You are on page 1of 14

Wang, Z. et al. (2020). Géotechnique 70, No. 10, 849–862 [https://doi.org/10.1680/jgeot.17.P.

300]

Unified thixotropic fluid model for soil liquefaction


ZHIHUA WANG , JINLONG MA , HONGMEI GAO†, ARMIN W. STUEDLEIN‡, JIAN HE 
and BINGHUI WANG§

The geotechnical earthquake engineering profession has struggled with the inherent complexity of
the multiphase soil response to cyclic loading owing to the progressive nature of the generation of
excess pore pressure (EPP) and degradation of soil stiffness and strength. One approach to improve
understanding of the cyclic response and correlate the transition from a two-phase saturated soil to a
single, fluid-phase liquefied soil is to treat the soil as a non-Newtonian viscous liquid. However, the
work to date suggests that the viscous fluid model approach can only be implemented following
the onset of sustained soil liquefaction. This paper presents a unified thixotropic fluid model and
framework that effectively links the pre-shear soil fabric and its progressive cyclic response to the onset
and maintained state of soil liquefaction. The framework treats the soil fabric as a fluid net-type
structure proposed for use with thixotropic fluids, and presents the constitutive state and rate equations
describing the deconstruction of the liquefiable soil fabric in response to cyclic loading. The unified
framework uses physically meaningful soil parameters that can be obtained from common cyclic
laboratory tests to seamlessly link the state-dependent and shear-strain-rate-dependent nature of soils to
the generation of EPP, the latter of which is shown to increase in significance as EPP accumulates. The
proposed thixotropic-induced excess pore pressure model should prove advantageous for use in forward
modelling of the stress–strain rate response of liquefiable soil and generation of EPP.

KEYWORDS: laboratory tests; liquefaction; pore pressures; theoretical analysis; time dependence

INTRODUCTION possibly strength if the magnitude and/or duration is


Significant progress in the evaluation of the triggering of sufficient (Marcuson, 1978). The onset of soil liquefaction
earthquake-induced soil liquefaction has been made since the in undrained loading can be readily evaluated using the
1964 Niigata, Japan and Good Friday, Alaska earthquakes principle of effective stress, and initial liquefaction has been
(e.g. Seed, 1979; Seed et al., 1985; Youd & Idriss, 2001; sufficiently defined as the case when the magnitude of excess
Boulanger & Idriss, 2014, 2015). However, soil liquefaction pore pressure equals the initial mean effective stress (Seed &
and its consequences, which include ground failure, excessive Lee, 1966). Thus, the first question regarding the conditions
displacement and rotation of sub- and superstructures, as for liquefaction of relatively clean granular soils may be
well as flotation of buried structures, continue to represent a considered answered. However, this definition does not
formidable engineering concern for civil infrastructure allow for ready correlation between the magnitude of
(Tokimatsu et al., 2012; Konagai et al., 2013; Cubrinovski excess pore pressure and the characteristics of the liquefied
et al., 2014). Recent earthquakes in Japan and New Zealand soil. Poulos et al. (1985) sought to describe soil liquefaction
point to pertinent questions that remain outstanding, includ- in terms of its flow-type behaviour and relationship to
ing: what are the conditions (e.g. susceptibility, loading) for strength degradation, noting that significant deformations
triggering of liquefaction? What are the consequences of are possible in the absence of zero mean effective stress.
liquefaction? And, how can liquefaction be economically Notably, both perspectives (i.e. zero mean effective stress
mitigated. The unifying principle guiding development of against strength degradation) have been adopted in method-
solutions to these questions rests with the need to understand ologies developed to simulate the response of liquefied soils
the mechanisms associated with the onset and persistence (e.g. Davis & Berrill, 1982; Ishihara, 1993). Indeed, many
of the liquefied soil state. state-of-the-practice approaches continue to rely on classical,
Earthquake-induced soil liquefaction may be described as dynamic, rigid–perfectly plastic sliding block analyses
the process whereby cyclic loading, generally in the form (e.g. Newmark, 1965; Makdisi & Seed, 1979; Bray &
of shear stresses, τ, produced from vertically propagating Travasarou, 2007; Saygili & Rathje, 2008) or elasto-plastic
horizontal shear waves, results in the generation of excess constitutive models implemented into continuum represen-
pore pressures that can lead to loss of soil stiffness and tations of the near- and free-field (e.g. Yang et al., 2003;
Dafalias & Manzari, 2004; Tasiopoulou & Gerolymos,
2016). However, these simplified analyses are either inap-
Manuscript received 1 December 2017; revised manuscript accepted propriate for modelling liquefied soil behaviour (in the case
21 August 2019. Published online ahead of print 24 September 2019. of rigid sliding block analyses) or generally only appropriate
Discussion on this paper closes on 1 February 2021, for further during that portion of shaking where deformations begin
details see p. ii. to accumulate in response to the generation of excess pore
 Research Center of Urban Underground Space, Nanjing Tech
pressure and onset of liquefaction, requiring in some cases
University, Nanjing, P. R. China.
artificial means (e.g. local rules for switching to residual
† Research Center of Urban Underground Space, Nanjing Tech
University, Nanjing, P. R. China (Orcid:0000-0003-0096-4419). strength-type constitutive relations) to simulate large defor-
‡ School of Civil and Construction Engineering, Oregon State mation behaviour (Wang et al., 2014). Typical continuum-
University, Corvallis, OR, USA. type analyses further suffer from the inability to incorporate
§ School of Architecture and Civil Engineering, Jiangsu University shear strain rate, γ̇, dependence of soils, a critical character-
of Science and Technology, Zhenjiang, P. R. China. istic during periods of high excess pore pressure (Towhata

849

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
850 WANG, MA, GAO, STUEDLEIN, HE AND WANG
et al., 2010), the result of which leads to inaccuracies in are presented along with their respective physical material
predicting the mechanisms and consequences of sustained, parameters. Set within the framework of viscous fluid
near-liquefied or fully liquefied soil. behaviour, the response of the liquefied soil is sufficiently
An alternative approach to analyse and simulate liquefac- described using the thixotropic-induced excess pore pressure
tion behaviour that has received significantly less attention is (TEPP), developed and validated in observation of over
to adopt fluid dynamics principles and treat liquefied soil as a 40 original experiments and re-analysis of data previously
viscous fluid with viscosity coefficient η. Towhata et al. (1992) reported by others.
proposed a closed-form solution for the displacement of
liquefied soil by assuming that deformations observed in 1g
shake table tests may be described using a sinusoidal function
determined using the minimum energy principle. Hamada & CONCEPT OF THIXOTROPIC-INDUCED EXCESS
Wakamatsu (1998) treated liquefied soil as a pseudo-plastic PORE PRESSURE
fluid with flow velocity set proportionally to the square root of Constitutive relationships for thixotropic fluids
the liquefied soil thickness. Naturally, the impact of the fluid Non-Newtonian fluids, defined by their non-constant
model selected to represent the viscous behaviour of the viscosity coefficients, may exhibit a range in behaviour
liquefied soil on the deformation response became apparent as including elasticity, inelasticity with or without a yield
others adopted the fluid dynamics approach to the problem stress, and/or time dependence. Fig. 1 demonstrates some
of deformation behaviour (e.g. Uzuoka et al., 1998; Hadush common non-Newtonian responses of viscous fluids: the
et al., 2001; Sawicki & Mierczynski, 2009), as implied by the power-law fluid model captures non-linear inelastic shear-
differences in fluid response illustrated in Fig. 1. thinning behaviour, whereas the Bingham fluid model treats
Upon liquefying, soil behaves similarly to a non- a fluid as a rigid body prior to the onset of the yield shear
Newtonian fluid with a viscosity that is related to the shear- stress. These two fluid models have been widely used in
ing rate (Towhata et al., 1999; Hadush et al., 2001). analysing the large deformation characteristics of liquefied
Experimental methods used to quantify the viscosity coeffi- soil (Towhata et al., 1992; Hadush et al., 2001). Some typical
cient η of liquefied soil have ranged from cyclic triaxial com- fluids such as suspensions, liquid foods and crude oils belong
pression tests to object-pulling tests (e.g. bars, spheres, to a class of fluids that may be defined by their thixotropy, a
coupons; Towhata et al., 1999; Nishimura et al., 2002; time-dependent variation in viscosity during shearing. An
Hwang et al., 2006; Sawicki & Mierczynski, 2009; Chen important characteristic of thixotropic fluids is that they can
et al., 2016; Dewoolkar et al., 2016). These studies have achieve a stable viscous state under a constant shearing rate.
shown that a properly calibrated viscous fluid can indeed Cheng (1973) established a structural theory of thixotropic
capture the deformation characteristics of liquefied soils. fluid assuming that the fluid could be characterised with an
However, the questions regarding the correlation between initial, inner net-like structure. The variation of the viscosity
generation of excess pore pressure and the transition of coefficient with shearing time associated with thixotropy
saturated soil from a two-phase material with an established was attributed to structural changes in the assumed net-like
soil fabric (i.e. arrangement of grains in the soil skeleton; structure, which could be continuously destroyed and
Zlatovic & Ishihara, 1997) to a single, fluid-phase material reconstructed during certain shearing actions. Upon reaching
remain unanswered, and the viscous fluid model approach an equilibrium state under Cheng’s theory, the inner structure
can only be implemented following the onset of sustained of the thixotropic fluid no longer changes.
soil liquefaction. Using λ to represent the integrity of the inner net-like
Owing to the progressive nature of the generation of excess structure, the general form of the constitutive state and rate
pore pressure and degradation of soil strength with increased equations for the thixotropic fluid could be expressed as
cyclic loading, the profession has experienced significant (Cheng, 1973)
difficulty in modelling the spectrum of the possible phases of τ ¼ ηðλ; γ̇Þγ̇ ð1aÞ
soil and corresponding deformation response under a unified
approach. This study implements the classical structural dλ
theory of thixotropic fluids to model the de-structuring of ¼ gðλ; γ̇Þ ð1bÞ
dt
an imaginary fluid (the intact, two-phase soil) at an initial
equilibrium state to achieve the characteristics of a viscous where γ̇ may be expressed in s1, and η(λ,γ̇) is the shear
fluid (the single-phase liquefied soil) in a limit equilibrium strain-rate-dependent and structural-integrity-dependent vis-
state under cyclic loading. The state and rate equations cosity coefficient expressed in kPa s or Pa s. The structural
necessary for describing liquefied soil as a thixotropic fluid parameter λ ranges from 0 to 1; the lower extreme corres-
ponds to the case where the fluid structure is fully destroyed,
whereas the upper extreme corresponds to an intact state. The
state equation (equations (1a) and (1b)) can be replaced by
the classical Moore thixotropic fluid constitutive equations,
τ = τ0 + ηγ⋅
Shear stress, τ

given by (Meris, 1979)


τ ¼ ðηe þ aλÞγ̇ ð2aÞ
τ = Kγ⋅ n–1γ⋅

¼ bð1  λÞ  cλγ̇ ð2bÞ
dt
τ0 where ηe is the viscosity coefficient for the fully destroyed
fluid, and constants a, b and c are experimentally derived as
τ = ηeqγ⋅ described subsequently. Equations (2a) and (2b) describe the
shear stress–strain rate and structural instability rate relation-
Shear strain rate, γ⋅ ships for a Moore thixotropic fluid, respectively, in which the
products b(1  λ) and cλγ̇ represent the fluid reconstruction
Fig. 1. Commonly used viscous fluid models used for analysis of and destruction rate terms, respectively. Application of these
large-deformation response constitutive state and rate equations to liquefiable soils

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
UNIFIED THIXOTROPIC FLUID MODEL FOR SOIL LIQUEFACTION 851
requires the linkage of the thixotropic fluid structure concept rate coefficient c ¼ 0) or has not been subjected to any
to soil fabric that obeys the widely accepted and more shearing action (γ̇ ¼ 0), and is intact.
familiar concept of effective stress. (b) For b = 0, λe = 0. The reconstruction rate coefficient b is
equal to zero indicating that the only phenomenon
occurring is the destruction of the inner structure of the
Application of thixotropy theory to soil liquefaction liquefiable soil during shear. In this condition, the
This study introduces the structural theory for thixotropic liquefiable soil eventually enters the limit equilibrium
fluids to soil liquefaction analysis, assuming that liquefiable state (i.e. full liquefaction) due to shearing action of
soils can be treated as a thixotropic fluid. In this framework, sufficient time.
the soil skeleton or fabric can be considered analogous to the (c) For γ̇ ! 1, λe ! 0, demonstrating that as long as the
inner, net-like structure of the thixotropic fluid as shown in shear rate exerted on the soil is sufficiently large, the soil
Fig. 2. Prior to cyclic loading, the inner net-like structure of fabric will be completely destroyed and associated with
the liquefiable soil consists of intact and stable soil particles full liquefaction.
and pore water. The intact, stable state is defined as the initial
equilibrium state (Fig. 2), with excess pore pressure ratio Cases (b) and (c) correspond to λe = 0, and therefore sub-
ru = 0 and inner structural integrity parameter λ = 1. Cyclic stitution of these criteria into equation (2a) yields
loading to liquefaction results in the generation of excess τ e ¼ ηe γ̇ ð6Þ
pore pressures and in the progressive de-structuring of the
soil fabric. When the inner structure of the liquefiable soil is where ηe and τe are the viscosity coefficient and shear stress in
completely destroyed – that is, the soil is at a fully liquefied the limit equilibrium state, respectively. An appropriate
state with ru = 1·0 and λ = 0 – the liquefiable soil has entered thixotropic fluid framework for liquefaction analyses would
the limit equilibrium state (Fig. 2). In this framework, the capture the soil response and generation of thixotropic-
inner structural integrity parameter λ may be related to ru to induced excess pore pressure (TEPP) across all magnitudes
satisfy the following conditions in shear strain. For implementation of TEPP into soil
liquefaction problems, the specific relationship between the
λ ¼ 1 and ru ¼ 0 for initial equilibrium state ð3aÞ simultaneous, progressive reduction of the inner structural
λ ¼ 0 and ru ¼ 1 for limit equilibrium state ð3bÞ integrity parameter and generation of excess pore pressure
ratio must be established and the relationship should satisfy
Whether or not the soil is in an initial or limit equilibrium the conditions of equations (3a) and 3(b).
state, a shared characteristic of these states is that the inner
net-like structure of the liquefiable soil does not change with
the time – that is VALIDATION OF THE TEPP FLUID MODEL
Experimental basis: soils and laboratory tests
dλ In order to provide a broad and deep experimental basis for
¼0 ð4Þ
dt establishing the concept of thixotropic-induced excess pore
pressure generation, an experimental laboratory programme
It is assumed here that the liquefiable soil is a Moore-type
devised to produce original measurements of the cyclic
thixotropic fluid as described in equations (2a) and (2b).
response of sands was undertaken and coupled with previously
Substitution of equation (2b) into equation (4) yields
reported data on other soils. The new tests performed for this
b study included undrained stress- and strain-controlled cyclic
λe ¼ ð5Þ triaxial (CTC) tests on 24 specimens of schistose, siliceous
b þ cγ̇
Nanjing fine sand and 12 specimens of angular, calcareous
where λe is the inner structural integrity parameter for soils in South China Sea medium sand. Nanjing fine sand is an
an equilibrium state. Equation (5) indicates that the soil can alluvial sediment distributed along the floodplains of the lower
be in an equilibrium state for the following three cases. reaches of the Yangtze River in Nanjing, China. The South
China Sea medium sand is a marine deposit derived from coral
(a) For c = 0 or γ̇ ¼ 0, λe = 1, corresponding to the initial reefs. The physical properties of each sand are presented in
equilibrium state shown in Fig. 2. At this moment, the Table 1, and the corresponding grain size distributions are
soil fabric has not undergone any damage (destruction shown in Fig. 3. The initial effective confining stresses applied

Excess pore pressure ratio of the liquefiable soil

ru = 0 ru = 1
Initial equilibrium state

Limit equilibrium state

λ=1 λ=0

Inner structure integrity parameter of the imaginary fluid

Fig. 2. Conceptual link between the TEPP fluid model and changes in soil fabric

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
852 WANG, MA, GAO, STUEDLEIN, HE AND WANG
Table 1. Physical properties of the sands tested for this study

Soil Mean diameter, Uniformity coefficient, Curvature coefficient, Maximum void Minimum void Specific
d50: mm cu ¼ d60/d10 cc ¼ d230/(d60d10) ratio, emax ratio, emin gravity, Gs

Nanjing sand 0·16 2·31 1·07 1·14 0·62 2·70


Coral sand 0·40 4·47 0·91 1·14 0·69 2·77

100 Equation (8) indicates that the viscosity coefficient evolves


over the course of cyclic loading since the maximum shear
Per cent passing: %

80 strain rate and maximum shear stress varies continuously for


stress- and strain-controlled CTC tests, respectively. In using
60 the maximum shear strain rate or shear stress for a given
cycle to represent the entire cycle of loading, the dynamic
40
Fine Nanjing sand response can be treated as a monotonic process; accordingly,
Coarse to medium the viscosity coefficient expressed by equation (8) reduces
20
coral sand monotonically with time as the strength and stiffness of a
0
specimen degrades.
10 1 0·1 0·01
Grain size diameter : mm
State equation for the TEPP fluid model
Fig. 3. Grain size distributions for fine Nanjing sand and coarse to Experimental validation of the state equation. The appli-
medium coral sand cation of the state equation for a Moore thixotropic fluid to
liquefiable soils requires that equation (2a) be experimentally
validated for cyclic loading. Focusing first on the stress-
to the Nanjing fine sand specimens ranged from 30 to 150 kPa, controlled undrained CTC tests (e.g. Fig. 4(a)), the maxi-
and 50 and 100 kPa for the South China Sea sand specimens. mum excess pore pressure and the maximum shear strain rate
In order to supplement the new CTC data, data from cyclic continuously increased to a full state of liquefaction. The
experiments reported in the literature were also evaluated, relationship between the maximum γ̇ and ru can be obtained
including two strain- and five stress-controlled CTC tests, and as shown in Fig. 5. For each cycle, the relationship can be
three cyclic simple shear (CSS) tests. Together, the range of described by a hyperbolic model given by
soils and conditions evaluated to form the basis for the
validation of the TEPP model include sand–gravel mixtures, 1
clean sand, silty sand, and silt, with relative densities ranging γ̇ ¼ ð9Þ
A þ Bð1  ru Þ
from 20 to 92% and effective confining pressures ranging from
30 to 207 kPa. Table 2 summarises the laboratory tests used to where A and B are fitting parameters determined using
form the basis for the validation of TEPP model. ordinary least squares (OLS) and are summarised in Table 2.
Under the conditions of a limit equilibrium state (i.e. case (b)
with ru = 1·0), parameter A can be defined in terms of ηe and
Computation of the viscosity coefficient the dynamic shear stress amplitude, τd, by substitution of
Although uniform amplitude stress- and strain-controlled equation (8) into equation (9)
cyclic tests each have their advantages and limitations,
A ¼ ηe =τ d ð10Þ
neither test simulates the transient shear strain rates
associated with strong ground motion. Nonetheless, these and the sum of A and B can be inferred equal to the reci-
tests can serve to form the fundamental basis for the TEPP procal of the initial shear strain rate, by substitution of ru = 0
fluid model by facilitating computation of the viscosity into equation (9)
coefficient. Based on the shear strain–time histories
measured or imposed in the CTC tests, the shear strain rate A þ B ¼ η1 =τ d ð11Þ
may be calculated by (Chen et al., 2016)
  Substitution of equation (10) into equation (11) yields
1 γiþ1  γi γi  γi1
γ̇i ¼ þ ð7Þ
2 tiþ1  ti ti  ti1 B ¼ ðη1  ηe Þ=τ d ð12Þ
where γ̇i is shear strain rate at time ti, and γiþ1, γi and γi1 are
the shear strains at time tiþ1, ti and ti1, respectively. The where η∞ is the viscosity coefficient in the initial equilibrium
shear strain rate associated with stress- and strain-controlled state (kPa s). Substitution of equations (10) and (12) into
cyclic loading as represented by the maximum shear strain equation (9) yields
rate is shown in Fig. 4. Clear correlation of the maximum
excess pore pressures to changes in the maximum shear τ d ¼ ½ηe þ ðη1  ηe Þð1  ru Þγ̇ ð13Þ
strain rate can be observed for the stress-controlled tests,
whereas the invariant maximum shear strain rate in the Letting λ ¼ 1  ru and a ¼ η∞  ηe, equation (13) is shown
strain-controlled test prevents similar comparison. Regarding equal to equation (2a), the state equation for a Moore-type
liquefiable soil as a thixotropic fluid, the viscosity coefficient thixotropic fluid.
at the Nth cycle is expressed as Focusing now on the strain-controlled CTC tests, the
τN maximum shear strain rate for any given cycle was constant
ηNi ¼ i ð8Þ owing to the use of uniform strain amplitudes and loading
γ̇Ni frequencies (see Fig. 4(b)), whereas τd reduced with increases
where τ Ni and γ̇Ni are the maximum shear stress and shear in ru. Fig. 6 shows that the relationship between τd and
strain rate for the corresponding cycle, respectively. (1  ru) exhibits an approximately linear relationship for both

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
UNIFIED THIXOTROPIC FLUID MODEL FOR SOIL LIQUEFACTION 853
CTC and CSS tests that can be expressed as and C in Table 2. Yet, a relationship between ηe and the shear
strain rate at the limit equilibrium state, γ̇L , equal to 1/A and
τ d ¼ C þ Dð1  ru Þ ð14Þ γ̇ for stress- and strain-controlled tests, respectively, must
where C and D are the intercept and slope of the fitted line, be established to fully characterise the shear strain rate
respectively, and summarised in Table 2 for each test. dependency. Fig. 7 demonstrates the experimental relation-
Substitution of equations (3a) and 3(b) into equation (14) ship between ηe and γ̇L . The experimental data in Fig. 7
shows that contain all the experimental tests in Table 2 and some shake
table tests conducted by others (Hamada & Wakamatsu,
(a) when ru = 1, C = τd, implying that parameter C has a 1998; Hwang et al., 2006; Huang et al., 2012). The relation-
physical meaning (i.e. the dynamic shear stress applied ship between normalised ηe/σc′ and γ̇L is shown as follows
on the soil) when the liquefied soil has reached the limit
ηe
equilibrium state; in other words, C ¼ ηe γ̇ ¼ kðγ̇L Þn ð15Þ
(b) when ru = 0, τd = C + D, indicating that the sum of σ′c
parameters C and D also has a physical meaning, equal where σc′ is the effective confining pressure, and k and n are
to the dynamic shear stress applied on the soil in the empirically derived fitting parameters determined using
initial equilibrium state. OLS, equal to 0·235 and 0·71, and which indicates that
Stated formulaically in terms of established viscous fluid liquefied soils can indeed be characterised by their pressure
behaviour, C ¼ ηe γ̇ and C þ D ¼ η1 γ̇. Substitution of these and rate dependence and shear-thinning behaviour over four
terms into equation (14) produces the same state equation as orders of magnitude in shear strain rate. Although the
equation (13), and thus it may be concluded that the shear effective stresses may be rather low in the shake table tests, it
stress–strain rate relationship of liquefiable soil is not is notable that the effective confining pressure–normalised
influenced by the mode of loading when they are treated as viscosity coefficient at the moment of liquefaction is well
a Moore-type thixotropic fluid. with the scatter of normalised viscosity coefficients derived
In the foregoing validation, it was noted that the state from laboratory tests at significantly greater confining
equation for the Moore-type thixotropic fluid (equation (2a)) pressures. Substitution of equation (8) into equation (15)
has proved suitable for describing the shear stress–strain produces
rate relationship of liquefiable soil when expressed using
CSR ¼ kðγ̇L Þnþ1 ð16Þ
equation (13). The following observations may be concluded.
where CSR is the cyclic stress ratio, represented using the
(a) Based on its suitability for capturing the salient response widely accepted power law (e.g. Chen et al., 2016; Price et al.,
of liquefiable soils under various cyclic stress paths, 2017) and CSR ¼ 2σd/σc′ ¼ τd/σc′ . As an empirical equation,
transient and constant maximum shear strain rates, and equation (16) can be adopted to estimate the possible shear
soils of various solid constituents (e.g. sand, silty sand, strain rate of soils when liquefied under a certain cyclic stress
clayey sand and silt), equation (13) can serve as an ratio in stress-controlled tests, as well as to evaluate the
appropriate state equation describing the constitutive required dynamic shear stress to maintain a constant shear
response of liquefiable soil as a Moore-type thixotropic strain rate γ̇L for soils entering the liquefied state in
fluid. strain-controlled tests.
(b) The inner structural integrity parameter λ describing the Equation (15) shows that the value of ηe is proportional to
destruction of the soil fabric was independent of stress the value of σc′ for a constant shear strain rate. Since γ̇L varies
path or soil type. The relationship λ ¼ 1  ru illustrates in stress-controlled tests, it is more convenient to discuss
the progressive failure of the inner structure of a the relationships between ηe, σc′ and relative density using
liquefiable soil and corresponds to the increase in excess strain-controlled tests. Fig. 8 shows that ηe increases with
pore pressure, and validates the concept of the TEPP increases in σc′ and Dr for the liquefied coral sands, and is in
fluid model proposed in this study. general agreement with observations reported by Nishimura
(c) The viscosity coefficients ηe and η∞ have clear physical et al. (2002).
meaning that relates liquefiable soil to their initial and One conclusion stemming from the above-mentioned
limit equilibrium states, respectively, and can be derived liquefaction analysis using thixotropy theory is that the
based on experiments. The derived state equation greater the initial σc′ and Dr, the greater the shear stress
(equation (13)) illustrates that the initial viscosity required to maintain a constant γ̇L in liquefied soils. This
coefficient (and therefore the initial, stable state of the conclusion has been verified by many laboratory experiments
soil fabric) progressively reduces in response to the (e.g. Towhata et al., 1992; Hadush et al., 2001; Chen et al.,
generation of excess pore pressure. As soil approaches 2016), and it is demonstrated by Fig. 9 that the required CSR
the liquefaction state, the influence of the initial, intact increases when shearing occurs at a higher γ̇L. This is a
soil fabric (represented by η∞) is small because the soil typical, although often unrecognised, rate-dependency effect
fabric is nearly destroyed and λ ¼ 1  ru is near zero. of liquefied soils (Towhata et al., 2010). For example, when γ̇L
increases from 0·5 to 1·5 s1, the dynamic shear stress
increases from 3·9 and 7·8 kPa for confining pressures of
Viscosity coefficient in the limit equilibrium state, ηe. The 50 kPa and 100 kPa, respectively (equation (16) and Fig. 9).
limit equilibrium state for a liquefiable soil is assumed to
occur at ru = 1·0; thus, the key to the implementation of the
TEPP model is to establish the shear strain rate dependency Viscosity coefficient in the initial equilibrium state, η∞. As
of liquefied soils at the moment of liquefaction. It is widely interpreted above, η∞ denotes the viscosity coefficient of
recognised that the viscosity coefficient of liquefied soils liquefiable soils in the initial equilibrium state corresponding
decreases with increases in shear strain rate (i.e. the liquefied to ru ¼ 0. The values of η∞, shown in Table 2, are obtained
soil exhibits shear thinning; Nishimura et al., 2002; De Alba through equations (11) and (14) for stress-controlled and
& Ballestero, 2006; Towhata et al., 2010). The viscosity strain-controlled tests, respectively. Although the shear strain
coefficient at the limit equilibrium state ηe can be calculated rate of soil in its initial equilibrium state is zero or near zero
using equation (10) and empirically derived parameters A (i.e. below that required to exceed the threshold shear strain;

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
854
Table 2. Experimental cases used in this study

Test Soil types/test modes Loading conditions Max. ru Fitted or calculated parameters
no.
Dr: % σc′ : kPa CSR γ̇con : s1 f : Hz A B C D ηe: kPa s η∞: kPa s γ̇L : s1 β: kPa1

NJS1 Nanjing fine sand/stress-controlled 20 100 0·155 — 1·0 1·0 2·53 100·8 — — 39·2 1601·6 0·395 0·731
NJS2 (in this study) 30 100 0·155 — 1·0 1·0 2·01 108·4 — — 31·2 1711·4 0·498 0·529
NJS3 30 70 0·144 — 1·0 1·0 3·41 303·8 — — 34·4 3096·7 0·293 0·879
NJS4 30 30 0·162 — 1·0 1·0 4·69 116·1 — — 36·9 949·3 0·213 2·381
NJS5 30 70 0·186 — 1·0 1·0 2·95 132·9 — — 38·4 1768·8 0·339 0·909
NJS6 30 100 0·150 — 1·0 1·0 2·11 118·9 — — 31·7 1815·2 0·474 0·548
NJS7 30 90 0·135 — 1·0 1·0 2·76 148·3 — — 33·5 1835·4 0·362 0·624
NJS8 30 150 0·268 — 1·0 1·0 0·85 113·5 — — 21·4 2881·6 1·176 0·388
NJS9 46 150 0·218 — 1·0 1·0 1·10 103·9 — — 31·0 2961·3 0·909 0·206

WANG, MA, GAO, STUEDLEIN, HE AND WANG


NJS10 46 150 0·194 — 1·0 1·0 1·54 143·1 — — 35·6 3338·9 0·649 0·164
NJS11 70 70 0·194 — 1·0 1·0 2·14 193·3 — — 29·1 3057·4 0·467 0·507
NJS12 70 70 0·201 — 1·0 1·0 2·34 202·4 — — 32·9 2880·7 0·427 0·549
NJS13 70 100 0·166 — 1·0 1·0 2·49 210·2 — — 41·3 3527·3 0·402 0·254
NJS14 70 100 0·186 — 1·0 1·0 1·93 166·6 — — 35·9 3123·5 0·518 0·26
NJS15 70 100 0·203 — 1·0 1·0 1·88 147·4 — — 38·2 3022·3 0·532 0·277
NJS16 50 100 0·150 — 1·0 1·0 2·83 156·4 — — 42·5 2382·5 0·353 0·355
NJS17 50 100 0·170 — 1·0 1·0 2·55 127·7 — — 43·4 2202·4 0·392 0·351
NJS18 50 100 0·185 — 1·0 1·0 2·90 97·9 — — 53·7 1848·2 0·345 0·322
NJS19 50 100 0·211 — 1·0 1·0 2·12 71·6 — — 44·7 1542·8 0·472 0·305
NJS20 50 50 0·210 — 1·0 1·0 3·72 107·1 — — 39·1 1372·6 0·269 1·585
NJS21 50 150 0·215 — 1·0 1·0 1·48 92·6 — — 47·7 3034·1 0·676 0·143
NJS22 50 50 0·186 — 1·0 1·0 5·76 145·2 — — 53·6 1403·9 0·174 1·773
NJS23 50 150 0·187 — 1·0 1·0 1·85 122·3 — — 51·9 3474·0 0·541 0·134
NJS24 70 100 0·215 — 1·0 1·0 2·35 117·5 — — 50·5 2566·0 0·426 0·288
COS1 Coral sands/strain-controlled (in this 30 100 — 0·024 0·5 0·70 — — 10·55 87·8 448·9 4185·1 0·024 0·101
COS2 study) 30 100 — 0·033 0·5 0·79 — — 12·66 105·3 385·9 3586·1 0·033 0·091
COS3 30 100 — 0·047 0·5 0·88 — — 16·98 135·6 362·8 3246·8 0·047 0·080
COS4 45 50 — 0·024 0·5 0·84 — — 8·96 55·5 312·9 2622·9 0·024 0·347
COS5 45 50 — 0·033 0·5 0·91 — — 8·59 56·7 257·2 1953·3 0·033 0·363
COS6 45 50 — 0·047 0·5 0·97 — — 3·94 49·7 218·3 1282·8 0·047 0·305
COS7 60 100 — 0·024 0·5 0·75 — — 14·04 94·2 597·3 4607·0 0·024 0·122
COS8 60 100 — 0·033 0·5 0·83 — — 16·80 117·0 542·1 4108·6 0·033 0·101
COS9 60 100 — 0·047 0·5 0·90 — — 24·1 151·4 511·1 3725·2 0·047 0·091
COS10 45 100 — 0·024 0·5 0·82 — — 11·7 93·4 495·6 4451·0 0·024 0·099
COS11 45 100 — 0·033 0·5 0·89 — — 14·5 108·8 443·2 3760·4 0·033 0·074
COS12 45 100 — 0·047 0·5 0·96 — — 18·9 141·4 404·0 3426·4 0·047 0·200
REF1 Coarse silica sand/stress-controlled 35 40 0·21 — 0·2 0·99 2·25 128·9 — — 18·9 2919·2 0·4437 0·664
(Ghionna & Porcino, 2006)
REF2 Izmir Sand/stress-controlled (Eskisar et al., 30 100 0·18 — 0·1 0·99 2·92 2733·2 — — 52·6 61 000·0 0·3425 0·225
2014)
REF3 River sand/cyclic simple shear (Boulanger 45 207 0·15 — 0·2 0·98 1·88 133·0 — — 58·4 5645·5 0·5319 0·151
& Seed, 1995)
REF4 Izmir silt (FC = 74%)/cyclic simple shear 61 50 0·12 — 0·1 1·0 1·41 46·4 — — 8·4 3157·9 0·7112 0·523
(Momkul et al., 2015)
Continued

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
UNIFIED THIXOTROPIC FLUID MODEL FOR SOIL LIQUEFACTION 855
Dobry et al. 1982; Hazirbaba & Rathje, 2009), it is con-

β: kPa1
venient to adopt the shear strain rate corresponding to ru ¼ 0

0·135

0·128
0·645

0·101

0·257

0·884
as the reference shear strain rate, γ̇r . For stress-controlled
tests, γ̇r can be obtained by equation (9). For strain-controlled
tests, γ̇r ¼ γ̇L . Fig. 7(b) presents the relationship correspond-
γ̇L : s1
0·0006

0·1043
0·0028

0·4949

0·0993

0·6726
ing to the proposed γ̇r and the normalised initial viscosity
coefficient.
As an imaginary fluid, the viscosity coefficient of liquefi-
able soil in the initial equilibrium state is apparently depen-
2690·6
6839·3
111 666·7

63 777·8

33 333·4

684·9
η∞: kPa s
Fitted or calculated parameters

dent on its physical state (i.e. effective confining pressure and


relative density). Fig. 10(a) shows that the initial viscosity
coefficient increases linearly with increasing effective con-
fining pressure when evaluated at the same γ̇r and Dr, whereas
ηe: kPa s

Fig. 10(b) shows the near-parallel linear trend for constant γ̇r
141·7

46·3
5·0

79·3

240·0

27·9
and σc′ . Comparison of Figs 8(b) and 10(b) shows that the
relationship between Dr and viscosity coefficient is approxi-
mately linear for the initial equilibrium state but tends to be
non-linear at the limit equilibrium state. According to the
97·6

27·9


D

state equation for the TEPP fluid model, a larger initial visco-
sity coefficient will require more shearing time to destroy the
0·014

soil fabric. Fig. 10 illustrates that relative density and effective


1·90


C

confining pressure enhance the resistance of soils to lique-


faction within the thixotropic fluid framework, which is
1109·7

consistent with the existing understanding of liquefaction


115·6

551·6

65·7


B

resistance (e.g. Seed & Lee, 1966; Suazo et al., 2016).


2·02

1·38

4·01

2·79


A

Rate equation for the TEPP fluid model


Physical meaning of the rate equation. Because of the
Max. ru

simplification invoked by referencing only the maximum


0·99

0·99
0·89

0·94

0·95
1·0

shear strain rate or shear stress of a given loading cycle to


represent the dynamic process as shown in Fig. 4, the
maximum excess pore pressure and shear strain rate for a
f : Hz
0·01

given cycle rise monotonically. This simplification means


0·1
0·2

0·1

0·5

1·0

that the proposed TEPP fluid model can only capture the
destruction of the soil fabric, and cannot capture its recon-
γ̇con : s1
0·0317

0·0022

struction. Consequently, the parameter b in the reconstruc-


tion rate term of equation (2b) must equal zero. Accordingly,


Loading conditions

equation (2b) can be simplified as



CSR

0·25

0·29

0·30

0·10

¼ c γ̇t ð17Þ

dt
The integral of equation (17) yields
σc′ : kPa
166

90
100

200

200

100

λ ¼ λ0  expðc γ̇tÞ ð18Þ


where λ0 represents the initial inner structural integrity
parameter for t ¼ 0 and λ0 ¼ 1. Equation (18) shows that
Dr: %
73

70
50

92

50

60

the inner structural integrity parameter is a function of γ̇ and


shearing time, t. In the validation of state equation of the
TEPP fluid model, the relationship λ ¼ 1  ru has been
Silty sand/strain-controlled (Hazirbaba &
Toyoura sand/strain-controlled (Jafarian

Silt/stress-controlled (Wang et al., 2013)

validated for stress- and strain-controlled tests. Substitution


A sand–gravel mixture/stress-controlled
Shinias-Marathon/stress-controlled

of λ ¼ 1  ru into equation (17) and rearranging for excess


(Xenaki & Athanasopoulos, 2003)

(Xenaki & Athanasopoulos, 2008)

pore pressure yields for stress-controlled tests with varying γ̇


Sand-silt mixture/stress-controlled

 ðt 
ru;t ¼ 1  exp  cγ̇dt ð19aÞ
(Karim & Alam, 2014)

0
Soil types/test modes

and for strain-controlled tests with constant γ̇


ru;t ¼ 1  expðc γ̇tÞ ð19bÞ
Rathje, 2009)
Natural soil of
et al., 2012)

Equations (19a) and 19(b) form the rate equation for the
TEPP fluid model obtained on the basis that liquefied soil
Table 2. Continued

behaves as a Moore thixotropic fluid. The rate equation


captures the generation of excess pore pressure with shearing
time and shear strain rate, and provides a rational basis for
REF10

treating soil as a thixotropic fluid. Equations (19a) and (19b)


REF5

REF6
REF7

REF8

REF9
Test

can be represented by a unified incremental expression as


no.

follows

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
856 WANG, MA, GAO, STUEDLEIN, HE AND WANG
40 100
Dynamic shear stress,

Dynamic shear stress,


30 75
20 50
10 25
τd: kPa

τd: kPa
0 0
–10 –25
–20 –50
–30 –75
–40 –100
6 2·0
4
Shear strain, γ : %

Shear strain, γ : %
1·0
2
0 0
–2
–1·0
–4
–6 –2·0
0·6 0·06
Shear strain rate, γ⋅ : s–1

Shear strain rate, γ⋅ : s–1


0·5
0·4 0·05
0·3
0·2 0·04
0·1
0 0·03
1·0 1·0
Excess pore pressure

Excess pore pressure


0·8 0·8
ratio, ru

ratio, ru

0·6 0·6
Test NJS21 Test COS21
0·4 0·4
σ'c = 150 kPa σ'c = 100 kPa
0·2 Dr = 50% 0·2 Dr = 45%
CSR = 0·216 CSRmax= 0·95
0 0
0 5 10 15 20 25 30 35 0 4 8 12 16 20 24
Time, t: s Time, t: s
(a) (b)

Fig. 4. Comparison of the cyclic response of sands tested in this study: (a) stress-controlled test of fine Nanjing sand; (b) strain-controlled test of
coarse to medium coral sand

! where τd,j is the maximum shear stress of the jth cycle.


X
i
ru;i ¼ 1  exp c γ̇j Δt ð20Þ Substitution of equation (22) into equation (20) gives
j¼1
for stress  controlled tests and constant τ d ; ru;i
where ru,i is the maximum excess pore pressure ratio at the  
ΔWi
ith cycle; γ̇j is the maximum shear strain rate at the jth cycle; ¼ 1  exp c ð23aÞ
and Δt = 1/f is the loading time for a cycle of loading at τd
frequency, f. for strain  controlled tests and varying τ d ; ru;i
Nemat-Nasser & Shokooh (1979) noted that part of the !
energy dissipated during seismic wave propagation in the ΔWi
ground results from the generation of excess pore water ¼ 1  exp c Pi ð23bÞ
pressure, and that a quantitative relationship between ru and j¼1 τ d;j =i

the energy dissipated per unit volume of the liquefiable In comparison of equations (21), (23a) and (23b), it is
soil could be developed. The energy concept gained interest observed that the rate equation for the TEPP fluid model is
and was further evaluated by numerous researchers (e.g. similar to the energy-based pore pressure generation model
Davis & Berrill, 1982; Dief & Figuero, 2007; Polito et al., proposed by Davis & Berrill (2001). The rate equation thus has
2008; Kokusho, 2013). Consider the non-linear relationship a physical meaning which lends further support to the hypo-
between excess pore pressure ratio and accumulated strain theses that: (a) the selected thixotropic fluid theory is appro-
energy density proposed by Davis & Berrill (2001) priate for liquefied soils, and (b) excess pore water pressure is
  generated by energy that is dissipated under cyclic shearing.
ΔW
ru ¼ 1  exp α ð21Þ
σ′c
where ΔW is the accumulated strain energy per unit volume Validation of the rate equation of the TEPP fluid model.
of soil (J/m3) and α is a dimensionless fitting parameter. Letting c ¼ α  CSR, then equations (21), (23a) and (23b) are
For undrained cyclic triaxial experiments, the accumulated equivalent. Based on equations (23a) and (23b), the excess
strain energy density of soils at ith cycle can be expressed as pore pressure ratio at the ith cycle can be calculated by
follows
Xi ru;i ¼ 1  expðβ ΔWi Þ ð24Þ
ΔWi ¼ τ d;j γ̇j Δt ð22Þ where β ¼ α/σc′ . The undrained cyclic triaxial compression
j¼1
experiments summarised in Table 2 may be used to validate

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
UNIFIED THIXOTROPIC FLUID MODEL FOR SOIL LIQUEFACTION 857
0·6 0·6
Measured stress-controlled Measured stress-controlled
data: NJS1 data: NJS10
0·5 0·5
Shear strain rate, γ⋅ : s–1

Shear strain rate, γ⋅ : s–1


Fitted hyperbolic model Fitted hyperbolic model
(equation (10)) R2 = 0·98 (equation (10)) R2 = 0·99
0·4 0·4

0·3 0·3

0·2 0·2

0·1 0·1

0 0
0 0·2 0·4 0·6 0·8 1·0 0 0·2 0·4 0·6 0·8 1·0

0·6 0·6 Measured stress-controlled


Measured stress-controlled
data: NJS21 data: NJS23
0·5 0·5
Shear strain rate, γ⋅ : s–1

Shear strain rate, γ⋅ : s–1


Fitted hyperbolic model Fitted hyperbolic model
(equation (10)) R2 = 0·99 (equation (10)) R2 = 0·97
0·4 0·4

0·3 0·3

0·2 0·2

0·1 0·1

0 0
0 0·2 0·4 0·6 0·8 1·0 0 0·2 0·4 0·6 0·8 1·0

0·6 0·6
Measured strain-controlled Measured stress-controlled
data: REF3 data: REF10
0·5 0·5
Shear strain rate, γ⋅ : s–1

Shear strain rate, γ⋅ : s–1

Fitted hyperbolic model Fitted hyperbolic model


(equation (10)) R2 = 0·98 (equation (10)) R2 = 0·97
0·4 0·4

0·3 0·3

0·2 0·2

0·1 0·1

0 0
0 0·2 0·4 0·6 0·8 1·0 0 0·2 0·4 0·6 0·8 1·0
1 – ru 1 – ru

Fig. 5. Comparison of the shear strain rate γ̇ and 1 − ru for the stress-controlled CTC and CSS test data and the assumed hyperbolic relationship

125
COS9
τd = 148λ + 25·1
Dynamic shear stress, τd: kPa

100 τd = 117λ + 16·8 REF5 R2 = 0·98


R2 = 0·99
REF7
75
τd = 88λ + 10·5 τd = 98λ + 1·7
R2 = 1·0 R2 = 0·99
50
τd = 55λ + 7·9 τd = 27·9λ
R2 = 0·98 R2 = 0·96
25

COS1 COS5 COS8


0
0 0·2 0·4 0·6 0·8 1·0 0 0·2 0·4 0·6 0·8 1·0
1 – ru 1 – ru

Fig. 6. Comparison of the dynamic shear stress τd and 1 − ru for the strain-controlled CTC and CSS test data and the assumed linear relationship

equation (24). Parameter β varies from 0·074 to 2·381, as Fig. 11 compares the calculated and measured excess pore
determined by fitting equation (24) to experimental data pressure ratios showing that, for both stress-controlled and
using OLS, as shown in Table 2 for each case considered. strain-controlled experiments, the calculated generation of

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
858 WANG, MA, GAO, STUEDLEIN, HE AND WANG
100·00 10 000
ηe
= 0·235 γ⋅L–0·71 NJS1–NJS24
σ'c Stress-
Normalised viscosity coefficient: s R2 = 0·92 controlled REF1, 2, 6, 8–10
10·00 1000 COS1–COS12
Strain-
controlled REF5

1·00 100

Shaking table tests


0·10 10
NJS1–NJS24
COS1–COS12
REF1–REF10
0·01 1
0·0001 0·0010 0·0100 0·1000 1·0000 10·00 0·0001 0·0010 0·0100 0·1000 1·00
Shear strain rate at moment of liquefaction, γ⋅ : s–1
L Reference shear strain rate, γ⋅r: s–1
(a) (b)

Fig. 7. Relationship between the: (a) normalised viscosity coefficient, ηe/σc′ , and the shear strain rate at the moment of liquefaction, γ̇L ;
(b) normalised viscosity coefficient, ηe/σc′ , and the reference shear strain rate, for stress- and strain-controlled CTC and CSS tests

600
γ⋅L ≈ 0·024 s−1
Viscosity coefficient at moment of

γ⋅ ≈ 0·032 s−1
liquefaction, ηe: kPa s

500 L

γ⋅L ≈ 0·047 s−1

400

γ⋅L ≈ 0·024 s−1


300
γ⋅ ≈ 0·032 s−1
L

γ⋅L ≈ 0·047 s−1


200
40 60 80 100 120 20 30 40 50 60 70
Effective confining pressure, σ'c: kPa Relative density, Dr: %
(a) (b)

Fig. 8. Effect of initial conditions on the viscosity coefficient, ηe, at the moment of liquefaction observed during strain-controlled tests of medium
to coarse coral sand: (a) effective confining pressure; (b) relative density

excess pore pressure ratio is in good agreement with that 0·4


measured and corresponds to R 2 values greater than 0·95.
CSR = 0·235γ⋅L−0·29
Cyclic stress ratio, CSR

It may be concluded that the energy-based concept for


0·3 0·265
the generation of excess pore pressure fits within the TEPP
fluid model for liquefiable soils. Furthermore, differences ΔCSR = 0·078
0·187
in the characteristic shapes between the stress- and strain- 0·2
controlled experiments in Fig. 11 indicate the role of shear
strain rate in controlling the rate of excess pore pressure 0·1 Δτd = ΔCSRσ'c
generation; thus, the TEPP fluid model offers an improve- Δτd = 3·9 kPa, σ'c = 50 kPa
ment over typical energy-based pore pressure generation Δτd = 7·8 kPa, σ'c = 100 kPa
models (e.g. Berrill & Davis, 1985; Davis & Berrill, 2001; 0
0 0·5 1·0 1·5 2·0 2·5 3·0
Polito et al., 2008).
Shear strain rate at moment of liquefaction, γ⋅L: s–1
Forward modelling using the proposed rate equation (i.e.
equations (23a) and (23b)) requires that parameter β be
Fig. 9. Effect of shear strain rate at initial liquefaction on the cyclic
suitably estimated using experimental data and that it follows stress ratio and illustration of pressure-dependent changes in dynamic
established trends in soil mechanics. Fig. 12(a) presents the shear stress
variation of β with effective confining pressure for the
stress-controlled data in Table 2, along with a fitted trend
for the 24 Nanjing sands, which may be described by in Fig. 12(c), the strain-controlled experiments on the coral
β ¼ 1200σ′175
c ð25Þ sands indicate that the initial relative density has a similar
influence on β and that β decreases linearly with increasing γ̇
The scatter associated with the general variation of β with for the range in γ̇ investigated. It may be concluded that
effective confining pressure indicates that other factors parameter c in the rate equation effectively describes the rate
contribute to this destruction rate parameter, as may be of soil fabric destruction, which is in turn related to the rate
expected. Fig. 12(b) presents the variation in β with relative of excess pore pressure generation, described by parameter β,
density for the Nanjing sands and for a confining pressure of and both may be used within the Moore-type thixotropic
100 kPa; clearly, the state of the soil correlates to β. As shown fluid model proposed for liquefiable soils.

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
UNIFIED THIXOTROPIC FLUID MODEL FOR SOIL LIQUEFACTION 859

γ⋅L ≈ 0·024 s−1


5000 γ⋅L ≈ 0·024 s−1
Initial viscosity coefficient, η∞: kPa s
0·032 s−1
4000 0·032 s−1
0·047 s−1
0·047 s−1 0·0078 s−1
3000
0·008 s−1
0·0085 s−1

2000
COS, Dr = 45%
COS, σ'c = 100 kPa
1000
NJS, Dr = 50%
NJS, Dr= 30% NJS, σ'c = 100 kPa
0
0 50 100 150 200 0 20 40 60 80 100
Effective confining pressure, σ'c: kPa Relative density, Dr: %
(a) (b)

Fig. 10. Variation of the initial viscosity coefficient with (a) effective confining pressure and (b) relative density, for a given shear strain rate at the
moment of liquefaction

1·0
R2 = 1·00
R2 = 0·98
1·00
0·8 1·00
Excess pore pressure ratio, ru

1·00 0·99
0·98
0·99
0·6

0·4

COS1
NJS5
NJS9 COS5
0·2
NJS19 COS10

REF1 REF7
0
0 5 10 15 20 25 30 35 1 10 100
Number of cycles, N Number of cycles, N
(a) (b)

Fig. 11. Comparison of measured and calculated generation of excess pore pressure with loading cycles for (a) stress-controlled and
(b) strain-controlled tests

DISCUSSION (b) Differentiating from traditional constitutive


This and numerous other studies have shown that the loss equations of liquefiable soils, the proposed
of stiffness and strength is associated with the generation of constitutive equation under the TEPP framework is
excess pore pressure and the onset and sustained state of strain rate-dependent during the entire loading period.
liquefaction. The experimental data presented in this study The increased significance of the rate dependence
and reported by others have shown that the treatment of the with increased excess pore pressure is effectively
liquefiable soil as a thixotropic fluid is fully consistent with captured, and overcomes this difficulty in typical
previous findings. Further, this work shows that TEPP may constitutive models that are invoked within
be linked to dissipated energy in a manner that accounts for a continuum.
the soil state and shear strain rate. The advantages of (c) The proposed rate equation (i.e. equations (23a)
modelling liquefiable soil as a thixotropic fluid within the and (23b)) shows that the excess pore pressure ratio
constitutive equations for the proposed TEPP model are of liquefiable soils can approach but never achieve
evident. 100%. In addition, equations (23a) and (23b) show
that even if the shear strain rate or cyclic shear stress
(a) The constitutive equations for TEPP, given by equations is small, the excess pore pressure ratio will still approach
(13), (23a) and (23b), present a unified description of 100% if the time of shearing is sufficient. Recent
the dynamic response of soils before and during long-duration earthquakes have shown that severe
liquefaction, which distinguishes this approach from liquefaction can manifest at sites with epicentral
most of the existing dynamic constitutive models for soil distances greater than 300 km or with the seismic
liquefaction analysis. The inner structural integrity intensity below VI (Tshukamoto et al., 2012;
parameter in the rate equation tends to zero during Verdugo, 2012).
cyclic loading and the state equation transforms into a (d ) The TEPP fluid model provides a new, physically
power-law fluid equation upon liquefaction. justified framework for modelling the process of soil

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
860 WANG, MA, GAO, STUEDLEIN, HE AND WANG
3·0 CONCLUSIONS
Model fitted to NJS
Pore pressure generation parameter,
β = 1200σ'c–1·75
NJS This paper presents a theoretical and experimental basis
2·5 R2 = 0·81 REF1 for treating liquefiable soils within the framework of a non-
REF2
Newtonian thixotropic fluid, in which strain rate dependence
may be suitably captured before and during liquefaction. The
2·0 REF3 state and rate equations for traditional Moore-type thixo-
tropic fluids were validated using cyclic, stress- and strain-
β: kPa–1

REF4
1·5
REF6 controlled experiments representing a broad range of soil
types and stress paths. Specifically, this study demonstrated
REF8
1·0 the following points.
REF9

0·5
REF10 (a) Soils may be described using an inner structural
integrity parameter that serves as a proxy for the
stability of the soil fabric and is related to the excess pore
0 pressure generated. The proposed rate equation can
0 50 100 150 200 250
capture the accumulation of excess pore pressure
Effective confining pressure, σ'c: kPa
triggering the progressive failure of inner structure
(a)
of liquefiable soil, and is similar, and superior,
0·8 to established energy-based pore pressure generation
NJS, σ'c = 100 kPa models.
(b) The constitutive equation of the proposed TEPP model
Pore pressure generation parameter,

includes three rate-dependent parameters that have clear


0·6 physical justification. These parameters include: the
β = 8·3Dr–0·81
viscosity coefficient in the initial equilibrium state,
R2 = 0·97
η∞; the viscosity coefficient in the limit equilibrium
β: kPa–1

0·4 state, ηe; and the rate coefficient of structural damage,


β, which can be obtained by conventional undrained
cyclic triaxial or simple shear experiments. These
parameters were shown to be functions of relative
0·2 density and initial effective confining pressure (hence
state), as well as shear strain rate. The parameters have
been tabulated for use in forward modelling efforts.
0 (c) The treatment of liquefiable soils as a thixotropic fluid
10 20 30 40 50 60 70 80 that obeys established soil mechanics principles provides
Relative density, Dr: % a new and reliable framework for the unified modelling
(b) of liquefaction triggering and its consequences.
0·4 The physically justified and experimentally validated thixo-
σ'c = 50 kPa tropic fluid model presented here provides a promising new,
Pore pressure generation parameter,

Dr = 45% unified approach to the critical problem of earthquake-


0·3 σ'c = 100 kPa
induced liquefaction and its consequences. Nonetheless, the
Dr = 30% experimental data developed for this study and supplemented
from the literature are limited and further effort is needed
σ'c = 100 kPa
β: kPa–1

Dr = 45%
to demonstrate the model’s applicability to other types of
0·2 liquefiable soils and conditions. Furthermore, the influence
σ'c = 100 kPa of fines content or particle size distribution on the constitu-
Dr = 60%
tive model parameters has not been taken into consideration
0·1 at this time. Clearly, additional work to characterise factors
contributing to the viscous response of soils is warranted.

0
0 0·01 0·02 0·03 0·04 0·05 ACKNOWLEDGEMENTS
Shear strain rate, γ⋅ : s–1 The authors gratefully acknowledge the financial support
(c)
for this study from the National Natural Science Foundation
of China (grant no. 51378257; grant no. 51678300). The third
Fig. 12. Comparison of the pore pressure parameter β with author was supported by the National Science Foundation
(a) effective confining pressure for stress-controlled tests, (b) relative (grant CMMI 1663654) under program director Dr Rick
density and (c) shear strain rate for strain-controlled tests Fragaszy on related studies of soil liquefaction while con-
ducting this work; this support is gratefully acknowledged.
Special thanks are owed to Dr Sun Yifei at Hohai University
for his help and contributions to the preparation of the paper.
liquefaction. Moreover, the continuous dynamic
response (e.g. deformation) of liquefiable soils
before and after liquefaction can be realised by placing
NOTATION
the proposed constitutive equations within a A, B fitting parameters determined using ordinary least squares
computational fluid dynamics framework, shown to a state coefficient
present distinct advantages (e.g. Huang et al., 2012; b parameter denoting structure reconstruction
Jafarian et al., 2014) to capture large flow-type C, D intercept and slope of the fitted line, respectively
deformations. c parameter denoting structure destruction

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
UNIFIED THIXOTROPIC FLUID MODEL FOR SOIL LIQUEFACTION 861
cc curvature coefficient Davis, R. O. & Berrill, J. B. (2001). Pore pressure and dissipated
cu uniformity coefficient energy in earthquakes field verification. J. Geotech. Geoenviron.
Dr relative density Engng 127, No. 3, 269–274.
d50 mean diameter (mm) De Alba, P. & Ballestero, T. P. (2006). Residual strength after
emax maximum void ratio liquefaction: a rheological approach. Soil Dynamics Earthquake
emin minimum void ratio Engng. 26, No. 2, 143–151.
f loading frequency (Hz) Dewoolkar, M., Hargy, J., Anderson, I., Alba, P. D. & Olson, S. M.
Gs specific gravity (2016). Residual and post-liquefaction strength of a liquefiable
k, n constant parameters determined by the test sand. J. Geotech. Geoenviron. Engng 142, No. 2, 04015068.
ru excess pore pressure ratio Dief, H. M. & Figuero, J. L. (2007). Liquefaction assessment by the
ru,t excess pore pressure ratio at time t unit energy concept through centrifuge and torsional shear tests.
t shearing time (s) Can. Geotech. J. 44, No. 11, 1286–1297.
α dimensionless fitting parameter for strain energy density Dobry, R., Ladd, R. S., Yokel, F. Y., Chung, R. M. & Powell, D.
equation (1982). Prediction of pore water pressure buildup and liquefaction
β fitting parameter (kPa1) of sands during earthquakes by the cyclic strain method. NBS
γ̇ shear strain rate (s1) Building Science series 138, Report No. PB83111617.
γ̇i shear strain rate at time ti (s1) Washington, DC, USA: National Bureau of Standards,
γ̇j maximum shear strain rate of jth cycle (s1) Department of Commerce.
γ̇L shear strain rate in the limit equilibrium state (s1) Eskisar, T., Karakan, E. & Altun, S. (2014). Evaluation of cyclic
γ̇Ni maximum shear strain rate at the Nith cycle (s1) stress–strain and liquefaction behavior of Izmir sand.
γ̇r reference shear strain rate (s1) Arab. J. Sci. Engng 39, No. 11, 7513–7524.
Δt shearing time of each cycle (s) Ghionna, V. A. & Porcino, D. (2006). Liquefaction resistance of
ΔW accumulated strain energy per unit volume of soil (J/m3) undisturbed and reconstituted samples of a natural coarse sand
η viscosity coefficient (kPa s) from undrained cyclic triaxial tests. J. Geotech. Geoenviron.
ηe viscosity coefficient for the fully destructured fluid (kPa s) Engng 132, No. 2, 194–202.
ηNi maximum viscosity coefficient at the Nith cycle (kPa s) Hadush, S., Yashima, A., Uzuoka, R., Moriguchi, S. & Sawada, K.
η∞ viscosity coefficient in the initial equilibrium state (kPa s) (2001). Liquefaction induced lateral spread analysis using
λ scalar parameter expressing instantaneous degree of the CIP method. Comput. Geotech. 28, No. 8, 549–574.
thixotropic structure Hamada, M. & Wakamatsu, K. (1998). A study on ground
λ0 structural parameter for soils in the initial state displacement caused by soil liquefaction. Proc. Jap. Soc. Civ.
λe structural parameter for soils in the equilibrium state Engrs 596, 189–208.
σc′ initial effective confining pressure (kPa) Hazirbaba, K. & Rathje, E. M. (2009). Pore pressure generation of
σd axial dynamic stress (kPa) silty sands due to induced cyclic shear strains. J. Geotech. Engng
τ shear stress (kPa) 135, No. 12, 1892–1905.
τd dynamic shear stress amplitude (kPa) Huang, Y., Mao, W. W., Zheng, H. & Li, G. (2012). Computational
τd,j maximum shear stress of jth cycle (kPa) fluid dynamics modeling of post-liquefaction soil flow using the
τe shear stress in the limit equilibrium state (kPa) volume of fluid method. Bull. Engng. Geol. Environ. 71, No. 2,
τ Ni maximum shear stress at the Nith cycle (kPa) 359–366.
Hwang, J. I., Kim, C. Y., Chung, C. K. & Kim, M. M. (2006).
Viscous fluid characteristics of liquefied soils and behavior
of piles subjected to flow of liquefied soils. Soil Dynamics
Earthquake Engng 26, No. 2, 313–323.
REFERENCES Ishihara, K. (1993). Liquefaction and flow failure during earth-
Berrill, J. B. & Davis, R. O. (1985). Energy dissipation and seismic quakes. Géotechnique 43, No. 3, 351–415, https://doi.org/
liquefaction of sands: revised model. Soils Found. 25, No. 2, 10.1680/geot.1993.43.3.351.
106–118. Jafarian, Y., Towhata, I., Baziar, M. H., Noorzad, A. &
Boulanger, R. W. & Idriss, I. M. (2014). CPT and SPT based on Bahmanpour, A. (2012). Strain energy based evaluation of
liquefaction triggering procedures, Report UCD/CGM-14/01. liquefaction and residual pore water pressure in sands using
Davis, CA, USA: Center for Geotechnical Engineering, cyclic torsional shear experiments. Soil Dynamics Earthquake
University of California. Engng 35, No. 4, 13–28.
Boulanger, R. W. & Idriss, I. M. (2015). CPT-based liquefaction Jafarian, Y., Ghorbani, A. & Ahmadi, O. (2014). Simplified
triggering procedure. J. Geotech. Geoenviron. Engng 142, No. 2, dynamic analysis to evaluate liquefaction-induced lateral defor-
04015065. mation of earth slopes: a computational fluid dynamics
Boulanger, R. W. & Seed, R. B. (1995). Liquefaction of sand under approach. Earthquake Engng Engng Vibr. 13, No. 3, 555–568.
bidirectional monotonic and cyclic loading. J. Geotech. Engng Karim, M. E. & Alam, M. J. (2014). Effect of non-plastic silt content
121, No. 12, 870–878. on the liquefaction behavior of sand–silt mixture. Soil Dynamics
Bray, J. D. & Travasarou, T. (2007). Simplified procedure for Earthquake Engng. 65, 142–150.
estimating earthquake-induced deviator slope displacements. Kokusho, T. (2013). Liquefaction potential evaluations: energy-
J. Geotech. Geoenviron. Engng 133, No. 4, 381–392. based method versus stress-based method. Can. Geotech. J. 50,
Chen, G. X., Zhou, E. Q., Wang, Z. H., Wang, B. H. & Li, X. J. No. 10, 1088–1099.
(2016). Experimental investigate on fluid characteristics of Konagai, K., Kiyota, T., Suyama, S., Asakura, T., Shibuya, K. &
medium dense saturated fine sand in pre- and post-liquefaction. Eto, C. (2013). Maps of soil subsidence for Tokyo bay shore
Bull. Earthquake Engng. 14, No. 8, 2185–2212. areas liquefied in the March 11th, 2011 off the Pacific Coast of
Cheng, C. H. (1973). A differential form of constitutive relation for Tohoku Earthquake. Soil Dynamics Earthquake Engng. 53,
thixotropy. Rheol. Acta. 12, No. 2, 228–233. No. 10, 240–253.
Cubrinovski, M., Winkley, A., Haskell, J., Palermo, A., Makdisi, F. I. & Seed, H. B. (1979). Simplified procedures for
Wotherspoon, L., Robinson, K., Bradley, B., Brabhaharan, P. evaluating embankment response. ASCE J. Geotech. Engng 105,
& Hughes, M. (2014). Spreading-induced damage to short-span No. GT12, 1427–1434.
bridges in Chrischurch, New Zealand. Earthq. Spectra. 30, Marcuson, W. F. (1978). The committee on soil dynamics of the
No. 1, 57–83. geotechnical engineering division: definition of terms related
Dafalias, Y. F. & Manzari, M. T. (2004). Simple plasticity sand to liquefaction. ASCE J. Geotech. Engng 104, No. GT9,
model accounting for fabric change effects. ASCE J. Engng 1197–1200.
Mech. 130, No. 6, 622–634. Meris, J. (1979). Thixotropy – a general review. J. Non-Newton. Fluid
Davis, R. O. & Berrill, J. B. (1982). Energy dissipation and seismic 6, No. 1, 1–20.
liquefaction in sands. Earthquake Engng Struct. Dynamics 10, Momkul, M. M., Gultekin, C., Gülver, M., Akın, Ö. &
No. 1, 59–68. Eseller-Bayat, E. (2015). Estimation of liquefaction potential

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.
862 WANG, MA, GAO, STUEDLEIN, HE AND WANG
from dry and saturated sandy soils under drained constant Towhata, I., Vargas-Monge, W., Orense, R. P. & Yao, M. (1999).
volume cyclic simple shear loading. Soil Dynamics Earthquake Shaking table tests on subgrade reaction of pipe embedded in
Engng 75, 27–36. sandy liquefied subsoil. Soil Dynamics Earthquake Engng 18,
Nemat-Nasser, S. & Shokooh, A. (1979). A unified approach to No. 5, 347–361.
densification and liquefaction of cohesionless sand in cyclic Towhata, I., Anh, T. T. L., Yamada, S., Motamed, R. &
shearing. Can. Geotech. J. 16, No. 4, 659–678. Kobayashi, Y. (2010). Zero-gravity triaxial shear tests on
Newmark, N. M. (1965). Effects of earthquakes on dams and mechanical properties of liquefied sand and performance
embankments. Géotechnique 15, No. 2, 138–159, https://doi.org/ assessment of mitigations against large ground deformation.
10.1680/geot.1965.15.2.139. In Proceedings of the 5th international conference on recent
Nishimura, S., Towhata, I. & Honda, T. (2002). Laboratory shear advances in geotechnical earthquake engineering and soil
tests on viscous nature of liquefied sand. Soils Found. 42, No. 4, dynamics, paper OSP 3. Rolla, MO, USA: Missouri University
89–98. of Science and Technology.
Polito, C. P., Green, R. A. & Lee, J. (2008). Pore pressure generation Tshukamoto, Y., Kawabe, S. & Kokusho, T. (2012). Soil liquefaction
models for sands and silty soils subjected to cyclic loading. observed at the lower stream of Tonegama river during the 2011
J. Geotech. Geoenviron. Engng 134, No. 10, 1490–1500. off the Pacific Coast of Tohoku earthquake. Soils Found. 52,
Poulos, S. J., Castro, G. & France, J. W. (1985). Liquefaction eval- No. 5, 987–999.
uation procedure. ASCE J. Geotech. Engng 111, No. 6, 772–792. Uzuoka, R., Yashima, A., Kawakami, T. & Konrad, J. M.
Price, A. B., Dejong, J. T. & Boulanger, R. W. (2017). Cyclic loading (1998). Fluid dynamics based prediction of liquefaction
response of silt with multiple loading events. J. Geotech. induced lateral spreading. Comput. Geotech. 22, No. 3,
Geoenviron. Engng 143, No. 10, 04017080. 243–282.
Sawicki, A. & Mierczynski, J. (2009). On the behavior of liquefied Verdugo, R. (2012). Comparing liquefaction phenomena observed
soil. Comput. Geotech. 36, No. 4, 531–536. during the 2010 Maule, Chile earthquake and 2011 Great East
Saygili, G. & Rathje, E. M. (2008). Empirical predictive models for Japan earthquake. Proceedings of the international symposium
earthquake-induced sliding displacements of slopes. J. Geotech. on engineering lessons learned from the 2011 great east Japan
Geoenviron. Engng. 134, No. 6, 790–803. earthquake, Tokyo, Japan, pp. 707–718.
Seed, H. B. (1979). Soil liquefaction and cyclic mobility evaluation Wang, S. Y., Yang, J. S. & Onyejekwe, S. (2013). Effect of
for level ground during earthquakes. J. Geotech. Engng 105, previous cyclic shearing on liquefaction resistance of
No. 2, 201–255. Mississippi River valley silt. J. Mater. Civ. Engng 25, No. 10,
Seed, H. B. & Lee, K. L. (1966). Liquefaction of saturated sands 1415–1423.
during cyclic loading. ASCE J. Soil Mech. Found. Div. 92, Wang, R., Zhang, J. M. & Wang, G. (2014). A unified plasticity
No. SM6, 105–134. model for large post-liquefaction shear deformation of sand.
Seed, H. B., Tokimatsu, K., Harder, L. F. & Chung, R. M. (1985). Comput. Geotech. 59, No. 3, 54–66.
Influence of SPT procedures in soil liquefaction resistance Xenaki, V. C. & Athanasopoulos, G. A. (2003). Liquefaction
evaluations. J. Geotech. Engng 111, No. 12, 1425–1445. resistance of sand–silt mixtures: an experimental investigation
Suazo, G., Fourie, A., Doherty, J. & Hasan, A. (2016). Effects of of the effect of fines. Soil Dynamics Earthquake Engng. 23,
confining stress, density and initial static shear stress on the No. 3, 1–12.
cyclic shear response of fine-grained unclassified tailings. Xenaki, V. C. & Athanasopoulos, G. A. (2008). Dynamic properties
Géotechnique 66, No. 5, 401–412, https://doi.org/10.1680/jgeot. and liquefaction resistance of two soil materials in an earthfill
15.P.032. dam-laboratory test results. Soil Dynamics Earthquake Engng.
Tasiopoulou, P. & Gerolymos, N. (2016). Constitutive modeling of 28, No. 8, 605–620.
sand: formulation of a new plasticity approach. Soil Dynamics Yang, Z., Elgamal, A. & Parra, E. (2003). Computational model for
Earthquake Engng 82, 205–221. cyclic mobility and associated shear deformation. J. Geotech.
Tokimatsu, K., Tamura, S., Suzuki, H. & Katsumata, K. (2012). Geoenviron. Engng 129, No. 12, 1119–1127.
Building damage associated with geotechnical problems in the Youd, T. L. & Idriss, I. M. (2001). Liquefaction resistance of soils:
2011 Tohoku Pacific Earthquake. Soils Found. 52, No. 5, summary report from the 1996 NCEER and 1998 NCEER/NSF
956–974. workshops on evaluation of liquefaction resistance of soils.
Towhata, I., Sassaki, Y., Tokida, K., Matsumoto, H., Tamari, Y. & J. Geotech. Geoenviron. Engng 127, No. 4, 297–313.
Yamada, K. (1992). Prediction of permanent displacement of Zlatovic, S. & Ishihara, K. (1997). Normalized behavior of very
liquefied ground by means of minimum energy principle. Soils loose non-plastic soils: effects of fabric. Soils Found. 37, No. 4,
Found. 32, No. 3, 97–116. 47–56.

Downloaded by [ Indian Institute Of Technology - Kharagpur] on [07/05/21]. Copyright © ICE Publishing, all rights reserved.

You might also like