Finite element simulation of an Excavation-triggered landslide Using large
deformation theory

Samaneh Mohammadi, Hossein Taiebat

PII: S0013-7952(16)30044-8
DOI: doi: 10.1016/j.enggeo.2016.02.012
Reference: ENGEO 4240

To appear in: Engineering Geology

Received date: 11 October 2015

Revised date: 6 January 2016
Accepted date: 26 February 2016

Please cite this article as: Mohammadi, Samaneh, Taiebat, Hossein, Finite element sim-
ulation of an Excavation-triggered landslide Using large deformation theory, Engineering
Geology (2016), doi: 10.1016/j.enggeo.2016.02.012

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Title: FINITE ELEMENT SIMULATION OF AN EXCAVATION-TRIGGERED LANDSLIDE

USING LARGE DEFORMATION THEORY

Order of Authors: Samaneh Mohammadi; Hossein Taiebat

Corresponding Author: Dr. Samaneh Mohammadi, PhD

Corresponding Author's Institution: Western Sydney University,

T
Institute for infrastructure engineering, Sydney, Australia
Second author: Dr. Hossein Taiebat, PhD

IP
Second author institution: UNSW Australia, Sydney, Australia
E-mail addresses: s.mohammadi@westernsydney.edu.au (S. Mohammadi),
h.taiebat@unsw.edu.au (H. Taiebat)

R
SC
NU
MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

FINITE ELEMENT SIMULATION OF AN EXCAVATION-TRIGGERED

LANDSLIDE USING LARGE DEFORMATION THEORY

T
ABSTRACT

IP
Numerical simulation of an excavation induced landslide in a strain softening material is

R
presented and the results are compared with field measurements. The simulation is based on

SC
the methodology proposed to estimate the post-failure deformation of slopes in strain
softening materials. The method includes: a) the Updated Lagrangian formulation which is
essential in capturing the changing geometry and configuration of the slope during failure, b)

NU
a strain softening constitutive model which enables simulation of the progressive failure
mechanisms, c) a stable solution scheme to prevent problems associated with numerical
MA
convergence in strain softening materials, and d) the h-adaptive mesh refinement technique to
prevent excessive distortion of the finite element mesh due to large deformation and to
increase the accuracy of the numerical solution.
D
TE

For the slope considered here, it is shown that failure initiated due to the excavation at the toe
of the slope and propagated upward due to the strain softening behaviour of the geomaterials
P

which eventually led to the progressive failure of this slope. The failure surface is mainly
within a thin layer of soil with substantial strain softening behaviour but propagates to the
CE

surrounding soil as the excavation proceeds. The predicted crest settlement, toe movement,
and deformed shape of the slope are lower than the observed behaviour of the slope.
AC

However, the numerical analysis clearly predicts the triggering factor and the failure
mechanism of the landslide and the impact of the large deformation of the soil mass on the
houses at the toe of the slope.

1. INTRODUCTION
Landslides are natural phenomena induced either by natural factors such as erosion of
geomaterial, rain and changes in groundwater level, loss of vegetation and earthquakes or by
human activities such as extra loading and construction works at the top of slopes, vibrations,
blasting and earthworks and excavation at the toe. Landslides cause significant damage to
infrastructure and can lead to loss of life. In order to reduce the substantial social,
environmental and economic impacts of landslides, multi-disciplinary experts from
 ACCEPTED MANUSCRIPT

geotechnical engineering, geoscience, remote sensing and hydrology are required to design
predictive measures to reduce the vulnerability to landslides. Development of early warning
systems as well as advanced numerical modeling techniques is crucial in predicting the
possible initiation of a slide and the extent of its movement. The magnitude of deformations

T
after failure can be used to evaluate the extent of the failure and to design the remedial work

IP
required to limit the associated damage to infrastructure. It is therefore important to be able to
predict the post-failure deformation of slopes and to identify slopes with small post-failure

R
deformation from those with progressive failure which may lead to large landslides. This can

SC
be achieved by numerical simulation of slope failure, provided that a suitable numerical
algorithm and an appropriate constitutive relationship are available.

NU
The standard finite element method based on infinitesimal strain formulation has been
found to be very effective in slope stability analysis especially in finding the location of
MA
failure surfaces and in obtaining accurate factor of safety, FoS, of slopes in complex soil
strata. However, the standard finite element method can give the deformation of slopes only
up to the onset of failure. A large deformation finite element analysis based on finite strain
D

formulation can give the deformation of slopes after failure, provided that the geomaterials
TE

can be assumed to deform in a continuum. This assumption is valid if the extent of

deformation is relatively small. Geomaterials in landslides lose their integrity due to large
P

deformation and cannot be modeled by constitutive relationships that are based on continuum
CE

mechanics. However, a relatively large deformation predicted for a slope by a large

deformation analysis will be a good indicator of the possible occurrence of a landslide.
AC

In this paper, the results of a large deformation analysis of the failure of a slope are
presented and discussed. The geometry and material properties of the slope are presented
followed by a brief description of the large deformation finite element formulation. Then
detailed step-by-step procedures are used to simulate the failure of the slope and its post
failure behavior is presented. Finally the finite element results are compared with field
records and the observed behavior of the slope.
 ACCEPTED MANUSCRIPT

T
R IP
SC
NU
MA
D
TE

Figure 1. Bore hole results of the Senise area, adapted from Guerricchio & Melidoro (1988), soil
P

symbols: 1-sand with remolded gravel, 2-fine sand with gravel, 3-medium coarse sand, 4-weakly
CE

compact cemented fine sand, 5-yellow sand with thin layers of silty clay, 6-blue grey clay, 7- fossils

2. THE SENISE SLOPE FAILURE

AC

The Senise landslide occurred at Timpone on 26 July 1986 in southern Italy. The landslide
was triggered by an excavation at the toe of the slope and although the vertical cuts were
protected by 9–10 m concrete retaining walls, the slope movements could not be avoided.
The landslide affected a large area and led to loss of eight lives and destruction of seven
houses. The dimensions of the landslide were about 150 m in width and 230 m in length. The
depth of the soil mass moved in the landslide varied between 10 to 15 m except at the toe
which was about 5 m. The borehole investigation of the Senise area showed that the subsoil is
mainly composed of yellow sand with clayey silt levels and blue grey clay as shown in Figure
1. The comparison of the modified morphology of the area after the landslide with the initial
slope configuration, showed that the landslide occurred by a slipping translational mechanism
which occurred along a weak planar surface approximately parallel to the slope with little
rotation or backward tilting (Guerricchio and Melidoro 1988). The geological and
 ACCEPTED MANUSCRIPT

geomorphological mapping of the landslide area, as well as the subsurface investigations

indicated that the base of the slide slip surface was on the thin clayey silt layer interbedded
with a slightly cemented sand formation inclined 18 degrees to the horizontal (Del Prete and
Hutchinson 1988). The stability analyses conducted by Del Prete and Hutchinson (1988) and

T
Viggiani and Di Maio (1991) also suggested that the landslide occurred along the clayey

IP
levels parallel to the slope with shear strength parameters near residual values. The landslide
occurred during a very dry period (Guerricchio and Melidoro 1988) where the location of the

R
water table was estimated to be 23 m below the toe of the slope (Del Prete and Hutchinson

SC
1988). Therefore, water could not have contributed to the failure of the slope. A plan view of
the landslide is shown in Figure 2. The geometry of the Senise hill in section A–A and the

NU
location of the excavation at the toe of the slope is shown in Figure 3 and Figure 4. The
landslide caused movement of a large mass of soil by about 30 m downhill along section A–
MA
A (Figure 2 and Figure 3).
D
P TE
CE
AC

Figure 2. Plan of landslide area with the marked location of section A–A (after Del Prete and
Hutchinson, (1988))
 ACCEPTED MANUSCRIPT

T
R IP
SC
Figure 3. Schematic soil profile of the Senise hill at section A–A (adapted from Viggiani and Di

NU
MAMaio, (1991)
D
TE

Figure 4. Senise hill geometry and soil profile at section A–A and location of excavation
P
CE

Numerical analyses were performed to investigate the contributing factors leading to the
AC

Senise landslide (Conte et al. 2013; Tang 2008; Troncone 2005; Troncone et al. 2014).
Elasto-plastic finite element analyses of the landslide did not reach convergence when the
excavation was modeled to its maximum depth. According to Troncone (2005), the elasto
visco-plastic analysis captured the position of the failure surface very well and a converged
solution was found for the model. However, the predicted crest displacement was limited to a
maximum of 0.53 m at the onset of failure which is very small compared to that observed in
the field. Obviously this small deformation would not raise any concern about a possible
landslide. Tang (2008) applied the elasto-plastic Cosserat continuum model and resolved the
convergence issues of the classical continuum models when simulating the excavation to its
maximum excavation depth. The analysis captured the position of the failure surface
accurately. However, no attempt was made to predict post-failure deformations. The three-
dimensional numerical analysis of this landslide performed by Troncone et al. (2014) showed
 ACCEPTED MANUSCRIPT

the direction of the slip surface accurately but could only predict the crest settlements up to
the onset of failure as expected from any standard finite element analysis based on
infinitesimal strain formulation. The focus of all previous studies has been on simulation of
the slope up to failure rather than evaluating the post-failure deformations. The magnitude of

T
the post-failure deformations is important as it shows the extent of failure and differentiates

IP
between small slope failures and landslides.

R
3. SOIL PROFILE AND CONSTITUTIVE MODEL

SC
The geotechnical profile of the Senise hill in section A–A is shown in Figure 3. The subsoil
consists of yellow sand interbedded with clayey silt layers, which overtops blue grey clay

NU
(Troncone 2005). The clayey silt shows a pronounced strain softening characteristic. The
factor of safety of the slope was calculated for the peak and the residual strength parameters
as FoSpeak = 1.73 and FoSres = 0.60 (Troncone 2005), considering a slip surface similar to
MA
the one actually observed in the field with its base on the thin clayey silt layer. This implies
that the progressive failure occurred with a mobilized strength varying between the peak and
the residual values, depending on the magnitude of strains developed in the layer due to the
D

excavation of the toe. The three-dimensional back-analysis of the landslide also confirmed
TE

that a progressive failure occurred at Senise hill (Conte et al. 2013; Troncone et al. 2014). A
back analysis of the slide based on the Janbu method also gives an average mobilized friction
P

angle in between the peak and the residual values (Viggiani and Di Maio 1991).
CE

In this paper, the strain softening behavior of the Senise slope material is taken into
account using the framework of the Mohr-Coulomb failure criteria. In this model the shear
AC

strength parameters, such as cohesion and internal friction angle, are defined as functions of
the shear strain, or more generally the cumulative deviatoric strain invariant, . For plane
strain conditions can be defined as:

[1]
where and are the normal strains in and directions, and is the engineering shear
strain.

The process of strain softening begins when exceeds the strain corresponding to the peak shear
strength, , beyond which the strength parameters, gradually decrease from the peak value,
, to its residual value, , at a residual strain of . The relationship between strength and
 ACCEPTED MANUSCRIPT

strain is schematically shown in Figure 5

T
R IP
SC
NU
Figure 5. The strength parameters can be defined in terms of a piecewise tri-linear function of
the deviatoric strain invariant as:
MA

[2]
D

where corresponds to the strength parameters of soils, that is, friction angle, , and
TE

cohesion, . The simple strain softening model employed in this study can be regarded as an
extension of the standard elasto-plastic Mohr-Coulomb model, provided that the extra model
P

parameters are defined.

CE
AC

Figure 5. Variation of shear strength parameter, , with shear strain,

The subsoil peak strength parameters, and , and the residual values, and
, for the yellow sand and clayey silt are deduced from the direct shear tests performed by
Viggiani and Di Maio (1991). The peak and residual strains for yellow sand and clayey silt
 ACCEPTED MANUSCRIPT

are extracted from the results of CD triaxial as shown in Figure 6. The soil dilation angle was
assumed to be zero, implying a non-associated flow rule with zero volume change during
yielding. The choice of dilation angle is less important in the slope stability analysis since the
nature of this problem is unconfined; as proposed by Griffiths and Lane (1999) and Troncone

T
(2005). Details of soil parameters for all layers are given in Table 1. The Young’s modulus,

IP
, of different layers should be adopted with care in the finite element analysis of slopes with
strain softening materials, since progressive failure is more dramatic when the material is

R
stiffer (Mohammadi and Taiebat 2013). Values of Young’s moduli for different layers have

SC
been deduced from the results of the triaxial tests performed by Del Prete and Hutchinson
(1988), considering confining pressures that correspond approximately to those that existed in

NU
the slope. No specific experimental value for Poisson’s ratio is available. Generally, for the
drained case of slope analysis, the value of the Poisson's ratio is within the range of 0.2 < υ
MA
<0.3 (Griffiths and Lane 1999). Therefore, the value of Poisson's ratio is assumed to be 0.25
for all layers.

Table. 1 Soil parameters used in the analyses

D
TE

Material Model and element type Parameters

E = 70 MPa, , ν = 0.25, γ = 20 kN/m3, ψ=00
Strain-softening
Yellow Sand c’peak = 37 kPa, c'res = 0 kPa, ’peak = 430,’res = 350
(extended Mohr-Coulomb)
P

= 0%, = 15%
E = 25 MPa, ν = 0.25, γ = 20 kN/m3, ψ=00
CE

Strain-softening
Clayey silt c'peak = 15 kPa, c'res = 0 kPa, ’peak = 300,’res = 120
(extended Mohr-Coulomb)
=0%, = 10%
Blue Grey Elastic-perfectly plastic E = 70 MPa, ν = 0.25, γ = 20 kN/m3, ψ=00
Clay (Mohr-Coulomb) c’ = 150 kPa, ’ = 310
AC

Concrete Elastic E = 5000 MPa, ν = 0.20, γ = 20 kN/m3

 ACCEPTED MANUSCRIPT

T
IP
Figure 6. Experimental results of CD triaxial test: a) Sand b) Clayey Silt, (after Del Prete and

R
Hutchinson, (1988))

SC
4. SIMULATION OF INITIATION AND PROPAGATION OF THE FAILURE
4.1. Finite element modeling

NU
The initiation and propagation of the Senise landslide is simulated using the finite element
program, AFENA (Carter and Balaam 1995). AFENA has been upgraded here to include the
MA
large deformation formulations and enables simulation of the strain softening behavior of
soils and h-adaptive remeshing. The analysis is performed in a two-dimensional finite
element analysis under plane strain conditions. The element type used in the spatial
D

discretization of the slope is the isoparametric six-noded triangle. The initial finite element
TE

meshes used in the analysis before and after the excavation and the location of houses No 4
and 6 are shown in Figure 7. The finite element model of the slope is approximately 40 m
P

high resting on a 28 m soil stratum. The soil is assumed to obey the Mohr-Coulomb failure
CE

criterion, modified to include the strain softening behavior based on a non-associated flow
rule, as described previously in soil profile and constitutive model section.
AC
 ACCEPTED MANUSCRIPT

T
R IP
SC
NU
MA

Figure 7. Finite element mesh: a) before excavation and b) after excavation and location of houses
D

The slope is under self-weight gravitational loading. The initial stresses due to the soil gravity
TE

are distributed in the continuum by simulating staged construction of the slope, that is,
switching on the gravity in layers of elements sequentially. The strength parameters during
P

the construction of the slope are set to the peak values to prevent failure during the staged
CE

construction. An iterative solution scheme is used in the finite element analysis because of
the geometric nonlinearity, due to large displacements and strains, and the material
AC

nonlinearity, due to post-failure behavior. The solution procedure adopted in the analyses is
based on the modified Newton-Raphson method, using the elastic stiffness of the material.
Convergence of the finite element solution is established on the basis of the standard norm of
the out-of-balance force with a tolerance of 10-8 throughout the analyses.

The large deformation analysis performed here is based on the Updated Lagrangian (UL)
formulation which accounts for rigid body rotations. The large deformation analysis can
simulate the continuous change of the slope configuration during failure. This eliminates the
limitation of the conventional application of the finite element method in slope stability
analysis which can only estimate the slope deformations up to the onset of failure. The strain
softening model enables development and formation of the shear zone during failure.
However, attention is paid to introduce a stable numerical procedure to eliminate numerical
instabilities (Potts et al. 1990; Troncone 2005) when an elasto-plastic model is employed.
 ACCEPTED MANUSCRIPT

The pronounced strain dependency of the slope materials could lead to significantly large
deformations and mesh distortions during the simulation of the failure which reduces the
accuracy of the finite element analysis. Hence, frequent remeshing of the problem is required
when large distortion is detected in an element during the analysis. An h-adaptive remeshing

T
technique is employed in this study to increase the efficiency and accuracy of the solution.

IP
4.2. Simulation of Excavation and Progressive Failure

R
The excavation near the toe is simulated by switching off the self-weight and the stiffness of

SC
layers of elements near the toe and simultaneous installation of the retaining wall. The
excavation process is simulated in four equal increments. The retaining wall was modeled
using a stiff elastic material. The shear strength parameters of the soil reduce during the

NU
analysis according to the strain levels in slope materials following Equation (2). The strain
localizations induced by the excavation results in initiation of the progressive failure and
MA
slope movement. At this stage the corresponding FoS of the slope is equal to one. As the
shear strength parameters decrease further towards their residual values, the FoS drops below
one, the shear zone expands, and the slope deforms further which in turn results in further
D

reduction of shear strength parameters.

TE

As mentioned before, a realistic simulation of slope failure should take into account the
P

continuous change in the geometry and configuration of the slope in the analysis. The UL
large deformation analysis can capture the change in slope geometry, as well as simulating
CE

the post-failure deformations. In any large deformation analysis three steps should be taken to
simulate the geometrical nonlinearity properly: 1) updating nodal coordinates after each
AC

increment, 2) transforming stresses to account for rigid body rotations, and 3) regenerating
the finite element mesh when a large element distortion is detected to maintain the accuracy
of the finite element analysis and to ensure convergence of the solution. Details of this
procedure in the simulation of the failure of the Senise slope are given in the following
sections.

1) Updating Nodal Coordinates

A practical way to include the change in the geometry during a large deformation analysis of
a slope is to frequently update the nodal coordinates of the finite element mesh. Convergence
will not be achieved after the initial failure if the change in the geometry is not considered
during the analysis. This is due to the fact that the destabilizing forces and moments which
correspond to the initial geometry of the slope will not be reduced during the analysis.
 ACCEPTED MANUSCRIPT

However, when the nodal coordinates are updated during the course of the incremental
nonlinear analysis, the destabilizing forces and moments, corresponding to the updated
configuration of the slope, will be gradually reduced resulting in eventual stabilization of the
slope. Any change in the geometry of the slope requires stress transformation to prevent

T
generation of spurious stresses due to rigid body rotation.

IP
2) Updated Lagrangian governing equations

R
An accurate calculation of stresses can be made using large deformation theory. Large

SC
deformation finite element analysis using the UL formulation has been introduced to slope
instability problems by Mohammadi and Taiebat (2013) to determine the post-failure
deformation of slopes. Among the existing formulation, the UL (Bathe et al. 1975) is more

NU
compatible with the nature of slope failures. In a large deformation finite element analysis,
the formulation based on the infinitesimal strain is replaced by the finite strain formulation
MA
(Chen and Mizuno 1990). In addition, a stress transformation should be used in the stress-
strain relation to ensure the components of Cauchy stress do not change under rigid body
rotation (Belytschko et al. 2000). Here, a stress transformation based on Jaumann stress rates
D

(Jaumann 1911), , is incorporated into the stress–strain relationship to account for the
TE

objectivity of stresses. The stress increment in the finite strain formulation is given by:
P

[3]
CE

where is the rigid body rotation term that can be calculated based on the displacement
vector, , as:
AC

[4]
The stress transformation can be included in the stress equations either before or after the
return mapping in each increment (Nazem et al. 2009). A more detailed discussion on the
theoretical background of the formulation was presented by Mohammadi and Taiebat (2013).

The large deformation analysis of the Senise hill slope captures the characteristic features of the
failure during the excavation of the toe. For example, the development of cumulative shear strains
 ACCEPTED MANUSCRIPT

during different stages of the excavation obtained by the large deformation analysis is shown in

T
R IP
SC
NU
MA

Figure 8
D

Figure 8. The excavation at the toe increases the cumulative strains mainly in the clayey
TE

silt layer which has a pronounced strain softening behavior. Localization of shear strains in
this layer leads to the development of progressive failure and large crest settlements. The
P

gradual development of failure in the clayey silt layer is also shown in Figure 8. The zone
CE

with larger shear strains fails first due to lower shear strength (Figure 8-a). The failure of this
zone increases the post-peak strains within the shear zone which results in further reduction
AC

of the shear strength and failure of a larger section of the shear zone (Figure 8-b).
Redistribution of stresses increases the shear strains further and eventually causes failure of
the whole shear zone and the overlying soil mass (Figure 8-c). A clear shear band is formed
as a result of the sequential excavation at the toe which leads to the collapse of the slope.
 ACCEPTED MANUSCRIPT

T
R IP
SC
NU
MA

Figure 8. Sequence of progressive failure and development of cumulative shear strains during
D

excavation steps: a) h=0.50hmax, , b)h=0.75hmax, , c)h=hmax

TE

Finite element analysis of problems with strain softening materials is often accompanied
by lack of convergence due to the post-peak instabilities and localization of shear strains
P

(Bazant et al. 1984; Belytschko et al. 1988; Brinkgreve 1994; Khoei et al. 1997; Larsson et al.
CE

1991; Pietruszczak and Mroz 1981; Pietruszczak and Stolle 1985, 1987; Sterpi 1999; Wang et
al. 2011). The issues related to the convergence are eliminated in this study by the
AC

implementation of the Updated Lagrangian formulation and frequent updating of the elastic
stiffness matrix with the updated geometry. This method is effective in eliminating the
problem of oscillation of the residual forces and results in a stable and unique solution
regardless of the mesh size (Mohammadi and Taiebat 2013). In addition, application of the h-
adaptive remeshing technique increases the accuracy of the solution (Mohammadi and
Taiebat 2014) which is described in detail in next section.
 ACCEPTED MANUSCRIPT

3) h-adaptive mesh refinement

The final sequence of the excavation is accompanied by localization of shear strains and large crest
settlements and toe movements. The large deformations of the slope during failure results in distortion
of the finite element mesh within the shear zone which compromises the accuracy and convergence of
the solution. A typical deformed mesh at this stage is illustrated in Figure 9

T
R IP
SC
NU
Figure 9. To maintain the accuracy and ensure the convergence of the solution, the finite
element mesh should be updated at some intermediate stages. The main objective of
MA
remeshing is to generate a new and efficient finite element mesh which allows the analysis to
continue without disruption, although greater accuracy of the solution can also be achieved if
finer elements are generated around the shear band. This is achieved by employing the h-
D

adaptive remeshing technique.

P TE
CE
AC

Figure 9. Excessive mesh distortion

The adaptive strategy selected in this study is based on minimization of error in the
domain by reducing the element size in the zones of high error (Belytschko and Tabbara
1993; Deb et al. 1996; Hu and Randolph 1998a; Khoei and Lewis 2002; Lewis and Khoei
2001; Pastor et al. 1991; Peraire et al. 1987; Zienkiewicz et al. 1995). The error is estimated
based on the distribution of shear strains which are inherently larger in the shear zone. The
strain gradient criterion can capture the area of localization in a large deformation finite
element analysis (Belytschko and Tabbara 1993; Hu and Randolph 1998b; Mohammadi and
Taiebat 2014). The adaptive technique identifies the shear zone automatically without a priori
and refines the mesh around it. The process of mesh refinement starts after analyzing the
 ACCEPTED MANUSCRIPT

problem using an initial mesh. The initial analysis provides the first solution to the problem
that will be used to find parameters required for remeshing, such as the size of elements in
different locations. A new and improved mesh is generated based on the Advanced Front
Technique (Lo 1985) and all state parameters required for a non-linear elasto-plastic analysis,

T
such as stresses, strains and internal state variables like cohesion, friction angle and dilation

IP
for different soil layers, are transferred from the old mesh to the new mesh. The analysis is
continued until the solution errors become unacceptably large, that is, element distortions

R
become excessive, (Lee and Bathe 1994) which signals that remeshing is required again. This

SC
process is continued until the final solution is achieved.

NU
MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

Figure 10. Updated mesh: a) remeshing 1, b) remeshing 2, c) remeshing 3

During the analysis here the adaptive remeshing was performed three times. The finite
element meshes generated for the deformed shapes of the slope at different stages of the

T
R IP
SC
NU
MA
D
TE

analysis are plotted in Figure 10

P
CE

Figure 10. No further remeshing was necessary once the analysis converged to a stable
solution with an acceptable tolerance.
AC

5. COMPARISON WITH FIELD OBSERVATIONS

The location and the extent of the failure surface and the distribution of shear strains obtained
by the analysis are shown in Figure 11. The failure, which was induced by the excavation,
extends along the thin clayey silt layer within the yellow sand and does not interact with the
blue grey clay layer. The strain softening behavior of the clayey silt layer results in
progressive failure with large post-failure deformations. The analysis accurately captures the
location of the failure zone as reported by Del Prete and Hutchinson (1988), that is, the
failure surface base was on the clayey silt layer inclined 18 degrees to the horizontal. The
movement and sliding of the soil mass above the weak clayey silt layer is predicted by the
analysis in terms of the incremental displacement field and it is shown in Figure 12. This is
comparable to the field observations as well.
 ACCEPTED MANUSCRIPT

T
R IP
SC
NU
MA
D
TE

Figure 11. Cumulative shear strains: a) remeshing 1, b) remeshing 2, c) remeshing 3

The variation of the post-failure crest settlement at different stages of analysis is illustrated in
P

Figure 13. It shows that the crest settlement reaches 3.7 m at the end of the analysis. This
CE

corresponds to 4.8 m sliding of the soil mass along the slope, which is less than the observed
displacement but much larger than what has been obtained by other investigators so far. It
AC

should be noted that the dynamic and the inertia effects of the moving soil mass is not
considered in this study. Therefore it is expected that the post-failure deformation predicted
by this analysis to be less than the observed value. In addition, the finite element formulation
used here is based on continuum mechanics and cannot model the flow like deformation of
the disintegrated soil mass during the slide. It is believed that the soil mass was disintegrated
after such a large deformation and thus moved much further down the slope. The results of
the finite element analysis presented here show very large post-failure deformation which is a
good indication of the occurrence of landslide. Such a large deformation would clearly
impact the houses of the Senise hill below the toe where the excavation was performed.
 ACCEPTED MANUSCRIPT

a)

T
R IP
SC
NU
MA
b)

Figure 12. Incremental displacement field during slope failure: a) remeshing 1, b) remeshing
D

2, c) remeshing 3
P TE
CE

c)
AC
 ACCEPTED MANUSCRIPT

Vertical displacements Horizontal displacements

4.0

Post failure crest settlement (m) 3.0

remeshing 3

T
remeshing 2

IP
2.0

remeshing 1

R
1.0

SC
initial mesh
0.0
0 5000 10000 15000 20000 25000 30000

NU
No. of iterations

Vertical displacements Horizontal displacemnets

MA
2.0
Post failure toe movement (m)

remeshing 3
D

remeshing 2
TE

1.0
remeshing 1
P

initial mesh
CE

0.0
0 5000 10000 15000 20000 25000 30000

No. of iterations
AC

Figure 13. Post-failure crest settlements and toe movements with number of iterations

6. CONCLUSIONS
A novel approach has been presented in this paper that can be used to estimate the post-
failure deformation of slopes and embankments. This method is based on the large
deformation theory and the Updated Lagrangian formulation, incorporating a simple strain
softening constitutive model and an h-adaptive remeshing technique. The Updated
Lagrangian formulation accounts for the continuous changes of slope geometry during
failure. Frequent remeshing coupled with transformation of state parameters between the old
and the new meshes enables the simulation of post failure deformation of slopes by
 ACCEPTED MANUSCRIPT

maintaining the accuracy and the stability of the solution and preventing excessive distortion
of the finite element mesh as a result of large deformations. The analysis procedure
eliminates the problem of convergence associated with shear strain localization in strain
softening materials.

T
The large deformation analysis has been utilized to investigate the performance of the

IP
Senise slope after the excavation of its toe which resulted in a massive landslide in 1986. The

R
capability of the method was demonstrated through simulation of initiation and propagation
of progressive failure induced by the excavation and estimation of its post-failure

SC
deformation. The failure is shown to be clearly triggered by the excavation at the toe of the
slope. The finite element result showed that the retaining wall constructed at the toe was not

NU
able to prevent the failure. The numerical tool could simulate the location of the failure
surface and the sliding soil mass reasonably well. The failure surface has spread along the
MA
thin clayey silt layer which has a pronounced softening behavior. The post-failure
deformation of the sliding mass predicted by the finite element analysis was lower than the
actual deformation during the landslide. This is due to the limitation of the finite element
D

formulation based on continuum mechanics which cannot simulate the flow like deformation
TE

of materials during a landslide. However the large deformation predicted by the analysis
could be interpreted as a signal for the likelihood of landslide.
P

ACKNOWLEDGEMENTS
CE

This research was made possible through the support from the Institute for Infrastructure
Engineering in Western Sydney University. A special thanks to Professor Chin Leo and
AC

A/Prof Samanthika Liyanapathirana from the School of Computing Engineering and

Mathematics in Western Sydney University who provided insight and comments that greatly
improved the manuscript.

REFERENCES
Bathe, K.J., Ramm, E. & Wilson, E.L. 1975. Finite element formulations for large deformation dynamic
analysis. International Journal for Numerical Methods in Engineering, 9, 353-386.
Bazant, Z.P., Belytschko, T.B. & Chang, T.P. 1984. Continuum theory for strain softening. Journal of
Engineering Mechanics, 110, 1666-1692.
Belytschko, T., Fish, J. & Engelmann, B.E. 1988. A finite element with embedded localization zones. Computer
Methods in Applied Mechanics and Engineering, 70, 59-89.
Belytschko, T., Liu, W.K. & Moran, B. 2000. Nonlinear Finite Elements for Continua and Structures. Willey.
Belytschko, T. & Tabbara, M. 1993. H-Adaptive finite element methods for dynamic problems, with emphasis
on localization. International Journal for Numerical Methods in Engineering, 36, 4245-4265.
Brinkgreve, R.B.J. 1994. Geomaterial models and numerical analysis of softening. PhD, Delft University.
 ACCEPTED MANUSCRIPT

Carter, J.P. & Balaam, N.P. 1995. AFENA users’ manual. Centre for Geotechnical Research, University of
Sydney, Australia.
Chen, W.F. & Mizuno, E. 1990. Nonlinear analysis in soil mechanics, Theory and Implementation. Elsevier.
Conte, E., Donato, A. & Troncone, A. 2013. Back-analysis of the Senise landslide (Southern Italy). 4th
International Seminar on Forensic Geotechnical Engineering, Bengaluru, India.
Deb, A., Prevost, J.H. & Loret, B. 1996. Adaptive meshing for dynamic strain localization. Computer Methods
in Applied Mechanics and Engineering, 137, 285-306.

T
Del Prete, M. & Hutchinson, J.N. 1988. La frana di Senise del 26-7-1986 nel quadro morfologico del versante
meridionale della collina Timpone. Riv. Ital. Geotec., 22, 7-34.

IP
Griffiths, D.V. & Lane, P.A. 1999. Slope stability analysis by finite elements. Geotechnique, 49, 387-403.
Guerricchio, A. & Melidoro, G. 1988. Urbanizzazione e franositá delle formazioni sabbiose plio-pleistoceniche

R
in Basilicata: il caso di Senise. . Mem. Soc. Geol. It., 37, 745–774.
Hu, Y. & Randolph, M.F. 1998a. H-adaptive FE analysis of elasto-plastic non-homogeneous soil with large

SC
deformation. Computers and Geotechnics, 23, 61-83.
Hu, Y. & Randolph, M.F. 1998b. A practical numerical approach for large deformation problems in soil.
International Journal for Numerical and Analytical Methods in Geomechanics, 22, 327-350.
Jaumann, G. 1911. Geschlossenes system physikalischer und chemischer differenzialgesetze. Sitzungsberichte

NU
Akad. Wiss. Wien, 120.
Khoei, A.R. & Lewis, R.W. 2002. H-adaptive finite element analysis for localization phenomena with reference
to metal powder forming. Finite Elements in Analysis and Design, 38, 503-519.
Khoei, A.R., Lewis, R.W. & Zienkiewicz, O.C. 1997. Application of the finite element method for localized
MA
failure analysis in dynamic loading. Finite Elements in Analysis and Design, 27, 121-131.
Larsson, R., Runesson, K. & Sture, S. 1991. Finite element simulation of localized plastic deformation. Archive
of Applied Mechanics, 61, 305-317, doi: 10.1007/bf00787599.
Lee, N.S. & Bathe, K.J. 1994. Error indicators and adaptive remeshing in large deformation finite element
analysis. Finite Elements in Analysis and Design, 16, 99-139.
D

Lewis, R.W. & Khoei, A.R. 2001. Numerical analysis of strain localization in metal powder-forming processes.
International Journal for Numerical Methods in Engineering, 52, 489-501.
TE

Lo, S.H., , I. 1985. A new mesh generation scheme for arbitrary planar domains. International Journal of
Numerical Methods in Engineering, 21, 1403-1426.
Mohammadi, S. & Taiebat, H. 2014. H-Adaptive Updated Lagrangian Approach for LargenDeformation
P

Analysis of Slope Failure. International Journal of Geomechanics, ASCE, doi: 10.1061/(ASCE)GM.1943-

5622.0000440.
CE

Mohammadi, S. & Taiebat, H.A. 2013. A large deformation analysis for the assessment of failure induced
deformations of slopes in strain softening materials. Computers and Geotechnics, 49, 279-288.
Nazem, M., Carter, J.P., Sheng, D. & Sloan, S.W. 2009. Alternative stress integration schemes for large
deformation problems of solid mechanics. Finite Elements in Analysis and Design, 45, 934-943.
AC

Pastor, M., Peraire, J. & Zienkiewicz, O.C. 1991. Adaptive remeshing for shear band localization problems.
Archive of Applied Mechanics, 61, 30-39.
Peraire, J., Vahdati, M., Morgan, K. & Zienkiewicz, O.C. 1987. Adaptive remeshing for compressible flows.
Journal of Computational Physics, 72, 26-37.
Pietruszczak, S. & Mroz, Z. 1981. Finite element analysis of deformation of strain softening materials.
International Journal for Numerical Methods in Engineering, 17, 327-334.
Pietruszczak, S. & Stolle, D.F.E. 1985. Deformation of strain softening materials: Part I: Objectivity of finite
element solutions based on conventional strain softening formulations. Computers and Geotechnics, 1, 99-115.
Pietruszczak, S. & Stolle, D.F.E. 1987. Deformation of strain softening materials Part II: Modelling of strain
softening response. Computers and Geotechnics, 4, 109-123.
Potts, D.M., Dounias, G.T. & Vaughan, P.R. 1990. Finite element analysis of progressive failure of Carsington
embankment. Geotechnique, 40, 79-101.
Sterpi, D. 1999. An analysis of geotechnical problems involving strain softening effects. International Journal
for Numerical and Analytical Methods in Geomechanics, 23, 1427-1454.
Tang, H.X. 2008. Analysis for progressive failure of the Senise landslide based on Cosserat continuum model.
Landslides and Engineered Slopes. From the Past to the Future, Two Volumes + CD-ROM. CRC Press, 945-
950.
Troncone, A. 2005. Numerical analysis of a landslide in soils with strain softening behaviour. Geotechnique, 55,
585-596.
 ACCEPTED MANUSCRIPT

Troncone, A., Conte, E. & Donato, A. 2014. Two and three-dimensional numerical analysis of the progressive
failure that occurred in an excavation-induced landslide. Engineering Geology, 183, 265-275, doi:
http://dx.doi.org/10.1016/j.enggeo.2014.08.027.
Viggiani, C. & Di Maio, C. 1991. Convegno sul tema: Resistenza a taglio mobilitata nella frana di Senise.
Deformazioni in prossimita della rottura e resistenza dei terreni naturali e delle rocce., Ravello, Italy, 133–158.
Wang, S., Zheng, H., Li, C. & Ge, X. 2011. A finite element implementation of strain softening rock mass.
International Journal of Rock Mechanics and Mining Sciences, 48, 67-76.

T
Zienkiewicz, O.C., Huang, M. & Pastor, M. 1995. Localization problems in plasticity using finite elements with
adaptive remeshing. International Journal for Numerical and Analytical Methods in Geomechanics, 19, 127-148

R IP
SC
NU
MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

Figure 14. Bore hole results of the Senise area, adapted from Guerricchio & Melidoro (1988), soil
symbols: 1-sand with remolded gravel, 2-fine sand with gravel, 3-medium coarse sand, 4-weakly
compact cemented fine sand, 5-yellow sand with thin layers of silty clay, 6-blue grey clay, 7- fossils
Figure 15. Plan of landslide area with the marked location of section A–A (after Del Prete and
Hutchinson, (1988))

T
Figure 16. Schematic soil profile of the Senise hill at section A–A (adapted from Viggiani and Di

IP
Maio, (1991)

R
Figure 17. Senise hill geometry and soil profile at section A–A
Figure 18. Variation of shear strength parameter, , with shear strain,

SC
Figure 19. Experimental results of CD triaxial test: a) Sand b) Clayey Silt, (after Del Prete and
Hutchinson, (1988))

NU
Figure 20. Finite element mesh: a) before excavation and b) after excavation and location of houses
Figure 21. Sequence of progressive failure and development of cumulative shear strains during
MA
excavation steps: a) h=0.50hmax, , b)h=0.75hmax, , c)h=hmax
Figure 22. Excessive mesh distortion
Figure 23. Updated mesh: a) remeshing 1, b) remeshing 2, c) remeshing 3
D

Figure 24. Cumulative shear strains: a) remeshing 1, b) remeshing 2, c) remeshing 3

Figure 25. Incremental displacement field during slope failure: a) remeshing 1, b) remeshing 2, c)
TE

remeshing 3
Figure 26. Post-failure crest settlements and toe movements with number of iterations
P
CE
AC
 ACCEPTED MANUSCRIPT

Highlights

 A novel method to estimate the slopes post-failure deformations prone to landslide

T
 Post-failure deformations can be used as an indicator of landslides manifestation


IP
Large deformation theory, h-adaptive remeshing and a strain softening constitutive
model is combined


R
Predicted crest settlements show excavation has impacted houses on slope negatively
 Results agree with field observations: failure surface location, sliding soil mass

SC
NU
MA
D
P TE
CE
AC