You are on page 1of 11

Evaluation of Critical Slip Surface in Limit

Equilibrium Analysis of Slope Stability by Smoothed


Particle Hydrodynamics
Downloaded from ascelibrary.org by Chittagong University of Engineering and Technology on 11/14/19. Copyright ASCE. For personal use only; all rights reserved.

Liang Li1; Yu Wang, F.ASCE2; Limin Zhang, F.ASCE3; Clarence Choi4; and Charles W. W. Ng, F.ASCE5

Abstract: Limit equilibrium methods (LEMs) have been widely used for slope stability analysis in engineering practice. The critical slip sur-
face in LEM, which has the minimum factor of safety (FS), is often assumed as the most likely slip surface and subsequently used as the failure
slip surface in the analysis and design of slope stabilization measures, such as soil nails and anchors, for unstable slopes. This assumption has
not been validated systematically, probably due to the difficulty in simulating large displacement of soils during landslides. In this study, a
new mesh-free, particle-based numerical method in geotechnical engineering called smoothed particle hydrodynamics (SPH) is adopted to
simulate the entire process of landslides, including the large displacement of soils after a landslide initiates. A series of comparative studies on
the stability analysis of cohesive slopes is performed using both LEM and SPH. The comparative studies show that the assumption is reasona-
ble for stable slopes. However, for unstable slopes in which the location of failure slip surface is critically needed for the design of stabilization
measures, the LEM critical slip surface differs substantially from the failure slip surface, and the volumes of sliding soil masses are greatly
underestimated by the LEM critical slip surfaces. Using the LEM critical slip surface as the failure slip surface in the design leads to the unsafe
design of stabilization measures. A new method was proposed to properly locate the failure slip surfaces for unstable slopes when using LEM.
It is found that the failure slip surfaces for unstable slopes is not the slip surface with the minimum FS; instead, it is the one with FS = 1. DOI:
10.1061/(ASCE)GM.1943-5622.0001391. © 2019 American Society of Civil Engineers.
Author keywords: Slope stability; Landslide; Critical slip surface; Limit equilibrium method (LEM); Smoothed particle hydrodynamics
(SPH).

Introduction minimum FS is defined as the critical slip surface, along which slid-
ing is most likely to occur (Duncan and Wright 2005).
Limit equilibrium methods (LEMs) have been widely used for slope As presented in Fig. 1, five steps are often involved when iden-
stability analysis in engineering practice. Examples of commonly tifying the critical slip surface in LEM: (1) the type of potential
used LEMs include the Bishop method (Bishop 1955), methods con- slip surface (e.g., noncircular or circular slip surface) is assumed
sidering only force equilibrium (USACE 1970), Janbu’s generalized first and a potential slip surface is generated mathematically; (2)
procedure of slices (Janbu 1973), the Spencer method (Spencer after a potential slip surface is generated, the soil mass encom-
1967), and the Morgenstern and Price method (Morgenstern and passed by the ground surface and the slip surface is divided into a
Price 1965). In LEM, the factor of safety (FS) for a large number of finite number of vertical soil slices, each of which is treated as a
potential slip surfaces is calculated, and the slip surface with the rigid body for performing a free body diagram analysis (i.e., limit
equilibrium analysis with different forces acting on the slices,
such as gravitational force, interslice force, and forces from the
1
Associate Professor, School of Civil Engineering, Qingdao Univ. of underlain soil); (3) when the force and/or moment equilibrium
Technology, Qingdao 266033, PR China; formerly, Senior Research equations are solved, LEM provides an FS by which the shear
Associate, Dept. of Architecture and Civil Engineering, City Univ. of Hong strength parameters of soil (e.g., cohesion, c, and tangent of fric-
Kong, Tat Chee Ave., Kowloon, Hong Kong. Email: celldl@126.com tion angle, tan w ) are divided to bring the soil mass to satisfy the
2
Associate Professor, Dept. of Architecture and Civil Engineering, limit equilibrium condition; (4) Steps 2 and 3 are repeated for a
City Univ. of Hong Kong, Tat Chee Ave., Kowloon, Hong Kong (corre- large number of potential slip surfaces, and FS is obtained for
sponding author). Email: yuwang@cityu.edu.hk
3 each potential slip surface; and (5) the slip surface with the mini-
Professor, Dept. of Civil and Environmental Engineering, Hong Kong
Univ. of Science and Technology, Clear Water Bay, Kowloon, Hong mum FS, denoted as FSmin, is identified as the critical slip surface.
Kong. Email: cezhangl@ust.hk Many different methods have been developed in the LEM litera-
4
Research Assistant Professor, Dept. of Civil and Environmental ture for effective searching for the FSmin and rapid location of the
Engineering, Hong Kong Univ. of Science and Technology, Clear Water critical slip surface, such as variation principle based methods
Bay, Kowloon, Hong Kong. Email: ceclarence@ust.hk (Baker 1980; Leshchinsky and Ambauen 2015), random search
5
Chair Professor, Dept. of Civil and Environmental Engineering, Hong (Chen and Shao 1988), stability chart-based methods (Steward
Kong Univ. of Science and Technology, Clear Water Bay, Kowloon, et al. 2011; Duncan 1996), and heuristic algorithms (Cheng 2003;
Hong Kong. Email: cecwwng@ust.hk
Sun et al. 2008; Li et al. 2010a, b, 2013a, b; Li and Chu 2011; Bai
Note. This manuscript was submitted on April 3, 2018; approved on
October 18, 2018; published online on March 5, 2019. Discussion period et al. 2014; Gao 2016). LEMs, as described in Fig. 1, have been
open until August 5, 2019; separate discussions must be submitted for widely used in engineering practice and reported in the literature.
individual papers. This paper is part of the International Journal of Generally speaking, the value of FS for a potential slip surface
Geomechanics, © ASCE, ISSN 1532-3641. implies the probability that the slope may fail along this slip

© ASCE 04019032-1 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 04019032


Assume the possible slip surface type and
Step 1 generate a potential slip surface mathematically

Divide the soil mass between ground surface and


Step 2
slip surface into a finite number of vertical slices Fig. 2. (Color) Particle-based discretization for a slope domain.
Downloaded from ascelibrary.org by Chittagong University of Engineering and Technology on 11/14/19. Copyright ASCE. For personal use only; all rights reserved.

Calculate Factor of Safety (FS) for the slip surface 1977; Gingold and Monaghan 1977), hydrodynamics, and aero-
Step 3
using limit equilibrium equations nautics (Liu and Liu 2004). Because elastoplastic constitutive
models for soils have recently been implemented into the SPH
formulation (Bui et al. 2008), the formulation is able to simulate
Repeat Steps 1 to 3 for a large number of potential large deformation problems in geotechnical engineering, such as
Step 4 slip surfaces, and obtain FS for each potential slip landslides (An et al. 2016; He et al. 2017; Mao et al. 2017).
Unlike mesh-based numerical methods, such as the FEM, no ele-
surface ment mesh or small deformation assumption is needed for SPH
in which the problem domain is represented or discretized by
nodes or particles (see Fig. 2 for a slope example). This particle-
The slip surface with the minimum FS is based discretization of the problem domain makes SPH immune
Step 5 to the mesh distortion problem that occurs when using FEM to
identified as the critical slip surface in LEM
simulate large displacement problems. Therefore, SPH is able to
directly simulate large displacement problems in geotechnical
Fig. 1. Flowchart for identifying the critical slip surface in LEM. engineering, such as the entire process of a landslide, starting
from modeling of the in situ condition of the slope and initiation
of the landslide to transportation and deposition of the sliding
surface (Duncan and Wright 2005). A relatively low FS value soil masses.
suggests a relatively high probability of sliding, and vice versa. Fig. 2 illustrates the discretization of a slope domain into a fi-
Therefore, the critical slip surface, which has the FSmin, is treated nite number of particles with different radii, as indicated by the
as the most likely slip surface and subsequently used as the failure white and red particles, respectively. Each particle carries a set of
slip surface in engineering analysis and design, such as stabiliza- state variables (e.g., velocity, displacement, stress, strain) and
tion of unstable slopes. For example, the design of anchors or soil material properties (e.g., density, cohesion, friction angle,
nails often requires the location of the failure slip surface, and Young’s modulus, Poisson ratio). The interactions between par-
anchors or soil nails must exceed the failure slip surface ticles are simulated through a kernel function, W(.), acting within
(Sabhahit et al. 1995; Patra and Basudhar 2005). However, the a support domain. Consider, for example, the blue particle in
usage of the critical slip surface as the failure slip surface has not Fig. 2. It interacts with its surrounding yellow particles that col-
been validated systematically, particularly for unstable slopes (i. lectively form the support domain for the blue particle, and the
e., those with FS < 1), in which the location of the failure slip sur- interaction between the blue and yellow particles is quantified by
face is critically needed for design and analysis of their stabiliza- the W(.) function.
tion measures. Such validation has been challenging because of The motion of a particle in SPH is governed by the equations of
the difficulty in simulating the large displacement of soils that mass and momentum conservation (Liu and Liu 2004)
occurs after a landslide initiates. 8
>
>D r i ¼  r ∂va ¼ Xm vi  vk : ∂Wik
> i N
This paper systematically investigates the assumption of using >
>
> k a a
the critical slip surface as the failure slip surface by a new numerical < Dt i
∂xia k¼1 ∂xia
method in geotechnical engineering called smoothed particle hydro- !
>
> ∂riab X riab rkab
dynamics (SPH) (McDougall and Hungr 2004; Bui et al. 2008, >
> Dv i
a 1 N
∂Wik
>
> ¼ þ fa ¼
i
mk þ þ Cab
ik
i þ fa
i
2011). SPH is a mesh-free numerical method that is able to directly : Dt r i ∂xib ð r i Þ 2
ð r k Þ 2 ∂x b
k¼1
simulate the large displacement of soils that occurs during landslides
(An et al. 2016; He et al. 2017). A series of comparative studies on (1)
the stability of cohesive slopes is performed using both LEM and
SPH. The assumption of using the critical slip surface as the failure where N = number of particles in the support domain for a particle i;
slip surface is not appropriate for unstable slopes and leads to the r = soil density of particles; m = mass of particles;  = velocity of a
unsafe design of slope stabilization measures. Based on the compar- particle; t = time; Wik = kernel function value between particles i
ative study, a new method to identify the failure slip surface for and k; r = total stress tensor of a particle, the expression of which is
unstable slopes in the LEM is proposed. After this introduction, a determined in accordance with the soil constitutive model adopted;
brief review of SPH is provided, followed by the comparative study. x = coordinates of a particle that has two horizontal and one vertical
component in a Cartesian system; a and b = three Cartesian compo-
nents with the Einstein convention applied to repeated indices; fa =
Brief Review of SPH component of acceleration due to externally applied force; and
ik
Cab = artificial terms for improving the numerical stability in SPH.
ik
SPH is a mesh-free, particle-based numerical method that has Details for Cab can be found in Bui et al. (2008), An et al. (2016),
been widely used in many fields, such as astrophysics (Lucy and Mao et al. (2017).

© ASCE 04019032-2 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 04019032


The Drucker-Prager (D-P) constitutive model is adopted for soils
in this study. Detailed formulation of implementing the D-P model Slope domain discretization
in SPH, together with a detailed description of the SPH equations
used in this study, is found in Bui et al. (2008). In addition, a cubic
spline function can be used as the kernel function (Liu and Liu Input parameter configuration
2004).
SPH performs the numerical simulations through a series of time
steps. At time t = 0, the velocity, strain, and stress of each particle Indicator of time step, nt=0
are all set to zero, and a gravitational force is applied to each particle
Downloaded from ascelibrary.org by Chittagong University of Engineering and Technology on 11/14/19. Copyright ASCE. For personal use only; all rights reserved.

as the external force fa. Because of the gravitational force, all par-
ticles move downward with the same acceleration. As the time nt= nt+1, determine the support
increases to t = 0 þ Dt, where Dt = time increment, the particles near domains for all particles
the boundaries cannot move beyond the boundaries, resulting in
stresses among the particles. Subsequently, the strains are induced in
particles according to the elastic or elastoplastic stress–strain rela-
tionships specified by the D-P model. Then the velocity increments Calculate D ρ , Dv and D σ
and coordinates for all particles are recalculated and updated, fol- Dt Dt Dt
lowed by recalculation and further update of the stresses and strains
for all particles. This iteration continues one time increment after
another, until the particles reach a steady state or satisfy convergence Yes
Mod(nt-1,M)=0
criteria.
In this study, the Verlet time integration scheme (Verlet 1967;
An et al. 2016) is used for the simulations along the time steps in No
Update state variables
SPH. This scheme includes two evolution parts: (1) the state varia-
bles of particles evolve in accordance with Eq. (2); and (2) on every by Equation (3)
M (e.g., 50) time step, they are updated by Eq. (3)
8  t
>
> Dr Update state variables
>
>r ¼ r þ 2Dt
tþ1 t1
>
> by Equation (2)
>
>
Dt
>
>  t
>
>
>
> Drab
>r tþ1
< ab ¼ rt1
ab þ 2Dt
Dt Calculate stress and strain for all
>  t (2)
>
> Dv particles using D-P model
>
> vtþ1 ¼ vt1 þ 2Dt
>
> Dt
>
>
>
>  t
>
>
>
> ¼ þ Dtv þ ðDt Þ2 Dv
: x tþ1
x t t
0:5
Dt
No Convergence is
8 reached?
 t
>
> Dr
>
> r ¼ r þ Dt
tþ1 t
>
>
>
>
Dt Yes
>
>  t
>
>
>
> Drab
> rtþ1
< ab ¼ r t
ab þ Dt Termination of SPH
Dt
>  t (3)
>
> D
>
>  tþ1 ¼  t þ Dt Fig. 3. Flowchart of SPH simulation of landslides.
>
> Dt
>
>
>
>  t
>
>
>
> ð Þ 2 D
: x ¼ x þ Dt þ 0:5 Dt
tþ1 t t
Dt
An in-house Fortran-based SPH software package has been
Fig. 3 is a flowchart for SPH simulations of landslides. As pre- developed using the previously mentioned formulations and equa-
sented in Fig. 3, the SPH starts with the discretization of the slope tions (Li and Wang, Forthcoming). The in-house SPH software
domain into a large number of particles with a prespecified radius, package has been validated through comparison with SPH and FEM
followed by configuration of input parameters, including initializa- results, as well as physical model test results, reported in the litera-
tions of state variables (e.g., velocities, strains, stresses) and specifi- ture. Fig. 4 illustrates the SPH simulation of the entire process of a
cation of SPH simulation parameters (e.g., time step Dt and smooth- landslide. The slope simulated is the Unstable Slope U1 discussed
ing length h). For each time step, the support domains for all later in the section “Critical Slip Surfaces for Unstable Cohesive
particles are identified, and Eq. (1) is adopted to calculate the deriv- Slopes.” The original condition of the slope before the landslide is
atives of r , v, and r over time t. Then, either Eq. (2) or Eq. (3) is illustrated in Fig. 4(a) at t = 0 s. As time increases to t = 0.5 s, notice-
used to update the values for r , v, and r. The time step increases, able soil displacement (e.g., the maximum displacement up to about
and the SPH simulation continues until steady-state or convergent 0.2 m) is observed in Fig. 4(b), and the landslide initiates. After that,
criteria are reached for all particles. the soils in the slope continue to move. Both the maximum soil

© ASCE 04019032-3 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 04019032


2004) is used for LEM analysis, and the in-house SPH software is
used to evaluate the slope stability and landslide process of the
same slopes, if landslides occur. The slopes are divided into two cat-
egories: (1) stable slopes that have an FS ≥ 1 and small soil dis-
placements; and (2) unstable slopes that have FS < 1 and large soil
displacements or sliding of soil masses. This section focuses on sta-
ble slopes, and the unstable slopes are reported in the following
section.
Fig. 5 illustrates geometry of the cohesive slopes used in the
Downloaded from ascelibrary.org by Chittagong University of Engineering and Technology on 11/14/19. Copyright ASCE. For personal use only; all rights reserved.

comparative study. As presented in Fig. 5, consider a slope with a


height H and a slope angle a. Three coefficients n1, n2, and d are
defined in Fig. 4 to specify the slope boundaries in the left, right,
and bottom directions, respectively. Table 1 summarizes informa-
tion for 10 stable cohesive slopes (i.e., Slope ID: S1–S10 in
Table 1) studied in this section. For example, the slope with the ID
S1 has a slope height H = 5 m; slope angle a = 26.6°; soil cohesion
c = 30 kPa; soil unit weight g = 20 kN/m3; and other coefficients
n1 = 2.0, n2 = 5.0, and d = 0.5. Stability analysis of Slope S1 is first
evaluated using the LEM commercial software SLOPE/W
(GEOSLOPE International 2004). The Spencer method is used to
calculate the FS for a large number of circular and noncircular
potential slip surfaces. The noncircular slip surfaces are generated
in SLOPE/W by connecting a series of smooth points, and an auto-
matic search method in SLOPE/W is adopted to locate the critical
slip surface as the failure slip surface. SLOPE/W provides a mini-
mum FS value of 1.80, and the corresponding critical slip surface is
plotted in Fig. 6(a) by a red line, which is obviously a noncircular
composite slip surface (i.e., a horizontally straight slip surface in
the center plus two circular slip surfaces in left and right,
respectively).
The in-house SPH package is also used to evaluate the stability
of Slope S1. Fig. 7(a) gives a color contour plot of the soil displace-
ment obtained from SPH. The maximum soil displacement in the
slope is less than 0.01 m, and the slope is stable without any noticea-
ble deformation. The SPH result is consistent with the FS = 1.80
obtained from LEM. To obtain the FS value in SPH and to simulate
the initiation of slope failure, a shear strength reduction (SSR)
method is used in SPH. The SSR method has been commonly used
in the FEM simulation of slope stability (Dawson et al. 1999; Wei
et al. 2009; Zheng et al. 2009; An et al. 2016). The shear strength of
soil (i.e., cohesion c in this study) is reduced by a factor of FS =
1.80, and SPH simulation is reperformed using the factored c value
(i.e., 30/1.8 = 16.7 kPa). To achieve a consistent comparison with
LEM results, a reduction factor of 1.80, which is identical to the FS
value from LEM, is used in SPH.
Fig. 4. (Color) SPH simulation of the entire process of a landslide: Fig. 6(a) plots the soil displacement contours in the slope
(a) t = 0 s; (b) t = 0.5 s; (c) t = 1.0 s; (d) t = 1.5 s; (e) t = 2.0 s; and (f) t = obtained from the SPH analysis with the factored c value. The fail-
6.0 s. ure slip surface can be identified in accordance with the variation of
soil displacement magnitude in the slope [i.e., variation of color
contours in Fig. 6(a)]. The color contours in Fig. 6(a) can be divided
displacement and the volume of soils that have noticeable displace- into two zones: a sliding zone with noticeable soil displacements
ment increases (see the change of colors in Fig. 4), as the time and a stationary zone with fairly small or negligible soil displace-
gradually increases to t = 1.0 s [Fig. 4(c)], 1.5 s [Fig. 4(d)], 2.0 s ments. The boundary between the sliding and stationary zones is the
[Fig. 4(e)], and 6.0 s [Fig. 4(f)]. All soil particles reach a steady failure slip surface. For example, the boundary between the cyan
state at t = 6.0 s, and the SPH simulation is terminated. The final color zone and blue color zone in Fig. 6(a) may be identified as the
deposition of the sliding soil masses is presented in Fig. 4(f). failure slip surface, if the soils in the blue color zone are treated as
stationary. The failure slip surface agrees reasonably well with the
noncircular critical slip surface obtained from the LEM (see the red
Critical Slip Surfaces for Stable Cohesive Slopes line in Fig. 6(a)]. To quantify the difference between the failure slip
surface obtained from SPH and the critical slip surface obtained
A series of comparative studies on the stability analysis of cohesive from LEM, the volume, V, of the sliding soil mass is calculated for
slopes is performed in this study using both LEM and SPH. A com- these two slip surfaces. The V obtained from SPH is 76.4 m2 per me-
mercial software called SLOPE/W (GEOSLOPE International ter long of the slope; whereas the V obtained from LEM is 72.3 m2

© ASCE 04019032-4 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 04019032


n2H

H Soil cohesion: c
Soil unit weight: γ = 20 kN/m3
n1H α

dH
Downloaded from ascelibrary.org by Chittagong University of Engineering and Technology on 11/14/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Geometry of the cohesive slopes used in the comparative study.

Table 1. Summaries of information for stable cohesive slopes unit weight g = 20 kN/m3; and other coefficients n1 = 5.0, n2 = 8.0,
and d = 0.5. Similar to those reported in the previous section for sta-
Slope ID H (m) a (degrees) n1 n2 d c (kPa) FS
ble slopes, LEM analysis is performed using SLOPE/W to calculate
S1 5 26.6 2.00 5.00 0.50 30.0 1.80 the FS and locate the critical slip surface. An FS value of 0.49 is
S2 8 26.6 1.50 2.50 0.50 40.0 1.49 obtained from SLOPE/W for Slope U1, which is obviously unsta-
S3 2 45.0 1.50 2.50 0.50 10.0 1.25 ble. Fig. 8(a) illustrates the critical slip surface obtained from
S4 10 26.6 1.50 2.50 0.30 55.0 1.75 SLOPE/W with a black line. The critical slip surface for Slope U1 is
S5 6 26.6 1.50 2.33 0.33 30.0 1.58 noncircular and contains a horizontally straight slip surface along
S6 4 45.0 1.50 2.50 0.50 20.0 1.39 the slope bottom in the middle.
S7 4 33.7 2.50 6.00 1.00 20.0 1.30 The in-house SPH software is also used to evaluate the stability
S8 10 45.0 1.50 2.00 0.20 45.0 1.18 of Slope U1 and to simulate the slope failure process. Unlike the
S9 4 38.7 2.50 3.75 0.50 17.0 1.11 SPH simulation of stable slopes reported in the previous section, no
S10 5 26.6 2.00 5.00 0.50 17.0 1.02 SSR method is needed for unstable slopes (i.e., the cohesion value c
is directly used in the SPH simulation without being factored)
per meter long of the slope. The relative difference between these because relatively large soil displacements occur due to the low FS
two sliding soil mass volumes is about 5%. value. The entire process of slope failure is simulated by SPH, start-
Similar comparative studies are performed for the other nine sta- ing from modeling of the original site conditions and initiation of
ble slopes listed in Table 1, and similar results are observed. For slope failure to transportation and deposition of sliding soil masses.
example, the critical slip surfaces for Slopes S2–S6 obtained from Fig. 8(a) also illustrates the SPH results, including the deformed
SLOPE/W are plotted by red lines in Figs. 6(b–f), respectively. The slope geometry after the landslide and variation of the soil displace-
corresponding FS values are 1.49, 1.25, 1.75, 1.58, and 1.39, as ments in the slope. Different colors are used in Fig. 8(a) to represent
listed in Table 1. The soil displacement contours in these slopes different magnitudes of soil displacement. Similar to Fig. 6, the
with and without SSR are given in Figs. 6(b–f) and Figs. 7(b–f), SPH failure slip surface can be identified in accordance with the
respectively, for Slopes S2–S6. The maximum soil displacements variation of soil displacement magnitude in the slope [i.e., variation
in Figs. 7(b–f) are all less than 0.01 m, implying that Slopes S2–S6 of color contours in Fig. 8(a)]. For example, if soils with a displace-
are stable or have very small soil displacements. On the other hand, ment less than a threshold of 0.1 m are treated as stationary, the
the maximum soil displacements in slopes increase significantly, as boundary between the orange and green color zones in Fig. 8(a) can
illustrated in Figs. 6(b–f), when the SSR method is used to simulate be identified as the failure slip surface.
the initiation of slope failure. Similar to the observation in Fig. 6(a), Fig. 8(a) demonstrates that the LEM critical slip surface (i.e., the
good agreement between the critical slip surfaces from LEM [see black line) differs significantly from the soil sliding pattern that
the red lines in Figs. 6(b–f)] and the failure slip surfaces from SPH occurs during the landslide. The volume of sliding soil mass corre-
with SSR is observed in Figs. 6(b–f). The difference in the volume sponding to the LEM critical slip surface is substantially smaller
of sliding mass between the LEM critical slip surface and SPH fail- than the volume of soils that slide after the landslide occurs. To
ure slip surface is calculated and summarized in Table 2. About quantify the difference in the sliding mass volume, the V for the
2–8% of relative difference is obtained for the 10 stable slopes in LEM critical slip surface in Fig. 8(a) is estimated as 8.9 m2 per me-
Table 2. Generally speaking, the critical slip surfaces from LEM are ter long of the slope; whereas the V from the SPH failure slip surface
consistent with the failure slip surface from SPH with SSR. is 26.8 m2 per meter long of the slope. The V from the SPH failure
Therefore, it is reasonable to use the critical slip surfaces from LEM slip surface is about three times larger than that from the LEM criti-
as the failure slip surfaces for stable cohesive slopes, and LEM per- cal slip surface. The relative difference in V can be calculated as
forms reasonably well for stable cohesive slopes. (26.8 – 8.9)/26.8 = 66.8%. Using the LEM critical slip surface as
the failure slip surface for unstable slopes greatly underestimates
the volume of sliding soil mass. In addition, using the LEM critical
Critical Slip Surfaces for Unstable Cohesive Slopes slip surface as the failure slip surface in the design and analysis of
stabilization measures (e.g., soil nails or anchors) for the unstable
A series of comparative studies on the locations of critical slip slopes leads to unconservative results and unsafe design because
surfaces for unstable cohesive slopes is performed using LEM and the soil nails and anchors must extend beyond the failure slip surfa-
SPH in this section. Table 3 lists the details of 20 unstable slopes ces for effective mobilization of stabilizing resistance.
with IDs ranging from U1 to U20. Consider, for example, an unsta- Similar comparative studies are performed for the other 19
ble cohesive slope with a slope ID of U1. This slope has a slope unstable slopes listed in Table 3, and similar results are observed.
height H = 2 m; slope angle a = 45°; soil cohesion c = 4 kPa; soil For example, Figs. 8(b–f) demonstrate the results for Slopes U2–

© ASCE 04019032-5 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 04019032


Downloaded from ascelibrary.org by Chittagong University of Engineering and Technology on 11/14/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. (Color) Comparison between LEM critical slip surface and fail- Fig. 7. (Color) Contour of soil displacements in slope from SPH with-
ure slip surface from SPH with shear strength reduction: (a) stable out shear strength reduction: (a) stable Slope S1; (b) stable Slope S2;
Slope S1; (b) stable Slope S2; (c) stable Slope S3; (d) stable Slope S4; (c) stable Slope S3; (d) stable Slope S4; (e) stable Slope S5; and
(e) stable Slope S5; and (f) stable Slope S6. (f) stable Slope S6.

© ASCE 04019032-6 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 04019032


Table 2. Comparison of sliding mass volume obtained from LEM and
SPH for stable slopes

Sliding mass
volume [V (m2)]
Relative
Slope ID LEM SPH difference (%)
S1 72.3 76.4 5.37
S2 177.0 184.0 3.80
S3 9.1 8.6 5.81
Downloaded from ascelibrary.org by Chittagong University of Engineering and Technology on 11/14/19. Copyright ASCE. For personal use only; all rights reserved.

S4 205 216.8 5.44


S5 75.7 82.2 7.90
S6 35.7 38.4 7.03
S7 78.0 84.0 7.14
S8 142.0 144.3 1.59
S9 37.3 40.0 6.75
S10 71.22 76.10 6.41

Table 3. Summaries of information for unstable cohesive slopes

Slope ID H (m) a (degrees) n1 n2 d c (kPa) FS


U1 2 45.0 5.0 8.0 0.50 4.0 0.49
U2 3 31.0 3.3 6.7 0.33 5.0 0.49
U3 4 33.7 2.5 6.0 1.00 12.5 0.81
U4 5 26.6 2.0 5.0 0.50 11.7 0.70
U5 8 26.6 2.0 4.0 0.50 20.0 0.75
U6 10 45.0 2.0 5.0 0.50 30.0 0.71
U7 2 45.0 5.0 8.0 0.50 5.0 0.62
U8 3 26.6 3.0 5.0 0.33 5.0 0.53
U9 3 45.0 3.0 6.0 0.33 5.0 0.42
U10 3 36.8 3.3 7.0 0.33 5.0 0.45
U11 4 33.7 2.5 6.0 0.13 10.0 0.77
U12 5 26.6 2.0 5.0 0.50 15.0 0.90
U13 5 26.6 2.0 5.0 0.50 13.0 0.78
U14 6 36.8 2.0 3.3 0.40 20.5 0.90
U15 6 36.8 2.0 3.3 0.40 20.0 0.87
U16 7 41.2 2.0 3.3 0.43 21.0 0.77
U17 7 35.0 2.0 3.0 0.43 22.0 0.86
U18 8 26.6 2.0 4.0 0.50 22.2 0.83
U19 10 45.0 2.0 5.0 0.50 37.5 0.91
U20 10 45 2.0 5.0 0.5 34.61 0.87

U6, respectively. The LEM critical slip surfaces (see the black lines
in Fig. 8) are quite different from the slope sliding pattern that
occurs during landslides. For unstable slopes, only the initial failure
slip surface at the initiation of the landslide can be obtained by
LEM. In contrast, the final failure slip surface after transportation
and deposition of the sliding soil mass is determined in SPH. The
LEM critical slip surfaces significantly underestimate the volume of
sliding soil masses. Table 4 compares the sliding mass volumes
obtained from LEM and SPH for unstable slopes. The relative dif-
ference between them is up to 76%, which means that the V corre-
sponding to the LEM critical slip surface is just a quarter of the vol-
ume of soils that slide during landslides.
Fig. 9 plots a variation of relative difference in the sliding mass Fig. 8. (Color) Comparison between LEM and SPH results for unsta-
volume with FS. As the FS value for unstable slopes increases from ble slopes: (a) unstable Slope U1; (b) unstable Slope U2; (c) unstable
0.40 to 0.91, the relative difference decreases dramatically from 75 Slope U3; (d) unstable Slope U4; (e) unstable Slope U5; and
to 9%. The smaller the FS value from LEM, the larger the relative (f) unstable Slope U6.
difference. When FS is larger than 1 and the slopes are stable, the
relative difference only varies slightly and remains at about 5%. It is
evident that, for unstable slopes, using the LEM critical slip surfaces although it is a reasonable approximation to use the LEM critical
as failure slip surfaces leads to substantial underestimation of slid- slip surfaces as the failure slip surfaces for stable slopes. The
ing mass volume and unsafe design of slope stabilization measures, assumption of using the LEM critical slip surface as failure slip

© ASCE 04019032-7 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 04019032


surfaces is not valid for unstable slopes. A new method is therefore the boundary that separates the sliding soil mass from the stationary
needed to properly locate the failure slip surfaces for unstable slopes soil mass. Note that a slope will change from a stable slope to an
when using LEM, as proposed in the next section. unstable one when the FS decreases from a value larger than 1 to a
value smaller than 1. Therefore, it is logical to expect that a potential
slip surface with FS = 1 is the boundary that separates the sliding
New Method to Locate Failure Slip Surfaces for soil mass from the stationary soil mass for unstable slopes. In other
Unstable Cohesive Slopes in LEM words, in the search of the failure slip surface for unstable slopes
(i.e., FSmin < 1) when using LEM, effort should not be only spent
A new method is proposed in this section to properly locate the fail- on searching for the FSmin and its corresponding potential slip sur-
Downloaded from ascelibrary.org by Chittagong University of Engineering and Technology on 11/14/19. Copyright ASCE. For personal use only; all rights reserved.

ure slip surfaces for unstable slopes when using LEM. For unstable face. Instead, the location of the potential slip surfaces with FS = 1
slopes, the FSmin along the critical slip surface is smaller than 1, and is more important and should be identified. Furthermore, in the
the soil mass encompassed by the critical slip surface will slide dur- search for the potential slip surfaces with FS = 1 for unstable slopes,
ing landslides. However, as presented in Fig. 8, some soils beyond it is possible that multiple potential slip surfaces are found to have
the critical slip surface will also slide. The critical slip surface is not an FS = 1. In this case, among these potential slip surfaces with
FS = 1, the one with the maximum volume, V, of sliding soil mass
Table 4. Comparison of sliding mass volume obtained from LEM and might be used as the most likely failure slip surface, if only one fail-
SPH for unstable slopes ure slip surface is to be identified.
To validate the new method proposed to locate failure slip surfa-
Sliding mass volume ces for unstable slopes, results from a large-scale experimental model
from LEM [V (m2)] test reported by Jia et al. (2009) were used. A large-scale sandy silt
Relative
Slope ID LEM SPH difference (%) soil slope model, with a length of 15 m, height of 6 m, and width of
5 m, was constructed and adopted to simulate slope failure induced
U1 8.9 26.8 66.79
by rapid drawdown of the water level (Jia et al. 2009). Based on the
U2 16.8 61.1 72.50
results from isotropic consolidated, undrained triaxial compression
U3 75.6 104.3 27.52
tests, the soil has an effective friction angle of 30°. After the slope
U4 71.8 137.4 47.74
model was constructed, the slope was fully submerged in water for
U5 184.1 318.7 42.23
72 h. Then, the water level was rapidly drawn down by 1.9 m to
U6 218.0 351.2 37.93
U7 9.1 16.9 46.15
U8 20.3 60.6 66.50
U9 14.9 61.6 75.81
U10 15.0 60.4 75.20
U11 20.9 31.6 33.86
U12 70.8 84.9 16.61
U13 71.9 102.2 29.65
U14 73.4 85.0 13.65
U15 73.4 86.1 14.75
U16 101.4 146.1 30.60
U17 110.4 143.0 22.80
U18 185.6 240.7 22.89
U19 221.6 244.9 9.51 Fig. 10. Slope model geometry and observed failure slip surfaces.
U20 230.2 297.8 22.70 (Adapted from Jia et al. 2009.)

80
Relative difference in volume of sliding soil

Slip surface with minimum FS

Proposed method
60
masses (%)

40

20

0
0 0.5 1 1.5 2
Factor of Safety, FS

Fig. 9. Variation of relative difference in sliding mass volume with FS.

© ASCE 04019032-8 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 04019032


Downloaded from ascelibrary.org by Chittagong University of Engineering and Technology on 11/14/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. (Color) Failure slip surfaces obtained from the proposed method for unstable slopes. (a) unstable Slope U1; (b) unstable Slope U2;
(c) unstable Slope U3; (d) unstable Slope U4; (e) unstable Slope U5; and (f) unstable Slope U6.

© ASCE 04019032-9 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 04019032


Table 5. Comparison of sliding mass volume obtained from SPH and pro- slope. When compared with the V value of 26.8 m2/m down the
posed method in LEM for unstable slopes slope obtained from SPH, the relative difference is less than 1%.
Sliding mass volume The V values obtained from the proposed method are estimated for
[V (m2)] all the 20 unstable slopes in Table 3, and the results are summarized
Relative in Table 5 and plotted in Fig. 9. The V values obtained from the pro-
Slope ID LEM SPH difference (%)
posed method when using LEM agree reasonably well with those
U1 26.5 26.8 1.12 obtained from SPH. The proposed method significantly improves
U2 59.3 61.1 2.95 the performance of LEM when locating failure slip surfaces and
U3 108.8 104.3 4.31 estimating sliding mass volume for unstable slopes.
Downloaded from ascelibrary.org by Chittagong University of Engineering and Technology on 11/14/19. Copyright ASCE. For personal use only; all rights reserved.

U4 137.0 137.4 0.29


U5 303.7 318.7 4.71
U6 337.7 351.2 3.84 Conclusions
U7 17.3 16.9 2.37
U8 53.7 60.6 11.39 The LEM critical slip surface with the minimum FS has been long
U9 55.7 61.6 9.58
assumed in the literature, and widely used in engineering practice,
U10 57.5 60.4 4.80
as the failure slip surface for design and analysis of slope stabiliza-
U11 33.2 31.6 5.06
tion measures. However, this assumption has not been validated
U12 86.2 84.9 1.53
systematically, probably due to the difficulty in simulating the large
U13 105.0 102.2 2.74
displacement of soils that occur during landslides. A new numerical
U14 81.1 85.0 4.59
method in geotechnical engineering called SPH was used in this
U15 84.5 86.1 1.86
U16 139.4 146.1 4.59
study to simulate the entire process of landslides, including the large
U17 148.8 143.0 4.06
displacement of soils during landslides, and to systematically vali-
U18 256.4 240.7 6.52 date the assumption through a series of comparative studies using
U19 246.2 244.9 0.53 both LEM and SPH. The comparative studies show that using the
U20 306.8 297.8 3.02 LEM critical slip surface as the failure slip surface is reasonable for
stable slopes with FS > 1, but for unstable slopes with FS < 1, the
failure slip surface differs substantially from the LEM critical slip
induce slope failure. Fig. 10 presents three separate failure slip surfa- surface. The volumes of sliding soil mass are greatly underesti-
ces (i.e., L1–L3) observed from the large-scale experiment. As mated by the LEM critical slip surfaces, which provide estimates
reported by Jia et al. (2009), L1 was relatively shallow and occurred that might be as low as just a quarter of the volumes of sliding soil
first, followed by L2 and L3, which started to slide at the same time. mass observed. Using the LEM critical slip surface as the failure
A slope stability analysis of water level rapid drawdown is performed slip surface in the design and analysis of stabilization measures
using LEM software SLOPE/W, and the FS values for the observed (e.g., soil nails or anchors) for the unstable slopes leads to unconser-
slip surfaces L1–L3 are calculated as 0.764, 0.989, and 1.026, respec- vative results and unsafe design because the soil nails and anchors
tively. A careful examination of the results from SLOPE/W and the must extend beyond the failure slip surfaces for effective mobiliza-
model test indicates that the slip surface with the minimum FS is the tion of stabilizing resistance.
initiating sliding surface (i.e., slip surface L1 has the minimum FS of A new method was proposed to properly locate the failure slip
0.764, and it was observed as the first slip surface when the sliding surfaces for unstable slopes when using LEM. The failure slip surfa-
initiated during the model test). However, this initiating sliding sur- ces for unstable slopes with FSmin < 1 is not the slip surface with
face is different from the final sliding surface (i.e., slip surface L3), FSmin, but it is the one with FS = 1. In the search of the failure slip
which has an FS value of about 1.0. The final failure slip surface in a surface for unstable slopes when using LEM, effort should not be
landslide does not have the minimum FS, but it has an FS value of spent only on searching for the FSmin and its corresponding poten-
about 1.0. The large-scale model test results reported by Jia et al. tial slip surface. Instead, the location of the potential slip surfaces
(2009) validate the proposed method for locating failure slip surface with FS = 1 is more important and should be identified.
using FS = 1, instead of using minimum FS. Comparative studies were also performed and showed that the pro-
The proposed methods for locating the failure slip surface for
posed method significantly improves the performance of LEM
unstable slopes are used in LEM to reanalyze the 20 unstable slopes
when locating failure slip surfaces and estimating sliding mass vol-
in Table 3. The black lines in Fig. 11 illustrate the failure slip surfa-
ume for unstable slopes.
ces for Slopes U1–U6 using the proposed method. The original
LEM critical slip surfaces are also shown by cyan lines for compari-
son. For example, Fig. 11(a) illustrates the results for Unstable
Slope U1. The failure slip surface from the proposed method [see Acknowledgments
the black line in Fig. 11(a)] covers a much larger volume of soils
than the original LEM critical slip surface [see the cyan line in The author would like to thank the anonymous reviewers for their
Fig. 11(a)]. The black line is more consistent with the SPH results constructive comments that helped improve the quality of the
than the cyan line. Similar observations can also be made in Figs. manuscript. The second author would also like to thank Mr. Xin
11(b–f). The failure slip surfaces obtained from the new method are Liu at City University of Hong Kong for performing slope
more consistent with the sliding pattern of soils during landslides stability analysis during revision. The work described in this paper
than the original LEM critical slip surfaces with the FSmin. was supported by a grant from the Research Grants Council of the
To quantify the difference in volume of sliding soil mass, the V Hong Kong Special Administrative Region, China (Project T22-
values obtained from the proposed method are also calculated. For 603/15N), and a grant from National Natural Science Foundation
example, for Slope U1, the V value corresponding to the failure slip of China (Grant 51778313). The financial support is gratefully
surface obtained from the proposed method is 26.5 m2/m down the acknowledged.

© ASCE 04019032-10 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 04019032


References Li, L., and X. S. Chu. 2011. “An improved particle swarm optimization
algorithm with harmony strategy for the location of critical slip surface
An, Y., Q. Wu, C. Shi, and Q. Liu. 2016. “Three-dimensional smoothed- of slopes.” China Ocean Eng. 25 (2): 357–364. https://doi.org/10.1007
particle hydrodynamics simulation of deformation characteristics in /s13344-011-0030-9.
slope failure.” Geotechnique 66 (8): 670–680. https://doi.org/10.1680 Li, L., and Y. Wang. Forthcoming. “Identification of failure slip surfaces for
/jgeot.15.P.222. landslide risk assessment using smoothed particle hydrodynamics.”
Bai, T., T. Qiu, X. Huang, and C. Li. 2014. “Locating global critical slip sur- GeoRisk.
face using the Morgenstern-Price method and optimization technique.” Li, L., Y. Wang, Z. J. Cao, and X. S. Chu. 2013b. “Risk de-aggregation and
Int. J. Geomech. 14 (2): 319–325. https://doi.org/10.1061/(ASCE)GM system reliability analysis of slope stability using representative slip
.1943-5622.0000312. surfaces.” Comput. Geotech. 53 (Sep): 95–105. https://doi.org/10.1016/j
Downloaded from ascelibrary.org by Chittagong University of Engineering and Technology on 11/14/19. Copyright ASCE. For personal use only; all rights reserved.

Baker, R. 1980. “Determination of the critical slip surface in slope stability .compgeo.2013.05.004.
computations.” Int. J. Numer. Anal. Methods Geomech. 4 (4): 333–359. Li, L., G. M. Yu, Z. Y. Chen, and X. S. Chu. 2010a. “Discontinuous flying
https://doi.org/10.1002/nag.1610040405. particle swarm optimization algorithm and its application to slope stabil-
Bishop, A. W. 1955. “The use of the slip circle in the stability analysis ity analysis.” J. Central South Univ. Technol. 17 (4): 852–856. https://
of slopes.” Geotechnique 5 (1): 7–17. https://doi.org/10.1680/geot.1955 doi.org/10.1007/s11771-010-0566-5.
.5.1.7. Li, Y. C., Y. M. Chen, T. L. T. Zhan, D. S. Ling, and P. J. Cleall. 2010b.
Bui, H. H., R. Fukagawa, K. Sako, and S. Ohno. 2008. “Lagrangian mesh- “An efficient approach for locating the critical slip surface in slope sta-
free particles method (SPH) for large deformation and failure flows of bility analyses using a real-coded genetic algorithm.” Can. Geotech. J.
geomaterial using elastic-plastic soil constitutive model.” Int. J. Numer. 47 (7): 806–820. https://doi.org/10.1139/T09-124.
Anal. Methods Geomech. 32 (12): 1537–1570. https://doi.org/10.1002 Liu, G. R., and M. B. Liu. 2004. Smoothed particle hydrodynamics: A mesh-
/nag.688. free particle method. Singapore: World Scientific.
Bui, H. H., R. Fukagawa, K. Sako, and J. C. Wells. 2011. “Slope stability Lucy, L. B. 1977. “A numerical approach to testing the fission hypothesis.”
analysis and discontinuous slope failure simulation by elasto-plastic Astron. J. 82 (12): 1013–1024. https://doi.org/10.1086/112164.
smoothed particle hydrodynamics (SPH).” Geotechnique 61 (7): 565– Mao, Z., G. R. Liu, and X. Dong. 2017. “A comprehensive study on the pa-
574. https://doi.org/10.1680/geot.9.P.046. rameters setting in smoothed particle hydrodynamics (SPH) method
Chen, Z. Y., and C. M. Shao. 1988. “Evaluation of minimum factor of safety applied to hydrodynamics problems.” Comput. Geotech. 92 (Dec): 77–
in slope stability analysis.” Can. Geotech. J. 25 (4): 735–748. https://doi 95. https://doi.org/10.1016/j.compgeo.2017.07.024.
.org/10.1139/t88-084. McDougall, S., and O. Hungr. 2004. “A model for the analysis of rapid
Cheng, Y. M. 2003. “Location of critical failure surface and some further landslide motion across three-dimensional terrain.” Can. Geotech. J. 41
studies on slope stability analysis.” Comput. Geotech. 30 (3): 255–267. (6): 1084–1097. https://doi.org/10.1139/t04-052.
https://doi.org/10.1016/S0266-352X(03)00012-0. Morgenstern, N. R., and V. E. Price. 1965. “The analysis of the stability of
Dawson, E. M., W. H. Roth, and A. Drescher. 1999. “Slope stability analy- general slip surfaces.” Geotechnique 15 (1): 79–93. https://doi.org/10
sis by strength reduction.” Geotechnique 49 (6): 835–840. https://doi .1680/geot.1965.15.1.79.
.org/10.1680/geot.1999.49.6.835. Patra, C. R., and P. K. Basudhar. 2005. “Optimum design of nailed soil
Duncan, J. M. 1996. “State of the art: Limit equilibrium and finite-element slopes.” Geotech. Geol. Eng. 23 (3): 273–296. https://doi.org/10.1007
analysis of slopes.” J. Geotech. Eng. 122 (7): 77–596. /s10706-004-2146-7.
Duncan, J. M., and S. Wright. 2005. Soil strength and slope stability. Sabhahit, N., P. K. Basudhar, and M. R. Madhav. 1995. “A generalized pro-
Hoboken, NJ: John Wiley & Sons. cedure for the optimum design of nailed soil slopes.” Int. J. Numer.
Gao, W. 2016. “Determination of the noncircular critical slip surface in Anal. Methods Geomech. 19 (6): 437–452. https://doi.org/10.1002/nag
slope stability analysis by meeting ant colony optimization.” J. Comput. .1610190605.
Civ. Eng. 30 (2): 06015001. https://doi.org/10.1061/(ASCE)CP.1943 Spencer, E. 1967. “A method of analysis of the stability of embankments
-5487.0000475. assuming parallel inter-slice forces.” Geotechnique 17 (1): 11–26.
GEOSLOPE International. 2004. GeoStudio: User’s guide for SEEP/W and https://doi.org/10.1680/geot.1967.17.1.11.
SLOPE/W, version 6.0. Calgary, Canada: GEOSLOPE. Steward, T., N. Sivakugan, S. Shukla, and B. Das. 2011. “Taylor’s slope sta-
Gingold, R. A., and J. J. Monaghan. 1977. “Smoothed particle hydrodynam- bility charts revisited.” Int. J. Geomech. 11 (4): 348–352. https://doi.org
ics: Theory and application to non-spherical stars.” Mon. Not. R. Astron. /10.1061/(ASCE)GM.1943-5622.0000093.
Soc. 181 (3): 375–389. https://doi.org/10.1093/mnras/181.3.375. Sun, J., J. Li, and Q. Liu. 2008. “Search for critical slip surface in slope stabil-
He, X., D. Liang, and M. D. Bolton. 2017. “Run-out of cut-slope landslides: ity analysis by spline-based GA method.” J. Geotech. Geoenviron. Eng.
Mesh-free simulations.” Geotechnique 68 (1): 50–63. 134 (2): 252–256. https://doi.org/10.1061/(ASCE)1090-0241(2008)134:
Janbu, N. 1973. “Slope stability computations.” In Embankment–Dam engi- 2(252).
neering: Casagrande volume, edited by R. C. Hirschfeld and Steve J. USACE. 1970. Engineering and design: Slope stability of earth and rock
Poulos, 47–86. New York: John Wiley & Sons. fill dams. Washington, DC: USACE.
Jia, G. W., L. T. Zhan, Y. M. Chen, and D. G. Fredlund. 2009. Verlet, L. 1967. “Computer ‘experiments’ on classical fluids. I. Thermo dy-
“Performance of a large-scale slope model subjected to rising and lower- namical properties of Lennard-Jones molecules.” Phys. Rev. 159 (1): 98
ing water levels.” Eng. Geol. 106 (1–2): 92–103. https://doi.org/10.1016 –106. https://doi.org/10.1103/PhysRev.159.98.
/j.enggeo.2009.03.003. Wei, W. B., Y. M. Cheng, and L. Li. 2009. “Three-dimensional slope failure
Leshchinsky, B., and S. Ambauen. 2015. “Limit equilibrium and limit analysis by the strength reduction and limit equilibrium methods.”
analysis: Comparison of benchmark slope stability problems.” J. Comput. Geotech. 36 (1–2): 70–80. https://doi.org/10.1016/j.compgeo
Geotech. Geoenviron. Eng. 141 (10): 04015043. https://doi.org/10 .2008.03.003.
.1061/(ASCE)GT.1943-5606.0001347. Zheng, H., G. H. Sun, and D. F. Liu. 2009. “A practical procedure for
Li, L., Y. M. Cheng, and X. S. Chu. 2013a. “A new approach to the determi- searching critical slip surfaces of slopes based on the strength reduction
nation of the critical slip surfaces of slopes.” China Ocean Eng. 27 (1): technique.” Comput. Geotech. 36 (1–2): 1–5. https://doi.org/10.1016/j
51–64. https://doi.org/10.1007/s13344-013-0005-0. .compgeo.2008.06.002.

© ASCE 04019032-11 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 04019032

You might also like