Engineering Geology 237 (2018) 102–115

Contents lists available at ScienceDirect

Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo

Three-dimensional stability analysis of slope in unsaturated soils considering T

strength nonlinearity under water drawdown
⁎
Jing-Shu Xu , Xiao-Li Yang
School of Civil Engineering, Central South University, No. 68, South Shao-shan Street, Tian-xin District, Changsha City, Hu'nan Province 410075, People's Republic of
China

A R T I C L E I N F O A B S T R A C T

Keywords: In slope stability analyses, the soil has been frequently considered dry or saturated, whereas the soil is in fact
Slope stability unsaturated in many cases. Owing to the existence of matric suction, the strength properties of unsaturated soils
Unsaturated soils are greatly different from dry or saturated soil, and thereby leading to diverse stability conditions of the slope.
Matric suction Based on the kinematical approach of limit analysis, in this study, a three-dimensional (3D) stability analysis was
Strength nonlinearity
conducted for slopes in unsaturated soils. The effects of matric suction distribution patterns, the nonlinearity of
Water level drawdown
the strength under soil-water characteristic curves and the water drawdown were investigated to explore the
shear strength and stability of a slope in unsaturated soil. It was found that, when the matric suction is uniformly
distributed, the stability factors of the slope increase linearly along with the matric suction. In addition, when
the matric suction increases linearly with the depth, the enhancing effect of the matric suction on the slope
stability intensifies as the variation magnitude of the matric suction increases, especially for a gentle slope with a
narrower width. For nonlinear patterns, different soil-water characteristic curve (SWCC) models will lead to
diverse shear strengths and stability conditions of the slope. The water drawdown has significant effects on the
functions of the matric suction as well as slope stability.

1. Introduction
A slope stability analysis, which is an important and classical issue
in geotechnical engineering, has been mostly performed under two-
dimensional (2D) plane strain conditions under the assumption that the
soils is dry or saturated. However, soil is in fact generally unsaturated in
nature (Hoyos et al., 2015; Oh and Lu, 2015). In addition, a slope
stability assessment is typically a three-dimensional (3D) problem, and
a number of studies have shown that 2D failure modes will lead to
conservative estimations of the slope stability (Michalowski and
Drescher, 2009). Owing to the existence of matric suction, the strength
behavior and properties of unsaturated soils differ greatly from dry or
saturated soil, leading to diverse slope stability conditions. Thus, in
many cases, it is inappropriate to ignore the unsaturation of soils and
simply treat it as dry or saturated. The pore water pressures and water
drawdown are decisive factors on the slope stability, and when the
matric suction, in particular is taken into account, the water level
drawdown directly influences the slope stability, and more importantly,
changes the acting zone and magnitude of the matric suction inside the
slope. It is of both theoretical and practical significance to investigate
the stability of 3D slopes subjected to a water drawdown in unsaturated
soils.
Many methods can be used to investigate the 3D stability of a slope,
introducing the rigorous limit equilibrium method (RLE), the discrete
element method (DEM), the general particle dynamics numerical
method, the discontinuous deformation analysis (DDA) method and the
limit analysis method.
With regard to the rigorous limit equilibrium method for a 3D slope
stability analysis, which takes the inter-column forces into account, the
RLE method was established by Zhou and Cheng (2013) based on the
three-axial-direction force equilibrium conditions and three direction
moment equilibrium conditions around the three coordinate axes. The
relationship between the width of a sliding body and the safety factor as
well as the value of the safety factor itself, were determined. The pre-
sented method was proved to be valid through a practical application.
Zhou and Cheng (2014) investigated the stability of a 3D slope sub-
jected to seismic forces when applying a combination of the RLE and
the pseudo-dynamic method. A new limit equilibrium method to ana-
lyze the stability and displacement of three-dimensional creeping slopes
was proposed by Zhou and Cheng (2015).
For the discrete element method, Assefa et al. (2017) analyzed a
deep-seated slope movement in a complex rock formation. The DEM

⁎
Corresponding author.
E-mail addresses: csushdxjs@csu.edu.cn (J.-S. Xu), yangxl@csu.edu.cn (X.-L. Yang).

https://doi.org/10.1016/j.enggeo.2018.02.010
Received 8 August 2017; Received in revised form 5 January 2018; Accepted 10 February 2018
0013-7952/ © 2018 Elsevier B.V. All rights reserved.
J.-S. Xu, X.-L. Yang
approach was adopted to simulate the slope as a complex blocky
structure defined by bedding planes, inclined and sub-vertical joints.
Different hypotheses regarding the geometry of the slip surface, com-
patible with field evidences, are discussed. Camones et al. (2013) in-
vestigated the step-path failure mechanism in fractured rock masses
using the discrete element method (DEM). Crack propagation and
coalescence were simulated in biaxial and triaxial laboratory tests, and
simple examples of this mechanism were modeled for a better under-
standing of the failure mechanism. Salmi et al. (2016) analyzed the
impacts of underground mining on the stability of a slope using nu-
merical modeling based on a coalmining induced landslide at Nattai
North, Australia, as a typical case study. Discontinuous numerical
modeling was employed to analyze the mechanisms of the ground
movements leading to landslide. The role of geological and geo-
technical factors on the slope stability were discussed.
Regarding the general particle dynamics numerical method, Zhou
et al. (2015) proposed the general particle dynamics approach (GPD3D)
to simulate the failure process (initiation, growth, and coalescence of
cracks macro-failures) of a slope.
The discontinuous deformation analysis method (DDA) was pre-
sented by Jing (1998) to investigate the deformations of geometrically
complex blocks in rock mechanics and engineering geology. The ad-
vantages and shortcomings of DDA were illuminated, and the solutions
by DDA were compared with the explicit distinct element method and
the finite element method.
The limit analysis method, the approach adopted in this study is
effective and has been widely used to investigate the stability problems
of foundations, tunnels (Li and Yang, 2018; Yang and Zhang, 2018),
slopes (Yang and Li, 2018) and other geotechnical structures. Slope
analyzed using the limit analysis method can be divided into two parts
based on the standard of dimensionality, i.e., (1) a 2D stability analysis
and (2) a stability analysis under 3D conditions. In terms of a 2D sta-
bility analysis, considering the seismic forces, Qin and Chian (2017)
analyzed the stability of a two-stage slope in layered soil, proposed a
closed-form solution to the kinematic stability of a slope, and in-
vestigated the influences of the soil weight and cohesion, soil non-
homogeneity, and external factors such as seismic forces on the slope
stability. Through a numerical analysis approach, Gischig et al. (2015)
conducted a rock slope stability analysis to demonstrate the roles of
amplification in enhancing the co-seismic slope deformation.
For a slope stability analysis under 3D conditions, Michalowski and
Drescher (2009) proposed a 3D rotational failure mechanism of a slope
with a spiral conic shape in accordance with the normality rule, and
derived the expression of critical height of a 3D slope based on the limit
analysis method, and investigated the influences of the 3D geometry
characteristics and soil strength on the slope stability. Lim et al. (2016)
presented sets of stability charts of a slope in purely cohesive clay using
the finite element limit analysis method.
With regard to investigations on unsaturated soil, Fredlund et al.
(1978) proposed an extended expression of the Mohr-Coulomb failure
criterion to take into account the effects of matric suction on the shear
strength of various types of soil. Thereafter, Bao et al. (1998), Fredlund
et al. (1996), Khalili and Khabbaz (1998), Vanapalli et al. (1996), and
Vilar (2006) also proposed expressions to estimate functions of the
matric suction on the shear strength of unsaturated soils. Zhang et al.
(2014) performed a stability analysis of 2D slopes in unsaturated soil
considering the strength of the nonlinearity aiming at diverse soil types
based on the soil-water characteristic curve (SWCC, i.e., permeability
function), investigated the effects of soils strength nonlinearity on the
slope stability, and proposed relevant recommendations for practical
use. Houston et al. (2008) performed a series of unsaturated soil triaxial
tests to investigate the strength properties and shear-induced volume
change behavior of unsaturated soil, as well as the feasibility of un-
saturated soil property estimation functions. Based on the moment
equilibrium method, an investigation into the critical embedment depth
for a rigid retaining wall in unsaturated soils was conducted by Zhang
103
Engineering Geology 237 (2018) 102–115
et al. (2016a,b), and the effects of the matric suction distribution pat-
terns on the objectives were illustrated. Tarantino and Mountassir
(2013) investigated the stability of geotechnical structures in un-
saturated soils, adopting an accessible approach for a hydraulic and
strength behavior estimation of unsaturated soils.
It should be noted that investigations into the slope stability of a
slope, regardless of under 2D or 3D conditions, have been mainly
conducted in dry or saturated soil, even though soil is commonly un-
saturated in nature (Fredlund et al., 1978; Lu and Likos, 2004), and
only a few investigation into the stability of 3D unsaturated slopes have
been applied. Matric suction, which exists in unsaturated soils, makes
the physical properties and behavior of unsaturated soils greatly dif-
ferent from those of dry or saturated soils, resulting in different stability
conditions of various slopes. As a consequence, the question, which has
both theoretical and practical importance, is how to estimate the 3D
stability of a slope in various unsaturated soil under different matric
suction distributions and pore water levels. In this regard, based on the
kinematical approach of limit analysis theorem, a 3D stability analysis
of a slope subjected to water level drawdown in unsaturated soils was
conducted in this study. External work rates by soil weight, pore water
pressure and matric suction as well as the internal energy dissipation
were calculated, and expressions of the critical heights of a slope were
then derived using an energy equilibrium equation. The effects of ma-
tric suction distributions, soil shear strength nonlinearity under diverse
SWCC models and water level drawdown on the slope stability were
investigated after optimization, and some numerical solutions were
presented.
2. Upper bound theorem
The limit analysis method assumes that the soil deforms plastically
according to the normality rule. The present method also states that the
internal energy dissipations are no less than the work rates by external
forces (Pan and Dias, 2017; Yang and Yao, 2018), namely
∫V σ ∗ij ε ∗̇ ij dV ≥ ∫S Ti vi dS + ∫V Xi vi∗dV (1)
where εij̇∗ is the strain rate; σij∗ is the stress rate tensor, S and V are the
boundary and volume of a failure block, respectively; Ti is the surcharge
load on boundary; Xi is the body force in volume V; and vi is the velocity
along the potential sliding surface.
3. Shear strength of unsaturated soils
Equations for the estimation of unsaturated soils shear strength can
be classified into two categories, i.e., linear equations (Fredlund et al.,
1978; Zhang et al., 2016a,b) and nonlinear equations (Bao et al., 1998;
Fredlund et al., 1996; Vanapalli et al., 1996; Khalili and Khabbaz, 1998;
Vilar, 2006). In this study, both linear and nonlinear equations were
adopted to demonstrate the effects of matric suction on the stability of a
3D unsaturated slope.
3.1. Linear conditions
For the linear conditions, the uniformly distributed condition of the
matric suction and its linearly increased conditions, along with depth,
were adopted in this study, as shown in Fig. 1.
3.1.1. Unified shear strength
The equation for the shear strength estimation of unsaturated soils
proposed by Fredlund et al. (1978) in the form of the Mohr-Coulomb
failure criterion is the most representative function and was adopted in
this study to describe the shear strength of unsaturated soils under a
linear condition in which the matric suction is uniformly distributed, as
shown in Fig. 1(a)
J.-S. Xu, X.-L. Yang Engineering Geology 237 (2018) 102–115

O O Table 1
ua-uw ua-uw Soil cohesion c(ua−uw) due to matric suction summarized by Zhang et al. (2014).

Unsaturated shear strength Cohesion due to matric suction

equation

ρ Fredlund et al. (1978) c(ua−uw) = (ua − uw) tan φb

Fredlund et al. (1996) c(ua−uw) = (ua − uw)Θdκ tan φ′
1 where Θd = θw/θs, κ is a fitting parameter.
Vanapalli et al. (1996)
c(ua − uw ) = (ua − uw ) ⎡tan φ′
⎣
( θ w − θr
θs − θr ) ⎤⎦
or

Depth, z Depth, z
c(ua − uw ) = (ua − uw ) ⎡tan φ′
⎣
( S − Sr
100 − Sr ) ⎤⎦
Vilar (2006) c(ua − uw ) =
(ua − uw )
(a) (b) a + b (ua − uw )
1 1
where a = ,b=
Fig. 1. Distributions of matric suction in unsaturated soils: (a) uniform distribution of tan φ′ (cult − c′)
matric suction with depth; and (b) linearly increased distribution of matric suction with 1 a
or b = −
(cm − c′) ψm
depth.
Khalili and Khabbaz (1998) c(ua−uw) = (ua − uw)[λ′] tan φ′
where [λ′]=1.0 when (ua − uw) ≤ AEV
τ = c′ + (σ − ua) tan φ′ + (ua − u w ) tan φb (2) [λ′] = ( ua − uw −0.55
) when (ua − uw) > AEV
AEV
′
where c is the effective cohesion intercept, φ is the effective internal ′ Bao et al. (1998) c(ua−uw) = (ua − uw)[ζ] tan φ′
where [ζ] = 1.0 when (ua − uw) ≤ AEV
friction angle, ua − uw is the matric suction among which ua and uw are log (ua − uw )r − log (ua − uw )
[ζ ] =
the pore-air pressure and pore-water pressure, respectively; σ − ua is log (ua − uw )r − log (AEV )

the net normal stress on the failure plane, and φb is the suction angle,
which is commonly 15° based on a number of datasets available from
the 1970s.
The shear strength of unsaturated soils shown in Eq. (2) can be
transformed into the Mohr-Coulomb failure criterion formation as
τ = c + (σ − ua) tan φ′
′
where c = c + (ua − uw) tan φ . b
3.1.2. Linearly increased matric suction with depth
According to Zhang et al. (2016a,b), another linear estimation used
to describe the values of the matric suction was developed from
Fredlund et al. (1978), as shown in Fig. 3(b), which states that the
matric suction is linear increased with depth from the water table. This
expression was also adopted in the present study to demonstrate the
effects of the matric suction on the slope stability in a more compre-
hensive manner. The matric suction (ua − uw)z at depth z with regard to
the linearly increased distribution is
(ua − u w )z = ρz
where ρ is the magnitude of linearly distributed matric suction and z is
the vertical distance between the water table level and the calculation
point.
3.2. Shear strength nonlinearity of unsaturated soils
The shear strength of unsaturated soils was initially considered to
have a linear function with the matric suction, as presented by Fredlund
et al. (1978). However, it was later found that the relationships
when AEV < (ua − uw) < ψr
[ζ] = 0, when (ua − uw) ≥ ψr
between shear strengths and matric suctions are nonlinear, as de-
termined through by subsequent tests on unsaturated soils. Fig. 2 shows
(3) the relationships between the shear strength and matric suction for a
typical unsaturated soil and the SWCC, which was drawn based on a
significant amount of experiment data. As shown in Fig. 2(a) and from
Zhang et al. (2014), soil shows saturated behaviors when the matric
suction is less than the air-entry value (AEV). Afterward, the shear
strengths begins to increase once the matric suction exceeds the AEV
within the nonlinear zone, and in the residual zone, the envelopes of
plastic clay and soil then become generally horizontal, whereas the
envelope of a cohesion-less material (sands) shows a leveling off along
with the matric suction, indicating that the shear strength of the sand
tends to decrease along with the matric suction.
With regard to the nonlinearity of the shear strength of unsaturated
soil, a number of equations have been proposed to demonstrate the
(4) nonlinear relationships between shear strength and the matric suction
(Bao et al., 1998; Khalili and Khabbaz, 1998; Fredlund et al., 1978,
1996; Vanapalli et al., 1996 and Vilar, 2006), as shown in Table 1.
Detailed explanations on these equations can also be found in Zhang
et al. (2014).
Where θw and θs in Table 1 are the volumetric water content and
saturated volumetric water content, respectively; θr is the residual vo-
lumetric water content; S is the degree of saturation; Sr is the residual
degree of saturation; cult is the ultimate un-drained shear strength of an
air-dried soil sample; AEV is the air-entry value of the soil and ψr is the
residual suction.

Linear zone Nonlinear zone Residual zone AEV ψr Residual suction

100
Degree of saturation/%
Shear strength

Boundary
φb=φ' 80 effect zone
Plastic clays

Soils 60 a f (Inflection point)

Cohesionless 40 Transition
soils (Sands) effect zone
φb=φ'
20 Residual zone

AEV Residual suction 01 10 100 1000 10000 100000 1000000

Matric suction (ua-uw)/kPa Matric suction/kPa (Logarithm scale)

Fig. 2. Redrawing of the relationship between unsaturated soils and SWCC: (a) strength envelope of typical unsaturated soil and (b) SWCC for typical soil.

104
J.-S. Xu, X.-L. Yang Engineering Geology 237 (2018) 102–115

Table 2 r = r0 e (θ − θ0) tan φ ′ (5)

SWCC parameters by Fredlund and Xing (1994).
r ′ = r0′ e−(θ − θ0) tan φ ′ (6)
SWCC θs af(kPa) nf mr ψr AEV θr κ cult(kPa)
The distance from point O to cone axis rm and the radius of the circle
1 0.4 1 2 1 10 0.5 0.081 1.0 11.4
of a cone cross-section R, are
2 0.4 10 2 1 100 5 0.080 1.8 13.7
3 0.4 100 2 1 1000 50 0.078 2.2 28.4 r + r′
4 0.4 1000 2 1 10,000 500 0.073 2.5 107.3 rm = = r0 f1
2 (7)

r − r′
An investigation by Zhang et al. (2014) found that different soil R= = r0 f2
2 (8)
types show various strength envelopes under nonlinear estimations
equations. In this regard, in accordance with Zhang et al. (2014), four where f1 and f2 are as given in the Appendix A.
types of soil were selected for this study to investigate the nonlinear
functions of the matric suction on the stability of the slopes in un- 5. Calculations and optimization
saturated soil, as listed in Table 2, where No. 1 indicates sandy soil, No.
2 denotes fine-grained soil (silt), No. 3 is clay and No. 4 is extremely 5.1. Internal energy dissipation
fine-grained soil.
In Table 2, af is the inflection point of the shear strength envelope, The internal energy dissipations D0−B−3D of sections θ0 − θB and
as shown in Fig. 2(b), which equals 1, 10, 100, and 1000 for SWCC Nos. DB−h−3D of section θB − θh of the rotational body can be written as
1 through 4; nf is the slope of the SWCC at the inflection point; mf is a follows according to the limit analysis method:
fitting parameter related to residual water content, and κ is a fitting θB α1∗
parameter equals to 1.0, 1.8, 2.2, and 2.5, respectively.
D0 − B − 3D = 2ω ∫θ ∫0
0
cR (rm + R cos α )2dαdθ

= ωc′r03 g1 + ωρ tan φbr04 (g2 − g3) (9)

4. 3D failure mechanism of slope in unsaturated soils θh α2∗

DB − h − 3D = 2ω ∫θ ∫0
B
cR (rm + R cos α )2dαdθ

The 3D rotational failure mechanism of a slope in unsaturated soils = ωc′r03 g4 + ωρ tan φbr04 (g5 − g6) (10)
is illustrated in Fig. 3. The present failure mechanism of the slope has
an angle of β with a slope height of H. Log-spiral curves AC and A'C', are where ω is the angular velocity, α1∗ = arccos
(d1/R), arccos (d2/α2∗ =
the boundaries of the conic rotating around center O, among which AC R), d1 = r0f3, and d2 = r0f4. In addition, g1 through g6 are given in the
is the failure surface of the slope. The toe point C of the slope is as- Appendix A.
sumed to be located at the water level. The shape of the failure me- The sum of the internal energy dissipation of rotational body is
chanism is only related to three independent variables, i.e., θ0 and θh, D3D = D0 − B − 3D + DB − h − 3D (11)
which are the origination and termination rotating angles, respectively,
and the independent ratio, r'0/r0. Investigations show that the stability The internal energy dissipation of the insertion block is
factors are also related to the slope width, and as a result, an insertion θh r
block of width b is inserted into the conic, as shown in Fig. 3(b). When
Dinsert = b ∫θ 0
cv cos φ
cosφ
dθ = ωc′r03 g7 + ωρ tan φbr04 g8
(12)
b → ∞, the 3D failure mechanism degenerates to a 2D plane strain one
The total internal energy dissipation rate Dint can be expressed as
and consequently, the 3D solutions become very close to the 2D results.
It should be noted that in the failure mechanism shown in Fig. 3, the Dint = D3D + Dinsert = ωc′r03 (g1 + g4 + g7)
water table is assumed to locate at the slope toe. The assumption is
+ ωρ tan φbr04 (g2 + g3 + g5 + g6 + g8) (13)
adopted in this section for that the pore water level presented in
Fig. 3(a) is classical and representative and had also been adopted by where g1 to g8 are given in the Appendix A.
many other investigators (Eid 2014; Gao et al., 2014; Michalowski and
Nadukuru, 2013; Viratjandr and Michalowski, 2006). With regard to 5.2. Work rate by soil weight
these reasons, it is assumed in this section that the water table is located
at the slope toe. Besides, a stability analysis of a 3D slope with different Velocity v of a point in the rotational failure block shown in Fig. 2(a)
pore-water tables is also conducted in this study. is
As shown in Fig. 2(a), OA = r0 and O'A' = r'0. The expressions of the
discontinuity curves AC and A'C', are v = ω (rm + y ) (14)

O
(a) (b)
θ0
θB

2φ
θ

A'
θh
C' b B
A
x

B
ρ
R
α1 *

rh
v H
y
d1

x
R

Insertion
*
β

α2

C d2
block
y

Fig. 3. 3D failure mechanism of slope in unsaturated soils: (a) rotational failure mechanism, and (b) failure mechanism with plant insert.

105
J.-S. Xu, X.-L. Yang Engineering Geology 237 (2018) 102–115

Table 3 c′ (g1 + g4 + g7 )
Slope stability factors of a 3D unsaturated slope (γ = 20 kN/m3, β = 60°, B/H = 5.0,
H= [e (θh − θ0) tan φ ′ sin θh
γ (g9 + g10) − ρ tan φb (g2 + g5 + g8 − g3 − g6 )
c' = 10 kPa and φ′ = 34°).
− sin θ0] (22)
Soil types SWCC models Matric suction/kPa
For the conditions that the matric suction is nonlinear distributed
0 20 40 80 100 120 following the equations in Table 1,
No. 1 1 21.08 25.37 27.71 31.59 33.31 34.93 c (g1 + g4 + g7 )
2 21.08 21.08 21.08 21.08 21.08 21.08
H= [e (θh − θ0) tan φ ′ sin θh − sin θ0]
γ (g9 + g10 ) (23)
3 21.08 23.76 23.89 23.96 23.97 23.98
′
4 21.08 24.82 26.19 28.06 28.80 29.46 where c = c + c(ua−uw).
5 21.08 21.08 21.08 21.08 21.08 21.08
To make the failure mechanism geometrically meaningful, neces-
No. 2 1 21.08 29.68 28.77 28.71 28.91 29.13
2 21.08 32.27 30.25 24.36 21.08 21.08 sary constraints are derived from the geometry relationships in Fig. 3,
3 21.08 27.20 27.94 28.38 28.48 28.54 as listed in Eq. (24). Based on this, the minimum upper bound solutions
4 21.08 34.35 39.21 45.84 48.46 50.80 of the stability factor Ns = γHc/c′ can be obtained through optimization.
5 21.08 36.36 38.48 29.56 21.08 21.08
No. 3 1 21.08 48.46 70.79 93.87 96.81 96.56
⎧ 0 < θ0 < θB < θh < π
2 21.08 48.92 73.76 108.88 119.29 126.05 0 < r0′/ r0 < 1
3 21.08 37.49 44.15 50.01 51.56 52.69 ⎨
4 21.08 49.52 77.96 108.93 118.21 126.50 ⎩ 0 < (b + Bmax′ )/ H < B / H (24)
5 21.08 49.52 77.96 117.0 130.38 141.86
No. 4 1 21.08 49.48 77.76 133.69 161.23 188.41
2 21.08 49.50 77.86 134.28 162.28 190.09 6. Numerical solutions and parameter analysis
3 21.08 46.06 65.62 94.26 105.07 114.24
4 21.08 49.52 77.96 134.85 163.29 191.73
In this section, some numerical solutions of safety factors corre-
5 21.08 49.52 77.96 134.85 163.29 191.73
sponding to different soil types and SWCC models were presented.
Thereafter, a parameter analysis carried out to investigate the effects of
The infinitesimal work rate element is the 3D slope geometry characteristics, the matric suction distributions
and the shear strength nonlinearity on the 3D slope stability in un-
dw = γv cos θdV = γv cos θdxdy (rm + y ) dθ (15) saturated soil is described.
The work rate by the soil weight of the rotational body shown in
6.1. Numerical solutions
Fig. 3(a) is then obtained as follows:

θB x1∗ y∗ Table 3 shows the stability factors Ns of a 3D slope with respect to

Wγ − 3D = 2ωγ ⎡ ∫θ ∫0 ∫d (rm + y )2 cos θdxdydθ γ = 20 kN/m3, B/H = 5.0, c′ = 10kPa, and φ′ = 34° for different SWCC
⎣ 0 1

θh x2∗ y∗
models. In the row listing SWCC models, 1 indicates the Fredlund et al.'s
+ ∫θ ∫0 ∫d (rm + y )2 cos θdxdydθ⎤
(16)
(1978) model, 2 is Vanapalli et al.'s (1996) model, 3 is Vilar's (2006)
B 2 ⎦ model, 4 is Khalili and Khabbaz's (1998) model, and 5 is Bao et al.'s
where y∗ = R2 − x 2 and x1∗ = R2 − di2 (i = 1, 2), according to the (1998) model, as listed in Table 1. The stability factors can be easily
geometric relationships of the failure mechanism. obtained for practical use.
Angle θB can be obtained based on the trigonometric relations,
6.2. Comparison
sin θ0
θB = arctan
cos θ0 − ζ (17) A numerical simulation is conducted in this section to investigate
the validity of the solutions in this study, as follows: According to the
sin (θh − θ0) e (θh − θ0 ) tanφ ′ sin θh − sin θ0 definition of the stability factor Ns, the safety factor of the slope under
ζ= − sin(θh + β ) the corresponding stability factor should be equal to 1.0. In this regard,
sin θh sin θh sinβ (18)
the finite difference software FLAC3D is used to calculate the safety
Eq. (16) can be briefly written as factor. Taking the 3D slope with the stability factor of 21.08 in the first
row and the first column in Table 3 for example. With regard to the
Wγ − 3D = γωr04 g9 (19) slope, the unit weight γ = 20 kN/m3, the slope angle β = 60°, the ratio
B/H = 5.0 and the internal friction angle φ′ = 34°. The height of the
For the insertion block, the work rate of the soil weight can be c′N 10kPa × 21.08
slope can be obtained as Hc = γ s = 20 kN / m3 = 10.54 m and the co-
briefly written as
hesions of the slopes corresponding to Table 3 can be calculated from
Wγ − insert = γωr04 g10 (20) the equations in Table 1, as listed in Table 4.
Thereafter the numerical model of the 3D slope corresponding to
The total work rate by soil weight can be expressed as the stability factor in the first row and the first column in Table 3 is
established and the safety factor of the slope is then calculated in virtue
Wγ = Wγ − 3D + Wγ − insert = γωr04 (g9 + g10) (21) of FLAC3D, as shown in Fig. 4.
It is found from the Fig. 4 that the slope has the safety factor of
where g9 and g10 are as given in the Appendix A.
1.092, which is a slightly > 1.0. This is reasonable for that the stability
factors in Table 3 are the upper bound solutions. In such a way, the
5.3. Slope height and optimization solutions in this study can be proved valid.

Equating the external work rates to internal energy dissipation, the 6.3. Parameter analysis under uniform suction
height of a 3D slope in unsaturated soils can be written as follows.
For the condition that the matric suction is linearly distributed The effects of uniformly distributed matric suction on 3D slope
along with the depth, stability are illustrated herein. Fig. 5(a) and (b) show the stability

106
J.-S. Xu, X.-L. Yang Engineering Geology 237 (2018) 102–115

Table 4 6.4. Parameter analysis under linear suction with depth

Cohesions corresponding to the stability factors in Table 3.
A parameter analysis is described in this part to analyze the effects
SWCC Cohesion corresponding to the stability factors in Table 3/kPa
models of the linearly increased matric suction on the slope stability. Fig. 6(a)
0 20 40 80 100 120 and (b) are the stability factors along with B/H under a variation in
magnitude ρ for β = 60° and 75°, respectively. Fig. 6(c) and (d) illus-
No. 1 1 10.0 12.03 13.15 14.98 15.80 16.57
trate the variation rules of the stability factors along with ρ under dif-
2 10.0 10.0 10.0 10.0 10.0 10.0
3 10.0 11.27 11.33 11.37 11.37 11.38 ferent ratios B/H and slope angles β, respectively.
4 10.0 11.77 12.42 13.31 13.66 13.97 It is shown in Fig. 6(a) and (b) that, as expected, the stability factors
5 10.0 10.0 10.0 10.0 10.0 10.0 increase with the variation in magnitude ρ of the matric suction. In
No. 2 1 10.0 14.08 13.65 13.57 13.71 13.82 addition, an interesting phenomenon takes place in that the point of the
2 10.0 15.30 14.35 13.01 10.0 10.0
smallest curvature radius of Ns along with the ratio B/H moves forward
3 10.0 12.90 13.25 13.39 13.51 13.54
4 10.0 16.29 18.60 20.32 22.98 24.10 from B/H = 1.1 for ρ = 0 to B/H = 1.8 for ρ = 10.0, meaning that
5 10.0 17.25 18.25 16.90 10.0 10.0 magnitude ρ will influence the 3D effect of an unsaturated slope in
No. 3 1 10.0 22.98 33.58 40.69 45.92 45.80 addition to the stability of a 3D slope. Fig. 6(c) and (d) indicate that, the
2 10.0 23.20 35.0 44.53 56.58 59.79
changes in rules of Ns along with magnitude ρ are strongly influenced
3 10.0 17.78 20.94 22.65 24.46 24.99
4 10.0 23.49 36.98 46.61 56.07 60.0 by the slope angle and ratio B/H. The increasing trends in the stability
5 10.0 23.49 36.98 48.0 61.84 67.29 factors along with ρ become increasingly gentle, and even horizontal,
No. 4 1 10.0 23.47 36.88 50.21 76.48 89.37 with regard to the condition of B/H = 5.0 and β = 90°, indicating that
2 10.0 23.48 36.93 50.34 76.97 90.17 the influence of the linear increased pattern of matric suction on the
3 10.0 21.85 31.12 38.58 49.84 54.19
slope stability is much more significant on gentle slopes (β ≤ 45°) with
4 10.0 23.49 36.98 50.47 77.45 90.94
5 10.0 23.49 36.98 50.47 77.45 90.94 a narrower width, compared with steep 2D slopes.

6.5. Parameter analysis under shear strength nonlinearity

factors along with the ratio B/H under different matric suction values of
0, 25 and 50 kPa for β = 60° and β = 75°, respectively. Fig. 5(c) and (d)
demonstrate the stability factors versus the matric suction under dif-
ferent ratios B/H and slope angles β, respectively.
Fig. 5(a) and (b) show that the stability factors decrease sharply
within B/H ≤ 1.5, the variation rules of which then become gentle and
eventually barely change for B/H ≥ 10.0. It is clear that the matric
suction (ua − uw) has a positive effect on the slope stability. For the
conditions of β = 60°, φ = 15°, φb = 15°, and B/H = 3.0, the stability
factor increases by 44.67% and 89.34% as the matric suction increases
from 0 to 25 and 50 kPa, respectively. Fig. 5(a) and (b) also show that
the existence of the matric suction has little influence on the effects of
the ratio B/H on the slope stability, and that the varying patterns of
stability factors under (ua − uw) = 25 kPa and (ua − uw) = 50 kPa in
unsaturated soils and the rules by Michalowski and Drescher (2009) are
the same. From Fig. 5(c) and (d), it is clear that the stability factors
increase linearly along with the matric suction, and that the variations
in slope angle β results in little change in the impact of the matric
suction on the stability factors of a slope. It can also be seen from
Fig. 4(c) that, under the condition of B/H = 10.0, solutions under a 3D
failure mechanism are in good agreement with the 2D solutions,
meaning that 3D solutions will degenerate into 2D solutions once the
ratio B/H exceeds 10.0.
The effects of shear strength nonlinearity on slope stability are de-
monstrated in Fig. 7, with regard to the soil types in Table 2 and the
nonlinear strength estimation functions in Table 1. Fig. 7(a) through (d)
are illustrations of stability factors of a slope as a function of the matric
suction in (a) sandy soil, (b) fine-grained soil, (c) clay, and (d) ex-
tremely fine-grained soils, respectively. In Fig. 7, at β = 60° and B/
H = 5.0, effective cohesion and internal frictional angle are selected as
c' = 10 kPa and φ' = 34°.
From Fig. 7, it is clear that the soil types and shear strength esti-
mation equations can both influence the values of the stability factors of
the slopes in unsaturated soil. More specifically, for sandy soil, as
shown in Fig. 7(a), the stability factors are lower than those under the
commonly adopted φb = 15°. The curves of the stability factors ob-
tained by Vanapalli et al. (1996) and Bao et al. (1998) are horizontal
and overlapped, meaning that sandy soil shows saturated behaviors
under these two models, but nonlinear functions under the other three
models within (ua − uw) ≤ 40 kPa, and then degenerates into a linear
form. The differences in the stability factors obtained under the models
are limited, and the effects of the matric suction on the stability analysis
of soil with a small AEV (AEV = 0.5 kPa) can be ignored, which is in
accordance with the results by Zhang et al. (2014).
As for the fine-grained soil, the curves under Bao et al.'s (1998)
model and Vanapalli et al.'s (1996) model show similar trends in that

Fig. 4. Geometrical parameters and the safety factor of the 3D slope.

107
J.-S. Xu, X.-L. Yang Engineering Geology 237 (2018) 102–115

50 40
Present solution
(a) Present solution
Solution by Michalowski and Drescher , 2009 (b)
40 Solution by Michalowski and Drescher, 2009
30
β=60°, φ'=15°, φb=15° β=75°, φ'=15°, φb=15°
ua-uw=0 ua-uw=0
30
25kPa 25kPa
Ns

Ns
50kPa 20
50kPa
20

10
10

0 0
0 2 4 6 8 10 0 2 4 6 8 10
B/H B/H
80 80

(c) β=60°, φ'=15°, φb=15° (d) B/H=5.0, φ'=15°, φb=15°

60 60
Ns

Ns
40 40
1.5 3.0 10.0
B/H=0.6 β=30°
1.0
20 20 45°
60°
75°
90°
2D solutions
0 0
0 20 40 60 80 100 0 20 40 60 80 100
ua-uw/(kPa) ua-uw/(kPa)

Fig. 5. Effects of uniformly distributed matric suction on 3D slope stability.

60 60
Present solution (a) Present solution (b)
Solution by Michalowski and Drescher , 2009 Solution by Michalowski and Drescher , 2009

40 β=60°, φ'=15°, φb=15° 40 β=75°, φ'=15°, φb=15°

Ns

Ns

ρ= 0 ρ= 0
4.0 6.0
2.0 6.0
2.0 4.0
20 20
10.0
10.0
-1
ρ/(kPa·m ) ρ/(kPa·m-1)
0 0
0 2 4 6 8 10 0 2 4 6 8 10
B/H B/H
60 100

(c) β=75°, φ'=15°, φb=15°

(d) B/H=5.0, φ'=15°, φb=15°
75
40
Ns
Ns

50
1.5 3.0 10.0
B/H=0.6
20 75° 90°
25 β=30°
1.0
45°
60°

0 0
0 2 4 6 8 10 0 2 4 6 8 10
ρ/(kPa·m-1) ρ/(kPa·m-1)

Fig. 6. Slope stability based on change in magnitude of matric suction in unsaturated soil.

108
J.-S. Xu, X.-L. Yang Engineering Geology 237 (2018) 102–115

50 80
(a) Sandy soils (b) A fine-grained soils (silt)
b
φ =φ'=34°
φb=φ'=34° φb=15°
40 b
60
Ns φ =15°

Ns
30 40

20 20
af=1, nf=2, mf=1, ψr=10kPa, κ=1.0, cult=11.4kPa
af=10, nf=2, mf=1, ψr=100kPa, κ=1.8, cult=13.7 kPa
10 0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
(ua-uw)/kPa (ua-uw)/kPa
140 160
(c) Clays (d) Extremely fine-grained soils
120
φb=φ'=34° 120
100
φb=φ'=34°
80
φb=15°
Ns

Ns
80
φb=15°
60

40
40
af=1000, nf=2, mf=1, ψr=10000kPa,
20
af=100, nf=2, mf=1, ψr=1000kPa, κ=2.2, cult=28.4 kPa κ=2.5, cult=107.3kPa
0 0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
(ua-uw)/kPa (ua-uw)/kPa
Fredlund et al., (1996) Vanapalli et al., (1996)
Vilar, (2006) Khalili and Khabbaz, (1998)
Bao et al., (1998) φb=15°
φb=φ'=34°

Fig. 7. Soil strength nonlinearity on slope stability in unsaturated soils.

the stability factors increase within (ua − uw) ≤ 40 kPa, and then de- 7.1. Pore-water pressure in limit analysis
crease nonlinearly for 40 kPa ≤ (ua − uw) ≤ 100 kPa, finally reaching a
constant value. The stability factors obtained under Fredlund et al.'s In virtue of the limit analysis method, the presence of water inside a
(1996) model and Vilar's (2006) model are very close to each other, slope requires two additional terms of work rates, i.e., Wu′, the work
where those under Khalili and Khabbaz's (1998) model are the largest. rate by pore-water pressure on the volumetric deformation of the soil
The nonlinearity of the stability factors along with the matric suction and Wu′′, the work rate by the water pressures on the boundaries of the
becomes distinct and non-negligible for fine-grained soil. slope (Gao et al., 2014; Michalowski and Nadukuru, 2013; Viratjandr
With regard to clays, the curves of the stability factors approach the and Michalowski, 2006), as follows
line φb = φ′ = 34° within the range (ua − uw) ≤ 60 kPa, after which the
increasing trend decreases. For the special Vilar's (2006) model, the Wu = Wu′ + Wu′ ′ = − (∫ uε ̇ dV + ∫ un v dS )
V
∗
ii u
S
i i u
(25)
stability factors increase slowly when (ua − uw) ≤ 40 kPa, and then
become constantly gradual along with the matric suction when where u = ruγh is the pore-water pressure, ru is the pore water coeffi-
(ua − uw) ≥ 60 kPa. It can be concluded that clays can be treated as cient and h is the vertical distance between the pore-water table and a
soils with linear functions of matric suction φb = φ′ = 34° with smaller point on the failure surface of the slope, Vu is the volume of soil under
matric suctions. When it comes to extremely fine-grained soil, the sta- the pore-water table and Su is the slope surface under water table. For a
bility factor curves become linear, and the stability factors are in good case with no submerged condition or other boundary loads, the work
agreement with those under φb = φ′ = 34°. rate by water pressure on the boundaries of the slope is equal to 0, and
the work rate by pore-water pressure can be expressed as

7. Stability analysis of slope under water drawdown Wu = Wu′ = − ∫V uεii̇∗ dVu (26)

The stability conditions of a totally or partially submerged slope

may change when an inner pore water level drawdown takes place. In
addition, the acting zone of the matric suction is also related to the
water level. Furthermore, the water table shown in Fig. 3(a) is a special
case, it is worth investigating the stability of a slope subjected to a
water drawdown in unsaturated soil.
7.2. Failure mechanism of a slope subjected to water drawdown
Fig. 8 is the illustration of the failure mechanism of a 3D slope in
unsaturated soil subjected to water drawdown. In Fig. 8, the pore water
level is represented as Lw, which is in agreement with the condition
shown in Fig. 3(a). It is assumed that the soil under the water table is
completely saturated, and thus the matric suction can be taken into

109
J.-S. Xu, X.-L. Yang Engineering Geology 237 (2018) 102–115

O θB α1∗
D3D = 2ω ∫θ ∫0 c′R (rm + R cos α )2dαdθ

θ0
0

θB
θh α1∗
+ 2ω ∫θ ∫0 c′R (rm + R cos α )2dαdθ

θF
θE
B

θh
θB α1∗
r0 + 2ω ∫θ ∫0 0
(ua − u w ) tan φbR (rm + R cos α )2dαdθ
θF α2∗
B A + 2ω ∫θ ∫0 B
(ua − u w ) tan φbR (rm + R cos α )2dαdθ
(30)

L w /H
7.3.2. Linearly increased matric suction
rh D If θF < θB, then
F H θF α1∗
z1
D3D = 2ω ∫θ ∫0
0
cR (rm + R cos α )2dαdθ
z2 θB α1∗

β E
+ 2ω ∫θ ∫0 F
c′R (rm + R cos α )2dαdθ
θh α2∗
C + 2ω ∫θ ∫0 B
c′R (rm + R cos α )2dαdθ
(31)
Fig. 8. Illustration of water level drawdown inside a slope in unsaturated soils.
Otherwise, if θF ≥ θB, then
θB α1∗
account within the range of θ0 − θF, and in particular Lw = 1.0 in- D3D = 2ω ∫θ ∫0
0
cR (rm + R cos α )2dαdθ
dicates that the slope is completely submerged and there is no matric θF α2∗
suctions inside. In addition, it should be noted here that when the water + 2ω ∫θ ∫0 B
cR (rm + R cos α )2dαdθ
table is lower than the slope toe, a water level change can only change θh α2∗
the acting range of the matric suction, while when the water table is + 2ω ∫θ ∫0 F
c′R (rm + R cos α )2dαdθ
(32)
higher than the slope toe, the variation of water table can both change
the acting range of the matric suction and the work rate by pore-water θF r θh α2∗ r
pressure inside the slope. Therefore, only the conditions that the water
Dinsert = b ∫θ 0
cv cos φ
cos φ
dθ + b ∫θ ∫0
F
c′v cos φ
cos φ
dθ

table is equal or higher than the slope toe is considered in this section. (33)
In accordance with the studies by Viratjandr and Michalowski
(2006) and Gao et al. (2014), in this study, the pore water table is
assumed uniform throughout the slope, and an instantaneous draining 7.3.3. Work rates by pore-water pressure
process is taken place and during which the water table remains vertical If θF < θB, then
within a portion of the slope, as shown in Fig. 8. The assumption of an θB α1∗
instantaneous drawdown will lead to overestimation estimations of the Wu − 3D = 2ωγw tan φ ⎡
⎣
∫θ ∫0 F
h1 − 1 (rm + R cos α )2 dθdxRdα
hydraulic gradients, and it is acceptable for practical calculations. θE α2∗
Consequently, according to the studies by (Michalowski and Nadukuru, + ∫θ ∫0 B
h1 − 2 (rm + R cos α )2 dθdxRdα
2013; Saada et al., 2012), the pore water pressures inside a slope can be θh α2∗
treated as an external force, taking into account the energy limit + ∫θ ∫0 E
h2 (rm + R cos α )2 dθdxRdα⎤
⎦ (34)
equilibrium equation.
In virtue of the water drawdown illustration in Fig. 8, the work rate Otherwise, if θLW21 ≥ θB then
by pore-water pressures and variations of the integration bounds for
θE α2∗
matric suction calculations can be calculated: According to the geo- Wu − 3D = 2ωγw tan φ ⎡
⎣
∫θ ∫0 F
h1 − 2 (rm + R cos α )2 dθdxRdα
metrical relationships shown in Fig. 8, z1, the vertical distance from the
θh α2∗
water level to the failure surface of the slope, and z2, the vertical dis- + ∫θ ∫0 E
h2 (rm + R cos α )2 dθdxRdα⎤
⎦ (35)
tance from slope surface to the failure surface of the slope, can be
written as z1 ( R2 − x 2 − d1 ) z1 ( R2 − x 2 − d2 )
where h1 − 1 = R − d1
, h1 − 2 = R − d2
and h2 =
z1 = rsinθ − r0 sin θ0 − L w / H (27) z2 ( R2 − x 2 − d2 )
.
R − d2
The work rate by pore water pressure on the insertion block can be
z2 = rsinθ − rh sin θh + (rcosθ − rh cos θh) tan β (28) written as

Wu − insert = ωγw r04 A (36)

7.3. Internal energy dissipation where A = b

H
tanφ {∫ θE z1
θF r 0
exp[2(θ − θ0 ) tan φ]dθ + ∫
θh z2
θE r 0 }
exp[2(θ − θ0 ) tan φ]dθ .

{exp[(θh − θ0 ) tan φ] sin θh − sin θ0}

7.3.1. Uniformly distributed matric suction The stability factors Ns = γHc/c′ for a 3D slope under uniformly
If θF < θB, then distributed and linear increased condition can be obtained with respect
θB α1∗
to the parameters θ0, θh, r′0/r0, and b/B following the restraint condi-
D3D = 2ω ∫θ ∫0
0
c′R (rm + R cos α )2dαdθ tions in Eq. (24) and 0 ≤ Lw ≤ 1.0, as follows:
θh α1∗ For a linearly increased condition,
+ 2ω ∫θ ∫0 c′R (rm + R cos α )2dαdθ
B Ns = minf (θ0 , θh , r0′/ r0, b/ H | c′, φ′, φb , γ , B, β , ρ) (37)
θF α1∗
+ 2ω ∫θ ∫0 0
(ua − u w ) tan φbR (rm + R cos α )2dαdθ
(29) For the nonlinearly distributed conditions,

Otherwise, if θF ≥ θB, then Ns = minf (θ0 , θh , r0′/ r0, b/ H | c′, φ′, φb , γ , B, β , (ua − u w ) ) (38)

110
J.-S. Xu, X.-L. Yang Engineering Geology 237 (2018) 102–115

Table 5 drawdown significantly influences the stability of a slope, particularly

Stability factors of slope subjected to water drawdown (γ = 20kN/m3, β = 60°, φ′ = 34°, for conditions of higher matric suctions and suction angles under a
B/H = 5.0, and ua − uw = 140 kPa).
lower pore water level. This makes sense because the greater the matric
SWCC Models Water level suction and suction angle are, the stronger the enhancing impacts of the
matric suction on the slope stability will be. A drawdown of the water
0 0.2 0.4 0.6 0.8 1.0 level indicates a decrease in the pore water pressure acting zone
(θF − θh) and an increase in the acting zone of the matric suction
No. 1 1 6.99 9.30 13.20 19.94 28.77 41.49
2 6.99 8.62 11.14 14.83 18.93 26.08 (θ0 − θF), as shown in Fig. 8, which is beneficial to the slope stability
3 6.99 8.75 11.53 15.79 20.79 28.99 and enhancing effects of the matric suction on this stability.
4 6.99 9.02 12.34 17.81 24.67 35.06
5 6.99 8.62 11.14 14.83 18.93 26.08
No. 2 1 6.99 8.99 12.25 17.58 24.23 34.37
8.3. Parameter analysis under linear suction with depth
2 6.99 8.62 11.14 14.83 18.93 26.08
3 6.99 8.95 12.14 17.32 23.73 33.59 When the matric suction is linearly increased with the depth, as
4 6.99 10.04 15.40 25.40 39.27 57.93 shown in Fig. 10, the effects of the water level drawdown on the sta-
5 6.99 8.62 11.14 14.83 18.94 26.08
bility factors are similar to those under uniformly distributed condi-
No. 3 1 6.99 11.77 20.97 39.23 65.87 99.60
2 6.99 13.19 25.72 51.01 88.54 135.10 tions. A decrease in the water level expands the acing zone (θ0 − θF) of
3 6.99 10.06 15.48 25.6 39.66 58.55 the matric suction and consequently its magnitude, resulting in a
4 6.99 13.35 26.25 52.33 91.09 139.10 greater impacts of the matric suctions on the slope stability and better
5 6.99 14.05 28.61 58.19 102.36 156.74 stability conditions. In addition, it is also clear that an increase in the
No. 4 1 6.99 16.42 37.09 79.24 142.86 220.17
2 6.99 16.50 37.43 80.09 144.48 222.70
matric suction angle φb will accelerate the effects of the matric suction
3 6.99 12.87 24.65 48.36 83.44 127.12 on the slope stability.
4 6.99 16.57 37.76 80.90 146.05 225.17
5 6.99 16.57 37.76 80.90 146.05 225.17
8.4. Parameter analysis under shear strength nonlinearity

8. Numerical results and parameter analysis
In this section, some numerical solutions of safety factors of the
slope subjected to water drawdown were presented. Thereafter, the
effects of matric suction distributions of a 3D unsaturated slope sub-
jected to a water level drawdown are demonstrated: First, the linear
distributions, i.e., the uniform and linear increased distributions will be
dealt with, followed by the nonlinear functions under different SWCC
models.
8.1. Numerical solutions
Table 5 shows the stability factors Ns of a 3D unsaturated slope of
different pore water levels with respect to γ = 20 kN/m3, β = 60°, B/
H = 5.0, c' = 10 kPa and φ′ = 34° under different SWCC models.
8.2. Parameter analysis under uniform suction
The effects of the uniformly distributed matric suction of a slope
under a water level drawdown are illustrated herein. Fig. 9(a) and (b)
show the stability factors as a function of the pore water level Lw under
matric suction and the effective suction angle φb, respectively.
It can be clearly seen in Fig. 9(a) and (b) that the water level
The stability of slopes under different SWCC models and soil types
are described herein. Fig. 11 illustrates the stability factors of a slope
along with the matric suction under diverse pore water levels. Fig. 12
shows the rules of the stability factors as a function of the pore water
level under different SWCC models.
Fig. 11 shows that the variations in water level will not change the
effects of the shear strength equations on the stability factors as a
function of the matric suction. However, it should be noted that when
the water level is relatively high (Lw is low, as illustrated in Fig. 8), the
variation in the stability factors is limited and the trends are gentle.
This is due to the fact that when the water table inside a slope increases,
a saturated soil area expands, and the acting zone of matric suction
inside the slope shrinks, thereby reducing the effects of the matric
suction on slope stability. With a constant decrease in the water level,
the acting zone of matric suction continues to increase, and the effect of
the matric suction on the slope stability becomes more significant.
Fig. 12 shows an illustration of the stability factors corresponding to
diverse SWCC models of a slope along with a water level drawdown. In
Fig. 11, c' = 10 kPa, φ′ = 34° and ua − uw = 140 kPa.
Fig. 12 shows the variation trends of the stability factors of four
classes of soils along with water level Lw under different SWCC models.
It is clear that the stability factors of an unsaturated slope are directly
related to the water level, and simultaneously, the SWCC models and
suction angle. The variation trends of the stability factors along with the

30 25
(a) b
β=60°, B/H=5.0, φ =15°, φ =15°
b
β=60°, B/H=5.0, φ'=25°, φb=20°
(b)
25 ua-uw=30kPa
ua-uw=100 20

20
Ns
Ns

40 15
60
15 20
6°
0 8°
10
10 10°
ua-uw/(kPa) 14°
5 5
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Lw Lw

Fig. 9. Effects of water drawdown on slope stability under uniformly distributed pattern of matric suction.

111
J.-S. Xu, X.-L. Yang Engineering Geology 237 (2018) 102–115

30 80

b
(b) β=45°, B/H=5.0, φ'=25°, ρ=5.0
25 (a) β=45°, B/H=5.0, φ'=15°, φ =15° φb=20°
ρ=10.0 60 16°
8.0 14°
20

Ns
Ns

6.0 40 10°
15 6°

0
20
10
2.0
4.0
ρ/(kPa/m)
5 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Lw Lw

Fig. 10. Effects of water drawdown on slope stability under linearly increased pattern of matric suction.

40 60
(a) Sandy soils (b) A fine-grained soils (silt)

30
Lw=1.0 Lw=1.0
40
Ns

Ns
20
Lw=0.6

20
10 Lw=0.4

Lw=0.4 Lw=0.6
0 0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
(ua-uw)/kPa (ua-uw)/kPa

150 120
(c) Clays (d) Extremely fine-grained soils
120
90

90
Lw=1.0
Ns

Ns

60
Lw=1.0
60

30
30

0 0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
(ua-uw)/kPa (ua-uw)/kPa
Fredlund et al., (1996) Vanapalli et al., (1996)
Vilar, (2006) Khalili and Khabbaz, (1998)
Bao et al., (1998) Totally saturated (Lw=0)

Fig. 11. Effects of matric suction nonlinearity on stability of slope subjected to water drawdown.

water level are aggravated as Lw increases, which is due to the increase
in the acting zone of the matric suction. For sandy and fine-grained soils
with af smaller than 10 kPa and AEV smaller than 5 kPa, the stability
factors are lower than those obtained under the average value suction
angle of φb = 15°, and the effects of the matric suction on the slope
stability are not crucial and can be ignored for conditions under high
water levels (Lw ≤ 0.5). At the same time, stability factors obtained
under Vanapalli et al.'s (1996) model, Bao et al.'s (1998) model and the
curve under (ua − uw) = 0 overlap, indicating that these two models
are non-effective for sandy soil, which is in accordance with Fig. 7. As
for clays and extremely fine-grained soil, the matric suction plays a
dominant role in the stability of an unsaturated soils slope, particularly
for a slope of a lower water level (higher Lw), whose stability factors are
far above those under φb = 15°, meaning that the average value of
φb = 15° yields conservative estimations on the slope stability for clay
and extremely fine-grained soil. Furthermore, the variation in magni-
tudes of the stability factors along with water level Lw can be predicted
to increase as the matric suction (ua − uw) increases.
9. Case study
As reported by Rahardjo et al. (2001) and Travis et al. (2010), a
number of failures were triggered by rainfall in February 1995 at the
Nanyang Technological University (NTU) campus in Jurong, Western

112
J.-S. Xu, X.-L. Yang Engineering Geology 237 (2018) 102–115

40 40
(b) A fine-grained soils (silt)
(a) Sandy soils
30 b
φ =φ' =34° 30 φb=φ' =34°

φb=15° φb=15°

Ns
Ns

20 20

10 10

af=1, nf=2, mf=1, ψr=10kPa, κ=1.0, cult=11.4kPa af=10, nf=2, mf=1, ψr=100kPa, κ=1.8, cult=13.7 kPa
0 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Lw Lw

120 160
(d) Extremely fine-grained soils
(c) Clays
100 b
φ =φ'=34° af=1000, nf=2, mf=1, ψr=10000kPa,
af=100, nf=2, mf=1, 120
κ=2.5, cult=107.3kPa
80 ψr=1000kPa, κ=2.2,
cult=28.4 kPa
φb=φ'=34°
Ns

Ns
60 80

40 φb=15°
b
φ =15° 40
20

0 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Lw Lw
Fredlund et al., (1996) Vanapalli et al., (1996)
Vilar, (2006) Khalili and Khabbaz, (1998)
Bao et al., (1998) φb=15°
φb=φ'

Fig. 12. Effects of water level drawdown on slope stability.

Table 6
Parameters for NTU failure number 4.

Property Value Units

γs 19.6 kN/m3
c' 2.0 kPa
φ' 26.0 deg
φb 13.0 deg
γw 9.8 kN/m3
γd 16.1 kN/m3
θr 13.0 %
θs 35.5 %

Singapore. Among the failures, the Slip number 4 was used by Travis
et al. (2010) to investigate the effects of matric suction on the stability
of the slope in unsaturated clay.
The Slip number 4 had a slope angle of 45°, a width of 7.8 m, a
length of 7 m and a depth of 1 m. With regard to the strength of the
Fig. 13. NTU Slip Number 4 safety factor versus the water table.
unsaturated clay, the failed slope had a saturated unit weight γs of
19.6 kN/m3, a dry unit weight γd of 16.1 kN/m3, and a saturated water
content θs of 35.5%. The effective internal friction angle φ′, the suction To investigate the stability of the slope in unsaturated clay, based on
angle φb, and the cohesion c' were 26°, 13° and 2 kPa respectively, the limit analysis method, four types of calculation models i.e., the liner
found by strength testing. The parameters are as given as in Table 6. increased model, the Fredlund et al.'s (1996) model, and the Vilar's
According to the investigation by Travis et al. (2010), the water (2006) model, are adopted in this section to investigate the effect of the
table inner the Slip number 4 changed rapidly from 5 m to < 1 m owing matric suction on slope stability.
to the storm, resulting in the variation change of the stability status of Fig. 13 shows the safety factors of the Slip number 4 under the di-
the slope. The safety factors of the slope both with and without the verse models versus the storm progression. As investigated by Travis
consideration of the matric suction were calculated under the quasi- et al. (2010), it is clear in Fig. 13 that the slope failed (Fdry/sat = 0.68)
unsteady model by Travis et al. (2010), as shown in Fig. 13. even before the storm under the dry/saturated model. However, when

113
J.-S. Xu, X.-L. Yang Engineering Geology 237 (2018) 102–115

the matric suction was taken into consideration, the slope kept a steady
state at the beginning of the storm, and then reduced sharply to about
0.5.
With regard to the models adopted in this study, in accordance with
the investigations by Travis et al. (2010), this study reveals that the
matric suction has favorable effects on slope stability. When the water
table rises to about 3.5 m, the safety factors of the slope reduce to 1.0,
representing the limit state of failure. Thereafter, the safety factors of
the slope continue to decrease during the storm and eventually reach to
about 0.5. All of the models indicate a high probability of failure, and
the matric suction should be considered in practical engineering for a
more precise prediction of the slope stability condition. It is worth
noting that, even if the most of the safety factors in Fig. 13 are < 1.0,
these safety factors are still meaningful and necessary for landslide
dynamics and risk analysis, as explained by Travis et al. (2010).
10. Conclusions
In this study, the limit analysis method was employed to investigate
the stability of 3D slopes in unsaturated soils considering different
matric suction distributions, strength nonlinearity, and water draw-
down. By introducing different matric suction distribution patterns, the
work rates and internal energy dissipations of 3D slopes were calcu-
lated, and the critical heights were derived using the energy equili-
brium equation. The effects of the matric suction distributions, shear
strength nonlinearity and water drawdown on the slope stability were
investigated. Some numerical solutions were presented for con-
venience. The following conclusions from this study can be made:
(2) When the matric suction is uniformly distributed, an increase in the
matric suction provides stability factors with linear increasing rules.
When the matric suction linearly increases with depth, the im-
proving effects of the matric suction on the slope stability increase
as the magnitude of the matric suction variation increases, parti-
cularly for gentle slopes with a narrower width.
(3) With regard to nonlinear conditions, different SWCC models will
yield diverse shear strength envelopes and striking differences in
the stability factor solutions for different soil types. The average
value of the suction angle φb = 15° is not reasonable to all types of
soil. For sandy soil with lower AEV values, the influences of the
matric suction can be ignored within(ua − uw) ≤ 40 kPa; for ex-
tremely fine-grained soil with a high AEV values, the effects of the
matric suction on the slope stability can be seen as linear functions
of φ′ = φb, and for fine-grained soils and clay, the influences of the
matric suction corresponding to different estimation equations on
the slope stability are diverse and tremendous, and the functions of
the matric suction on the stability of an unsaturated slope should be
seriously considered.
(4) The water level drawdown inside a slope significantly influences
the intensity of the matric suction on the slope stability because the
acting zone of the matric suction inside a slope is directly related to
such a drawdown. When the high water table nears the slope crest,
the effects of the matric suction on the slope stability are negligible.
As the water level decreases during the water drawdown process,
the effects of the matric suction on the slope stability become more
remarkable.

(1) 3D geometry characteristics have a remarkable impact on the sta- Acknowledgement

bility of 3D unsaturated slopes in that the stability factors experi-
ence a rapid decrease along with B/H within B/H ≤ 1.5, after
which the variation rules become smooth and eventually show little
change. When B/H ≥ 10.0, the solutions can be treated as 2D plane
strain solutions.
The preparation of this paper has received financial supports from
Innovation Foundation for Postgraduate of Central South University,
China (2015zzts061), and National Natural Science Foundation
(51378510). The financial supports are greatly appreciated.

Appendix A

1 ⎡ (θ − θ0) tan φ ′ r′
f1 = e + 0 e−(θ − θ0) tan φ ′⎤
2⎢⎣ r0 ⎥
⎦

1 ⎡ (θ − θ0) tan φ ′ r′
f2 = e − 0 e−(θ − θ0) tan φ ′⎤
2⎣⎢ r0 ⎥
⎦

sin θ0 1 r′ sin θ0
f3 = − ⎡e (θ − θ0) tan φ ′ + 0 e−(θ − θ0) tan φ ′⎤ = − f1
sin θ 2⎢⎣ r0 ⎥
⎦ sin θ

sin θ0 sin(θB + β1) 1 r′ sin θ0 sin(θB + β1)

f4 = − ⎡e (θ − θ0) tan φ ′ + 0 e−(θ − θ0) tan φ ′⎤ = − f1
sin θB sin(θ + β1 ) 2⎣⎢ r0 ⎥
⎦ sin θB sin(θ + β1 )

1
f5 = [(3 tan φ′ cos θh + sin θh ) e3(θh − θ0) tan φ ′ − (3 tan φ′ cos θ0 + sin θ0)]
3(1 + 9tan2 φ′)

1 L⎛ L
f6 = 2 cos θ0 − ⎞ sin θ0
⎜ ⎟

6 r0 ⎝ r0 ⎠

f7 = 1 + sin θ0 + [f2 − e (θ − θ0) tan φ ′] sin θ

f8 = 1 + sin θ0 − e (θ − θ0) tan φ ′ sin θ

⎡f ⎛f 2 + 1 f 2 ⎞ arccos(f / f ) + f ⎛2f + 1 f ⎞ f 2 − f 2 ⎤ dθ
θB
g1 = 2 ∫θ 2 1
2 2⎠ 3 2 2 1
2 3⎠ 2 3
0 ⎣ ⎝ ⎝ ⎦
θB 1 1
g2 = 2 ∫θ f7 ⎡f2 ⎛f12 + f22 ⎞ arccos(f3 / f2 ) + f2 ⎛2f1 + f3 ⎞ f22 − f32 ⎤ dθ
2 ⎠ 2 ⎠
0 ⎣ ⎝ ⎝ ⎦

114
J.-S. Xu, X.-L. Yang Engineering Geology 237 (2018) 102–115

⎡f f 3 arccos(f / f ) + f ⎛f 2 + f f + 2 f 2 + 1 f 2 ⎞ f 2 − f 2 ⎤ sinθdθ
θB
g3 = 2 ∫θ 1 2 3 2 2 1
⎝
1 3
3 2 3 3⎠ 2 3
0 ⎣ ⎦

⎡f ⎛f 2 + 1 f 2 ⎞ arccos(f / f ) + f ⎛2f + 1 f ⎞ f 2 − f 2 ⎤ dθ
θh
g4 = 2 ∫θ 2 1
2 2⎠ 4 2 2 1
2 4⎠ 2 4
B ⎣ ⎝ ⎝ ⎦
θh 1 1
g5 = 2 ∫θ f7 ⎡f2 ⎛f12 + f22 ⎞ arccos(f4 / f2 ) + f2 ⎛2f1 + f4 ⎞ f22 − f42 ⎤ dθ
2 ⎠ 2 ⎠
B ⎣ ⎝ ⎝ ⎦

⎡f f 3 arccos(f / f ) + f ⎛f 2 + f f + 2 f 2 + 1 f 2 ⎞ f 2 − f 2 ⎤ sin θdθ

θh
g6 = 2 ∫θ 1 2 4 2 2 1
⎝
1 4
3 2 3 4⎠ 2 4
B ⎣ ⎦

b (θh − θ0) tan φ ′ θh

g7 = [e sin θh − sin θ0 ] ∫θ e 2(θ − θ0) tan φ ′⎞⎟ ⎤ dθ
H 0 ⎥
⎠⎦

b (θh − θ0) tan φ ′ θh

g8 = [e sin θh − sin θ0 ] ∫θ f8 e 2(θ − θ0) tan φ ′⎟⎞ ⎤ dθ
H 0 ⎥
⎠⎦

4 2 2
θB ⎡ ⎛f f2f2⎞ ⎛ f2 − f3 ⎞
g9 = 2 ∫θ ⎢ (f22 f3 /8 − f33 /4 − 2f1 f32 /3 − f3 f12 /2 + 2f1 f22 /3) f22 − f32 + ⎜ 2 + 1 2 ⎟ arcsin ⎜
8 2 ⎠ f2 ⎟ cos θdθ
0 ⎢ ⎝
⎣ ⎝ ⎠
4 2 2
θh ⎡ ⎛f f2f2⎞ ⎛ f2 − f 4 ⎞
+2 ∫θ ⎢ (f22 f4 /8 − f43 /4 − 2f1 f42 /3 − f4 f12 /2 + 2f1 f22 /3) f22 − f42 + ⎜ 2 + 1 2 ⎟ arcsin ⎜ ⎟ cos θdθ
⎢ ⎝8 2 ⎠ f2
B
⎣ ⎝ ⎠
b
g10 = (f − f6 − f7 )[e (θh − θ0) tan φ ′ sin θh − sin θ0]
H 5

References rainfall-induced failure of cut slopes. Eng. Geol. 184, 96–103.

Assefa, S., Graziani, A., Lembo-Fazio, A., 2017. A slope movement in a complex rock
formation: deformation measurements and DEM modelling. Eng. Geol. 219, 74–91.
Bao, C., Gong, B., Zhan, L., 1998. Properties of unsaturated soils and slope stability of
expansive soils. In: Proceedings of the Second International Conference on
Unsaturated Soils (UNSAT 98), Beijing. 1, pp. 71–98.
Camones, L.A.M., Vargas, E.A., Figueiredo, R.P.D., Velloso, R.Q., 2013. Application of the
discrete element method for modeling of rock crack propagation and coalescence in
the step-path failure mechanism. Eng. Geol. 153, 80–94.
Eid, H.T., 2014. Stability charts for uniform slopes in soils with nonlinear failure envel-
opes. Eng. Geol. 168, 38–45.
Fredlund, D.G., Xing, A., 1994. Equations for the soil-water characteristic curve. Can.
Geotech. J. 31 (4), 521–532.
Fredlund, D.G., Morgenstern, N.R., Widger, R.A., 1978. The shear strength of unsaturated
soils. Can. Geotech. J. 15 (3), 313–321.
Fredlund, D.G., Xing, A., Fredlund, M.D., Barbour, S.L., 1996. The relationship of the
unsaturated soil shear strength to the soil water characteristic curve. Can. Geotech. J.
32 (3), 440–448.
Gao, Y.F., Zhu, D.S., Zhang, F., Lei, G.H., Qin, H.Y., 2014. Stability analysis of three-
dimensional slopes under water drawdown conditions. Can. Geotech. J. 51 (6),
1355–1364.
Gischig, V.S., Eberhardt, E., Moore, J.R., Hungr, O., 2015. On the seismic response of
deep-seated rock slope instabilities-insights from numerical modeling. Eng. Geol.
193, 1–18.
Houston, S.L., Perez-garcia, N., Houston, W.N., 2008. Shear strength and shear-induced
volume change behavior of unsaturated soils from a triaxial test program. J. Geotech.
Geoenviron. 134 (11), 1619–1632.
Hoyos, L.R., Suescún-Florez, E.A., Puppala, A.J., 2015. Stiffness of intermediate un-
saturated soil from simultaneous suction-controlled resonant column and bender
element testing. Eng. Geol. 188, 10–28.
Jing, L., 1998. Formulation of discontinuous deformation analysis (DDA)-an implicit
discrete element model for block systems. Eng. Geol. 49, 371–381.
Khalili, N., Khabbaz, M.H., 1998. A unique relationship for χ for the determination of the
shear strength of unsaturated soils. Géotechnique 48 (5), 681–687.
Li, T.Z., Yang, X.L., 2018. Reliability analysis of tunnel face in broken soft rocks using
improved response surface method. Int. J. Geomech. 18. http://dx.doi.org/10.1061/
(ASCE)GM.1943-5622.0001129.
Lim, K., Lyamin, A.V., Cassidy, M.J., Li, A.J., 2016. Three-dimensional slope stability
charts for frictional fill materials placed on purely cohesive clay. Int. J. Geomech. 16
(2), 04015042.
Lu, N., Likos, W.J., 2004. Unsaturated Soil Mechanics. John Wiley and Sons, New York.
Michalowski, R.L., Drescher, A., 2009. Three-dimensional stability of slopes and ex-
cavations. Géotechnique 59 (10), 839–850.
Michalowski, R.L., Nadukuru, S.S., 2013. Three-dimensional limit analysis of slopes with
pore pressure. J. Geotech. Geoenviron. 139 (9), 1604–1610.
Oh, S., Lu, N., 2015. Slope stability analysis under unsaturated conditions: case studies of
Pan, Q.J., Dias, D., 2017. Upper-bound analysis on the face stability of a non-circular
tunnel. Tunn. Undergr. Space Technol. 62, 96–102.
Qin, C.B., Chian, S.C., 2017. Kinematic stability of a two-stage slope in layered soils. Int.
J. Geomech. 17 (9), 06017006.
Rahardjo, H., Li, X.W., Toll, D.G., Leong, E.C., 2001. The effect of antecedent rainfall on
slope stability. Geotech. Geol. Eng. 19, 371–399.
Saada, Z., Maghous, S., Garnier, D., 2012. Stability analysis of rock slopes subjected to
seepage forces using the modified Hoek-Brown criterion. Int. J. Rock Mech. Min. Sci.
55, 45–54.
Salmi, E.F., Nazem, M., Karakus, M., 2016. Numerical analysis of a large landslide in-
duced by coal mining subsidence. Eng. Geol. 217, 141–152.
Tarantino, A., Mountassir, G.E., 2013. Making unsaturated soil mechanics accessible for
engineers: preliminary hydraulic-mechanical characterisation and stability assess-
ment. Eng. Geol. 165, 89–104.
Travis, Q.B., Houston, S.L., Marinho, F.A.M., Schmeeckle, M., 2010. Unsaturated infinite
slope stability considering surface flux conditions. J. Geotech. Geoenviron. 136 (7),
963–974.
Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E., Clifton, A.W., 1996. Model for the prediction
of shear strength with respect to soil suction. Can. Geotech. J. 33 (3), 379–392.
Vilar, O.M., 2006. A simplified procedure to estimate the shear strength envelope of
unsaturated soils. Can. Geotech. J. 43 (10), 1088–1095.
Viratjandr, C., Michalowski, R.L., 2006. Limit analysis of submerged slopes subjected to
water drawdown. Can. Geotech. J. 43 (8), 802–814.
Yang, X.L., Li, Z.W., 2018. Factor of safety of three-dimensional stepped slope. Int. J.
Geomech. 18. http://dx.doi.org/10.1061/(ASCE)GM.1943-5622.0001154.
Yang, X.L., Yao, C., 2018. Stability of tunnel roof in nonhomogeneous soils. Int. J.
Geomech. 18 (3), 06018002.
Yang, X.L., Zhang, R., 2018. Limit analysis of stability of twin shallow tunnels considering
surface settlement. KSCE J. Civ. Eng. 22. http://dx.doi.org/10.1007/s12205-017-
1398-8.
Zhang, L.L., Fredlund, D.G., Fredlund, M.D., Wilson, G.W., 2014. Modeling the un-
saturated soil zone in slope stability analysis. Can. Geotech. J. 51 (12), 1384–1398.
Zhang, C.G., Chen, X.D., Fan, W., 2016a. Critical embedment depth of a rigid retaining
wall against overturning in unsaturated soils considering intermediate principal
stress and strength nonlinearity. J. Cent. South Univ. 23 (4), 944–954.
Zhang, C.G., Chen, X.D., Fan, W., 2016b. Overturning stability of a rigid retaining wall for
foundation pits in unsaturated soils. Int. J. Geomech. 16 (4), 06015013.
Zhou, X.P., Cheng, H., 2013. Analysis of stability of three-dimensional slopes using the
rigorous limit equilibrium method. Eng. Geol. 160, 21–33.
Zhou, X.P., Cheng, H., 2014. Stability analysis of three-dimensional seismic landslides
using the rigorous limit equilibrium method. Eng. Geol. 174, 87–102.
Zhou, X.P., Cheng, H., 2015. The long-term stability analysis of 3D creeping slopes using
the displacement-based rigorous limit equilibrium method. Eng. Geol. 195, 292–300.
Zhou, X.P., Zhao, Y., Qian, Q.H., 2015. A novel meshless numerical method for modeling
progressive failure processes of slopes. Eng. Geol. 192, 139–153.

115