Finite Elements in Analysis and Design 58 (2012) 10–19

Contents lists available at SciVerse ScienceDirect

Finite Elements in Analysis and Design

journal homepage: www.elsevier.com/locate/finel

Stability analysis of soil slope based on deformation reinforcement theory

Yaoru Liu n, Chuanqi Wang, Qiang Yang
State Key Laboratory of Hydroscience and Hydraulic Engineering, Tsinghua University, Beijing 100084, China

a r t i c l e i n f o a b s t r a c t

Article history: The traditional rigid-body limit equilibrium method cannot provide actual stress distribution in slope
Received 20 October 2009 stability analysis, neither finite-element-based strength reduction method could accurately identify
Received in revised form convergence points in related nonlinear calculation. Deformation Reinforcement Theory (DRT) is
13 February 2012
elaborated and developed with a definition of instability that an elasto-plastic structure is not stable
Accepted 15 April 2012
Available online 9 May 2012
if it is unable to satisfy simultaneously equilibrium condition, kinematical admissibility and constitu-
tive equations under given external loads. It stated that Plastic Complementary Energy (PCE) and
Keywords: unbalanced forces can be used to evaluate the stability of the slope and indicate reinforcement if
Stability analysis required. FEM expression of DRT was deduced and implemented in TFINE, a nonlinear three
Slope
dimensional finite element method program, and furthermore, successfully applied in slope stability
Plastic complementary energy
computation and analysis. The results of classic case studies showed that many key issues in slope
Unbalanced forces
Nonlinear FEM analysis could be well solved within the framework of the DRT. The stability of the slope can be
determined by the curve of the PCE norm vs. Strength Reduction Factor (SRF). The unbalanced-force
derived by elasto-plastic FEM analysis could be the basis of reinforcement design and evaluation.
& 2012 Elsevier B.V. All rights reserved.

1. Introduction which could be described as follows:

c
Stability analysis is the precondition of the reinforcement c0 ¼ ð1Þ
FOS
design and engineering practice for the slope. It tells whether  
the slope is stable or not, and if not, how much reinforcement is tan f
f0 ¼ arctan ð2Þ
required, and in what manner. This theoretical and engineering FOS
issue has not been perfectly solved yet. Because not only the on- where c and f refer to the effective friction angle and cohesion of
site geologic background and slope property is usually hard to the soil. FOS is the factor of safety.
investigate, but also a reasonable and powerful analysis approach The elasto-plastic SSR-FE approach offers a number of signifi-
is pending to develop. cant advantages over traditional limit equilibrium method. It
The traditional limit equilibrium method [1–4] is still widely eliminates the necessity for pre-assumptions on failure modes,
used so far due to its relative simplicity and the experience which instead automatically provides the critical failure mechanism
geotechnical engineers have accumulated over years. However, its (Zienkiewicz et al., 1975, Matsui and San 1992, Griffiths and Lane
fundamental assumption is that the slope should be divided into 1999, Zhao Shangyi 2002) [5–8].
slides, what worse is, the mechanism, shape, and location of the Historically both the traditional limit equilibrium and SSR-FE
failure surface was artificially pre-assumed. Besides, it neglects analysis contribute a lot in the slope stability analysis. However,
stress–strain behavior of the material. neither of them could deal with the failure phenomena after local
From 1970s, FEM has been applied comprehensively in the failure occurs (the non-linear iteration has stopped before that
stability analysis of the slope in accordance with the rapid during the FEM calculation). In other words, no guideline or
development of computer technology. It could accommodate suggestion was proposed for the reinforcement to satisfy given
practically all kinds of geometry, and simulate various properties FOS, which is practically essential in engineering. Besides, a large
of material behavior, while yielding realistic results. In 1992, number of traditional calculations indicate that the variation of
Matsui and San [5] firstly proposed the Shear Strength Reduction displacement, stress and yielding zone was not well observed by
(SSR) technique based on the Mohr–Coulomb (M–C) criteria, reinforcement measures, which led to the wrong conclusion that
the reinforcement measures were not effective. In order to clarify
the paradox, Yang et al. [9–12] proposed the Deformation Rein-
n
Corresponding author. Tel.: þ86 1062796954. forcement Theory (DRT) for the perfectly elasto-plastic structures
E-mail address: [email protected] (Y. Liu). with small deformation similar to limit analysis. DRT mainly

0168-874X/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.finel.2012.04.003
 Y. Liu et al. / Finite Elements in Analysis and Design 58 (2012) 10–19 11

studies the structural behavior after the external loads exceed the Definition 2, sometimes it is not easy to find out the ‘sharp break’
limit loads which the structure could undertake. The post-failure point immediately when the slope failed, especially when the
behaviors are essential for determination of the reinforcement displacement curve is comparatively smooth. For Definition 3, the
measures, however were not discussed in detail for the classic powerful post-processing software is required for the defined
limit analysis theory. variables plots, which may also influenced by artificial factor or
In this paper, a solid theoretic basis for DRT is illustrated, and a not compatible with the current yield surface in a fair accuracy.
clear definition of instability and FEM expression of DRT gets Besides, the spread of yield is affected by Poisson’s ratio u. When u
established. Then DRT is applied in slope stability and reinforce- is small, the spread area of yield is usually very large, which is an
ment analysis. Furthermore, a comparison with some classic SSR- unfavorable factor to the contour plots.
FE example models has been performed.

3. Deformation reinforcement theory and its FEM expression

2. Determinations of the factor of safety (FOS) for slope
stability and slope failure 3.1. Increment form of elasto-plastic constitutive relation

There were three determinations of factor of safety in slope The linear elasticity can be described in total form, r ¼ D:e or
stability analysis so far. e ¼C:r, where r and e are the second-order stress and strain
tensors, respectively, D and C is the fourth-order elasticity and
(1) The ratio of the shear strength of soil over the shear stress compliance tensors, respectively. The elasto-plastic constitutive
required for equilibrium. (Duncan, 1996) [13]. equation is given in differential form [17]
(2) By how much the original shear strength parameters must be
reduced to bring the slope to the point of failure. (Matsui and dr ¼ D : ðdedep Þ ð3Þ
San, 1992) [5]. p
where e is the plastic strain tensor.
(3) The ratio of the resistant shear force over the driving shear Following consistency condition and associated normality flow
force along a certain slip surface. rule, an important equation is obtained for loading state
dep : dr ¼ 0 ð4Þ
The 2nd method was derived from the SSR, and was the same
as the first one actually. Zheng Hong et al. compared those Since dep  0 in unloading and elastic states, Eq. (4) is valid for
determinations of FOS and pointed out that both the FOS and all the states.
the potential slip surface based on the 3rd method are different DRT is based on incremental elasto-plastic constitutive equa-
from those on the 2nd method. Some irrational results might be tions, so the above differential constitutive equations should be
obtained when the 3rd method is used in FE slope stability recast in incremental form. A typical incremental loading process
analysis [14]. is illustrated in Fig. 1. The initial stress state r0 is stable, i.e., f
For slope failure identification in SSR-FE analysis, there were (r0) r0.
three widely used definitions. By imposing a strain increment De on initial state, the elastic
loading stress state is r1 ¼ r0 þ Dre where Dre ¼D:De. The strain
(1) The non-convergence of FEM (Griffiths and Dawson, 1999) increment De results in plastic loading if
When the algorithm does not converge within a user-speci-
f ðr1 Þ 40 ð5Þ
fied maximum number of iteration (e.g. 1000 times), the
implication is that no stress distribution can be found able otherwise the material response is elastic. For a loading process,
to simultaneously satisfy equilibrium condition, kinematical the final stress state is r ¼ r1  Drp where f (r)¼0. If f (r1)o0,
admissibility and constitutive equations. Under this circum- then r ¼ r1. Eq. (3) can be rewritten in incremental form:
stance, ‘failure’ is said to have occurred [7,15,17].
Dr ¼ D : ðDeDep Þ, ð6Þ
(2) The jump of displacement (Griffith, 1999)
Perceptually when the slope collapsed, the slide began to The associated normality flow rule can be written in incre-
move and glided along the slip surface. Usually the displace- mental form:
ment of the slide has a dramatic ‘‘jump’’ at the same time,
@f
especially for the upper area. Some researchers picked the Dep ¼ Dl : ð7Þ
@r
maximum nodal displacement dmax or dimensionless displa-
cement Edmax =gH2 as the indicator of the slope stability (E is Eq. (6) holds exactly. According to mean value theorems for
the deformation modulus, g is the unit weight and H is the integration, Eq. (7) can also hold exactly if @f/qs is determined by
height of the slope). When they have dramatic increase, the a proper stress state from r^ to s. In this paper, @f/@s is determined
slope was said to fail [7]. by the final stress state r, so the incremental normality flow rule
(3) The transfixion of spread of certain variables within the slope is approximate. Due to Drp ¼ r1  r, the plastic strain increment
(Luan Maotian, 2003).

The variations and distributions of some physical variables,

such as generalized shear strain or equivalent plastic strain,
showed the location and shape of the potential failure surface. If
the distribution of those variables with a given threshold is
progressed from the toe to the top, the slope will approach a
critical instable condition [16].
The disadvantages of the above definitions are also obvious.
For Definition 1, the evaluation criteria by pre-assuming an
allowable iteration number, in some cases, was influenced by
artificial factor or led by the deficiency of the software itself. For Fig. 1. Elasto-plastic stress adjustment.
12 Y. Liu et al. / Finite Elements in Analysis and Design 58 (2012) 10–19

can be expressed as yield conditions under the prescribed loading, so the structure is
Dep ¼ C : Drp ¼ C : ðr1 rÞ: ð8Þ stable. Therefore, structural stability can be defined based on
whether or not structural solutions exist. This principle also be used
Substituting formula (8) into normality flow rule (7), and implicitly to find the point of impending instability [7,15,17].
under premise that r is in the yielding surface, the final stress Let us elaborate the definition based on set operations.
state can be determined as Considering a structure whose volume is V, assumes that u and

@f  up are compatible displacement-field and residual plastic displa-
C : ðr1 rÞ ¼ Dl  , f ðrÞ ¼ 0: ð9Þ cement-field corresponds to a strain field e and plastic strain field
@r r
ep. A stress field can be obtained by r ¼ D:(e  ep), which is called
If @f/@r is determined by the final stress state, it is easy to show the compatible stress field set denoted by Sk. Two stress fields can
that Eq. (4) can be extended for incremental loading and elastic be defined: the equilibrium stress-field set S1 of which each
processes, element satisfies equilibrium condition throughout the structure
dr : Dep ¼ 0 ð10Þ and the stable stress-field set S of which each element satisfies
yield criterions throughout the structure.
Drucker–Prager yield criterion is used in this paper:
qffiffiffiffiffiffiffiffiffiffiffi Considering two arbitrary stress fields, r1(r1AS1) and r(rAS),
f ¼ aI1 þ J2 k r0 ð11Þ their difference is the plastic-stress increment field Drp
Drp ¼ r1 r ð15Þ
where
h i which leads to the plastic-strain increment field Dep ¼C:Drp.
I1 ¼ s1 þ s2 þ s3 ; J 2 ¼ 16 ðs1 s2 Þ2 þ ðs2 s3 Þ2 þ ðs3 s1 Þ2 ð12Þ Structural stability can be defined based on the intersection of
S1 and S: a structure is stable if S1\Sa / (/ is null set), and the
s1, s2, s3 are the principal stresses, a and k are the translation
structure is instable if S1\S¼ f, as shown in Fig. 2.
parameters between Mohr–Coulomb and Drucker–Prager criter-
A Euclidean space for stress field can be established. In this
ion. For equivalent area circle,
pffiffiffi pffiffiffi space, one point presents one stress field. Take C/2 as measuring
2 3sin f 6 3ccos f tensor, and the distance L between the two stress fields, r1 and r,
a ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffi , k ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffi ð13Þ
can be defined as
2 3pð9sin2 fÞ 2 3pð9sin2 fÞ
Z Z
2 1
f and c refer to the effective friction angle and cohesion of L2 ¼ l dV ¼ ðr1 rÞ : C : ðr1 rÞdV: ð16Þ
V 2 V
the soil.
The final stress state can be obtained by where, V presents the structure.
If L¼0, the two stress fields r1 and r are the same, i.e., r1 ¼ r
1
r ¼ sij ¼ ð1nÞs ij þ pdij ð14aÞ anywhere in V. The distance between S1 and S is denoted by L: ^
where, L^ ¼ minL, ðr1 A S 1 , r A SÞ: ð17Þ
9 s1 , s
wm
n¼ pffiffiffi , p ¼ mw þ 3nI1 =
J2
ð14bÞ Therefore, the distance L^ can be taken as an index to evaluate
m ¼ að3l þ 2mÞ, w¼ f ; the stability of structure:
3am þ m

in which, J2, I1 and f are all determined by r1, and l and m are (a) If L^ ¼ 0, a stress field satisfying equilibrium condition and
Lame constants. yield criterions throughout the structure exists and the
As from the aspect of the constitutive relation integration structure is stable.
strategy, the above method is equivalent to closest point projec- ^
(b) If L40, a stress field satisfying equilibrium condition and
tion algorithm [18]. The closest point projection algorithm has yield criterions throughout the structure does not exist and
first-order precision and is unconditionally stable, which is good the structure is instable. The more L^ is, the more unstable the
for problem with big strain increment. structure is.

3.2. Definition of instability of elasto-plastic structures The physical meaning of L2 is complementary energy, there-
fore it can be regarded as plastic complementary energy (PCE) DE:
Solutions of elastic structures always exist and are unique, but Z Z
1 1
solutions of elasto-plastic structures may not exist. The existence of DE ¼ L2 ¼ Drp : C : Drp dV ¼ ðr1 rÞ : C : ðr1 rÞdV: ð18Þ
2 V 2 V
structural solutions implies that the displacement and stress fields
throughout the structure satisfy simultaneously equilibrium condi- This equation shows that DE is also the norm of plastic-stress
tion, kinematical admissibility and constitutive equations including increment field DrP. The shortest distance can be defined as

Fig. 2. Sketch map of elasto-plastic solution in vector space.

 Y. Liu et al. / Finite Elements in Analysis and Design 58 (2012) 10–19 13

2
DEmin ¼ L^ : If DEmin ¼0, then Drp is always zero and the structure Eq. (21) can be understood as
is stable. If DEmin 40, the structure is unstable.
structural selfbearing force ¼ external loads

3.3. FEM expression of deformation reinforcement theory þ reinforcement force ð23Þ

or
Assume that the body-force field is f¼fi. Its stress field is
r ¼ sij, which satisfies the equilibrium equations, sij, j þfi ¼0, and structural selfbearing force þunbalance force
constitutive equations including yield condition f (r) r0. The ¼ external loads ð24Þ
displacement field u¼ui is kinematically admissible and leads to a
strain field e ¼ eij ¼(ui, j þuj, i)/2. The total boundary is Su þSs, The above two expressions are equivalent, but the meanings
where Su is the displacement boundary with ui ¼ ui ; Ss is the are different. The former shows that the reinforcement force is
stress boundary with T ¼Ti ¼ sijnj. the external force, and the latter shows that the unbalanced force
For an equilibrium stress field r1 and arbitrary virtual dis- is the internal force. In fact, the unstable state of a structure can
placement du, the principle of virtual displacements reads be viewed as constrained equilibrium states in thermodynamics
Z Z Z with internal variables [19–22]. DQ can be taken as opposite
deij s1ij dV ¼ dui f i dV þ dui T i dS ð19Þ conjugate force to internal variables.
V V Ss
Plastic stress field Drp is determined by displacement field
The governing equation of FEM can be deduced from Eq. (19) which is satisfied with deformation compatibility condition. So
after the structure is discretized. reinforcement force calculated with  Drp is self-equilibrium
X Z T
X Z T
X Z force field. The general reinforcement force is self-equilibrium
e
B r1 dV ¼ e
N f dV þ e
N T TdS ð20Þ indeed, such as anchorage force. The distribution of anchorage
Ve Ve Ser
force is complicated along anchor cable, but anchor cable is
Considering r1 ¼ r þ Drp, we can get equilibrium, therefore anchor cable is self-equilibrium (the grav-
X Z X Z ity of anchor cable is neglected herein). According to self-equili-
e
BT rdV ¼ F e BT Drp dV ¼ FDQ ð21Þ brium character and Saint-Venant’s principle, reinforcement force
Ve Ve
has little effect to deformation and stress for stable structure.
where, N is the matrix for shape function, B is the strain matrix, F
is the equivalent nodal force vector of external loads. DQ is the
equivalent nodal force of the plastic stress field Drp which is 3.4. Implement in nonlinear FEM calculation
called unbalanced forces in FEM:
X Z In general elasto-plastic FEM iterative calculation, from an
DQ ¼ e
BT Drp dV ð22Þ arbitrary compatibility stability stress field rð1Þ 1 , the nearest
Ve
compatibility equilibrium stress field rð2Þ 1 can be calculated, then
As mentioned above, the equivalent nodal force of reinforce- the nearest compatibility stability stress field r(2), etc. And finally
ment force is  DQ, that is to say, for a specified deformation the nearest two stress fields r and r1 of the structure have been
state, reinforcement force has the same size but opposite direc- obtained, as shown in Fig. 2. This calculating strategy is just the
tion with unbalanced force. Therefore at the level of nodal force, solving method of elasto-plastic FEM analysis based on displace-
ment method factually. Therefore, the general elasto-plastic FEM
analysis can be directly used to solve the problem for instable
structure. The residual unbalanced force which cannot be elimi-
nated after complete iteration is the minimum reinforcement
PCE
(K, ΔE) force needed by structure stability, and the final compatibility
stability stress field is the real stress field r of instable structure.
Calculating convergent criterion is that the complementary
energy norm DE(n) tends to stable value DE0 (or little change) in
iteration process:

DEðnÞ -DE0 Z0 ðn b 1Þ ð25Þ

O K0 SRF
0
Ideally, DE should be the minimum complementary energy
Fig. 3. Strength reduce factor to complementary energy. norm DEmin (DEmin ¼ L2 , referred to Eq. (18)). Since the effect of

Fig. 4. Example 1: The FE model of the slope.

14 Y. Liu et al. / Finite Elements in Analysis and Design 58 (2012) 10–19

calculating error and algorithm, DE0 is more than DEmin in general 4. Stability and reinforcement of slope based on DRT

DE0 Z DEmin ð26Þ It is not an easy task for traditional SSR-FE to design and
evaluate reinforcement measures because most of the calculation
Eq. (26) shows reinforcement force calculated by elasto-plastic results did not reflect the actual effects of those measures: the
FEM is always upper bound of real solution, or calculating result is displacement and stress distribution influence caused by reinfor-
always conservative. This conclusion is meaningful. The details of cement was not obviously observed. However in engineering
elasto-plastic FEM calculation, such as mesh partition, loading practice, the retaining walls and anti-slide piles were proved to
step size, integral strategy, iteration times and so on, have be very effective for slope reinforcement.
influence to convergence. So DE0 40 may appear for stable According to DRT, an instable structure could not satisfy
structure. For DE0 40 in calculation, there are three cases: simultaneously equilibrium condition, kinematical admissibility
and constitutive equations under given external loads, which
(a) structure is stable; means that there are un-balanced forces in some regions of the
(b) structure is instable, DE0 ¼ DEmin ; structure at this time. Those regions should be reinforced in order
(c) structure is instable, DE0 4 DEmin . to rectify the breach. Furthermore, the magnitude of reinforce-
ment force is the same as un-balanced force with opposite
direction. This principle indicates that an instable structure
But anyway imposing reinforcement force system  DQ which always tries to maximize its self-bearing force and minimize
is corresponded to DE0 on structure can maintain the stability of reinforcement force under given external loads. The un-balanced
structure. force gives the most efficient reinforcement in order to maintain
In practice, the calculation is through numerical iteration the structure stable with given FOS. It is beyond the capacity of
process, during which the plastic deformation will be accumu- traditional SSR-FE analysis.
lated after failure. The unbalanced force is corresponding to the The structure’s global stability could be described by PCE (vs.
residual plastic deformation in excess of yield. In other words, the SRF or overloading) as shown in Fig. 3. The discontinuity of PCE
structure could undertake more loads itself if more elasto-plastic norm corresponding to SRF (or overloading) represents the
deformation is accepted. Consequently the required reinforce- transition point between two fundamental mechanisms: stable
ment could be less under this circumstance. and instable. The PCE norm remained 0 when the structure is
stable and had sharp breaks when the structure fails. Compara-
tively it is much easier to get FOS (the sharp SRF break point, K0 in
Table 1 Fig. 3) through this process than the 2nd method [5] mentioned
Safety factor by different methods (g ¼ 20 kN/m3). above, which needs to determine the catastrophe point that
sometimes is not so clear. As a result, it is more reasonable and
Methods/Slope angle 301 351 401 451 501 easier to use PCE norm as the criteria of the global stability.
In fact, the unstable condition can also be considered as a
ANSYS 1.56 1.42 1.31 1.21 1.12
Z_Soil 1.56 1.42 1.31 1.22 1.13 generalized limit state, which can be characterized by a coordi-
Bishop 1.56 1.42 1.30 1.20 1.12 nate (K, DE) on the curve where K reflects the ultimate load and
Spencer 1.55 1.41 1.30 1.20 1.12 DE is the measurement of required reinforcement forces.
TFINE(DRT) 1.52 1.34 1.14 1.02 0.94

5. Examples
Table 2
Safety factor by different methods (g ¼ 25 kN/m3). The DRT for slope stability analysis has been implemented in
TFINE, a nonlinear three dimensional finite element method program,
Methods/Slopeangle 301 351 401 451 501 and successfully applied in slope stability computation and analysis.
ANSYS 1.39 1.27 1.16 1.060 1.00
Z_Soil 1.39 1.26 1.15 1.070 1.00 5.1. Material parameters of the soil model
Bishop 1.40 1.26 1.15 1.064 0.99
Spencer 1.39 1.26 1.15 1.063 0.99
In the following soil slope models, the unit weight g assigned
TFINE(DRT) 1.34 1.16 1.00 0.88 0.80
to the soil is the main external load.

0.004
30
0.0035 35
0.003 40
45
0.0025 50
PCE / (N*m)

0.002
0.0015
0.001
0.0005
0
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6
-0.0005
SRF

Fig. 5. The PCE curve of slope vs. SRF (g ¼20 kN/m3).

 Y. Liu et al. / Finite Elements in Analysis and Design 58 (2012) 10–19 15

Though Young’s modulus E and Poisson’s ratio n have a The dilation angle c affects the volume change of the soil
profound influence on the deformation calculation prior to failure, during yielding. Usually as the slope stability analysis is
they do not affect the factor of safety at all in slope stability relatively unconfined, the value of dilation angle is less
analysis. To simplify the research process, in this paper, they were important. The main objective of this paper’s study is to get the
given nominal values (e.g. E¼104 kN/m2, n ¼ 0.3). accurate prediction of slope factors of safety, so a compromise

0.006
30
0.005 35
40
0.004 45
50
PCE / (N*m)

0.003

0.002

0.001

0
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4
-0.001 SRF

Fig. 6. The PCE curve of slope vs. SRF (g ¼ 25 kN/m3).

Fig. 7. Unbalanced force distribution of homogeneous slope (b ¼ 501, f ¼ 171, c ¼42 kPa, g ¼25 kN/m3). (a) SRF¼ 0.86, (b) SRF ¼ 0.94, (c) SRF ¼1.
16 Y. Liu et al. / Finite Elements in Analysis and Design 58 (2012) 10–19

value of c ¼ f, corresponding to a associated perfectly flow rule, opposite direction. Besides, this mechanism of reinforcement is
has been adopted. the most efficient one because at this time the slope’s self-bearing
In summary, the most important parameters in a finite capacity was fully utilized.
element model for slope stability analysis are shear strength Fig. 7 showed an interesting reinforcement progress along
parameters f, c, and unit weight g. with SRF continues to increasing after failure. When the SRF
(4FOS) is low, it is only necessary to reinforce the slope’s toe. As
SRF increased, more reinforcement (both magnitude and distribu-
5.2. Example 1: homogeneous slope tion area) was required to maintain stability. The path and
development of un-balanced force was just the same as the
To illustrate the DRT application in slope stability analysis, a spread of yield before the slope collapsed observed by traditional
simple, dry homogeneous slope (Fig. 4) is induced, whose inclined SSR-FE analysis.
angle increased from 301 to 501 gradually with 51 increment in
the serial of FE models. The boundary conditions are given as
5.3. Example 2: an undrained clay slope with a thin weak layer
vertical rollers on the sides and full fixity at the base. Gravity load
is applied to the mesh.
For comparison, this slope model was taken from literature
The material parameters of the slope are as follows:
example 3 prepared by Griffiths and Lane (1999) [7]. Fig. 8 shows
E¼104 kN/m2, n ¼ 0.3, f ¼171, c ¼42 kPa, g ¼20 and 25 kN/m3.
a slope (f ¼01) containing a thin layer made of weaker material
Tables 1 and 2 list the FOS calculation results of several existing
and initially running parallel to the surface. The FOS of the slope
solutions, including Bishop and Spencer (based on limit equili-
was estimated by FE analysis for a serial of values of SRF by
brium methods), ANSYS and Z_Soil (based on conventional
changing the thin layer’s cu2, while keeping the strength of the
SSR-FE) [7]. TFINE’s result (based on DRT) is also provided for
surrounding soil as cu1 ¼ 100 kPa.
comparison.
The comparison clearly demonstrated that the calculation
results of TFINE agree well with those of limit equilibrium
analysis and traditional SSR-FEM solutions, although there are
slight differences. The result of TFINE was gained from the PCE
curve vs. SRF as shown in Figs. 5 and 6.
From Figs. 5 and 6 we could easily get the SRF sharp break
points (where SRF¼FOS) for different angle slope models. Those
points are the cut-off index of stable and instable condition of the
corresponding slopes.
Fig. 7 gives the initiation, development and redistribution of
un-balanced forces in the slope cross section throughout strength
reduction process (SRF4FOS). The un-balanced force firstly
appeared at the toe, which was the stress concentration area,
and then, extended along a ‘slip surface’ to the top with SRF
increasing.
According to DRT, when SRF is less than FOS, the slope was
stable and could maintain this state by its self-bearing force,
during which its PCE norm remained 0 and there is no external
reinforcement (un-balanced force) needed. Then along with SRF
increased to FOS, the slope’s self-bearing capacity was gradually
fully utilized and could not bear more loads. When SRF was above
FOS, the slope could not maintain stable by its self-bearing
capacity any longer (failed). At the same time, the PCE appeared,
together with un-balanced force. In other words, reinforcement
is required under this circumstance. Importantly the magnitude Fig. 9. Deformed meshes at failure for three different values of cu2/cu1. (a) cu2/
of reinforcement force is the same as un-balanced force with cu1 ¼1.0, (b) cu2/cu1 ¼ 0.6, (c) cu2/cu1 ¼ 0.2

Fig. 8. Example 2: Undrained clay slope including a thin layer.

 Y. Liu et al. / Finite Elements in Analysis and Design 58 (2012) 10–19 17

The initial calculation result is demonstrated in Fig. 9, which of cu2/cu1 ¼1.0. That is, if the thin layer is stronger than the
shows the deformed mesh at failure under three different scenar- surrounding soil, the factor of safety is a little larger than that of
ios by changing the ratio cu2/cu1. homogeneous slope. With the failure surface located below the thin
Figure (a) and (b), when cu2/cu1 ¼1.5 and cu2/cu1 ¼1, an layer and not crossing it, the thin layer has a little reinforcing
essential circular failure mechanism tangent to the firm base influence.
occurs.
Figure (d), when cu2/cu1 ¼0.2, a highly concentrated non- 5.4. Example 3: an undrained clay slope with a weak foundation
circular mechanism closely following the path of the thin weak layer
layer occurs.
Figure (c), when cu2/cu1 ¼0.6, considerable complexity and Like Example 2, another slope model (f ¼01) was taken from
ambiguity was observed. At least two conflicting mechanisms Griffiths and Lane’s (1999) [7] example 4. In this case, the same
are apparent at the same time. On the one hand, there is a base slope geometry as in the previous example has been adopted but
failure mechanism merging with the weak layer beyond the toe of with a different type of inhomogeneity, which is shown in Fig. 12.
the slope, on the other hand, there is a mechanism running along The shear strength of the slope upper area material has been kept
the weak layer parallel to the slope and outcropping at the toe. at a constant value of cu1/gH¼0.25, while adjusting the SRF of the
Fig. 10 showed the distributions and development of foundation layer. The relative shear strength of the two layers has
un-balanced force corresponding to different cu2/cu1, which recon- again been expressed as the ratio cu2/cu1.
ciled well with the failure mechanisms of the previous researches. The calculation result carried out by Griffiths and Lane is
PCE curve vs. SRF under different cu2/cu1 are shown in Fig. 11. For demonstrated in Fig. 13, which shows the deformed mesh at
an additional scenario, when cu2/cu1 ¼1.5, the failure is similar to that failure under three different values of the ratio cu2/cu1.

Fig. 10. Unbalanced force distribution with different values of cu2/cu1. (a) cu2/cu1 ¼ 1.5, SRF¼ 1.50, (b) cu2/cu1 ¼1.0, SRF¼ 1.45, (c) cu2/cu1 ¼0.6, SRF ¼1.35, (d) cu2/cu1 ¼ 0.2,
SRF ¼0.65.
18 Y. Liu et al. / Finite Elements in Analysis and Design 58 (2012) 10–19

4.5

3.5 Cu1/Cu2=0.2
Cu1/Cu2=0.5
3
Cu1/Cu2=1.0

PCE / (N*m)
2.5
Cu1/Cu2=1.5
2

1.5

0.5

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
SRF

Fig. 11. PCE curve with SRF under different cu2/cu1.

Fig. 12. Example 3: Undrained clay slope with a weak foundation layer (f ¼ 01,cu1 =gH ¼ 0:25).

Fig. 13. Deformed meshes at failure for three different values of cu2/cu1. (a) cu2/ Fig. 14. Un-balanced force for three different values of cu2/cu1. (a) cu2/cu1 ¼ 0.6,
cu1 ¼ 0.6, (b) cu2/cu1 ¼ 1.5, (c) cu2/cu1 ¼2. SRF¼ 1.35, (b) cu2/cu1 ¼1.5, SRF ¼1.65, (c) cu2/cu1 ¼ 2.0, SRF ¼2.1.

When cu2 5 cu1(a), a deep-seated base mechanism is observed,

whereas a shallow ‘toe’ mechanism is seen when cu2 b cu1(c). same time. The discontinuity corresponding to cu2/cu1 E1.5
Especially the scenario cu2 E1.5cu1(b) shows an ambiguous situa- obviously represents the transition between these two funda-
tion in which both failure mechanisms are trying to form at the mental mechanisms.
 Y. Liu et al. / Finite Elements in Analysis and Design 58 (2012) 10–19 19

Fig. 14 gives another picture of the distributions of un- [3] E. Spencer, A method of analysis of the stability of embankments assuming
balanced force under different cu2/cu1, which agreed well with parallel interslice forces, Geotechnique 17 (1) (1967) 11–26.
[4] Z.Y. Chen, X.G. Wang, J. Yang, et al., Rock Slope Stability Analysis: Theory
the failure mechanisms of the previous researches showed in Methods and Programs, China Water Power Press, Beijing, 2005 (in Chinese).
Fig. 13, except the transition point is SRF ¼1.65, a bit larger. [5] T. Matsui, K.C. San, Finite element slope stability analysis by shear strength
reduction technique, Soils Found. 32 (1) (1992) 59–70.
[6] O.C. Zienkiewicz, C. Humpheson, R.W. Lewis, Associated and non-associated
viscoplasticity and plasticity in soil mechanics, Geotechnique 25 (1975)
6. Conclusion 671–689.
[7] D.V. Griffiths, P.A. Lane, Slope stability analysis by finite elements, Geotech-
A clear definition of instability that an elasto-plastic structure nique 49 (3) (1999) 387–403.
[8] Zhao Shang-yi, Zheng Ying-ren, Shi Wei-ming, et al., Analysis on safety factor
is not stable if it cannot satisfy simultaneously equilibrium of slope by strength reduction FEM, Chin. J. Geotech. Eng. 24 (3) (2002)
condition, kinematical admissibility and constitutive equations 343–346. (in Chinese).
under given external loads, is introduced from the Deformation [9] Q. Yang, W.Y. Zhou, X. Chen, The principle of minimum complementary
energy and upper bound theorem in geotechnical reinforcement analysis,
Reinforcement Theory (DRT). Based on the definition, it is demon- Geomech. Geotech. Eng. 21st Century (2003) 158–166. (in Chinese).
strated that the PCE determines the stability condition of a [10] Q. Yang, X. Chen, W.Y. Zhou, et al., On unbalanced forces in 3D elastoplastic
complex structural system and can be used to evaluate the global finite element analysis, Geotech. Eng. 26 (3) (2004) 323–326. in Chinese.
[11] Q. Yang, X. Chen, W.Y. Zhou, Elastoplastic basis of geotechnical engineering
stability of the slope. Besides, unbalanced forces can be used to
reinforcement analysis, Rock Soil Mech. 26 (4) (2005) 553–557. (in Chinese).
evaluate the slope’s stability and reinforcement. [12] Q. Yang, L.J. Xue, R.K. Wang, et al., Reinforcement theory considering
It is demonstrated by comparison with classic case studies that deformation mechanism of rock mass and non-equlibrium eleastoplastic
slope stability and reinforcement design can be well solved mechanics, Chin. J. Rock Mech. Eng. 24 (20) (2005) 3704–3712. (in Chinese).
[13] J.M. Duncan, Limit equilibrium and finite element analysis of slopes, J.
within the framework of the DRT. The slope’s stability can be Geotech. Eng. 122 (7) (1996) 577–596.
determined by the curve of the PCE norm vs. Strength Reduction [14] Zheng Hong, Tian Bin, et al., On definitions of safety factor of slope stability
Factor (SRF). The progress and distribution of unbalanced force is analysis with finite element method, Chin. J. Rock Mech. Eng. 24 (13) (2005)
2225–2230. (in Chinese).
similar to the spread of yield, and could be the basis of analysis of [15] E.M. Dawson, W.H. Roth, A. Drescher, Slope stability analysis by strength
reinforcement design. reduction, Geotechnique 49 (6) (1999) 835–840.
[16] Luan Maotian, Wu Yajun, Nian Tingkai, A criterion for evaluating slope
stability based on development of plastic zone by shear strength reduction
FEM, J. Hydraul. Eng. 23 (3) (2003) 1–8.
Acknowledgments [17] J. Lubliner, Plasticity Theory, Macmillan Publishing Company, New York,
1990.
[18] M. Ortiz, E.P. Popov, Accuracy and stability of integration algorithms for
The work reported here was supported by the National Science
elastoplastic constitutive relations, Int. J. Numer. Methods Eng. 21 (1) . 561-1
Foundation of China with Grant no. 50823005 and no. 50709014 576.
and subsidized by National Basic Research Program of China with [19] J.R. Rice, Inelastic constitutive relations for solids: an integral variable theory
Grant no. 2009CB724604. and its application to metal plasticity, J. Mech. Phys. Solids 19 (1) (1971)
433–455.
[20] Q. Yang, X. Chen, W.Y. Zhou, Thermodynamic relationship between creep
References crack growth and creep deformation, J. Non-equilib. Thermodyn. 30 (1)
(2005) 81–94.
[21] Q. Yang, X. Chen, W.Y. Zhou, Multiscale thermodynamic significance of the
[1] A.W. Bishop, The use of the slip circle in the stability analysis of slopes, scale invariance approach in continuum inelasticity, J. Eng. Mater. Technol.
Geotechnique 5 (1) (1955) 7–17. 128 (4) (2006) 125–132.
[2] J.M. Duncan, State of the art: limit equilibrium and finite-element analysis of [22] Q. Yang, R.K. Wang, L.J. Xue, Normality structures with thermo-dynamic
slopes, J. Geotech. Eng. ASCE 122 (7) (1955) 577–596. equilibrium points, J. Appl. Mech. 74 (5) (2007) 965–971.