Tunnelling and Underground Space Technology 52 (2016) 147–156

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

journal homepage: www.elsevier .com/ locate/ tust

Evaluation of common TBM performance prediction models basedon field data from the second lot of Zagros water conveyance tunnel(ZWCT2)

http://dx.doi.org/10.1016/j.tust.2015.12.0060886-7798/� 2015 Elsevier Ltd. All rights reserved.

⇑ Corresponding author.E-mail address: hassanpour@ut.ac.ir (J. Hassanpour).

J. Hassanpour a,⇑, A.A. Ghaedi Vanani a, J. Rostami b, A. Cheshomi a

aUniversity of Tehran, College of Science, School of Geology, Dept. of Engineering Geology, Tehran, Iranb Pennsylvania State University, Dept. of Energy and Mineral Engineering, University Park, PA, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 10 June 2015Received in revised form 20 October 2015Accepted 11 December 2015Available online 17 December 2015

Keywords:TBM performanceRate of penetrationEmpirical modelsTheoretical models

Zagros water conveyance tunnel (ZWCT) is a 49 km tunnel designed for conveying 70 m3/s water fromSirvan River southward to Dashte Zahab plain in western Iran. This long tunnel has been divided in 3Lots namely 1A, 1B, 2. By November 2014, about 22 km of the Lot 2 (with a total length of 26 km) hasbeen excavated by two double shield TBMs from two southern and northern portals. The bored sectionof tunnel passed through different geological units of 3 main formations of Zagros mountain rangeswhich mainly consist of weak to moderately strong argillaceous-carbonate sedimentary rocks. In thispaper, the operating and as-built geological data collected during construction phase of the Lot 2 ofZWCT project was used to compare the calculated machine performance by empirical methods such asthe Hassanpour et al. (2011), QTBM, NTNU, Palmstrom, and theoretical model of Colorado School ofMines or CSM. The predicted penetration rates were then compared with the observed field performanceof the machine and the variations of predicted rates were examined by statistical analysis. The resultsshowed that the site-specific model, which was based on TBM performance in similar formations can pro-vide estimates closer to actual machine performance.

� 2015 Elsevier Ltd. All rights reserved.

1. Introduction

In last two decades, many performance prediction models havebeen developed by various researchers to estimate penetration rateof hard rock tunnel boring machines (TBMs) in new tunneling pro-jects. Some of the most important TBM prognosis models includeCSM (Rostami, 1997), NTNU (Bruland, 1998), QTBM (Barton, 2000),and models proposed by Palmstrom (1995), Yagiz (2002), Gongand Zhao (2009), Hassanpour et al. (2009, 2010, 2011a,b),Hassanpour (2010), and Delisio and Zhao (2014). On the otherhand due to observation of some shortcomings in existing predic-tion models and their application, it was necessary to continueresearch on the subject and for development of adjustments forthe existing models and correction factors for special circum-stances that would allowmore accurate prediction of machine per-formance. Many researchers have conducted studies to modify andadjust common prediction models based on data collected fromdifferent projects. Bruland (1998) updated and improved NTNUmodel (originally introduced by Blindheim (1979)) based on field

data collected from Norwegian tunnels. Yagiz (2002) has proposednew formulas involving additional rock parameters to modify CSMmodel by analyzing data from Queens’s. Ramezanzadeh (2005)analyzed data from projects in Norway, Hong Kong and Switzer-land and attempted to develop new equations for adjusting resultsof CSM model.

The current study is an attempt to compare the results of abovementioned prediction models, including Hassanpour et al., Palm-strom, NTNU, QTBM and CSM, with actual TBM performance, basedon collected data from 9.5 km bored section of southern portal ofZWCT Lot2. This tunnel is under construction in geological zoneof Zagros Mountains in western Iran. This paper will describe thetunneling project, geological conditions, estimated rates of pene-tration (ROP) obtained from these models, and compares the esti-mates with actual ROPs in different tunnel sections. Accuracy andreliability of each model in geological conditions of this project willbe examined.

2. TBM performance prediction models

A wide variety of performance prediction methods and princi-ples are used to estimate performance (penetration and advance

148 J. Hassanpour et al. / Tunnelling and Underground Space Technology 52 (2016) 147–156

rate) of a TBM in hard rock. Different models are used in differentcountries, contractors, engineers, and by various TBM manufactur-ers based on their experience and available information. Some ofthe methods are based mainly on one or two rock parameters(for instance uniaxial compressive strength and a rock abrasionparameter) while others are based on a combination of compre-hensive laboratory, field and machine parameters. In general,methods for TBM performance prediction are classified in the fol-lowing categories:

1. Theoretical/Experimental models (based on laboratory testingand cutting forces).

2. Empirical methods (based on field performance of TBMs andsome rock properties).

In this research, CSMmodel from first category and NTNU, QTBM,Palmstrom and Hassanpour et al. models from second categorywere used to analyze field performance data related to the TBMapplication in Zagros tunnel. Table 1 presents an overview of theapplied models in this paper.

Among the models presented in Table 1, Hassanpour et al.model has been developed based on data collected from some tun-neling projects completed in Iran in recent years including first5.3 km of ZWCT2 project. Actual machine performance andas-built geological data from three major hard rock mechanizedtunneling projects in Iran (Karaj, Zagros and Ghomrood water con-veyance tunnels) and Manapouri second tailrace tunnel in NewZealand have been collected and analyzed to develop relationshipsbetween machine performance and geological parameters. Table 2lists empirical equations proposed in Hassanpour et al. model forestimating TBM performance in different geological conditions.

As can be seen in Table 2, among the machine performanceparameters, field penetration index or FPI, which is a compositeparameter based on penetration and cutter load, have beenselected for developing empirical equations. Actually, this parame-ter showed the best correlations with geological parameters(Hassanpour et al. 2009, 2010, 2011a,b). This special parametercan be calculated from machine operating parameters using thefollowing equation:

FPI ¼ Fn

P¼ 60FnRPM

1000ROPð6Þ

where Fn is average cutter load (kN/cutter), P is penetration(mm/revolution), ROP is rate of penetration (m/h) and RPM iscutterhead speed (revolution per minute). Fn is calculated bydividing operational thrust by number of disc cutters aftersubtracting the shield friction.

Table 1Input parameters in applied prediction models.

Predictionmodel

Required input parameters

Rock mass parameters Machine p

CSM Uniaxial compressive strength (UCS), Brazilian tensilestrength (BTS) and Cerchar Abrasivity Index (CAI)

Cutter loacutter diamTBM Thru

NTNU Fracturing: frequency and orientation, Drilling Rate Index(DRI), Bit Wear Index (BWI) and Cutter Life Index (CLI),Porosity, and other parameters

Cutter thrdiameter

QTBM RQD0, Jn, Jr, Ja, Jw, SRF, rock mass strength, cutter lifeindex (CLI), UCS, induced biaxial stress

Average cu

Palmstrom Elastic modulus, UCS, block size, roughness, length anddirection of joints

Thrust, cuof the cuttcutterhead

Hassanpouret al.

UCS, RQD Thrust, rothead, num

3. Project description

Zagros water conveyance tunnel with total length of 49 km and6.73 m diameter has been designed for conveying 70 m3/s waterfrom the Sirvan River in South of Nowsood city to Dasht e Zahabplain in Kermanshah province in western Iran. The tunnel wasdivided into three sections including Lot 1A (14 km) as northeastsection, Lot 1B (9 km) as middle section, and Lot 2 (26 km) assouthwest section (Fig. 1). Currently, Lot 2 is under constructionfrom two portals at south and north ends. Two other sections havebeen constructed from portals located in two deep valleys crossingthe tunnel route.

At southern portal of Lot 2, a TBM was launched from a 200 mstarter tunnel excavated by drill and blast method. The machinecommenced excavation on March 2006 and at end of 2014, morethan 19 km of tunnel was completed from southern portal. In July2011, the TBM work was temporarily suspended to carry out acomprehensive overhaul and optional improvements on themachine. The overhaul was conducted in an excavated cavern atthe intersection with an existing access tunnel at the chainage of14,850 m of the main tunnel.

After encountering adverse geological conditions and dramaticreduction in TBM performance at beginning of the southern sec-tion, another TBM was mobilized and started to excavate the tun-nel from northern portal in mid-2012. This machine has completedaround 3 km of the tunnel in very difficult hydrogeological condi-tions. This tunnel is lined with pre-cast concrete segments withhexagonal arrangement and thickness of 25 cm.

4. Site geology

The area around the ZWCT2 project is located in Zagros zone, asone of the main geological and structural zones of Iran and in ‘‘sim-ply folded” subzone. The main geological formations outcropped inthe project area are Pabdeh, Gurpi and Ilam Formations. These arethe major formations of the Zagros Mountain range which starts innorthwest and extends to southeast of Iran. These formations arecomposed of a variety of carbonate and argillaceous rock units.Geological section in Fig. 2 shows the distribution of these strati-graphic units along the tunnel. Pictures (a) to (c) of Fig. 3 show dif-ferent identified engineering geological units in outcrops andtunnel faces.

Due to considerable difference in engineering properties of thestratigraphic units, they can be considered as different engineeringgeological units. As shown in the geological section, due to foldedstructure of the formations, the geological units may be presentthrough the tunnel alignment in different sections repetitively.

Output parameters Limitations

arameters

d capacity, cutter spacing,eter, cutter tip width, and

st and Torque

Penetration The original model isbased only on intactrock properties

ust, cutter spacing, cutter Penetration rate,advance rate,utilization factor,tunnel cost

Determination ofinput parametersneeds special tests

tter load, TBM diameter Penetration rate,advance rate

The model Uses manyparameters

tter distance, size and shapeer, round the cutter head,power

Penetration rate,advance rate

ation speed of the cutterber of disc cutters

Penetration rate,boreability class

Table 2Empirical equations developed by the first author and his colleagues for different geological conditions.

No. Equation Eq. no Project Application range Reference

1 FPI ¼ exp ð0:005UCSþ 0:002RQDþ 2:129Þ (1) KWCT1 Pyroclastic rocks Hassanpour et al. (2010)2 FPI ¼ exp ð0:004UCSþ 0:008RQDþ 2:077Þ (2) ZWCT2 Carbonate-argillaceous rocks Hassanpour et al. (2009)3 FPI ¼ exp ð0:004UCSþ 0:023RQDþ 1:003Þ (3) GWCT Low grade metamorphic rocks Hassanpour (2010)4 FPI ¼ exp ð0:005UCSþ 0:020RQDþ 1:644Þ (4) MSTT Igneous and metamorphic rocks Hassanpour et al. (2011b)5 FPI ¼ exp ð0:008UCSþ 0:015RQDþ 1:384Þ (5) All projects General Hassanpour et al. (2011a)

Fig. 1. Details of Zagros water conveyance tunnel scheme, west of Iran.

Fig. 2. Geological cross section along the tunnel.

J. Hassanpour et al. / Tunnelling and Underground Space Technology 52 (2016) 147–156 149

(a) (b) (c)

PEPd

2`

PEPd

2`

KGu

5`

Fig. 3. (A) A view of Pabdeh and Gurpi contact and stratigraphic units identified along the tunnel alignment, (b) alternation of shale and shaly limestone in tunnel face, (c)views of limestone blocks in tunnel face.

Table 3Main characteristics of lithotypes (Hassanpour et al., 2009).

No Lithotype Major minerals Quartz content (%) Porosity (%) UCS (MPa) BTS (MPa) Drilling rate index (DRI) Cutter life index (CLI)

1 Limestone Calcite, Clay minerals <5 10–12 100–150 8–10 55–60 75–852 Shaly limestone <5 5–8 50–100 6–8 60–65 70–753 Limy Shale Clay minerals, Calcite 5–15 8–10 20–50 4–6 65–70 60–704 Shale 10–15 5–10 15–30 3–5 70–75 50–60

150 J. Hassanpour et al. / Tunnelling and Underground Space Technology 52 (2016) 147–156

Structurally, the area around the tunnel is moderately foldedand gently faulted. As shown in Fig. 2, tunnel in the bored sectionhas passed through some minor synclines and anticlines. There areno important faults in bored section of tunnel route, but someminor faults and shear zones have been identified as crossing thetunnel line. The thickness of these fault zones, are estimatedthrough back mapping of tunnel and it is recognized that generallyranges between 10 and 25 m.

Results of petrographic analyses show that considering mineral-ogy and texture, there are fourmain Lithotypes in tunnel alignment.These lithotypes include (1) Limestone, (2) Shaly limestone, (3)Limy shale and (4) Shale. Two first lithotypes are competent andmore brittle rocks with developed joint systems and two othersare incompetent rocks with plastic behavior and weathered fea-tures at outcrops. Main petrographic, physical and mechanicalcharacteristics of these lithotypes are summarized in Table 3.

In this study, to determine drillability indices, 12 sets of Norwe-gian tests were performed by SINTEF laboratory on samples takenfrom different boreholes along tunnel alignment. Also a series oftests were performed using devices in a local laboratory. To inves-tigate relationship between drillability indices and mechanicalparameters of rock, some common rock mechanic tests (such asUCS, BTS. . .) were conducted on same samples used for drillabilitytests and results have been analyzed separately. Using these anal-yses drillability indices for each lithotype were estimated and sum-marized in Table 3.

In the course of tunneling, the machine encountered manyextraordinary situations related to intrusion of hydrogen sulfide(H2S) and some methane gases released in the tunnel as theyexolved from groundwater. These events resulted in significantdowntimes for mitigation of the toxic and hazardous gases andmodification of the equipment, reduction in TBM utilization rate,and increase in construction delays.

Groundwater conditions along the tunnel varies from dry tocontinued flow of water in carbonate aquifers. During constructionwherever tunnel passed through Pabdeh and Gurpi formations, nomajor water inflow was observed. Groundwater condition along

this parts of tunnel was ‘‘damp to wet” and infrequently ‘‘drip-ping”. In contrast, while entering the Ilam limestone, water inflowincreased immediately and exceeded 100 l/s in some sections. Asshown in Fig. 4, the H2S gas was first encountered at chainage3700 m and reoccurred at several geological zones. The main haz-ards of the gasses were toxicity to human and corrosive effects onmechanical and electrical equipment. Presence of hydrogencyanide (HCN) was also detected by field and laboratory tests atchainage 4446 m which resulted in a 3 week TBM shutdown. Atchainage 4157 m, a first significant water ingress was encountered(Q � 110 l/s), which gradually accumulated to Q = 730 l/s with fur-ther advancement to chainage 8256 m. At chainage 13,846 m,155 l/s of gas rich water intruded into the tunnel, totaling the tun-nel accumulative discharge flow at the portal to 900 l/s. Thisamount of water released 700 ppm H2S gas into atmosphere(Jalilian Khaveh, 2013).

5. TBMs specifications

As mentioned before, a double shield machine was selected forexcavating the tunnel from southern portal. Main technical speci-fications of this machine are summarized in Table 4. Fig. 5 showstwo pictures of the machine which excavated the tunnel fromsouthern portal. The TBM mining the tunnel from northern portalis a different double shield machine with specifications similar tothe machine in Fig. 5.

6. Actual machine performance

This study is focused on TBM performance for the first 14.8 kmof southern section of the ZWCT Lot 2 project. This part of tunnelwas completed in July 2011, before machine overhauling at anexcavated cavern in chainage 14,850 m. According to Fig. 6, exca-vated length of this section of tunnel was completed in 285 fullworking weeks at an average advance of 50 m/week and best weekadvance of more than 180 m. As shown in Fig. 6, occurrence of geo-

Fig. 4. Gas concentration and total water discharge rate measured in the tunnel (Jalilian Khaveh, 2013).

Table 4Main specifications of double shield TBM (model S-157) used for constructing thetunnel from southern portal (Hassanpour et al., 2009).

Parameter TBM (southwestern portal)

Machine type Double shieldManufacturer HerrenknechtMachine Diameter (m) 6.73Cutters diameters (mm) 432Number of Disc Cutters 42Cutter Tip Width (mm) 19Disc nominal spacing (mm) 90Max. Thrust per Cutter (KN) 350 NominalAvailable Cutterhead Thrust (KN/Cutter) 590Number of Thrust Cylinders 10Power of each Drive Motor (KW) 350Number of Drive Motors 6Cutterhead RPM 0–9Cutterhead Torque (KN m) 2967 (Nominal), 4450 (Overload)TBM Stroke (m) 0.8

J. Hassanpour et al. / Tunnelling and Underground Space Technology 52 (2016) 147–156 151

logical problems (gas and water inflow toward the tunnel) causedmany long delays in TBM operation and considerable reduction inaverage advance rate and utilization factor (Fig. 7).

Fig. 7 shows the utilization factor for TBM operation in this pro-ject and while rates higher than 25% have been realized in manyindividual weeks, due to long TBM stoppages, average utilizationrate for the whole tunnel was less than 14%.

Fig. 5. Double shield TBM at southern portal, (a) during assemb

Fig. 8 shows recorded rate of penetration, averaged on aweekly basis. As shown in this figure, the normal range of ROPwas 1.5–3 m/h with an average of 2.5 m/h for the total length ofbored section.

7. Development of TBM performance database

As mentioned earlier, the main objective of this study wasevaluation of accuracy of most common TBM performanceprediction models in a given geological condition encountered atthis site (Foliated, Jointed and blocky argillaceous and carbonaterocks). As the first step to achieve this goal, a database has beendeveloped to record and organize collected geological and machineperformance data. This database includes three main sections asfollows:

1: Operational and machine performance data measured or cal-culated in tunnel;

2: As-built geological and geomechanical data;3: Results of performance prediction by different models.

Each row in the database is related to an engineering geologicalunit identified along the tunnel and recorded and calculatedparameters have been averaged for each unit. It should be notedthat the data from first 5.3 km of tunnel has been previously used

ling and (b) a view of cutterhead (Hassanpour et al., 2009).

Fig. 6. Weekly advance of machine during construction of southern section of ZWCT2.

Fig. 7. Utilization factor of machine during construction of southern section of ZWCT2.

Fig. 8. Weekly average of ROP during construction of southern section of ZWCT2.

152 J. Hassanpour et al. / Tunnelling and Underground Space Technology 52 (2016) 147–156

for development of Eq. (2) in Table 2 (Hassanpour et al., 2009),therefore it was excluded from this analysis. The database for cur-rent paper includes only the remaining 9.5 km of the available TBMfield performance data from chainage 5300–14,800 m.

7.1. TBM performance data

First section of database includes parameters such as net boringtime, length of the unit, and average of machine operational

J. Hassanpour et al. / Tunnelling and Underground Space Technology 52 (2016) 147–156 153

parameters (thrust, RPM, power and applied torque) through theunit. These parameters were obtained from daily operatingsheets and TBM data logger. Also the most important performanceparameters including average penetration per revolution, rate ofpenetration (ROP), field penetration index (FPI), and specificenergy have been calculated and are recorded in Section 1 ofdatabase.

Graphs presented in Fig. 9(a–d) show the variations of the FPI, Por penetration (mm/rev), total TBM thrust and cutterhead RPM atdifferent identified geological units along the bored section of tun-nel used in this study. As can be seen in the graphs, the minimumand maximum FPI and ROP in different geological units are 11 and22 kN/cutter/mm/rev and 2 and 12 mm/rev, respectively.

7.2. Ground characteristics

In Section 2 of the database, results of ground characterization(laboratory and field measured parameters) are compiled. To char-acterize the ground along the tunnel, two sets of parametersincluding intact rock properties and rock mass parameters wereconsidered. Using these two sets of parameters, rock masses iden-tified along the bored section of tunnel were subdivided into 41units with uniform characteristics related to TBM performance,tunnel stability and groundwater inflow (engineering geologicalunits).

Intact rock properties: In general, the TBM performance is con-trolled by intact rock properties such as petrographic characteris-tics, physical and mechanical properties and drillabilty. Each ofthese parameters have been evaluated in the laboratory using avariety of testing methods. In addition to tests performed onpre-construction phase, many samples were obtained and tested

Fig. 9. Variations of average TBM performance parameters along the tunnel; (a) RPM; (b

during construction phase. Average values of intact rock propertiesfor each tunnel section (engineering geological units) fromdifferent phases of testing were calculated and recorded in thedatabase.

Rock mass parameters: Fractures, joints and discontinuitiesinfluence boreability of rock mass depending on their spacing, sur-face condition and orientation with respect to the direction ofmachine advance. In this study, these parameters for each engi-neering geological unit have been evaluated using measurementsand observations within the tunnel as well as surface outcrops.Average values of rock mass parameters (and intact rock proper-ties) have been used to determine geomechanical conditions ofthe identified engineering geological units by some empirical rockmass classification systems including RQD (Deer, 1964), RMR(Bieniawski, 1989), GSI (Hoek et al., 2002) and Q-system (Bartonet al., 1974). The variations of rock mass parameters in identifiedengineering geological units in the database are plotted againstrespective unit name in Fig. 10.

7.3. Estimation of TBM performance using common prediction models

Based on data arranged in Sections 1 and 2, and using selectedTBM performance prediction models, rate of penetration at eachtunnel section has been calculated and recorded in Section 3 ofthe database. In this research, recorded machine parameters andas-built geological and geomechanical data are used as input datafor usual prognosis models introduced in Table 1. Main outputparameter from these models is rate of penetration (ROP) at eachtunnel section (or engineering geological unit). Calculation proce-dure for each model has been explained in related references pre-sented in Table 1.

) Total TBM thrust; (c) Field penetration index (FPI); (d) Disc cutter penetration (P).

Fig. 10. Variations of (a) UCS and RQD0 and (b) rock mass parameters of engineering geological units identified along the tunnel alignment.

Table 5Descriptive statistics of absolute errors computed for the prediction models.

Model N Range Minimum Maximum Mean Std. deviation

NTNU 41 58.75 1.55 60.30 30.74 16.43CSM 41 112.94 2.09 115.03 41.66 28.60QTBM 41 258.72 1.00 259.72 69.26 66.59Palmstrom 41 52.22 1.16 53.38 21.24 14.84Eq. (2) 41 39.10 0.91 40.01 16.94 11.05Eq. (5) 41 87.23 0.45 87.68 23.92 22.78

154 J. Hassanpour et al. / Tunnelling and Underground Space Technology 52 (2016) 147–156

8. Comparison between the results of different TBMperformance models

Fig. 11 shows estimated ROP (m/h) for the TBM in each unitusing above mentioned prediction models. Also, variations of abso-lute errors or E(%) for each model in each tunnel section, are shownin the graphs of Fig. 12. Absolute errors were calculated accordingto following formulae:

Eð%Þ ¼ 100 � ActualROP � EstimatedROPActualROP

����

����

ð7Þ

A summary of the statistical analysis performed on calculatedrates and respective errors are presented in Table 5. Following con-clusions can be made by reviewing Figs. 11 and 12 and Table 5:

1. There are significant differences in the ROPs estimated usingdifferent prediction models. The NTNU and Palmstrom models

show lower values than measured ones, while two models ofCSM and particularly QTBM show very higher values.

2. The results obtained from QTBM model are very far from actualvalues. The main reason is consideration of some parameters(like in-situ stress) in the model which their influence on TBMperformance in this project have not been proved yet. The high-est values of E (%) are related to this model (Table 5 and Fig. 12).

3. The results obtained from the original CSM model do not matchthe measured values, because this model has been basicallydeveloped for massive rocks with no significant fracturing. Thisstudy shows that this model needs modification by adding cor-rection factors for jointing condition. In recent years, someresearchers (Yagiz, 2002; Ramezanzadeh, 2005] have proposedthe correction factors to update the model for jointed rockmasses.

4. The results obtained from Palmestrom and NTNU models indi-cate relatively good matches with the actual results. Thesetwo empirical models apply rock mass parameters which arevery important in boring process. Difference in geology of pro-jects used in the original database of the models which wassubsequently used for development of this method and the casestudy for this paper seem to be the weakness point of thesemodels.

5. As expected, among the studied models the best resultsobtained from Eq. (2) which has been developed based on datacollected from first 5.3 km of this project. As shown in Table 5and Fig. 12, the lowest values of E (%) are related to this model.

0

1

2

3

4

5

6

7

8

9

10

SH-L

S3SH

-LS2

SH-L

S1M

L-SH

5C

Z FZSH

-LS2

SH-L

S3 CZ

SH-L

S3SH

-LS2

SH-L

S1M

L-SH

5C

ZSH

-LS3

SH-L

S2SH

-LS1

SH-L

S2SH

-LS3

SH-L

S4 CZ

SH-L

S4SH

-LS3

SH-L

S2SH

-LS3

SH-L

S4 CZ

SH-L

S4SH

-LS3

SH-L

S2SH

-LS1

ML-

SH5

SH-M

L1M

L-SH

5SH

-LS1

SH-L

S2SH

-LS3

SH-L

S4 LI2

CZ

LI2

ROP

(m/h

)

Eng. Geological Units

QTBM model CSM model Actual ROP

0

1

2

3

4

5

6

7

8

9

10

SH-L

S3SH

-LS2

SH-L

S1M

L-SH

5C

Z FZSH

-LS2

SH-L

S3 CZ

SH-L

S3SH

-LS2

SH-L

S1M

L-SH

5C

ZSH

-LS3

SH-L

S2SH

-LS1

SH-L

S2SH

-LS3

SH-L

S4 CZ

SH-L

S4SH

-LS3

SH-L

S2SH

-LS3

SH-L

S4 CZ

SH-L

S4SH

-LS3

SH-L

S2SH

-LS1

ML-

SH5

SH-M

L1M

L-SH

5SH

-LS1

SH-L

S2SH

-LS3

SH-L

S4 LI2

CZ

LI2

ROP

(m/h

)

Eng. Geological Units

NTNU model Palmstrom model Actual ROP

0

1

2

3

4

5

6

7

8

9

10

SH-L

S3SH

-LS2

SH-L

S1M

L-SH

5C

Z FZSH

-LS2

SH-L

S3 CZ

SH-L

S3SH

-LS2

SH-L

S1M

L-SH

5C

ZSH

-LS3

SH-L

S2SH

-LS1

SH-L

S2SH

-LS3

SH-L

S4 CZ

SH-L

S4SH

-LS3

SH-L

S2SH

-LS3

SH-L

S4 CZ

SH-L

S4SH

-LS3

SH-L

S2SH

-LS1

ML-

SH5

SH-M

L1M

L-SH

5SH

-LS1

SH-L

S2SH

-LS3

SH-L

S4 LI2

CZ

LI2

ROP

(m/h

)

Eng. Geological Units

Hassanpour et al. (2011) Hassanpour et al. (2009) Actual ROP

0

1

2

3

4

5

6

7

0 1 2 3 4 5 6 7

Es�m

ated

RO

P (m

/h)

Actual ROP (m/h)

Hassanpour et al.…Hassanpour et al.…

0

1

2

3

4

5

0 1 2 3 4 5

Es�m

ated

RO

P (m

/h)

Actual ROP (m/h)

Palmstrom modelNTNU model

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10

Es�m

ated

RO

P (m

/h)

Actual ROP (m/h)

QTBM modelCSM model

Fig. 11. Comparison of actual ROP with calculated ROP using different models.

Fig. 12. Variations of absolute error calculated for different models in each tunnel section.

J. Hassanpour et al. / Tunnelling and Underground Space Technology 52 (2016) 147–156 155

6. To calculate the performance parameters of the machine usingHassanpour et al. (2011) model, the general equation (Eq. (5),Table 2) was also applied. As shown in Fig. 12, although trendof variations matches the actual values, there are significanterrors in some tunnel sections.

7. As an important conclusion, it must be emphasized that sitespecific prediction models are the best models for estimatingTBM performance in a given project. This requires real timeanalysis of the machine performance and establishing a rele-vant site specific database to capture the ranges of variation

156 J. Hassanpour et al. / Tunnelling and Underground Space Technology 52 (2016) 147–156

and relationship between machine performance and geologicalunits. This is due to the fact that machine related parametersare constant and as such, the variation in geological conditionsalong the tunnel can be singled out as the input parameter forthe modeling. The downside of this approach is that the modelsdeveloped in this manner cannot be used elsewhere.

8. Using more comprehensive and established models should beconsidered when there is no track record of site-specificmachine performance and for project design purposes. In suchcases, it is recommended to use various models to observe therange of variations and sensitivity of these models to inputparameters for the specific project under study.

9. Conclusion

In this study, accuracy and validity of some common TBM per-formance prediction models were examined based on recordedmachine performance from a long tunnel under construction inwestern Iran. Reliable data was available for 9.5 km of this tunneland was used for this study. The geological units along the tunnelwere subdivided into 41 engineering geological units and rock androck mass properties were measured and complied in a TBM per-formance database along with actual TBM performance and oper-ational parameters for each tunnel section. This information wassubsequently used to calculate TBM performance using commonprediction models.

The main conclusion of this study is that, for selection of anempirical model to predict TBM performance in any case oneshould pay attention to the application range of the model andgeological conditions that the original model is based on. Thisstudy indicated that the site specific models with similar geologicalparameters and machine specification, show best matches withactual TBM performance. In this case, as expected, the results ofperformance predictions by empirical equation developed fromfirst 5.3 km of the ZWCT2 project showed the best match withTBM performance in continuation of the project. So, it is recom-mended that engineers and contractors establish a TBM field per-formance database early in the project and use it to develop asite specific model and predict the anticipated machine perfor-mance for the rest of the project to minimize the errors of generalTBM performance models and to provide a more accurate projec-tion of the project completion time.

References

Barton N., Lien, R., Lunde, J., 1974. Engineering classification of rock masses for thedesign of tunnel support. Rock Mechanics 6(4), 189–236, Springer – Verlag.

Barton, N., 2000. TBM Tunneling In Jointed and Faulted Rock. Balkema, Rotterdam,p. 173.

Bieniawski, Z.T., 1989. Engineering Rock Mass Classifications: A Complete Manualfor Engineers and Geologists in Mining Civil and Petroleum Engineering. JohnWiley, p. 251.

Blindheim, O.T., 1979. Boreability Predictions for Tunneling, Ph.D. thesis.Department of geological engineering, The Norwegian Institute of Technology,p. 406.

Bruland, A., 1998. Hard rock tunnel boring, vol. 1-10, Ph.D. thesis, Norwegianuniversity of science and technology (NTNU), Trondheim, Norway.

Deer, D.U., 1964. Technical description of rock cores for engineering purpose. RockMech. Eng. Geol. 1 (1), 17–22.

Delisio, A., Zhao, J., 2014. A new model for TBM performance prediction in blockyrock conditions. Tunnel. Undergr. Space Technol. 43, 440–452.

Gong, Q.M., Zhao, J., 2009. Development of a rock mass characteristics model forTBM penetration rate prediction. Int. J. Rock Mech. Mining. Sci. 46 (1), 8–18.

Hassanpour, J., Rostami, J., Khamehchiyan, M., Bruland, A., 2009. Development newequations for performance prediction. Geo Mech. Geoeng.: An Int. J. 4 (4), 287–297.

Hassanpour, J., Rostami, J., Khamehchiyan, M., Bruland, A., Tavakoli, H.R., 2010. TBMperformance analysis in pyroclastic rocks, a case history of Karaj WaterConveyance Tunnel (KWCT). J. Rock Mech. Rock Eng. 4, 427–445.

Hassanpour, J., 2010. Analysis of actual TBM performance in Ghomrood project,Bulletin of Iranian Tunneling Association, No 9, (article in Persian).

Hassanpour, J., Rostami, J., Zhao, J., 2011a. A new hard rock TBM performanceprediction model for project planning. Tunnel. Undergr. Space Technol. 26, 595–603.

Hassanpour J., Rostami J., Tarigh Azali S., 2011b. Investigating relationshipsbetween TBM performance and engineering rock mass parameters inmetamorphic and igneous rocks of Manapouri project (New Zealand), FourthIranian Rock Mechanics Conf., Tehran, Iran, (article in Persian).

Hoek, E., Caranza-Torres, C.T., Corkum, B., 2002. Hoek-Brown failure criterion-2002edition. In: Bawden, H.R.W., Curran, J., Telsenicki, M. (Eds.), Proceedings of theNorth American Rock Mechanics Society (NARMS-TAC 2002). MiningInnovation and Technology, Toronto, pp. 267–273.

Jalilian Khaveh, G.H., 2013. TBM tunnelling in hydrogen sulfide gas bearing groundand its solutions. Geotechn. Geol. Eng. 31, 1621–1638.

Palmstrom, A., 1995. RMi, a rock mass characterization system for rock engineeringpurposes, chapter 7, Ph.D. thesis, Oslo University, Norway.

Ramezanzadeh, A., 2005. Performance analysis and Development of new models forperformance prediction of hard rock TBMs in rock mass. Ph.D. thesis, INSA,Lyon, France.

Rostami, J., 1997. Development of a force estimation model for rock fragmentationwith disccutters through theoretical modeling and physical measurement ofcrushed zone pressure, Ph.D. thesis, Colorado School of Mines, Golden, Colorado,USA, p. 249.

Yagiz S., 2002. Development of rock fracture and brittleness indices to quantify theeffects of rock mass features and toughness in the CSM model basic penetrationfor hard rock tunneling machines. Ph.D. thesis, Department of Mining and Earthsystems Engineering, Colorado school of Mines, Golden, Colorado, USA, p. 289.