Computers and Geotechnics 92 (2017) 6876

Contents lists available at ScienceDirect

Computers and Geotechnics

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

Research Paper

Computation of the safety factor for slope stability using discontinuous

deformation analysis and the vector sum method
Xiaodong Fu a,, Qian Sheng a,b, Yonghui Zhang a, Jian Chen a, Shankai Zhang a, Zhenping Zhang a
a
State Key Laboratory of Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics, Chinese Academy of Sciences, Wuhan 430071, China
b
School of Engineering Science, University of Chinese Academy of Sciences, Beijing 100049, China

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

Article history: This study introduces the vector sum method into discontinuum-based methods by considering the slid-
Received 8 June 2017 ing vector and the stress state of the discrete block system. The sliding direction computation and force
Received in revised form 20 July 2017 projection in the new approach are detailed, and the safety factor is solved by explicit equations. The vec-
Accepted 31 July 2017
tor sum method is implemented in the discontinuous deformation analysis (DDA) program and is used to
compute the safety factors for two numerical examples. A comparison of the solutions obtained with the
theoretical analysis and limit equilibrium analysis demonstrates that the new method is suitable for cal-
Keywords:
culating the safety factor of a slope.
Slope
Discontinuous deformation analysis (DDA)
2017 Published by Elsevier Ltd.
Vector sum method
Safety factor
Sliding direction
Force projection

1. Introduction accuracy than the analytical solutions or the results of conven-

There are two common calculation methods in slope stability
analysis: the limit equilibrium method (LEM) [1] and the finite ele-
ment method (FEM) [2]. Compared to the FEM, the LEM is simple
and has high computational efficiency, which have made it the fun-
damental method for slope stability analysis [3]. However, the LEM
is a static analysis method and ignores the movement of the slope
under the action of external forces. Large displacements, discontin-
uous contacts, a precise friction law and stabilized time-step
dynamic computation are required to reproduce the slope move-
ment until the equilibrium state. Discontinuous deformation anal-
ysis (DDA) is a new efficient numerical analysis method that can
meet the above requirements [4].
Since it was first proposed by Shi [5], the DDA method has been
developed rapidly in many respects, such as contact and 3D analy-
sis [614]. DDA can accurately simulate the translation, rotation
and deformation of a block based on time-step calculations. There-
fore, it is able to not only analyze the sliding instability but also
simulate slope failure, such as toppling and collapse. Yeung [15]
verified the form of sliding and toppling failure of blocks.
MacLaughlin et al. [16] simulated both the plane and circular
failure modes of a slope and found that the results have better
Corresponding author.
E-mail address: fxd_scu@163.com (X. Fu).
tional analytical methods. Cheng et al. [17] analyzed the accuracy
of DDA in simulating the four basic motion modes of a rock,
namely, free-falling, rolling, sliding and bouncing. The DDA
method has also been used to analyze the stability of practical
slopes. Hatzor et al. [18] investigated the dynamic deformation of
the upper terrace of King Herods Palace in Masada. Jiao et al. [19]
and Xu et al. [20] simulated the entire processes for three practical
landslides, namely, the Qianjiangping, Majiagou and Dongmiaojia
landslides. Wu et al. [21,22] applied DDA to simulate the kinematic
behavior of sliding rock blocks in the Tsaoling and Chiu-fen-erh-
shan landslides, which were induced by the 1999 Chi-Chi earth-
quake. Zhang et al. [23] and Wu et al. [24] presented a run-out
analysis of the Daguangbao landslide subjected to near-fault
multi-direction earthquake forces. Huang et al. [25] used DDA to
model the Donghekou landslide, which was triggered by the
2008 Wenchuan earthquake.
In addition to being used to simulate the failure process of a
slope, DDA can be used to compute the safety factor. There are
three ways to define the safety factor in DDA. The first approach
is the fictitious force method [26], where the safety factor is
defined as the magnification of the fictitious force, which puts
the slope in a state of critical instability. The second approach is
the strength reduction method [27], where the safety factor is
defined as the ratio of the reduced shear strength parameters,
which puts the slope in a state of critical instability. The third

http://dx.doi.org/10.1016/j.compgeo.2017.07.026
0266-352X/ 2017 Published by Elsevier Ltd.
 X. Fu et al. / Computers and Geotechnics 92 (2017) 6876 69

approach is the contact force method [28], where the safety factor
is defined as the ratio of the resistant and sliding forces, which are
calculated with the contact forces of the sliding surface. In the for-
mer two approaches, the stress states are obtained by multiple cal-
culations under hypothetical conditions, and the safety factor
cannot truly represent the actual state of the slope. The third
approach is actually the LEM, and it considers the dynamic itera-
tive process using DDA. However, in this approach, integrations
of the sliding or resistant forces are the scalar sum of the contact
force components, which cannot reflect the concept of the sliding
vector. To overcome this problem, Ge et al. [29] proposed the vec-
tor sum method (VSM), in which the safety factor is computed
based on the real stress state and the vector sum algorithm, and
thus, the stress field only needs to be calculated once and the
mechanical meaning is clear [30,31].
In this paper, the VSM has been introduced into DDA to deter-
mine the safety factor, and two key stepssliding direction compu-
tation and force projectionare detailed considering the dynamic
Fig. 1. Angle-edge contact in DDA.
iterations and contact forces.

2. A brief description of DDA Then, the total equilibrium equations are solved, and the
motion parameters in the current time step are obtained. The con-
A discrete deformable block is the basic unit of the DDA tact modes and penetrations are updated, and the hypothetical
method, and the individual blocks are connected and form the contact forces are modified to obtain the true contact forces. Rn
block system based on the boundary conditions. DDA uses time- and Rs are the normal and tangential components of the true con-
step calculations. For each time step, based on the simplex integra- tact forces, respectively, and they are confirmed by the following
tion method, the stiffness, inertia, initial stresses, loading, and con- conditions:
tact matrices are collected to form the following total equilibrium
equations: (a) If R0n is positive (it is assumed that tension is positive and
2 38 9 8 9 pressure is negative),
K 11 K 12


K 1n > DD1 > > F 1 >
>
> >
> > >
6K


K 2n 7> > > > > > R0n K n dN P 0
6 21 K 22 7< DD2 = < F 2 = 3
6 . .. .. 7 1
6 . ... 7
5>
> ... >
> >
>
.. >
4 . . . >
> >
> > . > > In this case, the contact mode is open, and no normal or tangential
: ; > : > ; springs are added. The true contact forces are expressed as
K n1 K n2


K nn DD n Fn

Rn 0
where [Kij] is a stiffness sub-matrix of order 6  6, {Fi} is a load sub- 4
vector of order 6  1, and {DDi} is the incremental displacement Rs 0
solution of a block in a time step: i, j = 1, 2, . . . , n, where n is the (b) If R0n is negative and R0s is less than the shear strength,
block number. )
R0n K n dN < 0
The most important task for solving the total equilibrium equa- 5
tions is accurately calculating the contact forces between the R0s < R0n tan u cl
blocks. In the DDA method, three contact modes are defined: open,
where u and c are the friction angle and cohesion strength, respec-
lock and slide. These contact modes are converted in the calcula-
tively, of the contact surface and l is the contact length.
tion process by the addition or subtraction of normal or tangential
In this case, the contact mode is lock, and both the normal and
springs at each contact position. We call this approach open-
tangential springs are added. The true contact forces are equal to
close iterations, which should satisfy two principles: no penetra-
the hypothetical contact forces, which are expressed as
tion or tension in the normal direction and Coulombs law in the
)
tangential direction. Rn R0n
In the two-dimensional DDA method, all contacts between the 6
Rs R0s
blocks are treated in the angle-edge form. As shown in Fig. 1, P1
is a vertex of block i, and P2P3 is an edge of block j. In the current (c) If R0n is negative and R0s is not less than the shear strength,
time step, before solving the total equilibrium equations, the )
R0n K n dN < 0
motion parameters in the previous time step are used to estimate 7
the movements, and P1 is assumed to move to P0, which may be R0s P R0n tan u cl
located in block j; then, and the penetrations are generated. dN
In this case, the contact mode is slide, and only the normal
and dS are the normal and tangential components of the penetra-
spring is added. The true contact forces are expressed as
tions, respectively, and the normal and tangential components of )
the hypothetical contact forces are expressed as Rn R0n
) 8
Rs R0n tan u
R0n K n dN
2
R0s K s dS In the original DDA program written by Shi [5], if slide has
occurred, the cohesive force is reduced to zero and the friction
where R0n and R0s are the normal and tangential components of the angle remains unchanged in the next time-step calculation, which
hypothetical contact forces, respectively, and Kn and Ks are the nor- reflects the sliding friction of the contact surface. A literature
mal and tangential spring stiffness coefficients, respectively. review shows that the errors are generally lower than 1% if the
70 X. Fu et al. / Computers and Geotechnics 92 (2017) 6876
strength of the contact surface is only from friction, and the accu-
racy will decrease dramatically once the strength of the contact
surface contains cohesion [32,33]. Thus, the original DDA program
may provide an underestimated safety factor when applied to
cohesive materials. To avoid this problem, the cohesive force
remains unchanged in this paper, and this treatment is similar to
the rigid plastic model adopted by the LEM or the ideal elastoplas-
tic model used in the FEM [1,2].
3. Safety factor computation in the DDA method
As described in the Introduction, the fictitious force method,
strength reduction method and contact force method are used to
compute the safety factor in DDA. The fictitious force method
and strength reduction method are introduced in view of the finite
element method, and cannot truly represent the actual state of the
slope. The contact force method is a specific method for a discrete
block system and employs stress states obtained by dynamic iter-
ative analysis.
As shown in Fig. 2, it is assumed that there is a distinct sliding
surface in the block system. The sliding block is defined as the
block above the sliding surface, and the contact block is defined
as the block along the sliding surface. All of the contact modes of
the contact blocks can be determined by the open-close itera-
tions. It is assumed that the number of the contact block with lock
mode is mi, and the number of the contact block with slide mode is
mj. Then, the safety factor of the block system along the sliding sur-
face can be computed as
Pmi Pmj
tan ui ci li j1
i1 Rni Rnj tan uj
Fs Pm i Pmj
i1 Rsi j1 Rnj tan uj
where Fs is the safety factor, which is defined as the ratio of the
resistant force to the sliding force; Rni and Rsi are the normal and
tangential components, respectively, of the contact force for the
contact block with lock mode; Rnj is the normal contact force for
the contact block with slide mode; ui and ci are the friction angle
and cohesion force, respectively, of the contact surface for the con-
tact block with lock mode; and li is the contact length for the con-
tact block with lock mode.
Eq. (9) can only be used for the equilibrium state, which means
that all of the sliding blocks should be static after the DDA calcula-
tion. The inertia forces of the blocks should be considered when
computing the safety factor in each time step. The number of slid-
ing blocks with inertia forces is assumed to be bp; then, the safety
factor of the block system along the sliding surface can be
expressed as
Fig. 2. Block system sliding along a distinct surface.
Pmi Pmj
i1 Rni tan ui c i li j1 Rnj tan uj
Fs Pm i Pmj Pbp 10
i1 Rsi j1 Rnj tan uj p1 Rp
where Rp is the inertia force of the sliding block.
4. Safety factor computation using the VSM
4.1. VSM
In Eqs. (9) and (10), both the sliding and resistant forces are sca-
lar sums of the contact force components, and the computation
method does not reflect the concept of the sliding vector. The
VSM is used to overcome this limitation. In this method, first, the
sliding direction of the slope is determined by analyzing the force
system of the sliding blocks; second, the sliding/resistant force of
each sliding block is projected in the sliding direction of the slope
and is added to the total sliding/resistant force; and third, the
safety factor of the block system is computed using Eq. (11). At pre-
sent, the VSM is mainly used in continuum-based methods [34,35],
such as the FEM. This paper introduces the VSM into
discontinuum-based methods. There are two key stepssliding
direction computation and force projectionwhich are detailed
in the following sections.
P
Rh
F s h P 11
Th
where h is the sliding direction of the slope; Rh is the component
of the resistant force of a sliding block in the sliding direction; and
Th is the component of the sliding force of a sliding block in the
9 sliding direction.
4.2. Computation of the sliding direction
In Fig. 3, the two-dimensional standard coordinate system is
used, where the X-axis is horizontal and points to the right and
the Y-axis is vertical and points upward. The boundary conditions,
sliding surface and mechanical parameters of the rock mass are
known. The sliding direction of the VSM is the potential sliding
direction of the slope, which is related to the force conditions of
each sliding block. The motion parameters of each block and the
contact forces between them can be calculated using
discontinuum-based methods, and these results are used to deter-
mine the sliding direction.
Regarding the motion parameters of each block, the incremen-
tal displacement is solved using Eq. (1), and the velocity and accel-
eration of a block can be computed with the transform formula of
the Newmark method as follows:
Fig. 3. Schematic of the determination of the sliding direction.
 X. Fu et al. / Computers and Geotechnics 92 (2017) 6876 71

    
_ c c _ c Rn K n dN
fDg fDDgn1 1  fDg 1  DtfDg 17
bDt b Rs K n dN tan u
n1 n n
2b
  12
1 1 _ 1
fDg fDDgn1  fDgn   1 fDg The normal and tangential contact forces in Eqs. (16) and (17)
n1
bDt 2 bDt 2b n
are projected in the sliding direction and then summed to obtain
where fDg is the acceleration, {D}
_ is the velocity, {DD} is the incre- the resistant force:
mental displacement, the subscripts (n and n + 1) represent two X X
mi X
mi
adjacent calculation time steps, Dt is the time interval for each cal- Rh Rni tan ui ci li cosh  ai Rni sinh  ai
culation step, and b and c are calculation parameters of the New- i1 i1

mark method. Various integration methods can be generated with Xmj

X
mj

different values of b and c. The central difference method (b = 0, Rnj tan uj cosh  aj Rnj sinh  aj 18
c = 1/2) and constant acceleration integration method (b = 1/2, j1 j1

c = 1) are common methods used in discontinuum-based methods. Another aspect is the sliding force, which is provided by the
The numerical stability of these integration methods has been dis- contact force and inertia force. For an arbitrary contact point of a
cussed by many researchers [3638], and the conventional DDA block on the sliding surface, if the contact mode is lock, the sliding
method uses the constant-acceleration integral method to fully uti- force includes the normal and tangential contact forces, and
lize the unconditionally stable condition. according to Eq. (6),
As shown in Fig. 3, for an arbitrary sliding block P, the inertial 
force can be expressed by the product of the acceleration and mass, Rn K n dN
19
and the X- and Y-axis components are Rs K s dS

Ipx mp apx If the contact mode is slide, the sliding force is equal to the
13
Ipy mp apy resistant force, which can be computed by Eq. (17). For an arbitrary
sliding block P, the inertial force can be expressed by the product of
where Ipx and Ipy are the X- and Y-axis components, respectively, of the acceleration and mass, and the X- and Y-axis components can
the inertial force; mp is the mass of block P; and apx and apy are the be computed by Eq. (13).
X- and Y-axis components, respectively, of the acceleration. The contact forces and inertia forces in Eqs. (13), (17) and (19)
Another aspect is the contact forces between the blocks. As are projected in the sliding direction and then summed to obtain
shown in Fig. 3, for an arbitrary contact point K of a block on the the sliding force:
sliding surface, according to friction theory, the sliding direction
is the tangential direction of the sliding surface, and thus, the slid- X X
mi X
mi X
mj

ing force of contact point K should be the contact force component Th Rsi cosh  ai Rni sinh  ai Rnj
i1 i1 j1
in the tangential direction of the sliding surface. The tangent line of
contact point K is the contact edge, and it is assumed that the inter- X
mj
X
bp

section angle between the contact edge and the positive X-axis is  tan uj cosh  aj Rnj sinh  aj Ipx
ak, where the angle from the X-axis counterclockwise to the Y- j1 p1

axis is positive. In Fig. 3, ak is negative. X

bp  p
As detailed in Section 2, the tangential force of each contact is  cos h Ipy cos h 20
p1
2
computed and is projected to the X- and Y-axes as

F kx F kt cos ak Finally, the safety factor of the VSM can be obtained by substi-
14 tuting Eqs. (18) and (20) into Eq. (11).
F ky F kt sin ak
where the Fkt is the tangential force of a contact and Fkx and Fky are
5. Implementation of the VSM in the DDA program
the X-and Y-axis components, respectively, of the tangential contact
force.
Based on the above analysis, as shown in Fig. 4, the VSM is
The X- and Y-axis components of all forces are computed using
implemented in the DDA program as follows:
Eqs. (13) and (14), and the sliding direction can be obtained as
Step 1. Input the geometric and mechanical parameters of the
P P
FY F ky Ipy blocks and determine the computational parameters, including
h arctan arctan P P 15 the total analytical steps N, time interval Dt and maximum step
FX F kx Ipx
displacement dm.
Step 2. Identify whether a block is a sliding block or not. A block
4.3. Projections of the resistant/sliding force is marked with 1 if it is sliding and with 0 otherwise.
Step 3. Begin the calculation in step-i (1  i  N).
According to the principle of VSM, all of the forces acting on the Step 4. Contact judgment, inheritance, and initialization.
blocks should be projected in the sliding direction to obtain the Step 5. Determine the stiffness matrix size and non-zero storage
resistant and sliding forces. location.
The resistant force is provided by the contact force between the Step 6. Calculate all types of stiffness sub-matrixes and the load
sliding block and sliding surface and includes the frictional force, sub-vector.
cohesive force and normal contact reaction force. For an arbitrary Step 7. Add or remove the contact spring according to the con-
contact point of a block on the sliding surface, if the contact mode tact state and calculate the stiffness sub-matrixes and load sub-
is lock, the resistant force includes the normal and tangential con- vector of the contact spring.
tact forces: Step 8. Form and solve the total equilibrium equations.
 Step 9. Assess whether the open-close iterations are conver-
Rn K n dN
16 gent or not, and if they are not convergent, return to Step 7. If
Rs K n dN tan u cl
the iteration number is more than 5 and the time interval is
If the contact mode is slide, according to Eq. (8), reduced, return to Step 6.
72 X. Fu et al. / Computers and Geotechnics 92 (2017) 6876

geometric mechanical computational Step 15. Compute the safety factor.

parameters parameters parameters Step 16. i = i + 1, if 1  i  N, return to Step 3.
Step 17. END.

Identify the sliding blocks 6. Numerical examples

Begin the step-

6.1. Slope with a planar sliding surface

Fig. 5 is the analytical model of a slope with a planar sliding sur-

Contact judgment, inheritance, and initialization face. Theoretically, the resistance for this model is
R mg cos a tan / cl 21
Stiffness matrix size and non-zero storage location
where a is the slope angle; m is the sliding mass, which can be com-
Reduce time interval puted using the sliding volume and rock density; l is the sliding
length; u and c are the friction angle and cohesion strength, respec-
Calculate stiffness sub-matrixes
tively, of the sliding surface; and g is the gravitational acceleration,
and load sub-vector
which is 10 m/s2 in this paper.
The sliding force for this model is

Add or remove the contact spring T mg sin a 22

Solve the total equilibrium
equations
open-close iterations are
convergent?
Yes
maximum step displacement is
satisfied?
Yes
Store accelerations of the sliding
blocks
Store contact forces between the
sliding and non-sliding blocks
Compute the sliding direction
Then, the safety factor for the slope can be computed as
Yes
No tan / cl
Fs 23
tan a mg sin a
In this example, the slope angle is 30, the sliding mass is
No 9.4  106 kg, the sliding length is 69.282 m, and the mechanical
Iteration number parameters of the bedrock and sliding mass are provided in Table 1.
is more than 5? The frictional angle and cohesion strength of the sliding surface are
varied to design 18 different calculation cases, as shown in Tables 2
No and 3.
For each calculation case, the safety factor for the slope can be
computed by Eq. (23). In contrast, the safety factors are computed
using the DDA program implemented with the VSM. The results
are presented in Tables 2 and 3, which demonstrate that the sliding
directions of the DDA results are approximately 30, which is
consistent with the slope angle. Furthermore, the errors of the
safety factors between the simulation results and theoretical solu-
tions are small, with a maximum value of less than 0.3%, thus val-
idating the improved DDA method.

6.2. Slope with a circular sliding surface

Projections of the forces
Fig. 6 shows the analytical model of a slope with a circular slid-
ing surface. The mechanical parameters of the bedrock and sliding
Compute the safety factor mass are shown in Table 4. The frictional angle and cohesion
strength of a sliding surface are varied to design three different cal-
Yes
culation cases. In calculation case one, the frictional angle is 40,
the cohesion strength is 60 kPa, the slope is completely stable
No and the safety factor is 1.507 based on the LEM. In calculation case
two, the frictional angle is 30, the cohesion strength is 30 kPa, the
END
slope is critically stable and the safety factor is 1.012 based on the
Fig. 4. Implementation of the vector sum method in the DDA program.

Step 10. Judge whether the maximum step displacement is sat-

isfied or not. If it is not satisfied, reduce the time interval and
return to Step 6.
Height/m

Step 11. Store the accelerations of all sliding blocks marked as 1.

Step 12. Find all contacts between a sliding block and non-
sliding block based on the marked numbers and store the contact
forces. For two blocks that come into contact, one block is marked 30
as 1 and the other is marked as 0.
Step 13. Compute the sliding direction based on the stored
accelerations and contact forces. Length/m
Step 14. Project the stored accelerations and contact forces in
the sliding direction and compute the resistant/sliding forces. Fig. 5. Analytical model of a slope with a planar sliding surface.
 X. Fu et al. / Computers and Geotechnics 92 (2017) 6876 73

Table 1
Mechanical parameters of the rock and planar sliding surface.

Materials Density Youngs modulus Poissons Frictional Cohesion Tensile strength Normal springs Shear springs
(kg/m3) (GPa) ratio angle () strength (kPa) (kPa) coefficient (GPa) coefficient (GPa)
Bedrock 2700 50 0.25
Sliding 2100 5 0.25
mass
Sliding 1550 090 0 10 4
surface

Table 2
Comparisons of the DDA results and analytical solutions with different friction angles (c = 0 kPa).

Frictional angle () Theoretical safety factor DDA results with the vector sum method
Safety factor Sliding direction Error of the safety factor (%)
15 0.4641 0.4641 29.9999 0.0000
20 0.6304 0.6304 30.0000 0.0000
25 0.8077 0.8077 29.9977 0.0000
30 1.0000 0.9972 30.0924 0.2800
35 1.2128 1.2111 30.2440 0.1402
40 1.4534 1.4534 30.0036 0.0000
45 1.7321 1.7320 29.9815 0.0058
50 2.0642 2.0641 29.9944 0.0048

Table 3
Comparisons of the DDA results and analytical solutions with different cohesive strengths (u = 20).

Cohesive strength (kPa) Theoretical safety factor DDA results with the vector sum method
Safety factor Sliding direction Error of the safety factor (%)
0 0.6304 0.6304 30.0000 0.0000
10 0.7778 0.7767 30.0571 0.1414
20 0.9252 0.9246 30.0180 0.0649
30 1.0726 1.0723 30.0435 0.0280
40 1.2200 1.2171 29.5384 0.2377
50 1.3674 1.3644 29.9343 0.2194
60 1.5149 1.5148 30.0000 0.0066
70 1.6623 1.6622 30.0000 0.0060
80 1.8097 1.8097 30.0000 0.0000
90 1.9571 1.9571 30.0000 0.0000

(2) For case two, the slope starts to move during the DDA sim-
ulation when the slope is critically stable when using the
LEM. This result occurs because the LEM ignores the move-
ment of the slope under gravity, whereas DDA reproduces
Height/m

the slope movement by a time-step dynamic computation.

Length/m
Fig. 6. Analytical model of a slope with a circular sliding surface.
LEM. In calculation case three, the frictional angle is 25, the cohe-
sion strength is 20 kPa, the slope is unstable and the safety factor is
0.816 based on the LEM. For each calculation case, the improved
DDA method is used to compute the safety factor of the slope,
and the results are presented in Table 5. Comparisons between
the DDA results and LEM solutions yield the following conclusions:
(1) For cases one and two, the maximum error is less than 1.5%,
which proves the effectiveness of the improved DDA
method.
Reinforcement measures are recommended for a critically
stable slope even if its safety factor is slightly larger than 1.
(3) For case three, the LEM only calculates a static safety factor,
whereas the DDA can provide the time history of the safety
factor and sliding direction during the movement of the
slope. As shown in Fig. 7, the safety factor increases contin-
uously from 0.800 to 0.937 from 0 s to 50 s and then remains
unchanged; several fluctuations in the safety factor and slid-
ing direction are observed, which should be caused by the
unstable contacts between the sliding blocks. In addition,
DDA can reproduce the entire failure process of the slope,
as shown in Fig. 8.
7. Conclusions and discussion
DDA is a new, efficient numerical analysis method that satisfies
three convergences: convergence of the equilibrium equations,
convergence of open-close iterations for all contacts and conver-
gence of the maximum displacement for static computations. For
slope stability problems, the DDA method not only simulates the
failure process but also calculates the safety factor. In this paper,
74 X. Fu et al. / Computers and Geotechnics 92 (2017) 6876

Table 4
Mechanical parameters of the rock and circular sliding surface.

Materials Density Youngs Poissons Frictional Cohesion Tensile Normal springs Shear springs
(kg/m3) modulus (GPa) ratio angle () strength (kPa) strength (kPa) coefficient (GPa) coefficient (GPa)
Bedrock 2700 50 0.25
Sliding mass 2100 5 0.25
Sliding surface-case 1 40 60 0 10 4
Sliding surface-case 2 30 30 0 10 4
Sliding surface-case 3 25 20 0 10 4

Table 5
Comparisons of the DDA results and LEM solutions under different calculation cases.

Cases Safety factor for limit equilibrium method DDA results with the vector sum method
Safety factor Sliding direction Error of the safety factor (%)
1 1.507 1.4902 29.4846 1.1148
2 1.012 0.9990 30.0009 1.2846
3 0.816

Fig. 7. Time history of the safety factor and sliding direction by DDA calculation.

0s 100s

175s 250s
Fig. 8. Failure process of the slope by DDA simulation.
 X. Fu et al. / Computers and Geotechnics 92 (2017) 6876
we discuss the applicability of the three existing methods and
introduce a new approach.
(a) The fictitious force method and strength reduction method
use the stress distributions resulting from the fictitious con-
ditions by artificially increasing the load or reducing the
strength of the material, which cannot truly represent the
actual state of the slope. The contact force method is a
method specific to the discrete block system and uses the
actual stress states obtained by dynamic iterative analysis.
However, the sliding and resistant forces of the contact force
method are only the scalar sum of the contact force compo-
nents, which does not reflect the concept of the sliding
vector.
(b) According to the sliding vector and stress state of the dis-
crete block system, the VSM is introduced into
discontinuum-based methods, and a safety factor that
reflects the potential slide of the slope under the actual force
conditions is defined. A method of determining the sliding
direction is presented, and the solution expression for the
safety factor is derived by the force projection. The safety
factor of this method can be directly solved by an explicit
equation without iterative calculation, and the computation
procedure is simple and convenient for engineering applica-
tions when the contact forces and inertia forces are known.
(c) Combining the principle of the DDA method, the VSM is
implemented in the DDA program, and 17 steps are detailed.
Twenty calculation cases are designed for two numerical
examples of a slopea planar sliding surface and circular
sliding surfaceby changing the frictional angle and cohe-
sion strength of the sliding surface. The safety factor of each
case is computed using the improved DDA method. The
results of a comparison with the theoretical analysis and
LEM illustrate that the new method is suitable for calculat-
ing the safety factor of the slope.
(d) The principle of VSM indicates that an accurate stress field of
the slope is the key factor for determining the safety factor.
Specifically, the stress field is represented by precise simula-
tions of the contact force and block stress in discontinuum-
based methods. With regard to contact analysis, the DDA
method first checks the contact modes by a rigorous geo-
metric operation and then calculates the contact forces using
the penalty function method and applies Coulombs Law to
assess the shear failure of the contact surface. This analysis
process is accurate for most simulations, but three problems
have been noted by previous research. The first problem is
the contact judgment for certain cases, the second problem
is the selection of the penalty parameters, and the third
problem is the contact mechanics model. Shi [39] and other
researchers [4043] have made significant progress in solv-
ing these problems by using general contact theory, comple-
mentary theory and the displacement-dependent interface
shear strength. Regarding block stress analysis, the original
DDA program uses a first-order assumption and assumes
that all blocks have constant stresses and constant strains.
The DDA method with higher-order [44] or sub-block [45]
algorithms has been proven to be effective at improving
the simulation accuracy of the stress field inside a block.
Acknowledgements
The authors thank Dr. Shi G.H. for his guidance in understand-
ing the DDA method and providing the original DDA program.
The work reported in this paper is financially supported by the
National Natural Science Foundation of China (No. 51509241, No.
75
U1402231, No. 51679232) and the National Key Basic Research
Program of China (973 Program) under Grant No. 2015CB057905.
References
[1] Morgenstern NR, Price VE. The analysis of the stability of general slip surfaces.
Geotechnique 1965;15(1):7993.
[2] Zeinkiewicz O, Humpheson C, Lewis R. Associated and non-associated visco-
plasticity in soils mechanics. Geotechnique 1975;25(4):67189.
[3] Duncan JM. State of the art: limit equilibrium and finite element analysis of
slopes. J Geotech Eng (ASCE) 1996;122(7):57796.
[4] Shi GH. Rock stability analysis and three convergences of discontinuous
deformation analysis (DDA). In: Ma G, Zhou Y, editors. Proceedings of ICADD-9:
the 9th international conference on analysis of discontinuous
deformation. Singapore: Dover; 2009. p. 111.
[5] Shi GH, Goodman RE. Two dimensional discontinuous deformation analysis.
Int J Numer Anal Meth Geomech 1985;9:54156.
[6] MacLaughlin MM, Doolin DM. Review of validation of the discontinuous
deformation analysis (DDA) method. Int J Numer Anal Meth Geomech 2006;30
(4):271305.
[7] Yagoda-Biran G, Hatzor YH. Benchmarking the numerical discontinuous
deformation analysis method. Comput Geotech 2016;71:3046.
[8] Bao HR, Zhao ZY. The vertex-to-vertex contact analysis in the two-dimensional
discontinuous deformation analysis. Adv Eng Softw 2012;45(1):110.
[9] Zhang H, Liu SG, Zheng L, Zhong GH, Lou S, Wu YQ, et al. Extensions of edge-to-
edge contact model in three-dimensional discontinuous deformation analysis
for friction analysis. Comput Geotech 2016;71:26175.
[10] Beyabanaki SAR, Mikola RG, Hatami K. Three-dimensional discontinuous
deformation analysis (3-D DDA) using a new contact resolution algorithm.
Comput Geotech 2008;35:34656.
[11] Ning YJ, Zhao ZY. A detailed investigation of block dynamic sliding by the
discontinuous deformation analysis. Int J Numer Anal Meth Geomech
2012;37:237393.
[12] Fu XD, Sheng Q, Zhang YH, Chen J. Application of the DDA to stress wave
propagation through a one-dimensional rock mass. Int J Rock Mech Min Sci
2015;80:15570.
[13] Choo LQ, Zhao ZY, Chen HM, Tian Q. Hydraulic fracturing modeling using the
discontinuous deformation analysis (DDA) method. Comput Geotech
2016;76:1222.
[14] Ma SQ, Zhao ZY, Nie W, Nemcik J, Zhang ZY, Zhu X. Implementation of
displacement-dependent Barton-Bandis rock joint model into discontinuous
deformation analysis. Comput Geotech 2017;86:18.
[15] Yeung MR. Application of Shis discontinuous deformation analysis to the
study of rock behavior. Berkeley: Department of Civil Engineering, University
of California; 1991.
[16] MacLaughlin M, Sitar N, Doolin D, Abbot T. Investigation of slope-stability
kinematics using discontinuous deformation analysis. Int J Rock Mech Min Sci
2001;38(5):75362.
[17] Cheng GQ. Numerical simulation in rockfall analysis: a close comparison of 2-
D and 3-D DDA. Rock Mech Rock Eng 2013;46:52741.
[18] Hatzor YH, Arzi AA, Zaslavsky Y, Shapira A. Dynamic stability analysis of
jointed rock slopes using the DDA method: King Herods Palace, Masada, Israel.
Int J Rock Mech Min Sci 2004;41:81332.
[19] Jiao YY, Zhang HQ, Tang HM, Zhang XL, Adoko AC, Tian HN. Simulating the
process of reservoir-impoundment-induced landslide using the extended DDA
method. Eng Geol 2014;182:3748.
[20] Xu WJ, Jie YX, Li QB, Wang XB, Yu YZ. Genesis, mechanism, and stability of the
Dongmiaojia landslide, yellow river, China. Int J Rock Mech Min Sci
2014;67:5768.
[21] Wu JH. Seismic landslide simulations in discontinuous deformation analysis.
Comput Geotech 2010;37(5):594601.
[22] Wu JH, Chen CH. Application of DDA to simulate characteristics of the Tsaoling
landslide. Comput Geotech 2011;38(5):74150.
[23] Zhang YB, Chen GQ, Zheng L, Li YG, Wu J. Effects of near-fault seismic loadings
on run-out of large-scale landslide: a case study. Eng Geol 2013;166:21636.
[24] Wu JH, Do TN, Chen CH, Wang GH. New geometric restriction for the
displacement-constraint points in discontinuous deformation analysis. Int J
Geomech 2016;17(5), doi:http://dx.doi.org/10.1061/(ASCE)GM.1943-5622.
0000648.
[25] Huang D, Song YX, Cen DF, Fu GY. Numerical modeling of earthquake-induced
landslide using an improved discontinuous deformation analysis considering
dynamic friction degradation of joints. Rock Mech Rock Eng 2016;49:476786.
[26] Shi GH. Applications of discontinuous deformation analysis (DDA) to rock
stability analysis. In: Ju Y, Xisheng F, Bian H, editors. Proceedings of ICADD-8,
the 8th international conference on the analysis of discontinuous deformation.
Beijing (China); 2007. p. 114.
[27] Lin P, Liu XL, Hu SY, Li PJ. Large deformation analysis of a high steep slope
relating to the Laxiwa reservoir, China. Rock Mech Rock Eng 2016;49:225376.
[28] Wu AQ, Ding XL, Lu B. Validation for rock stability kinematics and its
application to rock slope stability evaluation using discontinuous deformation
analysis. In: Ju Y, Xisheng F, Bian H, editors. Proceedings of ICADD-8, the 8th
international conference on the analysis of discontinuous deformation. Beijing
(China); 2007. p. 2731.
[29] Ge XR, Feng DX, Gu XR, et al. Stability and deformation analysis of complex
rock foundations of several large dams and hydropower stations in China. In:
76 X. Fu et al. / Computers and Geotechnics 92 (2017) 6876
R. Yoshinaka, K. Kikuchi, editor. Proceedings of international workshop on rock
foundation, Tokyo, Japan; 1995.
[30] Wu JH. The elastic distortion problem with large rotation in discontinuous
deformation analysis. Comput Geotech 2015;69:35264.
[31] Jiang W, Zheng H. An efficient remedy for the false volume expansion of DDA
when simulating large rotation. Comput Geotech 2015;70:1823.
[32] Tsesarsky M, Hatzor YH, Sitar N. Dynamic displacement of a block on an
inclined plane: analytical, experimental and DDA results. Rock Mech Rock Eng
2005;38:15367.
[33] MacLaughlin MM, Doolin DM. Review of validation of the discontinuous
deformation analysis (DDA) method. Int J Numer Anal Meth Geomech
2006;30:271305.
[34] Ge XR. The vector sum method: a new approach to calculating the safety factor
of stability against sliding for slope engineering and dam foundation problems.
In: Chen Y, Zhan L, Tang X, editors. Advances in environmental geotechnics:
Proceedings of the international symposium on geoenvironmental engineering
in Hangzhou, China; 2010. p. 99110.
[35] Liu GY, Zhuang XY, Cui ZQ. Three-dimensional slope stability analysis using
independent cover based numerical manifold and vector method. Eng Geol
2017;225:8395.
[36] Qu XL, Fu GY, Ma GW. An explicit time integration scheme of numerical
manifold method. Eng Anal Bound Elem 2014;48(6):5362.
[37] Otsubo M, OSullivan C, Shire T. Empirical assessment of the critical time
increment in explicit particulate discrete element method simulations.
Comput Geotech 2017;86:6779.
[38] Lin SZ, Xie ZQ. Performance of DDA time integration. Sci China Tech Sci
2015;58(9):155866.
[39] Shi GH. Contact theory. Sci China Tech Sci 2015;58(9):145096.
[40] Wu JH, Ohnishi Y, Nishiyama S. Simulation of the mechanical behavior of
inclined jointed rock masses during tunnel constructions using discontinuous
deformation analysis (DDA). Int J Rock Mech Min Sci 2004;41(5):73143.
[41] Zheng H, Jiang W. Discontinuous deformation analysis based on
complementary theory. Sci China Ser E-Tech Sci 2009;52(9):254754.
[42] Wang LZ, Jiang HY, Yang ZX, et al. Development of discontinuous deformation
analysis with displacement-dependent interface shear strength. Comput
Geotech 2013;47:91101.
[43] Zhang YB, Xu Q, Chen GQ, Zhao JX, Zheng L. Extension of discontinuous
deformation analysis and application in cohesive-frictional slope analysis. Int J
Rock Mech Min Sci 2014;70(9):53345.
[44] Koo CY, Chern JC. The development of DDA with third order displacement
function. In: Salami MR, Banks D, editors. Proceedings of the 1st international
forum on discontinuous deformation analysis (DDA) and simulations of
discontinuous media. Berkeley: TSI Press; 1996. p. 3429.
[45] Lin CT, Amadei B, Jung J, Dwyer J. Extensions of discontinuous deformation
analysis for jointed rock mass. Int J Rock Mech Min Sci 1996;33(1):67194.