Doboku Gakkai Ronbunshuu C Vol.65 No.1, 151-161, 2009.

EXPERIMENTAL VALIDATION OF A
NUMERICAL MODEL: REVERSE FAULT
RUPTURE PROPAGATION THROUGH SAND

Md. ROKONUZZAMAN1, Toshinori SAKAI2, Ala’a El NAHAS3,

Tadatsugu TANAKA4 and Md. Zakaria HOSSAIN5
1
Ph.D. Student, Dept. of Env. Sc. and Tech., Grad. School of Bioresources, Mie University
(1577 Kurimamachiya, Tsu, Mie 514-8507, Japan)
E-mail: rokoncekuet@yahoo.com
2
Member of JSCE, Professor, Dept. of Env. Sc. and Tech., Grad. School of Bioresources, Mie University
(1577 Kurimamachiya, Tsu, Mie 514-8507, Japan)
E-mail: sakai@bio.mie-u.ac.jp
3
Principal Geotechnical Engineer, Haskoning UK Ltd., U.K.
(17, Hornbeam Drive, Tile Hill, Coventry CV4 9UJ, U.K.)
E-mail : elnahas24@hotmail.com
4
Member of JSCE, Professor, Dept. of Bio. and Env. Eng., Grad. School of Agr. and Life Science, University of Tokyo
(1-1-1 Yayoi, Bunkyo-ku, Tokyo 113-8657, Japan)
E-mail: atanak@mail.ecc.u-tokyo.ac.jp
5
Associate Professor, Dept. of Env. Sc. and Tech., Grad. School of Bioresources, Mie University
(1577 Kurimamachiya, Tsu, Mie 514-8507, Japan)
E-mail: zakaria@bio.mie-u.ac.jp

In this study, a sophisticated numerical model incorporating a hardening-softening constitutive model

with shear band is calibrated from the direct shear test results, and validated for the prediction of the be-
haviour of medium dense Fotainebleau sand bed for quasi-static displacement induced by reverse fault’s
base rock with dip angle of 60°. The vertical displacements profile of the ground surface, minimum vertical
base displacement for the rupture to reach the ground, the average dip angle propagated into the soil as well
as the horizontal extent of the deformed surface ground due to different stress fields (centrifuge and 1g
tests) are evaluated by having close agreement between experiments and numerical analyses.

Key Words: sand deposit, reverse fault, elasto-plastic, progressive failure, shear band, finite element,
mesh size, scale effect

1. INTRODUCTION
In recent years, the localization of deformation
into shear bands or shear zones has received much
attention. The proper understanding of shear band
mechanism provides a useful basis for soil-structure
interaction problems. The failure of sand mass is,
usually, progressive in nature. It is related to the
development of a shear band of localized deforma-
tion. It is very important for the geotechnical engi-
neers to understand the fault rupture propagation
through overlaying sand mass to site and design
structure near or across fault plane.
Cole & Lade1) performed extensive sandbox tests
using dry loose and dense sands. They predicted the
shape of the failure surface over dip-slip fault as a
function of the depth of the soil, the angle of dilation,
and the dip angles. The downsides to this approach
are the introduction of adhoc assumptions for the
shape of failure surface, overlooking the progressive
failure behavior of soil. Scott & Schoustra2) per-
formed numerical simulation of 800 m-deep soil
mass over vertical fault by two-dimensional finite
element method (FEM), assuming a linear-perfectly
plastic relation. They overlooked the hardening sof-
tening nature of the soil. Consequently, their results
showed the rupture zone bending over to the
up-thrown side, which was not consistent with ex-

151

perimental results. Roth et al.3) compared the cen-
trifuge tests and the shear rupture in 6m deposits with
their finite difference simulation. They concluded
that the simulation could duplicate the experiments
only qualitatively. Walters & Thomas4) performed
sandbox experiment and numerical simulation of
their experiment by FEM. They found that
non-associated flow rule and strain softening were
essential in modeling the location, development, and
propagation of localized failure surfaces in the gra-
nular material. But in their FE analysis, failure sur-
faces propagated through the sand and broke the
ground surface with only a fraction of the displace-
ment observed in experiments. Bray et al.5) per-
formed FE analyses and compared the results with
the clay-box experiments and anchor pullout expe-
riments. The analogous trapdoor and anchor prob-
lems with sand deposits were modeled using FEM by
Sakai & Tanaka6) and Tanaka & Sakai7). Where, they
incorporated the strain hardening-softening behavior
of sand with shear band effect.
In this study, a sophisticated numerical modeling
procedure is discussed and verified for its usefulness
in direct shear tests with medium dense Fontaineb-
leau sand. Then, the objective failure mechanisms
and soil deformation patterns of the overlaying me-
dium dense Fontainebleau sand deposit over a 60°
dip angled reverse fault are evaluated by comparing
the results of a conventional 1g-model test with those
of numerical analyses. Afterward, the scale effect is
also evaluated between the 1g and 115g centrifuge
tests using the results obtained by this numerical
model due to the difference in stress level.
2. TESTING PROCEDURES
For the experiment of fault rupture propagation
through Fontainebleau sand (D50=0.24mm, U0=1.33,
Gs=2.59, emax=0.833, emin=0.55, fines content=0%)
deposit, University of Dundee’s beam centrifuge was
used. The strong box internal model dimensions were
800 mm × 500 mm × 500 mm (Fig. 1), with front and
back transparent Perspex plates, through which the
models were monitored during the tests. Two hy-
draulic cylinders were used to push the hanging or
right part up or down to simulate reverse and normal
faulting. A central guidance (G) and three wedges
(A1-A3) were used to guide the imposed displace-
ment at the desired dip angle (60°, Fig. 1). Sand was
pluviated in the strong box on 20-30 mm thick layers
to fill up to desired depth. On top of each layer, a line
of dyed sand was laid behind each Perspex wall to
clearly visualize the shear bands. The corner and
internal cans were placed to verify the sand unit
weights inside the strong box and near the edges at
152
Doboku Gakkai Ronbunshuu C Vol.65 No.1, 151-161, 2009. 2
the bottom and in the middle of the model ground. A
series of digital images were taken for the displaced
model ground after each stepwise fault dip slip of
about 0.5 to 1.5 mm (time: 10 to 30 seconds) till to
the total machine-allowable dip slip or maximum
vertical base dislocation (hmax, in Table 1). The dis-
placement vectors and shear strains in the model
ground were analyzed using the deformation mea-
surement system (Geo-PIV program of White et al.8)).
In addition, linearly variable differential transfor-
mers (LVDTs) were used to monitor the vertical
settlements of the model ground surface, and the
vertical component of the base dislocation (h, shown
in Fig. 2). All the other definitions of the physical
model used in this work are shown in Fig. 2. The
strong box was mounted on the centrifuge, and spun
to the predetermined g-levels. The prototype dimen-
sions and parameters used in the experiments are
given in Table 1. The detailed technical description
of this facility and the testing procedures can be
found in El Nahas et al.9), 10).
3. NUMERICAL MODELING
This FE model uses an elasto-plastic framework
with non-associated flow rule and strain harden-
ing/softening law. An explicit dynamic relaxation
method11) is used for the solution of the nonlinear
equations.
The modeling of the materials having softening
properties is full of serious difficulties both in mod-
eling strain localization and from the view point of
numerical analysis. The straightforward use of the
material softening model in a classical continuum,
generally, does not result in a well-posed problem.
The standard finite element solution of strain loca-
lization in a rate-dependent material results in solu-
tions that is strongly mesh-sensitive. Higher order
constitutive models can solve this problem: viscop-
lastic model12), non local theory13), gradient elas-
to-plastic model14), otherwise, Gudehus & Nubel15),
showed the size of elements has to be in the order of
3D50. Such fine mesh size prohibits the rigorous ap-
plication of FE method to real-scaled problems. So,
objectively, the shear band effect is introduced into
the constitutive equation. The shear band effect
Table 1 Prototype dimensions and basic parameters for the
centrifuge experiments.
Test Centrif. Dr H L W hmax
name accel. (%) (m) (m) (m) (m)
Test8_1g_R 1g 60 0.22 0.66 0.21 0.03
Test 8 115g 60.9 25.3 75.9 24.2 2.56
 Doboku Gakkai Ronbunshuu C Vol.65 No.1, 151-161, 2009. 2

660 in stresses can cause an associated change in the

elastic strain given by
dε e = D −1 (σ B − σ A ) (3)
where, D is the elasticity matrix.
In the present study, it is also assumed that the
570 deformation of a given sand element under uniform
boundary stress conditions is homogeneous in the
600
pre-peak regime, and the strain localization in a shear
band starts suddenly at the peak stress state. The rate
70
of post-peak strain softening associated with shear
50 banding depends on the strain localization parameter
900 S which is a function of the shear band width. The
Fig.1 Basic dimensions of the experimental apparatus shear band thickness is known to be about 16-30
installed in the Dundee University centrifuge (not to times the mean particle diameter (D50) 19), 20). As the
scale and all dimensions are in mm). total strain does not change during the relaxation
process, and, thus, the plastic strain change is ba-
Tracing line of main
lanced by an equal and opposite change in the elastic
Boundary of soil
surface deformation shear band strain:
O′+ d δy Sdε p = −dε e = −D −1 (σ B − σ A ) (4)
(3)
The plastic strain increments can be obtained by
(2) Displaced soil the following equation:
H

θmax (1) ∂Ψ
h

Stationary soil △
O(focus)
Hanging dε p = λ (5)
Footwall Dip angle=600 wall ∂σ
Dip slip
direction where, λ is a positive scalar multiplier to be deter-
W
L mined with the aid of the loading-unloading crite-
( ) Shear band sequence number rion.
Combining equations (4) and (5) σ B can be
Fig.2 Definitions of the physical model (not to scale). solved for to obtain:
∂Ψ
is introduced in the form of a parameter S , which is σ B = σ A − SλD (6)
∂σ
the ratio of shear band area to finite element area and The plastic strain and internal variable (“kappa”)
is introduced into the elasto-plastic constitutive are given by the following equation, respectively:
model as a characteristic length. This method is ∂Ψ
similar to the one proposed by Pietruszczak & εB = ε A + λ (7)
Mroz16). Unlike their method, however, no direction
∂σ
of shear banding is specified in the present study. κ B = κ A + dκ (8)
Rather, it is implicitly assumed that the direction of The value of λ is
shear band coincides in a broad sense with the di- f (σ A , κ A ) (9)
λ=
rection of the maximum shear strain. ∂f ∂Ψ ∂f
S D −
F ∂σ ∂σ ∂κ
S= b (1)
Fe A yield function （ f ） corresponding to the
where, F b is the area of shear band in one element Mohr-Coulomb model and a plastic potential func-
and Fe is the total area of an element. Decision of tion （ Ψ ） , geometrically, represented by the
Drucker-Prager model, are employed:
Fb is based on the shear band thickness ( SB ). An
J2
approximated form of SB used in the present study f = −3α(κ)σ m + =0 (10)
can be expressed as g (θ )
SB Ψ = −3α ′(κ )σ m + J 2 = 0 (11)
S= (2)
Fe κ = ∫ dε p (12a)
The return mapping algorithm17), 18) is used, where (dε )
p
2
{ 2 2 2
}
= 2 (dexp ) + (deyp ) + (dezp ) + (dγ xyp ) = λ (12b)
2 2

the elastically predicted stresses ( σ A ) are relaxed

where, σ m is the mean stress (positive in compres-
onto a suitably updated yield surface ( σ B ). A change
sion), J2 is the second invariant of deviatoric

153
 Doboku Gakkai Ronbunshuu C Vol.65 No.1, 151-161, 2009. 2

stresses, θ is the Lode angle and dexp , deyp , de zp , ⎡ ⎧ ⎛ κ ⎞ 2 ⎫⎪ ⎤

⎪
φ r′ = φ r ⎢1 − β exp ⎨ − ⎜⎜ ⎟⎟ ⎬ ⎥ (19b)
dγ xyp are incremental deviatoric plastic strains in the ⎢ ⎪⎩ ⎝ ε d ⎠ ⎪⎭ ⎥⎦
⎣
coordinate axes x, y, and z. where, β and ε d are the stress-dilatancy material
In the case of the Mohr-Coulomb model, the Lode
angle function, g (θ ) and θ are given by following parameters.
The elastic moduli are estimated from modified
equations: equation proposed by Hardin & Black22) and are
3 − sin φ mob given the following equations in the case of clean
g (θ ) = (13a)
2 3 cos θ − 2 sin θ sin φ mob sand:
(2.17 − e0 )2 ⎛⎜ σ m ⎞⎟
0.5
1 ⎡ 3 3 J3 ⎤
θ = cos−1 ⎢ Gmax = G0 (20a)
3/ 2 ⎥ (13b) 1 + e0 ⎜ P ⎟ Pa ( Pa = 98kPa )
3 ⎢⎣ 2 J 2 ⎥⎦ ⎝ a ⎠
where, J3 is the third invariant of deviatoric stresses. 2(1 + ν )
K= Gmax (20b)
The mobilized friction angle of φ mob is given the 3(1 − 2ν )
equation: where, ν is Poisson’s ratio ， e0 is the initial void
⎧⎪ 3 3α (k ) ⎫⎪ ratio , G0 is the initial-shear-modulus constant and Pa
φ mob = sin −1 ⎨ ⎬ (14)
⎪⎩ 2 + 3α (k ) ⎪⎭ is atmospheric pressure.
The frictional hardening (when κ ≤ ε f ) and sof-
tening (when κ > ε f ) functions α (κ ) are expressed 4. VALIDATION OF THE NUMERICAL
as MODEL
m
⎧ 2 κε f ⎫
⎪ ⎪
α (κ ) = ⎨ ⎬ α p (hardening-regime) (15a) (1) Direct shear test box
⎪⎩ κ + ε f ⎪⎭ To justify the usefulness of the constitutive model
⎧⎪ ⎛ κ − ε ⎞ 2 ⎫⎪ incorporated in the numerical model discussed ear-
α (κ ) = α r + (α p − α r ) exp ⎨− ⎜⎜ f
⎟ ⎬ (softening-regime) (15b)
⎟
⎪⎩ ⎝ ε r ⎠ ⎪⎭
lier, the experimental global response of the direct
shear (DS) test box (shear stress-vertical displace-
where, m , ε f and ε r are the hardening/softening ments-horizontal displacements) with medium dense
material parameters and the parameters of α p and Fontainebleu sand is compared with predicted results
α r are estimated using the following equations: by the numerical model. The DS tests closely mimics
2 sin φ p the shearing from the faults, though it has been se-
αp = (16a) verely criticized for the nonuniformity of stresses and
3 (3 − sin φ p ) strains inside the soil sample23). The sand was plu-
2 sin φ r viated in the shear box and four Aluminum cans
αr = (16b) around the box, to check the average unit weight of
3 (3 − sin φ r )
the pluviated sand (Dr=63±4%). Then, direct shear
where, φ p , the peak friction angle, is estimated from (DS) tests were
the empirical relations21) to consider the stress level
effect, obtaining φ r , the residual friction angle, from
DS tests: Table 2 Material parameters of the numerical model.
I r = Dr {10 − ln (σ m )} − 1 (17a) Density (kN/m3) 15.57
Initial void ratio ( e0 ) 0.64
φ p = 3I r + φ r (17b)
Initial earth pressure coefficient (K0) 0.5
The plastic potential’s function α ′(κ ) is defined Coefficient of shear modulus ( G 0 , kPa ) 50
for plane strain conditions as,
Residual friction angle ( φ r : °) 30.2
tan ψ
α ′(κ ) = (18) Poisson’s ratio ( ν ) 0.3
9 + 12 tan 2 ψ Shear band thickness 3.84
( SB : mm, model scale)
The dilatancy angle of ψ is estimated from mod-
ified Rowe’s stress-dilatancy relationship, εf 0.2

sin φ mob − sin φ r′ εr 0.6

sin ψ = (19a)
1 − sin φ mob sin φ r′ εd 0.3
m 0.8
β 0.2

154
 Doboku Gakkai Ronbunshuu C Vol.65 No.1, 151-161, 2009. 2

conducted at a quasistatic displacement rate of 0.01 cretisation is displayed in Fig. 5. It refers to a uni-
mm/sec, to determine the sand’s shear strength pa- form soil deposit of thickness, H, at the base of
rameters. The specimen’s length and thickness were which a reverse fault, dipping at angle of 60°
60 and 30.4 mm, respectively. The test procedures (measured from the horizontal), ruptures and
followed the test standards BS 1377-724). The details
can be found in El Nahas et al.9). The finite element
mesh used for analyzing the direct shear box test is
shown Fig. 3. It contains 360 elements. The analysis
is performed for plane strain, with the following
boundary conditions. The nodes along the bottom of
“A” were fixed. Nodes along the upper box sides
were given a prescribed displacement in the hori-
Fig. 3 Finite element mesh and constituents for direct shear test
zontal direction. The center of an element row “C” in box analysis (A=soil, B=loading plate, C=gap, not to
the sample represents the gap between the upper box scale).
and the lower box. The upper box was free to move
relative to the lower box in the vertical direction. The 200 2.5
Stress Displacement
solid top element row “B” was assumed to be a cap 175 εf=0.2, m=0.8, εr=0.6, εd=0.3, β=0.2 εf=0.2, m=0.8, εr= 0.6, εd=0.3, β=0.2

for transmitting the normal stress. The solid top cap εf=0.1, m=0.8, εr=0.6, εd=0.3, β =0.2 εf=0.1, m=0.8, εr=0.6, εd=0.3, β =0.2

Vertical displacement (mm)

2.0
εf=0.2, m=0.2, εr=0.6, εd=0.3, β =0.2 εf=0.2, m=0.2, εr=0.6, εd=0.3, β =0.2
150
is free to rotate and to move in the vertical direction.

Shear stress (kPa)

εf=0.2, m=0.8, εr=0.1, εd=0.3, β=0.2 εf=0.2, m=0.8, εr= 0.1, εd=0.3, β=0.2
εf=0.2, m=0.8, εr=0.6, εd=0.05, β =0.2 εf=0.2, m=0.8, εr= 0.6, εd=0.05, β =0.2
The top row and side box of elements model a solid 125 εf=0.2, m=0.8, εr=0.6, εd=0.3, β=0.4 εf=0.2, m=0.8, εr= 0.6, εd=0.3, β=0.4 1.5

material, with linear elastic properties, Young’s 100

modulus of 2.1×104 MPa and Poisson’s ratio of 0.3. 75

1.0

Interfaces were assumed to be a soil sample attached

to all steel walls with angle of wall friction of a 50 0.5

maximum of 6 degrees (determined from DS tests). 25

To understand the effect of the parameters of such a 0

0.0

sophisticated constitutive model on the relationship 0.0 0.5 1.0 1.5 2.0
Horizontal displacement (mm)
2.5 3.0 3.5 4.0

between average shear stress of elements along the (a)

prescribed horizontal shear plane (Fig. 3) and aver- 400 1.5
age vertical displacements of the loading plate with Experiment
Shear stress (385 kPa)
Analysis
Shear stress (385kPa)
350
respect to relative prescribed horizontal displace- Vertical disp. (385 kPa) Vertical disp. (385 kPa)
Shear stress (200 kPa)
Shear stress (200 kPa)

Vertical displacement (mm)

ments in the DS tests, a detailed parametric study was 300 Vertical disp. (200 kPa) Vertical disp. (200 kPa)
1.0
conducted and some are shown in Fig. 4(a). It can be
Shear stress (kPa)

250
observed that the hardening/softening material pa-
200
rameters εf, εr and m influence the pre and post-peak
of that relationship. The parameter εf is related to εr. 150 0.5

The parameter m influences the hardening-regime 100

and εr influences the softening-regime. Also, the
50
stress-dilatancy material paremeters β and εd control 0.0
the mobilization of dilatancy and shear stress in the 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
DS test box. Fig. 4(b) shows close agreement be- Horizontal displacement (mm)

tween numerical and experimental results of some (b)

DS tests with effective normal stress=200 and 385 Fig. 4 Parametric study on direct sheat test box with normal
kPa, using the calibrated material parameters given effective stress of 100 kPa (a), and comparison of expe-
in Table 2. So, the calibrated numerical model has rimental and numerical results (using calibrated material
derived it’s validation for stress-level effect, strain parameters) with normal effective stress of 200 and 385
hardening-softening nature, and dilation in the DS kPa (b).
tests. Thus, the material parameters of the numerical
model (summarized in Table 2) have been calibrated
H 2H H
for medium dense Fontainebleu sand and will be
used for the fault analyses in the ensuing sections.
H
(2) Mesh size sensitivity
The problems of fault rupture propagation are
analyzed in this paper, using the quadrilateral iso-
parametric finite elements. The finite element dis- Fig. 5 Finite element mesh and boundary conditions.

155
 Doboku Gakkai Ronbunshuu C Vol.65 No.1, 151-161, 2009. 2

produces upward displacement, with a vertical 16

Normalized vertical displacement, δy/H(%)

component, h. Following the recommendation of 14

Bray25) and to minimize undesired boundary effects, 12 h/H= 2%

the width, L, of the FE model was set equal to 4H. 10 Experiment

The discretisation is finer in the central part of the 8

Analysis (FE=1 m)
Analysis (FE=1.5 m)
model with the quadrilateral elements than those at 6 Analysis (FE=2 m)
the two edges where limited deformation is expected. 4
The differential quasielastic displacement is applied 2
to the right part of the model (hanging-wall) in small 0
consecutive increments as shown in Fig. 5. Such a -160 -120 -80 -40 0 40 80
Normalized horizontal distance, d/H (%)
numerical model incorporating hardening-softening 16
model must be verified before to apply to the real-

Normalized vertical displacement, δy/H(%)

14
world fault problems, as strain softening makes the
12
analysis sensitive to mesh size. For this purpose, h/H= 6%
10 Experiment
finite elements of sizes: 1m×1m (width×height), Analysis (FE=1 m)
8
1.5 m×1.5 m, 2 m×2m, respectively, are used in the Analysis (FE=1.5 m)

central part. At the two edges, 2m×1m, 3m×1.5m, 6 Analysis (FE=2 m)

4m×2m, respectively, are used (Fig. 5). For the all 4

numerical analyses in this study, the used model 2

parameters are shown in Table 2. In this study, the 0

-160 -120 -80 -40 0 40 80
deformation field is normalized by the soil thickness, Normalized horizontal distance, d/H (%)
H, as suggested by Cole and Lade1), Bray25). The Fig. 6 Analysis of sensitivity of mesh density (compared with
results are compared with an arbitrary set experi- Test 8) for vertical displacements of ground surface.
mental result (Test 8). Fig. 6 shows that the norma-
lized vertical displacements (δy/H) are mesh size
independent (where, d is measured from the point of
application of the base dislocation, Fig. 2, and nor-
malized by soil thickness, H). Fig. 7 shows the plots (2)
for apparent maximum shear strain averaged at ele- (1
ment level for different mesh sizes, and it is observed
that the widening of the shear zone is dependent on
mesh size, but the orientation of the progressive path Ground image from experiment (h/H≈7.7%)
is less affected by the mesh size: reducing the size of
mesh leads slight shifting of out cropping location
towards the foot wall. So, the inclusion of shear band
effect into the constitutive relation makes the nu-
merical model insensitive to mesh size.
FE=1 m
5. RESULTS AND DISCUSSION

These two reverse faults were conducted at 1g

(Test8_1g_R) and 115g (Test 8) centrifuge accelera-
tion. However, the sand (D50=0.24 mm), modeled in
the centrifuge sand box, corresponds to a prototype FE=1.5 m
material with mean particle size diameter equals to
nD50 (where, n is scale factor). So, the shear band
thicknesses used for 1g (n=1) and 115g (n=115)
model test analyses are 16D50 (≈3.84 mm) and
115×16D50 (≈441.5 mm), respectively. Mesh with
FE=2 m and 2/115(≈0.0174) m are used for 115g and
band thicknesses used for 1g (n=1) and 115g (n=115) FE=2 m
model test analyses are 16D50 (≈3.84 mm) and Fig. 7 Comparison of experimental ground image (Test 8)
115×16D50 (≈441.5 mm), respectively. Mesh with with deformed mesh and shear strain contour (darker
FE=2 m and 2/115 (≈0.0174) m are used for 115g regions denote higher shear strain) for FE sizes: 1m,
and 1g tests, respectively. Numerical predictions 1.5 m and 2 m.

156
 Doboku Gakkai Ronbunshuu C Vol.65 No.1, 151-161, 2009. 2

are compared with the experiments in terms of nor-
malized vertical displacements on the ground surface.
Additionally, the photographs of the deformed
ground and a characteristic shear strain contour are
compared with the numerical results.
Figs. 8 and 9 show that the numerically predicted
normalized vertical displacements on the ground
surface for 1g (for relative base dislocations: h/H≈2.1
to 10.3%) and 115g (h/H≈2 to 10.1%), respectively,
are, satisfactorily, closed to the experimental data
points. The development of failure mechanism in the
model ground during reverse fault rupture in free
field is shown in the images in Figs. 10(a) to 10(d)
and 12(a) to 12(d) for Test 8_1g_R and Tests 8, re-
spectively, indicating with the sequence number of
shear band formation. The shear bands are drawn by
naked eye observation from the deformed colored
sand layers, while the average shear strain on the
strain plot was about 20% or more. The strain loca-
lization in narrow shear bands starts at relative ver-
tical base displacement, h/H=4.1% (experiment:
h/H≈3.9%) and h/H=5% (experiment: h/H≈5.1%) at
1 and 115g level, respectively. Increasing the g-level
causes late start of shear bands. First, a shear band
propagated in the upward direction from the fault
respectively. Additionaly, Figs. 13 and 14 show the
satisfactory comparison between the experimentally
and numerically obtained characteristic average
shear strain contours in the 1g and centrifuge, re-
spectively. The inclination of the line between the
faults rupture point and the ends of the soil defor-
mations at ground surface at the end of each test (θmax,
Fig.2) is 33° (experiment: 40.3°) and 35° (experi-
ment: 44°) at 1g and 115g levels, respectively. Here,
the numerical model satisfactorily predicts the ex-
tension of the ground surface deformation. So, the
increasing of the centrifuge acceleration to produce
the prototype stess fields in the model ground results
increased ∆ and θmax.
6. CONCLUSIONS
The emphasis of this study is on the verification
of a sophisticated numerical model for prediction of
progressive failure in dry cohesionless material
above reverse fault with dip angle of 60° in
Normalized vertical displacement, δy/H(%)

16
point, and later became inactive. Thereafter, the soil Experiment Analysis
14 h/H=2.1% h/H=2.2%
shear strains were localized in one or more shear h/H=4.4% h/H=4.2%
h/H=6% h/H=6%
bands which had a steeper inclination to the hori- 12 h/H=8.2% h/H=8.2%
h/H=10.3% h/H=10.5%
zontal than the first shear band and represented the 10
active fault rupture shear bands. This type of pro- 8
gressive nature of shear band formation is predicted
6
also by the the proposed numerical model to a satis-
factory extent, shown in Figs. 10(e) to 10(h) and 4

12(e) to 12(h) for 1g and centrifuge, respectively. 2

The fault rupture lines gradually decreased their in- 0
clinations to the horizontal, as they propagated up- -160 -120 -80 -40 0 40 80
Normalized horizontal distance, d/H (%)
ward. At 1g level, the fault reached the ground sur-
face and developed a scarp after h/H=7.3% (expe- Fig. 8 Comparison of experimental and numerical vertical
riment: h/H≈7.4%). In the centrifuge tests, the fault displacements of ground surface for Test8_1g_R.
rupture didn’t reach the ground surface, even at the
test’s maximum vertical base displacement, hmax (22
mm, in model scale), whereas the analysis predicts
Normalized vertical displacement, δy/H(%)

16
Experiment Analysis
the outermost shear band outcrops at h/H=11.2%, in 14 h/H=2% h/H=2 %
h/H=4.1% h/H=4 %
model scale. Now, for the calculation of the inclina- 12
h/H=6.1% h/H=6 %
h/H=8.5% h/H=8.4 %
tion of the fault main slip surface in the model ground h/H=10.1% h/H=10 %

to the horizontal (∆, Fig. 2), an operative definition26) 10

for the location of the fault rupture at the ground 8

surface is used to identify it by the point with the 6
maximum absolute value of the second derivative of
4
the vertical displacement along the horizontal di-
rection. Fig. 11(a) shows such a point on the cha- 2

racteristic normalized vertical displacement plot of 0

-160 -120 -80 -40 0 40 80
the ground surface and that lies within the corres- Normalized horizontal distance, d/H (%)
ponding shear bands (Fig. 11(b)). Thus, the average
dip angle (∆) at maximum h, is 48° (experiment: 50°) Fig.9 Comparison of experimental and numerical vertical
displacements of ground surface for Test 8.
and 55° (experiment: 60°) at 1g and 115g levels,

157
 Doboku Gakkai Ronbunshuu C Vol.65 No.1, 151-161, 2009. 2

Experiment Analysis

(a) Before experiment (e) Before analysis

(1)

(b) h/H≈4.2% (f) h/H=4.2%

(3)
(2)

(1)

(c) h/H≈6% (g) h/H=6%

(3)

(2)

(1)

(d) h/H≈7.4% (h) h/H=7.4%

Fig. 10 Comparison of experimental results with numerical prediction for Test8_1g_R: (a) to (d) are ground images and (e) to (h)
are numerical average shear strain plot on deformed FE mesh (the darkest color indicates shear strain equal or more than
20%).

10 Outcropping location
Normalized vertical displacement, δy/H(%)

9
8
7
6 Outcropping location
5
4
3
2
1
0
-120 -80 -40 0 40 80
Normalized horizontal distance, d/H (%)

(a) (b)
Fig. 11 Identification of fault outcropping location: (a) a characteristic vertical displacement on the ground surface and the out-
cropping location (b) corresponding average shear strain plot.

158
 Doboku Gakkai Ronbunshuu C Vol.65 No.1, 151-161, 2009. 2

Experiment Analysis

(a) Before experiment (e) Before analysis

(1)

(b) h/H≈5.1%
(f) h/H=5.1%

(2)

(1)

(c) h/H≈7.7%
(g) h/H=7.7%

(2)

(3)
(1)

(d) h/H≈10.1%
(h) h/H=10.1%

Fig. 12 Comparison of experimental results with numerical predictions for Test 8: (a) to (d) are ground images and (e) to (h) are
numerical average shear strain plot on deformed FE mesh (the darkest color indicates shear strain equal or more than 20%).

-10
Normalized depth (%)

-20
-30
-40
-50
-60
-70
-80 Distance O
-90

-99 -75 -50 -25 0 25 50

O (focus)
Normalized distance, d/H(%)

(a) h/H≈10.1% (b) h/H=10.1%

Fig. 13 Characteristic experimental (a) and numerical (b) average shear strain contour for Test8_1g_R.

159

-10
Normalized depth (%)
-30
-50
-70
-90
-110 -90 -70 -60 -40 -20 0 20
Normalized distance, d/H(%)
(c) h/H≈10.1%
Fig. 14 Characteristic experimental (a) and numerical (b) average shear strain contour for Test 8.
relation to soil mechanics and displacement based
finite element method. In this regards, the analytical
results and extensive experimental results at 1g and
115g level in terms of some important design factors,
generally considered for the design of structures near
or above the active faults: the normalized vertical
displacements profile of the ground surface, mini-
mum normalized vertical base displacement for the
rupture to reach the ground, the average dip angle
propagated into the soil, inclination of the line join-
ing the faults rupture point and the ends of the soil
deformations at ground surface at the end of each test
to measure the horizontal extent of the deformed
surface ground, are compared. From the results, the
scale effects are found as: in the 1g tests, the shear
bands start and reach the ground faster, the average
dip angle is shallower, as well as, the horizontal ex-
tent of soil displacement on the ground surface is
larger than that in the centrifuge. So, such a numer-
ical model, including the factors: effect of confining
pressure, non-linear strain hardening and softening
relationship, stress-dilatancy relationship and shear
band effect, can take care when projecting the model
test results to prototype scale, like the centrifuge
does.
ACKNOWLEDGMENT: The experimental part of
this work was done at the University of Dundee un-
der the research project, “QUAKER”, funded by the
European Commission under the contract number of
EVG1-CT-2002-00064. Special thanks to the all
research members of that project.
REFERENCES
1) Cole, D.A. Jr. and Lade, P.V.: Influence zones in alluvium
over dip-slip faults, J. Geotech. Engrg., ASCE, Vol. 110, No.
5, pp. 599-615, 1984.
2) Scott, R.F. and Schoustra, J.J.: Nuclear power plant sitting on
deep alluvium, J. Geotech. Engrg., ASCE, Vol.100, pp.
449-459, 1974.
Doboku Gakkai Ronbunshuu C Vol.65 No.1, 151-161, 2009. 2
40 60 O (focus)
(d) h/H=10.1%
3) Roth, W.H., Kalsi, G., Papastamatiou, O. and Cundall, P.A.:
Numerical modeling of fault propagation in soils, Proc. 4th
Int. Conf. on Num. Meth. Geomech., pp. 487-494, 1982.
4) Walters, J.V. and Thomas, J.N.: Shear zone development in
granular materials, Proc. 4th Int.Conf. Num. Meth. Goemech.,
Part I, pp. 263-274, 1982.
5) Bray, J.D., Seed, R.B. and Seed, H.B.: Analysis of earth-
quake fault rupture propagation through cohesive soil, J.
Geotech. Engrg., ASCE, Vol. 120, No. 3, pp. 562-580, 1994.
6) Sakai, T. and Tanaka, T.: Scale effect of a shallow circular
anchor in dense sand, Soils and Foundations, Vol. 38, pp.
93-99, 1998.
7) Tanaka, T. and Sakai, T.: Progressive failure and scale effect
of trap-door problems with granular materials, Soils and
Foundations, Vol. 33, No.1, pp.11-22, 1993.
8) White, D. J., Take, W. A., and Bolton, M. D.: Soil deforma-
tion measurement using particle image velocimetry (PIV)
and photogrammetry, Geotechnique, Vol. 53, No.7, pp.
619-631, 2003.
9) El Nahas, A., Bransby, M. F., Davies, M. C. R.: Centrifuge
modelling of the interaction between normal fault rupture
and rigid, strong raft foundations, Proc. International Con-
ference on Physical Modelling in Geotechnics, Hong Kong,
pp. 317-323, 2006.
10) El Nahas, A., Bransby, M.F. and Davies, M.C.R.: Centrifuge
Modelling of Dip Slip Fault Rupture in Alluvium, submitted
to ASTM, 2008.
11) Tanaka, T. and Kawamoto, O.: Three dimensional finite
element collapse analysis for foundations and slopes using
dynamic relaxation, Proc. of Numerical Methods in Geo-
mechanics, Innsbruch, pp. 1213-1218, 1988.
12) Loret, B., and Prevost, H. J.: Dynamic strain localization in
elasto-(visco)-plastic solids. Part 1: General formulation and
one-dimensional examples, Comput. Methods Appl. Mech.
Eng., Vol. 83, No. 3, pp. 247-273, 1990.
13) Bazant, Z. P., and Tsang, T. P.: Nonlocal finite element
analysis of strain softening solids, J. Eng. Mech., Vol. 110,
No. 12, pp. 1441-1450, 1984.
14) Muhlhaus, H. B., Vardoulakis, I.: The thickness of shear
bands in granular materials, Geotechnique, Vol. 37, No. 3,
pp. 271-283, 1987.
15) Gudehus, G., and Nubel, K.: Evolution of shear bands in
sand, Geotechnique, Vol. 54, No. 3, pp. 187-201, 2004.
16) Pietruszczak, S., Mróz, Z.: Finite element analysis of de-
formation of strain-softening materials, Int. J. Numer.
Methods Eng., Vol. 17, No. 3, pp. 327-334, 1981.
17) Simo, J. C. and Ortiz, M.: A unified approach to finite de-
formation elastoplastic analysis based on the use of hyper-
elastic constitutive equations, Computer Methods in Applied
Mechanics and Engineering, Vol. 49, pp. 221-245, 1985.
160

18) Ortiz, M. and Simo, J. C.: An analysis of a new class of
integration algorithms for elastoplastic constitutive relations,
Int. J. Numer. Meth. In Engrg., Vol. 23, pp. 353-366, 1986.
19) Vardulakis, I., Graf, B. and Gudehus, G.: Trap-door problem
with dry sand : a statical approach based upon model test
kinematics, Int. J. for Numerical and Analytical Methods in
Geomechanics, Vol. 5, pp. 57-78, 1981.
20) Yoshida, T., Tatsuoka, F., Siddiquee, M. S. A., Kamegai, Y.
and Pak, C. S.: Some observation of zone of localization in
model tests on dry sand, Proc., 3rd Int. Workshop on Loca-
lisation and Bifurcation Theory for Soils and Rocks, Gre-
noble, pp. 165-180, 1993.
21) Bolton, M. D.: The strength and dilatancy of sands, Geo-
technique, Vol. 36, pp. 65-78, 1986.
161
Doboku Gakkai Ronbunshuu C Vol.65 No.1, 151-161, 2009. 2
22) Hardin, B. O. and Black, W. L.: Vibration modulus of nor-
mally consolidated clay, ASCE, 94(SM2), pp. 353-369, 1968.
23) Budhu, M.: Nonuniformities imposed by simple shear ap-
paratus, Can. Geotech. J., Vol. 21, No. 1, pp. 125-137, 1984.
24) British Standards Institution: Methods of test for soils for
civil engineering purposes-shear strength tests (total
stress)-BS1377-7, London, UK, 1990.
25) Bray, J. D.: The effects of tectonic movements on stresses
and deformations in earth embankments, Ph.D. thesis, Uni-
versity of California, Berkeley, 1990.
26) Yilmaz, M., T., and Paolucci, R.: Earthquake fault rup-
ture-shallow foundation interaction in undrained soils: a
simplified analytical approach, Earthquake Eng. Struct.
Dyn., Vol. 36, pp. 101-118, 2007.
(Received February 27, 2008)