m at the edge of the road towards Tanjung Priok. The time required to reach the amount of settlement is 6800 days.

Therefore, some proposed trial designs were analyzed to fulfill minimum FS as follow:


3.1.1. Corrugated prestressed concrete sheet piles 3.1.1. Corrugated prestressed concrete sheet piles The first trial design to strengthen the road is by using corrugated prestressed concrete sheet piles (see Table 1) to a depth of 16m (up to silty clay soil, firm to very firm).

The first trial design to strengthen the road is by using corrugated prestressed concrete sheet piles (see Table 1) to a depth of 16m (up to silty clay soil, firm to very firm). TTable 1. Design parameters of corrugated concrete type 5 able 1. Design parameters of corrugated concrete type 5

Sheet pile type

Sheet pile type Type Type

CrackingMoment(kN.m/m)

CrackingMoment(kN.m/m)

I I (m4) (m4)

A A (m2) (m2)

Corrugated concrete type 5

Elastic 269 3.5x10-3 0.1835

To simulate the corresponding field condition i.e. water level fluctuation, two finite element models were developed to evaluate the infuence of the highest and lowest water level on the road stability. The results of stability analysis indicated FS = 1.22 at the highest water level, while FS at the lowest water level is 1.16 with large moments working on the sheet pile.

Predicted magnitude of settlement was 0.36 m at the center of the road, 0.41 m at the middle and of the road towards to Tanjung Priok, and 0.47 m at the edge of the road towards Tanjung Priok. The time required to reach the amount of settlement is 6800 days.

3.1.2 Concrete sheet pile and ground anchor Strengthening of the road with a combination of concrete sheet piles (see Table 1) and ground anchor. The results of the analysis by considering the road reinforced with sheetpile (type 5) combined with additional ground anchors with an inclination from the horizontal of 40o. Giving prestressed ground anchor on presstress modeled with extreme force drawn from the total normal stress / FK, i.e. 91.15kPa / 3 = 30.38 kN/m/m. Extreme total normal stress obtained based on the forces acting on the parts that have the greatest displacement. The results of the analysis showed that the stability of the road which is reinforced with sheetpile and supplementary reinforcement combined with ground anchors provide FS = 1.3.


3.1.3 Secant pile walls Another strengthening method is by installing secant pile walls with diameter 0.8 m (see Table 2) to a depth of 25 m from the surface of the existing road. The results of stability analysis provides FS = 1.67.

1331


Table 2. Design parameter of secant pile walls (plate model)

Element Type EA(kN/m)

EI(kN.m2/m)

W(kN/m2)

n

Secant pile Elastic 1.2x107 3.6x105 14.4 0.15


Summary of settlement and time rate of 3 observation points as mentioned above can be seen in Table 3.

Table 3. Comparation of settlement value at Sta. 3+750.

Settlement ExistingRoad

Corrugated concrete Secant Piles

At center line 0.36 m 0.29 m 0.39 m

At middle 0.41 m 0.33 m 0.40 m

At edge 0.47 m 0.36 m 0.38 m

Time (days) 6800 6900 6500

3.2 Zone 2: Sta 2+376 – Sta. 3+401

In this Zone, there is a section of the road that had been strengthened by using sheet piles and stone works, while the other have not exist handling. Construction sheet pile implemented in 2011. This Zone is very vulnerable to both submerged from overflow of Japat River and from the surrounding environment (see Fig. 3), so raising works should be considered to be implemented.

Analysis of proposed design in this zone is based on the road cross-section at approximately Sta. 3+275. Design parameters such as the friction angle, modulus of elasticity obtained by correlating the CPT #5 on shore and CPT #17 off shore and evaluation of laboratory tests on undisturbed soil samples in BH6 at Sta. 3+150 and BH7 at Sta. 3+350.

The results of stability analysis of the existing road condition indicating the road is relatively in critical condition due to FS = 1.09. Predicted magnitude of settlement were 0.40 m at the center of the road, 0.51 m at the middle and of the road towards to Tanjung Priok, and 0.64 m at the edge of the road towards Tanjung Priok. The time required to reach the amount of settlement is 580 days.


3.2.1 Corrugated prestressed concrete sheet piles The first trial design to strengthen the road is by using corrugated prestressed concrete sheet piles (see Table 1) to a depth of 14 m (to sandy silt clay soil, firm to very firm).

To simulate the corresponding field condition i.e. water level fluctuation, two modeling analysis were conducted, which are to evaluate the influance of the highest water levels and the lowest water level measured on the tides. The results of stability analysis indicated FS = 1.15 at the highest water level, while FS at the lowest water level is 1.08 with large moments working on sheet piles 230 kN.m/m.

3.2.2 Corrugated prestressed concrete sheet piles + horizontal bars

Alternative proposed design is by using concrete sheet piles (see Table 1) and horizontal steel bars reinforcement (see Table 4) placed across the road which binds sheet piles with continuous slab constructed on the left side of the road. Stability analysis indicates FS = 1.4.

Table 4. Design parameter of horizontal steel bars

Steel grade Nominaldiameter

(mm)

Ultimatestress(MPa)

Crosssection

area (mm)

Ultimatestrength

(kN)

150 45 1035 1716 1779

Raising 0.7m should be performed at this Zone to prevent the road submerged from overflow of Japat River and from the surrounding environment. Stability analysis indicates FS after raising is 1.3 with tensile force on a horizontal rebar is 176.7 kN/m. Predicted magnitude of settlement were 0.40 m at the center of the road, 0.35 m at the middle and of the road towards to Tanjung Priok, and 0.25 m at the edge of the road towards Tanjung Priok. The time required to reach the amount of settlement is 677 days.

3.2.3 Concrete sheet pile and ground anchor Strengthening of the road with a combination of concrete sheet piles (see Table 1) and ground anchor. The results of the analysis by considering the road reinforced with sheetpile (type 5) combined with additional ground anchors with an inclination from the horizontal of 40o. Giving prestressed ground anchor on presstress modeled with extreme force drawn from the total normal stress / FK, i.e. 211kPa / 3 = 70.33 kN/m/m. Extreme total normal stress obtained based on the forces acting on the parts that have the greatest displacement. The results of the analysis showed that the stability of the road which is reinforced with sheetpile and supplementary reinforcement combined with ground anchors provide FS = 1.55.


3.3 Zone 3: Sta. 3+400 – Sta. 3+860

Analysis of proposed design in this zone is based on the road cross-section at approximately Sta. 3+750. There is a temporary construction using bamboo and stone masonry. This location is expected to potentially experience stability problems and settlement.

Design parameters such as the friction angle, modulus of elasticity obtained by correlating the CPT #9 on shore and evaluation of laboratory tests on undisturbed soil samples in BH9 at Sta. 3+750.

The results of stability analysis of the existing road condition indicating the road is relatively in critical condition due to FS = 1.02. Predicted value of settlement during 4600 days was 0.60 m at the centerline, 0.68 m in the middle of the road directions to Tanjung Priok, and 0.75 m at the edge of the road way to Tanjung Priok. The time required to reach settlement is 4600 days.

As mentioned above, the existing road condition at Sta. 3+750 is in relatively unstable condition due to FS value = 1.02. Therefore, some proposed alternative designs were analyzed to fulfill minimum FS as follow:

3.3.1 Concrete sheet piles and ground anchor Strengthening of the road with a combination of concrete sheet piles (see Table 1) and ground anchor. The results of the analysis by considering the road reinforced with sheetpile (type 5) combined with additional ground anchors with an inclination from the horizontal of 40o. Giving prestressed ground anchor on presstress modeled with extreme force drawn from the total normal stress / FK, ie 91.15kPa / 3 = 30.38 kN/m/m. Extreme total normal stress obtained based on the forces acting on the parts that have the greatest displacement. The results of the analysis showed that the stability of the road which is reinforced

1332


with sheetpile and supplementary reinforcement combined with ground anchors provide FS = 1.3.

3.3.2 Secant pile walls

-0.50

-0.45

-0.40

-0.35

-0.30

-0.25

-0.20

-0.15

-0.10

-0.05

0.000 50 100 150 200 250 300 350 400 450 500 550 600

Sett

lem

ent (

m)

Days

Another strengthening method is by installing secant pile walls with diameter 0.8 m (see Table 2) to a depth of 25 m from the surface of the existing road. The results of stability analysis provides FS = 1.65. The improvement of vertical geometry was also considered in this analysis. There are two raising level modeled in the models which are raising 0.7m consisting of ± 0.25 m selected material, ± 0.15 m subbase foundation and ± 0.3 m concrete pavement; and raising 1.2m consisting of 1m raising using lightweight material (see Table 5) and 0.2m asphalt. The FS of the road which strengthened by secant pile walls combined with raising 0.70 m and 1.20 m are 1.31 and 1.37, respectively.

Settlement at point A, raising 1.2m using selected material

Settlement at point B, raising 1.2m using selected material

Settlement at point C, raising 1.2m using selected material

Settlement at point A, raising 0.7 m using lightweight material

Settlement at point B, raising 0.7 m using lightweight material

Settlement at point C, raising 0.7 m using lightweight material

Table 5. Design parameter of raising material

Element TypeE

(kN/m2)c’

(kN/m2) ’ (deg) n

Lightweight Non-porous 3.0x104 50 45 0.15

Settlement analysis were performed at some points over the pavement surface for all raising cases, i.e. at road centerline, at the middle and the edge of the road directions to Tanjung Priok, respectively. The result of settlement analysis can be seen in Table 6.

Table 6. Predicted settlement value at Sta. 3+750.

Settlement Raising 0.7 m Raising 1.2 m

Center line 1.24 m 0.79 m

Middle 1.29 m 0.67 m

Edge 1.30 m 0.51 m

3.4 Zone 4: STA 6+497 - STA 7+333

The condition of this zone is similiar to that of in Zone 2 where is very vulnerable to both submerged from overflow of Japat River and from the surrounding environment, so raising works should be considered to be implemented. Analysis of proposed design in this zone is based on the road cross-section at approximately Sta. 7+000.

Design parameters such as the friction angle, modulus of elasticity obtained by correlating the CPT #13 on shore and CPT #37 off shore and evaluation of laboratory tests on undisturbed soil samples in BH13 at Sta. 7+000.

Based on analysis result, the existing road remains relative stable with FS = 1.95. Eventhough the road seems to be stable, a potential problem in this zone is flood. There are two raising level modeled in the models which are raising 0.7m consisting of ± 0.5 m selected material, and ± 0.2 m concrete pavement; and raising 1.2 m consisting of ± 1m selected material and 0.2m asphalt.

The results of stability analysis shows that the road is relatively stable at the time of the raising 0.7 m and 1.2 m, with FS = 1.54 and FS = 1.37, respectively. If lightweight material is used to replace selected material in both raising 0.7 m and 1.2 m, FS = 1.93 and FS = 1.56.

The results of settlement analysis show that the predicted settlement of point A at the center line of the road after raising 0.7 m and 1.2 m using selected material are 0.35 m and 0.47 m within 449 days and 508 days, respectively. If lightweight material is used in raising, the predicted settlement will be

0.27 m and 0.31 m within 420 days and 452 days, respectively (see Fig. 4).

Figure 4. Time – settlement curve of raising using selected material and lightweight material at point A, B and C at Sta. 7+000

4 CONCLUSIONS

Based on the observations of field conditions, soil test results, evaluation of the existing condition of the road, and analysis result, performance of road embankments over North Jakarta-soft soil can be summarized as follows : 1. To fulfil stability and settlement analysis, the road at Zone

1, and 3 should be strengthened by secant pile walls combined with raising 0.7 m and in some places raising 1.2 m should be implemented. If this strengthening method is applied at Zone 1 and Zone 3, FS is 1.67 and 1.31-1.37, respectively.

2. To fulfil stability and settlement analysis, the road at Zone 2 should be strengthened by concrete sheet piles and ground anchor. If this strengthening method is applied, FS is 1.55.

3. Although the results of the analysis of the stability of the existing road at Zone 4 shows that the road is still in a stable condition, a potential problem in this zone is flood. Therefore, raising 0.7 m and in some places raising 1.2 m should be implemented. If this strengthening method is applied, FS is 1.31-1.37.

5 ACKNOWLEDGEMENTS

The authors very gratefully acknowledge to Mr. Bambang Hartadi, Chief of Sub Directorate of Freeways and Urban Road, for giving all data used in this paper.

6 REFERENCES

Rahadian H., Hendarto and Prasetya B. 2011. The failure of a road embankment over north java soils. Geotechnical Engineering for Disaster Mitigation and Rehabilitation, Semarang.

Institute of Road Engineering, Ministry of Public Works. 2011. Laporan evaluasi teknik Jalan RE. Martadinata, Jakarta.

1333

Retrofit Technique for Asphalt Concrete Pavements after seismic damage

Technique de réhabilitation pour chaussée en béton d'asphalte après dommage sismique

Ohta H. Chuo University Research and Development Initiative

Ishigaki T. NIPPO Corporation Research Institute

Tatta N. Maeda Kosen Co.Ltd.

ABSTRACT: Reducing the risk of earthquake-induced damage to road is needed to promote safety and disaster mitigation andrecovery. It is strongly needed for pavement performance to keep the emergency traffic remain in service despite severe earthquake. This paper presents a retrofit technique of asphalt pavements using Confined-Reinforced Earth (CRE) consisting of 1) compacted soil, 2) geosynthetics and 3) post-tensioning anchors. Confining by the anchors is the application of both compressive and confining forceto the compacted soil layers, and gives a pre-tensile force to geosynthetics. The high flexural rigidity of CRE is for overcomingweakness of base course or subgrade in tension and flex/bending. In this paper, 1) structure of the retrofit technique of asphaltpavements using CRE, 2) construction method, 3) the results of full scale in-situ tests are presented.

RÉSUMÉ : La réduction du risque de dommage routier induit par un tremblement de terre est nécessaire pour la sécurité, l'atténuation de l’effet de la catastrophe et la remise en service. La performance du pavement pour permettre le trafic d'urgence qui doit rester en service est visée. L’article présente une technique de rehabilitation des chaussées d’asphalte à l’aide d’un sol renforcé confiné (CRE) par 1) un sol compacté, 2) géosynthétiques, et 3) post-pension d’ancrages. Les ancrages en acier rigide sont placés verticalement du haut vers la couche de base et verrouillé à la base des géosynthétiques. Le confinement des sols compactés s’effectue par l'application des deux forces (de compression axiale et latérale) qui applique une pré-tension aux géosynthétiques. La granderigidité à la flexion de CRE contre balance la faiblesse de la couche de base ou de la couche de fondation en traction et flexion.L’article présente 1) la structure de chaussées, 2) la méthode de construction, et 3) les résultats des essais « in situ » pleine échelle.

KEYWORDS: pavement, earthquake, seismic retrofit, confined-reinforced earth, geosynthetics.

1 INTRODUCTION.

Reducing the risk of earthquake-induced damage to road is strongly recquired to promote safety, disaster mitigation and recovery. Road pavements which are adjacent to highway structures such as bridges and culvert boxes are often damaged due to the differential settlement of highway embankments around bridge abutments, edge of culvert boxes and wing walls during and after severe earthquakes. Traffic is easily intercepted by the earthquake damage to road pavements. In one of the precept of the Great East Japan earthquake (2011), to keep the emergency traffic remain in service after severe earthquake is the most important subject especially for emergency activity.

This paper presents a newly developed seismic retrofit technique of asphalt concrete pavements using Confined-Reinforced Earth (CRE). CRE is composed by compacted soil, geosynthetics and post-tensioning rigid anchors. Confining by the anchors is the application of both compressive and confining force to the compacted soil layers, and gives a pre-tensile force to geosynthetics.

The basic idea of implementing pre-stresses in reinforced earth was advanced technology even from a gloval perspective as follows (Uchimura et al. 1996, 2003, 2005). The high rigidity of CRE is apparently useful in preventing excessive differential settlement of the road pavement despite severe earthquake.

In this paper, 1) structure of the seismic retrofit technique of asphalt pavements using CRE, 2) construction method, 3) application of actual highway embankment, 4) the results of full scale in-situ tests are presented.

2 STRUCTURE

Figure 1 shows the structure of CRE applied to road subgrade of asphalt concrete pavement. CRE is a composite structure

consisting of compacted crushed stone, geosynthetics and post-tensioning anchors. The high flexural rigidity of CRE is for overcoming weakness of base course or subgrade intension and bending.

Figure 1. Structure of Confined-Reinforced Earth applied to asphalt concrete pavement.

Compacted soil is a key material for use in CRE. Selection of soil material is very important to keep the reinforced performance of CRE. Crushed stone for mechanical stabilization is the best material due to the high compression and shear strength. High degree of compaction is also effective to the reinforced effect of CRE. The crusshed stone layers are sandwitched by four layers of geogrid and confined by confining rigid steel anchors.

Photograph 1 (1) shows geosyntheyics used in this method. The geosynthetics for use in CRE has high tensile strength of 200kN/m with low strain of 4.5%. The width of a sheet of geosynthetics is maintain as same as road lane width.

1334


Photograph 1 (2) shows the newly developed post-tensioning rigid steel anchor used in CRE. This anchor was improved from a slope reinforcement anchor. The anchors is vertically penetrated from the top to the bottom layer and locked to the lower geosynthetics.

Photographs 1. Geosynthetics and confining rigid steel anchors using Confined-Reinforced Earth.

3 CONSTRUCTION METHOD

Photographs 2 show the construction sequence of CRE. After preparing the lower subgrade, the 1st layer of geosynthetics is laid on the area to be reinforced (Photograph 2 (1)).

Then crushed stones are carefully laid by a bulldozer or by a motor grader (Photograph 2 (2)) on the 1st layer of geosynthetics and the layer of crushed stone is fully compacted by vibrating rollers (Photograph 2 (2)). After placing three layers of compacted crushed stone and laying of 2nd ,3rd and 4th

layers of geosynthetics (Photograph 5 (3)), rigid steel anchors are vertically penetrated from the top to the bottom layer by the small pile driving equipment (Photograph 2 (4)) and mechanically locked to the lower geosynthetics employing a small hydraulic jack ( Photograph 2 (5)).

It should be noted that the construction time of setting anchors is very short (about 40 to 50 anchors per hour). Finally, a top steel plate is set through a rod and is fixed to the rod

Photographs 2. Construction method of Confined-Reinforced Earth. (Construction in Joban Highway in Fukushima, Japan, 2011)

with a nut using a torque wrench (Photograph 2 (6)). Confining load of 30kN can be exactly maintained by setting torque. By use of this construction method, rapid construction of CRE becomes possible making it practical enough to apply this CRE for road in service. Photograph 3 shows the application of CRE for seismic retrofit of asphalt pavement on the actual highway

embankment in Joban Highway, Fukushima, Japan, constructed in 2011.

Photograph 3. Application of Confined-Reinforced Earth for actual highway embankment in Joban Highway, Fukushima, Japan, 2011.

4 FULL SCALE IN-SITU TEST

4.1 Trial embankment

A full-scale test of this high rigidity reinforced earth was carried out in the field in Ibaraki, Japan, from 9th to 16th March 2011 as shown in Photograph 4. The constructed trial embankment was of 25m length, 4m width and 2.5m height at the top of embankment. Full-scale asphalt pavements were placed on the trial embankment. The asphalt concrete pavement consisted of asphalt concrete of 50mm thickness and base course of 300mm thickness. Two types of asphalt pavement were constructed. The first type was conventional asphalt concrete pavement placed on the compacted soil subgrade, while the second was asphalt pavement placed on the high rigidity confined-reinforced earth consisting of the crushed stone sandwiched by four layers of geosynthetics and confined by confining rigid anchors.

We aimed at direct comparison of performance of the two pavement types by artificially generating the differential settlement of trial embankment such as often seen during severe earthquakes. The forced differential settlement of the embankment was realized by using 10 multi-controlled large hydraulic jacks supporting the steel deck (10m long) placed under the embankment body. The layout of the trial embankment is shown in Figure 2.

Photograph 4. Trial embankment after testing of 550mm differential settlement. (Confined –Reinforced Earth tested in Ibaraki, Japan, 2011)

1335


Figure 2. Layout of the trial embankment (Constructed and tested in Ibaraki, Japan, 2011)

- 2 0 2 4 6- 600

- 500

- 400

- 300

- 200

- 100

0

Distance（m）

Roa

d he

ight（

mm）

2550

100

150

Settlement (mm)

200

250

300

350

400

450

500550

Settlement of 250mm○ ：Before Great East

Japan Earthquake (9th March 2011)

● ： After Great East Japan Earthquake (16th March 2011)

Photographs 5. Deformation performance of Confined–Reinforced Earth (550mm differential settlement, tested in Ibaraki, Japan, 2011).

4.2 Deformation performance

Trial embankment after testing looks as shown in Photographs 5. It is clearly observed that the central part of road is artificially lowered by 550mm. Experimental results indicate that use of the newly developed high rigidity confined-reinforced earth (CRE) can maintain the minimum road usability for vehicles even when the amount of differential settlement reaches approximately 600 mm.

Although the asphalt pavement settles together with the embankment, there is no crack or sharp gap in the CRE pavement shown in Photograph 5 (2). Instead, the CRE asphalt pavement settled in a gentle curve as shown in Figure 3. Conversely, in the case of conventional asphalt pavement, the asphalt pavement splits resulting in a step-like gap that started to appear when the forced settlement reached approximately 200 mm. The gap at the stage of differential settlement of 550 mm is shown in Photograph 5 (3).

Figure 3. Surface deformation profile of asphalt concrete pavement reinforced by Confined-Reinforced Earth.Figure 3. Surface deformation profile of asphalt concrete pavement reinforced by Confined-Reinforced Earth.

4.3 Seismic performance 4.3 Seismic performance

On 11th March, 2011, the Great East Japan Earthquake ofMagnitude 9 occurred during the test when the settlement was 250mm. About 200gal of seismic power acted on the trial embankment. Deformation of CRE being compared with conventional pavement before and after the earthquake is shown in Photographs 6. Figure 3 also shows that excessive deformation of CRE did not occur after the earthquake.

On 11th March, 2011, the Great East Japan Earthquake of Magnitude 9 occurred during the test when the settlement was 250mm. About 200gal of seismic power acted on the trial embankment. Deformation of CRE being compared with conventional pavement before and after the earthquake is shown in Photographs 6. Figure 3 also shows that excessive deformation of CRE did not occur after the earthquake.

There was no additional crack or gap in the CRE pavementafter the earthquake as seen in Photographs 6. Conversely, in the case of conventional asphalt pavement, the cracks of asphalt pavement further opened after the earthquake.

There was no additional crack or gap in the CRE pavementafter the earthquake as seen in Photographs 6. Conversely, in the case of conventional asphalt pavement, the cracks of asphalt pavement further opened after the earthquake.

Photograph 7 shows the perfect soundness of CRE as seen in the structure of the asphalt pavement after the earthquake. Only small amount of additional deformation of the CRE structure was observed after the severe earthquake as shown in Figure 3.

Photograph 7 shows the perfect soundness of CRE as seen in the structure of the asphalt pavement after the earthquake. Only small amount of additional deformation of the CRE structure was observed after the severe earthquake as shown in Figure 3.

4.4 Traficability 4.4 Traficability

Photographs 8 show the trafficability tests of asphalt concrete pavement reinforced by CRE after 550mm differential settlement. The longitudinal curve of CRE is gentle enough for all types of vehicles to drive at low speed. Conversely, in the case of conventional asphalt pavement, it is obviously impossible for vehicles to drive in such conditions.

Photographs 8 show the trafficability tests of asphalt concrete pavement reinforced by CRE after 550mm differential settlement. The longitudinal curve of CRE is gentle enough for all types of vehicles to drive at low speed. Conversely, in the case of conventional asphalt pavement, it is obviously impossible for vehicles to drive in such conditions. The authors consider that the high rigidity subgrade using the confined-reinforced earth can contribute to the construction of The authors consider that the high rigidity subgrade using the confined-reinforced earth can contribute to the construction of

1336


Photographs 6. Seismic performance of Confined–Reinforced Earth (Tested in Ibaraki, Japan, 2011).

ter the Great East Japan Earthquake (550mm ifferential settlement).

t excessive differential settlement of road embankments placed in vicinity to box

butments.

einforced earth str

, as well as for emergency restoration of plant functionality and other aspects of

ns (BCP).

Ts

it for Geotechnical Engineering in Practice of the enter for Research and Development Initiatives, Chuo

Ishi

Oht

th

Uch

th

Uch

Photograph 7. Longitudinal cross section of Confined–Reinforced Earth and asphalt pavement afd

Photographs 8. Traficability tests of Confined-Reinforced Earth

safer and more durable roads and can be used as an anti-earthquake measure to prepare roads for damage from major earthquakes as well as a measure agains

culverts and/or bridge a

5 CONCLUSIONS

The structure of the seismic retrofit technique for asphalt concrete pavements using Confined-Reinforced Earth (CRE), construction method and the results of full scale in-situ tests were described. Full scale in-situ tests show the acceptable performance of CRE after the forced settlement to simulate the severe earthquake-induced damage.

After experiencing the acceptable performance of high rigidity confined-reinforced earth observed at trial embankment, we still had two major technical problems related to (1) relaxation of pre-stresses and (2) materials unfavourable to be used. Apparently the degree of time dependency of the relaxation of pre-stresses depends on the kind of compacted material and degree of compaction. Distinct definition and complete rejection of unfavourable materials are also unavoidable factors to guarantee successful performance. To avoid the difficulties arising from these two problems, the authors have decided to use only crushed stones (and crushed concrete) traditionally used as the base course materials. The authors have also decided not to make the design of confined-reinforced earth too much relying on the high level of pre-stresses being implemented, i.e., the authors intend to make design procedure that requires just confinement on the movement of crushed stone particles and requires moderate pre-stressing. In the course of developing reinforced earth technology, the authors have tried to find out practical methods of specifying the material parameters of compacted materials and compacted crushed stones used in the r

uctures. Some of these methods are concisely summarized by Ohta et al. (2007) and by Ishigaki et al. (2008).

The seismic retrofit technique for asphalt concrete pavements using CRE gives minimum functionality of roads for vehicle access which is essential for initial emergency response such as lifesaving and firefighting activities

corporate Business Continuity Pla

6 ACKNOWLEDGEMENTS

The authors greately indebted to Mr. M. Kondou (East Nippon Expressway Co. Ltd.), Dr. M. Tonogaito (West Nippon Expressway Co. Ltd.), Mr. H. Yamauchi & Mr. T. Suzuki (NIPPO Corporation Co. Ltd.), Dr. S. Omoto (NIPPO Corporation Research Institute), Mr. T. Nishimoto & Dr. S.

uji (Maeda Kosen Co. Ltd.) and Prof. A. Iizuka (Kobe University) for their encouragement and their helpful suggestion.

The asphalt pavement using confined-reinforced earth (CRE) was developed through a joint industry-academia research with Maeda Kosen and NIPPO Corporation conducted at the Research UnCUniversity.

7 REFERENCES

gaki, T., Watanabe, S., Omoto, S. and Ohta, H. 2008. Constant volume direct shear behavior of statically compacted granular materials, Journal of Pavement Engineering, JSCE, Vol. 13. 115-123. (in Japanese)

a, H., Yoshikoshi, H., Uchita, Y., Ishiguro, T. and Hayashi, Y. 2007. Geo-material characterization in simulating the performance of large dams under construction, Proc. 16 Southeast Asian Geotechnical Conference, Edited by K. Yee, Ooi TeikAun, Ting Wen Hui and Chan Sin Fatt, 39-50. imura, T., Tatsuoka, F., Tateyama, M., Koseki, J., Maeda, T. and Tsuru, H. 1996.Mechanisms, element tests, full-scale model tests and construction of preloaded and prestressed geothinthetic reinforced soil structure, Proc. 11 Geosynthetics Symposium, Japanese Chapter of the International Geosynthetics Society, 72-81. (in Japanese) imura, T., Tateyama, M., Tanaka, I. and Tatsuoka, F. 2003. UchPerformance of a preloaded-prestressed geogrid-reinforced soil pier for a railway bridge, Soils and Foundations, Vol. 43, No.6, 155-171 imura, T., Tamura, Y., Tateyama, M., Tanaka,I. and Tatsuoka, F. 2005. Vertical and horizontal loading tests on full-scale preloaded and prestressesd geogrid-reinforced soil structures, Soils And Foundations, Vol. 45, No. 6, 75-88.

1337

Simultaneous interpretation of CPT/DMT tests to ground characterisation

L'interprétation simultanée des essais CPT/DMT pour la caractérisation du sol

Rabarijoely S., Garbulewski K. Department of Geotechnical Engineering, Warsaw University of Life Sciences, Poland

ABSTRACT: This paper addresses the simultaneous interpretation of two well known in situ tests, namely CPT and DMT, to characterize the geotechnical conditions, particularly the stress history of clay sediments. The CPT/DMT tests had been carried out in order to recognize the geotechnical conditions in the foundations of design buildings at SGGW (WULS) Campus in Warsaw. The datafor two layers of glacial boulder clays have been analysed simultaneously using CPT and DMT tests. Based on the results of analyses new formulas for determination of the OCR have been developed. The suggested relationships should be considered in future analyses for their improvement.

RÉSUMÉ : Cet article traite de l'interprétation simultanée de deux essais in situ bien connus, à savoir CPT et DMT, pour caractériser les conditions géotechniques, en particulier l'histoire des contraintes de sédiments argileux. Les essais CPT/DMT ont été réalisés dans le but de reconnaître les conditions géotechniques pour le dimensionnement des fondations des bâtiments SGGW (WULS) Campus à Varsovie. Les données de deux couches d'argiles glaciaires ont été analysées simultanément en utilisant les essais CPT et DMT. Sur la base des résultats des analyses, de nouvelles formules pour la détermination de l'OCR ont été développées. Les relations proposées devraient être pris en compte dans les analyses futures en vue de leur amélioration.

KEYWORDS: ground characterisation, CPT/DMT tests, join interpretation

1 INTRODUCTION

In engineering practice the cone penetration and dilatometer of Marchetti tests (see Figure 1) are commonly recommended to identify the geotechnical conditions for design structures (Lutenegger and Kabir 1988, Briaud and Miran. 1991, Lunne et al. 1997, Marchetti 1980). Although the CPT/DMT tests have been used for over 30 years in the same purposes, generally to recognize geotechnical condition in ground (Robertson 1990, Młynarek 2007), to date relatively little published regarding join interpretation (Mayne and Liao 2004, Robertson 2009). One of not numerous comparison between measured DMT parameters (ID, KD, and ED) with depth and predicted parameters using the CPT at Moss Landing site (California) is showing in figure 2 (Robertson 2009). This site provides a good test for the proposed correlations since the soils range from soft to firm clay and loose to dense sand. The site is composed of about 2.6 m of silty sand to silt over about 4.4 m of sand. Below the sand is a deposit of firm plastic clay extending to a depth of 13.4 m. The ground water level is at a depth of about 2.2 m below ground surface but fluctuates somewhat with the tide. After an analysis of CPT/DMT correlations Robertson concluded that horizontal earth pressure index KD from DMT tests correlate to normalized cone resistance (qc - σv0)/σ’v0 and proposed formula as follows:

80

80.

'vo

vocD

q.K

(1)

Knowing the horizontal index KD from DMT test and normalised cone resistance, the overconsolidation ratio OCR can be obtained using following formulas (Marchetti 1980, Robertson 2009):

561.

DK 0.5OCR (2)

251240

.'vo

votq.OCR

(3)

According to Robertson (2009) the resistance of cone qc can be determined based on the following equation:

vo1.25D

'voc K.q 251 (4)

This paper presents the CPT and DMT tests carried out in order to recognize the geotechnical conditions for the design structures in frame of the SGGW (WULS) Campus development. Based on the results obtained from the extensive research (Rabarijoely et al. 2011) the relationships between CPT and DMT tests have been developed. For the determination of the cone resistance qc new procedure was developed considering dilatometer index of horizontal stress KD. Also, for determination of KD index new procedure was proposed based on the cone resistance qc. Moreover, the simultaneous interpretation of CPT/DMT tests was applied to determine the overconsolidation ratio of clay sediments.

a) b)

Figure 1. View of CPTu tip (a) and DMT blade (b) actually used in geotechnical investigation

1338


Figure 2. Comparison between measured DMT parameters and predicted parameters using the CPT at Moss Landing site (Robertson 2009)

2 CPT/DMT TESTS AT SGGW CAMPUS

In order to determine geotechnical conditions in the foundations of design buildings at SGGW Campus a total of 69 of CPT and DMT tests were conducted (see Figure 3). Analysing data gathered in the Ground Investigation Report, five geotechnical layers were identified in the Campus test site (see Figure 4), including a layer of brown glacial boulder clay noted in this paper as layer No. III (acc. to geotechnical classification sandy clay - saCL and sasiCL) of the Warta glaciation (gQpW), for which liquidity index values IL = (0.0÷0.11) and a layer of grey glacial boulder clay of the Odra glaciation (gQpO), sandy clay with boulders as layer No. IV, for which IL = (0.0÷0.12). The layers III and IV were pointed out as layers with suitable geotechnical conditions for foundation of the Campus buildings. Typical distributions of the cone resistance qc from CPT tests and the horizontal stress index KD from DMT tests for III and IV layers (boulder clay sediments) are shown in figure 5. Relationships between measured values of qc and KD using CPT and DMT tests respectively are shown in figure 6. These relationships were obtained using statistical analysis (Solver modulus).

Figure 3. Map of the SGGW Campus with locations of CPT (▼) and DMT (■) tests (Rabarijoely et al. 2011)

Figure 4. Typical geological conditions at the SGGW (WULS) Campus

Figure 5. Profiles of cone resistance qc from CPT tests and horizontal stress index KD from DMT tests for III and IV layers of SGGW (WULS) Campus

0,08

0,10

0,12

0,14

0,16

0,18

0,20

s'vo[MPa]

810

1214

1618

2022

2426

28

KD_Test[-]

4

6

8

10

12

14

16

18

20

qc_obl [-]= f( KD)

Figure 6. Comparison between calculated qc (eq. 8) and KD measured by DMT tests (’vo - effective vertical stress in boulder clay layer)

1339


3 CPT/DMT CORRELATIONS FOR OCR DETERMINATION

It is obvious that the values of overconsolidation ratio determined from both CPT and DMT tests should be the same. Therefore, the following equation is valid:

56150240 .D

1,25

vo

voc K.σ'σq.

(5)

In order to determine CPT/DMT correlations the statistical analysis (Solver modulus) to obtain the best fitting between calculated according to equation (1) and measured values of KD in the SGGW Campus was applied. From statistical analysis the following relationship was estimated with Mean and Maximum Relative Deviations respectively MSRD=9,0%, MRD=20,0%:

0,40

vo

vocD σ'

σq,1K

2 (6)

The similar statistical analysis was carried out for obtaining the best fitting of qc distribution in profile of the SGGW Campus between calculated according to eq. (6) and measured values. Equation (6) was rearranged to determine the qc values. The following relationship was determined with MSRD=20,0%, MRD=30,0%:

vo2,0Dvoc σKσ'0,45q (7)

Finally, introducing to formula (eq.1) the proposed relationships (eq.6 and eq.7) the overcondsolidation ratio can be calculated using the following formulas:

0.82

vo

voc

σ'σq.0CRO

28 (8)

2148 .DK.0CRO (9)

Comparison between KD measured in the foundation of building No 34 in SGGW Campus and calculated using equations (eq.1) and (6) is presented in figure 7, whereas between qc measured and calculated according to equation 7 in figure 8. The values of KD and qc calculated according to eq. (6) (7) are similar to measured. The distribution of OCR values in profile analysed is shown in figure 9.

4 FINAL CONCLUSIONS

The objective of this paper is to compare the results of CPT and DMT tests obtained for boulder clay distinguished in SGGW Campus at Warsaw. Based on the statistical analyses of 69 CPT/DMT profiles the formulas for the cone resistance qc as a function of horizontal stress index KD and for KD as a function of qc were suggested. Moreover, the new formulas for determination of overconsolidation ratio are also proposed.

In general, the formulas proposed for boulder clay in foundation of SGGW Campus give smaller values of OCR to comparison with Robertson’s proposal.

The suggested relationships should be considered in future analyses for their improvement.

Figure Profile of KD under SGGW building No 34 in Warsaw SGGW Campus

Figure Profile of qc under SGGW building No 34 in Warsaw SGGW Campus

1340


6 REFERENCES

Briaud J., Miran J. 1991: The flat dilatometer test, TX, 77843-3136 USA for The Federal Higway Administration.

Rabarijoely S., Garbulewski K., Rajtar J., Jabłonowski S. 2011: Geotechnical mapping of the SGGW Campus in Warsaw applying the Bayesian approach. Proc. of XV European Conference on SMGE, Athens, Greece, Part 1, p. 447-452.

Marchetti S. 1980: In situ Tests by Flat Dilatometer. Journal Geotechnical Engineering Division, ASCE 106, GT3: p. 299-321.

Lutenegger A. J., Kabir M.G. 1988: Dilatometer C-reading to help determine stratigraphy. Proc. Int. Sym. on Penetration Testing ISOPT-1, Orlando, 1: p. 549-553.

Robertson P. 1990: Soil classification using the cone penetration test. Canadian Geotechnical Journal 27(1), p. 151-158.

Lunne T., Robertson P.K, Powell J.J.M. 1997: Cone Penetration Testing in Geotechnical Practice; London.

Mayne P. W., Liao T. 2004: CPT-DMT interrelationship in Piedmont residium. Proc., ISC’2, Vol. 1, Millpress, Rotterdam, The Netherlands. p. 345-350.

Młynarek Z. 2007: Site investigation and mapping in urban area. Proc. XIV European Conference on Soil Mechanics and Geotechnical Engineering, Rotterdam Vol. 1. p. 175 – 202.

Robertson P.K. 2009: CPT-DMT Correlations. Journal of Geotechnical and Geoenvironmental Engineering, Vol. 135, No 11. p. 1762-1771.

Figure. Profile of OCR under SGGW building No 34 in Warsaw SGGW Campus

5 ACKNOWLEDGEMENT

This research was supported by two Grants: (1) N N506 218039 and (2) UMO-2011/03/D/ST8/04309 from the National Science Centre, Kraków, Poland.

1341

Modélisation numérique 3D d’un système de fondation d’un complexe immobilier

3D numerical modeling of a foundation system of a building complex

Reynaud S., Allagnat D., Mazaré B. EGIS Géotechnique, Grenoble, France

Julien T. EIFFAGE CONSTRUCTION, Clermont-Ferrand, France

RÉSUMÉ : Pour la construction d’un nouveau complexe immobilier au centre-ville de Clermont-Ferrand, la société EIFFAGE CONSTRUCTION a confié à EGIS Géotechnique les études d’exécution du système de fondation. Le projet est situé sur une ancienne cheminée volcanique, dans des silts lâches de forte épaisseur. Le système de fondation consiste en un radier général fondésur un sol renforcé par des colonnes de transfert de charges en jet grouting, elles-mêmes reliées à un bouchon général injecté. Conformément au souhait du client, le système de fondation a été modélisé de façon numérique en 3 dimensions, comprenant chaquecolonne de transfert. Le logiciel FLAC 3D a été utilisé pour constituer ce modèle. La première difficulté est de parvenir à constituer un modèle numérique viable et à maîtriser les temps de calcul. La seconde difficulté est de modéliser correctement le comportementmécanique du système, notamment des colonnes de transfert et de leur liaison avec le radier. Plusieurs itérations ont été nécessaires pour parvenir à un couplage correct entre la modélisation de la structure et celle du sol (sol+fondations+radier). L’objectif de cettemodélisation est de vérifier les tassements, les contraintes dans les éléments modélisés et les efforts dans les colonnes de transfert.

ABSTRACT: For the construction of a new buildings complex in the city center of Clermont-Ferrand, EIFFAGE CONSTRUCTIONsociety entrusted to EGIS Géotechnique the design studies for the execution of the foundation system. The project is located on an ancient volcanic chimney, within loose thick silts. The system consists of a foundation raft constructed upon a soil reinforced by jet grouted columns for load transfer; the columns are interconnected within a jet grouted bedding layer. According to the client’s needs, the foundation system was numerically modeled in three dimensions, with each load transfer columns. The software FLAC 3D was used for this model. The first challenge is to establish a viable numerical model and to control the computation time. The second challenge is to correctly model the mechanical behaviour of the system, notably the load transfer columns and connections with the foundation raft. Several iterations are necessary to achieve a good correlation between the structure and soil models (soil+foundations+foundation raft). The objective of this modelisation is to verify the settlements, the stresses within the modeledelements and the forces within the load transfer columns.

MOTS-CLES : Modélisation numérique 3D, couplage des modèles, fondations, sol renforcé, jet grouting,

KEYWORDS: 3D numerical modelisation, correlation of models, foundations, reinforced soil, jet grouting, 1 INTRODUCTION

Pour la construction d’un nouveau complexe immobilier situé au centre-ville de Clermont-Ferrand, à proximité de la place de Jaude, EIFFAGE Immobilier (promoteur) et EIFFAGE Construction Auvergne (constructeur) ont confié à EGIS Géotechnique les études d’exécution du système de fondation de ce complexe. L’objet de cette étude est la modélisation complète du système de fondation. Ce complexe est composé de bureaux, de bâtiments d’habitation, d’un centre commercial, d’un cinéma, d’un hôtel et de deux niveaux de sous-sol, sur une surface de plus d’un hectare (11 500 m²).

La particularité du site étudié est sa nature géologique : ancienne cheminée volcanique ayant entaillé le substratum marneux et marno-calcaire de l’Oligocène, puis ayant été comblée de sédiments lacustres voire fluviatiles du type silt, argile molle et sable de très faible compacité. Cette particularité a conduit les concepteurs à opter pour un renforcement général du sol par colonnes sécantes de jet grouting, constituant un bouchon, disposé au-dessous du radier. Ce bouchon injecté est connecté au radier par des colonnes de transfert de charges en jet grouting selon un maillage particulier.

L’objectif des études d’exécution est d’optimiser le système de fondation tout en garantissant son fonctionnement en statique et en dynamique (sous chargement sismique). Dans ce cadre, Egis Géotechnique a proposé une modélisation numérique 3D avec le logiciel FLAC 3D, permettant la modélisation du

renforcement général et de chaque colonne de transfert (1650 unités au total).

Les études d’exécution ont repris les hypothèses géotechniques et les principaux résultats d’une étude de projet (de type G2) réalisée par la société FUGRO. Les études d’exécution ont été menées en concertation avec le bureau d’études structure SIGMA et le bureau de contrôle VERITAS. L’élaboration des modèles a été réalisée avec l’appui de la société ITASCA (commercialisant FLAC 3D).

Sont présentées dans le présent article les principales hypothèses géomécaniques, la méthodologie de calcul mise en œuvre et quelques illustrations des résultats.

2 CONTEXTE

2.1 Nature du sous-sol

La succession des terrains est la suivante : des remblais divers et des limons sur une épaisseur de 3 m environ, de compacité faible en tête puis la présence de calcaires travertinisés avec des pics de compacité dont le niveau correspond avec celui de la nappe. Ensuite sont rencontrés des silts de compacité plutôt faible sur une épaisseur de l’ordre de 5m, puis des sables assez compacts sur une épaisseur de l’ordre de 10 m. Enfin, sont rencontrés des silts sur une forte épaisseur (de plasticité moyenne à forte), de compacité faible

1342


mais croissante avec la profondeur. Le niveau moyen de la nappe se situe à 3m de profondeur.

Figure 1. Coupe du sous-sol (extrait étude G2, FUGRO).

2.2 Système de fondation

Le système de fondation est composé d’un « bouchon injecté » d’une épaisseur variable entre 2 et 3m en fonction de la profondeur du radier et de son épaisseur (4 zones définissent le radier). Il est constitué de colonnes de jet grouting sécantes. Sa profondeur et son épaisseur sont fixées de manière à assurer la stabilité provisoire à la sous-pression hydraulique (état-limite UPL selon les Eurocodes 7) : l’arase supérieure du bouchon est placée entre 3.0 et 3.2m sous le radier. Le bouchon est relié au radier par des colonnes de transfert de charge en jet grouting de 1.2m et 1.4m de diamètre disposées selon une maille régulière en partie courante et selon une maille renforcée dite en « marguerite » au droit des porteurs (voiles et poteaux).

Figure 2. Extrait du plan d’implantation des colonnes de transfert

Le système de fondation sera analysé dans un premier temps avec des modèles locaux exploratoires comprenant quelques colonnes seulement, puis par un modèle global intégrant toute la superficie du bâtiment. C’est à partir du modèle global que le dimensionnement final a été réalisé.

2.3 Modèle géo-mécanique

Le modèle géo-mécanique retenu est présenté dans le tableau 1.

T ableau 1. Modèle géo-mécanique du remblai, silt et sable

Paramètres Remblai Silt Sable

γsat (kN/m3) 20.1 18.2 20.1

φ’(°) 25 25 33

c’ (kPa) 2 0 0

Modèle élastique linéaire :

Ey (kPa) 5 000 1 800 20 000

υ 0.3 0.3 0.3

Angle de dilatance ψ (°) - - 3

Modèle de sol avec écrouissage (Hardening soft-soil model) :

Module d’élasticité volumique, K (kPa) - - 16 700

Contrainte moy. max. pc (kPa) - - 60 / 500

Déformation volumique à la contrainte pc (%)

- - 0 / 2.6

Kmax = 3.k (kPa) - - 50 000

RadierRemblai

Colonnes de transfert

Silt

Bouchon injecté

Sable

Silt profond

Concernant les silts profonds, il a été retenu le modèle de

Roscoe, « Cam-Clay » modifié, adapté pour les limons plastiques à très plastiques, normalement consolidés. Les paramètres de ce modèle ont été calés sur la base d’essais en laboratoire (essais oedométriques) et d’essais in situ (pressiométrique et prénétrométrique). Deux hypothèses dites « raisonnables » et représentatives de la variation des mesures ont été envisagées ; elles sont présentées dans le tableau 2. T ableau 2. Modèle géo-mécanique des silts profonds

Paramètres Silts profonds (hyp.1)

Silts profonds (hyp.2)

γsat (kN/m3) 18.2 18.2

φ’(°) / M = 6.sinφ/(3-sinφ) 25 / 0.984 25 / 0.984

c’ (kPa) 2 2

Ko 0.58 0.58

Modèle Cam-Clay :

Cc / λ 0.7 / 0.304 0.7 / 0.304

Cs / κ 0.06 / 0.026 0.07 / 0.030

υ 0.3 0.3

OCR_max (cote 375.5) * OCR_min (cote 350.0)*

1.30 1.05

1.25 1.00

*Cote TN : 385.5 NGF Aucun phénomène de consolidation secondaire n’a été pris

en compte, les études de projet ayant montré que ces terrains ne sont pas sujets au fluage dans le domaine des contraintes appliquées.

2.4 Mode de représentation des colonnes de jet grouting sous FLAC3D et caractéristiques mécaniques

Le bouchon injecté a été représenté comme une couche de sol améliorée (éléments volumiques). Les colonnes de transfert ont été représentées sous forme d’éléments volumiques pour les modèles locaux et d’éléments structurels pour le modèle global.

Les caractéristiques mécaniques des colonnes de jet grouting dans les conditions statiques (colonnes de transfert et bouchon injecté) sont indiquées dans le tableau 3.

Les contraintes admissibles sont indiquées dans le tableau 4. Il n’est admis aucune résistance à la traction dans les

éléments de jet grouting.

1343


T ableau 3. Caractéristiques mécaniques des colonnes de jet-grouting

Paramètres Colonne de jet grouting

γsat (kN/m3) 18

fc (MPa) 6

Ey (MPa) 2 000

υ 0.2

T ableau 4. Critères mécaniques pour les colonnes de jet-grouting

Paramètres Sous G SousG+0.3Q+E

Contrainte moyenne 0.30fc -

Contrainte maximale 0.60fc 0.85fc/1.15

Section comprimée 100% 10%

Tension de coupure (bouchon injecté) 0 0 Pour le modèle global, la condition de liaison entre les

colonnes de transfert et le radier a fait l’objet d’un développement particulier du modèle. Afin de simuler un encastrement non « parfait » aux extrémités des colonnes de transfert, une souplesse est introduite sous forme d’une rotule plastique : le moment dans la colonne est alors borné par la valeur du moment plastique. La prise en compte d’une telle rotule nécessite des calculs itératifs car le moment plastique est fonction de la contrainte axiale. Le processus a été le suivant : première itération sans rotule (encastrement) pour évaluer les contraintes axiales et les moments plastiques correspondants, seconde itération avec les rotules plastiques pour vérification du respect des contraintes admissibles pour chaque colonne et le cas échéant, définition de nouvelles rotules plastiques, puis nouvelles itérations jusqu’à vérification complète.

L’objectif de ces calculs est de vérifier que l’introduction des moments plastiques ne perturbent ni l’équilibre global du système de fondation (analyse des déplacements), ni les contraintes dans le bouchon.

3 MODÈLES LOCAUX EXPLORATOIRES – CALAGE DU MODÈLE

Trois modèles « locaux » exploratoires ont permis d’analyser, pour des cas de chargements représentatifs, la répartition des charges dans le bouchon injecté via les colonnes de transfert. L’intérêt de ces modèles est de présenter une taille limitée et par conséquent d’être de résolution rapide. L’intérêt est également de valider la géométrie du système et de caler certains paramètres avant de lancer la simulation numérique du modèle global.

Les modèles locaux représentent un volume limité mais jugé représentatif du radier. Chaque élément est modélisé comme un élément volumique (élément maillé). Les chargements étudiés sont les suivants : le cas de chargement le plus défavorable dans la zone renforcée (calcul 1), le cas de chargement le plus défavorable en zone non renforcée (calcul 3) et le cas du chargement uniforme égale à la moyenne du chargement global appliqué sur le radier (calcul 2).

Les résultats mettent en évidence que le maillage renforcé permet de respecter les critères de contraintes dans les éléments de jet grouting. Ils valident le diamètre minimal des colonnes de transfert (1.2m en zone renforcée, et 1.0m en zone non renforcée) et l’épaisseur minimale du bouchon injecté définie pour assurer la stabilité de la fouille en phase provisoire (2m).

Il est observé que ce système de fondation permet une diffusion rapide des charges dans le bouchon injecté et une répartition quasi uniforme du chargement à la base du bouchon.

Nous présentons dans la figure 3 les résultats du calcul 1.

Figure 3. Modèle local, calcul n°1, zone renforcée (Nu=15113 kN) – Représentation des contraintes verticales effectives (Valeur moyenne : 1630 kPa, Valeur maximale : 2320 kPa) - Unité : Pa

4 SIMULATION STATIQUE DU MODÈLE GLOBAL

4.1 Couplage des modèles

Le dimensionnement de l’ensemble du projet est basé sur un système composé de deux modèles en interaction : le modèle 1 de la superstructure avec un logiciel propre (Advance Design et ANSYS) et le modèle 2 du sous-sol intégrant le système de fondation avec un autre logiciel (FLAC 3D). La difficulté de ce système est d’obtenir un couplage cohérent entre les deux modèles et une simulation correcte de l’interaction sol-structure. Cet objectif a été atteint en procédant à une série d’itérations entre les deux modèles.

Une première évaluation analytique des raideurs verticales du sol (raideurs surfaciques) a été effectuée sur la base des résultats de déformation du radier issus de l’étude de projet (simulation aux éléments finis en 2D). Ce premier champ de raideurs surfaciques a permis une première simulation de la structure et l’obtention d’une première descente de charges (3 zones principales de raideur et une zone périphérique d’ajustement nécessaire à l’approche « Winkler », raideur majorée). Cette première descente de charges a permis de lancer la première simulation numérique avec FLAC 3D. De cette première itération, il a été déduit un nouveau champ de raideurs verticales, cette fois-ci ponctuelles, calculées au droit de chaque colonne de transfert. Ce deuxième champ de raideurs a fait l’objet d’une nouvelle simulation de la structure et l’obtention d’une nouvelle descente de charge cette fois-ci au droit de chaque colonne. A partir de cette seconde descente de charge, un troisième calcul des raideurs a été effectué et a permis de confirmer la convergence du système par adéquation des déformée du radier fournie par chaque modèle. Les résultats en termes de déformation, d’efforts et de contraintes ont alors pu être exploités.

4.2 Résultats du modèle 2

Nous analysons tout d’abord les efforts dans toutes les colonnes de transfert modélisées sous forme d’un élément structurel. La figure 4 est un extrait du modèle 3D global, en condition statique. Les efforts traduisent la répartition des charges appliquées, sachant que les files des colonnes à maille renforcée suivent les porteurs (voiles et poteaux). Un cas représentatif du porteur le plus chargé est un voile de section transversale 0.40mx2.0m chargé à 15 110 kN (résultante verticale). La charge répartie correspondant à la valeur moyenne du chargement global sur le radier est de 135 kPa.

1344


Figure 4.Modèle global - Efforts axiaux dans les colonnes de transfert (Valeur moyenne : 804 kN, Valeur maximale : 2680 kN) - Unité : N

La combinaison du critère de la contrainte maximale et de la section comprimée est représentée dans le graphe M=f(N) de la figure 5. Le critère de la contrainte moyenne est vérifiée par ailleurs. C’est l’introduction des rotules plastiques qui permet d’aboutir au respect des critères.

0 500 1000 1500 2000 2500 3000 3500 4000 45000

100

200

300

400Vérification sous G

(kN)

(kN

.m)

M xo( )

Ma

Mb

Mc

T i 2

N xo( ) Na Nb Nc T i 3

Figure 5. Modèle global - Valeurs M, N des colonnes de transfert (diamètre 1.2m)

Nous analysons ensuite la déformée du radier. L’introduction des rotules plastiques en tête des colonnes de transfert conduit à un état d’équilibre stable qui ne modifie pas la déformée globale ni du radier, ni du bouchon, ce qui valide la vérification des efforts. La répartition des tassements est cohérente avec la descente de charges et les conditions géotechniques. En effet, on observe une augmentation des tassements au niveau de la partie Est, caractérisée par une remontée des silts profonds lâches (cf. figure n°6). La valeur maximale calculée s’élève à 11 cm (valeur proche de la valeur de projet : 13cm).

Figure 6. Modèle global - Répartition des tassements sous le radier (Valeur maximale bleu: 11cm, Valeur minimale rouge: 4cm) - Unité : m

Ce premier résultat a permis de fournir une carte des différentiels de tassement, critère fondamental pour le dimensionnement du radier. Ce critère est fixé à ∆W/∆L < 1/500ème. Seule une zone limitée du radier fait apparaître un gradient compris entre 400 et 500. Cette zone a donc fait l’objet d’une vérification particulière de la section béton-armé du

radier. Il a également été vérifié les tassements aux avoisinants. Les calculs ont montré qu’ils sont d’amplitude modérée et que le gradient reste supérieur à 500 et à 1000 au-delà de 15m de distance à la paroi.

D’autre part, sont analysées les contraintes dans le bouchon injecté. Il a été procédé à deux étapes de calcul pour mieux simuler le phénomène physique de mise en tension du bouchon. La première étape a consisté à élever volontairement la traction de coupure dans le bouchon (Rc/2=3000 kPa), ce qui a permis de localiser les zones en traction : deux zones limitées ont été mises en évidences (traction inférieure à 50 kPa). La seconde étape de calcul a consisté à annuler la tension de coupure dans les zones en traction, ce qui a permis d’évaluer l’incrément de déplacements verticaux. Cette étape simule l’apparition de fissures dans les zones tendues du bouchon. On observe que les incréments de déplacement sont très faibles (inférieurs au millimètre) ce qui a permis de conclure que la fissuration obtenue par l’absence de résistance à la traction ne modifie pas l’état d’équilibre du système de fondation.

5 CONCLUSION

La modélisation numérique en 3 dimensions d’un système de fondation pour un radier de 11500m² peut s’avérer complexe et sa fiabilité toute relative sans vérification préalable.

Avant de simuler un modèle global, il est apparu indispensable de simuler des modèles locaux dans des configurations représentatives qui permettent de fiabiliser la géométrie et le fonctionnement du système radier / colonne de transfert / bouchon / sols.

Afin d’assurer la fiabilité des résultats, il est fondamental de procéder à des itérations successives nécessaires au couplage du modèle de structure avec le modèle de sol (couplage effort / déplacement, c’est à dire ajustement de la raideur du sol).

Un effort particulier a été mené sur les modèles mécaniques du sol, des éléments de structure et sur les liaisons entre éléments afin de représenter au mieux les phénomènes physiques. L’atteinte de cet objectif nécessite également des tests et des vérifications.

Enfin, ce projet montre que la modélisation numérique en 3D permet de représenter des phénomènes physiques qui n’auraient pas pu être appréhendés « intuitivement » ou simplement par une analyse 2D (forme des déformations du radier, effet des liaisons des colonnes, effet de la traction de coupure dans le bouchon). Le premier contrôle des tassements, effectué sur une période de 12 à 18 mois depuis l’achèvement des bâtiments (construction phasée et terminée au ¾ environ), confirme les résultats obtenus par la modélisation 3D. Les mesures finales du tassement permettront de mieux apprécier ces résultats.

6 RÉFÉRENCES

Itasca Consulting Group, Inc.2006. Fast Lagrangian Analysis of Continua in 3 Dimensions, Minneapolis, Minnesota, USA.

Koscielny M., Briançon L., Dias D.,2007, Projet national A.S.I.RI, Synthèse benchmark tranche 1, thèmes 1&4, CNAM, Paris

Bagagli Y., Vincens E., Fry J.J., 2010.a model for the computation of engineering earth structures to a seismic motion. EJECE Volume 14 – No.5/2010, 599-616

Leroueil S., Magnan J.P., Taveans F. 1985. Remblais sur argiles molles. Technique et Documentation Lavoisier, Paris.

Université Joseph Fourier, Laboratoire “3S”, 2006, Pratique éclairée des éléments finis en Géotechnique, Les Modèles de comportement, Marseille,

Cordary D., 1995, Mécanique des sols, Lavoisier, Paris

1345

Comportement du viaduc élevé de la ligne 12 du métro de la Ville de Mexico, autour de la Sierra de Santa Catarina

Elevated Viaduct behavior of Metro Line 12 Mexico City in the nearness of the Santa Catarina

Rodríguez G. L.B. Directeur de l’Enterprise IPISA de C.V.

Soria C.B. Chef de Géotechnique, IPISA de C.V.

RÉSUMÉ: Cet article présente l’évolution des mouvements observés pour la fondation du viaduc élevé de la ligne 12 du métro dans les premiers 20 mois après sa construction. La partie observée se trouve dans la zone sud de la Ville de Mexico. Cette zone est formée par la présence de collines souterraines de roches et de sol dur au voisinage des vallées d’argile volcanique lacustre, exposé auxeffondrements régionaux, typiques de la zone du lac de la Ville de Mexico. L’article décrit aussi les solutions alternatives utilisées pour les fondations du viaduc et les solutions considérées pour limiter les différences de tassement entre la fondation appuyée sur sol dur et la fondation appuyée sur argile afin d’obtenir un comportement compatible pour l’exploitation de la Ligne.

ABSTRACT: This paper presents the evolution of the movements that have been the foundations of Elevated Viaduct of Metro Line12 in the first 20 months of construction. The observed section is located in the south-east of Mexico City, formed by the presence ofunderground rock outcrops and hard floors in the nearness of which are volcanic lake valleys of very soft clay, subject to regionalsubsidence, typical lake areas of the City. It also describes the foundations used in this viaduct and considered solutions to mitigatesubsidence differences between foundations supported on hard and soft clays supported to ensure appropriate behavior duringoperation of Line.

MOTS-CLÉS: viaduc, fondation, tassement, solution

KEYWORDS: viaduct, foundations , settlements, solution

1 DESCRIPTION DU PROJET ET DE SA COMPLEXITÉ GÉOTECHNIQUE

La Ligne 12 du métro de la Ville de Mexico traverse la partie sud de la Ville d’est en ouest, sur une longueur approximative de 24 km et possède 20 stations, une zone d'ateliers et un dépôt. La choix de conception de la Ligne 12 décrite d’est en ouest correspond à une structure de surface pour les deux premières stations, un viaduc élevé sur le tronçon suivant sur une longueur d’environ 8,5 km, tranchée à ciel ouvert de 2,5 km de longueur, tunnel construit avec l’insigne EPB de 8,5 km de long, pour terminer sur un tronçon d’une longueur de 2,0 km correspondant à un tunnel construit selon des méthodes conventionnelles. La Ligne a été mise en service le 30 octobre 2012.Du point de vue géotechnique, la Ligne 12 passe par une zone de sols durs à l’Ouest qui correspondent aux reliefs montagneux qui entourent la Ville de Mexico, une zone de transition entre les sols dur et les sols lacustres, une zone d'argiles tendres lacustres typiques de la Ville de Mexico ;dans la région du sud-est nous avons une zone complexe formée d’affleurements rocheux et collines rocheuses souterraines proches de la surface avec à leurs côtés des vallées lacustres formées d’argiles volcaniques tendres sujettes à des tassements localisés typiques de la Ville de Mexico. Figure N ° 1.

Figure 1. Localisation de la Ligne 12 sur les sols de la Ville de Mexico

Cette dernière zone correspond aux contreforts de la Sierra de Santa Catarina et sa complexité repose sur le fait que la Ligne de métro repose sur des zones dures et sur des zones lacustres d’argiles tendres sujettes à des tassements localisés, proches des premières. Dans ces conditions, la Ligne 12 aura des zones avec beaucoup de tassements dans les vallées lacustres avoisinant des zones fixes, ce qui pose la problématique d’absorber ces grandes différences de subsidence sur de courtes distances tout en garantissant un comportement adéquat des structures devant ces tassements à court et à long terme. Il est important de mentionner que toute alternative pour

1346


accueillir la Ligne dans les environs de la Sierra de Santa Catarina, que ce soit en surface, avec un viaduc élevé, ou encore un tunnel ou une tranchée, sera sujette aux fortes subsidences différentielles et devront donc être prises en compte pour un comportement adéquat.

Pour le tronçon du viaduc élevé, la conception de la Ligne 12 a choisi l’utilisation de structures construites à l'aide de colonnes et de têtes en béton renforcé, avec des séparations variant entre 25 et 30 m, sur lesquelles s’appuient des poutres préfabriquées en béton et des poutres métalliques ne faisant que reposer sur la structure. La hauteur de ces structures laissent un espace libre de 5,50 m sauf dans la zone spéciale de croisement avec un pont routier au nord de la station Periférico où la hauteur du viaduc élevé atteint 15,00 m. Les fondations de chacun des supports ont été définies en accord avec les caractéristiques du type de sol dans lequel ils sont situés, comme décrit ultérieurement.

Du point de vue géotechnique, les formations rocheuses de la Sierra de Santa Catarina sont formées de roches volcaniques basaltiques, très crevassées et mélangées à de la mousse

volcanique "tezontle" dans sa partie la plus proche de la surface, améliorant ainsi les indices de qualité du massif rocheux par rapport aux zones plus profondes. Lorsque le bassin de la vallée a cessé son drainage à cause de la formation des chaînes de montagnes du Sud (ère cénozoïque, référence 1), se sont formés les lacs où se sont déposées les cendres volcaniques, transportées parle vent, qui ont donné origine aux argiles lacustres de la Ville de Mexico.

Au milieu du siècle dernier, il y a eu une forte exploitation des nappes phréatiques profondes pour fournir la Ville en eau, ce qui dans une moindre mesure continue aujourd'hui et qui favorise les affaissements localisés de ces zones lacustres.

Dans les zones des contreforts de la Sierra de Santa Catalina, les vallées lacustres sont constituées d'argiles tendres d'épaisseurs variant entre 20 et 60 m, avec des teneurs en eau variant entre 150 à 450 % et des résistances allant de 2,50 tonnes/m2 à 5,00 tonnes/m2 de cohésion pour des essais non consolidés et non drainés. Figure 2. Les niveaux d'eau souterraine dans ces sols argileux sont situés entre 3,0 m et 4,0 m profondeur.

Figure 2. Sols d’assise des fondations du Viaduc Élevé de la Ligne 12.

2 AFFAISSEMENTS LOCALISÉS

Les 8,5 km du viaduc élevé sont situés entre les stations Pueblo Culhuacán et Estación Zapotitlán et le long de cette zone se trouvent trois vallées lacustres limitées de la manière suivante: la première vallée s'étend de la station Zapotitlán à la station Los Olivos, la deuxième vallée de la station San Lorenzo Tezonco jusqu'à la station Calle 11 et la troisième de cette dernière jusqu'à la station San Andrés Tomatlán. Figure 2.

Les vitesses d’affaissement local dans ces vallées ont été obtenues grâce aux nivellements effectués par l'Autorité du Système des Eaux de la Ville de Mexico entre 1970 et 1992 et interprétées par l'Institut d'ingénierie de l'Université Nationale Autonome de Mexico, UNAM (Référence 3). Il est important de préciser que les vitesses d’affaissement local ne peuvent pas être extrapolées de façon linéaire, étant donné que les modèles d'extraction de l'eau de la Ville sont très variables.

Compte tenu de ce qui précède, dans la première vallée les vitesses d’affaissement local enregistrées atteignent des valeurs de 4,0 cm/an, dans la deuxième vallée 6,0 cm/an et dans la troisième vallée 3,00cm/an. Ces vitesses signifient un affaissement total futur de 2,0 m, 3,0 m et 1,50 m pour chacune des vallées respectivement, lesquels correspondent à une période de 50 ans, durée minimum devant être considérée dans la conception. A ces valeurs doivent être ajoutés les affaissements causés par la surcharge des structures du viaduc élevé.

3 MESURES PRISES EN COMPTE POUR LA CONCEPTION

Afin d'obtenir une conception qui garantisse un comportement approprié à long terme pour le Viaduc Élevé et compatible avec son environnement, certains concepts, tels que les suivants, ont été pris en considération.

Nous avons pris en compte le fait que les inclinaisons des voies pourraient souffrir quelques changements qui, sans avoir encore été déterminés, ne dépasseraient pas les 10% pour une période de 50 ans pour ne pas modifier la conception de la Ligne.

Nous avons pris en compte le fait que les déformations des supports de la Ligne devaient être compatibles avec celles de leur environnement pour éviter des problèmes futures. Par exemple, en cas d’emploi de fondations fixes comme des piliers ou des pilots reposant sur les strates dures, par effet de l’affaissement local dans les vallées lacustres argileuses, nous assisterions à une "émersion" des fondations par rapport au sol voisin, ce qui affecterait’ inclinaison des voieries parallèles, ainsi que le comportement des équipements municipaux, particulièrement la présence d'un tuyau d’eau potable de 1,80 m de diamètre qui suit le tracé du métro et qui se trouve à 10,00 m de l’axe de la Ligne et à 4,0 m de profondeur, ce qui pourrait causer des dégâts considérables.

La Réglementation des Constructions du District Fédéral de la Ville de Mexico ne prévoit que la conception de bâtiments et indique que les tassements, entre deux supports consécutifs pour des structures en béton est: = 0,006*l où 'l' est l’espace entre les supports, en conséquence de quoi les assises admissibles

1347


pour des espaces de 30 m, ce qui est le cas typique de la Ligne 12, seront de 18 cm.

La même réglementation limite les déformations totales par consolidation des fondations isolées a une valeur de 30 cm pour les sols lacustres.

En ce qui concerne les spécifications AASHTO, celles-ci indiquent que pour des solutions structurelles de ponts juste posés, la subsidence différentielle «» entre deux supports consécutifs ne doit pas excéder: = 0,008*l qui correspond aux valeurs de 24 cm.

Pour nous conformer aux limitations décrites ci-dessus, les mesures décrites ci-dessous ont été prises. Dans le projet des voies, nous avons envisagé des

contrepentes pour les transitions critiques entre les zones fixes et les zones mobiles qui permettraient avec le temps d’atteindre les valeurs de conception et qui plus est de supporter les tolérances décrites antérieurement.

Espace suffisant pour la mise en place de cales entre la tête et la poutre de la structure située sur les transitions critiques entre les zones fixes et les zones mobiles afin de pouvoir corriger une partie de la déformation à long terme.

L’installation de pilots d’inclusion, au niveau des fondations et de leur environnement dans le but d’échelonner les déformations produites par les tassements local au niveau des transitions critiques entre les zones fixes et les zones mobiles. Ces inclusions ont été installées entre une certaine profondeur et les strates dures de façon à agir de façon "entrelacée" avec le système de fondations et à réduire progressivement les déformations causées par les tassements local. Figure 3. Cette solution de pilots entrelacés a donné de bons résultats dans la réduction des tassements des fondations dans les sols de la Ville de Mexico.

L'utilisation de poutres simplement posées entredeux supports consécutifs.

L'utilisation de fondations "flottantes" ou compensées se déplaçant en fonction de l’affaissement local.

L'emploi éventuel d'épaisseurs de ballast supérieures aux normes.

4 CHOIX DE CONCEPTION DES FONDATIONS EMPLOYÉES

Les Fondations utilisées pour les sections du viaduc élevé de la Ligne 12 (Figure 3) ont été des différents types suivants : ________________________________________________________ Zone Type de fondation ________________________________________________________ Zone rocheuse Semelles isolées reposant sur la roche.

Semelles isolées avec pilier en béton incorporé dans la roche saine.

Zones de vallée lacustre Caissons compensés de fondation avec pilot de friction Cellules structurées travaillant par friction.

Zones de transition Cellules structurées travaillant par friction et caissons compensés de fondation avec pilot de friction, dans les deux cas avec pilots d’inclusion situés longitudinalement et transversalement au tracé de la Ligne, interagissant avec les fondations. ________________________________________________________

5 ANALYSES DE TASSEMENTS FUTURS

La détermination des tassements à long terme, concernant les fondations situées dans les vallées lacustres, tant en ce qui concerne les caissons compensés que les cellules structurées, est basée sur les théories classiques pour définir les tassements des fondations avec pilots dans les sols tendres. Dans le cas des

fondations situées dans les régions critiques de transition, nous avons eu recours en plus à des programmes d'éléments finis qui ont pris en considération les caractéristiques de rigidité et de déformabilité des sols, les fondations et les inclusions, ainsi que certains artifices pour représenter le phénomène des tassements local. Avec cette méthode, la quantité, la distribution et les caractéristiques des pilots d'inclusion ont été précisés.

Figure 3. Différent types de fondations utilisées en tronçons de argile lacustre et de transition de Viaduc Élevé de Ligne 12.

6 CARACTÉRISTIQUES GÉOTECHNIQUES DE CHAQUE SECTION ET ENREGISTREMENT DES TASSEMENTS

Nous présentons ici l’évolution et l’interprétation des mouvements des supports du viaduc élevé de la Ligne 12 du Métro, correspondant aux tronçons compris entre les stations

1348


Zapotitlán - Nopalera, Nopalera – Los Olivos, San Lorenzo - Periférico Oriente et Calle 11 - Santa Maria Tomatlán. Les périodes de nivellement vont de janvier 2010 jusqu'aux premiers mois de 2012.

Les déformations enregistrées entre les supports constitués pour différentes solutions sont résumées dans le tableau suivant. Deux fondations tassements type représenté sur la figure 4.

Vue d'ensemble des assises sur les supports et des assises différentielles du Viaduc Élevé de la Ligne 12 par type de fondation Type de support Tronçon Assise

Minimum (mm)

AssiseMaximum (mm)

Assisedifférentiellemaximum entresupports contigus(mm)

Caissonpartiellement compensé

Zapotitlán – Nopalera

2 10 6


Nopalera – Los Olivos

2 10 6


Calle 11 – Lomas Estrella

3 6 3

Cellule structuré Zapotitlán – Nopalera

10 70 28

Cellule structuré Calle 11 – Lomas Estrella

16 25 9

Piliers ou Semelles reposant sur roche saine

Nopalera – Los Olivos

2 8 4

Piliers ou Semelles reposant sur roche saine

Calle 11 – Lomas Estrella

4 4 0

Figure 4. Évolution des tassements en fondations des zones d’argiles (a) et en zones de transition avec inclusions (b).

Le comportement des supports dans les zones de transition critique et les supports contigus avec des fondations différentes sont donnés ci-dessous : Sur le tronçon Zapotitlán - Nopalera au niveau du passage de

cellule structuré à caisson compensé, nous avons observé des assises 10 mm et 5 mm respectivement.

Sur le tronçon Nopalera – Los Olivos au niveau de la zone de transition critique entre caisson compensé et pilier sur roche, nous trouvons des assises de 3 mm et 5 mm respectivement.

Sur le tronçon de transition critique San Lorenzo-Periférico, entre caissons compensés et piliers sur roche, les assises enregistrées varient entre 6 et 8 mm

Sur le tronçon Calle 11 – Lomas Estrella nous trouvons trois solutions de fondation. Les assises entre caisson compensé et

cellule structuré sont de 6 mm et 23 mm respectivement et les assises entre cellule structuré et pilier sur roche sont respectivement de 20 mm et 4 mm.

7 CONCLUSIONS

Dans tous les cas, les graphiques d’assise correspondant auxsupports déviés sur argile montrent une tendance de stabilisation avec des vitesses d’affaissement très petites sur les derniers mois; et l'ampleur des déformations pour chaque support ainsi que les déformations différentielles entre les supports contigus, restent dans l’échelle des normes spécifiées dans les règlements susmentionnés.Avec la construction du viaduc élevé déjà conclue, on remarque que pendant les premiers mois, les affaissements enregistrés sont petits et ne mettent pas en danger le fonctionnement de la Ligne ni les installations voisines, mais il est nécessaire de contrôler l'affaissement de la Ligne avec une période d’observation plus longue pour contrôler et garantir son comportement futur.

8 REFERENCES

Raúl J. Marsal y Marcos Mazari. (1955). “El Subsuelo de la Ciudad de México” Facultad de Ingeniería, UNAM, México.

Instituto de Ingeniería, UNAM. (2009). Informe final Línea 12 del Sistema de Transporte Colectivo Metro. D.F. Non publié.

Gobierno del Distrito Federal (2004). “Normas Técnicas Complementarias para Diseño y Construcción de Cimentaciones”, Gaceta del Distrito Federal, México.

AASHTO, (2012). “Standard Specifications for Highway Bridges”, 17th Edition - 2012.

Consorcio ICA – ALSTOM – CARSO. (2010). Campaña de instrumentación geotécnica de la Línea 12 del Metro. D.F. Non publié.

EIBS GmbH, 2010, Planfeststellungsunterlage zum Ersatz der Bundesstraße B 176 zwischen Pödelwitz und Neukieritzsch, Bauabschnitt 1 (unpublished)

Ahner, C., Kirstein, J., Uhlemann, S., Röder, K., Uhlig, P.: Baugrundverbesserungsverfahren zur Gründung der Bundesstraße B 176 auf einer jungen Hochkippe im Braunkohlenrevier der MIBRAG. Baugrundtagung 2012

Ahner, C., Kirstein, J., Uhlemann, S. Uhlig, P.: Ground improvement methods for establishment of the federal road B 176 on a new elevated dumb in the brown coal area of MIBRAG, ISSMGE - TC 211 International Symposium on Ground Improvement IS-GI Brussels 2012

1349

Influence of installation damage on the tensile strength of asphalt reinforcement products

Influence de l’endommagement de mise en place sur la traction des produits de renforcement en asphalte

Sakou Touole L., Thesseling B. Huesker Synthetic GmbH, Gescher, Germany

ABSTRACT: One of the major problems associated with the use of asphaltic pavements is reflective cracking. This phenomenon iscommonly defined as the propagation of cracks from an existing pavement or base course into and through a new asphalt overlay, resulting from load- and/or temperature-induced stresses. One of the proven techniques to reduce/delay reflective cracking is the useof asphalt reinforcement. The properties of asphalt reinforcement (e.g. tensile strength, tensile strain) are influenced during the paving and compaction procedures of the asphalt. The loss of tensile strength during those procedures is known as installation damage. Thedegree of installation damage largely depends on the raw material used, the number of passes of trucks and compactors. There iscurrently a lack of experience and knowledge concerning the real residual properties (what could be termed “effective tensilestrength”) of asphalt reinforcement products after the paving and compaction procedures. This paper describes a test procedure developed at the University of Aachen. The research illustrates a considerable difference in loss of tensile strength, due to the effectsof installation damage.

RÉSUMÉ : Un des problèmes majeurs liés à l'utilisation d'asphalte pour les chaussées est la fissuration réflective. Ce phénomène estdéfini comme la propagation des fissures à partir de la couche dúsure ou de base existante à travers la nouvelle couche dúsured'asphalte, résultant des contraintes induites par le trafic et/ou la température. Une des techniques prouvées pour réduire/retarder la fissuration réflective est l'utilisation d’armature d'asphalte. Les propriétés de lármature d'asphalte (notamment résistance à la traction,allongement à la rupture) sont influencées durant la pose et le compactage de l'asphalte. La perte de résistance à la traction lors de cesprocédures, connue sous le terme «Endommagement mécanique», dépend en grande partie de la matière première utilisée, du nombre de passage de camions et compacteurs. Il existe actuellement un manque d'expérience et de connaissances concernant les propriétésréelles résiduelles «résistance à la traction efficace» des produits dármature d'asphalte après la pose et le compactage de lásphalte.Le présent article décrit la méthode d'essai développée à l'Université de Aachen. La recherche révèle une différence notable de perte de résistance à la traction, en raison des effets des endommagements mécaniques.

KEYWORD: installation damage, asphalt reinforcement, effective tensile strength

MOTS-CLÉS : Endommagement mécanique, Armature dásphalte, résistance à la traction efficace

1 INTRODUCTION

It is well known that cracks appear due to external forces, such as traffic loads and temperature variations. The temperature influence leads to the binder content in the asphalt becoming brittle; cracking starts at the top of a pavement and propagates down (top-down cracking). On the other hand, high stresses at the bottom of a pavement, from external dynamic loads, such as traffic, lead to cracks that propagate from the bottom to the top of a pavement (bottom-up cracking). A conventional rehabilitation of a cracked pavement involves milling off the existing top layer and installing a new asphalt course, but cracks are still present in the existing (old) asphalt layers. As a result of stress concentrations at the crack tips caused by external forces from traffic and natural temperature variations, the cracks will propagate rapidly to the top of the rehabilitated pavement. Deteriorated concrete pavements are typically rehabilitated by installing new asphalt layers over the old concrete slabs. Temperature variations lead to a rapid crack propagation especially at the expansion joints to the top of the new asphalt overlay. Asphalt reinforcement has been used worldwide for many years to delay or even prevent the development of those reflective cracks in asphalt layers. Currently there are a number of different asphalt reinforcement and systems made of different

raw materials (e.g. polyester, fiberglass, carbon fiber, polypropylene...) available in the market. It is not disputed that each of these systems has a positive effect in the battle against reflective cracking; however there are differences concerning the real residual properties “effective tensile strength” of each asphalt reinforcement after the paving installation procedure. The properties of asphalt reinforcement (e.g. tensile strength, tensile strain) are influenced during their installation, the paving procedure (paver and truck passes) and the compaction of the asphalt (Figure 1). The result, specifically the loss of tensile strength of the asphalt reinforcement grid during the paving procedure, is known as installation damage. After an asphalt reinforcement product is placed, many asphalt delivery trucks have to pass over the grid. Additionally there is the compaction of the hot mix asphalt, during which the individual filaments or strands of the asphalt reinforcement are largely influenced by the movement of aggregates, in particular of coarse and sharp-edged aggregates. Next to the reinforcement characteristics (flexible or brittle raw materials), the degree of installation damage by roller compaction not only depends on the number of passes and the type of compaction (e.g. rubber tired, static, dynamic), but the weight of the compactor and the condition of the base layer (e.g. smooth, rough/milled) as well.

1350


To successfully counteract reflective cracking, installed reinforcement products must resist the installation influences without damage and without significant loss of strength. There is currently a lack of experience and knowledge concerning the real residual properties “effective tensile strength” of asphalt reinforcement products following their installation and the subsequent paving installation procedure.

In the context of a diploma thesis at the RWTH Aachen University, a test procedure to describe installation damage was developed. One of the aims was to analyze and quantify the “effective tensile strength” for two different asphalt reinforcement products with different raw materials (polyester and fiberglass) after the influence of installation damage.

Figure 1: Influences on asphalt reinforcement products during the asphalt installation

2 INVESTIGATION AT THE RWTH AACHEN UNIVERSITY

As part of the work to assess the resistance of asphalt reinforcement products to installation influences, site-appropriate tests were performed at the institute's installation test track. As one goal, comparable tensile strengths of the tested products after the following influences were intended to be achieved:

• Only the influence of asphalt truck passes

• Only the influence of asphalt compaction passes

• Combination: The influence of asphalt truck and compaction passes.

2.1. Test procedure

To determine the impact of truck traffic only, undamaged specimens of the reinforcement products were placed on a clean and even road and loaded by a truck. The applied load was carried out by 35 passes with a speed of 20 ± 5 km/h without any steering movements or braking activity (Figure 2). Considering a truck with 5 axles who drives backwards to the paver and forward again this corresponds to 3.5 delivery trucks driving over the grid.

Figure 2: Influences due to truck traffic only

In preparation for the tests, an asphalt binder course (AC 16 B S) was installed on the base of the test track first. Onto the binder course, each reinforcement grid has been placed according to the manufacturer´s installation guidelines. Some of the pre-damaged specimen (truck passes) have also been used in the test-track for further exposure to simulate the double load-effect, truck passes and compaction. To differ between undamaged, pre-damaged and the different loading types, the specimen had been placed into separate sections. Subsequent to the installation of the reinforcement specimen a 50mm asphalt wearing course was installed (Figure 3) and compacted with 6 roller passes (Figure 4).

Figure 3: Wearing course installation

Figure 4: Wearing course compaction

To test their tensile strength after the asphalt installation and compaction, some of the specimen had to be removed after the installation of the wearing course. For this reason these specimen have been wrapped into an aluminium foil and coated with a separating agent to create a very bad interlayer bond. To investigate the different influences the test track was divided into different sections:

A - An undamaged fiberglass reinforcement was installed, followed by the installation and compaction of an asphalt wearing course. ( Load influence: Compaction only)

B - An undamaged Polyester reinforcement was installed, followed by the installation and compaction of an asphalt wearing course. ( Like „A“, load influence: Compaction only)

C - A pre-damaged fiberglass reinforcement was installed. Subsequently the asphalt wearing course was installed and compacted. (Load influence: Truck passes and compaction)

D - A pre-damaged Polyester reinforcement was placed. Subsequently the asphalt wearing course was installed and compacted. ( Like „C“, load influence: Truck passes and compaction)

1351


2.2. Results In the context of the research the final material

characteristics (e.g. tensile strength) have been tested with the wide width tensile test according to EN ISO 10319. To compare the separate types of tests (variants) the property “residual strength” was chosen. The residual strength is defined as the maximum tensile strength after the installation damage tests, expressed as a percentage of the maximum tensile strength of the undamaged reference material. The detailed result can be found in chart 1.

After the load influence “truck passes” only, the polyester grid showed a residual strength of 85%, while the fiberglass grid had only 44% residual strength. After the load influence “compaction” only, the polyester grid showed a residual strength of 71%, while the fiberglass grid had only 21%.

The polyester grid specimens which were subjected to both loading influences - asphalt compaction and truck passes still had a residual strength of 70% while the fiberglass grid subjected to both loading influences revealed further damage with a residual strength of only 11%. The results revealed the considerable difference between the influence from truck traffic and asphalt compaction on the tensile strength of the specimens.

Figure 5: Results of installation damage test

3 CONCLUSIONS 2.3. Summary

After a series of testing, it can be safely concluded that installation damage plays an important role on the “effective tensile strength” of asphalt reinforcement products. It was found that the polyester grid lost max. 30% of its tensile strength after loading from truck passes and asphalt compaction. In contrast to the performance of the polyester grid, the fiberglass grid showed a loss of strength up to approximately 90%. The fiberglass grid was damaged significantly more than the polyester grid reinforcement (Figure 5).

As previously mentioned, asphalt reinforcement products must resist as much damage as possible from the stresses and strains applied during installation and compaction of the asphalt. Very high forces can also be applied to the individual strands of the reinforcement by aggregate movement within the hot asphalt during compaction.

In the research at the RWTH Aachen University, it was shown, that products with a brittle raw material (like fiberglass) can lose a significant part of their tensile strength when loaded by asphalt delivery trucks and after the asphalt compaction (Figure 5). The results of the research are validated by results of tests performed according to EN ISO 10722-1 “Geosynthetics: Procedure for simulating damage during installation” (tBU 2003). Furthermore, it is expected that for fiberglass grids the results would be worse on a milled surface.

Chart 1: Results of the wide width tensile test before and after installation damage tests

05101520253035404550556065707580

0‐P 0‐G a b c d e f

The tests reveal that polyester grids exhibit a very good resistance to installation damage compared to other products made with more brittle raw materials.

Tens

ile S

treng

th [k

N/m

]

4 REFERENCES

[1] Sakou L., 2011, Überprüfung der Wirksamkeit von Asphaltbewehrungssystemen unter Berücksichtigung der Einbaubedingungen, Diploma Thesis, RWTH Aachen, Institute of Road and Traffic Engineering, unpublished.

[2] Institut für textile Bau- und Umwelttechnik GmbH (tBU), Test Report Nr. 1.1/17810/493-2003e and 1.1/17810/494-2003e, Test method: Procedure for simulating damage during installation (acc. to ISO EN DIN 10722-1)

[3] FGSV- AP Nr.69, FGSV-Arbeitspapier Nr.69: "Verwendung von Vliesstoffen, Gittern und Verbundstoffen im Asphaltstraßenbau", Arbeitsausschuss: Asphaltbauweisen, Ad-hoc-Gruppe - Asphalteinlage, Köln, Ausgabe 2006

Type of installation damage

[4] FGSV, 2005, Merkblatt über die Anwendung von Geokunststoffen im Erdbau des Straßenbaus (M Geok E) O-P: Polyester grid - undamaged reference

O-G: Glass-fibre grid - undamaged reference [5] DIN EN ISO 10319, Geokunststoffe Zugversuch am breiten Streifen (ISO 10319:2008), Oktober 2008 a: Polyester grid - Only truck passes

b: Glass-fiber grid - Only truck passes c: Polyester grid - Only compaction d: Glass-fibre grid - Only compaction e: Polyester grid- Combination (truck passes and compaction)

[6] DIN EN ISO 10722-1, Geotextilien und geotextilverwandte Produkte - Verfahren zur Simulation von beim Einbau auftretenden Beschädigungen, August 2004

f: Glass-fibre grid- Combination (truck passes and compaction)

[7] Schmalz 2005, Schmalz M.: Vorstellung des FGSV Arbeitspapiers Nr.69 - Verwendung von Vliesstoffen,Gittern und Verbundstoffen im Asphaltstraßenbau, Asphalt- und Straßenbau. Seminar 13. Regenstaufer, 2005

1353

Influence of Anti-freezing layer on the Frost Penetration Depth for Paved Road Design

Influence d’une couche anti-gel sur la profondeur de pénétration du gel dans la conception des chaussées

Shin E.C. Dept. of Civil & Environ. Engrg, University of Incheon, Republic of Korea

Cho G.T., Lee J.S. Research Ins. For Engrg. & Tech., University of Incheon, Republic of Korea

ABSTRACT : Design of pavement in seasonal freezing areas should consider the environmental conditions in case of design andconstruction for pavement thickness. There are a lot of conditions of climate, soil, and material among the environmental conditions.One of that is caused from effect of these conditions is frost heaving. The frost penetration depth of paved road in Korea is usuallyestimated from the freezing index that made temperature data analysis of 30 years period and decided the thickness of anti-frost layer. It may be caused of over-design for pavement design with using the current estimation method of the frost penetration depth.Therefore, this study analyzed the depth of frost penetration for pavement design and the depth of frost penetration of paved roadusing field monitoring data. This paper presents the field monitoring results of frost penetration depth with anti-freezing layer andwithout anti-freezing layer. The analysis on the influence of anti-freezing layer to the pavement road was presented for the region ofSouth Korea.

RÉSUMÉ : La conception des chaussées dans les régions connaissant des périodes de gel doit tenir compte des conditions environnementales. Parmi les conditions environnementales, nombreuses sont celles relatives au climat, au sol et aux matériaux. Maisl’une d’entre elles est causée par l’impact de toutes ces conditions, c’est le déchaussement. En Corée du Sud, la profondeur depénétration du gel d’une route goudronnée est généralement évaluée à partir de son indice de gel. Celui-ci repose sur une analyse de données de température sur une période de 30 ans, et sa valeur détermine l’épaisseur de la couche anti-gel. La méthode de calculactuelle de la profondeur de pénétration du gel peut néanmoins provoquer une exagération de la profondeur des chaussées. Cette étudeanalyse le phénomène à l’aide de données de surveillance obtenues sur le terrain. L’article présente également les résultats de lasurveillance sur terrain de la profondeur de pénétration du gel sur des routes ayant une couche d’anti-gel et sur des routes n’en ayantpas. Les résultats de l’analyse de l’action de la couche d’anti-gel sur une route correspondent au cas de la Corée du Sud.

KEYWORDS : Field Frost Penetration Depth, Frost Index, Frost Penetration Map, Pavement, Air Temperature

1 INTRODUCTION

Desing of pavement in seasonal freezing areas should consider the environmental conditions including climate, soil, and moisture content of soil. One of that is caused from effect of these conditions is frost heaving. Jiang and Tayabji(1999) studied on the influence of in-situ moisture content on the seasonal monitoring site. Tomasz(2009), Wu, Zang and Liu(2010) evaluated the factors influenced on the freezing point in the soil water system and long-term thermal effect of the asphalt pavement. The frost penetration depth of paved road in Korea is usually estimated from the freezing index that made temperature data analysis of 30 years period and decided the thickness of anti-frost layer. It may be caused of over-design for pavement design with using the current estimation method of the frost penetration depth. Therefore, this study analyzed the depth of frost penetration for pavement design and the depth of frost penetration of paved road using field monitoring data.

2 FIELD MONITORING SYSTEM

2.1 Measuring instruments and their position

Field monitoring system is a system that measures field environmental parameters such as moisture and temperature. Purpose of field monitoring system is to measure the moisture content and temperature of the pavement automatically, continuously and objectively. Sensors of field monitoring system are shown in Fig. 1. Temperature sensors measure the

internal temperature of the pavement and moisture sensors measure the moisture content of the roadbed.

Figure 1. Field monitoring system and sensors

2.2 Field monitoring region

The field monitoring region is divided into three regions by freezing index 550~650°C·day, 450~550°C · day, and 350~450°C · day. Each region has three-section of road pavement such as cutting area, boundary area of cutting and banking, and lower area of banking. A total 25 sections for three regions (9 section in cutting area, 9 section in boundary area of cutting and banking, and 7 section in lower area of banking) were constructed as tabulated in Table 1.

1354


Table. 1. Field monitoring system section Freezing

index(℃·day)

Cutting section

Boundarysection

Banking section

No. 1 1 1 1No. 2 1 1 1

550-

650 No. 3 1 1 1No. 4 1 1 1No. 5 1 1 -

450–

550 No. 6 1 1 1No. 7 1 1 1No. 8 1 1 - No. 9 1 1 1

350–

450No. 10 1 1 1No. 11 1 1 1No. 12 1 1 -

200–

350 No. 13 1 1 1No. 14 1 1 - Below

200 No. 15 1 1 1 Total 15 15 11

Figure 2. Field monitoring region

2.3 Process of Field monitoring system construction

The placement of measurement sensors accompanies the calibration of sensors before the placement. To construct a field measurement system as shown in Fig. 3, the measurement sensors were laid at the center of each layer. After the placement is completed and then installed modem, battery, solar panel and data logger, consecutively, to collect data through wireless communication in console box. Fig. 4 shows the view of the installation of automatic measuring system.

Figure 3. Placement of measuring instrument on the compacted roadbed materials

Figure 4. Construction of field automatic monitoring system

3 FROST PENETRATION DEPTH OF PAVED ROAD WITH FIELD MONITORING DATA

3.1 Analysis of frost penetration depth for the region with freezing index 550~650 ·day

The analytical study of temperature distribution along the paved road profile was carried out by using the measured lowest daily temperature in the field. Figure 5 shows the temperature variations and correlation for pavement section for three areas with anti-frost layer existence and non-existence.

(a) Existence of anti-frost layer

(b) Non-existence of anti-frost layer

Figure 5. Frost penetration depth of No.1 region

3.2 Analysis of frost penetration depth for the region with freezing index 450~550 ℃·day

The analytical study of temperature distribution along the paved road profile was carried out by using the measured lowest daily temperature in the field. Figure 6 shows the temperature variations and correlation for pavement section of three areas with anti-frost layer existence and non-existence. The temperatures of No.4 or No.5 cases went down below 0℃ to subbase course and base course regardless of anti-frost payer

1355


existence or non-existence, and temperature of subgrade of No. 6 case went down below 0℃ in non-existence of anti-frost layer.




3.3 Analysis of frost penetration depth for the region with freezing index 350~450 ℃·day

The temperatures at No.7, No.8 and No. 9 regions went down below 0℃ to subbase course and base course regardless of anti-frost payer existence or non-existence.




3.4 Analysis of frost penetration depth according to regional

In this study, analysis of temperature distribution along the paved road profile was carried out with the measured the lowest daily temperature. Table 2 describes the frost penetration depth of cutting section with anti-frost layer for three years.

Table 2. Estimation results of frost penetration depth for three years

YearRegion

2009~2010 (cm)

2010~2011 (cm)

2011~2012 (cm)

No. 1 110 More than 110 70No. 2 120 More than 130 More than 130 No. 3 92 100 100No. 4 80 80 50

No. 5 55 55 40

No. 6 103 More than 120 120

No. 7 28 50 22No. 8 60 More than 60 45

No. 9 40 50 10

No. 10 70 50 50

No. 11 40 50 36

No. 12 17 17 17

No. 13 10 18 10

No. 14 18 37 29No. 15 50 70 52

4 COMPARISON OF FROST PENETRATION DEPTH OF FIELD RESULTS WITH EXISTING THEORY

The empirical equations developed by the Korea Institute of Construction Technology (KICT) and Japan Road Association (JRA) were used to determine the frost penetration depth. The results of frost penetration depth measured in the field and estimated by the empirical equations are tabulated in Table 3. Generally the frost penetration depths determined by the empirical equations are greater than those of field measurement with the exception of No.1, No.2 and No.3 sites. In particular, the results calculated by the Japan Road Association show no significant differences among the regional category of frost index.

1356


Table 3. Results of comparison for frost penetration depth Researchinstitute

Region

Fielddata(cm)

U.S. Army Corps of

Engineers(cm)

KICT (cm)

JRA(cm)

No. 1 110 95 105 85

No. 2 120 115 99 78

No. 3 92 110 96 74

No. 4 80 110 84 60

No. 5 55 104 94 72

No. 6 103 107 105 85

No. 7 28 90 92 69

No. 8 60 97 92 69

No. 9 40 94 84 61

No. 10 65 80 93 70

No. 11 60 70 80 56

No. 12 17 80 73 49

No. 13 10 71 61 38

No. 14 18 70 78 54

No. 15 50 67 54 40

(b) KICT (c) JRA Figure 8. Maps of the frost penetration depth by field measurement and empirical equations

6 CONCLUSION

The frost penetration depth of paved road was determined by the field measurement. The moisture content and temperature are measured and stored the data through solar panel data transmission system and manual system, respectively. The results of field monitoring for determination of frost penetration depth are summarized as below.

5 MAPS OF FROST PENETRATION DEPTH WITH UING FIELD AND EMPIRICAL RESULTS

(1) The paved road constructed with inclusion of anti-frost layer, the temperature at subgrade for five field regions does not go down below 0℃ with the exception of site No.1, No.2 and No.6. However, without inclusion of anti-frost layer, the temperature at subgrade with the region which has the frost index 550-650℃·day goes down below 0℃.

The maps of frost penetration depth were made based on 15 field monitoring datum using ArcGIS program. Interpolation method was adopted to make frost penetration maps with IDW (Inverse Distance Weighting). IDW is one of the most commonly used interpolation techniques. Fig. 8 shows the maps of the frost penetration depth made by using ArcGIS. Frost penetration depths by the field measurements show significantly different. However, the frost penetration depths using the empirical equation of KICT show a similar trend in lower frost index. In particular, the results by the empirical equation of JRA show no significant differences among the regional category of the frost index.

(2) The subbase and base courses are influenced by the temperature below 0℃ regardless of anti-frost layer is existed or not for all of the freezing index categories. (3) Frost penetration depth of field shows significant difference by the regional frost index. However, the frost penetration depth estimated by the empirical equation proposed by KICT shows a similar trend in lower frost index. In particular, the results calculated by the JRA empirical equation show no significant differences among the regional category of frost index. (4) Based on the analysis of field frost penetration depth measurement, the reasonable design concept can be available for road design.

7 ACKNOWLEDGEMENTS

This research was supported by a grant (11 Technology Innovation F01) from Construction Technology Innovation Program (CTIP) funded by Ministry of Land, Transportation and Maritime Affairs (MLTM) of Korean government.

8 REFERENCES

Asphalt Institute., 1995, Performance grade asphalt binder specification and testing, Superave Serise No. 1(SP-1)

Jiang, Y.J., Tayabji, S.D., 1999, Evaluation of in-situ moisture content at LTPP seasonal monitoring program sites, TRB 78th Annual Meeting, No. 990395

The Ministry of Land, Transport and Maritime Affairs, 2012, Evaluation of validity for frost protection layer and development of its construction criteria, Construction & Transportation R&D Report

Tomasz, K., 2009, Some factors affecting supercooling and the equilibrium freezing point in soil-water systems, Cold Regions Science and Technology 59, 25-33

Yoder, E.J., Witczak, M.W., 1973, Principles of pavement design, Second Edition, John Wiley and Sons, New York.

(a) Field measurements Wu. Q., Zhang Z., Liu Y., 2010, Long-term thermal effect of asphalt pavement on permafrost under an embankment, Cold Regions Science and Technology 60, 221-229

1357

Evaluation of roadbed potential damage induced by swelling/shrinkage of the subgrade

Effet du retrait-gonflement des sols sur les structures de chaussées

Simic D. Head of Geotechnical department. Ferrovial-Agromán

ABSTRACT: The expansive soils in arid and semi-arid regions are subject to seasonal moisture variations that trigger changes involume. These movements are reflected in swellings along the wet months and shrinkages along the dry months; seasonal movementsthat induce significant damages in the pavements. Traditionally, the construction of pavements on expansive clays has resulted inroads with a poor comfort level for the customers and a great maintenance cost for the administration. Such facts make veryproblematic the construction of road pavements in expansive soils. This paper analyzes the behavior of the pavement subject to deformations due to swelling and shrinkage of the subgrade, in order to evaluate some of the published design methods for theprotection of the pavement against the swelling phenomena of underlying clays. To introduce the design methods, this paper willdescribe first the usual pathologies due to swelling and shrinkage, and their explanation by means of the analysis of someinstrumented sections in existing roads. The different design methods will be summarized, showing also some limitations of theassumptions adopted in each analysis method.

RÉSUMÉ : Les sols gonflants situés dans des régions au climat aride sont soumis à des variations en teneur d’eau accompagnées de changements volumétriques : des gonflements en période humide et rétraction en période sec. Ces déformations se propageant au niveau de la chaussée donnent lieu à d’importants coûts de maintenance. Ces coûts rendent problématique la construction si ces problèmes ne sont pas correctement cernés et gérés. Dans cet article, le comportement de la chaussée soumise aux déformations de gonflement est décrit et les pathologies et méthodes d’analyse existantes dans la littérature sont évaluées. Des exemples sont montrés ainsi que les limitations des hypothèses retenues dans les procédés analysés

KEYWORDS: expansive soils, roads, semi-arid regions.

MOTS-CLÉS: sols expansifs, routes, regions semi-arides.

1 INTRODUCTION.

Expansive soil is a term usually applied to any soil that has a potential for shrinking or swelling due to changes in its moisture content. It is recognized that there are two main factors that provides the potential of the soil to swell and/or shrink: the properties of the soil and the environmental conditions of the area. The main soil parameters that are included within the first factor are the clay mineralogy, the soil water chemistry, the soil suction, the structure of the soil (fabric) and its dry density. Within the environmental conditions of the area the initial moisture condition, the moisture variations and the stress conditions are the factors believed to control the soil movement.

2 MECHANISM OF SWELLING/SHRINKING

The mechanism of the development of longitudinal cracks at the pavement in arid environments has been described by Zornberg, J. G.; Gupta, R. and Ferreira, J. A. Z. (2010). Tensile stresses induced by flexion of the pavement during settlements caused by the dry season leads to the development of longitudinal cracks. See Figures 1 and 2 below.

During the dry season there is a drop off in moisture content

of the soil in the shoulders of the pavement structure. The consequence of this reduction in moisture is a settlement in the shoulders that does not take place in the centre of the pavement where the moisture of the soil remains stable thorough the year. The appearance of cracks in the shoulder of the pavement accelerates the evaporation of the interstitial water of the soil reaching also greater depths.

Figure 1. Mechanism of longitudinal crack development on pavement over expansive clays during dry season. (Modified from Zornberg, Gupta & Ferreira, 2010).

Figure 2. Longitudinal cracks near the edge of pavement

Topographical surveys along the longitudinal cracks show that there is a tendency of the soil to settle when the cracks develop, confirming the model explained above.

1358


3 EVALUATION OF THE SWELLING DEFORMATION

The great volumetric expansion and contraction potential of these soils can be explained empirically through a direct correlation to plasticity index (from Texas Department of Transportation method TEX-124-E, “Vertical Rise Potential”) or from a theoretical approach (Little 2012). In the latter approach matrix suction is related to volume change. The change from the matric suction that exists under a current or existing moisture regime to a state of drying (where the matric suction increases) or to a state of wetting (where the matric suction decreases) is the trigger of volume change. This volumetric change was determined by Mitchell (1980) as a function of depth, soil type, and climatic conditions using the following relationship:

0.5

0 expz enU U U z

[1]

Where Ue is equilibrium suction, U0 is the amplitude of suction variation, n is the number of cycles of wetting and drying within a year, α is diffusion coefficient, and z is depth.

The difference between a current or initial (Ui) and final suction (Uf) as determined from equation [1] can be used to estimate the range of volume changes of the natural soils below the pavement structure. The volumetric strain is calculated using the following relationship (after Hong et al, 2006 and Bulut, 2001):

10 10

10 10

log log

log log

f fh

swelling i i

fh

shrinkage i

UVV U

UVV U

f

i

[2]

Where V

V

is volumetric strain, γh is suction

compressibility index, and σi and σf are initial and final overburden stress, respectively. From this relationship it is important to note that, first, volume change, whether shrinkage or swelling is driven by a difference between initial and final matric suction, U. Second, the impact of the driving force for volume change, ∆U, on volume change is determined by the suction compressibility factor, which operates in this constitutive relationship like a modulus in stress-strain constitutive relationships.

The result of the physico-chemical changes achieved through lime treatment of the clay soils had the practical effect of making the most highly susceptible soils to volume change within the active zone practically non-susceptible to volume change. As shown by equation 1, this active zone depth is influenced by climatic variables such as n and soil variables such as diffusivity, α. As one can visualize from equation 1, the upper portion of the active zone provides the greatest driving force, ∆U. Since it is the active, natural clay in this upper zone which is contributing more to the pavement movement, an envident remedial measure to replace this layer in large portion by an inert soil or the same natural clay treated with lime. In doing so, the swell and shrink volume change potential is greatly mitigated.

Suction values at depth for the application of equation 1 have normally a minimum suction value of U = 2.0pF and a maximum suction value of U = 4.5pF as measured in semi-arid zones. The suction values at the surface do not have limits and depend solely in the climatic region.

4 POTENTIAL VERTICAL RISE (PVR)

Texas method (TEX-124-E), is widely used in Texas to determine the required depth of replacement of expansive soils with inert soils, based on the expansion characteristics of the soils. This standard determines the Potential Vertical Rise (PVR) in soil strata, which is described as the “latent or potential ability of a soil material to swell, at a given density, moisture, and loading condition, when exposed to capillary or surface water, and thereby increase the elevation of its upper surface, along with anything resting on it”. Figure 3 shows the correlation between the PI of the soil and the volumetric change due to swelling.

Figure 3. Graph Plasticity Index vs. Percent Volumetric Change. From Tex-124-E

However, this method has a series of shortcomings:

1. Soil at all depths has access to water in capillary moisture conditions.

2. Vertical swelling strain is assumed as one-third of the volume change at all depths.

3. Remolded and compacted soils adequately represent soils in the field.

4. PVR of 0.5 inch (or 1 inch) produces unsatisfactory riding quality.

5. Volume change can be predicted by use if the plasticity index alone.

5 LABORATORY EVALUATION OF SWELLING

Twelve samples from five boreholes were collected from a

project in south Austin. The samples were selected to provide three replicate samples within a lower (<40%), intermediate (40 to 60%) and high (>60%) range of plasticity indices.

5.1 Comparison with the PVR analysis

The following laboratory tests were completed: Material passing 75 microns. Oedometer tests and free swell. Atterberg limits. Suction potential by pressure plate method.

1359


The Atterberg limits have been plotted in the plasticity chart of Figure 4.

Figure 4. Plasticity chart. SH-130 samples.

Swell deformations were obtained from the oedometer tests. The results are shown in the following Figure 5.

Figure 5. One-dimensional, free swell results.

The suction water characteristic curves from the pressure plate suction tests are represented in Figure 6 below.

Figure 6. Pressure plate suction vs. water content characteristic curves

The swell deformation values obtained from the oedometers

tests have been plotted in the graph of Tex-124-E for comparison purposes. Figure 7 shows that the samples with a moisture content in the very dry side have greater swell deformations than predicted by Tex-124-E standard.

ensional swell

.2 Suction based

determine the SWCC of soils sampled in the south Austin area.

ational Research Council

rel e of the samples. The SWCC and the γh can then be plotted.

Figure 7. Graph Plasticity Index vs. one dim

5 method for swell evaluation

There are several approaches to determine the suction compression index (γh). The soil water characteristic curve (SWCC) and volume measurements determined in the laboratory, can be used to calculate the γh.The Texas A&M University carried out pressure plate tests (see Figure 8) to

Figure 8. Pressure platCanada website).

e apparatus (from N

In the test, the weight and volume of the soil samples are

recorded at the end of each pressure cycle.The volume is measured using the Ottawa sand displacement method. The mass of the sand displaced is measured to calculate the increments in volume of the samples.

Three PVC samples blocks with smooth surface are used to calibrate the volume measurement equipment and obtain a

ation between the change in mass and volum

1360


6 CONCLUSIONS. COMPARISON OF THE SWELLING DEFORMATION OF PVR WITH THE SUCTION BASED METHODS

il water charac

The average suction compression index of the plate load

tests and the routine soil parameters were adopted to carry out a comparison between the methods of estimating swelling deformation (See Figure 11). The active moisture depth is the depth below ground level where the shrinkage and swelling movements of the soil are zero. The weather conditions and the properties of the soil are the most important parameters that determine the active moisture depth in a specific location.As it is already known, the PVR method is very dependent of the active moisture depth, which should be adopted based on the local experience. In this example, different depths have been adopted in the calculations.

Figure 9. Example of so teristic curve from the SH-

following equation proposed by tton (1977) can be applied.

d ratio tio

= total suction

entioned has been carried out for the samples of the test.

Figure 10. Com ressure plate tests

al strain or movement, can be calculated from the fol

l movement can be cal lated from the following equation:

130 samples . The suction compression index (γh) can then be determined

for agiven range of suction values. TheLy

Where: e = difference of voie0 = initial void rah The suction compression index (γh) can also be estimated

based on routine soil testing as the Atterberg Limits, % passing sieve #200 and % passing 2m. In 2004, Lytton proposed alternative charts that are implemented in the WinPRES software. The comparison of the γh calculated in the laboratory and the values estimated from the two authors aforem

Figure 11. Comparison of vertical movements calculated with

different methods The results showed that the relationships between one

dimensional swell and PI presented in the Tex-124-E can be considered acceptable if the “dry” condition is adopted. Based upon the calculations from soils obtained at South Austin, PVR calculations with an active moisture depth of 12ft would result in swelling values comparable to those calculated using suction based methods.

7 REFERENCES

Hong, G. T., Bulut, R., Aubeny, C., Jayatiaka, R. and Lytton, R. L. 2006 “Design Model for Roughness and Serviceability of Pavements on Expansive Soils”, TRR No. 1967. Litlle, D. 2012. “Background for predicting roughness and/or serviceability loss due to expansive soils”, Internal Memorandum.

parison of γh as calculated from p Mitchell, P. W. 1979. “Structural Analysis of Footings on Expansive Soil”, Research Report No. 1, Kenneth W. G. Smith and Associates, Newton, South Australia.

vs. estimated from routine tests. The vertic

lowing: Finally the summation of vertica Texas Department of Transportation 2011. “Pavement Design

Guide”. cu

1361

The performance of shale as fill and embankment material for a trunk roadin Ghana

La performance du schiste comme matériau de remblai pour une route destinée au traficde camions au Ghana

Solomon K.M.Gauff Ingineiures GmbA, Accra, Ghana

Oddei J.K.Architectural & Engineering Services Ltd, Accra, Ghana

Gawu S.K.Kwame Nkrumah University of Science & Technology, Kumasi, Ghana

ABSTRACT: As part of the Millennium Challenge Account (MCA) Compact, some roads in the Afram plains, includingAgogo – Dome trunk road of Ghana was constructed. Since it was uneconomical to haul suitable material over longerdistance for the project, shale which abounds in the area was evaluated during construction for use as embankment and fillmaterial. Representative samples were subjected to index property test, compaction, California Bearing Ratio (CBR) andtriaxial protocols at six different laboratories. The results, especially CBR values indicated variations between 8% and 12%,which are below the contract special specification minimum value of 15%. This was therefore considered as having marginalquality for its intended purposes. However a 100 meter, 1.00m thick, field trial road section was constructed using the shaleas fill under normal environmental conditions for traffic flow and monitored for a period of two months. The monitoringevaluation comprises conducting in situ CBR test, using the dynamic cone penetrometer and plate load bearing test includingpetrographic laboratory studies. Results obtained from the field evaluation indicated high CBR and bearing resistance valuesincluding insignificant settlement. Pavement performance indicators such as rutting and potholing were not evident. Theseperformance based characteristic of the shale merited its selection for use in the project. This paper also presents the resultsof laboratory tests as well as field studies.

RÉSUMÉ : Dans le cadre du Millénium Challenge Account (MCA) Compact, un certain nombre de routes ont étéconstruites dans les plaines d'Afram, dont celle d’Agogo – Dome. Puisque pour ce projet, il était trop coûteuxd’approvisionner des matériaux de remblai adéquat sur de longues distances, l’utilisation du schiste qui abonde dans cetterégion a été évaluée en tant que matériau de remblai. Des échantillons représentatifs ont été soumis à différents tests, indicede propriété, compactage, essai CBR et essais triaxiaux, essais effectués par six laboratoires différents. Les résultats, enparticulier les valeurs CBR ont montré des valeurs variables comprises entre 8 % et 12 %, inférieures à la spécificationminimale de ce contrat qui était de 15 %. Ce matériau a donc été considéré comme ayant des propriétés trop faibles pour sonutilisation envisagée. Cependant un tronçon d’essai de 100 mètres de longueur et 1 mètre d’épaisseur a été réalisé en utilisantle schiste comme matériau de remblai dans un environnement normal de conditions de trafic, et suivi pendant une période dedeux mois. Ce suivi a comporté des essais CBR en place, en utilisant le pénétromètre dynamique, et des essais dechargement à la plaque, avec des études pétrographiques en laboratoire. Les résultats obtenus lors de l'évaluation sur leterrain ont indiqué de hautes valeurs CBR et de portance, ainsi que des tassements insignifiants. Les indicateurs deperformance du revêtement tels que l’orniérage et la cavitation n'étaient pas apparents. Ces caractéristiques du schiste, tiréesde l’expérimentation sur site, ont donc permis son utilisation pour ce projet. Cet article présente les résultats des tests delaboratoire aussi bien que les études de terrain.

KEYWORD: MCA, Shale, embankment, fill, CBR, plate loading

1362


1 INTRODUCTION

The Millennium Challenge Account (MCA) Compact is anintervention muted by the United States of America tosupport the alleviation of poverty stricken areas in theGhanaian economy. The intervention focused mainly onagriculture infrastructure, which included some roadconstruction. The Agogo – Dome trunk road, 75.21km, wasone of such that received attention to the Afram Plains, amajor agriculture area. The project spanned between October2009 and October 2011.

The geology of the area is a well-defined regionalsedimentary terrain of the Voltaian System of Ghana, whichabounds in consolidated and unconsolidated sediments ormaterials with other low degree metamorphism associations[Ghana Geological Survey Department]. Locally the mainlithological units underlying the area are the Upper to LowerPaleozoic Voltaian Afram shale and sandstone. Lateriticdepositions (residual in nature) that are commonly used forroad fill, sub-base and base layers including embankment forthis project were scarce within the geological environmentduring the feasibility studies. Also earlier geotechnicalappraisal of materials within the project area was silent aboutthe shale.

The project special specification requested the use ofgeogrid membrane to provide reinforcement for soft andweak in situ materials, and lateritic materials of class G30 forfill and embankment, G40 for sub-base and G80 for basecourses. During the construction stages shale, which aboundsin the area was also evaluated in the laboratory withconventional test protocols such grading, Atterberg’s limits and soak California Bearing Ratio (96hr-CBR). Howeverlaboratory test results classified the shale as marginal materialto be utilized as a fill for road and embankment structures.Due to this marginal condition, there was the need to evaluateits performance from a trial section. A 100m length, 1.0mthick, trial section was constructed as part of the length of theroad under normal environmental and traffic conditions fortwo months. The field evaluation processes included dynamiccone penetrometer (DCP) test for CBR determination andplate loading for settlement and shear strength coupled withpetrographic and X-ray fluorescence laboratory studies. Fieldtest results were convincing enough for the shale to be used asintended.

This paper presents the laboratory and field resultsobtained. The observed field test results merited the use of theshale as fill and embankment structures for about 30.0kmstretch of the road.

1.1 Shale

Shale is generally a clastic water depositional materialcomposed chiefly of silt and clay. There are varyingclassification of shale depending on the mineral content,fossil content and depositional history.

The use of shale as road construction material is not verycommon. Sethi and Schieber (1998) have commented on theuse of the Ordovician Martinsburg Shale and DevonianBrailler Shale, in West Virginia, as lightweight (expanded)aggregate for concrete, brick, asphalt, railroad, ballast, roadbase and fill. Okogbue and Aghamelu (2010) also comparedthe geotechnical properties of three shale Formations fromsoutheastern Nigeria. In summary they concluded that theyare likely to be satisfactory as fill and embankment materials.

Afram Shale in the Eastern part of Ghana has not beenutilized as construction material probably due to perceivedgeotechnical challenges associated with Shale in general. TheShale within the project corridor is mostly moderatelyweathered and deep seated; physically observed to be friablewhen expose to the atmosphere.

1.1.1 Design Method for test section

The 100m long test section spans between, chainage,ch29+675 and ch29+775. The intended pavement design forsoft and weak zones of the road was to place a biaxial geogridreinforcement followed by a 200mm thick layer of selectedgranular material complying with the requirements formaterial class G30 of the standard specification. However thegeogrid was eliminated at the test section and insteadsubstituted with the 1.00m thick moderately weathered shale.Conventional procedure included clearing the vegetativecover and topsoil followed by 200mm thick lifts of the shale.Each lifts was compacted using the vibratory roller for twenty(20) passes and, visual indication that, close contact wasbeing obtained between aggregates. In order to check onadequate compaction proof-rolling was implemented. Atwenty ton fully loaded truck was engaged to move slowly onthe compacted surface and concurrently observing anymovements made by wheels (tyres) for the proof-rolling. Anyobserved movement was corrected by further compaction.This test section provided the platform as formation level forsub-base, base and bituminous layers at the referencechainage. The compacted section is presented at Figure 2.

1.1.2 Laboratory test results

Tests were carried out in five laboratories in Ghana and onelaboratory in Nairobi, Kenya while the construction was inprogress. The soil and aggregate results obtained from thelaboratories is shown in Table 1. Table 2 shows thepetrography of the Afram Shale and their major oxidescomposition obtained from X-ray fluorescence analysis isshown in Table 3.

Table 1. Summary of soil and aggregate laboratory results on AframShale.

Test

LabMDDg/cm3

LL%

PL%

PI%

CBR%

<425μm %

PM

1.88 30.9 19.9 11.0 11 16 1761

2.11 27.7 11.0 16.7 12 21 350.7

1.77 45.0 20.7 24.3 10 19 461.72

1.79 44.2 19.6 24.6 10 1.2 24.0

1.78 33.8 20.1 13.7 1 13.73

1.85 40.5 19.4 21.1 1 21.1

4 1.73 45.0 21.0 24.0 8 10 240.0

5 1.93 38.8 21.4 17.4 1 17.4

6 1.90 36.0 19.0 17.0 10 1 17.0

Reference laboratories: 1 – Ghana Highway Authority (GHA) Kumasi Lab,Ghana; 2 – Building and Road Research Institute, Kumasi, Ghana; 3 – NairobiLab, Kenya; 4 – GHA, Accra Lab, Ghana; 5 – China WE Suhum Lab, Ghana; 6– China Jiansu Jianda Corporation Lab, Agogo, Ghana.

1363


Table 2. Summary of Petrographic thin section result on Afram Shale(Source: Department of Geology, Univ. of Ghana).

Sample (%)

Mineralogy 1 2

Quartz 60 - 70 55 - 65

Feldspar 5 - 15 10 - 20

Clay 10 - 20 15 - 20

Organic Nil Nil

Table 3. Summary of X-ray fluorescence results based on majoroxides. Source: Ghana Geological Survey Department

Element Percentages

Na2O 1.39

MgO 1.52

Al2O3 8.57

SiO2 74.96

Fe2O3 2.63

CaO 0.26

others 2.57

L.O.I 8.10

1.1.3 Field test

The field test carried out on the 100m stretch were thedynamic cone penetrometer (DCP) based TRRL specificationand plate load bearing based on BS 1377:part9 1990

A total of eight (8) test points were carried out for theDCP between chainage 29+662 and 29+712. Summary ofresults is presented in Table 4.

Table 4. Summary of DCP test results

ChainageThickness,

mmAve CBR, %

29+625 131 150

29+637 43 150

29+650 103.5 140.42

29+662 35 150

29+675 221 130.33

29+687 154 150.14

29+700 69.8 150

29+712 73.5 145.07

Graphical presentation of plate load bearing test and typicalfield test performance are shown in Figure 1 and Figure 3respectively.

0

0.5

1

1.5

2

2.5

0 200 400 600 800 1000

Ave

rage

sett

lem

ent,

mm

Bearing Pressure, kN/m2

Load - settlement curves for plate bearing test

Ch29+637 Ch29+662

Ch29+675 Ch29+712

Figure 1. Plate load bearing graphs

1.1.4 Discussion

For quality assurance and statistical representation purposes,the shale was tested in six different laboratories to evaluate itsgeotechnical properties in order to understand its intendedutilization. The conventional laboratory tests that arenormally used for classifying materials in pavement worksare gradation, Atterberg’s limits, plasticity modulus and CBR.

Atterberg’s limits results, in Table 1, indicate that finecontent is mostly clay and inorganic of low to mediumplasticity. The plasticity modulus (PM) as per the contractspecification stated that values of, at most, 200 are assignedas base course; those between 200 and 250 are assigned sub-base course. The PM values obtained from the laboratoriesvaried between 13.7 and 461.7. These observed valuesindicate that the shale could be used as base, sub-base and fillmaterials. The CBR, four day soaking, values obtained rangebetween 8% and 12%, which falls below the expected 15%contract specification for fill.

During the construction stages an exposed section of theshale material on the road corridor, under traffic andenvironmental conditions, gave an interesting outlook. Thisprompted the implementation and evaluation of the trialsection using conventional road construction methods.

From Table 2, the petrographic analyses revealed thatthe subject material is mostly siliceous. The study revealedthat the quartz had been re-worked, an indication of atransported clastic sediment or an occurrence of a low grademetamorphism of the material. The high SiO2 content ofabout 75% shown in Table 3 and the low clay content indicatethe siliceous nature of the shale material.

The eight DCP test points were selected randomly withinthe 100m stretch. CBR values were obtained using the Kleynand Van Harden equation below.

Log (Cbr) = 2.628 – 1.273log (DCP)

Table 4 presents the DCP test results varying thicknesses forCBR evaluation. The thickness column, in the Table 4,represents total thicknesses run-by each DCP test at thevarious chainage. CBR values represent average values

1364


within the total thicknesses. Table 4 also shows that withinthe chainage explored CBR values estimated were very highand these were confined between the upper 35mm and221mm of the 1.0m thick layer. This is an indication of highblow counts as well as refusal to penetration.

Four plate loading tests were carried out within the100m stretch. Results are presented in Figure 1. Generally,the modulus of subgrade reaction (k – kN/m3) is the ratio ofloading stress (p) at 1.25mm average settlement to 1.25mm.The loading stress is determined from the load – settlementgraph at average settlement of 1.25mm. It is observed fromFigure 1 that 3 out of 4 (75%) of the test did not attain the1.25mm settlement criteria at the ultimate stress of 900kN/m2.However at chainage 29+637, k value is estimated as272MN/m3.This value may be considered as lower boundvalue for the test section and gives an indication of very lowor negligible settlement of a compacted layer. The graphs alsoindicate that no distinct shear failure occurred and so the finalload could be considered as the ultimate load. The finishedroad of the trial section is presented in Figure 4.

Figure 2. Compacted shale at trial section

Figure 3. Plate load bearing test at trial section

Figure 4. Finished road at trial section

2 CONCLUSIONS

The four day soaking CBR values from the laboratories werebelow required specification. However field implementationindicated a very good geotechnical strength measure for theshale as fill and embankment material within the 100m testsection for the two months observations. The performancecharacteristics of the shale encouraged its utilization as fill for30km stretch of the road including a 2.0km length, 5m high,of embankment. For nearly two and half years, including thedefect liability period, that the road has been in service therehas been continuous monitoring and observations. Pavementfailure indicators such as rutting and potholing have not beenobserved as well as significant shear failures in embankments.

3 ACKNOWLEDGEMENT

The writers acknowledge Millennium DevelopmentAuthority, Ghana, the project implementation authority; theproject consultant, Gauff Ingenieure (GmbH); the six soil andaggregate laboratories; the department of geology, Universityof Ghana, Legon and Ghana Geological Survey Department.The Ghana Geotechnical Society is also acknowledged for itstechnical advice.

4 REFERENCES

Geological Survey Department. 2009. Geological mapexplanation – map sheets 0601A/4, 0601B/3 & 0601D/1,0601D/2 & 0601D/3. 115 – 117.

Sethi, S. and Schieber, J. 1998. Economic aspects of shalesand clays: an overview. Chapter 7. 14

Okogbue, C.O. and Aghamelu, O.P. 2010. Comparison of thegeotechnical properties of crushed shales from easternNigeria. Bulletin of engineering geology and theenvironment.

1365

Influence of Mechanical Indices for Soil Basement on Strength of Road Structure

Influence des paramètres mécaniques de la couche de fondation sur la résistance d’une structure de chaussée

Teltayev B. JSC “Kazakhstan Highway Research Institute”, Almaty, Kazakhstan

ABSTRACT: It has been observed that at present insufficient attention has been paid to the consideration of moisture and mechanicalindices variety for soil basement while designing of highways. Peculiarities have been analyzed for moisture variety of soil basement for the highway “Astana-Burabai”. Peculiarities for indices variety of stress-deformed condition of road structure with cement concrete pavement have been calculated and determined.

RÉSUMÉ : Il a été observé que les variations de teneurs en eau et des paramètres mécaniques des couches de base sont généralementinsuffisamment prises en compte lorsqu’on dimensionne des chaussées. Des particularités ont été analysées pour différentes teneurs en eau du sol de fondation de la route "Astana-Burabaï". Des variations des paramètres de comportement contraintes/déformations de la couche de fondation ont été également prises en compte et analysées lors du dimensionnement de la chaussée en béton de cette route.

KEYWORDS: highway, soil basement, cement concrete pavement, moisture, deformation, stress.

1. INTRODUCTION

Main elements for modern highways are pavement and soil basement. They have been operated jointly in the highway. Therefore while designing of highways calculation for pavement and soil basement strength has been carried out jointly (SN RK 2005).

As mechanical impact from car wheels, as well as climatic factors (temperature, moisture and so on) has been made over longevity and strength of road structures. Moisture of soil basement can be one of the main climatic factors, influencing greatly over highways with cement concrete pavement.

The following assumptions have been made in Kazakhstan present standard provisions for calculation of rigid pavements for strength (SN RK 2005):

- maximum moisture of soil basement has been observed in spring, when its complete thawing occurs, and the weakest condition of road structure complies with that time moment;

- theoretically soil basement has constant moisture value during overall service period for road structure (from 16 to 20 years for highways of I-III technical categories).

Fundamentals of present method of road structures for strength (SN RK 2005) were established in 60s-70s of the last century. At that time the above assumptions were accepted to simplify calculations. At present there are high performance computers and powerful software’s, which allow to carry out very complicated calculations, considering seasonal moisture varieties, therefore, deformation and strength characteristics of soil basement.

2. HIGHWAY “ASTANA-BURABAI”

Express highway “Astana-Burabai” connects the capital of Kazakhstan, Astana city, with resort zone Burabai, and is a part of international automobile route “Atana-Kokshetau-Petropavlovsk-border of Russian Federation”. Total length of the road is 224 km. It has 6 lanes with the width of 3.75 m each. Start (50 km) and final (47 km) road sections have cement concrete pavement, road section in the middle of the road (127

km) is pavement from crushed stone mastic asphalt concrete. Reconstruction of the road was fully completed in November 2009.

Pavement structure consists of: cement concrete, 25 cm; crushed stone optimum mix, 25 cm; coarse-grained sand, 30 cm.

To investigate the peculiarities of temperature and moisture variety temperature and moisture sensors were placed into soil basement of that road section in October 2010. Depth for placement of those sensors from the surface of cement concrete pavement was 80, 115, 150, 185 and 220 cm.

Soil basement of the road consists of light pulverescent clay loam: moisture at the border of rolling Wр = 15.0 %; moisture at the border of fluidity Wт = 25.4 %. Granulometric content of soil has been shown in the Table 1.

Table 1. Granulometric content of light pulverescent clay loam

Full residue (%) in sieves with size of cells (mm)

2.0 1.0 0.5 0.25 0.1 0.05

6.8 8.4 10.9 17.3 28.2 29.4

Underground waters lay in the depth (below than 3.0 m

from the land surface).

3. MOISTURE OF SOIL BASEMENT

Diagrams have been represented in the Figures 1 and 2, showing moisture variety during time period in different points of soil basement. Analysis of the mentioned diagrams has shown that character of moisture variety in various depths of soil basement is different. The surface of soil basement (80 cm) has proved to be very sensitive to seasonal climate variety: the biggest values of moisture (21-23 %) have been observed in spring and autumn seasons of the year, and in summer, due to the dry weather moisture reduces to 8-8.5 %, which can be explained by transferring of part of the liquid moisture into soil condition (ice) with negative temperatures.

1366


Figure 1. Moisture variety in soil basement of the highway “Astana-Burabai”, road section with cement concrete pavement (2010-2011)

Figure 2. Moisture variety in soil basement of the highway “Astana-Burabai”, road section with cement concrete pavement (2010-2011)

Moisture increase in the depths of 115 cm and 150 cm occurs in summer and autumn seasons of the year (17-22 %) compared with spring (10-13 %). Moisture reduces to 6-9 % in winter season.

Moisture has not been reduced evidently in the depths of 185 cm and 220 cm in winter season. It can be explained by the fact that in the depths, mentioned above, negative temperature has small values. Moisture value was 15-16% in the depth of 220 cm in autumn 2010, moisture increase up to 20-21% occurred in spring 2011, which in future was practically constant. Moisture was of similar character of variety in the depth of 185 cm: moisture was 6-7% in autumn 2010; it was increased to 12% in spring 2011, and was practically constant in summer and autumn 2011. One can see from Figure 2, that the difference in moisture values of soil in the depth of 185 cm and 220 cm (8-10 %) during overall annual cycle is practically constant. Meanwhile, moisture in the depth of 220 cm is larger than in the depth of 185 cm, which can be explained by close location of the depth in 220 cm to underground waters.

Qualitative and quantitative analysis, carried out above, for moisture variety in points, located in different depths, has shown complicated character and non-homogeneous of moisture distribution in soil basement of the highway. In such cases it is convenient to use averaged moisture indices. Therefore, Figure 3 represents diagram, showing average moisture variety for soil basement of the highway. It reflects the following important peculiarities for moisture variety of soil basement according to its depth and time:

- it represents average value of liquid moisture at any reviewed time moment;

- moisture decrease occurs on winter season due to transfer of liquid moisture into solid condition (ice) with negative temperatures;

- initial moisture, observed in autumn 2010,with coming spring in 2011 has gradually been increased and further (in summer and autumn) remains practically constant.

4

6

8

10

12

14

16

18

20

22

24

01.1

0.20

10

31.1

0.20

10

30.1

1.20

10

30.1

2.20

10

29.0

1.20

11

28.0

2.20

11

30.0

3.20

11

29.0

4.20

11

29.0

5.20

11

28.0

6.20

11

28.0

7.20

11

27.0

8.20

11

26.0

9.20

11

26.1

0.20

11

25.1

1.20

11

Months

Moi

stur

e, %

- 80 сm; - 115 сm; - 150 сm

0

2

4

6

8

10

12

14

16

18

20

01.1

0.20

10

31.1

0.20

10

30.1

1.20

10

30.1

2.20

10

29.0

1.20

11

28.0

2.20

11

30.0

3.20

11

29.0

4.20

11

29.0

5.20

11

28.0

6.20

11

28.0

7.20

11

27.0

8.20

11

26.0

9.20

11

26.1

0.20

11

25.1

1.20

11

Date

Moi

stur

e,%

Figure 3. Average moisture variety in soil basement of the highway “Astana-Burabai”, road section with cement concrete pavement (2010-2011)

4

6

8

10

12

14

16

18

20

22

24

01.1

0.20

10

31.1

0.20

10

30.1

1.20

10

30.1

2.20

10

29.0

1.20

11

28.0

2.20

11

30.0

3.20

11

29.0

4.20

11

29.0

5.20

11

28.0

6.20

11

28.0

7.20

11

27.0

8.20

11

26.0

9.20

11

26.1

0.20

11

25.1

1.20

11

Months

Moi

stur

e

- 185 сm; - 220 сm

4. STRENGTH-DEFORMED CONDITION OF ROAD

STRUCTURE

4.1. Mechanical characteristics

Mechanical characteristics for materials of pavement layers have been represented in the Table 2.

Table 2. Calculated values of elasticity modulus and Poisson’s coefficient for pavement layers

Material Thickness h, cm

Elasticity modulus Е, МPa

Poisson’s coefficient

Cement concrete 25 37500 0.2

Crushed stone sand mix С 6 (40)

25 200 0.3

Coarse-grained sand

30 120 0.3

Mechanical characteristic of soil basement vary for a

year. Their calculated values (Table 3) has been determined considering average moisture variety of soil basement (Figure 3). Meanwhile elasticity modulus value for soil in winter season has been determined based on the results of Prof. N.A.Tsytovich (Tsytovich 1973), and during other seasons of the year – based on recommendations of standard provisions (SN RK 2005). Due to the absence of reliable data, Poisson’s coefficient value for soil has been accepted constant and equal to 0.35.

4.2. Calculated scheme and mathematical model for road structure

Calculated scheme for road structure has been represented in the Figure 4. This figure shows components for stress-deformed condition of road structure, which causes its destruction (sagging of the surface for cement concrete pavement ℓ, tensile stress r and tensile deformation r in cement concrete pavement, vertical compressed deformation for the surface of soil basement z0).

1367


Table 3. Calculated values of elasticity modulus for soil – light pulverescent clay loam

Calculated date

Average moisture

W, %

Average temperature,

0С

Elasticity modulus Е, МPa

19.11.2010 13.9 7.0 90

24.01.2011 9.5 -4.6 6500

26.02.2011 9.4 -5.2 7200

14.04.2011 13.7 -0.4 500

16.06.2011 16.3 12.3 58

29.09.2011 16.9 14.7 51

Note: Only content of unfrozen moisture has been shown in winter season

Figure 4. Calculated scheme of road structure

Mathematical model of multi-layer elastic half-space under influence of axle-symmetric load has been used to calculate the above components for stress-deformed condition of road structure (Privarnikov 1973). Load q = 0.6 МPa within circle with diameter of D = 42 cm influenced over the surface of concrete pavement. Such load corresponds to axle load of the truck Q = 13 tons.

4.3. Stress-deformed condition

The results of calculation of indices for stress-deformed condition of the road structure have been represented in the Figures 5 and 6. Analysis of the obtained dependencies has shown that sagging of the surface for cement concrete pavement, tensile stress and vertical deformation on the surface of soil basement have qualitatively similar character of variety in annual cycle. Thus, their least values have been determined in winter months (January, February), when upper part of soil basement is in the most frozen condition (average temperature in January and February is -4.6 0С and -5.2 0С respectively), due to this fact soil basement has the increased rigidity (ranging from 6500 MPa to 7200 MPa). With coming of warm season the rigidity of soil basement reduces, which results in the value increase of the above indices for stress-deformed condition (SDC) of pavement. Their biggest values were obtained during the period from June to September. Their values, obtained in the month of November, proved to be comparable with those ones obtained during the period from June to September. Therefore, the reviewed case shows that sagging of cement concrete pavement and vertical deformation of the surface for soil basement, during long period in annual cycle – from June to November – preserves the biggest values, which are higher in 6.0-6.5 times, compared with winter season.

020406080

100120140160

01.1

0.20

10

31.1

0.20

10

30.1

1.20

10

30.1

2.20

10

29.0

1.20

11

28.0

2.20

11

30.0

3.20

11

29.0

4.20

11

29.0

5.20

11

28.0

6.20

11

28.0

7.20

11

27.0

8.20

11

26.0

9.20

11

26.1

0.20

11

▲ - r10-6; ● - z0 10-6

Figure 5. Tensile deformation (r) in cement concrete pavement and compression deformation (z0) on the surface of soil basement

● - ℓ, мм; ▲ - r, МPа

00,20,40,60,8

11,21,41,61,8

01.1

0.20

10

31.1

0.20

10

30.1

1.20

10

30.1

2.20

10

29.0

1.20

11

28.0

2.20

11

30.0

3.20

11

29.0

4.20

11

29.0

5.20

11

28.0

6.20

11

28.0

7.20

11

27.0

8.20

11

26.0

9.20

11

26.1

0.20

11

Figure 6. Sagging (ℓ) and tensile stress (r) in cement concrete pavement

It should be mentioned that tensile stress value varies little in annual cycle. Thus, tensile stress value (1.6 MPa at average), occurring during overall summer-autumn seasons, is bigger only for 19 % of its least value, occurring in winter season (1.3 MPa at average). It can be explained by the fact that with big rigidity of cement concrete slab (Е=37 500 MPa), which is bigger than rigidity of soil basement in 735 times in autumn season (Е=51 MPa), rigidity increase in winter in 141 times (up to 7200 MPa) has not greatly influenced over tensile stress value.

Character of tensile deformation variety in cement concrete pavement differs substantially from the indices analyzed above. Tensile deformation value varies little in annual cycle (2.9 – 3.510-5), which also can be explained by big rigidity of cement concrete slab.

It should be mentioned that cement concrete slab has been calculated for sagging of pavement in spring, and sagging increase in summer and autumn seasons can be the cause of destruction of cement concrete slabs in the form of longitudinal, transversal and oblique cracks (Figures 7-9).

5. CONCLUSIONS

The results of experimental investigation for moisture variety of soil basement and calculations for indices of stress-deformed condition for road structure, obtained the work, allow making the conclusions as follows:

1. Moisture value and its phase content in soil basement of the highway vary substantially in annual cycle and according to the depth of basement.

2. Qualitative character of variety for sagging, tensile stress in cement concrete pavement and vertical compressed deformation of the surface for soil basement in annual cycle

1368


4. Calculated values for mechanical characteristics of soils should be determined in laboratory and field conditions.

5. Carrying out of profound analysis for stress-deformed condition of road structures using reliable calculated values for mechanical characteristics of soils and considering value variety and phase content of moisture for soil basement in different climatic conditions is important for practical purposes.

REFERENCES

SN RK 3.03-34-2006. 2005. Instructions for design of non-rigid pavement. Astana.

Tsytovich N.A. 1973. Mechanics of frozen soils. Moscow. Privarnikov A.K. 1973. Volumetric deformations of multilayer

basement. Stability and strength of structural elements. Dnepropetrovsk, 27-45.

Figure 7. Longitudinal crack in cement concrete pavement of the highway “Astana-Burabai”

Figure 8. Transversal crack in cement concrete pavement of the highway “Astana-Burabai”

Figure 9. Oblique crack in cement concrete pavement of the highway “Astana-Burabai”

is similar. Their maximum values occur in summer and autumn seasons, which are substantially bigger than in spring calculation season and can be the cause for appearing of longitudinal, transversal and oblique cracks in cement concrete pavement. Tensile deformation of cement concrete pavement varies little in annual cycle.

3. The issue of further investigation of peculiarities for moisture variety of soil basement for highways in different climatic conditions is topical.

1369

Design and performance of a jet grout retaining wall in a railway embankment on soft soil

Dimensionnement et performance d’une paroi de soutènement réalisée à l’aide de la technique de jet grouting dans un remblai ferroviaire sur sol mou

Verstraelen J., Maekelberg W., Lejeune C., De Clercq E. TUC RAIL

De Vos L. Geotechnics Division Flemish Government

ABSTRACT: Within the framework of the regional express network around Brussels, two new railway tracks will be added besidestwo existing tracks linking Brussels to Ghent. The existing railway crosses the river Senne in Brussels with a geotechnical complex lithography. Beside this complexity, other project boundary conditions, such as the height of the railway embankment, the very shortdistance to the existing railway tracks in service, the small work platform to realize the retaining wall and safety issues linked toworking next to railway tracks in service, were of great influence on the design of the retaining wall. The excavation has a total depth of 12.0m. The retaining wall consists of grout columns reinforced with steel beams. The wall is secured by grouted nails. The jet grout columns were monitored and the resulting displacements and bending moments compared well with the designcalculations.

RÉSUMÉ : Dans le cadre du projet du réseau express régional autour de Bruxelles, deux nouvelles voies de chemin de fer serontajoutées le long des deux voies existantes reliant Bruxelles à Gand. Le chemin de fer croise la Senne, rivière de Bruxelles, dans unelithographie géotechnique complexe. Outre cette complexité, d’autres conditions limites du projet, comme la hauteur du remblaiferroviaire, la très courte distance par rapport aux voies en services, l’étroitesse de la plateforme de travail pour la réalisation de la paroi et les consignes de sécurité liées au travail à proximité de voies de chemin de fer en service, ont eu une grande influence sur ledimensionnement de cette paroi de soutènement. La profondeur d’excavation a atteint 12,0m. La paroi de soutènement est constituée de colonnes de jet-grouting armées à l’aide de poutrelles métalliques. La paroi est retenue par différents lits de clous de jet-grouting. Les colonnes de jet grouting ont été suivies par monitoring et les résultats des déplacements et des moments fléchissants ont été comparés avec les résultats des calculs du dimensionnement.

KEYWORDS: VHP jet grout columns, retaining wall, soil nails, railway embankment, alluvium, peat, monitoring.

1 INTRODUCTION

The regional express network around Brussels aims to increase the capacity of the train traffic in and around Brussels. Therefore, the existing railway lines linking Brussels to the most important surrounding cities, have to be widened from 2 to 4 railway tracks. The enlargement of the railway platform has to be done in highly urbanized zones and with restrictive project constraints. These constraints led in some cases to very complex and innovative solutions.

The retaining wall discussed in this paper is situated along the existing railway line 50A linking Brussels to Ghent, near to the Senne river at the south of Brussels.

A retaining wall is needed to create an abutment for a new integral arch bridge which is situated next to an existing bridge abutment. The excavation removes the existing embankment fill and goes up to 1 m underneath the natural ground level, as shown in figure 1.

The realisation of the retaining wall for the abutments is restricted by several constrains such as: Railway traffic must remain undisturbed during the whole

construction period, Excessive movement of the railway tracks in service must

be avoided and if they occur, they must be rectified directly; these movements can lead to reduced exploitation speeds and passenger delays,

For safety reasons, the use of large foundation machinery is very much restricted next to the railway tracks in service,

No expropriation was possible, so a working platform had to be created within the embankment contours,

Since the embankment is situated on highly compressible alluvium, an enlargement or steepening of the embankment to create a larger working platform could not be realised.

Figure 1. Cross section of the retaining wall

Due to these constraints, in a first working phase, a Berliner wall with 3 m retaining height was executed close to the tracks at the top of the embankment in order to create the working platform that will be used to install the retaining wall to the final excavation level, 12.0 m beneath rail level (see figure 1).

1370


The use of secant piles wall, diaphragm walls, sheet piles etc was excluded for safety reasons, so a VHP-jet grout wall was chosen as retaining wall.

Since both the Berliner wall and VHP-wall are closely together and influence each other, their global design was carried out simultaneously using the finite difference program FLAC.

2 SITE GEOLOGY

The site is located within the alluvial basin of the river Senne, and contains highly compressible alluvium up to a depth of 10 m. Underneath this alluvium, a gravel layer is situated with a thickness ranging from 2 to 5 m. The deeper tertiary deposits consist of Yperian clay, an over-consolidated clay with a thickness of up to 16 m. Figure 2 shows the results of an electrical CPT together with the results of a pressiometer test at the same location. The Ménard modulus from the pressiometer test shows the different degree of consolidation in the upper clay layer from +8.0 to +4.0 mTAW and in the lower clay layer.

Figure 2. Cone resistance from CPT-E and Ménard modulus from pressiometer.

3 DESIGN

3.1 Site investigation and soil parameters

Besides the extensive in situ site testing, such as cone penetration tests (CPT’s), pressiometer tests and core drillings, also extensive laboratory testing was carried out on undisturbed samples. The results of these tests were used to determine the effective strength parameters summarized in Table 1. A number of CPT’s were performed directly from the train tracks and through the embankment fill. The railway embankment is a poorly compacted and silty fill.

3.2 Finite difference model

The geometry as shown in figure 1 and all of its construction phases was modelled with the finite difference program FLAC.

Special consideration was given to the compressibility of the alluvium. To model the stress dependant stiffness of this layer, a “double yield” soil mechanical model was used. The double yield model allows for plastic volumetric strain hardening, although independent of shear strain level. Also, unloading/reloading is taken into account with a user specified constant ratio between loading/unloading stiffness. The model is specifically designed for the use of isotropic compression tests, but the results of an oedometer can be converted to fit the input parameters. Very specific to this model is that a table of values

serves as input, and the model interpolates linearly between these values. As a result, the (converted) oedometric test results can serve as direct input. To check the model, a separate load test model was used to check the input versus the modelling results (Figure 3). Since it is very difficult to combine the results of different oedometer tests, one representative oedometer was chosen as single input. T able 1. Soil parameters.

Parameter alluvium gravel Ycclay

fill

coh. (kPa) 5 0 20 4

friction (°) 22 32 25 27

EM (MPa) 2.5 10 27 /

pl (MPa) 0.5 1.9 1.6 /

Alluvial Clay

Gravel Layer

Yperian Clay

Figure 3. Comparison of model response with data from reference oedometer test.

The jet grout wall was modelled as a continuous wall with interface friction angle δ equal to the internal friction angle of the surrounding soil, and with the stiffness and strength of only the beam reinforcement of the jet grout piles. The wall is stabilized with jet grout soil nails, which are modelled as pile elements (with tensile and flexural strength) with an interface cohesion (cu = qsu based on literature values).

To limit excessive swelling of the embankment, a simple double yield model was used with constant loading and reloading moduli.

3.3 Calculation results

The model showed that it was necessary to install the VHP jetgrout piles till the depth of the base gravel. Otherwise, vertical settlement of the wall would reverse wall interface friction and would lead to excessive horizontal displacement. Also, squeezing of the alluvium would occur when the embedded length is further reduced.

Table 2 summarizes some results of the design calculations.

4 EXECUTION

The Berliner wall was placed with an excavator mounted vibratory pile driver and anchored to an opposing Berliner wall with tiebacks between the rail sleepers as shown in figure 1. T able 2. Results of calculations

1371


Bending moment embedded 63 kNm

Bending moment retaining 65 kNm

Horizontal displacement 21 mm

Vertical displacement top of wall 4 mm

Max vertical displacement behind wall 9 mm

Load in nails row 1 140 kN




After excavation of a retaining height of 3 m, a guide wall for the jet grout columns was casted against the berliner’s soldier piles (HEB 300 beams). This guide wall also serves as temporary waler, since execution of jet grout columns decreases significantly the passive soil resistance in front of the soldier piles. To counter this effect, the installation of the jet grout columns was also carried out in a specified alternating sequence. guide wall also serves as temporary waler,

Figure 4 shows the finite difference model geometry with bending moments in the wall and axial loads in the nails.

Figure 4. The distribution of the nail forces along the nails and bending moments within the jet grout wall.

The VHP jet grout columns only serve as a mean to install the beam reinforcement at depth and to transfer earth and water pressure to the reinforcement beams. For safety reasons, the 21 m long reinforcement beams (HEB 280) had to be installed in 3 m long sections which were bolted together. This installation is tedious and time consuming, and has to take place before the grout starts to harden. Due to the column length, installing the beams can only start about 2 h after commencing the (water)pre-cutting and 1 h after the (watercement) grouting. Including the bolting of the different beam sections, which takes up about 45 min in total, the last section reaches the bottom of the column 2 h after the start of the jet grouting at that depth.

Furthermore, the VHP-piles are installed in alluvial soil. Although the strength of the grout was not an issue (since it is of minor importance in the design), it is challenging to realise a reasonable sized jet grout column in this alluvium, especially when peat is encountered. Test columns were installed prior to the wall installation, in which diameter measurements were carried out with a calliper in the wet column. During these tests, chunks of more than 10 cm diameter of compacted peat were found in the spoil (Figure 5). To aid in the realisation of the required diameter to install the beam reinforcement, a reamer of

30 cm diameter was placed above the jet nozzles. Even with this reamer, the minimum diameter realized was equal to that of the reamer (Figure 6).

Figure 5. Pieces of peat found in jet grout spoil.

Figure 6. Cone resistance from CPT-E and measured diameter in wet jet grout column.

After testing different jetting pressures, flow rates and nozzle diameters, a suitable set of parameters was chosen. Even with this most suitable set, it was difficult to install the last few meters of the reinforcement, due to decantation of soil inclusion (clay/peat) in the grout. To ease the installation, the jet grout columns were deepened 1 m to allow for a 1 m unreinforced length. The retaining wall was executed as primary (reinforced) and secondary (unreinforced) columns with hart-to-hart distance of 1 m between reinforcements. The secondary columns serve to fill up the gap between primary columns.

Due to the nature of the fill and natural soil, a larger diameter than conventional soil nail diameters was necessary to provide sufficient bearing capacity of the nails. The soil nails were also executed as jet grout nails with a diameter of 30 cm. Pull-out tests were performed on sacrificial nails to check the design assumptions.

Figure 7 gives a view on the retaining wall after excavation and creation of a working platform for the new pile foundation.

1372


Figure 7. Photo of retaining wall after final excavation phase, after backfill for piling rig

5 MONITORING

An extensive monitoring program was set up to measure the vertical and horizontal displacements of the wall during the different excavation stages. Specially instrumented beams were placed in the columns to measure deformations and strains in the reinforcement (L. De Vos, 2013). These reinforcement beams were exceptionally put into place in one piece and installed when the train tracks were out of service.

Figure 9. Comparison between calculated and measured (derived) bending moments.

6 CONCLUSION

The measurements lie in close approximation to the calculated values. Figure 8 compares the measured and calculated horizontal displacements of the wall. Displacements are measured with an inclinometer, in which the bottom measurement is considered to remain unchanged.

For the construction of a retaining wall adjacent to railway tracks, the restrictions in available space and allowable height in machinery led to a combined retaining wall consisting of a small Berliner wall and deep VHP jet grout wall. The design was based on different in situ and laboratory tests, and was checked through monitoring of the excavation and performing preliminary true scale measurements. The execution of jet grout piles turned out to be difficult due to peat layers and the installation of 21 m long reinforcement beams in 3 m long bolted sections was challenging. Nevertheless, execution difficulties could be resolved by taking special measurements to ease the installation of the reinforcement beams. A monitoring campaign showed that the resulting retaining wall performed close to the design and train traffic remained undisturbed during the works.

7 ACKNOWLEDGEMENTS

The author whishes to acknowledge the Geotechnical Division of the Flemish Government, which allowed for this extensive monitoring campaign.

8 REFERENCES

Figure 8. Comparison between calculated and measured horizontal displacements.

De Vos L., Van Alboom G. and Haelterman K. 2013. Comparison of monitoring techniques for measuring deformations in an excavation. Proceedings of the 18th International Conference on Soil Mechanics and Geotechnical Engineering, Paris.

Based on these inclinations, bending moments can be calculated as the second derivative (Figure 9). Only the stiffness of the steel reinforcement beams was used, as was the case in the design.

Verstraelen J., 2011, Green terraces for the regional express network in Brussels, EYGEC Rotterdam

Behind the wall, vertical displacements of up to 5 cm were measured, which is considerably more than the calculated values. When comparing the construction phasing with the continuous measurements, it was clear that the main part of the settlements could be related to the execution of the jet grout nails. Since the nails are realized by the jet grouting technique, soil is firstly cut away with water and further on replaced by a mixture of cementgrout and soil. This mixture takes a certain time to harden in which unconfined convergence of the drilled hole can occur. Once the mixture is hardened, the settlements stop. This effect will be investigated later in further detail.

Verstraelen J., Lejeune C., De Clercq E., 2012, Realisation of integrated steep landscape slopes within existing railway embankments ISSMGE- TC 211 International Symposium on Ground Improvement, Brussels, IV-169 – IV-179.

Maekelberg W., Verstraelen J., De Clercq E., 2012 Realization of a railway enlargement in unstable excavations alongside the existing line at Dilbeek (Belgium) ISSMGE- TC 211 International Symposium on Ground Improvement, Brussels, IV-107 – IV-121.

Van Alboom G., De Vos L., Haelterman K. and Maekelberg W. 2012. Innovative monitoring tools for on line monitoring of building excavations. A monitoring test site. ISSMGE- TC 211 International Symposium on Ground Improvement, Brussels, IV-327 – IV-338.

1373

Laboratory characterization and model calibration of a cemented aggregate for application in transportation infrastructures

Caractérisation en laboratoire et calibration d'un modèle d'agrégat cimenté pour une utilisation dans les infrastructures de transport

Viana da Fonseca A., Rios S., Domingues A.M., Silva A. University of Porto, Dep. of Civil Engineering – FEUP, Porto

Fortunato E. National Laboratory for Civil Engineering - LNEC, Lisbon

ABSTRACT: Research on increasingly stiffer and more resistant artificially stabilized geomaterials, such as soil-cement mixtures has frequently revealed interesting properties. The knowledge of such materials behaviour is as important as they are increasingly used inseveral layers of transportation infrastructures, as well as in transition zones between embankments and rigid structures. Most of theselast situations involve zones close to sensitive prefabricated structures, where compaction of soils or aggregates demand for moderateenergies, being necessary to increase the content of the hydraulic binders to increase their stiffness and strength. The present workreports some of the most notorious results obtained in some laboratory studies aiming to characterize different mixtures of cement and limestone aggregate. Seismic wave measurements, indirect tensile strength tests and triaxial compression tests were performed. Theresults indicated some relevant differences on dynamic and static stiffness properties and shear strength Mohr-Coulomb parameters, directly associated to the variation of porosity/cement ratio. Based on the triaxial test results, a calibration of the geo-mechanical parameters of the Hardening Soil Model available on commercial software was made.

RÉSUMÉ: La recherche sur des géomatériaux de plus en plus rigides et plus résistants artificiellement stabilisés, comme les mélanges sol-ciment, a souvent révélée des propriétés intéressantes. La connaissance du comportement de ces matériaux est importante car ils sont de plus en plus utilisés en plusieurs couches dans les infrastructures de transport, ainsi que dans les zones de transition entre remblais et structures rigides. Dans la plupart de ces dernières situations, on trouve des zones sensibles proches de structures préfabriquées, où le compactage des sols ou d'agrégats doit être réalisé à énergie modérée. En conséquence, il est nécessaire d'augmenter la teneur en liants hydrauliques pour augmenter leur rigidité et résistance. Ce travail présente des résultats remarquables obtenus dans certaines études en laboratoire visant caractériser différents mélanges de ciment et de granulats calcaires. Des mesures d'ondes sismiques, des essais de résistance à la traction indirecte et des essais de compression triaxiale ont été réalisés. Les résultats ont montré des différences intéressantes sur les propriétés de rigidité statique et dynamique aussi bien que sur les paramètres de résistance au cisaillement de Mohr-Coulomb, directement liées à la variation du ratio porosité/ciment. Sur la base des résultats d’essais triaxiaux, une calibration des paramètres géo-mécaniques du Hardening Soil Model, disponible sur logiciels commerciaux, a été réalisée.

KEYWORDS: Aggregate-cement mixtures, Hardening Soil Model, Parametric calibration, Porosity cement ratio.

1 INTRODUCTION

The research on increasingly stiffer and more resistant artificially stabilized geomaterials, such as aggregate-cement mixtures, has frequently revealed interesting properties. The knowledge of such materials behaviour is as important as they are regularly used in several layers of transportation infrastructures, as well as in transition zones between embankments and rigid structures. Most of these last situations involve zones close to sensitive prefabricated structures, where compaction of soils or aggregates demand for moderate energies, being necessary to increase the content of the hydraulic binders to increase their stiffness and strength.

Despite the widespread use of Portland cement in the improvement of soils and aggregates, there seems to be no dosage methodologies based on rational criteria.

However, the relationship between the porosity of the mixture (n) and the volumetric cement content (ie, the ratio between the cement volume and the total volume - Civ) adjusted by an exponent x, (x Є [0, 1]) has become a good parameter to evaluate the strength and stiffness of artificially cemented soils.

This parameter, designated as adjusted porosity/cement ratio (n/Civ

x) has been related with the compressive strength determined in uniaxial compression tests (Consoli et al., 2007) and with the parameters of strength and deformability obtained in triaxial compression tests (Consoli et al., 2009). More recently, the ratio was applied to the stress–dilatancy relation of

an artificially cemented sand (Rios et al., 2012) and even more recently in stress–strain and strength-dilatancy relationships on a cemented aggregate (Viana da Fonseca et al. 2012).

The present paper reports some of the most notorious results obtained with the scope of optimization of mixtures of aggregates and Portland cement. A laboratory program was developed to define the geomechanical characteristics of those mixtures, which includes indirect tensile strength tests, seismic wave measurements and triaxial compression tests. Based on the results obtained, the relationships between the mechanical properties and the n/Civ

x parameter were evaluated. A constitutive law was calibrated, taking into account the behaviour of these mixtures - in laboratory tests, and then evaluated in numerical modelling of triaxial compression tests.

2 LABORATORY TESTS

2.1 Tested materials

The aggregate tested is a well graded material which grain size is shown in Figure 1. This material has a plasticity index of 10% and a liquid limit of 22%. The Los Angeles Abrasion Index is 30%. The maximum dry unit weight obtained by the Modified Proctor test is 21.4 kN/m3 and the corresponding optimum water content is 6.6%. This aggregate was mixed with different percentages of cement, namely 2%, 3%, 4% and 5%. Portland cement of very high initial strength (CEM I 52.5 R) was used as

1374


binder. The maximum dry unit weight determined by the Modified Proctor test ranged between 21.3 and 21.8 kN/m3 and the optimum water content ranged between 6.8 and 7.2%, for the four mixtures.

Figure 1. Grain size distribution curve of the aggregate without cement.

This study sought to evaluate the mechanical characteristics of mixtures with low and medium compaction. This kind of materials is usually placed in difficult compaction zones such in the borders of concrete structures (e.g. underpasses).

2.2 Indirect tensile tests

Usually the characterization of the tensile strength of aggregate concrete mixtures is made using indirect tensile tests. In this case the standards EN 13286-42 (CEN, 2003) were used. Test specimens were compacted with 150 mm diameter and 145 mm high, with low compaction (LC) and medium compaction (MC), using the Modified Proctor test. The degree of compaction (DC) of the MC specimens ranged between 91% and 93% and the one of the LC specimens varied between 81% and 85% of maximum dry unit weight from Modified Proctor test. According to CEN (2003), the tensile strength, qt, is computed by:

where Q is the maximum applied force during diametrical compression and Ф and H are the specimens diameter and height. In the performed tests the tensile strength varied significantly with the degree of compaction and the cement content, the values ranging between 35 kPa for samples with 2% cement content with low compaction and 440 kPa for samples with 5% cement content and medium compaction. Figure 2 shows the values of the tensile strength, obtained in these tests, depending on the adjusted porosity/cement ratio (n/Civ

x). The relation shows a relatively high determination coefficient (R2=0.92) with an exponent of 0.27.

R² = 0.92

0

100

200

300

400

500

15 20 25 30 35

q t[k

Pa]

n / Civ0.27

LCMC

qt = 2E+08 (n / Civ0.27)-4.396

Figure 2. Relationship between indirect tensile strength and n/Civ

x.

2.3 Seismic wave tests

Seismic wave tests are an easy and economic technique to measure dynamic properties (Amaral et al. 2012). Aiming at

determining materials elastic properties, seismic wave tests were performed on several specimens: five specimens with degree of compaction ranging from 95% to 98% (MC-medium compaction); seven specimens with degree of compaction from 83% to 86% (LC-low compaction). The wave velocity propagation was determined with ultrasonic piezoelectric transducers, namely compression transducers and shear transducers (Figure 3).

Figure 3. Measurement of compression (left) and shear (right) waves.

The dynamic parameters of the mixtures were computed taking into account the following relations:

where: E0 - dynamic deformability modulus VS - shear wave velocityVL - longitudinal wave velocity ρ - density G0 - dynamic shear modulus υ0 – Poisson ratio

The dynamic shear modulus values range from about 2 GPa to 7 GPa. There was a significant increase in the dynamic modulus with increasing cement content and compaction effort. In general, the values of Poisson's ratio () decreased with increasing cement content, assuming values of 0.25, 0.23, 0.21 and 0.20, for cement content of 2%, 3%, 4% and 5%, respectively. The Figure 4 shows the values of the dynamic shear modulus as a function of n/Civ

x. The relation has a high determination coefficient (R2=0.96) having an empirical exponent with a value of 1.0, which shows the possibility of estimating G0 of the material based on that parameter.

R² = 0.96

0

2

4

6

8

10

0 5 10 15 20 25 30

G0

[GPa

]

n / Civ1.0

LC

MC

G0 = 109.11 (n / Civ1.0)-1.205

Figure 4. Relationship between dynamic shear modulus and n/Civ

x.

2.4 Triaxial compression tests

Monotonic triaxial tests were performed on specimens with 150 mm diameter and 280 mm height. These specimens were

1375


prepared with a cement content of 2%, 3%, 4% and 5% and a degree of compaction from 83% to 86% (LC-low compaction), and with a cement content of 2% and 3% and a degree of compaction from 95% to 98% (MC-medium compaction). All of these specimens have been previously tested with seismic waves (see 1.3). For each cement content value three constant confining pressures (30, 50 and 100 kPa) were applied, leading to 18 tests. The triaxial tests were performed according to CEN ISO/TS 17892-9 (2004) standard, with saturation, consolidation and triaxial compression.

To measure the axial strain three linear variable differential transformers (LVDT) were fixed in the specimen, while for radial deformation, a system was developed for measuring the variation of the perimeter using one LVDT which is mounted between the ends of a wire that surrounds the specimen. The wire is kept under tension by two helical springs (Figure 5).

a) b)

Figure 5. Triaxial test equipment: a) cell and load frame apparatus; b) axial and radial deformation transducers installed on the specimen.

During shear compression, at 0.0016 mm/second, the specimens were submitted to two unload/reload cycles in order to define the quasi-elastic behaviour. Some of the stress-strain curves obtained in the tests are presented in Section 3 of this paper, when discussing the numerical modelling of the tests.

Figure 6 shows the relationship between the adjusted porosity/cement ratio and deformability modulus computed at 50% of ultimate shear strength for specimens with confining pressure of 100 kPa (E50

ref). The mixtures are referenced by the percentage of cement and the type of compaction (for example, 2_LC means a mixture with 2% of cement content and low compaction). In this analyses, the best correlation is also found to an exponent of 1.0, but it is associated with a lower coefficient of determination (R2=0.76) than those presented above.

R² = 0.760

1

2

3

4

5

0 5 10 15 20 25 30

E 50re

f[G

Pa]

n / Civ1.0

2_LC3_LC4_LC5_LC2_MC3_MC

E50ref = 97.71 (n / Civ

1.0)-1.308

Figure 6. Relationship between E50

ref and n/Civx.

It is important to point out that, with exception of 4_LC specimen, there was a significant increase in the deformability modulus with increasing cement content. Furthermore, it is also interesting to note the significant increase in deformability

modulus with degree of compaction, when comparing mixtures having the same cement content.

Strength parameters, such as the angle of shearing resistance (’) and the cohesion intercept (c’), were computed using the results of three specimens of each type of mixture, with similar compaction and the same cement content, for different isotropic consolidation pressures.

With regard to the angle of shearing resistance, there is a slight increase from 40° to 42° when the cement content increases from 2% to 5% in the samples with low compaction. In the samples with medium compaction it was computed a angle of shearing resistance of 58°, regardless of the cement content, which shows the great importance of the compaction on the mechanical characteristics of the mixtures.

The values of c' ranged from 250 kPa to 830 kPa, reflecting a significant increase in this parameter with the increase of the cement content. For specimens with low compaction, the increase from 2% to 5% in the cement content causes an increase of c' from 255 kPa to 835 kPa. Figure 7 shows the relationship between the cohesion intercept and the porosity/cement ratio. The best correlation is also achieved for an exponent of 1.0, with a coefficient of determination R2=0.88.

R² = 0.880

200

400

600

800

1000

0 5 10 15 20 25 30

c' [k

Pa]

n/ Civ1.0

2_LC3_LC4_LC5_LC2_MC3_MC

c' = 13607 (n / Civ1.0)-1.236

Figure 7. Relationship between cohesion intercept and n/Civ

x.

Considering the presented results, it could be concluded that it is possible to assume the value 1.0 for the exponent x, when one intend to relate the adjusted porosity/cement ratio with mechanical properties of these aggregates, with exception of the tensile strength.

3 MODELLING OF TRIAXIAL TESTS

Based on the triaxial test results, a calibration of the geo-mechanical parameters for the Hardening Soil Model available on the commercial software Plaxis was made. The model parameters that were considered for each aggregate-cement mixture are shown in Table 1.

This paper presents only the tests results and the modelling curves (mod) for the specimens with 2% of cement content, with low compaction (Figure 8) and with medium compaction (Figure 9). As previously mentioned, three different values of confining pressure were applied (30, 50 and 100 kPa). Further details can be seen in Viana da Fonseca et al. (2012).

The analysis of Figures 8 and 9 shows that: a) the curves that relate the deviatoric stress with the axial deformation are fairly well approximated by the modelling curves, in particular for mixtures with low compaction; b) it is rather difficult to model the curves that relate the volumetric deformation with the axial deformation, particularly for the higher values of the confining pressure. For the tests performed on other aggregate-cement mixtures similar trends were found.

1376


T able 1. Hardening Soil Model parameters

Materials Parameters 2_LC 3_LC 4_LC 5_LC 2_MC 3_MC

c' [kPa] 256 414 469 835 352 640 ' [º] 39.7 41.0 41.6 41.8 58.0 58.1

Failure parameters as in Mohr-Coulomb model

[º] 35.8 41.0 41.6 41.8 42.1 46.9

E50

ref

[GPa] 1.33 2.83 1.99 4.65 2.49 4.53

Eur

ref

[MPa] 4.0 8.48 5.96 13.95 7.46 13.60

Eoed

ref

[kPa] 1.33 2.83 1.99 4.65 2.49 4.53

Basic parameters for soil stiffness

m 0.60 0.40 0.16 0 0.50 0.30 ur [-] 0.2 0.2 0.2 0.2 0.2 0.2 pref [kPa] 100 100 100 100 100 100

K0

nc

[-] 1.0 1.0 1.0 1.0 1.0 1.0 Advanced parameters

Rf [-] 1.0 1.0 1.0 1.0 1.0 1.0 c' - Cohesion intercept ' - Angle of shearing resistance - Angle of dilatancy E50

ref – Secant stiffness in standard drained triaxial test

Eur

ref – Unloading / reloading stiffness (default Eur

ref = 3 E50

ref)

Eoed

ref – Tangent stiffness for primary oedometer loading

(default Eoed

ref = E50

ref )

m – Power for stress-level dependency of stiffness ur – Poisson's ratio for unloading-reloading (default ur = 0.2) pref – Reference stress for stiffnesses (default pref = 100 kPa) K0

nc – K0-value for normal consolidation (default K0

nc = 1)

Rf – Failure ratio qf / qa (default Rf = 1.0)

0

250

500

750

1000

1250

1500

0.0 0.2 0.4 0.6 0.8 1.0 1.2

q [k

Pa]

a [%]

2_LC_30

2_LC_50

2_LC_100

mod 2_LC_30

mod 2_LC_50

mod 2_LC_100

-3.0

-2.0

-1.0

0.0

v[%

]

Figure 8. Test results and modelling of low compaction mixtures with a cement content of 2%.

4 FINAL REMARKS

The results obtained in a laboratory experimental program over an aggregate mixed with high strength Portland cement were presented. The tensile strength, elastic stiffness parameters, and Mohr-Coulomb shear strength values were analysed by the porosity/cement ratio adjusted by an exponent x (n/Civ

x). Most parameters revealed that the best correlation was obtained with an exponent of 1.0, although a significant growth in stiffness and strength was obtained with increasing cement content and

degree of compaction. The Hardening Soil Model parameters calibrated from the triaxial tests results allowed a good adjustment of the stress-strain curve. The volumetric behaviour as well as the post-peak strain softening cannot be reproduced satisfactory due to model limitations.

0

500

1000

1500

2000

2500

3000

3500

4000

0.0 0.2 0.4 0.6 0.8 1.0 1.2

q [k

Pa]

a [%]

2_MC_30

2_MC_50

2_MC_100

mod 2_MC_30

mod 2_MC_50

mod 2_MC_100

-3.0

-2.0

-1.0

0.0

v[%

]

Figure 9. Test results and modelling of medium compaction mixtures with a cement content of 2%.

5 ACKNOWLEDGEMENTS

This research was developed under the activities of FCT (Portuguese Foundation for Science and Technology) research unit CEC, in FEUP through the projects PTDC/ECM/ 099475/2008, and [SIPAV: Innovative Precast Structural Solutions for High-Speed Railway (SI IDT – 3440/2008)], financed by the European Community (QREN/UE/FEDER), Operational Program for Competitive Factors "COMPETE".

6 REFERENCES

Amaral, M.F., Viana da Fonseca, A., Rios, S. 2012. Laboratory seismic measurements for dynamic characterization of cemented aggregates. Geotechnical Testing Journal. ASTM, Accepted for special number on Dynamic Testing of Soil and Rock

European Committee for Standardization (CEN) 2003. EN 13286-42:2003 - Unbound and hydraulically bound mixtures - Part 42: Test method for the determination of the indirect tensile strength of hydraulically bound mixtures.

European Committee for Standardization (CEN) 2004. CEN ISO/TS 17892-9:2004 - Geotechnical investigation and testing - Laboratory testing of soil - Part 9: Consolidated triaxial compression tests on water saturated soil.

Consoli, N.C., Foppa, D., Festugato, L., Heineck, K. 2007. Key Parameters for Strength Control of Artificially Cemented Soils. Journal of Geotechnical and Geoenvironmental Engineering 133(2), 197-205.

Consoli, N.C., Viana da Fonseca, A., Cruz, R. C.,Heineck, K. 2009. Fundamental Parameters for the Stiffness and Strength Control of Artificially Cemented Sand. Journal of Geotechnical and Geoenvironmental Engineering 135(9), 1347–1353.

Rios, S. Viana da Fonseca, A., Baudet, B. A. 2012. The effect of the prorosity/cement ratio on the compression of cemented soil, Journal of Geotechnical and Geoenvironmental Engineering, American Society of Civil Engineers (ASCE), In Print (GTENG-2072R4).http://dx.doi.org/10.1061/(ASCE)GT.1943-5606.0000698.

Viana da Fonseca, A., Amaral, M., Panico, F. 2012. Dynamic and static geotechnical characterization of a cemented limestone aggregate. Submitted to Soils & Foundations.

1377

Special Aspects for Building a Motorway on a 185 m Deep Dump

Aspects particuliers pour construire une autoroute sur un remblai de comblement de 185 m

Vogt N., Heyer D., Birle E., Vogt S. Zentrum Geotechnik, Technische Universität München, Munich

Dahmen D., Karcher C., Vinzelberg G., Eidam F. RWE Power AG, Cologne

ABSTRACT: The major part of the new motorway A 44 will be constructed on a deep dump of the Garzweiler open cast mine, reaching a maximum depth of 185 m. The soils are placed by large scale spreaders in a bulk flow. Thus, the uncompacted dumppossesses a pronounced compressibility and can exhibit long-term time-dependent deformations, mainly as a result of particle redistribution in the dump. The magnitude and rate of the deformations depend on numerous parameters, of which the soil type is the most important. With this background, the RWE Power AG has developed a concept for dumping the pit with selected soil-types on which the autobahn will be constructed. This concept only allows the dumping of mainly coarse- and mixed-grained soils beneath thefuture motorway. Extensive laboratory and field tests have confirmed that the settlements along the A 44 route, compared to similar dumps in the area of the Rhenish lignite mines, can be clearly reduced. Based on the laboratory and field tests a model that allows the calculation of the time- and stress-dependent deformations within the dump, as well as on the surface along the planned autobahn, has been developed.

RÉSUMÉ: La nouvelle autoroute A 44 sera construite sur la zone remblayé de la mine à ciel ouvert Garzweiler, zone qui atteint uneprofondeur maximale de 185 m. La terre du remblai est déversée par des gros tombereaux rigides en grande masse. Les sols présentent dès lors une densité relativement faible ce qui se traduit par une compressibilité prononcée lors du processus deremplissage et peut conduire à des déformations à long terme, principalement en raison de la redistribution des particules dans lecorps de remplissage. L'ampleur et le rythme des déformations dépendent de nombreux paramètres. Le type de sol est le plus important. Dans ce contexte, la société RWE Power AG a développé un concept de remplissage de la fosse sur laquelle l'autoroutesera construite qui utilise une sélection des granules et sables lors du déversement. Des essais à grande échelle, en laboratoire et enplace, ont confirmé que les déformations à la surface du remblai peuvent être nettement réduites par le choix ciblé des matériaux. Unmodèle de mécanique des sols permettant de représenter les déformations à la surface du sol a été développé à partir des expériencesen laboratoire et sur le terrain.

KEYWORDS: dump, mining, settlements, time-dependent deformation, creep, sand

MOTS-CLES: mine, comblement, tassement, déformation, fluage, sable

1 INTRODUCTION

The Rhenish lignite mining area covers the flat plains of the Cologne Lowland between the cities of Aachen, Monchengladbach and Cologne. With a total annual output of some 100 million tons of lignite the 3 currently active large-scale open cast mines contribute to 12 % of the electricity production in Germany. The area under consideration is part of the Garzweiler open cast mine, presently covering an area of 48 km2. In order to exploit the annual 35-40 million tons of lignite, a fivefold magnitude of overburden (150 million cubic meters) must be excavated and dumped. The current mining permit stipulates the year 2045 by which the exploitation will have to end east of the city of Erkelenz.

Because of the westwards progressing exploitation, the motorway A 44 has been closed since November 2005 over a stretch of about 10 km. Since then traffic has been redirected along the almost parallel running autobahn A 61. The area of the former autobahn is currently used for mining purposes. According to the planning schedule, the open cast mine will reach the autobahn A 61 right-of-way by the year 2017. Before the A 61 can be closed and dismantled, the A 44 that has been closed must be reopened to traffic. The major part of the new autobahn comprising 6 lanes within a cross section of 42.6 m in width will be constructed on dumped soils of the Garzweiler open cast mine, which have a maximum depth of 185 m (Köther and Reeh 2011). Figure 1 on the right gives an overview of the

opencast mine Garzweiler situated within the Rhenish lignite mining area as well as the autobahn A 44 and A 61.

The first soils under the planned route of the A 44 were dumped close to the proposed autobahn interchange Jackerath in the south in 1999. From this time on the dump has continuously progressed to restore the original topography. The dumping taking place in area close to the A 44 should be completed by the year 2016. An aerial photograph in figure 1 shows the Garzweiler open cast mine in the background as well as the extent of the area that has been reused for agricultural purposes in the foreground. The A 44 route can easily be seen between the farmland.

2025

2017

2006

AK Holz

A 44

A 44

A 46

A 61

A 61

A 44

dem

olis

hed

A 44 planned

A 46

AKJackerath

Figure 1. Mining area of Garzweiler (left) and an aerial photograph of the planned A 44 route (right)

1378


2 DUMPING CONCEPT

In order to reduce the deleterious effects on the future use of the mining area that may arise from dumped soils, the RWE Power AG developed a concept introducing restrictions regarding the content of fine particles (d < 0,06 mm) of the soils dumped below the future motorway. Figure 2 illustrates the cross section of the dump with requirements on the soils.

sand and gravelcontent of fines < 5 M.-%

surface

60°

possible: mixed grained soils witha content of fines < 20 M.-%

50 m (A 44 alignment)

not allowed: fine grained soils orsoils saturated with water

-10 m

-90 m

max. -185 m

no restrictions regarding the dumped materials

150 m

possible: mixed grained soils witha content of fines < 30 M.-%

not recommended: fine grained soil or soils saturated with water

Figure 2. Dumping concept

At depths below - 90 m under the planned surface on which the motorway will be constructed it is not recommended to dump fine-grained or saturated soils. In this bottommost part of the dump only soils with fines contents < 30 mass-% should be deposited. The width of this lower region is 3 times the width of the planned motorway. Between -90 m and - 10 m more stringent restrictions are set. Only soils with a fines content < 20 mass-% may be used to construct this part of the dump. The top part of the dump from - 10 m to the ground surface is made up of sand and gravel with a fines content < 5 mass-%.

To verify the described concept, soil samples were taken at different depths and positions from the dump along the A 44 route since the year 2008. A total of 716 samples were classified by the year 2011. Considering the total soil masses dumped into the areas described in figure 2 and the capacity of about 240000 m3 per day and per large scale spreader, sampling may be considered as a random test method. Despite this, the evaluated data provides essential information about the adherence to the defined dumping concept.

The evaluation is based on a macroscopic classification scheme of every single soil sample. The classification divides the grain-size fractions gravel, sand and fines. This gives a quick classification without conducting laboratory-based tests such as sieving. The macroscopic classification was verified by calibration phases at different points between 2008 and 2011. In this calibration the outcome of the macroscopic classification was compared and finally adjusted to the results of the grain size distribution gained from laboratory sieving. During the calibration phases a total of 106 grain size distributions shown in figure 3 were investigated.

0

10

20

30

40

50

60

70

80

90

100

0,001

Diameter of the particle [mm]

Mas

s pe

rcen

tage

[%]

0.002

Silt Sand GravelClay Stones

fine medium coarse fine medium fine medium

0.001 0.20.060.02 2 630.006 0.63 6.3 20

coarse coarse

Figure 3. Grain size distribution (106 soil samples)

The results show that for 91 % of the soil samples the sand fraction is dominant. 92 % of the soil samples have a content of

fines < 15 mass-%. Only 2.5 % of the samples are classified as fine-grained soils with a fines content > 40 mass-%. In the top layer of the dump sand and gravel is dominant.

3 TIME-DEPENDENT DEFORMATION OF THE DUMP

3.1 Geodetic measurements along the autobahn alignment

It is well known that large dumps within the Rhenish lignite mining area exhibit significant time-dependent deformations (Nehring 1968, Kothen and Knufinke 1990). This is mainly due to the comparably low densities resulting from the dumping method without compacting of the soil masses. As noticed in Section 2 the soil type and the water content mainly influence the magnitude of the time dependent deformations. Typically the rate of deformation decreases strongly according to the elapsed time since the end of the dumping process. Nevertheless, due to the large dump depths of by far more than 100 m settlement rates of the surface in a magnitude of several centimetres per year may be observed even a decade after completing the dump (Lange 1986).

For providing detailed information about the time-dependent deformation of the Garzweiler dump along the planned A 44 route a series of measurements were undertaken and analyzed. The survey department operated by RWE POWER AG provides a main database for surface deformations. The changing geometry of the dump resulting from the advancing dumping progress is monitored continuously throughout the entire dump. Additionally along the A 44 route discrete survey points are installed every 50 m to get as precise data as possible. In two sections (at station 4900 m and station 5900 m) the intervals of the survey points were reduced to 10 m apart in order to examine differential deformations over short distances.

Figure 4 plots calculated strains of the dump at the survey points against time. The diagram shows data collected from survey points near station 4900. Zero time is set at the end of the dumping process. The first measurement was taken on day 1 and then on day 7 day after the completion of the dump.

0.0

0.1

0.3

0.4

0.5

1 10 100 1000 10000Time [d]

Stra

in a

ccor

ding

to th

e he

ight

of t

he la

ndfil

l [%

]

Def

orm

atio

n [m

]

0.00

0.15

0.45

0.60

0.75

Area of observation at station +4900 mHeight of landfill: app. 150 mTime of first measurement:End of dumping

0.2 0.30

Figure 4. Time-dependent strain and deformation of the dump body

In the logarithmic time scale used in figure 4 the relationship between strain and time show a nonlinear behaviour during the first 100 day period. This shape as given in figure 4 is strongly dependent on the method by which the data is evaluated. The longer the interval between the end of dumping process and the first measurement, the stronger the curvature is of the data line. To ensure a uniform evaluation, the onset of the time dependent deformation is set by the end of dumping process for each survey point. In reality time dependent deformations already occur during dumping process. So start of the time-dependent deformation is a function of the speed of the dumping progress itself. Furthermore, the deformation rates decrease rapidly shortly after completion of the dump. This means that even once the dumping process is completed within a period of 1 day a significant curvature appears at the beginning of the time dependent deformation curve.

For the serviceability of the autobahn, deformations that occur once the superstructure of the roadways is built are of

1379


interest. As the construction of the motorway takes some month it is not necessary to describe the curvature of the time-dependent deformation, which is observed during the first period.

According to figure 4, the deformation characteristics after approximately 100 days may be described by a creep law given by Buisman 1939. This law was originally developed to describe the creep following consolidation of clays after a stepwise increase in effective stresses. In addition it is suitable to describe the deformation characteristic of the dump mainly made up of unsaturated coarse-grained sandy soils. The creep law demands two model parameters, a reference time (defined as the time when the deformation starts) and the magnitude of creep deformation. The deformation rate characterized by the slope in a semi-logarithmic time vs. deformation plot as given in figure 4 is described by the parameter CK in analogy to the parameter CB given in the original work of Buisman 1936 (“K” represents the German word “Kippe” meaning “dump”).

In figure 5, the parameter CK derived from the time dependent deformation curve of the survey points along the alignment of the planned A 44 autobahn is plotted. CK was evaluated from the slope of the deformation characteristics for a period of 1 year after the completion of dumping until mid 2011.

4000 4500 5000 5500 6000 6500 7000

Station [m]

0.00

0.05

0.10

0.15

0.20

analyzed from 365 days on after completion of the dumping process

Para

met

er C

K[%

]

south northdumping process

Figure 5. Parameter CK along the A 44 autobahn route

The plot in figure 5 gives a very uniform distribution of the parameter CK in between the station 4500 m and 6800 m of the A 44 route. CK varies in between 0.075 % and 0.125 % with a mean value of roughly 0.1 %. Only some points around 6700 m give smaller values of CK. Analyses of the data show that the depth of the dump along the observed survey points has no effect on the parameter CK. Nevertheless the depth of the dump within the area that can be observed until now changes from in the south 135 m to 155 m in the north.

In the future the dumping process will advance another 3 km to the north creating depths of dump up to 185 m. The shown geodetic measurements provide data for determining time dependent deformations. Therefore, special care is required in the evaluation of geodetic data on a continuous basis to verify that the values of CK will change in the northern region because of an increasing depth of dump. In the event the dumping concept described in chapter 2 is carried out until the dump beneath the A 44 alignment is completed, the measured CK of 0.1 % will provide valuable information for the prediction of deformations.

3.2 Effect of initial density on the time-dependent behaviour

The stress and time-dependent deformation behaviour of the soils was investigated using one-dimensional compression tests. Four different soils representing the majority of the curves plotted in figure 3 were chosen. In different test series the effect of varying initial density and loading rate on the time-dependent compression were examined in detail. Additionally, soaking at different stresses was evaluated.

All tests were carried out by increasing the stress stepwise while observing the axial deformation of the sample. Within the first few seconds after the stress was applied, comparatively

large strains were measured. The following strains reduce rapidly with elapsed time. This characteristic can be described by drawing a straight line in a diagram plotting strain versus natural logarithmic time (see figure 6). The slope of the straight line can be expressed by the Buisman constant CB (Buisman 1936).

time t

= CB·ln(t/t0)

strain [%]

logarithmic(base 2,718)

t0

0

t0

ti

t0ti ti

time t

strain [%]

stepwise increase of stress

Figure 6. Time-dependent deformation after a stepwise stress increase

Figure 7 illustrates the values of CB determined for a silty fine sand with a content of fines of 15 mass-%. The samples had heights of 2 cm and 10 cm with respective diameters of 10 cm and 30 cm. Different initial densities with density indices ID = 0 to 0.8 were examined. The initial water content was about 10 % for all tests. The tests show a clear dependence of the Buisman constant on the initial compaction index and governing stress. CB increases clearly with increasing stress. On comparing the results for different relative densities it can be seen that the CBvalue decreases with increasing density. On analysing all tests carried out on 4 soils samples, no significant influence of the soil type was recognized. Only a slightly higher CB value was determined for the silty sand (see figure 7) having a fines content of 15 mass-% opposed to the other investigated soils for which the fines content varies between 3 and 6 mass-%.

0 250 500 750 1000 1250 1500 1750 2000 2250 2500

Para

met

er C

B[%

]

Stress z [kN/m ]2

0.00

0.02

0.04

0.06

0.08

0.10Dimensions of the soil sample

height = 2 cmdiameter = 10 cm

height = 10 cmdiameter = 30 cm

Initial relative density ID

ID ≈ 0,2

ID ≈ 0,4

ID ≈ 0,6

ID ≈ 0,8

ID ≈ 0,0

Figure 7. Influence of the initial relative density ID on the Parameter CBfor a fine sand from the Garzweiler dump (silt and clay = 15 mass-%)

4 PREDICTION OF TIME-DEPENDENT DEFORMATIONS

To predict the future time-dependent deformation of the dump especially regarding the areas along the A 44 route that are not yet filled up, the validation of a model based on a soil mechanic theory was necessary. As a reference date for the model used, the completion of dumping was set to the 1.1.2017. The information seen in figure 8 was calculated using two basic equations describing the stress and time-dependent deformation. This simple model only allows the calculation of a one-dimensional deformation. On expanding the model for predicting more complex dumping processes (e.g. simulating unloading and reloading) within a three-dimensional geometry,

1380


a finite element code was used by applying a visco-plastic soil model on the dump body.

The time-dependent deformation is mainly governed by 2 parameters namely the reference time t0 and a creep parameter defined as CB or CK respectively. The magnitude of both parameters was determined using the geodetic measurements at the stations 4900 m, 5900 m and 6600 m where the dumping process was completed several years ago. A favourable outcome between the measurements and the results of the simple model can be found using CK = 0.1 % and setting t0 at the end of the dumping process. The stress dependent deformation (stiffness) was evaluated using the data of cone penetration tests reaching depths of 72 m below the surface of the dump.

0.0

0.2

0.4

0,6

0.8

1.0

1.2

4000 5000 6000 7000 8000 9000 10000 11000Station [m]

Def

orm

atio

n [m

]

CK = 0.05 % for 0 m to 45 m and CK = 0.10 % for 45 m down to the base of the landfill

CK = 0.10 % independent from the depth, deformations up to 2030

Start of observation1.1.2017CK = 0.05 % for 0 m to 90 m and CK = 0.10 %

for 90 m down to the base of the landfill

Start of observation1.7.2017

Start of observation31.12.2017

South North

Figure 8. Prediction of the deformation along the planned A 44 for different depths treated by soil compaction methods and the effect of an extended time period after dumping before observing deformations

The continuous lines in the diagram show the predicted settlements of the dump surface considering CK = 0.1 %. Calculations were undertaken until the year 2030 using 11 cross sections between station 4900 m and 10500 m. The cross sections were idealized according to a one-dimensional column of soil layers dumped during different periods in time.

For the given case that the soils within the dump do not differ regarding to parameter CK, the calculated deformation is mainly dependent on the depth of the dump and time when dumping is completed or is intended to be completed. According to the used logarithmic creep model the deformations plotted in figure 8, an increase along the route occurs mainly between 8000 m and 10500 m. In the case where the dumping process has already been completed since 2 or more years from the start of observation back to the past, the chosen time when the observation starts has a minor impact on the deformations since settlement rates are already small.

From stations 8000 m to 10500 m the period between the end of the dumping and the start of observation has a strong influence on the calculated deformations that will occur by the year 2030. Even in a period of 6 months between the end of the dumping and the start of observation reduces the calculated settlements significantly. This is of special interest because the planned construction progress of the A 44 is put on hold until the superstructure sensitive to deformations is constructed.

In the simulations the effect of different measures of soil improvement (mainly compaction methods) were examined. The dashed and dash-doted lines in figure 8 show the deformation until the year 2030 for compaction methods reaching a depth of 45 m and 90 m respectively. It is assumed that the compaction of the different soil layers will lead to a lower creep parameter of CK = 0.05 % (see chapter 3.2). The calculation show a 45 m deep treatment of the dump still leads to a comparably high deformation reaching a maximum of 0.95 m. Even a 90 m deep soil improvement will lead to a calculated deformation of 0.85 m.

The results of the simulations show that providing enough time between completing the dumping process and the construction of parts sensitive to settlement is far more efficient than soil improvement or treating deep soil layers in the dump. On waiting at least a 6 month period before the construction of

the A 44 autobahn, a major quantity of settlements will already have been developed. By the used models settlements of 0.4 m and 0.3 m are to be expected in between mid 2017 and end 2017 respectively until the year 2030. This magnitude will be covered by the so-called precautionary gradient providing sufficient drainage of water by the planned crossfall of the roadway and the embankment of the motorway (Köther and Reeh 2011).

5 SUMMARY

The paper presents the most important project-specific conditions including the dumping process and the properties of the dumped soils along the future A 44 routing. Furthermore, geodetic measurements have shown to give a good overview on the time-dependent deformation of the Garzweiler dump. By use of one-dimensional compression tests the deformation characteristics of the dump body, consisting mainly of poorly-graded sand were examined. During the test series main influencing parameters such as initial density and loading rate were examined.

Using a simple model for the description of the time-dependent deformation of the dump and its soils the effectiveness of soil compaction methods is discussed and evaluated. The parameters governing the stress- and time-dependent deformations were calibrated by means of field data from the geodetic measurements and cone penetration testing. The simulation results cover the period from the end of the dumping process to the year 2030. Different periods from the end of dumping until the observation of the beginning of surface deformation were considered. It is evident that the period between the end of dumping and the beginning of construction of parts of the motorway sensitive to settlement, has a large impact and can therefore impair the serviceability. By far less impact was predicted in the case that even deep soil layers of the dump would be compacted.

In general it can be concluded that the concept of dumping predominantly coarse-grained soils within the planned route of the A 44 motorway, is suitable for limiting time-dependent settlements. The simulation results and geodetic measurements have shown that by allowing the proposed period of at least 6 months between the end of the dumping process and the start of the construction work the settlements of structures or pavements sensitive to deformations are reduced significantly.

6 REFERENCES

Köther M. and Reeh F. 2011. New autobahn through an active opencast mine. World of Mining – Surface & Underground 63 (6), 334-343.

Nehring H. 1968. Markscheiderische Beobachtung von Kippensetzung-en im rheinischen Braunkohlenrevier. Braunkohle, Wärme und Energie 20 (3), 83-91.

Kothen H. and Knufinke H. 1990. Restsetzungen auf Neulandflächen. Braunkohle 1990 (10), 1990, 24-29.

Lange S. 1986. Building on uncompacted dumps in the Rhenish brown eoal area of the Federal Republic of Germany. Building on marginal and derelict land, Thomas Telford Ltd, London 137-153.

Buisman K. 1936. Results of Long Duration Settlement Tests. Proc. 1st Int. Conf. of Soil Mechanics and Foundation Engineering,Cambridge, 103-107.

Vogt, N., Heyer, D., Birle, E., Vogt, S., Dahmen, D., Karcher, C., Vinzelberg, G. 2012. Neubau der A44 auf einer frischen Tagebaukippe, Beiträge zum 11. Geotechnik-Tag in München Geotechnik und Energie, Zentrum Geotechnik TU München, Heft 52, 5-17.

1381

Performance verification of a geogrid mechanically stabilised layer

Vérification de la performance d’une couche stabilisée mécaniquement par une géogrille

Wayne M., Fraser I., Reall B., Kwon J. Tensar International Corporation

ABSTRACT: As part of the study to evaluate performance of a geogrid stabilised unpaved aggregate base overlying relatively weakand non-uniform subgrade soils, a controlled field study was conducted in Weirton, West Virginia, USA. A punched and drawnpolypropylene triaxial geogrid was installed at the interface between a soft clayey subgrade and crushed limestone. In-ground pressurecells were used to monitor horizontal stress within the subgrade and base throughout subsequent compaction and traffic loading. The results demonstrated that the lateral stress in the subgrade were approximately ¼ that of the control section and in addition, the geogrid confines unbound aggregate leading to an increased lateral stress and a higher resilient modulus for the stabilised base layer. To verify that these results are applicable to different subgrade and aggregate materials, additional full-scale field tests wereconducted in Salt Lake City, Utah, USA. A total of four pressure cells were installed in test section, at spacings of 2.6 and 2.9 metresfrom the centerline and offset 3 metre parallel to the direction of travel. Pit run material was used as a base layer in this study. Pit run gravel is unprocessed material that contains all sizes of rock. The results show that the horizontal pressures within the subgradecreated by both the static and live loading conditions were significantly reduced by using the geogrid whereas a >80% increase inhorizontal pressures were measured for the control section after placement of aggregate fill. Post-traffic trenching of the control section found significant mixing of the subgrade materials, whereas very little intermixing of the subgrade materials was reported forthe stabilised section. This paper presents the results and analysis from these field studies. The results confirm that the geogridpromotes improved aggregate confinement and interaction, leading to enhanced structural performance of the unpaved aggregate base.

RÉSUMÉ : Dans le cadre d’une étude pour évaluer la performance d'une géogrille pour stabiliser une couche granulaire non revêtuesur des sols de fondation de faible portance et hétérogènes, une étude expérimentale in-situ a été réalisée à Weirton, Virginie-Occidentale, Etats-Unis, Une géogrille triaxiale, fabriquée à partir d’une feuille de polypropylène perforée et étirée a été installée à l’interface entre un sol de fondation argileux mou et une couche de concassé calcaire. Des capteurs de pression ont été installés dans le sol pour mesurer la contrainte horizontale dans le sol de fondation et la couche de gravier pendant le compactage et la circulation. Les résultats ont montré que la raideur latérale dans le sol de fondation était approximativement le quart de celle de la section de contrôle et par ailleurs, la géogrille confinait les agrégats granulaires conduisant à une augmentation de la raideur latérale et de larigidité de la couche de base stabilisée. Pour vérifier que ces résultats sont applicables à différents types de sols de fondation et àd’autres matériaux granulaires, d'autres essais in-situ à grande échelle ont été réalisés à Salt Lake City, Utah, Etats-Unis. Un total de quatre capteurs de pression ont été installés dans les sections d'essai, espacés de 2,6 et de 2,9 mètres de l'axe et décalés de 3 mètres parallèlement à la direction de déplacement. Un matériau non traité a été utilisé comme couche de base dans cette étude. Les résultatsmontrent que les pressions horizontales dans le sol de fondation créées par les deux conditions de chargements statique et dynamique ont été significativement réduites par l'utilisation de la géogrille, alors qu'une augmentation de 80% des pressions horizontales a étémesurée pour la section de contrôle après la pose de matériaux d'agrégats, par rapport à la section stabilisée. Cet article présente les résultats et l'analyse de ces études in-situ. Les résultats confirment que la géogrille favorise le confinement et l'interaction desagrégats, conduisant à améliorer la performance structurelle de la couche granulaire non revêtue.

KEYWORDS: Field trafficking performance, Triaxial geogrid, Lateral stress, and Resilient modulus.

1 INTRODUCTION.

In cases where a gravel surfaced road is required over subgrade conditions that are unable to adequately support the traffic loads, geogrids are commonly used to stabilize the aggregate base course and improve pavement performance by decreasing the load distributed to the subgrade. The aggregate that is directly above the geogrid is laterally confined and the result of this enhanced confinement leads to an increase in the resilient modulus of aggregate adjacent to the geogrid. As a result, the stabilised aggregate spreads surface loads over a wider area of subgrade. In general, an equivalent stabilised road section thickness yields an increased allowable traffic load compared to an unstabilised road section.

Geogrids have been used successfully to improve the performance and increase the design life of unpaved roads since the 1970’s. Nonwoven geotextiles have been efficient in

applications that require the separation of aggregate layers from the underlying subgrade soil.

Throughout the history of geosynthetics, monitored full-scale field studies have been extensively utilized to study the performance of geogrid stabilised sections. Although more sophisticated and precise methods, (i.e., numerical modeling and laboratory test models) can be utilized to study specific variables and/or to optimize a geosynthetic, a basic field study remains as one of the most effective means of providing a definitive proof of performance.

Full scale research has provided guidance, basic criteria and information for the use of geogrids in roadway design (Tingle and Webster 2003). Subgrade bearing capacity factors of the unstabilised and stabilised sections were determined using empirical data from full scale testing performed by the Engineering Research and Development Center (ERDC). The

1382


calculated bearing capacity factor of the geogrid stabilised section was more than double that of the unstabilised section.

This paper presents the in-ground stress cell measurements from two full-scale field tests to validate the enhanced confinement effect associated with use of an integrally formed punched and drawn geogrid.

2 FIELD STUDY 1 - WEIRTON, WEST VIRGINIA, USA

2.1 Research background

A field study at a site located in Weirton, West Virginiawas developed to evaluate the support conditions of a mechanically stabilised crushed limestone layer on soft clayey subgrade (White et al. 2011). In-ground piezoelectric earth pressure cells (EPC) were used to measure horizontal stress below and above the geogrid location versus the passage of construction and truck traffic over the course of test pad construction and trafficking.

Goals of this field investigation were to: -Validate fitness for use of geosynthetic products in a

challenging subgrade improvement application for construction and trafficking of an unpaved road.

-Verify the enhanced confinement effect associated with the use of geogrid due to geogrid-aggregate interlock.

-Verify the degree of load spreading by recording lateral stresses within the subgrade.

2.2 Test section construction

The subgrade soils beneath the test tracks were excavated to a depth of 900mm below the surface. The excavated material was replaced with a uniform lean clay (CL) material. The clay material was placed in the test tracks in uniform 0.35 metre thick loose lifts and mixed thoroughly to a uniform consistency with a roto-tiller. Water was added and several passes of the tiller were used to arrive at a moisture content that produced a subgrade California Bearing Ratio (CBR) of approximately 2 to 3 %.

Geogrids were installed on top of a finished subgrade. Physical properties of geogrid are summarized in Table 1.

Table 1. Summary of geosynthetic treatments.

Type Physical Properties

Polypropylene triangular aperture geogrid

Radial Stiffness = 300 kN/m @0.5% strain

Vertical and horizontal stress measurements were taken in the subgrade and about 150mm above geogrid/base material. Figure 1 illustrates the layout for horizontal and vertical stress cells.

Figure 1. Cross Section of instrumentation installation.

Then, a base course aggregate (Ohio Department of Transportation 304, base course material) was placed in two compacted 300mm lift thicknesses. The crushed limestone was classified as a GP-GM with about 8 percent of fines passing the No. 200 sieve.

Cardboard is used as a temporary liner to contain the silica sand backfill around the EPC (See Figure 2). Use of the sand ensured a uniform stress was applied to the EPC surface.

Figure 2. Placing horizontal earth pressure cell at the bottom of the base layer.

2.3 Results.

A Ford L8000 dump truck was used for trafficking of the constructed test sections. The vehicle was loaded to a gross vehicle weight of 18,370 kg.

Figure 3 depicts the readings of dynamic horizontal stresses within the subgrade versus the passage of construction and truck traffic over the course of test pad construction and trafficking. Evident within Figure 3, is the minimal amount of horizontal post traffic stress remaining within the subgrade in comparison to the level found in the control section. The lateral stress below the geogrid is a little over 5 kPa versus 20 kPa for the control test section. This equates to a stress state value that is 25% of the control stress state thus indicating a high level of subgrade protection. This work demonstrated an enhanced fully confined zone above the geogrid resulting in uniform vertical stress across the subgrade resulting in less lateral stress.

Figure 4 depicts the horizontal stress state, post trafficking, exhibited above geogrid. In contrast to the control, the geogrid confines the unbound aggregate leading to an increased lateral stress within the aggregate. The results demonstrate the inclusion of geogrid at the interface of soft subgrade and aggregate layers affects the development of the “locked-in” horizontal stress following loading. A higher horizontal stress within the stabilised aggregate layer gives a direct indication of the lateral restraint mechanism. The result of increased aggregate stresses leads to an increase in the resilient modulus of aggregate adjacent to the geogrid.

25

20

15

10

5

050 100 150


Stabilised section


Stress (kPa)

Figure 3. Horizontal stress within the subgrade layer after roller compaction and test vehicle passes (White, et. al., 2011).

1383


25

20

15

10

5

050 100 150


Stabilised section


Stress (kPa)

Figure 4. Horizontal stress within the base layer after roller compaction and test vehicle passes (White, et. al., 2011).

Field determination of the relative density values for the second or upper subgrade lift after the completion of 21 truck passes shows 90.2% and 98.5% relative density values were achieved on the control section and stabilised section, respectively. These numbers demonstrated that the aggregate placed over the geogrid can be compacted to a much higher degree than the unstabilised control section.

Lateral stress ratio (K) is calculated as the ratio of total horizontal to total vertical stresses for the subgrade and subbase layers following roller and trafficking passes. Resulting values are presented in Table 2. The calculated K values demonstrate that during trafficking, the K values are about 0.3 to 0.7 for the subgrade and 0.5 to 0.7 for the subbase for all test sections. However, the K values after 75 trafficking passes show buildup of horizontal stresses with relatively high K values in the control section subgrade layer compared to the geogrid stabilised section. The stabilisation ratio provides a clear indication of degree of improvement. For this study the geogrid results in a section that is 8 times better than the control with regard to stress distribution.

Table 2. Performance comparison between test sections.

Section Ksubgrade Kbase Stabilisation Ratio

Control section 3.2 1.2 0.4

Stabilised section 1.0 3.2 3.2

3 FIELD STUDY 2 - SALT LAKE CITY, UTAH, USA

3.1 Research background

A field study at a site located in Salt Lake City, Utah, United Stated of America was conducted to evaluate the effect of a mechanically stabilised platform to bridge over challenging soft subgrade areas. All test sections are backfilled with 150-mm minus pit run (unprocessed) gravel. A total of four sections were constructed and trafficked and two of the sections were stabilised by a layer of integrally formed punched and drawn triangular aperture geogrid placed at the interface between the subgrade and bridging material.

In-ground piezoelectric earth pressure cells (EPC) were used to evaluate the support conditions of the test sections. Measurements of tire ruts were recorded during the survey between passes of the haul truck.

Goals of this field investigation were to: -Validate ability of a geogrid to reduce lateral pressures

within the subgrade under heavy loading conditions and very soft soils.

-Assess the ability of geogrid to stabilize pit run gravel and quantify the benefits for different conditions and loading scenarios.

- Provide surface and subgrade data on heavier loading scenarios.

3.2 Materials.

3.2.1 Subgrade soil and pit-run gravel Test beds consisted of two materials in this study – low plasticity clay subgrade and pit-run gravel material. A Dynamic Cone Penetrometre (DCP) test was performed in accordance with ASTM D6951-03 using a 4.6-kg single mass hammer. Results were used to determine the strength of subgrade with depth. The near surface California Bearing Ratio (CBR) for the subgrade material varied from about 0.2% to 0.4%.

Aggregate fill material consisted of pit run gravel with a maximum particle size of 150 mm. An enhanced, second generation University of Illinois Aggregate Image Analyzer (UIAIA) was used to determine morphological indices, such as angularity index, AI (Rao et al., 2002) and surface texture index, ST (Rao et al., 2003) of the pit run gravel used in the test. Angularity is critical for aggregate interlock and surface texture has been found to directly influence friction between aggregate particles as well as the strength of the aggregate. The AI and ST indices are determined based on the particle image outlines obtained from each of the top, side and front of coarse particles. Morphological index results of the aggregate samples are presented in Figure 5 and 6.

About 88% of the aggregate samples have angularity index values that are less than 325 and surface texture index values less than 1.375. These values indicate that the pit run gravel used in this study consists of rounded and very smooth surface aggregate particles.

0

50

100

150

200

250

300

350

400

450

500

550

0 5 10 15 20 25

Angularity Index

Particler No.

Retained on 4.75 mm sieve


Retained on 19 mm sieve

Figure 5. Angularity Indices of the Pit Run Gravel.

0

0.25

0.5

0.75

1

1.25

1.5

1.75

2

2.25

2.5

0 5 10 15 20 25

Surface Texture

Particler No.



Retained on 19 mm sieve

Figure 6. Surface Textures of the Pit Run Gravel.

3.2.2 Geosynthetics Two geosynthetic materials were used during this investigation. Some physical properties of each geosynthetic material are summarized in Table 3.

1384


A nonwoven geotextile served as a separation layer in conjunction with a triangular aperture geogrid below the aggregate layer. Based on our understanding of the site subgrade soils and overlying pit-run aggregate, a nonwoven geotextile was recommended for use as a separation layer.

11.36.3

59.6

33.2

0

20

40

60

80

100

120

0.9 meter from CL 1.0 meter from CL

Subgrade Stress (kPa)

Stabilised Section Unstabilised Section

Table 3. Summary of geosynthetic treatments.

Type Physical Properties

Polypropylene triangular aperture geogrid

Radial Stiffness = 300 kN/m @0.5% strain

Polypropylene nonwoven geotextile Weight = 8oz/yd2

3.3 Test section construction

Four test sections were tested during this field study. All test sections consisted of a 300mm thick layer of pit run materials placed over the subgrade. Two sections are stabilised by a layer of geogrid placed at the interface between the subgrade and pit run material. Each test section was approximately 7 metre wide by 10 metres in length. Two EPCs were placed in each test section subgrade to monitor horizontal stress in the subgrade. In general, the approximate angle of distribution of stress within a properly designed geogrid stabilised section is 45 degrees. Therefore, EPCs were placed at 0.9 metre and 1.0 metre from the edge of the wheel path. Pit runmaterial was placed on top of the geogrid by a CAT D8 dozer in a 900mm thick lift. After placement of pit run material, lines were painted on the surface of the road at 0.3 metre intervals from the centerline within the areas over the pressure plates.

Figure 7. Subgrade pressures of 900mm sections.

28.6

7.6

119.2

48.0

0

20

40

60

80

100

120

0.9 meter from CL 1.0 meter from CL

Subgrade Stress (kPa)

Stabilised Section Unstabilised Section

3.4 Results

Test sections were trafficked by a Volvo A40F articulated truck. The loaded truck produces a ground contact pressure of 176 kPa under each wheel. There was some surface movement of the fill material (due to the smooth rounded aggregate), but no significant deformation of the section was noticed within the wheel path over the course of 50 passes.

Figure 8. Subgrade pressures of 600mm sections.

The stresses as presented represent the change in ground stresses under the accumulated trafficking passes. As expected, higher stresses are recorded within the control section. Stress measurements in Figure 7 indicate that the stresses of the control section were in the range of 33 to 60 kPa depending upon the distance from wheel path. The stresses of triangular aperture geogrid stabilised section were in the range of 6 to 11 kPa.

4 CONCLUSIONS

The other two sections (control and stabilised) were lightly trafficked. As no significant surface deformation was noticed within 23 passes, sections were then cut down from 900mm in height to 600mm by a CAT 980H loader.

Traffic resumed post-cut. Significant deformation occurred after the first pass across the control section. The trafficking was stopped after 1 additional pass. There was no indication of structural distress in the stabilised section. Stress measurements taken from the first pass are shown in Figure 8. The results indicate that the stresses of the control section were in the range of 48 to 120 kPa, whereas the stresses within the stabilised section were in the range of 8 to 29 kPa.

The field tests have demonstrated benefits in terms of a dramatic reduction in subgrade stress. Rut depth measurements showed all geogrid stabilised sections performed significantly better than the unstabilised controls. In-ground stress cell measurements showed that higher horizontal stress developed within the stabilised aggregate layer during compaction and this was maintained throughout trafficking. The stabilisation ratio calculated as the ratio of horizontal stresses in the base and subgrade layers provides an indication of field trafficking performance. Results of the second study validate the performance of the triangular aperture geogrid examined in two full scale trafficking studies.

5 REFERENCES

After the trafficking test was completed, trenches were excavated to observe subgrade conditions. A significant amount of intermixing of the pit run material and subgrade interface occured within the 600mm thick control section. Very little intermixing of the subgrade materials was observed in all other sections including the 600mm thick triangular aperture geogrid stabilised section.

Tingle, J.S. and Webster, S.L. 2003. Review of Corps of Engineers Design of Geosynthetc Reinforced Unpaved Roads, Annual meeting CD-ROM, TRB, Washington, D.C.

Rao, C., Tutumluer, E., and Kim, I-T. 2002 Quantification of Coarse Aggregate Angularity based on Image Analysis. Transportation Research Record (TRB) No. 1787, 117-124,.

Rao, C., Pan, T., and Tutumluer, E. 2003. Determination of Coarse Aggregate Surface Texture Using Imaging Analysis. In Proceedings of the 16th ASCE Engineering Mechanics Conference

White, D.J., Vennapusa, P.K.R., Gieselman, H.H., Douglas, S.C., Zhang, J. and Wayne, M.H. 2011. In-Ground Dynamic Stress Measurements for Geosynthetic Reinforced Subgrade/Subbase. Geo-Frontiers, Dallas, Texas.

1385

Characterization of Soil-Geosynthetic Interaction under Small Displacements Conditions

Caractérisation de l'Interaction sol-géosynthétique sous des conditions de petits déplacements

Zornberg J.G., Roodi G.H., Ferreira J. The University of Texas at Austin, Austin, Texas, USA

Gupta R. Geosyntec Consultants, Columbia MD, USA

ABSTRACT: While ultimate failure governs the performance of some geosynthetic-reinforced systems (e.g. reinforced walls), the small displacement response governs the behavior of geosynthetic-reinforced pavement systems. Yet, quantification andcharacterization of the effectiveness of geosynthetic products under small displacement conditions has been limited. The purpose ofthis study is to develop a soil-geosynthetic interaction model that captures the stiffness of the soil-geosynthetic interaction under small displacement conditions. The proposed model assumes: (1) a linear relationship between the axial strain of the confined reinforcement and its unit tension, and (2) a uniform soil-geosynthetic interface shear over the active length of the geosynthetic. The resulting force equilibrium differential equation is solved using a force boundary condition at the free end of the geosynthetic, and a displacementboundary condition at the end of the active length of the geosynthetic. The solution results in a parameter, the stiffness of soil-geosynthetic interaction, which consolidates the tensile properties of geosynthetic with the interaction properties of the soil-geosynthetic interface. Results of laboratory pullout tests illustrate the validity of the soil-geosynthetic interaction model.

RÉSUMÉ : Alors que la rupture finale régit les performances de certains systèmes renforcés par des géosynthétiques (par exemple les murs renforcés), la réponse en petits déplacements régit le comportement de chaussées renforcées par des géosynthétiques. Pourtant, la quantification et la caractérisation de l'efficacité des produits géosynthétiques sous des conditions de petit déplacement ont été peuétudiées. Le but de cette étude est de développer un modèle d'interaction de sol-géosynthétique qui prenne en compte la rigidité de l'interaction sol-géosynthétique sous les conditions de petit déplacement. Le modèle proposé suppose: (1) une relation linéaire entre la déformation axiale du géosynthétique confiné et la contrainte de traction, et (2) un cisaillement uniforme à l’interface entre le sol et le géosynthétique sur la longueur active du géosynthétique. L’équation différentielle résultant de l’équilibre des forces est résolue àl'aide des conditions aux limites à l'extrémité libre du géosynthétique, ainsi qu’une condition aux limites de déplacement à la fin de la longueur active du géosynthétique. La solution met en évidence un paramètre, le coefficient de rigidité d'interaction sol-géosynthétique, qui combine les propriétés en traction des géosynthétiques avec les propriétés de l'interaction de l'interface sol-géosynthétique. Les résultats des essais d’arrachement en laboratoire illustrent la validité du modèle d'interaction sol-géosynthétique.

KEYWORDS: Geosynthetics, Interface Shear, Soil-Geosynthetic Interaction, Small Displacement Conditions, Reinforced Pavement.

1 INTRODUCTION

Geosynthetic reinforcements are widely used in two groups of geotechnical systems: 1) Retaining walls and slopes, and 2) Pavement systems. In retaining structures and slope stabilization projects, geosynthetic reinforcements are designed to prevent the development of failure surfaces within the soil mass. Accordingly, tensile forces develop within the geosynthetic reinforcements that contribute to the stability of geosynthetic-soil composite (e.g. Zornberg and Christopher 2007). Instead, geosynthetic reinforcements in pavement applications are used to improve the performance of the paved road under in-service conditions induced by traffic and environmental loads (e.g. Zornberg et al. 2012, Roodi and Zornberg 2012). While ultimate tensile failure is the condition of concern in the design of geosynthetic-reinforced retaining structures, the small displacement response governs the performance of geosynthetic-reinforced systems in pavement reinforcement applications.

Most of the methodologies and models developed for the analysis and design of the geosynthetic-reinforced structures have focused on the maximum strength or ultimate capacity of the geosynthetic layers (Gupta 2009). However, capturing the initial stiffness of soil-geosynthetic interface is central to accurately address the small displacement behavior of

geosynthetic reinforced pavement systems. In the absence of proper specifications to characterize the behavior of soil-geosynthetic interfaces under small displacements, designers have typically relied on the mechanical properties of geosynthetics in isolation (e.g. ultimate tensile strength or tensile stiffness/modulus) in an attempt to satisfy a certain level of performance (Archer and Wayne 2012). Studies have aimed at establishing correlations between geosynthetic index properties and their field performance. These index properties have included the rib strength, junction strength, aperture size, wide-width tensile strength, tensile modulus, tensile strength at 2% and 5%, and flexural rigidity (e.g. Perkins et al. 2004, Christopher et al. 2008, Cuelho and Perkins 2009, Mahmood et al. 2012, Chen and Abu-Farsakh 2012). However, most of these properties correspond to the behavior of the geosynthetics in-isolation rather than to the soil-geosynthetic interaction.

The purpose of this study is to introduce a soil-geosynthetic parameter capable of quantifying the performance of geosynthetic reinforcement under small displacement conditions. This parameter is defined as “Stiffness of Soil-Geosynthetic Interaction” or KSGI, which is expected to be constant for a given soil-geosynthetic system under specific confinement stress. This paper describes the assumptions and formulations used to derive the KSGI. The paper also reports on

1386


the results obtained suing a conventional pullout test setup conducted for validation of the model.

2 ASSUMPTIONS OF THE SOIL-GEOSYNTHETIC INTERACTION MODEL

The proposed model is based on two major assumptions. The first assumption concerns the Unit Tension - Strain relationship of geosynthetic products. Researchers have assumed different relationship between the unit tension in geosynthetics (T) and strain (). While Wilson-Fahmy et al. 1994 assumed a linear relationship between T and , Perkins and Cuelho, 1999 used a nonlinear relationship, and Ochiai et al. 1996 and Sieira et al. 2009 assumed it to be equal to unconfined stiffness of the geosynthetic obtained from the in-isolation wide-width tensile test. For the purpose of this study, it is assumed that the T-relationship of geosynthetic materials remains linear under soil confinement. However, the slope of this line would be not necessarily the same as (probably higher than) in the unconfined condition. As shown in Figure 1, the slope of T- line (Jc or Confined Stiffness of Geosynthetic ) is assumed constant for small displacement:

cJT (1)

T

J (kN/m)c

Figure 1. Tensile load-strain relationship for geosynthetic reinforcement under confinement

The second assumption addresses the relationship between soil-geosynthetics interface shear and the displacement of the geosynthetic, which is also known as interaction law. Various assumptions for the distribution of interface shear have been adopted in previous studies. For example, Sobhi and Wu, 1996 assumed a constant interface shear, while Abdelouhab et al., 2008 considered linear distribution of interface shear. In addition, a bi-linear distribution was used by Juran and Chen 1988 and Madhav et al. 1998, other non-linear distribution were used by Perkins and Cuelho, 1999, and an hyperbolic interface shear relationship was assumed by Gurung and Iwao, 1998. Sugimoto and Alagiyawanna (2003) showed that the direct evaluation of the interface properties from the ultimate state may not be appropriate to simulate the actual geosynthetic behavior in reinforced soil masses before failure in a pullout test. Sobhi and Wu (1996) defined the limit shear stress for pullout test, which was lower than the maximum shear stress and a function of overburden pressure applied to the soil-geosynthetic interface. They showed results from finite element analyses indicating the development of uniform shear stress independent of the frontal pullout force magnitude and length of the geosynthetic. In the study presented in this paper, a uniform distribution of interface shear is assumed over the active length of the reinforcement, as shown in Figure 2. The constant interface shear stress is defined as the yield shear stress (y), which is independent of the interface displacement at any point along the confined active length of geosynthetic.

3 FORMULATION

The model assumptions are considered in order to solve the governing differential equation of a confined geosynthetic. The

solution can be used to obtain the displacement, strain and force at any point x along the length of the geosynthetic.

u

y

u

Figure 2. Interface shear-displacement relationship

As shown in Figure 3, the force equilibrium of a differential segment of the confined geosynthetic can be written as:

0)2()()( dxdTTT (3)

Where:

icgeosynthettheandsoilbetweenshearInterfaceicgeosynthettheintensionUnitT

icgeosynthettheofsegmentaldifferentiAdx

:::

T T+dT

dx

y

Figure 3. Force equilibrium for a differential segment of geosynthetic

Rearranging this equation returns the force equilibrium differential equation governing soil-geosynthetic interaction:

2dxdT

(4)

According to the second assumption described in the previous section, the soil-geosynthetic interface shear is constant along the active length of the geosynthetic (i.e. = y).Also, using confined stiffness of geosynthetic system (Jc), the unit tension (T) can be replaced using Equation (1). Substituting accordingly into Equation (4) returns the following equation:

yc

dxJd

2)(

(5)

The axial strain in the geosynthetic can be replaced by the derivative of displacement. In addition, Jc is considered constant for a given normal pressure and under small displacements. Therefore, Equation (5) can be rewritten as follows:

yc xdudJ 22

2

(6)

where u is the interface displacement. Equivalently:

c

y

Ju

2 (7)

Integrating twice the differential Equation (7), returns equations for u and , respectively: u

1387


1

2cx

Jdxdu

c

y

(8)

212 cxcx

Ju

c

y

(9)

Taking into account thatdxdu

, the unit tension in the

geosynthetic, T , can be obtained by replacing Equation (8) into Equation (1):

cy JcxT 12 (10)

The constants c1 and c2 can be found using by two boundary conditions. Assuming geosynthetic reinforcement confined with aggregates, unit tension will be decreasing from one end to another (Figure 4). Conventional solutions have used two force boundary conditions at the two ends of the geosynthetic to solve the governing differential equation. However, under small displacement movements, these boundary conditions are not realistic because the entire geosynthetic length is not mobilized. In this study, and as presented in Figure 4, the geosynthetic length includes two portions: an “active portion” which moves under small displacement (i.e. portion AC in Figure 4), and a “non-moving part” (i.e. portion BC in Figure 4).

Ty

Active Length=L' Stationary Length

Ao

C

Figure 4. Boundary conditions differential segment of geosynthetic

In this study, two realistic boundary conditions are assumed to solve the differential equation under small displacement. A force boundary condition is assumed at Point A (TA = T0), and a displacement boundary condition is assumed at Point C (uc = 0).Using these boundary conditions leads to unit tension and displacement functions in the active length of geosynthetic reinforcement. According to this solution unit tension in the active length is related to the displacement of geosynthetic as follows:

)()4()( 2 xuJxT yc (11)

Since the confined stiffness of geosynthetic (Jc) and the yield shear stress (y) are assumed constant for specific soil-geosynthetic system for a given stress conditions, the multiplier (4Jcy) represents a key parameter in soil-geosynthetic interaction under small displacements. This parameter is defined as the “Stiffness of Soil-Geosynthetic Interaction” or KSGI.

ycSGI JK 4 (12)

Equations 11 and 12 establish a linear relationship between the interface displacement ( ) and the square of the unit tension (T(x)2) at any location within the active length (0 < x < L’). The slope of this line is KSGI. These equations also suggest a parabolic relationship between T and u under small displacement regime.

)(xu

4 EXPERIMENTAL EVALUATION

As an illustration of the extensive program conducted to validate the proposed model, the authors conducted a

conventional geosynthetic pullout test in a large pullout box with internal dimensions of 1.5 m (60 inches) length, 0.6 m (24 inches) width and 0.3 m (12 inches) height. The test involved a biaxial geosynthetic with dimensions of 300 x 600 mm. The fill material used was clean poorly graded sand, which classifies as SP in the unified system. The sand is composed of medium to fine, and sub-angular to sub-rounded particles. The mean particle size (d50) is 0.44 and the coefficient of uniformity, Cu,and the coefficient of curvature, Cc, are determined as 1.6 and 1.0, respectively. Figure 5 shows the gradation curve of this soil.

Figure 5. Gradation of the fill material used in the pullout test

Telltale wire cables were used to connect 5 linear variable differential transformers (LVDTs) to evenly spaced points along the geosynthetic length in order to accurately measure displacements of the geosynthetic during testing (Figure 6).

B

Pullout Force (T0)

Geosynthetic Specimen

LVDT 5LVDT 4

LVDT 3LVDT 2LVDT 1

L

u5

u3u2

u4

u1

Figure 6. Schematic of geosynthetic specimen and attached LVDTs

Results of the test are presented in Figures 7 and 8 up for the initial portion of the test, up to a displacement of 1 mm. In

re 7, square unit tension of geogrid (T ) is displayed versus lacement ( ) for telltale locations of LVDTs 2, 3, and 4.

This figure illustrates good consistency of the results obtained using at different locations (LVDTs 2, 3, and 4). KSGI values are obtained as 5.3, 7.9, and 8.6 (kN/m)2/mm. Figure 8 illustrates the parabolic relationship between

Figudisp u

T and .u

5 SUMMARY AND CONCLUSIONS

Most of the parameters used in the design of geosynthetic reinforced systems consider characterization of the ultimate failure, and typically using unconfined conditions. However, the actual performance of pavement reinforced systems governs by the interaction between surrounding soil and the geogrid in small displacement conditions. In this study, a new parameter, defined as “Stiffness of Soil-Geosynthetic Interaction” or KSGI,was introduced to address soil-geosynthetic interaction behavior under small displacements. KSGI combines the interface shear properties of the reinforced system with the load-strain properties of geosynthetic under confined conditions.

1388


Archer S. and Wayne M.H. 2012. Relevancy of Material Properties in Predicting the Performance of Geogrid-Stabilized Roadway. Proc. of the conference GeoFrontiers, Advances in Geotechnical Engineering, ASCE, Oakland, California.

Figure 7. Results of the pullout test for LVDTs 2, 3, and 4 in space)( 2 uT

Figure 8. Results of the pullout test for LVDTs 2, 3, and 4 in space)( uT

Chen Q. and Abu-Farsakh M. 2012. Structural Contribution of Geogrid Reinforcement in Pavement. Proc. of the conference GeoFrontiers, Advances in Geotechnical Engineering, ASCE, Oakland, California.

Christopher B.R., Cuelho E.V., and Perkins S.W. 2008. Development of Geogrid Junction Strength Requirement for Reinforced Roadway Base Design. Proc. of GeoAmericas 2008 Conference, Cancun, Mexico, 1003-1012.

Culeho E.V. and Perkins S.W. 2009. Field Investigation of Geosynthetics Used for Subgrade Stabilization. Report No. FHWA/MT-09-0003/8193, Montana Department of Transportation.

Gupta R. 2009. A Study of Geosynthetic Reinforced Flexible Pavement System. PhD Thesis, The University of Texas at Austin, USA, 281p.

Gurung, N. and Iwao. Y. 1999. Comparative Model Study of Geosynthetic Pull-out Response. Geosynthetics International, Vol. 6., No. 1. pp. 53-68.

Juran, I. and Chen. C.L. 1988. Soil-Geotextile Pull-Out Interaction Properties: Testing and Interpretation. Transportation Research Record 1188, 37-47.

Madhav M.R., Gurung N. and Iwao Y. 1998. A theoretical Model for the Pull-Out Response of Geosynthetic Reinforcement. Geosynthetics International, Vol. 5, No. 4, pp. 399-424.

Mahmood T., Hatami K., Ghabchi R., and Zaman M.M. 2012. Pullout Performance of Geogrids with Different Junction Strength. Proc. of the conference GeoFrontiers, Advances in Geotechnical Engineering, ASCE, Oakland, California.

Ochiai H., Otani J., Hayashic S., and Hirai T. 1996. The pullout resistance of geogrids in reinforced soil. Geotextiles and Geomembranes, Vol. 14, 19-42.

Perkins S.W. and Cuelho E.V. 1999. Soil-geosynthetic interface strength and stiffness relationships from pullout tests. Geosynthetics International 6(5), 321-346.

Perkins S.W., Christopher B.R., Cuelho E.V., Eiksund G.R., Hoff I., Schwartz C.W., Svanø G., and Want A. 2004. Development of design methods for geosynthetic reinforced flexile pavements.FHWA-DTFH61-01-X-00068, Final report, 263p.

The KSGI index was built on the basis of two major assumptions. The first assumption was linear relationship between unit tension and strain in geosynthetic reinforcement under small displacement. The slope of this line is defined as Jc,Confined Stiffness of geosynthetic. In the second assumption a uniform distribution of interface shear, defined as yield shear stress (y), is assumed over the active length of the reinforcement. Both parameters will be constant for a certain soil-geosynthetic system under specific confinement stress. Therefore, KSGI, which corresponds to 4Jcy, is constant for a defined geosynthetic reinforcement conditions. This characteristic can then be used as a basis to compare similar geosynthetic products to be placed under same working conditions in the field.

Roodi G.H. and Zornberg J.G. 2012. Effect of geosynthetic reinforcements on mitigation of environmentally induced cracks in pavements. 5th European Geosynthetics Conference, EuroGeo5, Valencia, Spain.

Sieira A.C.C.F., Gerscovich D.M.S., Sayao A.S.F.J. 2009. Displacement and load transfer mechanisms of geogrids under pullout condition. Geotextiles and Geomembranes, Vol. 27, pp. 241-253.

Sobhi S. and Wu J.T.H. 1996. An Interface Pullout Formula for Extensible Sheet Reinforcement. Geosynthetics International, Vol. 3, No. 5, pp. 565-582.

Sugimoto M. and Alagiyawanna A.N.M. 2003. Pullout behavior of Geogrid by Test and Numerical Analysis. Journal of Geotechnical and Geoenvironmental Engineering, Vol. 129, No. 4, April 1, 2003, pp. 361-371.

As an illustration, the results of a test conducted as a part of this study are presented to examine the assumptions and the outcome of the model. A biaxial geogrid was used in a conventional pullout box filled with a poorly graded sand. Five LVDTs were attached to evenly spaced nodes along the length of the geosynthetics to read the small displacements during the test. Readings from the three middle LVDTs were used to calculate the KSGI values for the system. The relationships are found to be linear, with the three values reasonably close to each other, providing evidence of validity of the model assumptions.

Wilson-Fahmy R.F., Koerner R.M., and Sansone L.J. 1994. Experimental behavior of polymeric geogrids in pullout. Journal of Geotechnical Engineering, 120(4), ASCE, USA, 661-677.

Zornberg J.G., Roodi G.H., Ferreira J., and Gupta R. 2012. MonitoringPerformance of Geosynthetic-Reinforced and Lime-Treated Low-Volume Roads under Traffic Loading and Environmental Conditions. Proc. of the conference GeoFrontiers, Advances in Geotechnical Engineering, ASCE, Oakland, California.

Zornberg J.G. and Christopher B.R. 2007. Chapter 37: Geosynthetics. In: The Handbook of Groundwater Engineering, 2nd Edition, Jacques W. Delleur (Editor-in-Chief), CRC Press, Taylor & Francis Group, Boca Raton, Florida.

6 REFERENCES

Abdelouhab A., Dias D., Freitag N., and Bennani Y. 2008. Pullout Tests Analytical Modeling to deduce the Constitutive Soil-Reinforcement Interface Behavior. Paper No. 63, 4th European Conference on Geosynthetics, EuroGeo4, Edinburgh.


Maintenance des ouvrages en terre sur Lignes à Grande Vitesse

High speed railways earthworks maintenance

V. Talfumière – C. Girier-Bichon & JB. Néel SNCF, Direction PSI, St Denis, France

RÉSUMÉ : Le réseau ferré LGV français comporte, aujourd’hui, 2030 km de plate-forme. Ces infrastructures linéaires conçues et

construites avec des moyens modernes d’études, de chantier et de contrôle, ont été construites pour une durée de vie supérieure à 100 ans : malgré ces dispositions, pendant les premières années de vie, les ouvrages en terre sont le siège de désordres nécessitant la mise en œuvre de surveillance, de reconnaissances de sols et de travaux de confortement. Cet article s’intéresse aux principaux incidents rencontrés sur les 7 LGV en service, leurs origines, leur traitement en mettant en exergue la nécessité de bien prendre en compte lesmodifications d’état hydrique induites par le projet et de suivre les référentiels de constructions qui ont été établis sur la base des retours d’expérience de la maintenance des ouvrages mis en service. Cette démarche est d’autant plus importante que le contexte ferroviaire de ces ouvrages impose des interventions complexes et coûteuses après à la mise en service.

ABSTRACT: The French HSR railroad network is composed by 2030 km of platform. These linear infrastructures designed and build with modern means of studies, construction works and monitoring, were built for more then a 100-year-old life cycle: in spite of these measures, during the first years of life, disorders appear on this earthworks; they require the implementation of surveillance, recognitions of grounds and reinforcement works. This article deals with the main incidents met on 7 HSR, their origins, their treatment by highlighting the necessity of taking into account well the modifications of hydric state inferred by the project and of following the reference tables of constructions which were established on the basis of the experience feedback of the maintenance ofthe works put in service. This approach is all the more important as the railroad context of these works imposes complex andexpensive interventions later on the starting.

MOTS CLEFS: Ouvrage en terre – géotechnique ferroviaire – maintenance – ligne à grande vitesse.

KEYWORDS: Earthworks – railroad geotechnics – maintenance – high speed line.

1 INTRODUCTION: LES LIGNES A GRANDE VITESSE

EN FRANCE

Depuis 1981, SNCF et RFF ont mis en service 7 Lignes à Grande Vitesse (LGV) :

- 1981-1983 : LGV Paris Sud Est entre Paris et Lyon - 1989-1990 : LGV Atlantique entre Paris, Le Mans et

Tours - 1993-1996 : LGV Nord entre Paris et Lille, avec

prolongement vers le tunnel sous la Manche et Bruxelles, à laquelle s’ajoutent 2 branches de l’Interconnexion en Ile de France

- 1992-1994 : LGV Rhône Alpes entre Lyon et Valence - 2001 : LGV Méditerranée entre Valence, Marseille et

Nîmes - 2007 : LGV Est (1ère phase) entre Paris et la Lorraine - 2011 : LGV Rhin-Rhône (partielle) entre Dijon et

Mulhouse. A ces 7 lignes, il faut ajouter la portion de ligne en France

entre Perpignan et Figueras dont l’exploitation et la maintenance sont déléguées à une entreprise privée.

Le génie civil et plus particulièrement les ouvrages en terre constituent une part financière importante dans la conception et la construction de l’infrastructure mais peuvent également avoir un impact significatif sur l’exploitation et la maintenance ultérieure de la ligne ferroviaire.

Ces éléments de l’infrastructure ont permis aux trains à grande vitesse de pouvoir circuler à des vitesses commerciales supérieures à 300 km/h en toute sécurité et ont participé à la grande aventure des records ferroviaires (515 km/h en 1990 et 574 km/h en 2007).

Les référentiels techniques ont été établis pour que l’infrastructure ait une durée de vie de plus de cent ans. Ils ont d’abord été conçus à partir de référentiels routiers puis se sont

différenciés sur la base de l’expérience acquise lors des travaux de constructions des premières lignes mais aussi grâce au retour d’expérience assuré lors de la maintenance des lignes après leur mise en service.

Photo 1: Traversée de la vallée de la Savoureuse sur la LGV Rhin-Rhône

2 INCIDENTOLOGIE SUR LGV:

Les lignes à grande vitesse représentent actuellement 6.5% du linéaire de l’ensemble du réseau ferroviaire national, soit 2030 km environ, sur 31 000 km. Ce réseau a moins de 30 ans mais contrairement à la croyance générale, un ouvrage en terre neuf n’est pas exempt de désordres plus ou moins significatifs pouvant entrainer des incidents engageant la sécurité et la régularité des circulations.

On recense depuis 1981 plus de 400 incidents sur LGV sur un total de plus de 5000 incidents connus sur l’ensemble du


réseau (à noter que ce recensement est exhaustif depuis 1998), soit plus de 8% des incidents.

Photo 2: Exemple d’incident peu grave sur la LGV Est Européenne – glissement d’un talus de déblai avec soulèvement de la piste et du fossé

La géométrie des plate-formes et des ouvrages en terre (déblai – remblai – tranchée rocheuse) fait que le nombre d’incidents impactant la sécurité ou la régularité des trains est très faible (largeur des accotements, présence de fossés ou de cunettes circulable, piège à cailloux en domaine rocheux, pente des talus, présence de berme en domaine meuble). Ainsi, depuis 1981, une trentaine d’incidents de ce type ont été observés dont la moitié sur la problématique particulière des fontis de la LGV Nord Europe. La proportion d’incident grave (impactant la sécurité ou la régularité des trains) est de l’ordre de 7.5% alors qu’elle est de 30% sur l’ensemble du réseau.

Photo 3 : Exemple d’incident sur la LGV Atlantique ayant eu des répercutions sur les circulations ferroviaires – érosion de talus aux abords d’une descente d’eau

On ne compte finalement qu’un seul incident, ayant entrainé le déraillement d’un train : un fontis sous voie lié aux tranchées de la guerre de 14/18 dont l’évolution a provoqué cet accident au passage d’un rame à 300 km/h.

Excepté sur la LGV Nord Europe, où la majorité des incidents sont des fontis, plus des deux tiers des incidents sur LGV sont des glissements, principalement de déblai. Ils se produisent à la faveur de périodes particulièrement pluvieuses : - les orages touchent des ouvrages ponctuels et occasionnent plutôt des coulées ou des entraînements de matériaux ; - les longues périodes pluvieuses avec ou sans apport de neige peuvent générer un grand nombre d’incidents peu graves dans des déblais sensibles aux évolutions de teneur en eau.

Les périodes des hivers et printemps 1995, 2001 et 2011 sont particulièrement représentatives.

Cette sensibilité diminue avec la maturité de la ligne ; une analyse a été menée sur la LGV Atlantique : elle montre une diminution de l’incidentologie pour ce type d’ouvrage dans le contexte géologique de ce secteur (voir graphique en Figure 1).

Figure 1 : Evolution du nombre d’incidents affectant la LGV Atlantique depuis sa construction

3 PRINCIPAUX TYPES DE DESORDRES SUR LGV:

Les principaux désordres affectant les ouvrages en terre sur LGV ne sont pas foncièrement différents de ceux des autres lignes, plus anciennes. Dans la majorité des cas, c’est l’eau qui est l’origine directe ou indirecte de leur évolution.

Les matériaux sensibles, soit aux phénomènes d’érosion (sables et limons), soit aux phénomènes d’évolution de teneur en eau (marnes et argiles) sont susceptibles de générer des désordres nécessitant par la suite un traitement. Par expérience, ce sont les parties des lignes à grande vitesse traversant le Bassin Parisien qui ont été le plus touchées (LGV Atlantique, LGV Est et LGV Nord).

3.1 Les glissements :

Ils affectent principalement les déblais mais peuvent aussi concerner quelques remblais dans des contextes particuliers. Ils se concentrent dans des matériaux de très mauvaise qualité, argiles vertes du Sannoisien, marnes supra-gypseuses, argiles à silex, argiles des Flandres, tuffeau de Touraine,…

3.1.1 Dans les déblais : Dans les ouvrages en déblai, la prise en compte des caractéristiques hydriques du site et de leur potentialité d’évolution par un apport extérieur est importante dans le cadre du projet. En effet, l’analyse des désordres affectant ces ouvrages nous montre que dans un contexte défavorable, toute augmentation de teneur en eau des matériaux est susceptible de provoqué une instabilité. Cette augmentation peut être liée à :

- un défaut de captage d’une nappe perchée en crête de déblai, en particulier quand les travaux de terrassement sont réalisés pendant une période particulièrement sèche,

- l’absence de ceinturage du déblai pour captage des drains agricoles, ou le maintien des drains entre le ceinturage et le talus,

- la géométrie mal adaptée des masques drainants mis en œuvre pour la prise en charge des eaux internes (masques perchés par exemple),

- la nature des matériaux de masque non adaptée pour un drainage optimal du corps de déblai (masque comportant trop de fines)

- le mauvais pentage des crêtes de talus qui favorise l’infiltration et l’alimentation en eau d’une nappe temporaire,

- l’absence de terre végétale sur les masques (superficiels, drainants ou poids) qui favorise l’infiltration des eaux de ruissellement et des eaux météoriques à l’arrière du masque pour venir altérer

0

5

10

15

20

25

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2007 2008 2009 2010 2011

Nom

bre

d'in

cide

nts

Ligne 431 000

Ligne 429 000


ou alimenter en eau le matériau qu’ils sont censés protéger.

3.1.2 Sur les remblais: Les remblais sont parfois le siège de glissement, la plupart du temps superficiels. Leur épaisseur peut varier de quelques décimètres (glissement de la terre végétale) jusqu’à quelques mètres, pouvant tangenter le plan fictif P1 (3/2 par rapport à la crête de la banquette de ballast) à partir duquel des mesures vis-à-vis de la sécurité des circulations peuvent être envisagées (voir photo 4).

Photo 4 : Glissement de bord de remblai sur la LGV Est Européenne

L’origine de ces désordres peut être multiple : - le mauvais compactage du bord du remblai ou

absence de retrait du remblai excédentaire après le compactage, qui peut ensuite se gorger d’eau

- la mise en œuvre du matériau à des teneurs en eau trop fortes : le matériau s’essore progressivement vers l’extérieur créant des instabilités de bord,

- l’alimentation du corps de remblai par des écoulements superficiels circulant au niveau de la plate-forme ferroviaire perméable ou dans les caniveaux à câbles.

3.2 Les coulées et les phénomènes d’érosion :

Ces phénomènes sont lies à un défaut de prise en charge des eaux de ruissellement par le système de drainage longitudinal ou transversal ; ces désordres peuvent s’expliquer par l’absence de revêtement des fossés (béton ou géosynthétiques), un défaut de continuité du drainage entre un fossé de crête par exemple et une descente d’eau, le mauvais positionnement d’une descente d’eau ou, plus simplement, l’absence de tout système de drainage en crête de déblai. Une problématique spécifique a été découverte à l’occasion d’un incident avec entrainement de ballast lors d’un orage intense au droit d’un pont route, au niveau duquel le drainage latéral était enterré. Ce drainage, compte tenu de sa capacité, de la configuration du site et des conditions d’entretien n’a pas permis de faire passer le débit qui transitait dans le fossé en amont, si bien que c’est l’accotement de la voie qui a vu s’écouler le débit supplémentaire, entraînant une érosion de la banquette de ballast. Plusieurs sites de configuration équivalente ont été le siège du même phénomène entrainant un diagnostic sur l’ensemble des lignes concernées par cette problématique. Les coulées et phénomènes d’érosion sont aussi liés à l’évolution de l’environnement de l’ouvrage en terre (hors emprises ferroviaires). Il peut s’agir d’aménagements urbains, de changement de cultures, de création d’infrastructures

linéaires parallèles ou perpendiculaires sans prise en compte de l’impact hydraulique du projet sur les infrastructures existantes.

3.3 Les fontis:

Les fontis ont affecté essentiellement la LGV Nord, où, suite au déraillement de 1993, des investigations, des traitements par injection, l’étanchement des drainages et une surveillance spécifique ont été mis en place. Mais l’ensemble des LGV sont potentiellement le siège d’incident de type fontis car toutes les lignes traversent des horizons géologiques pouvant contenir des cavités (calcaires, craie, gypses, sels). Depuis l’accident de 1993, une méthodologie d’études à toutes les phases du projet et lors des travaux a permis de traiter, si nécessaire, tout indice de cavité et de réduire ainsi de façon significative l’aléa cavité ; depuis cette période (construction de la LGV Rhône Alpes), aucun fontis d’origine naturelle n’a été découvert sur le réseau. Les seuls désordres significatifs sont liés à des ruptures ou des désordres au niveau de drainages longitudinaux ou transversaux enterrés.

3.4 Les chutes de blocs:

Les référentiels de conception demandaient dès l’origine la prise en charge de l’aléa « chute de bloc », de sorte que son impact sur la sécurité des circulations ferroviaires devait être nul.

Ainsi, dès la mise en service de la LGV Paris Sud Est, des pièges à cailloux et des confortements ponctuels ont été mis en œuvre ; depuis 30 ans, un seul incident a pu engager la sécurité des circulations, sur l’ensemble du réseau national. Mais de nombreuses chutes dans les accotements ou dans les pièges ont nécessité la mise en place d’un entretien régulier des dispositifs de protection, l’aménagement d’accès spécifiques, et le confortement préventif de certaines parois très productives en éléments rocheux de taille très diverse (revêtement grillagé, béton projeté, mur voile, barrières grillagées).

3.5 Les tassements et affaissements:

Ce phénomène est lié à la consolidation du sol support aux remblais dans les zones compressibles ou au tassement du corps de remblai lui-même. Il a été rarement à l’origine d’incidents, car cette problématique est prise en compte dès la conception dans le cadre du référentiel. Celui-ci demande un tassement total inférieur à 10 cm après la fin des terrassements (phase génie civil) avec une vitesse maximale inférieure à 1 cm par an. Par ailleurs, la vitesse des évolutions, même si elle est significative, reste faible.

Pour une dizaine de remblais sur l’ensemble des 7 lignes à grande vitesse construites depuis 1981, ces valeurs n’ont pas été respectées, entrainant une forte augmentation de la maintenance des voies (travaux de mise à niveau par bourrages mécaniques lourds) et générant des impacts collatéraux significatifs :

- augmentation de la surveillance et du suivi de ces ouvrages par topographie et inclinométrie,

- nécessité d’élargir les accotements, d’allonger les câbles de signalisation ou de rehausser les supports caténaires pour tenir compte du tassement à venir.

Dans certains cas plus critiques, la présence d’un ouvrage d’Art dans la zone (pont rail) ou d’une structure sur le remblai (tranchée couverte) a pu entrainer des travaux spécifiques. Après diagnostic, quatre types de causes principales se sont dégagées de cette problématique :

- la poursuite de tassements due à une consolidation secondaire du sol support très importante (5 cas),

- la rupture du sol support au remblai liée à un chargement trop rapide ou mal maitrisé (1 cas),

- le tassement du corps de remblai lié à un défaut de mise en œuvre (défaut de compactage ou remblai construit avec des teneurs en eau très importantes) (3 cas),


- le tassement du corps de remblai lié à une alimentation en eau externe du bassin versant (1 cas).

Dans 4 cas sur les 10, le tassement s’est accompagné d’un glissement qui a pu concerner ou non la stabilité de la plate-forme.

3.6 Les soulèvements ou gonflements:

Ces phénomènes sont liés à la nature des matériaux de l’assise : certaines argiles, de par leur structure sont susceptibles de gonfler avec un apport d’eau extérieur. Cet aléa est pris en compte dès la conception des projets : l’objectif est de limiter l’apport d’eau au maximum dans ces secteurs en étanchant la plate-forme ferroviaire et les drainages ; une seule zone est connue sur la LGV Paris Sud Est et une dizaine d’ouvrages sont suivis sur la LGV Est. Les gonflements sont difficiles à traiter par l’entretien car le nivellement des voies par abaissement du plan de roulement est impossible sans des travaux considérables ; l’entretien ne peut se faire que par augmentation du niveau de la voie de part et d’autre de la zone de soulèvement.

3.7 Les très grands ouvrages en terre:

Les très grands ouvrages en terre en terme de hauteur sont rares mais sont plus susceptibles d’être le siège de désordres de par leur géométrie ou leur géologie : 7 ouvrages dont la hauteur est supérieure à 20m ont posé des problèmes de maintenance importants, suite à la mise en service. Dans chacun des 7 cas, la problématique hydrogéologique des sites était essentielle ; une meilleure prise en compte, par une reconnaissance plus fine aurait permis de diminuer le risque d’apparition de désordres pendant ou après les travaux. Les phénomènes se sont déclarés très tôt dans la vie de l’ouvrage et ont généré pour la plupart une gêne pour les circulations.

Les deux principaux ouvrages se situent sur la LGV Méditerranée (déblai de Chabrillan et déblai des Ayasses) ; ils ont été le siège du glissement de deux collines, générant le soulèvement de la plate-forme ; le traitement a été et sera réalisé par terrassement avec amélioration du drainage interne et superficiel. Le premier ouvrage nommé a nécessité, lors d’une période de crise après un épisode pluvieux exceptionnel, la mise en place d’une surveillance particulière avec un suivi dont la fréquence a été très impactante pour les services de l’Infrastructure.

Photo 5 : Vue générale du déblai des Ayasses sur la LGV Méditerranée 4 PARTICULARITES DE LA MAINTENANCE SUR

L’INFRASTRUCTURE FERROVIAIRE

La maintenance de l’infrastructure ferroviaire est plus difficile que celle des ouvrages routiers car cette infrastructure est beaucoup plus sensible à tout mouvement ; ceci est d’autant plus vrai dans le domaine de la grande vitesse. En effet, faire circuler des trains à plus de 270 km/h nécessite de maintenir un nivellement des voies quasiment parfait (précisions de quelques millimètres). En cas de problème, la distance d’arrêt des circulations est beaucoup plus importante que dans le domaine routier. Il n’est en général pas possible de trouver des itinéraires de détournement, d’autant plus sur LGV où l’alternative est la

ligne classique avec des retards de trains très conséquents à l’arrivée. Enfin, même si c’est moins vrai sur le réseau à grande vitesse, les accès aux zones d’incident sont difficiles et nécessitent des coûts et des délais significatifs pour leur création. Pour toutes ces raisons, la surveillance doit permettre de détecter le plus tôt possible l’apparition d’indices de désordre.

L’exploitation de ce type de ligne entraine également des contraintes qui ont un impact sur le coût des interventions :

- travaux de nuit en cas d’accès par la voie ou en cas de risque de déstabilisation de la plate-forme,

- terrassement par plots de longueur réduite, - création d’accès de grande longueur.

Lors de la conception tout doit être fait pour réduire au maximum l’entretien courant (hors travaux de confortement) des ouvrages après leur mise en service.

Le retour d’expérience montre que deux points méritent d’être examinés plus particulièrement : l’entretien des drainages et la maîtrise de la végétation. Les drainages doivent rester efficaces pendant la durée de vie de l’ouvrage mais dès les premières années ont été constatés :

- des phénomènes de calcification dans les drainages enterrés,

- la dégradation des conditions d’écoulement dans les fossés en terre, liée, le plus souvent, à un manque de pente ou à la pousse dans les fossés d’une végétation herbacée ou arbustive excessive.

L’absence ou le déficit d’entretien de ces installations peut à terme entrainer des désordres avec impact sur la régularité voire ponctuellement sur la sécurité des circulations.

La maitrise de la végétation est nécessaire au niveau des drainages mais aussi pour une bonne surveillance des ouvrages les plus instables.

5 CONCLUSIONS

La majorité des désordres affectant les ouvrages sur ligne à grande vitesse en phase exploitation est liée à un déficit de prise en compte de la problématique hydrogéologique, hydrologique voire hydraulique, mais aussi aux écarts constatés par rapport aux référentiels de conception ; ces écarts sont parfois présents dès la conception mais le plus souvent apparaissent lors la phase « réalisation ». L’établissement de demandes de dérogation apparait évident, dès que le concepteur s’écarte du référentiel. En outre, le contrôle lors des travaux de génie civil et une vision maintenance tant en phase travaux que lors de la réception peut apporter un plus vis-à-vis de la vie future de l’ouvrage en phase exploitation.

Ces remarques n’empêchent pas, bien sûr, de proposer des innovations ou des technologies éprouvées en routier mais non mises en œuvre dans le ferroviaire, sous réserve d’en avoir pesé les conséquences en terme de durabilité, de maintenabilité et de sécurité vis-à-vis des circulations.

L’expérience de 7 lignes à grande vitesse mise en service depuis 1981, montre aussi que pour une maintenance optimisée des lignes, il est primordial d’établir en même temps que les travaux, des dossiers de récolement de bonne qualité, exhaustifs et représentatifs de ce qui a été réellement construit.

6 REFERENCES

Girier-Bichon C. 2008. Retour d’expérience des désordres affectant les ouvrages en terre sur ligne à grande vitesse. Rapport de fin d’études CESFA-BTP -SNCF.

Lambert L. et al. 2011. 40 ans d’expérience de ligne à grande vitesse-tracé, assise et hydraulique : pratique, interdits et développements. Géorail2011, Paris.

Deherripont JL. et al. 2010. Etude d’un glissement hors normes en déblai ferroviaire. JNGG2010, Grenoble.

Talfumière V. 2011. Maintenance des ouvrages en terre sur le réseau ferré national. Revue française de géotechnique n°134/135, Paris.

technical committee 202 transportation · pdf filetechnical committee 202 transportation...

Documents