cyclic behavior of fine coal refuse and seismic …
TRANSCRIPT
The Pennsylvania State University
The Graduate School
CYCLIC BEHAVIOR OF FINE COAL REFUSE AND SEISMIC
STABILITY OF COAL TAILINGS DAMS
A Dissertation in
Civil Engineering
by
Sajjad Salam
2020 Sajjad Salam
Submitted in Partial Fulfillment
of the Requirements
for the Degree of
Doctor of Philosophy
August 2020
ii
The dissertation of Sajjad Salam was reviewed and approved by the following:
Ming Xiao
Associate Professor of Civil Engineering
Dissertation Advisor
Chair of Committee
Patrick J. Fox
John A. and Harriette K. Shaw Professor
Department Head of the Department of Civil and Environmental Engineering
Tong Qiu
Associate Professor of Civil Engineering
Shimin Liu
Associate Professor of Energy and Mineral Engineering
Murali Haran
Professor
Department Head of the Department of Statistics
iii
ABSTRACT
Coal has been one of the main sources of energy in the world. The coal waste that
is produced through coal extraction and processing is typically stored in form of slurry in
tailings impoundments. The static and dynamic stability of these impoundments are of
great importance, as their failure can result in significant spill, loss of human lives, and
damages to the environment and infrastructure. The main objectives of this research were
to (1) characterize the static and dynamic geotechnical properties of fine coal refuse (FCR),
(2) investigate the cyclic behavior and liquefaction resistance of FCR and influencing
factors such as strain history and aging, and (3) numerically assess the seismic stability of
coal tailings dams incorporating the heterogeneity on FCR deposit in the field.
To characterize the static and dynamic geotechnical properties of in situ FCR
samples, representative FCR samples were taken from two coal slurry impoundments in
the Appalachian coalfields in the USA. Standard penetration tests (SPT) were conducted
in the field. Index properties, hydraulic conductivity, and classification of FCR were
determined. Staged triaxial tests under consolidated undrained (CU) state and consolidated
drained (CD) state were conducted to assess short-term and long-term shear behavior of
FCR, respectively. Torsional resonant column tests were performed to determine shear
stiffness properties of FCR. Cyclic direct simple shear (DSS) tests followed by static
shearing were adopted to evaluate the cyclic and post-cyclic behavior of FCR under various
cyclic stress ratios (CSR).
To overcome the shortcomings of element testing methods, large-scale shake table
testing was conducted. Furthermore, the effects of strain history and short-period aging on
iv
cyclic response and liquefaction resistance of FCR were investigated. The FCR specimen
was slurry-deposited in a membrane-lined laminar shear box (L×W×H: 2.29 m × 2.13 m ×
1.4 m). The FCR specimen was subjected to three shaking events. Instruments including
piezometers and linear variable differential transformers (LVDTs) were used to measure
the FCR’s dynamic response during shaking. A piezocone penetrometer (CPTu) was used
to measure soil resistance and estimate cyclic behavior of the FCR specimen before and
after each shaking test for time intervals up to 97 days. The cyclic behavior, liquefaction
resistance, aging rate, and strength gain within the FCR were studied and compared with
those of clean sands.
Dynamic loadings such as earthquakes and blasting are among the main threats to
the stability of tailings dams. Seismic stability analyses of tailings dams are further
challenged by the uncertainty and variability of tailings properties. The influence of input
motion characteristics and spatial variability in coal tailings (CT) properties on the seismic
stability of a typical upstream-construction CT dam was investigated. First, the
applicability of two advanced constitutive plasticity models, PM4Sand and PM4Silt, in
simulating the cyclic behavior of CT was evaluated and a suitable model was selected. The
undrained shear strength of CT was modeled as a spatially correlated Gaussian random
field. Six input motions representing a variety of peak ground accelerations (PGA),
equivalent number of cycles (ENC), and frequency content were selected for the dynamic
analyses. The seismic stability of the CT dam with uniform properties (i.e. uniform models)
was compared to the stochastic models. The uncertainty in seismic response of the studied
dam caused by spatial variability in geotechnical properties was investigated. The necessity
v
of post-seismic stability analysis in CT dams was discussed. The influencing factors on the
seismic stability of CT dams were also characterized.
vi
TABLE OF CONTENTS
LIST OF FIGURES ..................................................................................................... viii
LIST OF TABLES ....................................................................................................... xi
ACKNOWLEDGEMENTS ......................................................................................... xii
Chapter 1 Introduction ................................................................................................ 1
1.1 Problem statement .......................................................................................... 1
1.2 Research motivation ....................................................................................... 4
1.3 Research objectives ........................................................................................ 5
1.4 Organization of the dissertation ...................................................................... 5
Chapter 2 Literature Review ....................................................................................... 7
2.1 Geotechnical properties of coal tailings ......................................................... 7
2.2 Physical modeling of tailings .......................................................................... 12
2.2.1 Shake table testing ................................................................................ 12
2.2.2 Strain history effect of cyclic response of soils and tailings ................ 14
2.2.3 Aging effect on cyclic response of soils and tailings ........................... 16
2.3 Numerical Modeling Approaches ............................................................ 18
Chapter 3 Characterization of Static and Dynamic Geotechnical Properties and
Behaviors of Fine Coal Refuse ............................................................................. 23
3.1 Field Sampling and Laboratory Testing ......................................................... 27
3.2 Index Properties of the Samples ..................................................................... 31
3.3 Static Triaxial Test Results Analysis .............................................................. 35
3.4 Resonant Column Test Results and Analysis ................................................. 42
3.5 Liquefaction Susceptibility and Cyclic Behavior Characterization ................ 45
3.6 Conclusion and Summary ............................................................................... 58
Chapter 4 Strain History and Short-Period Aging Effects on the Strength and
Cyclic Response of Fine-Grained Coal Refuse .................................................... 60
4.1 Testing Method ............................................................................................... 65
4.1.1 Shake table system and deposition process .......................................... 65
4.1.2 CPTu device and testing locations ....................................................... 68
4.1.3 Shake table test plan ............................................................................. 70
4.2 Results and Discussion ................................................................................... 71
4.2.1 Pre-shake CPTu .................................................................................... 71
4.2.2 Shake table test results .......................................................................... 75
4.2.3 Effect of strain history .......................................................................... 86
vii
4.2.4 Effect of short-period aging .................................................................. 93
4.3 Summary and Conclusions ............................................................................. 98
Chapter 5 Seismic Stability of Coal Tailings Dams with Spatially Variable and
Liquefiable Coal Tailings using Pore Pressure Plasticity Models ........................ 101
5.1 Model Configuration ...................................................................................... 104
5.1.1 PM4Sand and PM4Silt calibration based on CT cyclic response ........ 106
5.1.2 Random fields generation for CT ......................................................... 112
5.1.3 Input motions and analysis approach ................................................... 117
5.2 Model Results and Discussion ........................................................................ 122
5.2.1 Representative dynamic responses ....................................................... 122
5.2.2 Dynamic responses of uniform models ................................................ 124
5.2.3 Post-seismic analysis significance ........................................................ 127
5.2.4 Dynamic response of stochastic models (co-seismic) .......................... 128
5.2.5 Dynamic response of stochastic models (post-seismic) ....................... 132
5.2.6 Failure probability analysis .................................................................. 135
5.2.7 Implications in practice ........................................................................ 139
5.3 Conclusions and summary .............................................................................. 140
Chapter 6 Summary and Conclusions .......................................................................... 142
6.1 Summary ......................................................................................................... 142
6.2 Conclusions..................................................................................................... 144
6.3 Limitations of this research ............................................................................ 146
6.4 Recommendations for future work ................................................................. 147
REFERENCES ............................................................................................................ 148
viii
LIST OF FIGURES
Figure 3-1 Areal view of the impoundments ............................................................... 27
Figure 3-2 The particle size distributions of the FCR samples .................................... 34
Figure 3-3 Staged CU triaxial test results (S1B2-U) ................................................... 37
Figure 3-4 Staged CU triaxial test results (S1B1-D) ................................................... 38
Figure 3-5 Staged CD triaxial tests results (S2B1-U and S2B1-D) ............................. 40
Figure 3-6 Normalized shear modulus of FCR ............................................................ 43
Figure 3-7 Damping ratio of the FCR samples and their initial index properties ........ 45
Figure 3-8 Liquefaction susceptibility assessment criteria proposed by a) Seed et
al. (2003) b) Bray and Sancio (2006) ................................................................... 46
Figure 3-9 Relationship of cyclic stress ratio (CSR) with number of cycles (N) to
reach 5% double-amplitude strain ........................................................................ 47
Figure 3-10 Cyclic DSS test results at CSR~0.15 ....................................................... 49
Figure 3-11 Cyclic DSS test results at CSR~0.12 ....................................................... 50
Figure 3-12 Cyclic DSS test results at CSR~0.1 ......................................................... 52
Figure 3-13 Cyclic behavior of the FCR based on Idriss and Boulanger (2008)
criterion ................................................................................................................. 55
Figure 3-14 Post-liquefaction shear strength characteristics of FCR .......................... 57
Figure 4-1 Liquefaction definitions for various scenarios (Seed 1979, Robertson
and Wride 1998, Youd and Idriss 1998) ............................................................... 61
Figure 4-2 Laminar shear box, specimen preparation, and instrumentation ................ 65
Figure 4-3 FCR gradations and Atterberg limits for 8 random FCR samples ............. 68
Figure 4-4 (a) FCR specimen plan view showing piezometer and CPTu test
locations; and (b) photograph of CPT testing process .......................................... 69
Figure 4-5 Acceleration-time input motions for the testing program .......................... 70
Figure 4-6 Pre-shake CPTu results .............................................................................. 72
ix
Figure 4-7 CPTu results plotted on soil behavior charts: (a) Robertson and Wride
(1998), (b) Robertson (2016) and (c) Robertson (2009) ....................................... 74
Figure 4-8 Pore pressures during and after the first shake (height of water above
each piezometer in parentheses) ........................................................................... 77
Figure 4-9 Developed shear strains within the FCR specimen during the first shake
.............................................................................................................................. 79
Figure 4-10 CPTu test results before and after the first shake ..................................... 81
Figure 4-11 Developed shear strains within the FCR specimen during the second
shake ..................................................................................................................... 83
Figure 4-12 CPTu results before and after the first shake and up to 97 days after the
second shake ......................................................................................................... 85
Figure 4-13 Developed shear strains within the FCR specimen during the third
shake ..................................................................................................................... 86
Figure 4-14 Maximum lateral displacement of the FCR specimen during the (a)
first shake (b) second shake (c) third shake (the horizontal line represents the
FCR surface) ......................................................................................................... 89
Figure 4-15 FCR specimen classification over the test plan........................................ 91
Figure 4-16 FCR specimen liquefaction behavior over the test plan........................... 92
Figure 4-17 Strength gain trend over time for clean sand and FCR ............................ 95
Figure 4-18 CD values for FCR in the shake table test ................................................ 97
Figure 5-1 Typical upstream-construction CT dam model generated in FLAC2D ..... 106
Figure 5-2 Experimental and numerically simulated CSR-N curves for the studied
CT ......................................................................................................................... 109
Figure 5-3 Cyclic responses of CT from cyclic DSS test and simulations at
CSR=0.12.............................................................................................................. 111
Figure 5-4 CSR versus number of cycles to reach 5% shear strain for CT (PM4Silt
simulations) ........................................................................................................... 114
Figure 5-5 su,cs,eq_Rat variation in Realizations A, B, C, and D ..................................... 117
Figure 5-6 Histograms of su,cs,eq_Rat of CT for Realizations A, B, C, and D ................ 117
x
Figure 5-7 Selected input motions for CT dam seismic stability analyses .................. 119
Figure 5-8 ENC and maximum CSR of the input motions and the tested CT CSR-
N curve .................................................................................................................. 120
Figure 5-9 Crest acceleration response spectra for Realizations A, B, C, and D and
the uniform model with su,cs,eq_Rat = 0.2 ................................................................ 122
Figure 5-10 Co-seismic performance of the CT dam in terms of excess pore
pressure and shear strain contours in a uniform model with su,cs,eq_Rat = 0.2 (The
unit of excess pore pressure is Pa) ........................................................................ 123
Figure 5-11 Co-seismic performance of the CT dam in terms of excess pore
pressure and shear strain contours in a stochastic model with su,cs,eq_Rat ranging
from 0.1 to 0.5 (The unit of excess pore pressure is Pa) ...................................... 124
Figure 5-12 Co-seismic and post-seismic crest settlements of the uniform models
(a) PGA effect (b) ENC effect .............................................................................. 126
Figure 5-13 Co-seismic and post-seismic performances of select models under EQ2
.............................................................................................................................. 128
Figure 5-14 Summary of co-seismic crest settlement for stochastic models (a) PGA
effect (b) ENC and frequency content effect ........................................................ 131
Figure 5-15 Summary of post-seismic crest settlement for stochastic models (a)
PGA effect (b) ENC and frequency content effect ............................................... 134
Figure 5-16 Probabilistic co-seismic performance of the CT dam under the
earthquake input motions ...................................................................................... 137
xi
LIST OF TABLES
Table 3-1 Locations and labels of the samples ............................................................ 28
Table 3-2 Index properties and hydraulic conductivity of the FCR samples .............. 31
Table 3-3 Staged triaxial test results ............................................................................ 41
xii
ACKNOWLEDGEMENTS
I am writing this acknowledgement in the midst of the hardship the pandemic (Covid-19)
has caused for all people on earth. It has been indeed a lesson for us to be even more grateful
than before. I want to first thank my adviser, Prof. Ming Xiao, for giving me the opportunity
in the first place to start my PhD study at Penn State. He continued his technical, financial,
and morale support through my research studies and helped me to have significant
accomplishments.
I would like to appreciate my committee members, Prof. Patrick Fox, Prof. Tong Qiu, Prof.
Shimin Liu, Prof. Murali Haran for their constructive comments, criticism, and guidance.
Without the constructive feedback and insights provided by my committee members this
research would not have been possible.
I also want to extend my appreciation to my collaborators at different stages of my research
project. Prof. Khosravifar from Portland State University shared so many insightful
comments with me and provided invaluable experimental and numerical data necessary for
my research. Prof. Jeff Evans provided me with CPTu device and priceless suggestions
through data interpretation. Prof. Katerina Ziotopoulou advised me through numerical
simulations and reviewed my work, which resulted in significant improvement of my work.
I am deeply grateful for the assistance provided by Mr. Dan Fura during my experimental
tests setup. I also want to acknowledge the help from my fellow graduate students Min
Liew, Dr. Jintai Wang, Dr. Pezhouhan Tavassoti-Kheiry, Mehrzad Rahimi, Maximilian
xiii
Ororbia, Khashayar Jafari, Jack Wang, and Rong Zhao, who helped me during my research
studies at Penn State.
I am grateful to have my companion and best friend, Alexis Schad, by me during this
journey. Her unconditional love and support was an endless source of motivation and
energy throughout my PhD study. I would also like to express my gratitude toward my
parents and my lovely sister for their faith in me, and being supportive. Their constant
source of encouragement and kindness made this accomplishment possible.
Sajjad Salam
06/2020
Chapter 1
Introduction
1.1 Problem statement
The National Inventory Dams (2005) reported the existence of 1448 tailings dams
in the United States and over 3500 worldwide (Davis et al. 2002). According to a study by
the International Commission of Large Dams (ICOLD 2001), 1 to 2 major tailings
impoundment failure occur per year. These failures have huge impact on infrastructures,
environment, economy, and human lives. Liquefaction has been found one of the common
causes of tailings dams’ failure.
The Cadia tailings dam failure in 2018 occurred in New South Wales State in
Australia. The trigger of the incident was the liquefaction of a highly loose and
compressible layer within the foundation. The failure released 1.33 million m3 of tailings
to the downstream dam; fortunately, there was no further destruction or casualty. The
Kingston Fossil Plant spill in 2008 in Kingstone Steam Plant, Tennessee, United States,
was initiated by liquefaction of a loose layer under the dikes. Consequently, 4.2 million m3
of tailings were released, which damaged over 180 properties and destroyed utility and
power lines in the area. Furthermore, the cleanup cost approximately $1.2 billion.
Brumadinho tailings dam failure in 2019 in Brazil occurred due to poor internal drainage
system. Accordingly, a heavy rain triggered the liquefaction of tailings and approximately
12 million m3 of tailings were released. This catastrophe resulted in over 300 life losses,
2
significant environmental loss, and $19 billion drop in the stock value of the owner of the
facility.
Past incidents indicated that the consequences of tailings dams’ failure could be
extremely significant and in some cases unaffordable (Rico et al. 2008). Therefore, extreme
measures should be practiced to ensure safety and sustainability of these facilities.
According to a report by the International Commission on Large Dams (ICOLD 2001),
there were over 700 coal tailings dams in the United States, amongst which 241 dams were
classified as high hazard facility. The high hazard facilities require special attention and
maintenance as their failure leads to loss of human life and infrastructures. In addition, 70
to 90 million tons of coal tailings are annually deposited in tailings dams (National
Research Council 2002), this results in growing the size of dams, subsequently, higher risk
and vulnerability to failure.
Geotechnical properties of coal tailings should be determined to evaluate the
stability of coal tailings dams under various loading scenarios. Coal tailings are typically
mixed with water to form a slurry, and then hydraulically deposited behind tailings dams.
Coal tailings consolidate under self-weight in field, therefore, they are found under- to
normally-consolidated with a loose structure. Accordingly, due to safety and accessibility
issues, it is little known about the geotechnical properties of coal tailings in the
impoundments.
The in-situ geotechnical properties are necessary to accurately assess the stability
of coal tailings dams. Although element laboratory testing may determine the geotechnical
properties of coal tailings, the fabric and structure of coal tailings in the field may not be
fully represented in laboratory. Furthermore, the significant heterogeneity and interlayered
3
medium of coal tailings challenge the reliability and accuracy of these methods. The
heterogeneity is due to the depositional process in the field, and it may have significant
effects on cyclic and post-cyclic behavior of material. For example, Kingston Fossil Plant
coal fly ash slurry spill in Tennessee in 2008 was reported to be due to the liquefaction of
a loose layer under the dikes (Plant and Harriman 2008). Therefore, large scale testing on
reconstituted samples, which have similar fabric, structure, and heterogeneity to coal
tailings in the field may better represent the behavior of coal tailings.
The sustainability of coal tailings dams should be evaluated over the operational
period, during which the size of dam may change, several seismic events or cyclic loadings
may take place, consolidation and aging is in process. These changes may affect the
stability of the facility. Therefore, the influence of factors such as strain history and aging
on cyclic behavior and liquefaction resistance of heterogeneous coal tailings need to be
assessed.
Numerical modeling in the field of liquefaction and cyclic loading has significantly
advanced in recent years and can be used in seismic stability evaluation of geosystems such
as dams (Boushehri et al. 2020). Several constitutive plasticity models that are commonly
used to simulate the cyclic and liquefaction behavior of soils are PDMY02 model (Yang et
al. 2008), UBCSAND model (Beaty and Byrne 2011), PM4Sand model (Boulanger and
Ziotopoulou 2017), and PM4Silt (Boulanger and Ziotopoulou 2018). These models have
been successfully adopted and calibrated for geotechnical earthquake engineering
applications such as dams, levees, foundations, etc. However, the suitability of these model
to approximate the cyclic behavior of coal tailings, which are different from natural soils
in terms of composition, is not adequately assessed. The abovementioned models are suited
4
for either clean sands or silts. The applicability of these models to estimate cyclic responses
of coal tailings that consist of water, coal fraction, silt, and sand is unclear and needs further
investigations.
The simulations for dams’ stability evaluation are usually conducted assuming
uniform properties for different sections. This approach may not represent the true
performance of geo-systems and may produce misleading results. The necessity of
considering heterogeneity and stochastic modeling is even more distinguished based on the
failure reason in the past tailings dams’ incidents such as the Kingstone Fossil Plant in
2008. A uniform model may not capture a critical failure mode. Mathematical and
statistical tools have provided the possibility of incorporating the heterogeneity and
stochastic modeling for tailings dams’ seismic stability evaluation.
1.2 Research motivation
First motivation was to comprehensively characterize the static and dynamic
geotechnical properties of fine coal refuse (FCR) by running in-situ tests and laboratory
tests on undisturbed samples. Second motivation was to address the limitations engaged
with element testing by using a 1-g shake table facility. Therefore, the cyclic response of a
heterogeneous FCR specimen could be investigated. Furthermore, the effects of strain
history and aging on cyclic response of a heterogeneous FCR specimen are quantified and
compared with those on clean sands. Third motivation was to assess the applicability of
novel plasticity models in approximating the cyclic behavior and liquefaction resistance of
FCR. Subsequently, the suitable plasticity models are used to evaluate seismic stability of
5
coal tailings dams. It was of interest to study the necessity of stochastic modeling and
considering the heterogeneity of FCR in seismic stability evaluation of coal tailings dams.
1.3 Research objectives
This study has three main objectives. The first objective is to provide
comprehensive geotechnical properties of FCR using laboratory tests. The second
objective is to investigate the cyclic characteristics and liquefaction resistance of
heterogeneous FCR specimen by a shaking program using 1-g shake table test. The strength
gain due to multiple shaking events and aging are assessed using CPTu tests on the FCR
specimen. The third objective is to investigate the necessity of stochastic modeling in
seismic stability evaluation under cyclic loading.
1.4 Organization of the dissertation
This dissertation consists of six chapters including this chapter. Chapter 2 includes the
literature review of tailings and FCR geotechnical properties, physical modeling of tailings and
similar soils, aging effect, strain history effect, and numerical modelling techniques for
liquefaction predictions and seismic stability of dams. Chapter 3 presents the results of
laboratory tests on the FCR samples taken from an Appalachian coalfields. This chapter
was earlier published as a technical paper in Canadian Geotechnical Journal, please see
Salam et al. (2019). Chapter 4 presents the results of the shake table tests on a slurry-
deposited FCR specimen along with CPTu test results. The content of this chapter was
6
accepted for publication in the ASCE Journal of Geotechnical and Geoenvironmental
Engineering. Chapter 5 presents the development of a numerical model, plasticity model
calibration, and stochastic modeling of the seismic stability of a coal tailings dam. This
chapter will soon be submitted to the Computers and Geotechnics for review and
publication. Lastly, Chapter 6 presents a summary of findings and conclusions derived from
this research followed by recommendations for future research.
Chapter 2
Literature Review
This chapter presents the literature review on (1) the static and dynamic
geotechnical properties of fine coal refuse (FCR) (2) large-scale testing to assess the
geotechnical properties of soils and tailings and also the effects of aging and strain history
on their cyclic response (3) numerical modeling techniques and plasticity models.
Additional relevant background studies are presented in Chapters 3, 4, and 5.
2.1 Geotechnical properties of coal tailings
The basic and advanced geotechnical properties of FCR are of great importance to
better analyze the mechanical response of FCR under various loading scenarios. Many
laboratory studies have been conducted on disturbed and relatively undisturbed mine
tailings samples including FCR. Although in-situ subsurface exploration methods are not
typically versatile and practical in the FCR impoundments due to the loose structure of
FCR, several studies have attempted to carry out in-situ testing and sampling with light-
weight equipment.
For instance, Busch et al. (1975) retrieved relatively undisturbed samples from a
coal tailings impoundment by light drilling equipment. Then, laboratory tests including
hydraulic conductivity, sieve analysis, compaction, Atterberg limits, specific gravity, and
direct shear were conducted. The shear strength of coal tailings was found low. The void
8
ratio of coal tailings was reported to be high, furthermore, the loose and saturated structure
of coal tailings were expected to be highly liquefaction susceptible.
Leventhal and Ambrosis (1985) conducted laboratory tests such as sieve analysis,
X-ray diffraction, shear strength, and consolidation test on reconstituted coal refuse
samples from the Sydney Basin in Australia. The gradation and classification of coal refuse
samples varied in different sampling location. Strain hardening behavior was seen for the
coal refuse samples during undrained triaxial tests. Furthermore, the obtained data were
compiled with the previous collected data over 10 years. The laboratory results were found
well-compared with the coal refuse behavior observed in the field. An equation to estimate
permeability of coal refuse based on void ratio and gradation was also proposed.
Qiu and Sego (2001) studied the geotechnical properties of four different mine
tailings including gold, coal, copper, and oil sand composite. The basic geotechnical
properties of the mine tailings along with consolidation characteristics, hydraulic
conductivity, water retention characteristics, and shear strength properties were studied. A
linear relationship between void ratio and logarithm of saturated hydraulic conductivity of
the mine tailings was noticed. Copper and gold tailings showed strain softening behavior,
while strain hardening behavior was observed for coal tailings and oil sand composite. The
pore pressure build up during loading was reported significant and necessary to be
considered in tailings impoundments design.
A comprehensive laboratory characterization of coal refuse focusing on flow
behavior under static and dynamic loading was carried out by Yu (2015). Representative
coal refuse samples were taken from different coalfields and characterized for gradation,
specific gravity, Atterberg limits, permeability, consolidation and shear behavior. The
9
studied coal refuse samples were classified as low-plasticity silt with specific gravity
ranged from 2 to 2.15. High compressibility and extremely low shear strength were
observed for the coal refuse samples. The samples with higher initial water content showed
higher compressibility. Viscosity and flow behavior of coal refuse was also found
significantly sensitive to water content.
Among physical and geotechnical properties of FCR, cyclic behavior and
liquefaction resistance of FCR have not been sufficiently investigated. To date, liquefaction
has been mostly studied for clean sands and low plasticity silts and clays. However, FCR
is identified as a “transitional soil”, as it is composed of sand and silt with low to no
plasticity, and its behavior may vary between sand and silt. For example, in a study by
Polito and Martin (2001), 25% to 45% non-plastic silt content, which is the typical fraction
in coal tailings, was reported as limiting silt content. Further, it was concluded that the
liquefaction resistance of soils with limiting silt content cannot be adequately predicted by
relative density and applicability of the current empirical methods such as penetration test
is uncertain, as the behavior is not dominated by neither silt content nor sand content.
The dynamic properties and liquefaction resistance of mine tailings such as FCR
has been the main concern of researchers lately, as dynamic loadings can trigger tailings
liquefaction, subsequently, failure of dams. For example, although FCR consists of
appreciable amount of fines content, it is not considered as liquefaction resistant material.
FCR has been found significantly contractive and liquefaction susceptible due to its loose
and saturated structure. In addition, high water content and low hydraulic conductivity
associated with FCR facilitate the liquefaction occurrence and generation of excess pore
pressure under static and dynamic loading (Zeng et al. 2008).
10
To evaluate the liquefaction potential, there are empirical and experimental
approaches. In the empirical methods, the index properties such as Atterberg limits and
moisture content are used to determine the liquefaction susceptibility of soil (Seed et al.
2003; Bray and Sancio 2006; Idriss and Boulanger 2008). The empirical criteria to assess
the liquefaction susceptibility of soils including Andrews and Martin, Seed et al. (2003),
and Bray et al. (2004) were adopted by Salehian (2013) to evaluate the liquefaction
susceptibility of coal tailings. The seed et al. (2003) criterion showed agreement with the
laboratory tests results. However, assessing the liquefaction potential solely based on index
properties was questioned by Ajmera et al. (2015), as the composition and mineralogy of
the material were found to be influencing factors.
Laboratory testing such as cyclic triaxial test and cyclic direct simple shear (DSS)
test and in-situ testing such as standard penetration test (SPT) and cone penetration test
(CPT) are common methods in evaluating the liquefaction resistance of soils. The
applicability of the common in-situ subsurface exploration methods, which are originally
developed for natural soils such as silts and sands, for characterization of tailings is not
well-proved yet. However, these methods have been used in several studies on coal tailings
(Kalinski and Philips 2008; Kalinski and Salehian 2016; Robertson et al. 2017).
Few studies have been conducted to characterize the FCR specifically under cyclic
loading (e.g., Ishihara et al. 1981; Zeng et al. 1998a; Zeng et al. 1998b; Castro 2003; Zeng
et al. 2008; James et al. 2011; Salehian 2013; Geremew and Yanful 2013). Zeng et al.
(2008) studied cyclic behavior of coarse and fine coal refuse by conducting resonant
column tests and cyclic traixial tests. The coal tailings were fist classified as a composition
of sand, silt and clay with low plasticity. Shear modulus and damping ratio of the coal
11
tailings at different shear strain levels were determined. The modulus reduction curve of
fine coal refuse conformed to the empirical equation proposed by Seed et al. (1984).
Comprehensive laboratory testing was highly encouraged as physical and mechanical
properties of coal refuse are scattered, depending on sampling location.
In an experimental study by Hu et al. (2016), the static and dynamic behavior of
copper and iron mine tailings were investigated. The fine tailings showed higher
compressibility and lower permeability compared to those of coarse tailings. Linear
relationships between void ratio and hydraulic conductivity and coefficient of
compressibility were observed. The empirical equation for excess pore pressure generation
during cyclic loading proposed by Seed et al. (1975) and Zhang et al. (2006) was found
applicable for fine mine tailings. The necessity of numerical modeling and centrifuge
testing for stability analysis was also emphasized.
Cyclic triaxial tests were conducted and the cyclic behavior of coal tailings were
reported to be sand-like, clay-like, and transition type based on the criterion developed by
Boulanger and Idriss (2004). In-situ tests such as SPT and CPT were conducted in field,
and the coal tailings were mostly described as fine-grained soils, with higher strength close
to the coal tailings discharge point.
Robertson et al. (2017) compiled the case history field data on mine tailings. The
laboratory test results such as shear wave velocity were compared against the in-situ tests
such as SPTu and CPTu. Effect of matric suction on stiffness properties of unsaturated
mine tailings was also investigated. Shear wave velocity of the studied mine tailings were
found to be quite sensitive to slight matric suction change. However, this feature was not
12
pronounced in cone penetration tests results, and the discrepancy was attributed to different
induced shear strain range in different tests.
Kalinski and Philips (2008) and Kalinski and Salehian (2016) attempted to establish
a correlation between cyclic and post-cyclic resistance and in-situ tests indices such as CPT
cone resistance. SPT and CPT tests were conducted in two coalfields in eastern Kentucky.
The use of CPT test over SPT test to characterize coal refuse was highly recommended as
SPT could not clearly delineate the strength properties variation for such a soft material.
2.2 Physical modeling of tailings
2.2.1 Shake table testing
Large scale physical modeling such as shake table testing and centrifuge modeling
have been commonly practiced to investigate the liquefaction behavior and seismic
response of soils deposit (Sharp et al. 2003; Okamura and Teraoka 2006; Haeri et al. 2012;
Otsubo et al. 2016, Prabhakaran et al. 2020). Liquefaction and lateral spreading mechanism
were investigated by Thevanayagam et al. (2009) using 1-g shake table testing method.
The limitations associated with shake table physical modeling technique as well as
reliability of the data acquired by the embedded instruments were discussed. Ottawa sand
was deposited in the box following a novel hydraulic filling approach, resembling alluvial
deposition mechanism. The viability of the 1-g shake table test to study liquefaction and
lateral spreading behavior was approved, as the shake table system performed well and the
instrumentation results were consistent.
13
The liquefaction behavior of mine tailings by physical modeling has been studied
in only a few studies. Benzaazoua et al. (2004) used physical modeling technique to study
the variation of index properties such as water content, matric suction, and chemical
composition through profile of surface paste disposal, which was a mixture of gold mine
tailings and cement. The surface paste disposal was deposited in a small box and the
behavior of the mixture in terms of binder leaching and volumetric water content was
monitored over time.
Pepin et al. (2012) conducted several seismic table tests on mine tailings with and
without drainage inclusion to determine the improving effects in terms of pore pressure
build up and liquefaction potential. Mine tailings were found substantially liquefaction
susceptible, furthermore, excess pore pressure magnitude was reported to be related to
initial relative density. Pepin et al. (2012) observed that drainage inclusion greatly
decreased the excess pore pressure generation during cyclic loading, and it also increased
the rate of pore pressure dissipation after cyclic loading.
Antanoki et al. (2018) carried out several centrifuge tests on mine tailings and
commixing of mine tailings and waste rock. Consolidation behavior, seismic response, and
liquefaction susceptibility of the different configurations of mine tailings mixtures were
determined. Increasing rock content was found a viable practice to reduce consolidation
time and settlement. A threshold for mine tailings to rock ratio was noticed, at which the
improving effect of rack content was pronounced.
There are advantages and disadvantages associated with both shake table testing
and centrifuge modeling. For example, although shake table testing is limited to only
shallow soil deposit testing, dense instrumentation can be practiced, which is not possible
14
in centrifuge modeling. Physical modeling such as shake table testing could be a great and
novel approach to observe seismic response of hydraulically deposited coal tailings, as the
field condition is better represented. The physical modeling also provides the possibility to
further investigate other factors such as the influence of multi-shake and strain history on
liquefaction behavior of mine tailings.
2.2.2 Strain history effect of cyclic response of soils and tailings
The effects of strain history on seismic response and liquefaction resistance of soils
have been studied by several researchers using conventional and advanced laboratory
testing methods. Finn et al. (1970) observed that small and large shear strains have opposite
effects on liquefaction resistance of sands. Liquefaction resistance decreased when large
shear strains were developed, while liquefaction resistance increased after inducing small
shear strains. Creating of non-uniform structure due to large shear strain development was
claimed to be the underlying reason for the observations. Oda et al. (2001) provided a
microstructural explanation for the effect of strain history on liquefaction resistance by
running cyclic triaxial tests on granular soils. Generation of a column like structure in the
soil after shaking was found to be the reason of reduced liquefaction resistance.
Ha et al. (2011) conducted 1-g shake table tests on sands with different gradations.
Each specimen was shaken several times to investigate the effect of shake history on the
liquefaction resistance. It was observed that void ratio, relative density, and gradation
characteristics are not correlated with liquefaction resistance. All the specimens showed
lower liquefaction resistance after the first shake, which eliminated the aging effects. The
15
soil deposit after the first shake was perceived as a normally consolidated fresh deposit. In
another study, Heidary and Andrus (2012) stated that complete liquefaction tends to fully
eliminate the aging effects, and the aged soil deposit turns into a fresh deposit. However,
shake events do not always have negative effect on liquefaction resistance, and the intensity
of shake events dominates the subsequent effects on liquefaction resistance.
In a case study by Dobry et al. (2015), field measurements on liquefaction
resistance of sands and silty sands in Imperial Valley of South California showed that the
recent seismic activities in the area had increased the liquefaction resistance of the
investigated soils. This observation was earlier noticed by El-Sekelly (2014), where a sand
deposit in centrifuge system posed higher liquefaction resistance after several shake events.
The previous observations were approved by El-sekelly et al. (2016a, 2016b), who
conducted centrifuge tests on loose saturated silty sand deposit. Two shaking events were
considered in the test plan: strong shake consisting of 15 cycles and weak shake consisting
of 5 cycles. The weak shaking events did not liquefy the soil deposit, even though excess
pore pressure was generated. It was observed that weak shakes increased the liquefaction
resistance of the soil. The strong shake almost “reset the clock” by destroying the
improving effects of weak shakes. It was concluded that extensive liquefaction may result
in substantial decrease in liquefaction resistance such that all beneficial effects of shaking
history and aging is vanished. The aforementioned observations were not well pronounced
in shear wave velocity measurements from bender elements embedded in the centrifuge
container.
Wang et al. (2020) presented the results of multiple shakes using a 1-g shake table
on a clean sand deposit. Although the relative density of the deposit increased through the
16
shaking events, the deposit was found liquefiable or experiencing significant excess pore
pressure generation under strong shakes. Despite the immediate drop in cone resistance of
the deposit, the densification cause by the shakes increased the cone resistance over time.
Significant post-shake pore pressure increase was also observed due the disturbance during
the shake. A relationship between magnitude of the post-shake excess pore pressure and
relative density and pore pressure at the end of shaking was noticed.
2.2.3 Aging effect on cyclic response of soils and tailings
The improving effect of aging on freshly deposited, densified, and recently
disturbed soils have been investigated in several studies. Anderson and Stokoe (1978) and
Kim and Novak (1981) conducted resonant column tests on sands and cohesive soils,
respectively. Shear modulus showed increasing trend at constant confining stress over time.
This behavior was attributed to primary consolidation and creep movement of particles
during secondary consolidation over time. In general, lower increase in shear modulus was
noticed for sands compared to cohesive soils. It was concluded, in both studies, that aging
effect is of great importance and should be considered in design and analysis.
Mitchel and Solymer (1984) stated that clean sands, fresh deposit or densified, may
keep gaining strength over a period of several months. Therefore, caution should be
exercised while analyzing reconstituted samples laboratory results. Inter-particle
cementation caused by dissolution and precipitation of silica was claimed to be the main
reason for strength and stiffness increase over time. Schmertmann (1991) also noticed
significant improvement in strength and stiffness properties of soils by aging.
17
Schmertmann (1991) suggested that increase in basic soil friction and dilatancy effects are
the major reasons for the aging improvement. He also observed a linear relationship
between improving properties and logarithm of time of aging. Consideration of aging effect
was highly encouraged in design applications in both studies.
Mesri et al. (1990) conducted cone penetration tests on sands over time and
observed significant increase in tip resistance at constant effective stress. This phenomenon
was attributed to the continuous creep movement of particles during secondary
consolidation, which results in higher stiffness. Micro interlocking of sand grains and
surface roughness is also greatly improved due to this process. A cone resistant prediction
model to estimate tip resistance over time after ground disturbance was proposed.
CPT testing method was also adopted by Charlie et al. (1992) to assess aging effects
on tip resistance of soil in a blasting site. The CPTs were conducted before and after the
blast events up to 18 weeks. The tip resistance decreased by 62%, while friction ratio
increased by 100%, after the blast. The tip resistance increased by 18% after 18 weeks.
Jorshi et al. (1995) conducted CPT tests to study the aging effect on freshly deposited sands
in dry and submerged conditions. Considerable increase in tip resistance was reported due
to aging. Higher strength gain rate was also noticed for the submerged sand compared to
dry sand.
The necessity of considering time effects in liquefaction triggering analyses was
further discussed by Wang et al. (2019). Cone penetration tests were conducted before and
after shaking a relatively loose dopiest of clean sand up to 4.5 months. Although, the cone
resistance decreased immediately after the liquefaction of the deposit, over 100% increase
18
was observed in the cone resistance after 4.5 months of aging. Therefore, the strength of
the deposit not only was recovered, but also reached beyond the pre-liquefaction level.
2.3 Numerical Modeling Approaches
The constitutive models are adopted to determine response and stability of the dams
under different loading scenarios, among which seismic or cyclic loading is the main
concern. Cyclic loadings such as earthquakes and blasting can result in liquefaction of
FCR. Therefore, plasticity models that are able to capture cyclic response and liquefaction
behavior of materials should be employed to conduct numerical stability analysis on the
tailings impoundments.
Of the constitutive models advanced to assess the mechanical response of soils
subject to dynamic loading, PDMY02 (Elgamal et al. 2002; Yang et al. 2008), UBCSAND
(Beaty and Byrne 1998), Daflias-Manzari (Dafalias and Manzari 2004), PM4Sand
(Boulanger and Ziotopoulou (2013), and PM4Silt (Boulanger and Ziotopoulou (2018) are
the common ones. Among the mentioned constitutive models, UBCSAND, PM4Sand, and
PM4Silt can be implemented in FLAC commercial software.
Pressure Dependent Multi Yield 02 (PDMY02) elastio-plastic model approximates
the cyclic response of granular materials. PDMY02 model was employed by Karimi and
Dashti (2016) to approximate the performance of a shallow foundation laid on a liquefiable
layer of sand in centrifuge model. The model estimated adequate results compared to the
laboratory results in case of low intensity seismic input motion. However, the PDMY02
19
could not accurately predict the excess pore pressure in case of strong seismic input motion,
as extreme densification and dilation were occurred during cycles.
UBCSAND is a non-linear effective stress plasticity model, which was proposed
by Beaty and Byrne (1998) to determine mechanical response of sand under cyclic loading.
UBCSAND is able to conduct fully coupled analysis including flow calculations.
UBCSAND has been successfully used to simulate dynamic behavior of sands and low
plasticity tailings in engineering and laboratory practices. Seid-karbasi and Byrne (2004)
studied the failure of Mochikochi tailings dams in Japan by UBCSAND model. The results
were in well agreement with the observed deformations after the dam’s failure. Castillo et
al. (2006) investigated the seismic response of a heap leach pad with high phreatic line
using UBCSAND model in a fully coupled analysis. Effect of drainage system to reduce
the liquefaction and failure potential was also discussed. UBCSAND model was adopted
by James (2009) and Ferdosi et al. (2015a) to evaluate the reinforcing effect of waste rock
inclusion on stability of liquefiable mine tailings impoundments.
Byrne et al. (2004) compared the results of Nevada sand centrifuge test with
simulated Nevada sand using UBCSAND. The liquefaction resistance of Nevada sand was
first characterized by cyclic DSS test and then the UBCSAND model was calibrated
according to cyclic DSS test results. UBCSAND model approximated close excess pore
pressure to the observed values in the centrifuge test. Ferdosi et al. (2015b) simulated the
seismic and post-seismic mechanical response of mine tailings tested in a rigid shake table.
The rate of excess pore pressure generation was close to the laboratory observations at
shallow depth, while UBCSAND estimated higher rate at greater depths. The post
20
liquefaction settlement of the mine tailings could be properly predicted by the modified
UBCSAND, which was enabled to update the elastic modulus after liquefaction.
The Dafalias-Manzari model is a stress state, elasto-plastic model based on critical
state and stress-ratio controlled framework (Dafalias and Manzari 2004). Predictive
capabilities of PDMY02 model and modified Manzari-Dafalias model to estimate the
seismic response of a layered soil in centrifuge tests were compared by Ramirez et al.
(2017). Cyclic triaxial test results were used to calibrate both models to simulate the layered
soil behavior during cyclic loading. The modified Manzari-Dafalias model performed
better in estimating the pore pressure ratio and volumetric strain developed during the
centrifuge tests. However, both models failed to adequately estimate damping, as it was
overestimated and underestimated by PDMY02 and modified Manzari-Dafalias model,
respectively.
PM4Sand and PM4Silt models are plasticity models developed by Boulanger and
Ziotopoulou (2013, 2018). PM4Sand assesses the undrained cyclic and monotonic
mechanical response of sand and non-plasticity silt, while PM4Silt assesses those of non-
plasticity to low plasticity silts and clays. Both PM4Sand and PM4Silt plasticity models
are based on the framework of the stress-ratio controlled, critical state compatible,
bounding-surface plasticity model for sand developed by Dafalias and Manzari (2004). The
behavior of soil is predicted based on three key surfaces: the bounding, dilation, and critical
state surfaces. The location of bounding and dilation surfaces is determined based on
relative state parameter index. The relative state parameter index is a function of current
relative density and the relative density at critical state under the same effective stress.
PM4Sand and PM4Silt models have less required input parameters compared to
21
UBCSAND. The last version of PM4Sand (Version 3.1) was released in 2017, while the
previous limitations were resolved (Boulanger and Ziotopoulou 2017). The PM4Silt
(Version 1) was also released in 2018 by Boulanger and Ziotopoulou (2018).
PM4Sand has been implemented in analysis of different geotechnical applications
such as dams, embankments, and foundations. Ziotopoulou and Montgomery (2017)
studied the post-liquefaction settlement prediction capability of PM4Sand for a shallow
foundation on a liquefiable soil. The simulation results were compared with the centrifuge
tests conducted by Dashti et al. (2010). Sufficient agreement between laboratory and
simulation results were observed. Furthermore, the effect of several factors including soil
layer thickness and foundation dimensions on the reconsolidation strains after liquefaction
was parametrically investigated.
In a study by Boulanger et al. (2014), accuracy of PM4sand to predict the
mechanical response of a layered soil deposit in centrifuge tests was assessed. It was found
that the void redistribution has considerable effects on the response of soil in the laboratory
tests. The simulations showed similar behavior in terms of slope deformations. However,
it was concluded that the current practices still pose high level of uncertainty in predicting
post-liquefaction strength and deformations.
In one hand, the constitutive models are only able to capture the mechanical
response when structure and loading state along with soil-specific parameters are known.
On the other hand, the discrepancies between physical modeling and actual soil deposit
specifically FCR deposits in field may lead to inaccurate estimations. The post-seismic
behavior of the impoundments are also of great importance. For example, the FCR peak
strength may deteriorate to a fraction of its previous peak strength or residual strength due
22
to excessive strain experienced during earthquake (Castro 2003). The post-liquefaction
strength of soil was observed to be strain dependent (Sivathayalan and Vaid 2004;
Wijewickreme et al. 2005). Castro and Troncoso (1989) studied the residual strength and
post-liquefaction strength of fine refuse by performing in-situ vane shear tests, indicating
considerable drop in strength of fine refuse after liquefaction. None of the above-mentioned
characteristics can be properly represented by small scale laboratory testing or numerical
modeling. Therefore, these issues need to be addressed by some more advanced and unique
testing methods.
Chapter 3
Characterization of Static and Dynamic Geotechnical Properties and
Behaviors of Fine Coal Refuse
Coal refuse is the waste product of coal processing and mining activities. Coal
refuse is different from fly ash, which is the residue of coal combustion. Depending on the
milling process and particles size distribution, coal refuse can be classified as either coarse
coal refuse (CCR) or fine coal refuse (FCR) (Zamiran et al. 2015), the latter of which is
typically hydraulically deposited in the form of slurry behind tailings impoundments
constructed by the former. Based on the National Inventory of Dams report, there are 1172
tailings dams in the U.S., and they are mostly classified as high hazard facilities (CEER
1985). FCR is commonly loose, saturated, and under- to normally consolidated in the field
(Ishihara et al. 1981; Vick 1990). Therefore, FCR has low strength and stiffness, resulting
in stability issues specifically under dynamic loading. Earthquake-induced cyclic loading
can cause significant reduction in stiffness and strength of contractive soils such as FCR.
Accordingly, one of the predominant causes of failure of FCR impoundments is
earthquake, which can result in liquefaction (Martin and Davis 2000; Rico et al. 2008).
Although FCR consists of appreciable amount of fines content, it is not considered as
liquefaction resistant material. FCR has been found significantly contractive and
liquefaction susceptible due to its loose and saturated structure. In addition, high water
content and low hydraulic conductivity associated with FCR facilitate the liquefaction
occurrence and generation of excess pore pressure under static and dynamic loading (Zeng
24
et al. 2008). The two most known tailings impoundment failures due to liquefaction are the
1965 El Cobre Dam failure in Chile and the 1978 Mochikoshi impoundment failure in
Japan (Dobry and Alvarez 1967; Ishihara 1984). The recent failure of the Kingston
Tennessee Valley Authority coal ash impoundment in 2008 was also claimed to be partially
due to the liquefaction of the coal ash slurry (Plant and Harriman 2008) that was caused by
rapid static loading on the slurry. The rapid static loading consisted of 10 days of heavy
rain before the day of failure and construction of retaining walls on top of the
impoundment, which contributed to rapid undrained loading and liquefaction of loose fly
ash layer under the dikes. It is worth mentioning that several other factors such as poor
construction and maintenance were also suspected to have contributed to the failure.
The high scatter in physical and geotechnical properties of FCR has been observed
in the past studies (Qiu and Sego 2001; Hegazy et al. 2004). FCR may show varying
characteristics depending on its sampling location, as the FCR near the discharge point
consists of larger particles, while the FCR becomes finer at farther distance from the
discharge point. Evaluating the strength and stiffness properties of FCR by in-situ testing
or using representative samples has been highly recommended, as these characteristics are
substantially affected by void ratio, degree of saturation, and density (Castro 2003). Slurry
deposition method has been developed and found to be a suitable approach to prepare
samples resembling the fabric and structure of hydraulically deposited soils such as FCR
when undisturbed samples are not available (Kuerbis and Vaid 1988).
Among physical and geotechnical properties of FCR, cyclic behavior and
liquefaction resistance of FCR have not been sufficiently investigated. To date, liquefaction
has been mostly studied for clean sands and low plasticity silts and clays. However, FCR
25
is identified as a “transitional soil”, as it is composed of sand and silt with low to no
plasticity, and its behavior may vary between sand and silt. For example, in a study by
Polito and Martin (2001), 25% to 45% non-plastic silt content, which is the typical fraction
in coal tailings, was reported as limiting silt content. Further, it was concluded that the
liquefaction resistance of soils with limiting silt content cannot be adequately predicted by
relative density, and applicability of the current empirical methods such as penetration test
is uncertain, as the behavior is not dominated by neither silt content nor sand content. To
evaluate the liquefaction potential, there are empirical and experimental approaches. In the
empirical methods, the index properties such as Atterberg limits and moisture content are
used to determine the liquefaction susceptibility of soil (Seed et al. 2003; Bray and Sancio
2006; Idriss and Boulanger 2008). However, assessing the liquefaction potential solely
based on index properties was questioned by Ajmera et al. (2015), as the composition and
mineralogy of the material were found to be influencing factors. Laboratory testing such
as cyclic triaxial test and cyclic direct simple shear (DSS) test and in-situ testing such as
standard penetration test (SPT) and cone penetration test (CPT) are the common methods
in evaluating the liquefaction resistance of soils. The applicability of the common in-situ
subsurface exploration methods, which are originally developed and calibrated for natural
soils such as silts and sands, for characterization of tailings is not well-proved yet.
However, these methods have been used in several studies on coal tailings (Kalinski and
Philips 2008; Kalinski and Salehian 2016).
Few studies have been conducted to characterize the FCR under cyclic loading
(e.g., Ishihara et al. 1981; Zeng et al. 1998a; Zeng et al. 1998b; Castro 2003; Zeng et al.
2008; James et al. 2011; Salehian 2013; Geremew and Yanful 2013). Although cyclic
26
triaxial test has been adopted in several studies to assess the dynamic properties of FCR
(Thacker et al. 1988; Ullrich et al. 1991), cyclic DSS test better simulates the mode of
loading during earthquake. Post-liquefaction strength of FCR is also important for stability
consideration. The FCR peak strength may deteriorate to a fraction of its previous peak
strength or residual strength due to excessive strain experienced during earthquake (Castro
2003). The post-liquefaction strength of soil was observed to be strain dependent
(Sivathayalan and Vaid 2004; Wijewickreme et al. 2005). Castro and Troncoso (1989)
studied the residual strength and post-liquefaction strength of fine refuse by performing in-
situ vane shear tests, indicating considerable drop in strength of fine refuse after
liquefaction. Caution should be exercised when the post-liquefaction characteristics of
material are evaluated by laboratory testing, as void redistribution and water film effect
after liquefaction are not perfectly represented (Kokusho 2003).
The main goal of this study was to further investigate the mechanical behavior of
FCR, as there are not many studies focusing on FCR behavior, which may significantly
vary from other types of tailings. Therefore, this chapter aimed to first comprehensively
characterize the physical and hydraulic properties of FCR using representative samples
from different locations and depths in two Appalachian coalfields. Second, geomechanical
behavior of the representative samples, including shear strength and stiffness properties,
were determined. The cyclic behavior and liquefaction resistance of the slurry-deposited
FCR samples, which sufficiently resembled the fabric and structure of in-situ FCR (Kuerbis
and Vaid 1988; Carraro and Prezzi 2007), were assessed by cyclic DSS tests. Third, the
cyclic behavior of the FCR was further evaluated by empirical approaches and in-situ data.
Accordingly, the applicability and limitations of the common empirical and experimental
27
methods, which are originally proposed for natural soils, for assessing the liquefaction
behavior of FCR were investigated. Lastly, the effect of liquefaction on static shear strength
of the liquefied FCR samples was determined by conducting monotonic shear loading at
the end of the cyclic DSS tests.
3.1 Field Sampling and Laboratory Testing
The sampling was conducted in two different FCR impoundments, which are
labeled as S1 and S2. SPTs were conducted at each impoundment at various depths. The
aerial view of the impoundments and the locations of the boreholes (denoted as B1 and B2)
and SPTs are shown in Figure 3-1. There are two boreholes in S1 impoundment and one
borehole in S2 impoundment. The field sampling and SPT testing were led by Dr. Ming
Xiao, Dr. Shimin Liu, and Min Liew.
Figure 3-1 Areal view of the impoundments
28
A light-weight, track-mounted drill rig (model: Geoprobe 6620DT) was used to
retrieve continuous FCR samples. This drill rig can descend down the steep slopes of the
slurry impoundment and travel on the deposited soft coal refuse, while conventional drill
rigs cannot access to such cite condition. The drill rig uses percussion technique at
percussion rate of 32 Hz to continuously push a split-spoon sampler (model: DT325) into
the subsoil. The diameter and length of each coal slurry sample were 47 mm and 1.5 m,
respectively. The sample disturbance, which depends on sampler dimensions, sampler
driving mechanism, and soil types and consistency, was not assessed. SPTs were conducted
using the same drill rig with DH-100 automatic drop hammer. The corrected SPT numbers
(i.e. (N1)60) in the S1 impoundment was 6.9. The (N1)60 in the S2 impoundment was 3.7.
The low (N1)60 can be attributed to the looseness of the FCR, which is typically under- or
normally consolidated. The SPT numbers were found within the typical SPT ranges
observed for very loose to loose granular soils (Teng 1962). Table 3-1 shows the depths at
which the specimens were retrieved for laboratory testing. A unique label shown on the
last column of Table 3-1 was assigned to each sample for simplicity.
Table 0-1 Locations and labels of the samples
Impoundment Borehole Depth (m) Label
S1
S1
S1
S1
S2
S2
B1
B1
B2
B2
B1
B1
4.5-6
10.5-12
4.5-6
10.5-12
4.5-6
7.5-9
S1B1-U
S1B1-D
S1B2-U
S1B2-D
S2B1-U
S2B1-D
29
Representative samples were collected and used for static triaxial test, resonant
column test, and density measurement. The static triaxial and resonant column tests were
conducted at several confining pressures. The hydraulic conductivity of all the specimens
were also measured using flexible-wall permeameter method. The hydraulic conductivity
of the samples was measured under 34.5 kPa consolidation stress. Although the 34.5 kPa
confining pressure may not accurately represent the consolidation stress of the samples in
the field, the resulting hydraulic conductivities are relevant references, which can help
compare samples in terms of hydraulic conductivity at the same consolidation stress. The
diameter of all the samples used for static triaxial and resonant column tests was 35.5 mm.
The retrieved samples were extruded and trimmed axially and longitudinally by a sharp
thin-bladed ring and trimming knife, respectively, with great caution to avoid disturbing
the samples. The requirement of height to diameter ratio of 2:1 was met for all the samples.
The index properties such as grain size distribution and Atterberg limits were determined
using the samples after they were used for static triaxial tests. The index properties were
verified by duplicate tests. The specific gravity was determined by two approaches: 1)
ASTM D854 “Standard Test Method for Specific Gravity of Soil Solids by Pycnometer”,
2) Micromeritics Accupyc II 1340 Gas Displacement Pycnometry System with a chamber
volume of 1.0 cm3 at room temperature (about 23.5°C). In the second approach, after filling
up the chamber with coal tailings sample, helium gas was released and allowed to displace
the sample pores; the absolute density of the specimen was then calculated using the
volume that was not displaced by helium; ten measurements were made and the average
absolute density was computed. The average values of the specific gravity are reported in
this chapter, as the measurements varied marginally.
30
The extruded representative samples could not be used for cyclic DSS tests because
the diameter of representative samples was smaller than the diameter of the cyclic DSS
mold. Of the sample preparation methods, wet pluviation and slurry deposition methods
were effective in mimicking the fabric of the in-situ hydraulically deposited materials such
as FCR (Carraro and Prezzi 2007). However, as wet pluviation method is not reliable for
silty sands, due to the possibility of particle segregation, slurry deposition method
developed by Kuerbis and Vaid (1988) was adopted to prepare the sample. A series of
cyclic DSS tests were conducted on reconstituted S1B2-D samples, with void ratio in the
range of 0.6~0.7, until repeatable and consistent results in terms of cyclic resistant ratio
were observed. The target void ratio of 0.6-0.7 was chosen as a matter of consistency since
it could be repeatedly achieved during sample preparation, even though the target void ratio
was slightly lower than the average void ratio of the representative samples (~0.9) obtained
in the field. The liquefaction resistance of FCR was obtained at different cyclic stress ratio
(CSR). The cyclic DSS device is made by GeoComp and applies cyclic shear loading under
the constant volume condition. Therefore, the pore-water pressure is back-calculated from
the change in vertical total stress. The soil samples are confined by Teflon rings lined with
latex membrane to ensure uniform shearing of the sample. The horizontal stress is unknown
throughout testing. The post-liquefaction strength characteristics of FCR were also studied
by statically shearing the liquefied samples. The reason to choose S1B2-D sample for the
cyclic DSS and post-liquefaction tests is that the corresponding location is close to the
potential failure plane, therefore, the geomechanical properties of this sample was more of
interest than other locations.
31
The dimensions of the sample in cyclic DSS tests were 63.5 mm in diameter and
12.7 mm in height. The consolidation pressure was 60 kPa, the frequency of cyclic loading
was 0.1 Hz, and the liquefaction criterion was set as 5% double amplitude shear strain
(DAS). The cyclic loading was stress-controlled, while the static loading was strain-
controlled with shear strain rate of 1.4% per hour.
3.2 Index Properties of the Samples
The groundwater depth in field was measured by lowering a measuring tape into
the borehole. The depths of groundwater table that were measured at the time of the field
sampling was 1.1 m at Site 1 and 4.9 m at Site 2. The fine coal refuse samples that were
retrieved from drilling and tested in this study were all below groundwater table. Table 3-
2 presents the basic index properties, hydraulic conductivity, and classification of the
samples that were used for the static triaxial tests.
Table 0-2 Index properties and hydraulic conductivity of the FCR samples
Sample
Moisture
Content
(%)
Density
(KN/m3) eini
Saturation
Degree
(%)
LL
(%)
PL
(%) PI Gs
k
(cm/s)
USCS
Classification
S1B1-U
S1B1-D
S1B2-U
S1B2-D
S2B1-U
S2B1-D
35
49
35
36
25
48
15.6 0.8 92 33 29 4
2.1
3.3e-6
5.1e-7
8.8e-7
1.3e-4
3.6e-7
1.0e-6
SM
12.9 1.4 73 34 30 4 ML
14.6 0.9 82 31 27 4 ML
16.1 0.8 94 27 25 2 SM
15.4 0.7 75 20 19 1 2.2
SM
15.6 1.0 100 35 33 2 ML
The moisture content of the samples retrieved from the greater depth (i.e. 10.5 m
from impoundment S1 and 7.5 m from impoundment S2) were higher as expected due to
32
the less exposure to evaporation and desiccation. The saturation degree of all the samples
except S1B1-D and S2B1-U are noticeably high and above 80%, which can be attributed
to the deposition method (i.e. slurry) of FCR in field. The Atterberg limits tests were
conducted several times until repetitive and consistent results for each sample were
achieved. The reported moisture contents are the average values with standard deviation of
3%.
Unit weight of the samples was determined using the representative samples before
they were mounted on triaxial base plate. The highest variation in unit weight was observed
in Borehole 1 in S1 impoundment. The unit weight of S1B1-U and S1B1-D were 15.6
kN/m3 and 12.9 kN/m3, respectively. The unit weight of the samples taken from Borehole
2 in S1 impoundment was 14.6 kN/m3 and 16.1 kN/m3 for S1B2-U and S1B2-D,
respectively. The sample S1B2-D showed the highest unit weight among the studied
samples. The location of S1B2-D is close to the discharge point where the coal slurry is
hydraulically deposited. Typically, the larger and heavier grains settle first near the
discharge point while the slurry with fines remains on top. The least variation in unit weight
was in the samples collected from the S2 impoundment, which was geographically located
in the middle of the impoundment. The unit weights of 15.4 kN/m3 and 15.6 kN/m3 were
measured for S2B1-U and S2B1-D, respectively. Initial void ratio of the FCR samples
ranged from 0.7 to 1.4 with an average of 0.9. All the samples can be considered as non-
plastic to low plasticity, as the plasticity indices were less than 5. The moisture content of
all the samples was higher than their liquid limit, which is representative of soils greatly
prone to liquefaction (Byrne and Seid-Karbasi 2003). Specific gravity (Gs) shown in Table
3-2 is the average values, with standard deviation of 0.03. The specific gravity of the FCR
33
samples was lower than typical values reported for fine-grained soils due to high carbon
content (Hegazy et al. 2004). The values for specific gravity obtained in this study were in
agreement with other studies (e.g., Huang et al. 1987; Ullrich et al. 1991; Cowherd and
Corda 1998).
Particle size distribution of the samples used for triaxial tests was determined by
conducting sieve analysis and hydrometer analysis, as per ASTM C136 and ASTM D422,
respectively. The classification of each sample was determined based on the Unified Soil
Classification System (USCS), and the particle size distributions of the samples are
presented in Figure 3-2. As shown in Table 3-2, the FCR samples were all classified as
either silty sand (SM) or sandy silt (ML). Silt and sand content of the FCR samples ranged
from 15% to 52% and 41% to 82%, respectively. The wide range associated with sand and
silt content further emphasizes the scattered physical properties of the FCR in the field.
The FCR samples approximately showed similar silt content to limiting silt content (i.e.
25% to 45%) defined by Polito and Martin (2001). Accordingly, the liquefaction resistance
of FCR samples may not follow typical behavior observed for most sandy soils.
34
Figure 3-2 The particle size distributions of the FCR samples
A narrow range of hydraulic conductivity from 1.0e-6 cm/s to 3.6e-7 cm/s was
observed for all the samples excluding S1B2-D. The higher hydraulic conductivity of
S1B2-D can be again attributed to the accumulation of coarser particles at this depth.
According to the empirical equation for calculating hydraulic conductivity (Taylor 1948),
the highest hydraulic conductivity was expected to be associated with S2B1-U, which has
the largest particles and D50. However, the hydraulic conductivity is highly sensitive to
homogeneity and voids arrangement inside the sample’s structure (Budhu 2015).
Therefore, the observed discrepancy can be attributed to heterogeneity inside the samples’
mass, which is common in tailings. It is also worth mentioning that the hydraulic
35
conductivities presented in Table 3-2 are the hydraulic conductivity in vertical
(gravitational) direction.
3.3 Static Triaxial Test Results Analysis
Static triaxial tests were conducted on representative samples to determine the shear
strength properties of the specimens under monotonic loading. Staged triaxial approach on
a single specimen was practiced in this study, as sufficient number of representative and
identical samples were not available. All the samples from S1 and S2 impoundments were
tested under consolidated-undrained (CU) condition and consolidated-drained (CD)
condition, respectively. It was of interest to evaluate shear strength properties of FCR in
both short-term (i.e., undrained) and long-term (i.e. drained) conditions. The index
properties of the tested samples, including initial void ratio and initial unit weight, were
earlier reported in Table 3-2.
Samples were initially saturated using the back pressure technique, as per ASTM
D 4767, until a minimum B-value of 96% was reached. Each staged triaxial test consisted
of two stages. Samples were first consolidated and then axially loaded under confining
pressure of 34.5 kPa in Stage 1. It is noteworthy to mention that the vertical compression
in the first stage (i.e., at 34.5 kPa confining pressure) was halted before failure. The first
loading stage was continued until the threshold of the maximum deviatoric stress, where
the change in deviatoric stress became minimal. This practice prevented complete failure
or disturbance of the sample. Then, the axial load was removed and the sample was
consolidated under 69 kPa confining pressure and vertically compressed again to reach the
36
maximum deviatoric stress. All the samples showed strain hardening behavior during
loading so that deviatoric stress kept slightly increasing at large strains. This behavior was
also observed in other studies on coal refuse material (Qiu and Sego 2001). The typical
behavior of coal slurry samples observed in CU and CD tests in terms of deviatoric stress
versus axial strain, excess pore pressure versus axial strain, stress path in q-p space, and
shear strength envelope are presented for four samples. Figure 3-3 and Figure 3-4 depict
the staged CU triaxial test results of S1B2-U and S1B1-D, respectively. Figure 3-5 shows
staged CD triaxial results of S2B1-U and S2B1-D. Therefore, drained and undrained
mechanical behavior of FCR at shallow and deep depth under shear could be investigated.
37
Figure 3-3 Staged CU triaxial test results (S1B2-U)
Figure 3-3 presents the mechanical response of S1B2-U under staged CU triaxial
static loading. Figure 3-3 (a) shows strain hardening behavior of the sample and reinforcing
effect of confining pressure, as higher confining pressure resulted in higher deviatoric
stress. The maximum deviatoric stress reached approximately 120 kPa and 165 kPa at 6%
and 11% axial strain under 34.5 kPa and 69 kPa confining pressure, respectively. Similarly,
higher pore pressure was developed within the sample at higher confining pressure, as
shown in Figure 3-3 (b). Peak pore pressure of 15 kPa and 33.5 kPa was observed below
4% axial strain under 34.5 kPa and 69 kPa confining pressure, respectively; then, pore
pressure began decreasing. The stress path was plotted in q-p space per Lambe’s (1964)
definition, as in Figure 3-3 (c). The slope of the shear envelope in q-p space is equal to
tan𝛼 = sinϕ′, while the intercept is equal to 𝑚 = 𝑐′cosϕ′. The stress path in the first stage
relatively resembled the typical stress path seen for over-consolidated soils, as the sample
was consolidated under higher effective stress than 34.5 kPa in field. However, the stress
path in the second stage showed the typical path seen for normally consolidated soils, as
38
69 kPa was higher than the consolidation stress in field for S1B2-U. The effective Mohr’s
circles along with shear envelope were also plotted, as in Figure 3-3 (d). The effective shear
strength properties, c’ and ϕ’, are later presented and discussed in Table 3-3.
Figure 3-4 Staged CU triaxial test results (S1B1-D)
The mechanical response of S1B1-D under staged CU triaxial static loading is
presented in Figure 3-4. The strain hardening behavior of the sample S1B1-D was less
39
intense than that of S1B2-U, see Figure 3-4 (a). The reason could be attributed to the larger
amount of small particles in S1B2-U that led to higher compressibility of the sample. The
maximum deviatoric stresses observed for S1B1-D were 120 kPa and 160 kPa both
occurred at approximately 11% axial strain under 34.5 kPa and 69 kPa confining pressure,
respectively. The peak pore pressures of 15 kPa and 30 kPa were observed under 34.5 kPa
and 69 kPa confining pressure, respectively. The stress path observed for S1B1-D was
similar to the stress path expected for over-consolidated soils, as the sample had been
consolidated by higher effective stress according to the depth of the sample in field, see
Figure 3-4 (c). Effective Mohr’s circles are also displayed in Figure 3-4 (d).
40
Figure 3-5 Staged CD triaxial tests results (S2B1-U and S2B1-D)
The mechanical response of S2B1-U and S2B1-D under staged CD triaxial static
loading is shown in Figure 3-5. According to Figure 3-5 (a) and (d), the higher confining
pressure was, the higher maximum deviatoric stress was achieved. Consolidated drained
condition led to higher maximum deviatoric stress compared to the results observed in
consolidated undrained tests. For example, the maximum deviatoric stress achieved by
S2B1-U and S2B1-D were approximately 290 kPa and 230 kPa, respectively, under 69 kPa
confining pressure. The stress paths shown in Figure 3-5 (b) and (e) represented the drained
41
path, which is a straight line. Lastly, Figure 3-5 (c) and (f) show the shear envelopes for
S2B1-U and S2B1-D, which show higher internal friction angle and lower cohesion
compared with those under CU condition.
Table 3-3 presents the effective shear strength properties of all the FCR specimens.
The cohesion for all the specimens tested under CU condition falls within 13.8 kPa to 25.5
kPa, while the cohesion of the samples tested under CD condition ranges from 13.1 kPa to
16.5 kPa. In terms of internal friction angle, two samples from impoundment S2, tested
under CD condition, showed higher values compared to CU test results. Furthermore,
higher friction angle (i.e. 38°) was observed for S2B1-D compared to that of S2B1-U (i.e.
36°). Among the results obtained from impoundment S1, S1B2-D showed the highest
internal friction angle (i.e. 31°), which can be attributed to the higher concentration of
coarse particles at this location. The slope (α) and intercept (𝑚) of the failure envelope in
q-p space are also provided in Table 3-3. The observations in this study were found within
the range reported by Hegazy et al. (2004), who conducted statistical analysis on shear
strength properties of coal refuse that were determined using laboratory and in-situ testing.
Considering all the results, the FCR’s shear strength properties are relatively scattered and
dependent on its location and depth.
Table 3-3 Staged triaxial test results
Test Sample 𝑐′(𝑘𝑃𝑎) ∅′(𝑑𝑒𝑔. ) 𝑚 (𝑘𝑃𝑎) 𝛼 (𝑑𝑒𝑔. )
CU S1B1-U 13.8 29 12 25.9
CU S1B1-D 25.5 26 22.9 23.7
CU S1B2-U 24.8 30 21.5 26.6
CU S1B2-D 20.7 31 17.7 27.2
CD S2B1-U 16.5 36 13.3 30.4
CD S2B1-D 13.1 38 10.3 31.6
42
3.4 Resonant Column Test Results and Analysis
The shear moduli of the FCR samples were determined by resonant column tests
on representative samples. The torsional resonant column approach is commonly used to
characterize the maximum shear modulus of soil at low shear strain. The shear strain
applied to samples in the torsional resonant column ranges from 10-5 % to 10-2 %. The
sample’s behavior is considered elastic within this low range of shear strain. Each sample
was tested under three confining pressures of 34.5, 69, and 103 kPa. The key outputs of the
resonant column tests are shear wave velocity, shear modulus, and damping ratio.
Figure 3-6 presents the normalized shear modulus, defined as the shear modulus
divided by the corresponding maximum shear modulus, at 69 kPa confining stress. The
hollow and solid markers represent the samples at deeper and shallower depth,
respectively. Normalized shear modulus showed minimal change over the shear strain
range, therefore, the results at 34.5 kPa and 103 kPa were not shown in order to avoid
overlap of the data points. The observations are in agreement with previous studies (e.g.
Seed and Idriss 1970; EPRI 1993; Darendeli 2001), which showed negligible influence of
confining pressure on normalized shear modulus at low shear strain level that was less than
10-3 %, see Figure 3-6. The modulus reduction curve could not be established for the FCR,
as the shear modulus was only examined for low shear strain that was less than 10-3 %. The
effect of aging and time were not investigated in this study. However, the observed shear
stiffness properties such as shear modulus might vary over time as shown by Kim and
Novak (1963) and Anderson and Stokoe (1978). The normalized shear modulus obtained
in cyclic DSS tests are also embedded in Figure 3-6 and will be discussed in the following
43
section. Initial void ratio and unit weight of the samples used in the resonant column tests
are presented in Figure 3-7 (d).
Figure 3-6 Normalized shear modulus of FCR
The samples collected from greater depth, S1B1-D, S1B2-D, and S2B1-D, shown
by the hollow markers in Figure 3-6, showed higher absolute shear modulus and shear
wave velocity than the samples at shallower depth. However, higher reduction was
observed for S1B1-D and S1B2-D samples, as shear strain increased, this behavior was
found out of the proposed limits shown in Figure 3-6. The sample S1B1-D showed
significantly greater shear modulus and shear wave velocity (i.e. 90 to 100 m/s) than those
of other samples. The sample S1B1-D seemed to have coarser particle size distribution
44
than other samples, resulting in stiffer material. However, the S1B1-D gradation shown in
Figure 3-2 is not the largest. The reason is that the samples used for resonant column and
particle size distribution analysis were within the same depth range (i.e. 10.5 m to 12 m),
but were not the same sample. This discrepancy emphasizes the scattered physical
properties of FCR in the field even in small ranges of depth and distance. However, the
higher stiffness observed for deeper samples was consistent with higher shear strength
observed for the samples taken from deeper depths in triaxial testing. Except for S1B1-D,
other examined samples demonstrated close values in terms of shear modulus in the range
of 4.1 to 6.9 MPa.
The damping ratios of all the samples under three different confining pressures are
displayed in Figure 3-7. The effect of confining pressure was found minimal due to the
large amount of fines content and low induced shear strain. The damping ratio also
increased by increasing shear strain regardless of applied confining pressure. Overall,
damping ratio of the FCR samples was found to be within the range of 0.6% to 2%, which
is in agreement with other studies on FCR (e.g., Zeng et al. 2008).
45
Figure 3-7 Damping ratio of the FCR samples and their initial index properties
3.5 Liquefaction Susceptibility and Cyclic Behavior Characterization
The liquefaction potential of FCR is of great importance, as liquefaction occurrence
can result in significant loss in strength and stability of coal slurry impoundments.
Accordingly, liquefaction susceptibility of FCR should be assessed under seismic loading
conditions. The liquefaction potential per soil type can be evaluated by index properties
such as Atterberg limits and water content. Seed et al. (2003), Bray and Sancio (2006), and
46
Idriss and Boulanger (2008) have presented empirical criteria to determine the liquefaction
potential of soils based on index properties, including liquid limit (LL), plasticity index
(PI), and water content (w). The index properties of the FCR samples, shown in Table 3-2,
were plotted in the recommended figures by Seed et al. (2003) and Bray and Sancio (2006),
see Figure 3-8. As shown in Figure 3-8, all the samples fall within the area marked as
potentially liquefiable by both criteria. Liquefaction potential assessment per soil type
using the approach by Idriss and Boulanger (2008) is described later in detail.
(a) (b)
Figure 3-8 Liquefaction susceptibility assessment criteria proposed by a) Seed et al.
(2003) b) Bray and Sancio (2006)
Cyclic DSS tests were conducted on reconstituted S1B2-D sample at different
CSRs to determine the liquefaction resistance of FCR and assess the undrained cyclic
behavior of FCR. Duplicate tests were conducted until validity of the results was ensured.
The void ratios of the tested samples, which were prepared by slurry deposition approach,
were approximately 0.6~0.7 after consolidation and before cyclic loading.
47
Figure 3-9 shows the cyclic resistance of FCR prepared by the slurry deposition
method and consolidated under a vertical stress of 60 kPa with void ratio after
consolidation ranging approximately from 0.6 to 0.7. Higher number of cycles were
required to liquefy the sample as the CSR decreased. The relation between CSR and
number of cycles (N) to failure (defined in this study as 5% DAS) can be expressed as
𝐶𝑆𝑅 = 𝑎 × (𝑁5%𝐷𝐴𝑆)−𝑏. The relation and corresponding fitted line are presented in Figure
3-9. The power fit (b-value) was found to be 0.16.
Figure 3-9 Relationship of cyclic stress ratio (CSR) with number of cycles (N) to reach
5% double-amplitude strain
Figure 3-10 to 3-12 present the results obtained in cyclic DSS test on the FCR
samples at CSR of 0.15, 0.12, and 0.1, respectively. The initial void ratio of the samples
before cyclic loading is also shown in the figures. The subfigures (a) and (b) summarize
the undrained cyclic response of the FCR samples, and the subfigure (c) clearly shows the
48
development of shear strain against number of cycles; the developed double amplitude
strain (DAS) can be calculated by summing the positive and negative peak shear strain at
each cycle. The subfigure (d) displays the shear modulus reduction during the cyclic
loading. The shear modulus, calculated based on the dissipated energy during each cycle,
is directly calculated and reported by the device. According to Figure 3-10, the FCR sample
reached 5% DAS in almost two cycles when cyclically loaded by CSR of 0.15. The void
ratio of the sample before and after consolidation was 0.94 and 0.73, respectively. The pore
pressure ratio, which is traditionally considered as a parameter of evaluating liquefaction
occurrence, also increased to 0.55 in approximately two cycles. The shear modulus in the
first two cycles were plotted in Figure 3-10 (d) to show the decreasing trend of shear
modulus.
49
Figure 3-10 Cyclic DSS test results at CSR~0.15
Figure 3-11 shows the results of cyclic DSS test at CSR of 0.12. The void ratio of
the sample before and after consolidation was 1.02 and 0.69, respectively. The sample
reached 5% DAS after 8 cycles. According to the shear stress-strain loops, the sample
behavior is relatively plastic, as large strain is developed rapidly in the first cycle. The
shear stress-strain loops are slightly shifted to the left direction. However, the 5% DAS
failure criterion was assumed to properly eliminate any potential dependence of the cyclic
resistance to the directionality and bias in the shear stress-strain loops (Price et al. 2017).
The pore pressure ratio (ru) increased to 0.7 after 8 cycles.
50
Figure 3-11 Cyclic DSS test results at CSR~0.12
Similar behavior was observed for the FCR sample under CSR of 0.1, as the shear
stress-strain loops were wide and shifted, and considerable amount of pore pressure was
developed in the first few cycles, as shown in Figure 3-12. The void ratio of the sample
before and after consolidation was 0.78 and 0.6, respectively. The 5% DAS was reached
after approximately 23 cycles of cyclic loading, and the final pore pressure ratio was equal
to 0.7. As shown in Figure 3-12 (d), significant shear modulus reduction occurred in the
last few cycles. The specimen generated a considerable amount of pore water pressure in
51
the first cycle, as shown in Figure 3-10 (a), Figure 3-11 (a), and Figure 3-12 (a), while the
rate of pore water pressure generation reduced in the following cycles. The axial strain
during the cyclic loading was smaller than 0.05%, which ensured that the device was able
to maintain the constant volume during cyclic loading. As far as the shear modulus obtained
from cyclic DSS testing, the average shear modulus calculated at the beginning of the
cyclic DSS tests was approximately 1.3 MPa. The results are embedded in Figure 3-6,
which shows the obtained shear modulus is within the proposed limits for sands (Seed and
Idriss 1970; EPRI 1993) and low plasticity silty sands (Darendeli 2001). In comparison
with the range of Gmax observed in resonant column tests (4.1 MPa – 6.9 MPa), lower shear
modulus was seen for FCR at higher shear strain.
52
Figure 3-12 Cyclic DSS test results at CSR~0.1
The wide shear stress-strain loops and large shear strain development without
reaching 100% pore pressure ratio (ru) is commonly observed in clay-like material. The b
value (i.e. the power fit on the CSR-N plot in Figure 3-9) is also within the range of clay-
like material (Idriss and Boulanger, 2008). However, the cyclic response was expected to
be sand-like because of the extremely low plasticity index of 2 associated with S1B2-D,
see Table 3-2. The uncertainties in characterizing the cyclic response of FCR compelled
the authors to try to assess this characteristic using empirical criteria.
In order to further investigate the cyclic behavior of the FCR sample, an empirical
criterion proposed by Idriss and Boulanger (2008) was adopted, as shown in Figure 3-13.
The transition of cyclic behavior from sand-like to clay-like is shown against plasticity
index. The hatched region is the transitional area where the cyclic behavior is between
sand-like and clay-like behavior. Furthermore, the solid lines are the conservative limits
proposed by Idriss and Boulanger (2008). The cyclic resistance ratios (CRRs) of material
assuming clay-like and sand-like behavior can be determined by the equations proposed by
53
Idriss and Boulanger (2008). Equations 1 and 2 determine the sand-like and clay-like CRR
of soils, respectively.
𝐶𝑅𝑅𝑀=7.5,𝜎′𝑣𝑐=1 = exp ((𝑁1)60𝑐𝑠
14.1+ (
(𝑁1)60𝑐𝑠
126)
2
− ((𝑁1)60𝑐𝑠
23.6)
3
+ ((𝑁1)60𝑐𝑠
25.4)
4
− 2.8)
Equation 3-1
𝐶𝑅𝑅𝑀=7.5,𝜎′𝑣𝑐=1 = 0.8 ×𝑆𝑢
𝜎′𝑣𝑐
Equation 3-2
In-situ and laboratory tests are needed to determine the properties required in the
equations. Then, the cyclic behavior of the material can be characterized according to
plasticity index. The CRR relationship in Equations 3-1 and 3-2 are empirical relationships
developed for a wide range of soils and stress conditions. These empirical correlations, also
known as “simplified” procedure, are easy to use. However, there is a considerable
uncertainty in the estimated cyclic resistance ratios (CRR) from these empirical
correlations. In particular, these correlations have been primarily developed for sand and
sand-like materials and their applicability to estimating CRR for FCR is uncertain. The
study presented in this chapter aims to give an insight on the accuracy of using the
simplified procedures to estimate CRR for FCR. This is achieved by comparing the CRR
estimated from the simplified procedures and the CRR from cyclic DSS tests.
The clean sand-equivalent, overburden-corrected SPT number ((N1)60cs) for the
sample tested in the cyclic DSS tests (i.e. S1B2-D) was adopted to empirically calculate
the sand-like CRR of the sample, per Equation 3-1. Although loading conditions are
different between SPT and cyclic DSS tests, this comparison could enable us to understand
the accuracy of the simplified SPT-based procedures in estimating CRR for the FCR. This
54
comparison may also help practicing engineers to determine whether the commonly used
SPT-based simplified procedures can be used for liquefaction triggering evaluation of
FCR. According to the SPTs conducted in the study, the corrected SPT number ((N1)60)
was estimated to be 6.9. Subsequently, the clean sand equivalent value ((N1)60cs) was
calculated based on the approach proposed by Idriss and Boulanger (2008) given 40% fines
content, according to Figure 3-2. Therefore, the (N1)60cs and corresponding CRRsand-like
were 12.5 and 0.13, respectively. Furthermore, the undrained shear strength of the sample
was estimated to be 80.4 kPa using the triaxial test results. Idriss and Boulanger (2008)
also proposed to decrease the 𝐶𝑅𝑅𝑀=7.5,𝜎′𝑣𝑐=1 calculated by Equation 3-2 by 20% for
tailings, therefore, 𝐶𝑅𝑅𝑀=7.5,𝜎′𝑣𝑐=1, which is CRRclay-like, was determined to be 0.51.
The CRR calculated from the above empirical correlations were compared against
the CRR obtained from the cyclic DSS tests. The CRRM=7.5 of the S1B2-D, which is the
CSR that liquefies the sample in 15 cycles, was calculated by adopting the power equation
developed in Figure 3-9. Therefore, the CRRM=7.5 of the S1B2-D with PI of 2 (as seen in
Table 3-2) was determined to be 0.1 based on the CSR-N power equation. The CRRM=7.5
was converted to 𝐶𝑅𝑅𝑀=7.5,𝜎′𝑣𝑐=1 by applying the overburden correction factor (Kσ). The
overburden correction factor (Kσ) was 1.04 using the correlations by Idriss and Boulanger
(2008) and the corresponding 𝐶𝑅𝑅𝑀=7.5,𝜎′𝑣𝑐=1 was almost 0.1. The CRRs of FCR from the
two methods is shown in Figure 3-13. The empirical correlations by Idriss and Boulanger
(2008) estimate the CRR of FCR generally well, assuming the FCR has sand-like behavior.
However, as described earlier, the stress-strain loops and the pore-water-pressure
generation resemble those of clay-like behavior. According to the estimated values, sand-
55
like cyclic behavior is expected for the tested FCR (Figure 3-13). As seen in Figure 3-13,
the estimated 𝐶𝑅𝑅𝑀=7.5,𝜎′𝑣𝑐=1 is noticeably close to the transitional zone where the
behavior of the material changes from sand-like to clay-like over a small range of PI.
Therefore, observing clay-like cyclic response of the FCR in cyclic DSS is not a surprise.
Eventually, the FCR can be classified as a material that has transitional cyclic behavior
from sand-like to clay-like behavior.
Figure 3-13 Cyclic behavior of the FCR based on Idriss and Boulanger (2008) criterion
Post-liquefaction shear strength characteristics of the liquefied FCR, which had
experienced 5% DAS, were evaluated by conducting a static shearing immediately after
the cyclic loading. The post-liquefaction shear strength and stiffness properties are key
characteristics to evaluate the stability of tailings dams after seismic events. Although the
static DSS is not the best approach to determine the post-liquefaction shear strength, as the
56
potential void redistribution after liquefaction in field cannot be sufficiently captured in a
relatively uniform sample in the DSS equipment, the basic shear behavior of the liquefied
material can still be characterized. The strain-controlled shear stress was applied to the
sample at the rate of 1.4% per hour, the test was continued up to 30% shear strain. Figure
3-14 shows the shear stress and pore pressure ratio (ru) against shear strain during the post-
cyclic static loading for all three samples that were previously tested under different CSRs.
The liquefied FCR samples were found considerably soft so that the post-liquefaction
modulus and shear strength were significantly low. For instance, the secant shear modulus
of the liquefied FCR samples, at 5% shear strain, was within 40 kPa to 70 kPa.
Further in the static loading, when the shear strain increased, shear strength began
to recover. Eventually, the peak post-liquefaction shear strength (Su,pl) were 12 kPa, 12.5
kPa, and 15 kPa for the samples cyclically loaded by CSR of 0.1, 0.12, and 0.15,
respectively. The increasing trend in Su,pl as the CSR increased can be attributed to higher
void redistribution, subsequently, higher densification induced to the sample that was
subjected to higher CSR. Considering the figures of pore pressure ratios, the samples
showed dilative behavior during the static shearing, as the pore pressure ratio that had
developed during the cyclic loading phase decreased during the static shear phase.
57
Figure 3-14 Post-liquefaction shear strength characteristics of FCR
58
3.6 Conclusion and Summary
Basic geotechnical properties in terms of unit weight, classification, Atterberg
limits, specific gravity, and hydraulic conductivity were determined to characterize the
FCR samples that were obtained from two Appalachian coalfields. All the studied samples
were classified as silty sand or sandy silt with plasticity index lower than 5. The measured
unit weight of the representative samples varied noticeably through depth. Hydraulic
conductivity of the tested FCR samples was mostly within a narrow range from 1.0e-6 cm/s
to 3.6e-7 cm/s. However, the FCR sample taken from the location close to the coal slurry
discharge point showed higher unit weight and hydraulic conductivity compared to other
tested samples, implying the higher accumulation of coarse particles.
Staged triaxial tests and resonant column tests were conducted on representative
samples taken from different locations and depths in the impoundment. The samples at
deeper depth consistently showed higher shear strength and stiffness. The effective
cohesion and internal friction angle of the samples tested under CU condition were from
13.8 kPa to 25.5 kPa and 26 to 31, respectively. Lower cohesion and higher internal
friction angle were also observed for the samples tested under CD condition compared to
those under CU condition. The effect of confining pressure was found to be negligible on
normalized shear modulus at shear strain level less than 10-3%. The damping ratio ranged
from 0.6% to 2% for the FCR samples.
The liquefaction resistance and cyclic behavior of FCR were assessed by cyclic
DSS testing on reconstituted samples. FCR samples, taken from deeper depth in the vicinity
of the coal slurry discharge point that may substantially contribute to the failure of
59
impoundments, were prepared per slurry deposition method that resembles the structure
and fabric of the FCR in the field. The CSR-N relationship was established. The cyclic
resistance ratio (CRR) was close to the values estimated from empirical correlations for
sand-like behavior material based on the procedures by Idriss and Boulanger (2008). On
the other hand, the shear stress-strain loops and pore pressure ratio exhibited clay-like
behavior. Therefore, the FCR cyclic behavior was perceived to be transitioning from sand-
like to clay-like. Furthermore, the post-liquefaction shear behavior of FCR showed a
dilative response, as pore pressure ratio showed a decreasing trend from the beginning of
the static loading. The undrained shear strength of FCR after liquefaction was found to
range from 12 kPa to 15 kPa. It was also noticed that higher CSR induced higher
densification, consequently, slightly higher post-liquefaction peak shear strength was
observed compared to those liquefied with lower CSR.
Chapter 4
Strain History and Short-Period Aging Effects on the Strength and
Cyclic Response of Fine-Grained Coal Refuse
Tailings are the residual materials (soils) left from mineral extraction and
processing and are mostly composed of fine-grained particles with high water content.
These soils are typically stored in tailings impoundment in slurry form and consolidated
under their own weight; such deposition process results in a loose and sensitive soil
structure. Tailings dams often are partially composed of the tailings material and less stable
than conventional engineered dams used for water storage, with an annual failure rate about
120 times higher than that for water storage dams (Azam and Li 2010). The most common
cause of tailings dam failure is liquefaction of the tailings material (Martin and Davis 2000;
Rico et al. 2008). As described by Youd and Idriss (1998), depending on the strain behavior
of the soil, there are generally two types of liquefaction: flow liquefaction and cyclic
softening, as shown in Figure 4-1. Flow liquefaction mainly results from strain softening
and shear strength loss and can be triggered by either static or cyclic loading. The shear
strength loss under static loading is the result of excess pore pressure generation or pore
pressure redistribution. Flow liquefaction leads to significant shear deformations and flow
failure. Cyclic softening is a common liquefaction mechanism that is mainly due to loss of
shear stiffness (Robertson and Wride 1998). Cyclic softening itself can be divided into
cyclic liquefaction and cyclic mobility, depending on shear strain behavior, excess pore
pressure generation, and post-cyclic loading evidences such as sand boils (Seed et al.
61
1975). If limited excess pore pressure is generated but there is incremental shear strain
accumulation during the cyclic loading, the phenomenon is called cyclic mobility; when
high excess pore pressure is generated, the phenomenon is called cyclic liquefaction (Seed
1979).
Figure 4-1 Liquefaction definitions for various scenarios (Seed 1979, Robertson and
Wride 1998, Youd and Idriss 1998)
Various types of tailings have been assessed for liquefaction susceptibility and
other geotechnical properties (e.g., Ishihara et al. 1981; Castro 2003; Zeng et al. 2008;
James et al. 2011; Salehian 2013; Geremew and Yanful 2013; Salam et al. 2019). The
62
results from the past research indicate that tailings are commonly characterized as low
plasticity and sandy fine-grained soils with high liquefaction potential, and conventional
laboratory tests such as cyclic direct simple shear and cyclic triaxial tests were used to
determine liquefaction behavior. However, these methods may have several shortcomings,
including: 1) small specimen tests cannot properly represent a stratified field deposit, 2)
water film effects and void redistribution inside a tailings deposit and their influence on
the mechanical behavior cannot be captured, and 3) reconstituted samples may not reflect
the fabric and structure of in-situ undisturbed tailings.
In-situ testing methods, such as the standard penetration test (SPT) and cone
penetration test (CPT), are usually not viable in tailings impoundments due to accessibility
and safety issues, and only a few such studies have been conducted. Salam et al. (2019)
performed SPT tests in two Appalachian coalfields using a light-weight track-mounted drill
rig and the corrected SPT blow counts ranging from 3 to 7 were reported. Shuttle and
Cunning (2007) performed piezocone (CPTu) tests at a mine tailings site and proposed an
effective stress framework to characterize liquefaction susceptibility for tailings with high
silt content. Kalinski and Salehian (2016) carried out CPT tests in coalfields in eastern
Kentucky and proposed correlations between in-situ test indices and coal refuse properties,
such as cyclic and post-cyclic resistance. Robertson et al. (2017) performed seismic CPTu
tests with compressional and shear wave velocity measurements at a mine tailings site and
compared the results with laboratory data; the results indicated that shear wave velocity
measurements can better characterize the mine tailings specifically at unsaturated or
cemented state, as it causes less disturbance compared to CPTu test.
63
Several studies have found a significant effect of strain history on mechanical
properties, particularly liquefaction resistance, for natural soils. Finn et al. (1970)
conducted the original study and reported that the magnitude of induced shear strain
dominates the subsequent changes in strength and stiffness. Similarly, Heidary and Andrus
(2012) concluded that the intensity of a shaking event is a key factor affecting the
liquefaction resistance in subsequent events. Teparaksa and Koseki (2018) observed that a
soil deposit with prior liquefaction has higher liquefaction resistance at the same relative
density. El-Sekelly et al. (2016a, 2016b) tested a loose silty sand deposit under strong and
weak shaking events in a centrifuge and found that weak shakes increased the liquefaction
resistance of the deposit, while strong shakes can eliminate the resistance gained from weak
shakes and reestablish low liquefaction resistance for the soil. The past studies on soil
liquefaction have mostly focused on granular soils, primarily clean sands to silty sands.
Price et al. (2017) investigated the effect of strain history on liquefaction resistance of non-
plastic to low-plasticity silts with different over-consolidation ratios using cyclic direct
simple shear tests with multiple cyclic loadings. They observed a progressive increase in
liquefaction resistance of the normally consolidated silts over multiple cyclic loadings,
while the over-consolidated silts lost liquefaction resistance after the first cyclic loading
event.
In addition to strain history, reconsolidation and aging can also influence the
liquefaction resistance of natural soils. Accordingly, consideration of aging effect in design
applications was recommended by Mitchell and Solymer (1984). The aging effect due to
primary consolidation and secondary compression, inter-particle cementation, and increase
in soil friction have been studied in terms of improvement in strength and stiffness
64
properties (Anderson and Stokoe 1978; Kim and Novak 1981; Mitchell and Solymer 1984).
The aging phenomenon consists of physical and chemical processes and commences with
the start of grain to grain contact formation after a major disturbance. The chemical process
involves cementation at particle contacts and greatly depends on time, presence of
cementing agent, contact area, and pore water chemistry. Chemical aging is expected to
play a minor role for coal tailings due to the presence of carbon dioxide, which delays the
chemical aging process (Boggs 2014). On the other hand, the physical aging process
involves the rearrangement of particles and increasing frictional resistance due to
secondary compression (Mesri et al. 1990) and is expected to dominate aging of coal
tailings. CPT was utilized in a few studies to determine aging effect in terms of increased
tip resistance over time (Mesri et al. 1990; Charlie et al. 1992; Jorshi et al. 1995; Wang et
al. 2019), the aging effect was found to be noticeable even during a time period of 2 to 3
months.
This chapter presents an experimental investigation in the effects of strain history
and aging on liquefaction resistance and cyclic response of FCR using a large-scale shaking
table and CPTu tests. The FCR specimen was slurry-deposited in a membrane-lined
laminar shear box (LSB) and characterized using CPTu for classification, homogeneity,
and liquefaction resistance. The FCR specimen was subjected to three shaking events with
resting periods in between to study the effect of short-term aging on undrained shear
strength. During shaking, the dynamic response of the FCR was evaluated in terms of
lateral shear strains and pore pressures. CPTu tests were also conducted before and after
each shake to assess the geotechnical properties and liquefaction susceptibility of the FCR
material over time.
65
4.1 Testing Method
4.1.1 Shake table system and deposition process
Piezo- The experimental program was conducted using a 1-g shake table system
with a single degree of freedom in the Civil Infrastructure Testing and Evaluation
Laboratory (CITEL) at Penn State University, as shown in Figure 4-2. The table has a
vertical payload capacity of 133 kN and is driven by a computer-controlled 245 kN
hydraulic actuator. The laminar shear box (LSB) consists of 10 steel frames (laminae) with
one degree of freedom and relatively frictionless motion in the back-and-forth direction.
The LSB has dimensions of 2.29 m (length) × 2.13 m (width) × 1.4 m (height), and the
maximum displacement of the top lamina relative to the table is 228 mm. A 1 mm-thick
flexible geomembrane liner was placed inside the LSB to contain the saturated FCR
specimen. Additional details regarding the shake table facility were provided by Wang et
al. (2019).
Figure 4-2 Laminar shear box, specimen preparation, and instrumentation
66
Instrumentation for the FCR specimen included piezometers and linear variable
differential transformers (LVDTs) to measure response during and after the shaking events.
Four LVDTs were attached to one side of the LSB at elevations of 180, 460, 800, and 1080
mm above the bottom of the specimen. The shear strain between the successive LVDTs
was measured and referred to as the average shear strain at elevations of 90 mm (denoted
as γ1), 320 mm (γ2), 630 mm (γ3), and 940 mm (γ4). An LVDT was also used to measure
the settlements on the top surface of the specimen during and after each shaking event.
Pore pressures were measured using five piezometers along the vertical centerline of the
LSB at elevations of 0, 250, 500, 750, and 1000 mm; two duplicate piezometers were also
embedded in the FCR specimen at elevations of 500 mm (PZ6) and 750 mm (PZ7),
respectively, as shown in Figure 4-2 (a). The duplicate piezometers were used to potentially
observe the extent of boundary effects and heterogeneity inside the specimen. All
piezometers had a measurement sensitivity of 0.1 kPa and were used to track the generation
and dissipation of excess pore pressure through the testing program. The piezometers were
held in place vertically by metal strings tied to a steel bar on a stationary frame above the
shake table.
Idriss and Boulanger (2008) found that the method of sample preparation (e.g.,
reconstitution) can have a great influence on the cyclic resistance of soils in laboratory
experiments. Accordingly, it was important to use a preparation method that produces an
FCR specimen with similar fabric and structure to FCR material in a field tailings facility.
In the field, FCR is usually hydraulically deposited in slurry form and consolidated under
self-weight. The hydraulic filling method introduced by Whitman (1970) was employed
in this study. Moist FCR material was obtained from a coal slurry impoundment in
67
Pennsylvania. The FCR specimen was prepared with a mixing ratio of 1 unit moist FCR
and 1.5 unit tap water (by volume). The slurry was deposited into the LSB in five lifts.
Each lift was allowed to consolidate under self-weight and the clear surface water up to
100 mm above the settled FCR was removed prior the placement of the next lift. In the
field, large particles tend to settle close to the slurry discharge point. To avoid such
excessive segregation in the laboratory FCR specimen but achieve the inherent
heterogeneity in the field, the discharge point was moved during the deposition process,
following two paths as shown in Figure 4-2 (b). The LSB area was divided into 6 zones,
the slurry deposition was performed by moving the discharge point from Zone 1 to Zone 6
(i.e. Path A), and then counterclockwise from Zone 6 to Zone 1 (i.e. Path B). The final
height of the FCR specimen, after two months of self-weight consolidation, was 1.15 m.
The completion of self-weight consolidation was determined by regularly measuring the
specimen’s height after slurry deposition until no further settlement was observed. The
weight of removed water was subtracted from the total weight of placed slurry to determine
the final saturated specimen weight. The calculated saturated unit weight of the FCR
specimen was 15.4 kN/m3 and the corresponding final void ratio was 0.86. Index tests
conducted on eight random FCR samples that were taken during the deposition process
indicated a fines content of 43% to 98%, with average of 70%, and an average plasticity
index of 7.0, as shown in Figure 4-3. Overall, the FCR was classified as sandy silt with low
plasticity based on the Unified Soil Classification System (USCS) and potentially
liquefiable based on the empirical method of Seed et al. (2003). Furthermore, Boulanger
and Idriss (2006) recommended that fine-grained soils with PI ≥ 7 is expected to show
clay-like behavior under cyclic loading.
68
Figure 4-3 FCR gradations and Atterberg limits for 8 random FCR samples
4.1.2 CPTu device and testing locations
An overhead frame and push system were used to conduct CPTu tests at different
locations on the FCR specimen (Figure 4-4). These tests were conducted using a standard
cone with diameter (D) of 35.7 mm, a cross-section area of 1000 mm2, an apex angle of
60°, and a penetration rate of 20 mm/s, as per ASTM D5778. CPTu tests were conducted
on a grid pattern, as shown in Figure 4-4, with each location identified by coordinates. For
example, CPTu-13 denotes a CPTu test conducted at column 1 and row 3 in Figure 4-4
(i.e., the upper left corner of the grid). Of the 12 possible locations in Figure 4-4, three
locations were used for piezometers and nine were used for CPTu tests. The spacing of
69
CPTu test locations in the current study exceeded a distance of 7D (~250 mm), and thus
was beyond the influence zone of CPT suggested in several studies. Yang (2006) showed
the influence zone is 0.5D to 3D for piles in compressible silty sands. Burns and Mayne
(1998) suggested that the shear stress influence zone of CPT cone in clays was about 1-10
mm, and the plasticized zone due to CPT penetration has a dimeter of 𝑟𝑝 = 𝑟0𝐼𝑟0.333, where
𝑟0 is the cone diameter and 𝐼𝑟 is the rigidity index, which is calculated as shear modulus
divided by undrained shear strength of the soil. Shear modulus and undrained shear
strength of the FCR were adopted from Salam et al. (2019), and 𝑟𝑝 was determined to be
less than 7D.
Figure 4-4 (a) FCR specimen plan view showing piezometer and CPTu test locations; and
(b) photograph of CPT testing process
70
4.1.3 Shake table test plan
Three shaking events were conducted on the FCR specimen to investigate the effect
of strain history on the strength and liquefaction resistance, with a static resting period
between two consecutive tests to assess short-period aging of the FCR material. Each
shaking event consisted of sinusoidal motion with a constant frequency of 1 Hz and
different peak horizontal acceleration (PHA), with a ramp-up and ramp-down cycle at the
beginning and end of each event, respectively. Figure 4-5 shows the acceleration time-
history of the shaking tests.
Figure 4-5 Acceleration-time input motions for the testing program
The FCR specimen was first subjected to a relatively weak shake with PHA of
0.16g and a duration of 24 seconds. This was followed by 7 days of reconsolidation, after
which the piezometers’ readings reached to their initial hydrostatic pressure and then
remained stable. The second shake was a strong shake with PHA of 0.4g and a duration of
24 seconds. The second shake was followed by 97 days of reconsolidation and aging. CPTu
71
tests were conducted at different times during the resting periods. Finally, the FCR
specimen was subjected to a strong shake with PHA of 0.4g and a longer duration of 62
seconds. The uniform sinusoidal loadings may not be adequate in representing an
earthquake loading in the field, as they are more destructive. The equivalent earthquake
PGA (PGAeq) was approximated by scaling the PHA value up by 1/0.65, as shown in Figure
4-5. Accordingly, the induced cyclic stress ratio (CSR) of the FCR specimen was
approximated using (PGAeq) based on the simplified approach of Seed and Idriss (1971).
This approach was adopted assuming negligible excess pore pressure generation during the
cyclic loading and that the FCR specimen is shallow enough to use PHA as peak surface-
ground acceleration. The induced CSR in physical models such as shake table and
centrifuge models could be determined by the approach proposed by Abdoun et al. (2013),
if acceleration data are available.
4.2 Results and Discussion
4.2.1 Pre-shake CPTu
Six initial CPTu tests were conducted after two months of self-consolidation and
before the first shaking event at locations CPTu-12, CPTu-13, CPTu-31, CPTu-33, CPTu-
51, and CPTu-52. Figure 4-6 presents the CPTu results in terms of corrected tip resistance
qt, sleeve friction fs, and soil behavior index Ic. The noticeable variation in the maximum
qt can be attributed to the heterogeneity and stratification of the FCR specimen. For
example, CPTu-31 shows a stratified medium, with higher qt observed at the middle depth.
72
The heterogeneity observed in CPTu results confirmed the earlier research findings of the
dependency of the tailings properties on the location and depth (Salam et al. 2019). Overall,
the maximum qt ranged from 20 kPa to 47 kPa.
Figure 4-6 Pre-shake CPTu results
The low values of qt indicate a weak and loose initial structure and the near-zero
values of fs indicate low plasticity and clay content of the FCR specimen. The soil behavior
index is defined as
0.52 2
3.47 log 1.22 log c tn rI Q F
Equation 4-1
73
where the normalized tip resistance Qtn is
n
t vo atn
a vo
q PQ
P Equation 4-2
the normalized friction ratio Fr is
100%
sr
t vo
fF
q Equation 4-3
the exponent n is,
0.381 0.05 0.15
voc
a
n IP
Equation 4-4
𝜎𝑣0 = initial vertical total stress, 𝜎′𝑣0 = initial vertical effective stress, 𝑃𝑎 =
atmospheric pressure, and the value of Ic was calculated using an iterative process
(Robertson 2009). The value of qt was very low (i.e., lower than the overburden stress)
such that Qtn, Fr, and Ic could not be determined at several locations (e.g., CPTu-52). Soil
behavior index (Ic) at the observed locations and depths was above 3.3.
The measured data were also used to characterize the FCR specimen using the soil
behavior type (SBT) charts proposed by Robertson and Wride (1998) (Figure 4-7 (a)),
Robertson (2009) (Figure 4-7 (b)), and Robertson 2016 (Figure 4-7 (c)). The SBT charts
can be used to determine various characteristics of a deposit such as classification,
behavior, and liquefaction susceptibility, as shown in Figure 4-7. Figure 4-7 (a) consists of
nine zones, each representing a soil classification. The CPTu results for the FCR show
three data points in Zone 1 (i.e., fine-grained sensitive soils), two data points in Zone 4
(i.e., clayey silt to silty clay), and one data point in Zone 3 (i.e., silty clay to clay).
74
Therefore, the FCR specimen can be assumed mostly as a normally consolidated fine-
grained soil, which is in agreement with the particle distribution data shown in Figure 4-3.
Figure 4-7 CPTu results plotted on soil behavior charts: (a) Robertson and Wride (1998),
(b) Robertson (2016) and (c) Robertson (2009)
The zones in Figure 4-7 (a), except for zones 1, 8, and 9, establish the value of Ic,
which increases as fines content and plasticity index increase. Ic=2.6 is the threshold,
beyond which soils transition from sand-like behavior to clay-like behavior. Accordingly,
a soil with Ic> 2.6 is identified as clay-like fine-grained soil (i.e. silt mixtures or clays).
Figure 4-7 (b) shows that all the FCR data points have Ic> 2.6 and are within a dashed box
defined by Qtn < 10 and Fr < 2, which denotes clay-like contractive sensitive (CCS) soil
(Robertson 2016). In terms of liquefaction susceptibility, soils found in the lower left of
the SBT charts have low liquefaction potential. However, sensitive soils may undergo
75
strain softening, and subsequently flow liquefaction. Further testing is required to analyze
liquefaction susceptibility of these soils.
Figure 4-7 (c) characterizes soil liquefaction behavior with zones A1, A2, B, and C
using two curves defined by Ic= 2.6 and the clean sand equivalent penetration resistance
Qtn,cs = 70. The CPTu results indicate that the FCR specimen is within Zone C (i.e.,
cohesive soils susceptible to cyclic softening and flow liquefaction). Salam et al. (2020)
analyzed the same CPTu data according to Jefferis and Davis (1991) and classified the
FCR specimen as susceptible to strain softening. The CPTu analysis indicates that the FCR
specimen is expected to show contractive and subsequently strain softening behavior under
dynamic loading.
4.2.2 Shake table test results
First shake (FS)
The first shake consisted of a uniform sinusoidal motion with PHA of 0.16g for 24
seconds. According to the liquefaction triggering analysis presented by Moss et al. (2006),
the maximum normalized tip resistance (qc,1), as defined in Moss et al. (2016), of the FCR
specimen ranged from 21 kPa to 91 kPa before the first shake. The modified normalized
tip resistance (qc,1,mod) was equal to qc,1, because the friction ratio was below 0.5% (Moss
et al. 2006) and no modification was required to account for frictional effects of fines in
the FCR specimen. Accordingly, the FCR specimen before the first shake was plotted at
the very left of the probabilistic CPT-based liquefaction triggering resistance curves, where
76
CSR as low as 0.1 can liquefy the specimen with probability higher than 95%. However,
Figure 4-8 showed that the pore pressure buildup during the shake was less than 1 kPa, and
pore pressure ratio (ru) was below 10% except for PZ-1, which was close to surface and
reached ru of 50%. Insignificant excess pore pressure generation in FCR under cyclic
loading has also been observed in several other studies (Zeng et al. 2008, Salam et al.
2019). After the shake, piezometers PZ-3, PZ-4, and PZ-5 showed a 1 kPa to 3 kPa increase
in pore pressure in response to the settlement of the FCR specimen. The settlement and
reconsolidation of the FCR specimen were accompanied by an upward seepage force that
was captured by the pore pressure sensors. The variable pore pressure readings can be
attributed to the heterogeneity of the FCR specimen and pore pressure redistribution. This
phenomenon can lead to flow liquefaction, consistent with the observation in Figure 4-7.
Flow liquefaction is likely if large deformations occur as result of pore pressure
redistribution (Robertson and Wride 1998). As shown in Figure 4-8, the pore pressure at
PZ-3 to PZ-5 slightly increased over sixteen minutes after the first shake and then
decreased to the initial value over seven days. The long dissipation period was due to the
high fines content in the FCR. Large vertical deformation (i.e. 50 mm) was observed as a
result of the first shake so that the FCR specimen thickness reduced from 1.15 m to 1.10 m
and its void ratio reduced from 0.86 to 0.78. The saturated unit weight of the FCR increased
from 15.4 kN/m3 to 15.6 kN/m3.
77
Figure 4-8 Pore pressures during and after the first shake (height of water above each
piezometer in parentheses)
The limited excess pore pressure during and after shaking indicates that the FCR
specimen should be examined for either flow liquefaction or cyclic mobility. The 3% peak
shear strain criterion has been adopted for clay-like soils and tailings in liquefaction
analysis (Boulanger and Idriss 2007). The developed shear strain throughout the FCR
specimen was determined at four depths, shown by γ1, γ2, γ3, and γ4 in Figure 4-2. The
shear strains are the relative displacement divided by the vertical spacing between two
adjacent LVDTs. The red dashed lines in Figure 4-9 show 3% shear strain at negative and
positive directions. Using the shear-strain-based criterion, the FCR specimen liquefied at
depths indicated by γ1, γ2, and γ3. The maximum shear strain quickly developed in the
beginning of the cyclic motion and remained constant until the end of the motion except
for γ4. The shear strain at the bottom of the specimen progressively decreased through the
78
motion, probably due to significant densification of the FCR at this depth. Figure 4-8
showed limited excess pore pressure generation during the shake and pore pressure buildup
after the shake due to upward flow and void redistribution. In addition, Figure 4-9 showed
relatively large maximum shear strain throughout the FCR specimen. The observations in
Figure 4-8 and Figure 4-9 implied the occurrence of cyclic mobility within the FCR
specimen. Although the incremental shear strain accumulation was not clearly observed
probably due to small number of cycles, the occurrence of cyclic mobility was further
confirmed as sand boils were observed at several locations on the surface of the FCR
specimen after the shake.
79
Figure 4-9 Developed shear strains within the FCR specimen during the first shake
Within 30 minutes after the first shake two CPTu tests at the locations of CPTu-12
and CPTu-13 were conducted. Seven days after the first shake, two CPTu tests at the
locations of CPTu-31 and CPTu-33 were conducted. According to the piezometers’ and
settlement readings the reconsolidation was completed after seven days. Figure 4-10 shows
the CPTu results before and after the first shake at the tested locations. The dashed lines
show the CPTu profile before the first shake, while the solid lines delineate the CPTu
profile after the first shake. The corrected tip resistance readings (qt) showed strength
reduction of the FCR specimen immediately after the first shake. The maximum qt at CPTu-
80
12 before the first shake was 48 kPa and it dropped to 30 kPa at 20 minutes after the shake.
Although qt at CPTu-13 below 0.8 m was slightly higher at 30 minutes after the shake, the
CPTu-13 profile at 30 minutes after the shake consistently showed lower tip resistance
from surface down to 0.8 m compared to CPTu-13 before the shake. The FCR specimen
showed strength gain at the end of the reconsolidation (i.e. after seven days). The CPTu
profiles indicated that although a shaking event can have immediate detrimental effects on
the FCR’s strength, densification and reconsolidation can gradually strengthen the
specimen. The strength reduction was caused by the excessive deformation and shear
strains induced by the shake. However, a denser particle interlocking was achieved due to
reconsolidation and densification of the FCR specimen. Therefore, the maximum qt at
CPTu-33, which was recorded as low as 23 kPa at 0.9 m below the surface before the shake
increased to 40 kPa after reconsolidation. The sleeve friction was noticeably low for the
FCR specimen, as shown in Figure 4-6, due to the absence of high plasticity clay. The
variation in the magnitude of sleeve friction was also observed to be small and negligible;
therefore, the sleeve friction profiles were excluded.
81
Figure 4-10 CPTu test results before and after the first shake
Second shake (SS)
Once the reconsolidation of the FCR specimen after the first shake was completed,
a uniform sinusoidal motion with PHA of 0.4g was imposed to the shake table for 24
seconds. The maximum modified normalized tip resistance (qc,1,mod) ranged from 216 kPa
to 222 kPa before the second shake. The friction ratio did not exceed 0.5% after the first
shake. Accordingly, the FCR specimen was still plotted at the far left of the probabilistic
CPT-based liquefaction triggering resistance curves, and the FCR specimen remained
susceptible to liquefaction based on the liquefaction triggering analysis by Moss et al.
82
(2006). However, less than 1 kPa excess pore pressure was generated during the second
shake, and the excess pore pressure rose up to 1 kPa after the shake due to settlement of
the FCR specimen. The smaller post-shake increase in piezometers’ reading was due to the
denser structure of the FCR specimen. The FCR specimen settlement after the second shake
was 90 mm; therefore, the height and void ratio of the FCR specimen reduced from 1.10
m and 0.78 to 1.01 m and 0.64, respectively. The saturated unit weight of the FCR specimen
also increased from 15.6 kN/m3 to 16.1 kN/m3.
Similar to the first shake, the FCR specimen was examined for the cyclic mobility
occurrence. Figure 4-11 shows the developed shear strain throughout the FCR specimen
during the second shake. The entire FCR specimen liquefied according to the 3% peak
shear strain criterion. The stratification of the FCR specimen was noticed, as the maximum
shear strain was not consistent through depth. The FCR at the bottom densified and showed
decreasing shear strain trend (γ4) during the second shake, similar to the first shake. The
maximum shear strain was developed at shallow depth (γ1) and remained constant until the
end of the second shake. The developed shear strains that are represented by γ2 and γ3
clearly showed cyclic softening, as the peak shear strain at both directions progressively
increased during the second shake. The double amplitude strains at γ2 and γ3 were 13% and
17% at the beginning and increased to 15.5% and 24.5% at the end of the second shake,
respectively. Although the increase in shear strain may be due to higher shear stresses
imparted to shallower depth due to densification of deeper layers, the lack of high excess
pore pressure generation during the shake and evidence of sand boils on surface after the
shake suggested that the progressive shear strain accumulation was due to the occurrence
83
of cyclic mobility (Castro 1987). Cyclic mobility, as a sub-category of cyclic softening is
particularly common in heterogeneous deposits (Seed et al. 1975).
Figure 4-11 Developed shear strains within the FCR specimen during the second shake
Ten CPTu tests were conducted at various elapsed time up to 97 days after the
second shake to capture the aging and strength characteristics over time. All the CPTu
results including those before the first shake, after the first shake, and after the second
shake are shown in Figure 4-12. The sleeve friction results were excluded due to negligible
magnitude and variation. The CPTu profiles before the first shake, after the first shake, and
84
after the second shake are plotted with dotted, dash-dotted, and solid lines, respectively.
Each color represents a single location in the FCR specimen. According to Figure 4-12, the
effects of densification and strain history were found substantial, as qt after the second
shake was equal to or larger than qt before the second shake at different timing of the CPTu
tests. The qt profiles within seven days after the second shake (CPTu-31, 33, 51, 52, 12,
and 13) did not show significant increase since the primary consolidation was still in
progress. After seven days, the secondary compression initiated and noticeable increase in
tip resistance was observable, as shown by the CPTu profiles at 14, 21, 48, and 97 days.
The qt of the FCR specimen consistently increased over time as demonstrated in Figure 4-
12. In terms of modified normalized tip resistance, the qc,1,mod of the FCR specimen
increased from 0.02 MPa before the first shake to 2 MPa after two shakes followed by 97
days of aging. Although the FCR’s qc,1,mod increased by 100 times, the specimen remained
at the far left side of the probabilistic CPT-based liquefaction triggering resistance curves
(Moss et al. 2006) and was still susceptible to liquefaction.
85
Figure 4-12 CPTu results before and after the first shake and up to 97 days after the
second shake
Third shake (TS)
The third shake with PHA of 0.4g and a duration of 62 seconds occurred 97 days
after the second shake. The dynamic response of the densified and aged FCR specimen
under the same PHA but longer duration was investigated. Less than 1 kPa excess pore
pressure was generated since the specimen was even denser compared to the previous
shakes. Figure 4-13 presents the developed shear strains through depth. γ1 is not shown as
LVDT1 was excluded in the third shake, since the settled FCR surface was below the
elevation of LVDT1. Compared with the second shake, γ3 had smaller values in the third
86
shake, supporting the assumption of densification and strength gain of the FCR specimen.
In addition, the progressive shear strain accumulation observed in the second shake (shown
in Figure 4-11) was not noticed in the third shake, shown in Figure 4-13. This observation
indicates that the specimen may not be susceptible to cyclic softening anymore, but
susceptible to strain softening, consequently flow liquefaction (Robertson and Wride
1998), due to the low strength of the FCR (i.e. qc,1,mod equal to 2 MPa) despite the preceding
shakes and aging. The FCR specimen liquefied again based on the 3% peak shear strain
criterion and sand boil evidence was noticed at a couple of spots on the FCR surface. The
strength evolution and variation in displacement and shear strain behavior over the shaking
events emphasized the need to study the effects of strain history and aging in more depth.
Figure 4-13 Developed shear strains within the FCR specimen during the third shake
4.2.3 Effect of strain history
The mode of response and displacement throughout the FCR specimen could
indicate the effect of strain history. Figure 4-14 shows the maximum lateral displacement
87
in negative and positive directions during the three shaking events. Although acceleration
data could validate the following inferences, possible scenarios leading to the observations
in Figure 4-14 are presented and discussed since acceleration data were not available in
these experiments. The lateral displacements were measured by the four horizontal LVDTs
(shown in Figure 4-2) at elevations of 180 mm, 460 mm, 800 mm, and 1080 mm, with
elevation zero at the bottom of the specimen. The dashed lines indicate the input
displacement. The maximum induced lateral displacement by the first shake was 41 mm,
as shown in Figure 4-14 (a). Displacements in the right and left directions refer to the
positive and negative displacements, respectively. During the first shake, the maximum
positive lateral displacements at elevations 180 mm and 460 mm were close to the input
motion, but the specimen showed softened behavior in negative direction since the
maximum negative lateral displacement was larger than the input motion. A de-
amplification was noticed in lateral displacement at elevation 800 mm, while an
amplification was observed at elevation 1080 mm. This observation could be attributed to
two probable scenarios: one reason could be the stratification of the FCR specimen, which
was revealed by the CPTu tests and piezometer readings, another reason could be the
liquefaction-induced de-amplification of the specimen below elevation 800 mm and the
amplification of the top section of the specimen caused by the responses of side walls. The
largest maximum lateral displacement during the first shake was 59 mm.
88
89
Figure 4-14 Maximum lateral displacement of the FCR specimen during the (a) first
shake (b) second shake (c) third shake (the horizontal line represents the FCR surface)
Figure 4-14 (b) presents the maximum lateral displacement at the four elevations
through the FCR specimen during the second shake. The maximum input lateral
displacement in the second shake was 74.1 mm, and it is marked by the dashed lines. The
maximum positive lateral displacement was close to the input motion, while the maximum
negative lateral displacement was larger than the input motion except for the surface,
indicating a softened behavior in negative direction. The largest maximum lateral
displacement was observed at elevation 460 mm and was equal to 89 mm. From elevation
460 mm toward the surface, the maximum lateral displacement decreased, showing de-
amplification of the motion within the FCR specimen that could be either due to
liquefaction of underlying deposit or stratification of the FCR specimen.
90
Figure 4-14 (c) shows the maximum lateral displacement at three elevations 180
mm, 460 mm, and 800 mm of the FCR specimen during the third shake. The LVDT1 was
excluded, as it was located above the specimen surface due to prior settlements of the FCR
specimen. The maximum input lateral displacement in the third shake was 74.1 mm,
indicated by the dashed lines in negative and positive directions. Similar to the previous
shakes, the FCR specimen was softened in negative direction, but with less intensity. The
maximum lateral displacement was less than the input motion at all elevations, indicating
densification and strength gain of the FCR specimen prior to the third shake, as shown in
Figure 4-10 and Figure 4-12. The de-amplification of the motion toward the FCR surface
was almost consistent up to the observed depth of LVDT2. One reason for this observation
could be the segregation of particles such that heavier particles settled and were overlain
by fines due to successive shaking events. Therefore, the motion experienced more de-
amplification as it moved upward through finer particles. The other possibility for the
consistent decrease in maximum lateral displacement is the decrease in CSR due to
liquefaction of underlying deposit; due to the unavailability of acceleration data, this could
not be verified. Smaller maximum lateral displacement observed during the third shake
clearly shows the significant effect of strain history on the dynamic response of the FCR
specimen.
In addition to the displacement observations, CPTu results can be examined to
study the strain history effect. The normalized CPTu results (Qtn and Fr) were compiled
and plotted in the SBT chart, as shown in Figure 4-15 and Figure 4-16. The data points’
evolvement on the SBT chart indicated the variation in strength characteristics and
liquefaction behavior due to strain history. In order to solely investigate the strain history
91
effect, the timing of the CPTu tests within each stage was ignored. The average (ave) and
standard deviation (std) of Qtn and Fr for three stages before the first shake (i.e. before-
FS), between the first shake and second shake (i.e. after-FS), and after the second shake
(i.e. after-SS) were calculated. The average and standard deviation were used to
establish a pocket for each stage. Each pocket is made of four ellipse’s quarters, as a
circular shape could not be established in the logarithmic scale. The vertices are at (Qtn,ave,
Fr,ave± Fr,std) and co-vertices are at (Qtn,ave± Qtn,std, Fr,ave).
Figure 4-15 FCR specimen classification over the test plan
According to the before-FS pocket, the FCR specimen was mostly located within
Zone 1, which represents sensitive fine-grained soils. The first shake that was a relatively
weak shake reduced the sensitivity of the FCR specimen. Therefore, the after-FS pocket
shifted toward Zone 3 that represents silty clay to clay. The CPTu data and the after-FS
92
pocket clearly showed how the shaking event strengthened the FCR specimen and
diminished sensitivity by densification. The second shake further densified the FCR
specimen, creating stronger structure. The after-SS pocket demonstrated an upward
movement out of the clay-like-contractive-sensitive (CCS) box. The after-SS pocket was
mostly located in Zone 4: clayey silt to silty clay soils, with some overlap with Zone 5,
which includes higher Qtn and belongs to silty sands to sandy silts. Although partial to full
liquefaction of the cyclic mobility type was observed in both FS and SS events, the strain
history overall resulted in strength evolution within the FCR specimen. Figure 4-15
demonstrates that the FCR specimen transitioned from a sensitive fine-grained soil toward
soils with higher resistance.
Figure 4-16 FCR specimen liquefaction behavior over the test plan
The influence of strain history and strength evolution noticed in Figure 4-15 was
further studied in Figure 4-16 in terms of liquefaction susceptibility and behavior. Figure
93
4-16 delineates the four distinctive zones of various types of liquefaction along with the
data points and established pockets. The FCR specimen remained susceptible to cyclic
softening and flow liquefaction even after the second shake. The second shake resulted in
transition of the CPTu results toward Zone A2, such transition indicates less susceptibility
to cyclic softening and flow liquefaction as suggested by Figure 4-16. Although less shear
strains were observed during the third shake, evidences of localized cyclic mobility (e.g.
sand boils at a few locations), as a type of cyclic softening, were observed, confirming the
SBT chart recommendation. The occurrence of flow liquefaction could not be adequately
investigated as the specimen was contained and flow failure was not observable.
4.2.4 Effect of short-period aging
Since Short-period aging effect is the strength gain after a disturbance event (e.g.
an earthquake or blasting). The strength gain after disturbance begins with primary
consolidation and continues with secondary compression. The secondary compression is
caused by creep movement and particles rearrangement to reach an energy equilibrium.
Considering the short-period investigation in this research and the presence of carbon
dioxide in the FCR specimen (Boggs 2014), chemical aging was assumed to be negligible
and was not considered in this study. The primary consolidation is a relatively short process
in tailings due to the relatively high sand content, low clay content, and low plasticity.
Therefore, the improving effect of secondary compression may be noticed in a short period
of time. The strength gain behavior and timing of this phenomenon are essential in stability
analysis of coal tailings dams specifically after a disturbance event. The FCR generally has
94
much less strength compared to natural soils and any strength gain is advantageous for
stability consideration.
Average qt versus time after the second shake was plotted in Figure 4-17 at effective
stress of 4 to 5 kPa (approximately at γ3 elevation), which was reasonably distant from the
surface and bottom boundaries. The liquefied FCR did not show noticeable strength gain
up to approximately 10,000 min (seven days) after the shake. The primary reconsolidation
was completed seven days after the shake, then the secondary compression commenced.
As a result, qt started to increase. A correlation between qt and elapsed time after the shake
was developed and shown in Figure 4-17.
95
Figure 4-17 Strength gain trend over time for clean sand and FCR
The aging effect on a shallow specimen of liquefied clean sand was previously
studied by Wang et al. (2019). The increasing trend of qt over time at the same effective
stress (i.e. 4 to 5 kPa) was included in Figure 4-17. A significant strength difference
between the shallow specimen of clean sand and shallow specimen of FCR can be noticed.
The excess pore pressure dissipation and primary consolidation in the clean sand took only
1000 min (less than a day); this can be attributed to the absence of fines in sand. The
increasing trend during the secondary compression for clean sand was significantly sharper
compared to that of FCR. This observation can be explained by the slower rate of secondary
96
compression for soils with fines compared to clean sand. Wang et al. (2019) investigated
the increasing trend of qt at higher effective stresses (up to 11.3 kPa) and noticed that
although the tip resistance was higher at higher effective stress, the increasing trend
remained the same. Although this might be true for FCR too, the increasing trend at
different effective stress for FCR should be further investigated.
The estimation of strength gain after a major disturbance is beneficial for post-
disturbance stability analysis. The equation originally developed by Mesri (1987) can be
used to estimate the undrained shear strength over time for the FCR after a disturbance
event. Mesri (1990) developed an equation to predict undrained shear strength during
secondary compression (Equation 4-5).
(𝑆𝑢)2
(𝑆𝑢)1= (
𝑡2
𝑡1)
𝐶𝐷×𝐶𝛼
𝐶𝐶⁄
(Equation 4 − 5)
The undrained shear strength at t2 ((𝑆𝑢)2) over the undrained shear strength at t1
((𝑆𝑢)1) is estimated by the time ratio to the power of compressibility ratio (𝐶𝛼
𝐶𝐶⁄ ), where
𝐶𝛼 is secondary compression index and 𝐶𝐶 is the compression index. The compressibility
ratio is constant and has a limited range (Terzaghi et al. 1996). CD accounts for the power
of disturbance event. The undrained shear strength can be estimated by Qtn (Robertson
2009). Accordingly, Equation (4-5) was modified by replacing the undrained shear strength
by Qtn. The compressibility ratio was estimated as 0.04 for the FCR (Xiao et al. 2019).
(𝑄𝑡𝑛)2
(𝑄𝑡𝑛)1= (
𝑡2
𝑡1)𝐶𝐷×0.04 (Equation 4 − 6)
Mesri et al. (1990) compiled and presented CD over a range of normalized void ratio
change (∆𝑒𝑅 = ∆𝑒 (𝑒𝑚𝑎𝑥 − 𝑒𝑚𝑖𝑛)⁄ ) based on various ground-modification projects such as
97
blasting and dynamic compaction (Solymer 1984; Schmertmann et al. 1986; Dowding and
Hryciw 1986), as shown in Figure 4-18. CD was fitted to match the CPTu data in this study.
For example, CD was calibrated to fit the observed ratio of Qtn at 7 days (secondary
compression initiation) to 97 days. The observed Qtn ratio was equal to 2.14, indicating
114% increase in the undrained shear strength of the FCR specimen. CD value of 7 matches
well with the calculated ratio of normalized tip resistance. Similarly, the CD value that was
determined for 21 days after the shake was equal to 8.
Figure 4-18 CD values for FCR in the shake table test
98
Although the change in void ratio (∆𝑒𝑅) of the FCR specimen was recorded over
time, the 𝑒𝑚𝑎𝑥 and 𝑒𝑚𝑖𝑛 of the FCR was not available and could not be accurately
determined. Therefore, a shaded box representing a range for ∆𝑒𝑅 (±2.5% of the estimated
value) was defined for each CD value, as indicated in Figure 4-18. The CD ranged from
approximately 7 to 8, decreasing with longer period; this was in agreement with the
Solymer (1984) observations after vibro-compaction. Overall, CD was found at the lower
bound of the proposed range by Mesri et al. (1990). This approximation may be used to
estimate the strength gain in terms of undrained shear strength within coal tailings retained
by tailings dams after disturbance events.
4.3 Summary and Conclusions
This chapter reported the effects of strain history and short-period aging on the
strength and cyclic resistance of fine-grained coal refuse (FCR) by physical modeling. A
testing program including three shaking events with rest period and CPTu tests between
the events was conducted in this study. FCR collected from an active coal refuse
impoundment was slurry-deposited into a laminar shear box (LSB) to reach similar
structure and fabric of the material in the field. During the deposition, eight batches were
randomly selected to determine classification and Atterberg limits of the FCR specimen.
CPTu test results and instrument readings were used to characterize the cyclic behavior of
the FCR specimen.
The FCR specimen was mostly sandy silt with low plasticity and potentially
liquefiable based on Atterberg limits. According to the CPTu results prior to the first shake,
99
the FCR specimen was classified as sensitive fine-grained soil susceptible to cyclic
softening and flow liquefaction. Although the FCR liquefied based on the 3% peak shear
strain criterion, the magnitude of excess pore pressure generated during the shakes was
limited due to large amount of fines in the FCR specimen. The lack of plasticity and
cohesion, soft and loose structure, and interlayered medium of FCR resulted in cyclic
mobility, which was found to be the dominant type of liquefaction for the FCR specimen
in this study. The residual excess pore pressure developed during the cyclic loadings caused
sand boils on the FCR surface after the shakes. The incremental shear strain accumulation,
limited excess pore pressure, and sand boils confirmed the occurrence of cyclic mobility,
which was facilitated by the heterogeneity and void redistribution within the FCR
specimen.
Shaking events progressively densified the FCR specimen and increased the
liquefaction resistance. The CPTu tests showed higher tip resistance and less sensitivity to
cyclic loading as a result of shaking events. The FCR specimen was classified as clayey
silt to silty clay after two shakes followed by a short-period aging based on CPTu results.
The densification and short-period aging resulted in smaller shear strains in the last shaking
event. However, the clean sand equivalent penetration resistance of the FCR specimen
remained below 70; hence, the FCR was still considered susceptible to cyclic softening and
flow liquefaction. The modified normalized tip resistance also suggested that the FCR
specimen was still liquefiable, as the FCR specimen was plotted at the left end of the
probabilistic CPT-based liquefaction triggering resistance curves.
The FCR specimen’s strength was significantly lower than a clean sand in an earlier
study using the same testing facility. In addition, the FCR specimen’s aging rate was found
100
to be slower compared with the rate observed for clean sand and other types of soil.
However, the aging effect of the FCR even within a short period was substantial, as the tip
resistance increased over 100% within 97 days.
Chapter 5
Seismic Stability of Coal Tailings Dams with Spatially Variable and
Liquefiable Coal Tailings using Pore Pressure Plasticity Models
Coal tailings (CT) are the residue as a result of mine extraction process and mostly
consist of water, coal fraction, and non-coal materials such as sand and silt. CT are
commonly characterized as low plasticity silty sand to sandy silt and are typically deposited
in the form of slurry behind tailings dams. Generally, tailings dams have more vulnerability
than conventional and engineered dams used for water storage, and their annual failure rate
is 120 times higher than that of water-storage dams (Azam and Li 2010). Tailings dams are
constructed by three methods: downstream, centerline, and upstream, amongst which the
upstream configuration has the least stability (Vick 1990). A recent example of a tailings
dam failure was the Vale’s Brumadinho iron ore tailings dam in Brazil in 2019, which was
the 11th most serious tailings dam failure in the last decade and resulted in over 300 life
losses and significant social, economic, and environmental impacts (Home 2019).
One of the most common causes of tailings dams’ failure is liquefaction (ICOLD
2001; Rico et al. 2008). Liquefaction of CT could lead to different forms of failure such as
failure of the dam’s slope due to weakened and liquefied underlying layers, overtopping of
the liquefied material, and increase of lateral pressure on the dikes (ICOLD 2001).
Engineering procedures and numerical modeling tools can be used to better approximate
these complex processes and consequently assess the seismic stability of CT dams for a
variety of demand and capacity scenarios. Various constitutive plasticity models such as
102
UBCSAND (Beaty and Byrne 1998), PM4Sand (Boulanger and Ziotopoulou 2017), and
PM4Silt (Boulanger and Ziotopoulou 2018) have been developed to approximate the
response of sand and low plasticity silt in earthquake engineering applications. However,
the applicability of these models to simulate cyclic behavior of CT has not been accurately
assessed. UBCSAND, a non-linear effective stress plasticity model, was proposed by
Beaty and Byrne (1998) to determine the mechanical response of sand under cyclic loading.
UBCSAND has been used to simulate the dynamic behavior of sand and low plasticity
tailings in engineering practices (Seid-Karbasi and Byrne 2004; Castillo et al. 2005; James
2009; Ferdosi et al. 2015). PM4Sand and PM4Silt are plane-strain bounding surface
plasticity models developed by Boulanger and Ziotopoulou (2017, 2018). PM4Sand
assesses the drained and undrained, and cyclic and monotonic mechanical responses of
sands and non-plastic silts, while PM4Silt assesses those of low plasticity silts and clays.
Both the PM4Sand and PM4Silt plasticity models are based on the framework of the stress-
ratio controlled, critical state compatible, bounding-surface plasticity model for sand
developed by Dafalias and Manzari (2004). PM4Sand and PM4Silt have been successfully
used to simulate both sandy materials (Ziotopoulou and Boulanger 2016; Ziotopoulou and
Montgomery 2017) and alluvial silty deposits (Boulanger and Montgomery 2016,
Boulanger 2019). Field and laboratory testing by Salam et al. (2019) showed that the cyclic
behavior of CT is complex and transitioning from clay-like to sand-like, because the
composition of CT is a mixture of sand and silt. Therefore, both PM4Sand and PM4Silt
could be potentially used for simulating the cyclic behavior of CT.
Coal tailings have noticeable heterogeneity and spatial variability. In a recent study,
Liew et al. (2020) showed the significant heterogeneity in coal tailings properties using in-
103
situ seismic investigations in an active Appalachian coalfield. Such spatial variability of
slurry’s geotechnical properties is caused by variations of slurry discharge locations and
extracted coal materials during the service time of a tailings impoundment. Therefore, the
spatial variability of properties should be considered in the stability analyses, as a model
with uniform properties may not capture the critical failure modes. For example, the failure
of the Kingston Tennessee Valley Authority (TVA) coal ash impoundment was partially
due to the liquefaction of a loose layer under the dikes (Plant and Harriman 2008). This
mode of failure cannot be estimated unless the stratified medium of tailings is accounted
for in the stability analysis.
In this chapter, the uncertainty in seismic response of a typical upstream-
construction CT dam is analyzed considering the variability in CT geotechnical properties.
A suitable pore pressure plasticity model for simulating the cyclic response of CT is
selected by single element simulations and calibrations against experimental results. A
representative number of realizations for the CT section of the dam are generated by the
Karhunen-Loeve expansion method. It is of interest to evaluate how system response and
its uncertainty are influenced by input motion characteristics such as peak ground
acceleration (PGA), equivalent number of cycles (ENC) as a proxy for duration, and
frequency content. Six input motions representing a variety of PGA, ENC, and frequency
content are selected for numerical simulations. The seismic performance of the CT dam is
analyzed under co-seismic stage and then post-seismic stage to consider the volumetric
strains due to reconsolidation after each shaking event. Uniform models are also studied
and compared to the stochastic models to illustrate the necessity of stochastic modeling.
104
The ability of the uniform models to capture the extent of variability in system response is
evaluated.
5.1 Model Configuration
A typical upstream-construction CT dam was generated in the Fast Lagrangian
Analysis of Continua (FLAC Version 8) commercial platform, as shown in Figure 5-1. The
geometry approximately followed the geometry of Mochikochi tailings dam, discussed in
Byrne and Seid-Karbasi (2003). As reported by Rico et al. (2008), 45% of failed tailings
dams had height less than 15 m. Accordingly, the generated model was 90 m long and 15
m tall including a 3-m thick bedrock and 12-m thick CT behind a 3:1 (H:V) slope formed
by four dikes, each 3 m high. The meshing was implemented such that the spatial element
size was small, particularly in the vertical direction, to ensure proper wave transmission
through the model (Itasca 2017).
The dikes and bedrock properties were adopted from studies where the cyclic
behavior and seismic stability of CT dams were evaluated. The bedrock was assumed to be
an elastic and homogeneous material with a density of 2,400 kg/m3, a shear modulus of
860 MPa, and a Poisson’s ratio of 0.3 in all simulations. The dikes, which are typically
constructed with gravelly sand, were modeled using the Mohr-Coulomb elastoplastic
model. The density, cohesion, and friction angle of the dikes were selected to be 1,700
kg/m3, 10 kPa, and 35°, respectively, based on previous studies (Byrne and Seid-Karbasi
2003; Zeng et al. 2008; Ferdosi et al. 2015). The shear modulus of the dikes was pressure-
dependent with Poisson’s ratio of 0.3 and calculated based on the Hardin (1978) equation
105
that was developed for granular materials. The default hysteresis model in FLAC2D was
used among the built-in tangent modulus functions to define the shear modulus reduction
curves for the dikes (Itasca 2017). The input parameters for the default hysteresis model
were adopted from Zeng et al. (2008).
The spatial variability in geotechnical properties of CT is significant mainly due to
the depositional processes in the field. A uniform model for the CT may not accurately
represent the system response under loading events. In this study, CT were studied as
uniform and spatially variable materials, respectively. Since CT are composed of sand and
low plasticity silt and demonstrate cyclic behaviors that could be interpreted as either cyclic
liquefaction or cyclic mobility, both PM4Sand and PM4Silt could be considered in the
design and analysis. The applicability and calibration of both models for the CT are
presented in the next section.
The hydrostatic pore pressure was established through the model, and the CT were
assumed to be fully saturated and the toe of the bottom dike was the drainage zone. The
boundaries were extended sufficiently far from the failure zone to minimize the influence
of boundaries on the model response. A free-field boundary condition was assigned to the
side boundaries and a quiet boundary was considered at the bottom boundary in both the
horizontal and vertical directions during the dynamic analyses. The outcrop input motions
were applied in a form of shear stress time series at the base of the model using the
compliant-base procedure by Mejia and Dawson (2006). A Rayleigh damping of 0.5% at a
center frequency of 3 Hz was considered for the CT to account for low-strain damping
(Boulanger and Montgomery 2016). Only the CT were considered liquefiable and flow
was not permitted during the dynamic analyses due to the low permeability of CT (Salam
106
et al. 2019). The first column of the zones at the far left boundary was considered non-
liquefiable to avoid inaccurate free-field boundary calculations, as recommended by the
FLAC2D manual (Itasca 2017).
The seismic performance of the CT dam was evaluated in two stages, 1) during the
cyclic loading (i.e. co-seismic) and 2) after the cyclic loading (i.e. post-seismic). Co-
seismic analysis included the non-linear effective stress analysis during the motion. Post-
seismic analysis considered the excess pore pressure dissipation and effective stress
increase after the motion. Accordingly, the dynamic analysis was continued after each
shake to determine the volumetric strains due to reconsolidation. An empirical approach of
reducing elastic shear modulus is used in PM4Silt and PM4Sand to calculate the volumetric
strains during the reconsolidation process (Boulanger and Ziotopoulou 2017 and 2018).
Figure 5-1 Typical upstream-construction CT dam model generated in FLAC2D
5.1.1 PM4Sand and PM4Silt calibration based on CT cyclic response
To Both PM4Sand and PM4Silt models require three primary input parameters.
The contraction rate parameter (hpo), which estimates the plastic volumetric strain rate, is
the first primary input parameter and required in both models. hpo is a soil specific input
107
parameter and should be calibrated based on the relationship of cyclic stress ratio versus
the number of cycles to reach liquefaction, i.e., the CSR-N curve determined by laboratory
testing. Shear modulus coefficient, G0, is the second primary input parameter and required
in both models. The elastic shear modulus is determined by G0 .The remaining primary
input parameters, relative density (Dr) for PM4Sand, and undrained shear strength at
critical state under earthquake loading (su,cs,eq) for PM4Silt, are determined by either
empirical relationships or in-situ and laboratory tests. Undrained shear strength ratio
(su,cs,eq_Rat), which is su,cs,eq normalized by vertical effective stress, is used in this study
instead of su,cs,eq. In addition to the primary input parameters, there are eighteen and twenty
secondary input parameters defined in the PM4Sand and PM4Silt models, respectively.
To evaluate the applicability of PM4Sand and PM4Silt in simulating the cyclic
behavior of CT, the cyclic response of CT in cyclic direct simple shear (CDSS) tests
reported by Salam et al. (2019) was simulated by both models. The CT showed a
transitional cyclic behavior from clay-like to sand-like in Salam et al. (2019). Accordingly,
either cyclic liquefaction or cyclic mobility is likely to be observed in CT during cyclic
loading. This characteristic is due to the composition of CT (i.e. mixture of sand and silt)
and plasticity index less than or equal to 7 (Salam et al. 2019). Therefore, it is necessary to
examine the abilities of both PM4Sand and PM4Silt in capturing the cyclic behavior of
CT.
The primary input parameters, Dr, su,cs,eq_Rat, and Go of the sample in Salam et al.
(2019) were 50%, 0.25, and 160, respectively. The secondary input parameters retained
their default values. The contraction rate parameter, hpo, was calibrated for both models to
match the CSR determined using CDSS tests at 15 cycles. The effective vertical stress in
108
the CDSS tests and the numerical calibration was 60 kPa. Using single element simulations,
the hpo parameter was calibrated for CT to a value of 0.21 and 0.83 in PM4Sand and
PM4Silt, respectively. The CSR-N curve on log-scale can be expressed by a power law of
𝐶𝑆𝑅 = 𝑎 × (𝑁𝑓𝑎𝑖𝑙𝑢𝑟𝑒)−𝑏, where 𝑁𝑓𝑎𝑖𝑙𝑢𝑟𝑒 is defined as 5% double amplitude shear strain
(DAS) in the CDSS tests. The experimental and simulated CSR-N curves and the
corresponding equations are shown in Figure 5-2. The estimated b-value in the PM4Sand
and PM4Silt simulations was 0.25 and 0.23, respectively, while the b-value from the CDSS
test was 0.17. Considering the CSR-N curves, both PM4Sand and PM4Silt performed
similarly in estimating the liquefaction resistance of the CT. Both models approximated
higher cyclic resistance at large CSR values compared to the CDSS test results. For
example, the simulated FCR reached to failure at larger number of cycles at CSR of 0.15
compared to the CDSS test result. To further investigate the applicability of PM4Sand and
PM4Silt in approximating the cyclic response of the CT, the shear stress-strain loops, shear
strain accumulation, and pore pressure ratio from the experiments and the simulations were
compared.
109
Figure 5-2 Experimental and numerically simulated CSR-N curves for the studied CT
The simulated stress-strain loops by PM4Sand and PM4Silt are compared against
the CDSS test results and are shown in Figure 5-3 (a) and (b) for CSR of 0.12. According
to Figures 5-3 (a) and (b), the plastic behavior of the CT and the wide shear stress-strain
loops observed in the experiment were better approximated by PM4Silt than by PM4Sand.
However, both models estimated the 5% DAS occurrence at comparable number of cycles
(i.e. N10). Figures 5-3 (c) and (d) show the accumulation of shear strain with cycles
approximated by PM4Sand and PM4Silt. The soil element simulated by PM4Sand did not
accumulate large shear strains until the last cycle, where the sample suddenly reached 5%
DAS. The soil element simulated by PM4Silt experienced progressive accumulation of
shear strain until failure, similar to the laboratory observation. In addition, the excess pore
110
pressure ratios (ru) estimated by PM4Sand and PM4Silt along with the observed ru in the
CDSS test are shown in Figures 5-3 (e) and (f). The difference in estimating excess pore
pressure ratio by PM4Sand and PM4Silt was small since the final ru was approximated as
0.8~0.9 by both PM4Sand and PM4Silt. The estimated ru by PM4Sand was found to
approximate the lower bound of ru from CDSS test before the last cycle, as shown in Figure
5-3 (e). According to Figure 5-3 (f), the trend of pore pressure ratio with the cycles was
slightly better approximated by the PM4Silt model. The transitional behavior of CT
between clay-like and sand-like behavior observed by Salam et al. (2019) was further
confirmed by noticing insignificant differences between PM4Sand and PM4Silt calibration
results. However, PM4Silt showed better approximation specifically in terms of strain
accumulation and cyclic mobility (i.e. softening) of the tested CT. In addition to the
calibration results, a shear strength related index (Su,cs,eq_Rat in PM4Silt) may better
represent the behavior and consistency of fine-grained material such as coal tailings
compared to relative density (Dr in PM4Sand). Therefore, PM4Silt was selected to model
the CT in the seismic stability simulations.
111
Figure 5-3 Cyclic responses of CT from cyclic DSS test and simulations at CSR=0.12
112
5.1.2 Random fields generation for CT
Among the three primary input parameters (su,cs,eq_Rat, Go, and hpo) that are required
to model the CT using PM4Silt, su,cs,eq_Rat was modeled as a spatially correlated Gaussian
random field. Random field representation approach has been adopted in several other
geotechnical engineering applications (e.g. Fenton and Griffiths 2003; Montgomery and
Boulanger 2016; Boulanger et al. 2019). Random fields are defined by a probability
distribution, including mean and standard deviation, and auto-correlation functions based
on available data. An auto-correlation function states the distance in vertical and horizontal
directions, within which soil properties are correlated. The Karhunen-Loeve (K-L)
expansion method was adopted to generate and discretize the random fields, as described
in Phoon and Ching (2014) and Equation 5-1:
𝑅(𝑥, 𝑦, 𝜃) ≈ 𝜇 + ∑ √𝜆𝑖Ф𝑖(𝑥, 𝑦)𝜉𝑖(𝜃)
𝑀
𝑖=1
(𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5 − 1)
where x and y are the coordinates of the points in the space, θ denotes the stochastic
characteristic of the random field such that 𝜉𝑖(𝜃) are uncorrelated standard random
variables with zero mean and unit standard deviation. 𝜆𝑖 and Ф𝑖 are the eigenvalues and
eigenfunctions, respectively, which are determined from the covariance function. 𝑀 is the
truncation order of the expansion series and determines the accuracy and smoothness of
the generated random field. The undrained shear strength of tailings and similar soils such
as silty alluvial soils reported in the literature (e.g., Ladd and Foott 1974; Phoon et al. 1995,
Phoon and Kulhawy 1999, Olson and Stark 2002, Castro 2003, Hegazy et al. 2004,
Robertson 2009, Kalinski and Salehian 2016, Salam et al. 2019, Yu et al. 2019) were used
113
to establish the probability distribution for su,cs,eq_Rat. Accordingly, a lognormal distribution
with mean value (μ) of 0.2 and coefficient of variation (COV) of 20% was found the best
estimate for su,cs,eq_Rat. An exponential autocorrelation function was also adopted for the
CT properties, the horizontal (lx) and vertical (ly) autocorrelation lengths were assumed to
be 15 m and 1.5 m, respectively (Ji et al. 2012). The series was terminated at M=10, where
sufficient accuracy and smoothness were achieved for the distribution of su,cs,eq_Rat within
the random fields.
The shear modulus coefficient, Go, was correlated to su,cs,eq_Rat by the equation
proposed by Dickenson (1994) with a slight adjustment to represent the CT shear modulus
(Equation 5-2). The main equation was developed for cohesive soils in the San Francisco
Bay Area with a constant factor equal to 23. However, the constant factor was scaled up to
28 to fit the available data for the shear modulus of the tested CT by Salam et al. (2019).
𝐺𝑚𝑎𝑥 = 𝜌 ∙ (28 ∙ (𝑠𝑢,𝑐𝑠,𝑒𝑞𝑅𝑎𝑡∙ 𝜎′
𝑣)0.475
)2
(𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5 − 2)
where 𝜌 is total density and 𝜎′𝑣 is vertical effective stress. Keeping hpo constant,
the CSR versus number of cycles to reach 5% shear strain for the CT was simulated for
three values of su,cs,eq_Rat (i.e. 0.15, 0.2, and 0.25) using PM4Silt. Figure 5-4 shows the
increasing trend in cyclic resistance of CT due to increase in su,cs,eq_Rat. For example, the
required number of cycles to reach 5% shear strain increased from approximately 8 to 25,
when su,cs,eq_Rat of CT increased from 0.15 to 0.25. Figure 5-4 signifies the necessity of
sensitivity analysis and stochastic modeling for the seismic stability of CT dams.
114
Figure 5-4 CSR versus number of cycles to reach 5% shear strain for CT (PM4Silt
simulations)
In order to stochastically evaluate seismic stability of the model, 66 realizations for
the CT section in the model were selected. The Latin Hypercube Sampling (LHS) method
was adopted to select the representative realizations (Betz et al. 2014). Figure 5-5 presents
four realizations (A, B, C, and D) out of the 66 selected realizations. The su,cs,eq_Rat range
varies among the realizations. For example, the maximum values for su,cs,eq_Rat are 0.25,
0.35, 0.5, and 0.3 in Realization A to D, respectively. As shown in Figure 5-5, the
variability of su,cs,eq_Rat forms extremely strong and weak pockets within the tailings, which
could contribute to co-seismic or post-seismic failure of the CT dam. The su,cs,eq_Rat
variation through depth and horizontal distance is also shown in Figures 5-5. The solid line
115
shows the average su,cs,eq_Rat, which fluctuates around the set average value (i.e. 0.2) in both
vertical and horizontal directions. Figure 5-6 shows the histogram of su,cs,eq_Rat for the four
realizations in Figure 5-5; it can be seen that they approximately have lognormal
distributions with a mean of 0.2, as defined during generating the random fields. Three
uniform models were also generated with su,cs,eq_Rat of 0.15 (lower bound), 0.2 (best
estimate), and 0.25 (upper bound). su,cs,eq_Rat = 0.15 was selected to represent a weak CT
dam; and su,cs,eq_Rat = 0.25 was selected to represent a strong CT dam.
116
117
Figure 5-5 su,cs,eq_Rat variation in Realizations A, B, C, and D
Figure 5-6 Histograms of su,cs,eq_Rat of CT for Realizations A, B, C, and D
5.1.3 Input motions and analysis approach
Six input motions were selected to investigate the effect of PGA, ENC as a proxy
for duration, and frequency content on seismic stability of the CT dam. Figure 5-7 presents
118
the acceleration time histories and the response spectra of the input motions. In order to
investigate the effect of PGA, the 2011 Mineral Virginia Earthquake (Mw = 5.8), a shallow
crustal event recorded at the Corbin station, was selected. The event was scaled to bedrock
outcrop PGAs of 0.24g, 0.37g, and 0.5g and referred to as EQ1, EQ2, and EQ3,
respectively. The bracketed duration (D5-95) of the Mineral Virginia Earthquake was
approximately 20 seconds. Accordingly, to reduce the simulation cost, only 20 seconds of
the motion, the significant duration, was used for the dynamic analysis. The response
spectra of EQ1, EQ2, and EQ3 in Figure 5-7 reflect the 20 seconds motion. A blast motion
(provided by Vibra Tech, Inc., Hazleton, PA), as a common cyclic loading around mine
sites and tailings dams, was adopted. The blast motion (denoted as B1) had a duration of 5
seconds and was scaled to an outcrop PGA of 0.24g using a scaling factor equal to 4.
Accordingly, the effect of frequency content and duration could be studied by comparing
the system response under EQ1 and B1.
The 1940 El Centro Earthquake (Mw = 6.9) recorded at the El Centro Array 9 station
and the 1992 Landers Earthquake (Mw = 7.3) recorded at the Yermo fire station both with
scaled outcrop PGA of 0.24g were selected and are referred to as EQ4 and EQ5,
respectively. The scaling factors for these motions were 0.75 and 0.96. The bracketed
duration of EQ4 and EQ5 was 24.3 seconds, and 18.9 seconds, respectively. These input
motions were chosen to represent earthquakes with longer duration that imposes larger
number of cycles on the CT dam. The effect of duration and frequency content could be
studied by investigating the system response under EQ1, EQ4, and EQ5, which have the
same PGA.
119
Figure 5-7 Selected input motions for CT dam seismic stability analyses
ENC was adopted as a proxy for duration in this study. The ENC of each input
motion was determined according to the criteria discussed by Verma et al. (2018). The
CSR of each input motion was also calculated based on the simplified approach by Seed
and Idriss (1971). Figure 5-8 presents the CSR and ENC of the selected input motions. The
ENC of the input motions of EQ1, EQ2, EQ3 and B1 is approximately 11. The ENC of
EQ4 and EQ5 is 30 and 38, respectively. The CDSS test results of the tested samples are
also shown in Figure 5-8 as a reference. Since the constitutive model was calibrated to
capture the CDSS results, this figure implies that for all input motions in this study, the soil
element is expected to liquefy. The dynamic response of the dam is more complex as
liquefaction at deeper depths could change the propagation of motions throughout the soil
120
profile and inevitably change the cyclic shear stress that the shallower soil elements are
subjected to. The effective-stress dynamic 2D analysis that is presented in the next section
enables us to look into the complex dynamic response of the dam subjected to soil
liquefaction.
Figure 5-8 ENC and maximum CSR of the input motions and the tested CT CSR-N curve
The effect of frequency content on the coal tailings dam’s seismic response can be
illuminated by investigating the input motion and system response spectra. Significant
vibration and deformations are likely to occur when input motion and system response
spectra are in tune with each other such that natural periods of the input motion are similar
to those of the system. The response spectra of the coal tailings varied among the
realizations due to stochastic modeling with varying shear modulus. To investigate this
121
source of variability, the natural periods of the selected realizations and the uniform model
with su,cs,eq_Rat = 0.2 were determined following the “sum of sines” approach (Chakraborty
et al. 2019). Figure 5-9 shows the acceleration response spectra at the crest for Realizations
A, B, C, D, and the uniform model. The discrepancy between the response spectra of the
realizations and the uniform model indicates that the uniform model is unable to capture
the variability in response. Realization A shows significantly larger spectral acceleration
at small periods compared to that of the uniform model and other realizations. In contrast,
Realization B, C, and D show smaller spectral acceleration at small periods in comparison
with that of the uniform model. The difference between the envelope response spectra of
the uniform model and the selected four realizations is considered small at periods longer
than 0.4 second. The peaks of the response spectral acceleration are within periods of 0.06
to 0.4 second. Figure 5-9 emphasizes the importance of stochastic modeling to capture the
uncertainty in seismic response of a coal tailings dam.
122
Figure 5-9 Crest acceleration response spectra for Realizations A, B, C, and D and the
uniform model with su,cs,eq_Rat = 0.2
5.2 Model Results and Discussion
5.2.1 Representative dynamic responses
Figures 5-10 and 5-11 show the co-seismic performance of the CT dam in terms of
excess pore pressure and shear strain contours under the input motion EQ2, as an example.
Figure 5-10 shows the results of the uniform model with su,cs,eq_Rat = 0.2, and Figure 5-11
shows the results of the stochastic model with su,cs,eq_Rat ranging from 0.1 to 0.5. The
maximum excess pore pressures generated during EQ2 were equal in both the uniform
123
model and the stochastic model. However, in the stochastic model larger area in the vicinity
of the dikes experienced residual excess pore pressure. Generation of excess pore pressure
leads to softening and consequently large deformation of the CT dam. Accordingly,
although the shear band and deformation pattern were similar in Figures 5-10 and 5-11, the
shear strain developed in the stochastic model was larger than that of the uniform model.
In addition, the residual excess pore pressure may result in post-seismic failure, which is a
common concern for CT dams. Accordingly, the larger area with residual excess pore
pressure in the stochastic model implied higher risk of post-seismic failure.
Figure 5-10 Co-seismic performance of the CT dam in terms of excess pore pressure and
shear strain contours in a uniform model with su,cs,eq_Rat = 0.2 (The unit of excess pore
pressure is Pa)
124
Figure 5-11 Co-seismic performance of the CT dam in terms of excess pore pressure and
shear strain contours in a stochastic model with su,cs,eq_Rat ranging from 0.1 to 0.5 (The
unit of excess pore pressure is Pa)
5.2.2 Dynamic responses of uniform models
Figure 5-12 presents the co-seismic and post-seismic crest settlements of the
uniform models under the selected input motions except for B1. The CT dam was found to
be safe under B1 input motion for both the stochastic and the uniform models, because of
the short duration of blast loading and the weak acceleration response spectra. The B1 input
motion resulted in combined co-seismic and post-seismic crest settlement less than 0.05 m,
which was significantly smaller than the crest settlements observed under the earthquake
input motions. Therefore, the results of this input motion were excluded from further
analyses.
In this study, complete failure was assumed if the crest settlement exceeded 3 m,
which is equal to height of a dike. Therefore, crest settlement larger than 3 m is not shown
125
in Figure 5-12. In general, the uniform models with higher su,cs,eq_Rat experienced smaller
crest settlement. Figure 5-12 (a) shows the effect of PGA (by comparing the results from
EQ1 to EQ3) on the crest settlement of the uniform models. The uniform model with
su,cs,eq_Rat of 0.25 showed the smallest crest settlement with post-seismic crest settlement
less than 10 mm under the input motions. The co-seismic crest settlement increased from
0.281 m to 0.663 m when the PGA increased from 0.24g to 0.5g in the uniform model with
su,cs,eq_Rat = 0.25. The uniform model with su,cs,eq_Rat = 0.2 experienced co-seismic crest
settlement of 0.464 m, 0.946 m, 1.07 m under EQ1 to EQ3. The post-seismic crest
settlement of the uniform model with su,cs,eq_Rat = 0.2 noticeably increased as PGA
increased. The uniform model with su,cs,eq_Rat = 0.15 failed under EQ2 and EQ3. The co-
seismic and post-seismic crest settlements of the uniform model with su,cs,eq_Rat = 0.15
under EQ1 were 0.75 m and 0.925 m, respectively.
Figure 5-12 (b) shows the effects of ENC and frequency content on crest settlement.
The increase of ENC from 11 to 38 (by comparing results from EQ1, EQ4, and EQ5)
increased the crest settlement from 0.281 m to 1.03 m in the uniform models. The
increasing trend in the crest settlement could also be due to richer response spectra of EQ5,
which had higher acceleration in a wider range of periods compared to those of EQ1 and
EQ4. The post-seismic crest settlement of the uniform model with su,cs,eq_Rat = 0.25 was
less than 10 mm. According to Figure 5-12, significant additional settlement up to failure
was observed for the uniform model with su,cs,eq_Rat = 0.2 during post-seismic analysis. The
crest settlement of the uniform model with su,cs,eq_Rat = 0.2 exceeded 3 m (i.e. failure)
during the post-seismic and co-seismic analysis of EQ4 and EQ5, respectively. The
uniform model with su,cs,eq_Rat = 0.15 showed crest settlement beyond 3 m (i.e. failure)
126
during co-seismic analysis under EQ4 and EQ5. Overall, the input motions with larger
ENC and richer response spectra were found causing larger crest settlement.
Figure 5-12 Co-seismic and post-seismic crest settlements of the uniform models (a)
PGA effect (b) ENC effect
127
5.2.3 Post-seismic analysis significance
The stochastic modeling revealed how the spatial variability of su,cs,eq_Rat within the
CT section affects the co-seismic and post-seismic crest settlements. Figure 5-13
demonstrates the co-seismic and post-seismic crest settlements under the input motion EQ2
for the uniform model with su,cs,eq_Rat = 0.2, Realizations A, B, C, D, and four other
realizations (named E, F, G, and H). Figure 5-13 shows the crest settlement of the uniform
model with a solid line; the co-seismic and post-seismic crest settlements were 0.946 m
and 1.306 m, respectively. The results of Realizations A to D are shown with dashed lines
in Figure 5-13. The extent of variability in both co-seismic and post-seismic settlements of
the CT dam with spatially variable su,cs,eq_Rat can be seen in Figure 5-13. While Realization
B showed comparable crest settlement to the uniform model, Realizations A, C, and D had
significantly different results. Realization C experienced small co-seismic and post-seismic
crest settlements of 0.588 m and 0.671 m, respectively. Realization A showed larger co-
seismic crest settlement of 1.217 m and the CT dam failed during post-seismic analysis.
Realization D was found to be the most vulnerable, as the failure occurred during co-
seismic analysis.
The results of Realizations E to H, shown in dotted lines in Figure 5-13, were
included to demonstrate that the co-seismic performance alone may not accurately
represent the dynamic deformations after cyclic loading. Therefore, post-seismic analysis
is necessary in order to characterize the overall deformation of the CT dam. This finding
is consistent with the findings of other studies using numerical simulations (e.g. Naesgaard
and Byrne 2007). Realizations E, F, and G exhibited a co-seismic crest settlement of around
128
0.764 m, but the overall crest settlement ranged from 0.842 m to 1.570 m. In another
example, although Realization H showed similar co-seismic crest settlement to that of
Realization A, complete failure was not observed and the final crest settlement was 2.970
m in Realization H. Considering the variability presented in Figure 5-13, co-seismic and
post-seismic settlements of the CT dam are separately discussed under the input motions
in Figures 5-14 and 5-15.
Figure 5-13 Co-seismic and post-seismic performances of select models under EQ2
5.2.4 Dynamic response of stochastic models (co-seismic)
The variation of the co-seismic crest settlement when CT properties are spatially
variable is shown in Figure 5-14. Figure 5-14 includes two subfigures to separately present
129
the influences of PGA, ENC and frequency content on the variation of the co-seismic crest
settlement. The co-seismic crest settlements of the uniform models are also shown in
Figure 5-14. The normalized settlement (NS) is defined as the crest settlement divided by
the dam’s height (12 m). The realizations that resulted in complete failure (NS>25%) are
excluded from Figure 5-14 to have better resolution for the rest of the realizations. Four
levels of crest settlements were considered in this study to evaluate the performance of the
CT dam subjected to liquefaction based on NS: stable (NS ≤ 5%), moderate damage (5%
< NS ≤ 10%), severe damage (10% < NS ≤ 25%), and failure (NS > 25%). Figure 5-14 also
shows the percentages of the realizations in each category under each input motion.
As shown in Figure 5-14 (a), the CT dam was found to be stable under EQ1 for all
the realizations. EQ1 caused larger crest settlement (0.447 m on average) compared to that
of B1 (below 0.050 m). The PGA and ENC of EQ1 and B1 were the same. Therefore, this
observation could be mainly due to the richer acceleration response spectra of EQ1, which
showed higher acceleration in a wider range of periods compared to B1. The majority of
the realizations under EQ2 experienced moderate damage. EQ2 caused failure in 17% of
the realizations and the crest settlements of the remaining realizations ranged from 0.588
m to 2.850 m. Approximately half of the realizations showed larger crest settlement than
that of the uniform model with su,cs,eq_Rat = 0.2. The input motion EQ3 resulted in similar
observations but more realizations (i.e. 23%) failed due to the higher PGA of this input
motion. More than half of the realizations experienced larger crest settlement than that of
the uniform model with su,cs,eq_Rat = 0.2. Overall, increasing the PGA from 0.24g to 0.5g
(i.e., from EQ1 to EQ3) increased failure probability. In addition, the discrepancy of the
stochastic models’ response from the response of the uniform model with su,cs,eq_Rat = 0.2
130
became more significant as PGA increased. All the stochastic models’ results were
enveloped by the results of the uniform models with su,cs,eq_Rat = 0.15 and su,cs,eq_Rat = 0.25.
131
Figure 5-14 Summary of co-seismic crest settlement for stochastic models (a) PGA effect
(b) ENC and frequency content effect
132
Figure 5-14 (b) shows the effect of ENC and frequency content by comparing the
results under EQ1, EQ4, and EQ5. EQ4 resulted in failure of 36% of the realizations, and
the remaining experienced severe damage. More than half of the realizations showed crest
settlement larger than that of the uniform model with su,cs,eq_Rat = 0.2. EQ5, which had the
largest ENC and magnitude among the input motions, caused failure in 69% of the
realizations as well as failure in the uniform models with su,cs,eq_Rat of 0.2 and 0.15.
However, 31% of the realizations under EQ5 showed NS ≤ 25% and experienced less crest
settlement compared with EQ4. This can be attributed to the magnitude and frequency of
the peaks in the acceleration spectra of EQ4 and EQ5 and their interaction with the natural
frequencies of the stochastic models. This observation emphasized the necessity of
stochastic modeling and frequency content analysis in seismic stability evaluation. The
stochastic models’ results were enveloped by the uniform models with su,cs,eq_Rat of 0.15
and 0.25. However, the majority of the results were within the range observed for the
uniform models with su,cs,eq_Rat of 0.2 to 0.15 (i.e. best estimate to lower bound), as shown
in Figure 5-14 (b).
5.2.5 Dynamic response of stochastic models (post-seismic)
Figure 5-15 presents the final crest settlements after post-seismic analysis for the
stochastic and the uniform models. The incremental crest settlement that occurred during
post-seismic analysis of EQ1 was negligible and less than 3% of the co-seismic crest
settlement for both the stochastic and uniform models. As shown in Figure 5-15 (a), the
stochastic models showed additional settlement and higher probability of failure in post-
133
seismic analysis. Under EQ2, approximately 75% of the realizations showed larger final
crest settlement compared to that of the uniform model with su,cs,eq_Rat = 0.2. Under EQ3,
the uniform models with su,cs,eq_Rat of 0.2 to 0.15 and 74% of the realizations failed.
According to Figure 5-15 (b), EQ4 was found to be the most destructive input
motion among the input motions since 94% of the realizations failed. The higher post-
seismic failure rate observed for EQ4 compared to EQ5, despite the smaller ENC and
magnitude of EQ4, signified the dependency of the seismic response on the interplay
between the system natural period and input motion acceleration spectra. Accordingly, the
input motion EQ4 acceleration spectra are found to be in tune with larger number of
realizations and resulted in higher failure probability. This observation indicated the
significance of all indices such that only one characteristic (e.g. ENC) may not be enough
to predict the seismic performance of the CT dam.
134
Figure 5-15 Summary of post-seismic crest settlement for stochastic models (a) PGA
effect (b) ENC and frequency content effect
135
5.2.6 Failure probability analysis
The probabilistic co-seismic performance of the CT dam was evaluated by the
Subset Simulation (SS) method, which is an adaptive simulation approach in engineering
systems. SS adopts conditional probability and Markov chain Monte Carlo (MCMC)
method to efficiently compute failure probability of the system (Au and Beck 2001; Au
and Wang 2014). The failure event is defined as 𝐹 = {𝒙: 𝐺(𝒙) < 𝟎}, where 𝐺(𝒙) is the
performance function of the standard Gaussian random variables 𝒙 = (𝑥1, … , 𝑥10), as 10
random variables were used to generate the random fields in this study. The performance
function,𝐺(𝒙), is defined as the maximum allowable system response (i.e. crest settlement)
minus the actual system response. Negative values of performance function indicate failure
of the system. The failure probability (𝑃𝐹) is calculated as a product of intermediate failure
events {F1, F2, …, Fm}, which have larger conditional probability, as shown in Equations
5-3 and 5-4.
𝐹1 ⊃ 𝐹2 ⊃ ⋯ ⊃ 𝐹𝑚 = 𝐹 𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5 − 3
𝑃𝐹 = 𝑃(𝐹1) ∏ 𝑃(𝐹𝑖+1|𝐹𝑖)
𝑛=𝑚−1
𝑛=1
𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5 − 4
Therefore, simulating failure events within the original probability space is
replaced by a sequence of simulations of more frequent events in the conditional
probability spaces. A surrogate model is developed from the numerical simulation results
in the previous sections for each input motion. Then, the surrogate model is used to
determine the system response in each level of SS analysis. The SS analysis begins by
generating a primary pool (N) of vectors with 10 random variables, which define N random
136
fields (i.e. realizations for the model). The performance function output are first determined
for each sample and then sorted in ascending order in a vector G0. The failure threshold in
the first level of SS analysis is equal to 𝐶1 = (𝑁 × 𝑃0 + 1) th value in G0, P0 is the
prescribed failure probability and assumed equal to 10% (Phoon and Ching 2014). The
random variable vectors of the failed simulation cases are used to generate the new pool of
vectors (N). The Modified Metropolis-Hastings (MMH) algorithm of MCMC (Santoso et
al. 2011) is used to generate the vectors in the next level of SS analysis. The performance
function values in the second level are sorted in increasing order and stored in a vector G1,
the new failure threshold (𝐶2) is the (𝑁 × 𝑃0 + 1)th value in G1. The conditional subsets of
these intermediate events are generated until the failure threshold (𝐶𝑚) is negative. The
conditional failure probability associated with the last level is computed by Equation 5-5,
the indicator function 𝐼𝐹𝑚 is 1.0 if the performance function is negative and zero otherwise.
𝑃(𝐹𝑚|𝐹𝑚−1) =1
𝑁∑ 𝐼𝐹𝑚
(𝒙𝑛(𝑚−1))
𝑁
𝑛=1
𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5 − 5
To be comprehensive, the co-seismic failure probability of the studied CT dam
under the earthquake input motions is presented for several failure limits including NS
exceeding 5%, 10%, and 25% (i.e. moderate damage, severe damage, and complete
failure). The number of samples evaluated at each level of SS analysis is critical to reach
small values of coefficient of variation (COV) for failure probability. Accordingly, N=3000
samples per SS analysis level was found reasonable to achieve COV less than 5% for the
failure probability value. Figure 5-16 (a) and (b) show the SS analysis results revealing the
effects of PGA, and ENC and frequency content, respectively. The CT dam’s co-seismic
failure probability was found significantly small and close to zero under EQ1.
137
(a) PGA effect
(b) ENC and frequency content effect
Figure 5-16 Probabilistic co-seismic performance of the CT dam under the earthquake
input motions
Figure 5-16 (a) presents the increase in co-seismic failure probability due to
increase in PGA at the same ENC for the three different failure criteria. Assuming 5% NS
as the failure criterion, the co-seismic failure probability of the CT dam increased from
zero to 98.1% and 99.8% when PGA increased from 0.24g to 0.37g and 0.5g, respectively.
The significant growth in failure probability indicated a high sensitivity to PGA at 5% NS
138
as the failure criterion. In the case of 10% NS as the failure criterion, the co-seismic failure
probability increased from zero to 51.1% and 65% when PGA increased from 0.24g to
0.37g and 0.5g, respectively. The PGA showed strong effect on the co-seismic failure
probability at 10% NS as the failure probability exceeded 50%. The CT dam was found to
be relatively stable under EQ1, EQ2, and EQ3 at 25% NS as the failure criterion, as the
failure probabilities were 0%, 8.8%, and 12.4%, respectively.
Figure 5-16 (b) shows the effect of ENC and frequency content on the co-seismic
failure probability for the three different failure criteria. At 5% NS failure criterion, the co-
seismic failure probability is near 100% when ENC increased from 11 to 30 and 38,
indicating the significant destructive effect of high ENC on the stability of the studied CT
dam. The destructive effect of high ENC was similarly noticed at 10% NS as the failure
criterion. At 25% NS failure criterion, the co-seismic failure probability increased from
zero to 45.8% and 53% when ENC increased from 11 to 30 and 38, respectively. Similar
failure probabilities for EQ4 and EQ5, despite the difference in ENC, highlighted the effect
frequency content on the co-seismic performance of the CT dam.
Overall, it was noticed that high ENC and rich frequency content are more
destructive than high PGA. For example, considering EQ1 with PGA of 0.24g and ENC of
11 as the reference motion, the rate of increase in co-seismic failure probability due to
increase in PGA was significantly slower than when ENC increased and frequency content
became richer. This difference was more pronounced at larger NS failure criteria (e.g. 25%).
This observation was in agreement with earlier findings on CT, which was detected as
susceptible to cyclic softening and progressive shear strain accumulation under cyclic
loading.
139
5.2.7 Implications in practice
The significance of PGA, ENC, and frequency content in the seismic performance
of the CT dam were assessed by paired t-test. The mean responses of the realizations under
two input motions are compared in the paired t-test. For example, a statistically significant
mean differences between EQ1 and EQ2, EQ1 and EQ3, EQ2 and EQ3 revealed the
significance of PGA in the seismic performance of the CT dam. Conducting paired t-test
on appropriate set of data, all the studied input motion (PGA, ENC, and frequency content)
characteristics were found statistically significant.
To reveal the significant uncertainty and variation in seismic response due to
variability in su,cs,eq_Rat, the one-sample t-test procedure was adopted. This approach
determines whether the average response (i.e. crest settlement) obtained from stochastic
models is significantly different from the crest settlement from the uniform model with
su,cs,eq_Rat = 0.2 (i.e. best estimate). This procedure was conducted for each earthquake input
motion. The one-sample t-test was conducted once for co-seismic crest settlements and
once for post-seismic settlements. The one-sample t-test showed that the mean response of
the stochastic models is statistically significantly different from the response of the uniform
model with su,cs,eq_Rat = 0.2 (i.e. best estimate), for both co-seismic and post-seismic
settlements.
Therefore, the uniform model with best estimate (su,cs,eq_Rat = 0.2) cannot properly
capture the uncertainty in response caused by heterogeneity in subsurface condition.
Furthermore, although the stochastic results were enveloped by the results of the uniform
models with the lower bound (su,cs,eq_Rat = 0.15) and the upper bound (su,cs,eq_Rat = 0.25)
140
properties, the probability of failure could not be estimated. For example, Figure 5-14 (b)
shows that the uniform model with lower bound properties (su,cs,eq_Rat = 0.15) failed under
EQ4, while the majority of the realizations (64%) did not experience failure. Therefore,
uniform modeling can lead to conservative results. Stochastic modeling on the other hand
can be more efficient and used to perform a probabilistic analysis on seismic stability of
CT dams.
5.3 Conclusions and summary
In this chapter, seismic stability of a typical upstream-construction CT dam was
investigated considering the spatial variability in geotechnical properties of CT under six
cyclic loadings. The cyclic behavior of CT was first approximated by PM4Sand and
PM4Silt using the primary input parameters. PM4Silt was evaluated to better approximate
the cyclic mobility and progressive shear strain accumulation in CT under cyclic loading.
Among the primary input parameters, the undrained shear strength ratio (su,cs,eq_Rat) was
modeled as a spatially correlated Gaussian random field. The effects of variability in CT’s
geotechnical properties and input motion characteristics (i.e. PGA, ENC, and frequency
content) on the seismic stability of the CT dam were assessed. Uniform models with three
different values, lower bound, best estimate, and upper bound for su,cs,eq_Rat were also
studied under the selected input motions.
Among the uniform models, only the model with su,cs,eq_Rat = 0.2 (i.e. best estimate)
showed the necessity of post-seismic analysis, as the stability status changed during the
post-seismic analysis. Post-seismic analysis was found critical for the stochastic models as
141
failure probability significantly increased. The significance of stochastic modeling was
statistically proved by comparing the results of the stochastic models and the uniform
model with su,cs,eq_Rat = 0.2 under the input motions. The discrepancy between stochastic
and uniform modeling was intensified under stronger input motions. The majority of
stochastic models experienced larger settlement than the uniform model with su,cs,eq_Rat =
0.2. However, the range of stochastic results was captured by the uniform models with
lower and upper bound values for su,cs,eq_Rat (i.e. 0.15 and 0.25). Stochastic modeling was
found superior to uniform modeling as probabilistic analysis can be conducted.
This study highlighted the importance of stochastic modeling and the consideration
of spatial variability in seismic stability analysis of CT dams. More investigations for
different geometries, seismic demands, statistical characteristics of the random fields, and
autocorrelation lengths are necessary, so that the findings of this study and the effects of
PGA, ENC, frequency content, and other potential characteristics on the seismic response
of CT dams can be further confirmed.
Chapter 6 Summary and Conclusions
6.1 Summary
This dissertation presented the findings on cyclic behavior of fine coal refuse (FCR)
using in-situ and laboratory tests. During the field investigation in two Appalachian
coalfields, several SPT tests were conducted and representative samples were transferred
to laboratory for further testing and characterization. Basic geotechnical properties of FCR
such as Atterberg Limits, sieve analysis, specific gravity, density, and hydraulic
conductivity were determined using undisturbed samples. Staged triaxial test approach was
followed to determine shear strength properties of FCR for short term and long term
loading scenarios. Shear stiffness properties were also assessed using torsional resonant
column testing under various confining stresses. Consequently, maximum shear modulus
and damping of FCR were calculated. The liquefaction susceptibility of FCR samples was
first evaluated by empirical approaches. Then, cyclic direct simple shear tests were
conducted on reconstituted samples to measure cyclic resistance of the FCR. Slurry
deposition method was adopted to prepare samples with comparable fabric and structure
to in situ FCR. Cyclic and post-cyclic behavior of FCR was studied. Cyclic tests were used
to establish the CSR-N curve of the FCR. The cyclic tests were followed by static shearing
up to 30% strain to investigate post-cyclic shear properties of liquefied FCR.
Shake table testing was conducted to overcome the limitations of element testing
and to investigate the effect of strain history and short-period aging on the cyclic resistance
143
and liquefaction behavior of FCR. A FCR specimen was slurry deposited into a laminar
shear box (LSB) to mimic the depositional process in the field. The specimen was
instrumented with piezometers and LVDTs to study the seismic response of the specimen
during and after each shake. Three shaking events were imposed to the FCR specimen to
study the effect of strain history. The FCR specimen was allowed to age for 97 days after
the second shake. The strength evolution within the FCR sample before and after each
shake, and during the aging period was measured by piezo-cone penetration testing
(CPTu). The cone resistance was used to calculate cyclic resistance and strength gain over
time.
The seismic stability of upstream-construction coal tailings dams was also assessed
by numerical modeling. The applicability of PM4Sand and PM4Silt to approximate the
cyclic behavior of FCR was evaluated. PM4Silt was selected and calibrated for FCR to
investigate the seismic stability of an upstream-construction coal tailings dam under six
input motions. The influence of input motion characteristics — peak ground acceleration
(PGA), equivalent number of cycles (ENC), and frequency content — on the seismic
response was investigated. The heterogeneity of FCR deposit in the field was also
represented by modeling the coal tailings section by random fields. Numerical simulations
were conducted for a representative number of realizations for the coal tailings section in
the finite difference model. The dynamic analyses were performed in co-seismic and post-
seismic stages. Uniform models with three values, lower bound, best estimate, and upper
bound for undrained shear strength of coal tailings were also analyzed under the input
motions. The necessity of stochastic modeling was studied by comparing the results of the
144
stochastic and uniform models. The importance of PGA, ENC, and frequency content in
interpreting the seismic performance of the coal tailings dam was assessed.
6.2 Conclusions
The following conclusions are based on the data, analysis and interpretation
presented in this dissertation.
(1) The FCR was found saturated and loose with high void ratio in the field. The
FCR specific gravity was also low due to its carbon content. The SPT numbers ((N1)60) in
the field were approximately from 3 to 7.
(2) FCR was classified as either silty sand or sandy silt with low plasticity index.
(3) The shear strength and shear stiffness properties were variable and highly
dependent on depth and location of the FCR sample in the field. The samples taken from
deeper depth and close to the discharge point showed the highest strength and stiffness.
(4) The FCR samples showed strain hardening behavior under static shear.
(5) The FCR was found to be liquefaction susceptible based on empirical
approaches. Reconstituted FCR samples showed a transitional behavior from sand-like to
clay-like under cyclic loading.
(6) Post-cyclic strength of FCR was significantly low but started to recover at large
shear strains.
(7) The FCR used in the shake table experiment was classified as either silty sand
or sandy silt with average plasticity index of 7.
145
(8) The CPTu test results on the slurry deposited FCR classified the specimen as a
fine-grained sensitive silty clay to clayey silt with significantly low cone resistance.
(9) The type of liquefaction observed for FCR in the shake table experiment was
cyclic mobility, during which limited excess pore pressure and progressive shear strain
accumulation were occurred.
(10) Heterogeneity of the FCR specimen caused localized excess pore pressure
generation during the shake. Sand boils were observed on the surface after the shakes and
further confirmed the occurrence of cyclic mobility in the FCR specimen.
(11) The strain history resulted in densification and increase of the cyclic resistance
of the FCR specimen. Although the cone resistant showed minor reduction immediately
after each shake, it improved and reached beyond the initial value after reconsolidation.
(12) Aging effect on the cyclic behavior and cone resistant of the FCR specimen
was significant such that over 100% strength gain was obtained in 97 days.
(13) The aging rate and strength range of the FCR specimen was lower than those
of clean sands. This observation was attributed to abundance of fines content in the FCR
specimen.
(14) The plastic behavior and progressive shear strain accumulation of the studied
FCR was better approximated by PM4Silt plasticity model.
(15) The post-seismic analysis was found to be critical in seismic stability
evaluation of coal tailings dams.
(16) The co-seismic deformation and performance of coal tailings dam was not
sufficient to predict the post-seismic performance and overall deformations.
146
(17) Although the extent of stochastic results were captured by the uniform models
with lower and upper bound properties, the failure probability could not be estimated.
(18) Stronger input motions caused more divergence of the stochastic results from
the results of the uniform model with the best estimate properties.
(19) All studied input motion characteristics (i.e. PGA, ENC, and frequency
content) were found significant and necessary to understand and interpret the seismic
performance of the coal tailings dam.
6.3 Limitations of this research
It is noteworthy to mention that the type of FCR studied was anthracite coal tailings,
which are hydrophobic and can affect the liquefaction susceptibility of the material. The
limitation of the single element testing in the second chapter of this study was that the
heterogeneity of FCR in the field is not captured. And the samples used in the cyclic tests
were reconstituted and uniform. A uniform sample may not be able to show the void
redistribution and water film phenomenon, which may occur during and after cyclic
loading.
The limitation associated with the third chapter of this study was the low effective
stress achieved by shake table testing. The findings and conclusions may not be valid for
FCR under higher effective stress. Although the heterogeneity was captured in the shake
table tests, the specimen was still considered small compared to the field scale. Therefore,
the boundary effects may have affected the seismic performance of the specimen.
147
The limitation of the work presented in the fourth chapter of this study was the
calibration of the model based on one set of laboratory results. More samples should be
tested and used for the plasticity model calibration. The best estimate for the undrained
shear strength of FCR was also based on limited data available in literature.
6.4 Recommendations for future work
The following are recommended for future study:
(1) More in situ tests should be conducted in various coalfields, the equipment
should also be ideally modified to conduct CPT so that higher quality results can be
obtained.
(2) A series of centrifuge tests should be performed to investigate the studied factors
at higher effective stress.
(3) The numerical simulations should be conducted for different geometries and
statistical settings used in random fields generation. More input motions should also be
used to further confirm the findings.
(5) A probabilistic analysis platform should be established for the seismic stability
analysis of coal tailings dams.
(6) These research approaches can be extended to other type of tailings, which may
have different behavior and characteristics.
REFERENCES
Ajmera, B., Brandon, T., and Tiwari, B. (2015). “Cyclic strength of clay-like materials.”
In Proc., 6th Int. Conf. on Earthquake Geotechnical Engineering, Christchurch, New
Zealand.
Anderson, D. G., and Stokoe, K. H., II, (1978), "Shear modulus, a time-dependent soil
property," Dynamic Geotechnical Testing, ASTM STP 654, pp. 66-90.
Andrews, D. C. A. and Martin, G. R. (2000). “Criteria for liquefaction of silty soils.”
Proceedings of the 12th World Conference on Earthquake Engineering, New Zealand,
Paper No. 0312.
Andrus, R. D., Stokoe, K. H., II, and Juang, C. H. (2004). “Guide for shear wave-based
liquefaction potential evaluation.” Earthquake Spectra, 20(2), 285-308.
Antonaki, N., Abdoun, T., and Sasanakul, I. (2017). “Centrifuge tests on comixing of mine
tailings and waste rock.” Journal of Geotechnical and Geoenvironmental
Engineering, 144(1), 04017099.
ASTM, D. (1963). Standard test method for particle-size analysis of soils. ASTM D422-63.
Philadelphia, USA.
ASTM, C. (2006). “Standard test method for sieve analysis of fine and coarse
aggregates.” ASTM C136-06. Philadelphia, USA.
ASTM. (2012). “Standard test method for electronic friction cone and piezocone
penetration testing of soils”. ASTM D5778-12. American Society for Testing and
Materials
149
Au, S. K., & Beck, J. L. (2001). “Estimation of small failure probabilities in high
dimensions by subset simulation.” Probabilistic engineering mechanics, 16(4), 263-
277.
Au, S. K., & Wang, Y. (2014). “Engineering risk assessment with subset simulation.” John
Wiley & Sons.
Azam, S., and Li, Q. (2010). “Tailings dam failures: a review of the last one hundred
years.” Geotechnical News, 28(4), 50-54.
Beaty, M., and Byrne, P. M. (1998). “An effective stress model for predicting liquefaction
behaviour of sand.” In Geotechnical Earthquake Engineering and Soil Dynamics III
(pp. 766-777). ASCE.
Benzaazoua, M., Perez, P., Belem, T., and Fall, M. (2004). “A laboratory study of the
behaviour of surface paste disposal.” In Proceedings of the 8th International Symposium
on Mining with Backfill, The Nonferrous Metals Society of China, Beijing (pp. 180-
192).
Betz, W., Papaioannou, I., and Straub, D. (2014). Numerical Methods for the Discretization
of Random Fields by Means of the Karhunen–Loève Expansion. Computer Methods in
Applied Mechanics and Engineering, 271, 109-129.
Bray, J. D., and Sancio, R. B. (2006). “Assessment of the liquefaction susceptibility of
fine-grained soils.” Journal of Geotechnical and Geoenvironmental
Engineering, 132(9), 1165-1177.
Bray, J. D., Sancio, R. B., Riemer, M. F., and Durgunoglu, T. (2004). “Liquefaction
susceptibility of fine-grained soils.” Proc., 11th Int. Conf. on Soil Dynamics and
150
Earthquake Engineering and 3rd Int. Conf. on Earthquake Geotechnical Engineering,
D. Doolin et al., eds., Stallion Press, Singapore, 655–662.
Blight, G.E. and Fourie, A.B., (2005). “Catastrophe revisited–disastrous flow failures of
mine and municipal solid waste. Geotechnical & Geological Engineering, 23(3),
pp.219-248.
Boulanger, R. W. (2019). “Nonlinear dynamic analyses of Austrian dam in the 1989 Loma
Prieta earthquake.” Journal of Geotechnical and Geoenvironmental
Engineering, 145(11), 05019011.
Boulanger, R. W., & Idriss, I. M. (2006). “Liquefaction susceptibility criteria for silts and
clays.” Journal of geotechnical and geoenvironmental engineering, 132(11), 1413-
1426.
Boulanger, R. W., & Idriss, I. M. (2007). “Evaluation of cyclic softening in silts and
clays.” Journal of Geotechnical and Geoenvironmental Engineering, 133(6), 641-652.
Boulanger, R. W., and Montgomery, J. (2016). “Nonlinear deformation analyses of an
embankment dam on a spatially variable liquefiable deposit.” Soil Dynamics and
Earthquake Engineering, 91, 222-233.
Boulanger, R. W., Munter, S. K., Krage, C. P., and DeJong, J. T. (2019). “Liquefaction
evaluation of interbedded soil deposit: Çark canal in 1999 M7. 5 Kocaeli
earthquake.” Journal of Geotechnical and Geoenvironmental Engineering, 145(9),
05019007.
Boulanger, R. W., and Idriss, I. M. (2014). “CPT and SPT based liquefaction triggering
procedures.” Rep. No. UCD/CGM-14/01, Univ. of California, Davis, CA.
151
Boulanger, R. W., and Idriss, I. M. (2015). “CPT-based liquefaction triggering procedure.”
J. Geotech. Geoenviron. Eng., 142(2), 04015065.
Boulanger, R. W., and Ziotopoulou, K. (2017). “PM4Sand (version 3.1): A sand plasticity
model for earthquake engineering applications.” Report No. UCD/CGM-17/01, Center
for Geotechnical Modeling, Department of Civil and Environmental Engineering,
University of California, Davis, CA, March, 114 pp.
Boulanger, R. W., and Ziotopoulou, K. (2018). “PM4Silt (Version 1): A silt plasticity
model for earthquake engineering applications.” Report No. UCD/CGM-18/01, Center
for Geotechnical Modeling, Department of Civil and Environmental Engineering,
University of California, Davis, CA.
Boulanger, R. W., and Ziotopoulou, K. (2013). “Formulation of a sand plasticity plane-
strain model for earthquake engineering applications.” Soil Dynamics and Earthquake
Engineering, 53, 254-267.
Boulanger, R. W., and Idriss, I. M. (2004). “Evaluating the potential for liquefaction or
cyclic failure of silts and clays.” Rep. No. UCD/CGM-04/01, Center for Geotechnical
Modeling, Dept. of Civil and Environmental Engineering, Univ. of California, Davis,
Calif.
Boushehri, R., Estahbanati, S. H., Hosseini, S. M. M. M., and Soroush, A. (2020).
“Influence of reinforcement parameters on the seismic response of reinforced earth
dams.” Journal of Geotechnical and Geoenvironmental Engineering.
Boggs Jr, S. (2014). “Principles of sedimentology and stratigraphy.” Pearson Education.
Budhu, M. (2015). “Soil mechanics fundamentals.” John Wiley & Sons.
152
Burns, S. E., & Mayne, P. W. (1998). “Monotonic and dilatory pore-pressure decay during
piezocone tests in clay.” Canadian Geotechnical Journal, 35(6), 1063-1073.
Busch, R.A., Backer, R.R., Atkins, L.A. and Kealy, C.D., (1975). “Physical property data
on fine coal refuse.” US Department of Interior Bureau of Mines Report of
Investigation, 8062.
Byrne, P. M., and Seid-Karbasi, M. (2003). “Seismic stability of impoundments.”
In Proceedings of the 17th Annual Symposium, Vancouver Geotechnical Society,
Vancouver, BC.
Carraro, J. A. H., and Prezzi, M. (2007). “A new slurry-based method of preparation of
specimens of sand containing fines.” Geotechnical Testing Journal, 31(1), 1-11.
Castro, G. (1987). “On the behavior of soils during earthquakes–liquefaction.”
In Developments in geotechnical engineering (Vol. 42, pp. 169-204). Elsevier.
Castro, G. (2003). “Evaluation of seismic stability of tailings dams.” In Proceedings of the
12th Pan-American Conference on Soil Mechanics and Geotechnical Engineering,
Cambridge, MA (pp. 16-23).
Castro, G., and Troncoso, J. (1989). “Effects of 1989 Chilean earthquake on three tailings
dams.” In Proceedings of the Fifth Chilean Conference on Seismology and Earthquake
Engineering, Santiago, Chile.
Castillo, J., Hallman, D., Byrne, P., and Parra, D. (2005). “Dynamic analysis of heap leach
pad under high phreatic levels.” In Proceedings of 27th Mining convention, Arequipa,
Peru.
Cetin, K. O., Seed, R. B., Der Kiureghian, A., Tokimatsu, K., Harder, L. F., Jr., Kayen, R.
E., and Moss, R. E. S. (2004). “Standard penetration test-based probabilistic and
153
deterministic assessment of seismic soil liquefaction potential.” J. Geotech.
Geoenviron. Eng., 130(12), 1314–1340.
Chakraborty, S., Das, J. T., Puppala, A. J., and Banerjee, A. (2019). “Natural frequency of
earthen dams at different induced strain levels. Engineering geology, 248, 330-345.
Charlie, W. A., Rwebyogo, M. F., and Doehring, D. O. (1992). “Time-dependent cone
penetration resistance due to blasting.” Journal of Geotechnical Engineering, 118(8),
1200-1215.
Cowherd, D.C. and I.J. Corda, (1998). “Seismic considerations for upstream construction
of coal refuse dams.” Proceedings, 1998 Annual Conference, Las Vegas, Association of
State Dam Safety Officials, Lexington, KY., pp. 523-534.
Davies, M.P., (2002). “Tailings impoundment failures: are geotechnical engineers
listening.” Geotechnical News, September, 2002, pp.31-36.
Dafalias, Y. F., and Manzari, M. T. (2004). "Simple plasticity sand model accounting for
fabric change." Journal of Engineering Mechanics, 130(6), 622-634.
Dashti S, Bray JD, Pestana JM, Riemer MR, Wilson D. (2010) “Mechanisms of
seismically-induced settlement of buildings with shallow foundations on liquefiable
soil.” Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 136 (1),
151-164.
D’appolonia Engineering, Inc. (2009) “Engineering and design manual: coal refuse
disposal facilities,” prepared for the U.S. Mine Safety and Health Administration,
Pittsburgh, PA, 826 p., plus appendices.
Darendeli, M. B. (2001). “Development of a new family of normalized modulus reduction
and material damping curves.” PhD Dissertation, University of Texas A&M.
154
Dobry, R., and Alvarez, L. (1967). “Seismic failures of Chilean tailings dams. Journal of
Soil Mechanics & Foundations Div.
Dobry, R., Abdoun, T., Stokoe, K., II, Moss, R., Hatton, M., and El Ganainy, H. (2015).
“Liquefaction potential of recent fills versus natural sands located in high-seismicity
regions using shear-wave velocity.” J. Geotech. Geoenviron. Eng.,
10.1061/(ASCE)GT.1943-5606.0001239, 04014112.
Dowding, C. H., and Hryciw, R. D. (1986). “A laboratory study of blast densification of
saturated sand.” Journal of Geotechnical Engineering, 112(2), 187-199.
Ecemis, N., Demirci, H. E., and Karaman, M. (2015). “Influence of consolidation
properties on the cyclic re-liquefaction potential of sands.” Bulletin of Earthquake
Engineering, 13(6), 1655-1673.
Electric Power Research Institute (1993). “Guidelines for site specific ground motions.”
Rept. TR-102293, 1-5, Pala Alto, California.
El-Sekelly, W., Dobry, R., Abdoun, T., and Steidl, J. H. (2016a). “Centrifuge modeling of
the effect of pre-shaking on the liquefaction resistance of silty sand deposits.” J.
Geotech. Geoenviron. Eng., 10.1061/(ASCE)GT.1943-5606.0001430.
El-Sekelly, W., Abdoun, T., and Dobry, R. (2016b). “Liquefaction resistance of a silty sand
deposit subjected to preshaking followed by extensive liquefaction.” J. Geotech.
Geoenviron. Eng., 10.1061/(ASCE)GT.1943-5606.0001444.
Fenton, G. A., and Griffiths, D. V. (2003). “Bearing-capacity prediction of spatially
random C-Φ soils.” Canadian Geotechnical Journal, 40(1), 54-65.
155
Ferdosi, B., James, M., and Aubertin, M. (2015). “Effect of waste rock inclusions on the
seismic stability of an upstream raised tailings impoundment: a numerical
investigation.” Canadian Geotechnical Journal, 52(12), 1930-1944.
Finn, W. D. L., Bransby, P. L., and Pickering, D. J. (1970). “Effect of strain history on
liquefaction of sand.” J. Soil Mech. Found. Div., 96(SM6), 1917–1934.
Geremew, A. M., and Yanful, E. K. (2012). “Dynamic properties and influence of clay
mineralogy types on the cyclic strength of mine tailings.” International Journal of
Geomechanics, 13(4), 441-453.
Ha, I.S., Olson, S. M., Seo, M. W., and Kim, M. M. (2011). “Evaluation of reliquefaction
resistance using shaking table tests.” Soil Dyn Earthq Eng 31(4):682–691.
Hegazy, Y. A., Cushing, A. G., and Lewis, C. J. (2004). “Physical, mechanical, and
hydraulic properties of coal refuse for slurry impoundment design.” D’Appolonia
Engineering.
Heidari, T., and Andrus, R. D. (2012). “Liquefaction potential assessment of Pleistocene
beach sands near Charleston, South Carolina.” J. Geotech. Geoenviron. Eng.,
10.1061/(ASCE)GT.1943 5606.0000686, 1196–1208.
Home, A. (2019). “Brazilian mine tragedy will not be the last tailings dam disaster: Andy
Home, <https://www.reuters.com/article/us-vale-sa-disaster-ahome/brazilian-mine-
tragedy-will-not-be-the-last-tailings-dam-disaster-andy-home-idUSKCN1Q405J>
Hu, L., Wu, H., Zhang, L., Zhang, P. and Wen, Q., (2016). “Geotechnical properties of
mine tailings.” Journal of Materials in Civil Engineering, 29(2), p.04016220.
Huang, Y., Li, J., and Gamini, W. (1987). “Strength and consolidation characteristics of
fine coal refuse.” Report to Office of Surface Mining, Department of Interior.
156
Idriss, I. M., and Boulanger, R. W. (2006). “Semi-empirical procedures for evaluating
liquefaction potential during earthquakes.” Soil Dynamics and Earthquake Engineering
26: 115–130.
Idriss, I. M., and Boulanger, R. W. (2008). “Soil liquefaction during earthquakes.”
Earthquake Engineering Research Institute.
International Commission on Large Dams (ICOLD). (2001). “Tailings dams - risk of
dangerous occurrences, lessons learnt from practical experiences.” (Bulletin 121). Paris:
International Commission on Large Dams (ICOLD); 2001.
Ishihara, K., Yasuda, S., and Yokota, K. (1981). “Cyclic strength of undisturbed mine
tailings.
Ishihara, K. (1984). “Post-earthquake failure of a tailings dam due to liquefaction of pond
deposit.
Ishihara, K., Troncoso, J. O. R. G. E., Kawase, Y. A. S. U. H. I. R. O., and Takahashi, Y.
(1980). “Cyclic strength characteristics of tailings materials.” Soils and
Foundations, 20(4), 127-142.
Itasca Consulting Group. (2017). Fast Lagrangian analysis of continua, Version 8.0.
James, M. (2009). “The use of waste rock inclusions to control the effect of liquefaction in
tailings impoundments.” Ph.D. thesis, Department of Civil, Geological, and Mining
Engineering, Ecole Polytechnique de Montreal, Montréal, Que.
James, M., Gomes, P., and Schwenger, R. (2007). “The liquefaction resistance of tailings
and the stability of impoundments.” In Proceedings of the 60th Canadian Geotechnical
Conference, Ottawa, Ont (pp. 21-24).
157
James, M., Aubertin, M., Wijewickreme, D., and Wilson, G. W. (2011). “A laboratory
investigation of the dynamic properties of tailings.” Canadian Geotechnical
Journal, 48(11), 1587-1600.
Jefferies, M. G., and Davies, M. P. (1991). “Soil classification by the cone penetration test:
Discussion.” Canadian Geotechnical Journal, 28(1), 173-176.
Ji, J., Liao, H. J., and Low, B. K. (2012). “Modeling 2-D spatial variation in slope reliability
analysis using interpolated autocorrelations.” Computers and Geotechnics, 40, 135-
146.
Joshi, R. C., Achari, A., Kaniraj, S.R., and Wijeweera, H. (1995). “Effect of aging on the
Penetration Resistance of Sands.” Canadian Geotechnical Journal, 32(5), 767-782.
Kalinski, M. E., and Phillips, J. L. (2008). “Development of methods to predict the dynamic
behavior of fine coal refuse: Preliminary results from two sites in Appalachia.”
In Geotechnical Earthquake Engineering and Soil Dynamics IV (pp. 1-10).
Kalinski, M. E., and Salehian, A. (2016). “Estimating the cyclic and post-earthquake
behavior of coal mine tailings at a site in eastern Kentucky. In Geo-Chicago, pp. 279-
288.
Karimi, Z., and Dashti, S. (2016). "Numerical and centrifuge modeling of seismic soil–
foundation–structure interaction on liquefiable ground." Journal of Geotechnical and
Geoenvironmental Engineering, 142(1), 04015061.
Kim, T. C., and Novak, M. (1981). “Dynamic properties of some cohesive soils of
Ontario.” Canadian Geotechnical Journal, 18(3), 371-389.
158
Kokusho, T. (2003). “Current state of research on flow failure considering void
redistribution in liquefied deposits.” Soil Dynamics and Earthquake
Engineering, 23(7), 585-603.
Kuerbis, R., and Vaid, Y. P. (1988). “Sand sample preparation-the slurry deposition
method.” Soils and Foundations, 28(4), 107-118.
Ladd, C. C., and Foott, R. (1974). “New design procedure for stability of soft clays. Journal
of Geotechnical and Geoenvironmental Engineering, 100 (Proc Paper 10064).
Lee, K. L., (1977). "Adhesion bonds in sands at high pressures," Journal of the
Geotechnical Engineering Division, ASCE, Vol. 103, No. GT8, pp. 908-913.
Leventhal, A. R., and De Ambrosis, L. P. (1985). “Waste disposal in coal mining—a
geotechnical analysis.” Engineering geology, 22(1), 83-96.
Liew, M., Xiao, M., Liu, S., and Rudenko, D. (2020). “In situ seismic investigations for
evaluating geotechnical properties and liquefaction potential of fine coal
tailings.” Journal of Geotechnical and Geoenvironmental Engineering, 146(5),
04020014.
Martin, T. E., and Davies, M. P. 2000. “Development and review of surveillance programs
for tailings dams. Tailings Dams 2000 Proceedings. Association of State Dam Safety
Officials. Las Vegas. March.
Mejia, L. H., and Dawson, E. M. (2006). “Earthquake deconvolution for FLAC.” In FLAC
and Numerical Modeling in Geomechanics (Proceedings of the 4th International FLAC
Symposium, Madrid, Spain (pp. 211-219).
Mesri, G., Feng, T. W., and Benak, J. M. (1990). “Postdensification penetration resistance
of clean sands.” Journal of Geotechnical Engineering, 116(7), 1095-1115.
159
Mitchell, J. K., and Solymar, Z. V. (1984). “Time-dependent strength gain in freshly
deposited or densified sand.” J. Geotech. Engrg., ASCE, 110(11), 1159–1576.
Mine Safety and Hazard Administration (MSHA), (2009). “Engineering and design manual
for coal refuse disposal facilities.” United States Department of Labor MSHA
Publication.
Moss, R. E., Seed, R. B., Kayen, R. E., Stewart, J. P., Der Kiureghian, A., and Cetin, K. O.
(2006). “CPT-based probabilistic and deterministic assessment of in situ seismic soil
liquefaction potential.” Journal of Geotechnical and Geoenvironmental
Engineering, 132(8), 1032-1051.
Naesgaard, E., & Byrne, P. M. (2007). “Flow liquefaction simulation using a combined
effective stress–total stress model.” In 60th. Canadian Geotechnical Conference,
Canadian Geotechnical Society, Ottawa, Ontario.
National Research Council (US). Committee on Earthquake Engineering, & National
Research Council (US). Committee on Earthquake Engineering Research. 1985.
Liquefaction of soils during earthquakes (Vol. 1). National Academies.
National Research Council. (2002), “Coal waste impoundments: risks, responses, and
alternatives.” Washington, D.C.: National Academies Press.
Oda, M., Kawamoto, K., Suzuki, K., Fujimori, H., and Sato, J. (2001). “Microstructural
interpretation on re-liquefaction of saturated granular soils under cyclic loading.” J.
Geotech. Geoenviron. Eng., 10.1061/ (ASCE) 1090-0241(2001)127(5), 416–423.
Okamura, M., and Teraoka, T. (2006). “Shaking table tests to investigate soil desaturation
as a liquefaction countermeasure”. In Seismic Performance and Simulation of Pile
Foundations in Liquefied and Laterally Spreading Ground (pp. 282-293).
160
Olson, S. M., and Stark, T. D. (2002). “Liquefied strength ratio from liquefaction flow
failure case histories.” Canadian Geotechnical Journal, 39(3), 629-647.
Otsubo, M., Towhata, I., Hayashida, T., Shimura, M., Uchimura, T., Liu, B., and Rattez,
H. (2016). “Shaking table tests on mitigation of liquefaction vulnerability for existing
embedded lifelines.” Soils and Foundations, 56(3), 348-364.
Pépin, N., Aubertin, M., and James, M. (2012). “Seismic table investigation of the effect
of inclusions on the cyclic behaviour of tailings.” Canadian Geotechnical
Journal, 49(4), 416-426.
Phoon, K. K., Kulhawy, F. H., and Grigoriu, M. D. (1995). “Reliability-based design of
foundations for transmission line structures.” Rep. No. TR-105000, Electric Power
Research Institute, Palo Alto, CA.
Phoon, K. K., and Kulhawy, F. H. (1999). “Characterization of geotechnical
variability.” Canadian geotechnical journal, 36(4), 612-624.
Phoon, K. K., and Ching, J. (2014). “Risk and reliability in geotechnical engineering.” CRC
Press.
Plant, K. F., and Harriman, T. (2008). “Root cause analysis of TVA Kingston dredge pond
failure on December 22, 2008.” Volume I–Summary Report, Volume II–Geological and
Field Explorations.
Polito, C. P., and Martin II, J. R. (2001). “Effects of nonplastic fines on the liquefaction
resistance of sands.” Journal of Geotechnical and Geoenvironmental Engineering,
127(5), 408-415.
Prabhakaran, A., Kim, K., Orang, M. J., Qiu, Z., Ebeido, A., Zayed, M., Boushehri, R.,
Motamed, R., Elgamal, A., and Frazao, C. (2020). Polymer Injection and Liquefaction-
161
Induced Foundation Settlement: a Shake Table Test Investigation. Journal of
Geotechnical and Geoenvironmental Engineering.
Price, A. B., DeJong, J. T., and Boulanger, R. W. (2017). “Cyclic loading response of silt
with multiple loading events.” Journal of Geotechnical and Geoenvironmental
Engineering, 143(10), 04017080.
Qiu, Y., & Sego, D. C. (2001). “Laboratory properties of mine tailings.” Canadian
Geotechnical Journal, 38(1), 183-190.
Ramirez, J., Badanagki, M., Rahimi Abkenar, M., ElGhoraiby, M., Manzari, M. T., Dashti,
S., Barrero, A., Taiebat, M., Ziotopoulou, K., Liel, A. (2017). "Seismic performance of
a layered liquefiable site: validation of numerical simulations using centrifuge
modeling." In Geotechnical Frontiers 2017, pp. 332-341.
Rico, M., Benito, G., Salgueiro, A. R., Díez-Herrero, A., and Pereira, H. G. (2008).
“Reported tailings dam failures: a review of the European incidents in the worldwide
context.” Journal of Hazardous Materials, 152(2), 846-852.
Robertson, P. K. (2009). “Evaluation of flow liquefaction and liquefied strength using the
cone penetration test.” Journal of Geotechnical and Geoenvironmental
Engineering, 136(6), 842-853.
Robertson, P. K. (2016). “Cone penetration test (CPT)-based soil behaviour type (SBT)
classification system—an update.” Canadian Geotechnical Journal, 53(12), 1910-
1927.
Robertson, P. K., and C. E. Wride. (1998). “Evaluating cyclic liquefaction potential using
the cone penetration test.” Canadian Geotechnical Journal. 35(3), 442-459.
162
Robertson, P.K., da Fonseca, A.V., Ulrich, B. and Coffin, J., (2017). “Characterization of
unsaturated mine waste: a case history.” Canadian Geotechnical Journal, 54(12),
pp.1752-1761.
Salam, S., Xiao, M., Khosravifar, A., Liew, M., Liu, S., and Rostami, J. (2019).
“Characterizations of static and dynamic geotechnical properties and behaviors of fine
coal refuse.” Canadian Geotechnical Journal, 56(12), 1901-1916.
Salam, S., Xiao, M., and Wang, J. (2020). “Physical modeling of fine coal refuse using
shake table testing.” In GeoCongress 2020: Earthquake Engineering and Soil
Dynamics. Reston, VA: American Society of Civil Engineers.
Santoso, A. M., Phoon, K. K., and Quek, S. T. (2011). “Modified Metropolis–Hastings
algorithm with reduced chain correlation for efficient subset simulation.” Probabilistic
Engineering Mechanics, 26(2), 331-341.
Salehian, A. (2013). “Predicting the dynamic behavior of coal mine tailings using state-of-
practice geotechnical field methods.” University of Kentucky.
Schmertmann, J.H. (1991) “The mechanical aging of soils.” Journal of Geotechnical
Engineering, ASCE, 117(9), 1288-1330.
Seed, B. (1979). “Soil liquefaction and cyclic mobility evalution for level ground during
earthquakes.” Journal of geotechnical and geoenvironmental engineering, 105(ASCE
14380).
Seed, H. B., and Idriss, I. M. (1970). “Soil moduli and damping factors for dynamic
response analyses.” Report No. EERC 70-10, Univ. of California, Berkeley, Calif.
Seed, H. B., and Idriss, I. M. (1971). “Simplified procedure for evaluating soil liquefaction
potential.” Journal of Soil Mechanics & Foundations Div.
163
Seed, H.B., Wong, R.T., Idriss, I.M., and K. Tokimatsu (1986). “Moduli and damping
factors for dynamic analyses of cohesionless soils.” ASCE Journal of Geotechnical
Engineering, Vol. 112, No. 11, 1016-1032.
Seed, H. B., Lee, K. L., Idriss, I. M., and Makdisi, F. I. (1975). “The slides in the San
Fernando Dams during the earthquake of February 9, 1971.” Journal of Geotechnical
and Geoenvironmental Engineering, 101(ASCE# 11449 Proceeding).
Seed, R. B.; Moss, R. E. S.; Kammerer, A. M.; Wu, J.; Pestana, J. M.; Riemer, M. F.;
Sancio, R. B.; Bray, J. D.; Kayen, R. E.; Faris, A.; and Çetin, K. Ö., (2003). “Recent
advances in soil liquefaction engineering: a unified and consistent framework,”
Earthquake Engineering Research Center, University of California, Berkeley, 71 pages
(400/E177/2003-06).
Seid-Karbasi, M., and Byrne, P. M. (2004). “Embankment dams and
earthquakes.” International Journal on Hydropower & Dams, 11(2), 96-102.
Shuttle, D. A., and Cunning, J. (2007). “Liquefaction potential of silts from
CPTu.” Canadian Geotechnical Journal, 44(1), 1-19.
Sivathayalan, S., and Vaid, Y. P. (2004). “Cyclic resistance and post liquefaction response
of undisturbed in-situ sands.” In Proceedings of the 13th world conference on
earthquake engineering.
Taylor, D. (1948). “Fundamentals of soil mechanics.” Chapman and Hall, Limited; New
York.
Teng, W.C. (1962). “Foundation design.” Prentice-Hall, Inc., Englewood Cliffs, N.Y.
Teparaksa, J., and Koseki, J. (2018). “Effect of past history on liquefaction resistance of
level ground in shaking table test.” Géotechnique Letters, 8(4), 256-261.
164
Terzaghi, K., Peck, R. B., and Mesri, G. (1996). “Soil mechanics in engineering practice.”
John Wiley & Sons.
Thacker, B. K., Ullrich, C. R., Athanasakes, J. G., and Smith, G. (1988). “Evaluation of a
coal refuse impoundment built by the upstream method.” In Hydraulic Fill Structures
ASCE pp. 730-749.
Thevanayagam, S., T. Kanagalingam, A. Reinhorn, R. Tharmendhira, Ricardo Dobry, M.
Pitman, Abdoun, T., (2009) "Laminar box system for 1-g physical modeling of
liquefaction and lateral spreading." Geotechnical Testing Journal. 32, no. 5: 438-449.
Ullrich, C. R., Thacker, B. K., and Roberts, N. R. (1991). “Dynamic properties of fine-
grained coal refuse.”
Verma, P., Seidalinova, A., and Wijewickreme, D. (2018). “Equivalent number of uniform
cycles versus earthquake magnitude relationships for fine-grained soils.” Canadian
Geotechnical Journal, 56(11), 1596-1608.
Vick, S. G. (1990). “Planning, design, and analysis of tailings dams.” BiTech.
Wang, J., Xiao, M., Evans, J. C., Qiu, T., and Salam, S. (2019). “Time-dependent cone
penetration resistance of a postliquefaction sand deposit at shallow depth.” Journal of
Geotechnical and Geoenvironmental Engineering, 145(6), 04019021.
Wang, J., Salam, S., & Xiao, M. (2020). “Evaluation of the effects of shaking history on
liquefaction and cone penetration resistance using shake table tests.” Soil Dynamics and
Earthquake Engineering, 131, 106025.
Whitman, R. V. (1970). “Hydraulic fills to support structural loads.” Journal of Soil
Mechanics & Foundations Div.
165
Wijewickreme, D., Sanin, M. V., and Greenaway, G. R. (2005). “Cyclic shear response of
fine-grained mine tailings.” Canadian Geotechnical Journal, 42(5), 1408-1421.
Xiao, M., Liew, M., Salam, S., Liu, S., and Rostami, J. (2019). “Field investigation and
stability analysis of coal slurry impoundments.” Rep. No. S16AC20074, U.S. Office of
Surface Mining Reclamation and Enforcement, PA.
Yang, J. (2006). “Influence zone for end bearing of piles in sand.” Journal of geotechnical
and geoenvironmental engineering, 132(9), 1229-1237.
Yamamuro, J. A., and Lade, P. V. (1998). “Steady-state concepts and static liquefaction of
silty sands.” Journal of geotechnical and geoenvironmental engineering, 124(9), 868-
877.
Youd, T. L., Idriss, I. M., Andrus, R. D., Arango, I., Castro, G., Christian, J. T., Dobry, R.,
Finn, W. D. L., Harder, L. F., Hynes, M. E., Ishihara, K., Koester, J. P., Liao, S. S. C.,
Marcuson, W. F., Martin, G. R., Mitchell, J. K., Moriwaki, Y., Power, M. S., Robertson,
P. K., Seed, R. B., and Stokoe, K. H. (2001). “Liquefaction resistance of soils: summary
report from the 1996 NCEER and 1998 NCEER/NSF workshops on evaluation of
liquefaction resistance of soils.” J. Geotechnical and Geoenvironmental Eng., ASCE
127(10), 817–33.
Yu, H., Zeng, X., and Michael, P. R. (2019). “Geotechnical properties and flow behavior
of coal refuse under static and impact loading.” Journal of Geotechnical and
Geoenvironmental Engineering, 145(7), 04019024.
Zamiran, S., Salam, S., Osouli, A., and Ostadhassan, M. (2015). “Underground disposal of
fine coal waste.” In 49th US Rock Mechanics/Geomechanics Symposium. American
Rock Mechanics Association.
166
Zeng, X., Wu, J., and Rohlf, R. (1998a). “Modeling the seismic response of coal-waste
tailings dams.” Geotechnical News, 6(6), 29-32.
Zeng, X., Wu, J., and Rohlf, R. A. (1998b). “Seismic stability of coal-waste tailing dams.”
ASCE.
Zeng, X., Goble, J. A., and Fu, L. (2008). “Dynamic properties of coal waste refuse in a
tailings dam.” In Geotechnical Earthquake Engineering and Soil Dynamics IV (pp. 1-
14).
Zhang, C., Yang, C. H., and Bai, S. W. (2006). “Experimental study on dynamic
characteristics of tailings material.” rock and soil mechanics-wuhan-, 27(1), 35.
Zhang, Y. (2010). “New Advances in Machine Learning.” Online publication by InTech.
Ziotopoulou K, Maharjan M, Boulanger RW, Beaty M. H., Armstrong, RJ, and Takahashi
A. (2014). “Constitutive modeling of liquefaction effects in sloping ground.” In
Proceedings, Tenth U.S. National Conference on Earthquake Engineering, Anchorage,
Alaska. 2014.
Ziotopoulou, K., and Boulanger, R. W. (2015). “Validation protocols for constitutive
modeling of liquefaction.” Technical report, 6th International Conference on
Earthquake Geotechnical Engineering, Christchurch, New Zealand.
Ziotopoulou, K., and Montgomery, J. (2017). “Numerical modeling of earthquake-induced
liquefaction effects on shallow foundations.” In Proceedings of the 16th World
Conference on Earthquake Engineering, Paper (No. 2979, pp. 9-3).
Ziotopoulou, K., and Boulanger, R. W. (2016). “Plasticity modeling of liquefaction effects
under sloping ground and irregular cyclic loading conditions.” Soil Dynamics and
Earthquake Engineering, 84, 269-283.
VITA
Sajjad Salam
EDUCATION
Doctor of Philosophy in Civil Engineering
Minor in Statistics
The Pennsylvania State University, University Park, PA, United States.
Master of Science in Civil Engineering (Emphasis in Geotechnical Engineering)
Southern Illinois University Edwardsville, Edwardsville, IL, United States.
Bachelor of Science in Civil Engineering
Sharif University of Technology, Tehran, Iran.
RESEARCH EXPERIENCE
Graduate Research Assistant, The Pennsylvania State University, University Park, PA
Graduate Research Assistant, Southern Illinois University (SIUE), Edwardsville, IL
Research Assistant, Sharif University of Technology, Tehran, Iran
SELECTED TEACHING EXPERIENCE
Instructor, Pennsylvania State University, University Park, PA
Adjunct Lecturer, Southern Illinois University Edwardsville, IL
SELECTED WORK EXPERIENCE
Geotechnical Engineer, Marino Engineering Associates, Inc., Saint Louis, MO
Staff Civil Engineer, Mahab Ghodss Consulting Engineering Co., Tehran, Iran
SELECTED PUBLICATIONS
Salam, S., Xiao, M., Khosravifar, A. Ziotopoulou, K., (2020). “Seismic Stability of
Spatially Variable Liquefiable Coal Tailings Dam using Pore Pressure Plasticity Models.”
Computers and Geotechnics, submitted.
Salam, S., Xiao, M., Evans, J., (2020). “Strain History and Aging Effects on the Strength
and Cyclic Response of Fine-Grained Coal Refuse.” Journal of Geotechnical and
Geoenvironmental Engineering, https://doi.org/10.1061/(ASCE)GT.1943-
5606.0002364.
Salam, S., Xiao, M., Khosravifar, A., Liew, M., Liu, S., Rostami, J. (2019).
“Characterizations of Static and Dynamic Geotechnical Properties and Behaviors of Fine
Coal Refuse.” Canadian Geotechnical Journal. https://doi.org/10.1139/cgj-2018-0630