Professional Documents
Culture Documents
Research Paper
a
Department of Civil Engineering, Shanghai University, Shanghai 200444, China
b
College of Civil Engineering and Architecture, Quzhou University, Zhejiang 324000, China
c
School of Engineering, RMIT University, Melbourne 3001, Australia
Keywords: Seismic stability analyses of soil slopes are routinely performed with dry and/or saturated assumptions under
Seismic stability two-dimensional (2D) conditions. However, in practice, soils are often unsaturated and slopes usually fail in 3D
Limit analysis fashion. Three-dimensional (3D) effect and suction-induced effect are essential to be considered to produce more
Unsaturated soil realistic solutions in the slope stability analyses. In this regard, an analytical framework is developed in this
Gentle soil slope
paper for assessing the 3D seismic stability of unsaturated soil slopes. A semi-analytical method including the
Semi-analytical method
simplified method (SM) and the layer-wise summation method (LSM) is developed to deal with the highly
variable nature of suction. The validity of the analytical framework in assessing the soil slope stability is checked
with published solutions. Examples are given to illustrate the 3D effect and the effect of suction on the slope
seismic stability. The described framework allows the user to quantify the effect of changes in moisture content
on effective stress distribution and effective unit weight distribution in the slope and thus, changes in the slope
stability.
1. Introduction LEMs (e.g., [14,47]) are still the classical approaches, and can be used
to the safety estimation of 3D slopes by extending the 2D cases. How-
Earthquake-related damages to the soil slopes or dams have been ever, the method is statically indeterminate as arbitrary assumptions
widely reported and investigated (e.g., [25,8]). Seismic excitation is are introduced to determine the location of the thrust line or surface to
one of the possible factors leading to failures of slopes or dams. Stability satisfy the force and moment equilibriums. Furthermore, the results
assessments of slopes subjected to seismic shaking are routinely per- given by the LEMs are neither a strict lower- nor upper-bound solutions
formed under two-dimensional (2D) conditions using a pseudo-static [19]. The numerical approaches, such as the finite element method
approach (e.g., [2,46]). In the pseudo-static approach, the seismic ex- (FEM) (e.g., [7,17]), etc., have gained significant attention in geo-
citation caused by seismic acceleration is treated as a steady uniformly technical engineering. FEM is more applicable for specific, well-defined
distributed inertial force (horizontal and/or vertical). Though this ap- slopes with accurate geotechnical parameters, but would become more
proach does not reflect the characteristics of seismic excitation (e.g., the elaborate and time-consuming to produce solutions for a wide range of
duration, periodicity and amplification, etc.) on the slope stability, it parameters. Compared with the LEM, the classical plasticity theorem-
allows for implementing a quantitative assessment of the slope stability based LAM proposed originally by Drucker and Prager [6] is a much
and is widely used in the preliminary stage for designing new slopes or more rigorous method for assessing the stability of geostructures (e.g.,
the safety assessment of existing slopes. For slopes failing with three- [3,24,35]). Meanwhile, with the advent and development of 3D ad-
dimensional (3D) fashion, the 2D approaches may be overly con- missible failure mechanisms in limit analysis (e.g., [5,27,10]), it is
servative (e.g., [7,4]). To produce more realistic solutions, it is of possible to evaluate the 3D seismic stability of slopes using the LAM.
practical significance to perform 3D stability analyses of slopes. In this Michalowski and Martel [28] adopted the pseudo-static approach to
regard, three main approaches, i.e., the traditional limit equilibrium assess the 3D seismic stability of steep soil slopes assuming a toe-failure
method (LEM), the robust numerical approaches and the limit analysis mechanism. Nadukuru and Michalowski [29] implemented a 3D dis-
method (LAM) can be used to assess the 3D seismic stability of slopes. placement analysis of slopes subjected to seismic loads by adopting a
The above approaches are used for their respective advantages. The below-toe failure mechanism. Zhang et al. [45] investigated the effect
⁎
Corresponding author at: 99 Shangda Road, Shanghai 200444, China.
E-mail address: sundean@shu.edu.cn (D. Sun).
https://doi.org/10.1016/j.compgeo.2019.02.008
Received 12 October 2018; Received in revised form 16 January 2019; Accepted 8 February 2019
0266-352X/ © 2019 Elsevier Ltd. All rights reserved.
L. Wang, et al. Computers and Geotechnics 110 (2019) 296–307
where f depicts the shear strength of unsaturated soils; c′ and ϕ′ are the
effective cohesion and internal friction angle, respectively; s represents
the suction stress; ua denotes the net stress on the failure plane; and
ua is the pore-air pressure and is assumed to be zero in this paper.
The second term on the right-hand side in Eq. (1), c = s tan ,
describes the shearing resistance arising from capillary effects and is
usually defined as capillary cohesion. The suction stress primarily arises
from the combination of the interparticle physicochemical stresses and
the capillary stresses caused by surface tension and negative pore-water
pressure. The equation for s can be expressed as a function of matric
suction (ua u w ) with two fitting parameters α and n and is given as
follows [22]:
(u a uw )
s = ua uw 0
{1 + [ (ua u w )]n }(n 1) n
(2)
297
L. Wang, et al. Computers and Geotechnics 110 (2019) 296–307
Table 1
Parameters for different soils under drying and wetting conditions (after [21]).
Soil USCS c' (kPa) ϕ′ (°) n α (kPa−1) Sr ksat (10–7 m/s) Gs
D W D W D W D W
A SC-SM 20.3 32 1.75 1.75 0.05 0.05 0.342 0.433 1.20 0.10 2.70
B SM 8.8 24 1.44 1.52 0.12 0.13 0.227 0.303 1.10 0.039 2.70
C ML 12.5 35 1.50 1.46 0.12 0.13 0.319 0.375 8.00 0.059 2.70
D SM 6.6 37 2.00 2.20 0.08 0.17 0.380 0.413 1.00 0.146 2.70
Note: D = drying; W = wetting; For lack of data in situ, the specific gravity (Gs) of each soil is assumed to be 2.70 in this analysis.
10 10
Distance from water table, z (m)
6 q/ks 6 q/ks
0 0
4 - 0.2 4 - 0.2
- 0.4 - 0.4
- 0.6 - 0.6
2 - 0.8 2 - 0.8
- 1.0 (NS) - 1.0 (NS)
drying and wetting conditions are listed in Table 1 [21]. Note that, for 1 ln[(1 + q ks ) e wz q ks ]
s =
unsaturated soils, the suction (or water content) obtained along any {1 + { ln[(1 + q k s ) e wz q k s ]}n}(n 1) n
(4)
wetting path is generally smaller than that obtained at the same water
In application of limit analysis to slopes without surface loads, the
content (or suction) along a drying path. This means that the effect of
self-weight of soil is usually identified as the external force. When the
suction on the slope stability is less pronounced when the soil para-
effect of suction is included in the slope safety assessment, the variation
meters along wetting paths are utilized in the slope stability analyses.
of moisture content would undoubtedly alter the unit weight of soils,
Hence, to yield a conservative estimation of the slope stability, only the
and thus affects the slope stability. In routine slope stability analyses,
wetting path is considered in present analysis.
the variation of moisture content is usually neglected (e.g., [40,30,20]),
Combining transient suction analysis with the LAM may bring dif-
which may yield unrealistic results. According to the proportional re-
ficulties in performing a 3D stability analysis of unsaturated soil slopes.
lationship between the gases, liquid and solid phases of unsaturated
As a preliminary study, this paper focuses on the stability of un-
soils, the effective unit weight of unsaturated soils can be derived as
saturated soil slopes under steady flow conditions. Combining Darcy’s
law and Gardner’s [11] model and considering the flow boundary ( sat w ) Gs + [Se (1 Se ) Sr ](Gs w sat )
=
conditions (zero suction at water table elevation), the matric suction Gs 1 (5)
versus depth relation under vertical steady flow condition can be ob-
tained analytically as follows [44,23]: where γsat represents the unit weight of saturated soil; Gs denotes spe-
cific gravity of soil; Se and Sr depicts the effective and residual degrees
1 q q of saturation, respectively. For most inorganic silts and clays, the value
(u a u w) = ln 1+ e wz
of Gs varies over a narrow range of 2.60 ∼ 2.80 [32]. For lack of data in
ks ks (3)
situ, the values of γsat and Gs, which can be easily ascertained in a la-
where q denotes the steady flow rate (negative value for infiltration, boratory or in situ, are assumed to be equal to 20 kN/m3 and 2.7 in this
positive value for evaporation and zero for hydrostatic condition); ks study.
describes the saturated hydraulic conductivity; γw is the unit weight of By using the soil water characteristic curve by Van Genuchten [38],
water; and z represents a distance from a generic point within the slope the relationship between the effective degree of saturation and matric
to the water table level. Eq. (3) implies that the matric suction in a soil suction can be written as
slope varies only in the vertical direction and is constant in the same 1
1 1 n
horizontal plane. Though the possible effect of transient flow near the Se =
1 + [ (u a uw )]n (6)
slope surface and the effect of slope shape on the suction distribution
cannot be considered, the vertical steady flow assumption allows for Fig. 3 shows the characteristics of Soil A under different infiltration
making a tractable analysis of the slope stability and is widely adopted conditions. The normalized flow rate q/ks is used and an unsaturated
in conventional slope stability analyses (e.g., [16,40,20]). soil layer with a depth of 10 m is investigated. Fig. 3(a) depicts the
Substituting Eq. (3) into Eq. (2), the suction stress profile under capillary cohesion profile and Fig. 3(b) describes the effective unit
vertical steady flow condition can be expressed as weight profile, respectively. As seen from Fig. 3, both the capillary
298
L. Wang, et al. Computers and Geotechnics 110 (2019) 296–307
cohesion profile and the effective unit weight profile exhibit nonlinear
shapes. Moreover, as the infiltration rate increases, the capillary co-
hesion profile decreases continually and approaches no suction case
(i.e., NS corresponds to q/ks = −1.0 in Eq. (4)) and the effective unit
weight profile increases continually and approaches that of saturated
condition (i.e., γsat = 20 kN/m3). More detailed description about suc-
tion can be found in the references (e.g., [23,22]).
Fig. 4. 3D ‘horn-shape’ rotational failure mechanism for a gentle unsaturated soil slope based on the concept of Nadukuru and Michalowski [29].
299
L. Wang, et al. Computers and Geotechnics 110 (2019) 296–307
Fig. 6. Calculation of the external work rate done by unsaturated soil weight: (a) for 2D plane insert; and (b) for 3D failure mechanism.
where k
ij and k
ij represent the stress and the strain rate tensor in a slope stability [45]. When kv ≤ 0.5 kh, the effect of vertical seismic
kinematically admissible mechanism, respectively; Wi( ) and Wi(s) de- acceleration on the 3D slope stability can nearly be ignored (e.g.,
scribe the work rates done by the gravity force and seismic force, re- [13,45]). Thus, only the effect of horizontal seismic acceleration on the
spectively; vik depicts the velocity vector consistent with the kinemati- 3D slope stability is considered in this paper.
cally admissible strain ijk ; and V is the volume of the rotating block. Andrianopoulos et al. [1] and Papadimitriou et al. [31] presented a
methodology for estimating the horizontal seismic coefficient kh of
earth dams and tall embankments. They suggested using an ‘‘effective’’
4. A semi-analytical method
value of the horizontal seismic coefficient, khE (approximately, khE/
khmax = 0.67 based on their literature), instead of the peak value of the
In this study, the external work rate is caused by the soil’s self-
seismic coefficient khmax in the design process. This is because that
weight and seismic load, while the internal energy is dissipated due to
using the value of khmax yields an overly conservative estimation of the
the soil cohesion and capillary cohesion. The energy balance equation
stability of geostructures. Hence, in this paper, only the ‘‘effective’’
can be expressed as
horizontal seismic coefficient khE is used in assessing the slope stability.
W + Ws = Dc' + Dc'' (8) Note that, for gentle soil slope subjected to stronger seismic ex-
citation, the critical failure surfaces tend to be very large and deep, and
where W and Ws represent the external work rates done by the un-
khE > 0.2 from Kramer [26]. To produce a rational result, as proposed
saturated soil weight and the seismic load, respectively; and Dc' and Dc''
by Gao et al. [9], the depth of the below-toe failure mechanism is
are the internal energy dissipation rates caused by the soil cohesion and
limited to a realistic value D below the slope toe elevation as shown in
the capillary cohesion, respectively. The expression of Dc' is available in
Fig. 4. In this paper, the depth factor D/H = 1.0 is used for the slopes
Nadukuru and Michalowski [29]. In the following, only detailed cal-
subjected to stronger seismic excitation.
culations of W , Ws and Dc'' are involved. Note that, in this paper the
effects of seismic excitation and infiltration on the slope stability are
4.2. A SM to calculate the external power rate
uncoupled. This will greatly facilitate the development of the analytical
framework.
Because of the highly variable nature of the effective unit weight
profile (Fig. 3b), it is hard to calculate the external work rate of the
4.1. Pseudo-static analysis unsaturated soil weight using a direct integrate method. Based on the
volume integral of the rotating body, a SM is developed to calculate the
The pseudo-static approach, though approximate, has the ad- work rate of the unsaturated soil weight. The volume element can be
vantages of accumulating experience and user friendliness and is widely written as
used in evaluating the seismic stability of geotechnical structures since
dV = A ( ) lc d (9)
it was first employed in 1950s [37]. Fig. 4 shows some significant
parameters of this approach, such as the peak values of the seismic where A(θ) is the area of soil represented by the line segment
acceleration at the outcropping bedrock, PGArock, at the ground surface, ¯ (m¯n , m ¯n ) for the plane insert (Fig. 6a) or the area of soil inter-
mn
PGA and at the slope crest, PGAcrest. In this approach, the seismic ex- sected by the curvilinear cone surface (the shaded region in Fig. 6b). lc
citation is treated as a uniformly distributed inertial force, Fh and Fv (h is a distance from the center point of area A(θ) to the rotation axis O.
and v denote the horizontal and vertical directions), which are applied The rotation velocity v regarding to the center point of area A(θ) is
at the sliding soil mass’s center of gravity as shown in Fig. 4. Hence, the perpendicular to the distance lc and can be given as
most important point of this approach is to determine the values of Fh
v = lc (10)
and Fv, which can be calculated by multiplying the weight of soil block,
Wγ', with a dimensionless seismic coefficients kh and kv, respectively. where ω is the angular velocity.
Vahedifard et al. [39] investigated the relationship between the To make the calculation manageable, the local nonlinear feature of
seismic coefficients kh and kv on the 2D stability of reinforced earth the effective unit profile in the volume element dV is neglected. The
structures. They found that when the direction of vertical seismic ac- weight of the volume element is calculated approximately by multi-
celeration is consistent with the direction of gravity, the stability of plying the volume and the effective unit weight at the center point (i.e.,
reinforced earth structures is enhanced. However, for a 3D approach, c (c′, c'') in Fig. 6). This assumption, though approximate, allows for
the vertical seismic acceleration with a downward direction reduces the implementing a relatively simple calculation of the external work rate
300
L. Wang, et al. Computers and Geotechnics 110 (2019) 296–307
done by the unsaturated soil weight. This is because that, in practice, cj = 0.5[ s (z )|z = jh1 + s (z )|z = (j 1) h1 ] tan (j = 1, 2 …m1) (14)
slopes or dams are usually constructed by fine-grained soils and the unit
weight profiles of soils generally present a linear distribution. The and
overall external work rate (including the work rate done by the soil ck = 0.5[ s (z )|z = (z 0 +
N ) + kh2 s (z )|z = (z 0 N ) + (k 1) h2 ] tan (k = 1,
weight and seismic load) for soil block ABCD (including the 2D and 3D
cases) are given as follows: 2 …m2) (15)
where s (z )|z = jh1 and s (z )|z = (j 1) h1 describe the suction stress at the top
W 2D + Ws2D
and bottom surfaces of layer j for soil block ABCG (Fig. 7b), respec-
1
= 4 br03 { 0
B
(z )|z = z1
(cos + k hE sin )
sin3
[sin e( 0 ) tan
+ sin 0]
2 tively; s (z )|z = (z 0 N ) + kh2 and s (z )|z = (z 0 N ) + (k 1) h2 capture the suction
stress at the top and bottom surfaces of layer k for soil block CDG
× [sin e ( 0) tan
sin 0]d (Fig. 7c), respectively. The expressions of s (z )|z = jh1, s (z )|z = (j 1) h1,
s (z )|z = (z 0 N ) + kh2 and s (z )|z = (z 0 N ) + (k 1) h2 can be easily got by sub-
C (cos + k hE sin )
+ (z )|z = z2 3
sin ( + ) sin3 B
stituting the variable z in Eq. (4) with jh1, (j − 1)h1, (z0 - N) + kh2 and
B
( 0 ) tan 2
[sin( + ) sin Be + sin( B + ) sin 0] (z0 - N) + (k − 1)h2, respectively. Note that, when the water table
× [sin( + ) sin Be
( 0 ) tan
sin( B + ) sin 0 ]d intersects the slip surface (i.e., z0 < N), the capillary cohesion below
(cos + k hE sin ) the water table is zero and the effective unit weight is equal to saturated
+ h
(z )|z = z3 [sin e( 0 ) tan
+ sin he
( h 0 ) tan
]2
C sin3 unit weight.
( ) tan ( h 0 ) tan
× [sin e 0
sin he ]d } For slopes in cohesive-frictional soils (ϕ′ > 0° and c′ > 0), the
(11) overall dissipated work rate due to the capillary cohesion of each soil
layer (e.g., layer j) (sum of the dissipation within volume DV and that
W 3D + Ws3D over the velocity discontinuity surfaces Dt) can be derived based on the
divergence theorem as
B
= (z )|z = z 4 (cos
0
Dc''3D - j = c j cot Srem - j
vi ni dSrem - j
2 2 2 2
R a R x
+ khE sin )(rm + x c1 )2 dydxd
0 a =c j cot vi ni dS + v n dS
Sj j 1 i i
+ vi ni dS
C R 2
d 2
R2 x 2
Sj j Sj 1j 1 (16)
+ (z )|z = z5 (cos + khE sin )(rm + x c2 )2 0 d
dydxd
where Srem-j = Sj′j + Sjj-1 + Sj-1j′-1 (Sj′j, Sjj-1 and Sj-1j′-1 are the upper
B
R2 e 2 R2 x 2
+ h
(z )|z = z 6 (cos + khE sin )(rm + x c3 ) 2 dydxd surface, slope surface and bottom surface of soil layer j, respectively, as
C 0 e
shown in Fig. 7b).
(12) The dissipation work rate Dc''3D caused by the capillary cohesion in
whole soil block ABCD can be divided into two parts (i.e., the energy
where khE is the effective horizontal seismic coefficient (typically,
dissipated in soil block ABCG, Dc''ABCG - 3D , and that dissipated in soil
0.1 ≤ khE ≤ 0.3) [26], (z )|z = z1, (z )|z = z2 and (z )|z = z3 capture the ef-
block CDG, Dc''CDG - 3D), which are given, respectively, as
fective unit weights corresponding to the center points of the areas
¯ , m¯n and m ¯n , respectively, as
represented by the line segment mn
shown in Fig. 6(a); (z )|z = z 4 , (z )|z = z5 and (z )|z = z 6 represent the ef-
Dc''3D - ABCG = cot { 0
B
c (z )|z = z7 vi ni dS +
B
C
c (z )|z = z 8 vi ni dS
and (6) by substituting the variable z with the variables z1, z2, z3, z4, z5
and z6, which are given in Appendix. The variables a, d, e, R, rm, θB and
+
sin2 ( + C) sin2 h 2( h
sin2 C
e 0 ) tan C
B
cos( + )
s (z )|z = z 8 sin3 ( + ) R2 d2 d }
m1 - 1
θC can be easily derived from the geometrical relationships as shown in 2 cot r j20 (c j c j + 1)sin
2
j
j cos
R2 a2j d
sin3
Fig. 4, which are available in the references (e.g., [29]). j =1
j
C
c (z )|z = z 0 cot vi ni dS
4.3. A LSM to calculate the internal energy dissipation rate
G
(17a)
Because of the highly variable nature of the capillary cohesion and
profile (Fig. 3b), a LSM is developed to calculate the internal energy m2
dissipation rate based on the divergence theorem. To make the calcu- Dc''3D - CDG = c k cot { Sk k
vi ni dS + v n dS
Sk 1k 1 i i }
lation manageable, the soil mass is divided into two parts, i.e., soil k=1
cos
block ABCG (above the slope toe elevation) and soil block CDG (below =2 r02sin2 h e 2( h 0) tan h
s (z )|z = z 0 sin3 R2 e2 d
C
the slope toe elevation) as shown in Fig. 7. Soil blocks ABCG and CDG m2 - 1
k cos
are divided into m1 and m2 discrete layers, respectively. The thickness 2 rk20 (c k c k + 1) cot sin2 k sin3
R2 ek2 d
k
k=1
of the soil layer in each part are h1 = H/m1 and h2 = N/m2. The value C
of m2 can be approximately determined according to the relationship + c (z )|z = z 0 cot vi ni dS (17b)
G
between N and H once the value of m1 is given. The relationship be-
where s (z )|z = z7 , s (z )|z = z 8 and s (z )|z = z 0 describe the suction stress at
tween N and H takes the following form
slope crest AB, slope surface BC and ground surface CD, respectively,
N cos e( 2 + 0) tan
sin which can be easily got by substituting the variable z in Eq. (4) with the
0
= ( 1 variables z7, z8 and z0; and r j0 and rk0 are shown in Fig. 7. The ex-
H e h 0 ) tan
sin h sin 0 (13)
pressions of the variables z7, z8, z0, aj, θj′, θj, ek, θk′ and θk are given in
The capillary cohesion of the soil layer in each part (e.g., layer j and Appendix. Note that, the last terms on the right-hand side in Eqs. (17a)
k) can be approximately written, respectively, as and (17b) are opposite in sign and equal in magnitude.
301
L. Wang, et al. Computers and Geotechnics 110 (2019) 296–307
Fig. 7. LSM to calculate the energy dissipation rate of a gentle unsaturated soil slope: (a) notation and convention for LSM; (b) 3D diagram of Layer j; and (c) 3D
diagram of Layer k.
For the 2D plane insert, the dissipation work rate Dc''2D can be derived factor, which is extremely time-consuming to solve under 3D condi-
as follows: tions. Conversely, the GIM has an advantage in giving an explicit ex-
h
pression of the safety factor, which may be preferred by geotechnical
Dc''2D = br02 tan s (z )|z = z9 e
2( 0) tan
d engineers. Thus, the GIM is adopted in this paper. For the gentle un-
0 (18)
saturated soil slope with a given height, the safety factor Fs can be
where s (z )|z = z9 describes the suction stress at a generic point on the slip expressed as
surface, which can be easily got by substituting the variable z in Eq. (4)
Dc' + Dc''
with the variable z9. The expression of the variable z9 is given in Ap- Fs =
pendix.
W + Ws (20)
For the special case of purely cohesive soils (ϕ′ = 0° and c′ > 0), the The fundamental inequality of the kinematical approach of limit
expression of Dc''3D is given as analysis theorem implies that the safety factor should not less than 1.0
(i.e., the internal energy dissipated rate Dc' + Dc'' is not less than the
Dc''3D = 0 (19)
external power rate W + Ws ) for a slope to be safe. When Fs < 1.0, the
In geotechnical engineering practice, the strength reduction method soil starts yielding and displacements of the sliding soil mass may take
(SRM) (e.g., [48]) and the gravity increase method (GIM) (e.g., [34,30]) place, which means the slope becomes unstable.
are widely used for slope safety estimation. The above two definitions The basic plasticity theorems of limit analysis can be cast as opti-
of the safety factor are valid although not necessarily the same [36]. mization problems. An optimization procedure is developed with in-
Nevertheless, when applied to the kinematical approach of limit ana- dependent variables θ0, θh, r′0/r0 and β′ to find the global minimum
lysis theorem, the SRM gives an implicit equation about the safety value of the safety factor. According to the 3D rotational failure
302
L. Wang, et al. Computers and Geotechnics 110 (2019) 296–307
4 1.6
Deviation of the safety factor, Dm1 (%)
() ()
Deviation of the safety factor, Dm1 (%)
0 20
5 25
3 10 1.2 30
15 35
2 0.8
1 0.4
0 0.0
5 10 50 100 5 10 50 100
Layer number, m1 Layer number, m1
303
L. Wang, et al. Computers and Geotechnics 110 (2019) 296–307
Table 3
Safety factors obtained using LSM for clayey soil slopes under no-flow condition (c′ = 20 kPa, z0 = 0.5H, q = 0 m/s, β = 30°, H = 10 m, B/H = 2.0 and khE = 0.0).
m1 ϕ′ = 0° ϕ′ = 5° ϕ′ = 10° ϕ′ = 15°
continuous and smooth transition for a soil going from an unsaturated 1.2
state to a saturated state indicates to some extent the validity of the
6. Examples
0.3 10
Parametric analyses about the effects of soil type and unsaturated
flow condition on the slope stability have already been studied by many 20
researchers (e.g., [20,40]). In this section, only examples with real soil 0.0 100
parameters are given to demonstrate the 3D effect and the seismic effect 2 4 6 8 10 12 14 16 18 20
on the stability of unsaturated soil slopes. The parameters for different
Slope height, H (m)
soils along wetting paths are given in Table 1.
Fig. 11 shows the change in the seismic stability of unsaturated soil Fig. 9. Deviation versus slope height relation for gentle unsaturated clayey soil
slopes with different constrains of the width-to-height ratio B/H for slopes (β = 30°, B/H = 2.0, ϕ′ = 5°, z0 = 0.5H, q = 0 m/s and khE = 0).
each soil. The parameters H = 10 m, β = 45°, z0 = 0.5H, q = 0 m/s and
m1 = 40 are used. For comparison, the cases considering and ignoring no suction case. When the value of B/H is increased to 10, the corre-
the suction are plotted with “WS” and “NS” in the figure, respectively. sponding safety factors are 2.420 and 1.317, respectively, which both
As expected, the slope stability generally decreases as the seismic in- approach those of the 2D cases (the safety factors are 2.322 and 1.267,
tensity increases. Meanwhile, the 2D stability approach gives a con- respectively). For Soil B (Fig. 11b), the 3D effect and the suction-in-
servative estimation of the slope safety. duced effect on the slope stability are similar with Soil A. However, the
To make a realistic estimation of the slope safety, the 3D effect and safety factors for no suction cases are almost less than 1.0, which are
the suction-induced effect are essential to be considered in the slope much smaller than those considering the suction-induced effect on the
stability analyses. The 3D effect of unsaturated soil slopes is nearly slope stability. The calculated results indicate that ignoring the 3D ef-
identical for different soils and is similar to that of no suction cases fect and the suction-induced effect on the slope stability may lead to
[27]. Conversely, the suction-induced effect varies with soils, is most errors.
pronounced for Soil A and Soil C and decreases in Soil B and Soil D, as
shown in Fig. 11. For Soil A with B/H = 2.0 and khE = 0.3 (Fig. 11a),
the safety factor is 2.933 when the suction is considered but is 1.578 for
Table 4
Safety factors obtained using LSM for clayey soil slopes under no-flow condition (c′ = 20 kPa, z0 = 0.5H, q = 0 m/s, β = 30°, B/H = 2.0, H = 10 m, ϕ′ = 5° and
khE = 0.0).
m1 β = 20° β = 25° β = 30° β = 35°
304
L. Wang, et al. Computers and Geotechnics 110 (2019) 296–307
1.8
q/ks
approach of limit analysis theorem. For this, the extended Mohr-
Coulomb failure criterion was introduced to describe the dependency of
0
1.6 - 0.2 shear strength in unsaturated soils on matric suction. Meanwhile, the
- 0.4 3D rotational failure mechanism [27] used for no suction case is ex-
- 0.6 tended to an unsaturated condition. The widely used pseudo-static
Safety factor, Fs
1.4
- 0.8 approach was adopted and the external work rate done by the seismic
- 1.0 (NS) load was incorporated in the energy balance equation to account for the
1.2
effect of seismic excitation on the slope stability. To deal with the
highly variable nature of suction, a semi-analytical method (including
1.0 the SM and the LSM) is developed. By comparing with the existing 3D
solutions, the validity of the analytical framework is demonstrated.
0.8 Based on the calculated results, it can be found that the LSM is ef-
fective in assessing the stability of gentle unsaturated soil slopes and the
0.6 slope height has insignificant effect on the application of the LSM. For a
0.00 0.05 0.10 0.15 0.20 0.25 0.30 gentle soil slope, the LSM is reasonable and effective (Dm1 < 0.15%)
Effective horizontal seismic coefficient, khE when the soil mass is divided into 20 layers or more in practice.
Accounting for the 3D effect and the suction-induced effect can both
Fig. 10. Stability of gentle unsaturated clayey soil slopes under different in- lead to more realistic results. The 3D effect is nearly identical for dif-
filtration conditions (H = 10 m, β = 30°, B/H = 2.0, z0 = 0.5H and m1 = 40) ferent soils while the suction-induced effect varies with soil type. When
(NS represents no suction case). subjected to seismic load, the slope stability is significantly reduced as
the seismic intensity increases. The described framework allows the
7. Conclusions user to quantify the effect of changes in moisture content on effective
stress distribution and effective unit weight distribution in the slope and
This paper develops an analytical framework for evaluating the 3D thus, changes in the slope stability.
seismic stability of gentle unsaturated soil slopes using the kinematical
6 6
WS NS
B/H B/H
5 5 1.0 1.0
2.0 2.0
5.0 5.0
Safety factor, Fs
Safety factor,Fs
4 4 10.0 10.0
2D 2D
3 3
2 2
1 1
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.05 0.10 0.15 0.20 0.25 0.30
Effective horizontal seismic coefficient, khE Effective horizontal seismic coefficient, khE
(a) Soil A (b) Soil B
6 6
5 5
Safety factor, Fs
Safety factor,Fs
4 4
3 3
2 2
1 1
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.05 0.10 0.15 0.20 0.25 0.30
Effective horizontal seismic coefficient, khE Effective horizontal seismic coefficient , khE
(c) Soil C (d) Soil D
Fig. 11. Safety factor versus horizontal seismic coefficient relation for gentle unsaturated soil slopes (H = 10 m, β = 45°, z0 = 0.5H, q = 0 m/s and m1 = 40): (a) Soil
A; (b) Soil B; (c) Soil C; and (d) Soil D (WS and NS represent with suction and no suction cases, respectively).
305
L. Wang, et al. Computers and Geotechnics 110 (2019) 296–307
Appendix
The variables z1, z2 and z3 in Eq. (11) are the vertical distances from the middle point of line segments mn¯ , m¯n and m ¯n to the slope toe
elevation, respectively, which can be derived from the geometric relationship in Fig. 6(a) and are given as follows:
1
z1 = z 0 + r0 {2e ( h 0 ) tan
sin e( 0 ) tan
sin sin 0}
2
h
(A1)
1 sin( h + )
z2 = z 0 + r0 {e ( h 0) tan
sin h e( 0 ) tan
sin + e( h 0 ) tan
sin h sin
2 sin( + ) (A2)
1
z3 = z 0 + r0 {e ( h 0) tan
sin e( 0 ) tan
sin }
2
h
(A3)
The variables xc1, xc2 and xc3 in Eq. (12) are given as
4R [1 (a R)2]3 2
x c1 =
3[2 arccos(a R) 2(a R) 1 (a R ) 2 ] (A4)
4R [1 (d R)2]3 2
x c2 =
3[2 arccos(d R) 2(d R) 1 (d R ) 2 ] (A5)
4R [1 (e R)2]3 2
x c3 =
3[2 arccos(e R) 2(e R) 1 (e R ) 2 ] (A6)
The variables z4, z5 and z6 in Eq. (12) are the vertical distances from the center points of areas A(θ) to the slope toe elevation, respectively, which
can be derived from the geometric relationship in Fig. 6(b) and are given as follows:
z7 z9
z 4 = z7 (x c1 a)
R a (A7)
z8 z9
z5 = z 8 (x c2 d)
R d (A8)
z0 z9
z6 = z0 (x c3 e)
R e (A9)
where z7 is the slope height, z8 is the vertical distance from a generic point (point n′) on the slope surface to the slope toe elevation, and z9 is the
vertical distance from a generic point (point m, m′ or m′′) on the slip surface to the slope toe elevation. The variables z7, z8 and z9 can be derived from
the geometric relationship in Fig. 6(a) and are given as follows:
z7 = z 0 + r0 e ( h 0 ) tan
sin h r0 sin 0 (A10)
sin( h + )
z 8 = z 0 + r0 e ( h 0 ) tan
sin h r0 e ( h 0 ) tan
sin
sin( + ) (A11)
z 9 = z 0 + r0 e ( h 0 ) tan
sin h r0 e ( 0 ) tan
sin (A12)
Expressions of the variables aj, θj′, θj, ek, θk′ and θk can be derived from the geometric relationship in Fig. 7, which are given as follows:
r0 sin 0 + (1 j m1 ) H
aj = rm
sin (A13)
r0 e ( j 0 ) tan
sin j (1 j m1 ) H = r0 sin 0 (A14)
sin j sin
j = arctan
cos j sin [sin( j + ) e( h j ) tan sin( h + )] (A15)
r0 sin 0 + H + (n k ) h2
ek = rm
sin (A16)
r0 e ( k 0 ) tan
sin k = r0 sin 0 + [H + (n k ) h2] (A17)
r0 e ( k 0 ) tan
sin k = r0 e ( h 0) tan
sin h + (n k ) h2 (A18)
306
L. Wang, et al. Computers and Geotechnics 110 (2019) 296–307
307