Вы находитесь на странице: 1из 61

Chapter 8

Circular failure

8.1 Introduction The approach adopted in this chapter is to

present a series of slope stability charts for circu-
Although this book is concerned primarily with the
lar failure. These charts enable the user to carry
stability of rock slopes containing well-defined
out a rapid check on the factor of safety of a
sets of discontinuities, it is also necessary to
slope, or upon the sensitivity of the factor of
design cuts in weak materials such as highly
safety to changes in ground water conditions,
weathered or closely fractured rock, and rock fills.
slope angle and material strength properties.
In such materials, failure occurs along a sur-face
These charts should only be used for the analysis
that approaches a circular shape (Figure 8.1),
of circular failure in slope materials that are
and this chapter is devoted to a discussion on the
homo-genous and where the conditions apply
stability analysis of these materials.
that were assumed in deriving the charts (see
In a review of the historical development of
Section 8.4). More comprehensive methods of
slope stability theories, Golder (1972) traced the
analysis are presented in Section 8.6. These
subject back almost 300 years. Much of the
methods can be used, for example, where the
development of circular failure analysis meth-ods
material properties vary within the slope, or
was carried out in the 1950s and 1960s, and
where part of the slide surface is at a soil/rock
these techniques have since been used to
interface and the shape of the slide surface
prepare computer programs that have the ver-
differs significantly from a simple circular arc.
satility to accommodate a wide range of geo-
This chapter primarily addresses the stability
logic, geometric, ground water and external
of slopes in two dimensions, and assumes that
loading conditions. This chapter discusses the
the slope can be modeled as a unit slice
principles of the theoretical work, and demon-
through an infinitely long slope, under plane-
strates their application in design charts and in
strain condi-tions. Section 8.6.5 discusses
the results of computer analyses. During the past
three-dimensional circular failure analysis, and
half century, a vast body of literature on the
Section 10.3.1 dis-cusses the influence of the
subject of circular failure has accumulated, and
radius of curvature of the slope on stability.
no attempt will be made to summarize the
material in this chapter. Standard soil mechan-ics
text books such as those by Taylor (1937),
8.2 Conditions for circular failure
Terzaghi (1943) and Lambe and Whitman (1969),
and methods of analysis
and papers by Skempton (1948), Bishop (1955),
Janbu (1954), Morgenstern and Price (1965), In the previous chapters, it has been assumed
Nonveiller (1965), Peck (1967), Spencer (1967, that the failure of rock slopes is controlled by
1969) and Duncan (1996) all contain excellent geological features such as bedding planes and
discussions on the stability of soil slopes. joints that divide the rock into a discontinuous
Circular failure 177

Figure 8.1 Circular

failure in highly
weathered, granitic
rock (on Highway 1,
near Devil’s Slide,
Pacifica, California).

mass. Under these conditions, one or more of the basalts, will also tend to fail in this manner. It is
discontinuities normally defines the slide sur-face. appropriate to design slopes in these materials
However, in the case of a closely fractured or on the assumption that a circular failure
highly weathered rock, a strongly defined process will develop.
structural pattern no longer exists, and the slide
surface is free to find the line of least resistance
8.2.1 Shape of slide surface
through the slope. Observations of slope failures
in these materials suggest that this slide surface The actual shape of the “circular” slide surface is
generally takes the form of a circle, and most influenced by the geological conditions in the
stability theories are based upon this observa- slope. For example, in a homogenous weak or
tion. Figure 8.1 shows a typical circular failure in weathered rock mass, or a rock fill, the failure is
a highly weathered rock slope above a high-way. likely to form as a shallow, large radius sur-face
The conditions under which circular failure will extending from a tension crack close behind the
occur arise when the individual particles in a soil crest to the toe of the slope (Figure 8.2(a)). This
or rock mass are very small compared with the contrasts with failures in high cohesion, low
size of the slope. Hence, broken rock in a fill will friction materials such as clays where the surface
tend to behave as a “soil” and fail in a circular may be deeper with a smaller radius that may exit
mode when the slope dimensions are beyond the toe of the slope. Figure 8.2(b) shows
substantially greater than the dimensions of the an example of conditions in which the shape of
rock fragments. Similarly, soil consisting of sand, the slide surface is modified by the slope geology.
silt and smaller particle sizes will exhibit circular Here the circular surface in the upper, weathered
slide surfaces, even in slopes only a few meters rock is truncated by the shallow dipping, stronger
in height. Highly altered and weathered rocks, as rock near the base. Stability analyses of both
well as rock with closely spaced, randomly ori- types of surface can be carried out using circular
ented discontinuities such as some rapidly cooled failure methods, although for the latter case it is
178 Circular failure


i –1

i –1
i –1
Wi b
h iE Ai
S = (ci + i tan i ) Ai
Circular sliding i Ai
surface Forces acting on slice, i


sliding surface
Figure 8.2 The shape of
typical sliding surfaces:
(a) large radius circular
surface in homogeneous,
weak material, with the
detail of forces on slice;
(b) non-circular surface in
weak, surficial material with
stronger rock at base.

necessary to use a procedure that allows the

8.2.2 Stability analysis procedure
shape of the surface to be defined.
For each combination of slope parameters The stability analysis of circular failure is carried
there will be a slide surface for which the factor out using the limit equilibrium procedure similar to
of safety is a minimum—this is usually termed that described in earlier chapters for plane and
the “critical surface.” The procedure to find the wedge failures. This procedure involves compar-
critical surface is to run a large number of ana- ing the available shear strength along the sliding
lyses in which the center co-ordinates and the surface with the force required to maintain the
radius of the circle are varied until the surface slope in equilibrium.
with the lowest factor of safety is found. This is The application of this procedure to circu-lar
an essential part of circular slope stability failures involves division of the slope into a series
analysis. of slices that are usually vertical, but may
Circular failure 179

be inclined to coincide with certain geological and rearranging this equation, we have
features. The base of each slice is inclined at angle
ψb and has an area A. In the simplest case, the c + σ tan φ
τ (8.2)
forces acting on the base of each slice are the shear e= FS
resistance S due to the shear strength of the rock
(cohesion c; friction angle φ), and forces E (dip The method of solution for the factor of safety
angle ψ; height h above base) acting on the sides of is to use an iterative process in which an initial
the slice (see detail Figure 8.2(a)). estimate is made for FS, and this is refined
The analysis procedure is to consider equilib- with each iteration.
rium conditions slice by slice, and if a condition of The influence of various normal stress distri-
equilibrium is satisfied for each slice, then it is butions upon the factor of safety of soil slopes
also satisfied for the entire sliding mass. The has been examined by Frohlich (1955) who found
number of equations of equilibrium avail-able that a lower bound for all factors of safety that
depends on the number of slices N, and the satisfy statics is given by the assumption that the
number of equilibrium conditions that are used. normal stress is concentrated at a single point on
The number of equations available is 2N if only the slide surface. Similarly, the upper bound is
force equilibrium is satisfied, and 3N if both force obtained by assuming that the normal load is
and moment equilibrium are satis-fied. If only concentrated at the two ends of the slide surface.
force equilibrium is satisfied, the number of The unreal nature of these stress distributions
unknowns is (3N − 1), while, if both force and is of no consequence since the object of the
moment equilibria are satisfied, the number of exer-cise, up to this point, is simply to determine
unknowns is (5N − 2). Usually between 10 and 40 the extremes between which the actual factor of
slices are required to realist-ically model the safety of the slope must lie. In an example con-
slope, and therefore, the number of unknowns sidered by Lambe and Whitman (1969), the
exceeds the number of equations. The excess of upper and lower bounds for the factor of safety of
unknowns over equations is (N − 1) for force a particular slope corresponded to 1.62 and 1.27,
equilibrium analysis, and (2N − 2) for analyses respectively. Analysis of the same problem by
that satisfy all conditions of equilib-rium. Thus, Bishop’s simplified method of slices gives a factor
the analyses are statically indeterm-inate and of safety of 1.30, which suggests that the actual
assumptions are required to make up the factor of safety may lie reasonably close to the
imbalance between equations and unknowns lower bound solution.
(Duncan, 1996). Further evidence that the lower bound solution
The various limit equilibrium analysis pro- is also a meaningful practical solution is provided
cedures either make assumptions to make up the by an examination of the analysis that assumed
balance between known and unknowns, or they the slide surface has the form of a logarithmic
do not satisfy all the conditions of equilib-rium. spiral (Spencer, 1969). In this case, the factor of
For example, the Spencer Method assumes that safety is independent of the normal stress distri-
the inclination of the side forces is the same for bution, and the upper and lower bounds coincide.
every slice, while the Fellenius and Bishop Taylor (1937) compared the results from a num-
methods do not satisfy all conditions of ber of logarithmic spiral analyses with results of
equilibrium. 1
lower bound solutions and found that the dif-
The factor of safety of the circular failure based
ference is negligible. Based on this comparison,
on limit equilibrium analysis is defined as

shear strength available to resist sliding (c + σ tan φ)

FS = 1 The lower bound solution discussed in this chapter is usu-
shear stress required for equilibrium on slip surface(τe) ally known as the Friction Circle Method and was used by
(8.1) Taylor (1937) for the derivation of his stability charts.
180 Circular failure

Taylor concluded that the lower bound solution (e) The locations of the tension crack and of
provides a value of the factor of safety that is the slide surface are such that the factor
sufficiently accurate for most practical of safety of the slope is a minimum for the
problems involving simple circular failure of slope geometry and ground water
homogeneous slopes. conditions considered.
The basic principles of these methods of (f) Ground water conditions vary from a dry
ana-lyses are discussed in Section 8.6. slope to a fully saturated slope under
heavy recharge; these conditions are
8.3 Derivation of circular failure charts defined in Figure 8.4.
(g) Circular failure charts are optimized for a
This section describes the use of a series of charts 3
rock mass density of 18.9 kN/m .
that can be used to determine rapidly the factor of
Densities higher than this give high factors
safety of circular failures. These charts have been of safety, densities lower than this give
developed by running many thousands of circular low factors of safety. Detailed circular
analyses from which a number of dimensionless analysis may be required for slopes in
parameters were derived that relate the factor of which the mate-rial density is significantly
safety to the material unit weight, friction angle and 3
different from 18.9 kN/m .
cohesion, and the slope height and face angle. It
has been found that these charts give a reliable The charts presented in this chapter corres-
estimate for the factor of safety, provided that the pond to the lower bound solution for the factor of
conditions in the slope meet the assumptions used safety, obtained by assuming that the normal
in developing the charts. In fact, the accuracy in load is concentrated on a single point on the slide
calculating the factor of safety from the charts is surface. These charts differ from those published
usually greater than the accuracy in determining the by Taylor in that they include the influence of a
shear strength of the rock mass. critical tension crack and of ground water.
Use of the stability charts presented in this
chapter requires that the conditions in the
8.3.1 Ground water flow assumptions
slope meet the following assumptions:
In order to calculate the forces due to water
(a) The material forming the slope is homogen- pres-sures acting on the slide surface and in
eous, with uniform shear strength properties the tension crack, it is necessary to assume a
along the slide surface. set of ground water flow patterns that coincide
(b) The shear strength τ of the material is char- as closely as pos-sible with conditions that are
acterized by cohesion: c and a friction believed to exist in the field.
angle φ, that are related by the equation τ In the analysis of rock slope failures discussed in
= c + σ tan φ (see Section 1.4). Chapters 6, 7 and 9, it is assumed that most of the
(c) Failure occurs on a circular slide surface, water flow takes place in discontinuities in the rock
2 and that the rock itself is practically imper-meable.
which passes through the toe of the slope.
(d) A vertical tension crack occurs in the In the case of slopes in soil or waste rock, the
upper surface or in the face of the slope. permeability of the mass of material is gener-ally
several orders of magnitude higher than that of
2 Terzaghi (1943: 170), shows that the toe failure assumed for intact rock and, hence, a general flow pattern will
this analysis gives the lowest factor of safety provided that φ develop in the material behind the slope.

> 5 . The φ = 0 analysis, involving failure below the toe of Figure 5.10(a) shows that, within the rock mass,
the slope through the base material has been discussed by the equipotentials are approximately perpendi-cular
Skempton (1948) and by Bishop and Bjerrum (1960) and is
applicable to failures which occur during or after the rapid to the phreatic surface. Consequently, the flow lines
construction of a slope. will be approximately parallel to the
Circular failure 181


Phreatic surface
Assumed equipotentials

Assumed flow lines

Sliding surface

Face angle Figure 8.3 Definition of ground

water flow patterns used in
(b) Surface recharge due to heavy rain circular failure analysis of slopes
in weak and closely fractured
Tension crack rock: (a) ground water flow
pattern under steady state
Sliding surface drawdown conditions where the
H phreatic surface coincides with
the ground surface at a distance
Assumed equipotentials x behind the toe of the slope.
The distance x is measured in
multiples of the slope height H ;
Assumed flow lines (b) ground water flow pattern in
a saturated slope subjected to
surface recharge by heavy rain.

phreatic surface for the condition of steady-state Figure 8.4 shows five ground water
drawdown. Figure 8.3 shows that this approx- conditions ranging from fully drained to
imation has been used for the analysis of the water saturated, based on the models shown in
pressure distribution in a slope under condi-tions of Figure 8.3. For con-ditions 2, 3 and 4, the
normal drawdown. Note that the phreatic surface is position of the ground water table is defined by
assumed to coincide with the ground surface at a the ratio x/H . These five ground water
distance x, measured in multiples of the slope conditions are used in conjunc-tion with the
height, behind the toe of the slope. This may circular failure charts discussed in Section 8.4.
correspond to the position of a surface water source,
or be the point where the phreatic surface is judged
to intersect the ground surface. 8.3.2 Production of circular failure charts
The phreatic surface itself has been obtained, The circular failure charts presented in this
for the range of the slope angles and values of x chapter were produced by running a search
considered, by solution of the equations proposed routine to find the most critical combination of
by L. Casagrande (1934), and discussed in the slide surface and tension crack for each of a wide
textbook by Taylor (1937). For the case of a sat- range of slope geometries and ground water con-
urated slope subjected to heavy surface ditions. Provision was made for the tension crack
recharge, the equipotentials and the associated to be located in either the upper surface, or in the
flow lines used in the stability analysis are based face of the slope. Detailed checks were carried
upon the work of Han (1972). This work involved out in the region surrounding the toe of the slope
the use of an electrical resistance analogue where the curvature of the equipotentials results
method to study ground water flow patterns in in local flow which differs from that illustrated in
slopes comprised of isotropic materials. Figure 8.3(a).
182 Circular failure

Ground water flow conditions Chart number

Fully drained slope

Surface water 8x slope height

behind toe of slope

Surface water 4x slope height

behind toe of slope

Surface water 2x slope height

behind toe of slope

Figure 8.4 Ground water flow
models used with circular
Saturated slope subjected to
failure analysis
heavy surface recharge charts—Figures 8.6–8.10.

The charts are numbered 1–5 (Figures 8.6– the slope and choose the chart which
8.10) to correspond with the ground water con- is closest to these conditions, using
ditions defined in Figure 8.4. Figure 8.4.
Step 2: Select rock strength parameters appli-
8.3.3 Use of the circular failure charts cable to the material forming the slope.
Step 3: Calculate the value of the dimensionless
In order to use the charts to determine the factor
ratio c/(γ H tan φ) and find this value on
of safety of a slope, the steps outlined here and
the outer circular scale of the chart.
shown in Figure 8.5 should be followed. Step 4: Follow the radial line from the value
found in step 3 to its intersection with
Step 1: Decide upon the ground water con- the curve which corresponds to the slope
ditions which are believed to exist in angle.
Circular failure 183

Step 5: Find the corresponding value of tan 2 1

φ/FS or c/(γ H FS), depending upon
which is more convenient, and
calculate the factor of safety.
Consider the following example: tan
H tan
◦ 3
A 15.2-m high cut with a face angle of 40 is FS
to be excavated in overburden soil with a dens-
ity γ = 15.7 kN/m , a cohesion of 38 kPa and a 4

friction angle of 30 . Find the factor of safety of 4
the slope, assuming that there is a surface
water source 61 m behind the toe of the slope.
The ground water conditions indicate the use c
of chart number 3 (61/15.2 ∼ 4). The value of H FS

c/(γ H tan φ) = 0.28 and the corresponding

Figure 8.5 Sequence of steps involved in using circular

value of tan φ/FS, for a 40 slope, is 0.32. failure charts to find the factor of safety of a slope.
Hence, the factor of safety of the slope of 1.80.

0 .01 .02 .03

2.0 .04 .05
.06 .07
1.8 .10
1.6 .14
.15 c
.17 H tan
1.4 .19

1.2 .25
FS .30
1.0 90
Slope angle .45
80 .50
0.6 .60
70 .70
60 .80
0.4 50 .90
40 1.0
30 1.5
0.2 20 2.0
0 8
0 .02 .04 .06 .08 .10 .12 .14 .16 .18 .20 .22 .24 .26 .28 .30 .32 .34

Figure 8.6 Circular failure chart number 1—fully drained slope.

184 Circular failure

0 .01 .02 .03 .04

2.0 .05 .06 .07
1.8 .10
1.6 .14
.15 c
.17 H tan
1.4 .19

1.2 .25
FS .30
1.0 90
Slope angle .45
0.6 60
70 .70
60 .80
0.4 50 .90
40 1.0
20 1.5
0.2 10 2.0
0 8
0 .02 .04 .06 .08 .10 .12 .14 .16 .18 .20 .22 .24 .26 .28 .30 .32 .34

Figure 8.7 Circular failure chart number 2—ground water condition 2 (Figure 8.5).

Because of the speed and simplicity of using

were determined for each slope analyzed.
these charts, they are ideal for checking the sens-
These locations are presented, in the form of
itivity of the factor of safety of a slope to a wide
charts, in Figures 8.11 and 8.12.
range of conditions. For example, if the cohesion
It was found that, once ground water is present
were to be halved to 20 kPa and the ground water
in the slope, the locations of the critical circle and
pressure increased to that represented by chart
the tension crack are not particularly sensitive to
number 2, the factor of safety drops to 1.28.
the position of the phreatic surface and hence
only one case, that for chart number 3, has been
plot-ted. It will be noted that the location of the
8.4 Location of critical slide
critical circle center given in Figure 8.12 differs
surface and tension crack signific-antly from that for the drained slope
During the production of the circular failure plotted in Figure 8.11.
charts presented in this chapter, the locations These charts are useful for the construction of
drawings of potential slides and for estimating the
of both the critical slide surface and the critical
friction angle when back-analyzing existing
tension crack for limiting equilibrium (FS = 1)
Circular failure 185

0 .01 .02
2.0 .03 .04 .05
.06 .07
.08 .09
1.8 .10
1.6 .14
.15 c
.17 H tan
1.4 .19

FS 90
Slope angle .40
80 .50
70 .60
60 .70
50 .80
0.4 40 .90
30  1.0
20 

0.2 2.0
0 8
0 .02 .04 .06 .08 .10 .12 .14 .16 .18 .20 .22 .24 .26 .28 .30 .32 .34

Figure 8.8 Circular failure chart number 3—ground water condition 3 (Figure 8.4).

circular slides. They also provide a start in 8.5.1 Example 1—China clay pit slope
locating the critical slide surface when carrying
Ley (1972) investigated the stability of a China
out more sophisticated circular failure analysis.
clay pit slope which was considered to be
As an example of the application of these
charts, consider the case of a drained slope potentially unstable, and that a circular failure
◦ was the likely type of instability. The slope pro-
having a face angle of 30 in a soil with a friction
file is illustrated in Figure 8.14 and the input

angle of 20 . Figure 8.11 shows that the critical data used for the analysis is included in this
slide circle center is located at X = 0.2H and Y = figure. The material, a heavily kaolinized
1.85H and that the critical tension crack is at a granite, was tested in direct shear to determine
distance b = 0.1H behind the crest of the slope. the friction angle and cohesion.
These dimensions are shown in Figure 8.13.
Two piezometers in the slope and a known
water source some distance behind the slope
8.5 Examples of circular failure analysis enabled an estimate to be made of the pos-
ition of the phreatic surface as shown in Figure
The following two examples illustrate the use of 8.14. From Figure 8.14, chart number 2
the circular failure charts for the study of the corresponds most closely to these ground
stability of slope in highly weathered rock. water conditions.
186 Circular failure

0 .01 .02 .03

2.0 .04
.05 .06
1.8 .10
1.6 .14
.15 c
.17 H tan
1.4 .19

1.2 .25
tan 90
FS .30

0.8 Slope angle .40

80 .45
0.6 70 .60
60 .70
50 .80
0.4 40 .90
0.2 2.0
0 8
0 .02 .04 .06 .08 .10 .12 .14 .16 .18 .20 .22 .24 .26 .28 .30 .32 .34

Figure 8.9 Circular failure chart number 4—ground water condition 4 (Figure 8.4).

From the information given in Figure 8.14, was required to check whether the cut would be
the value of the ratio c/(γ H tan φ) = 0.0056, stable. The slope was in weathered and altered
and the corresponding value of tan φ/FS from material, and failure, if it occurred, would be a
chart number 2, is 0.76. Hence, the factor of circular type. Insufficient time was available for
safety of the slope is 1.01. A number of trial ground water levels to be accurately established
calculations using Janbu’s method (Janbu et or for shear tests to be carried out. The stability
al., 1956) for the critical slide circle shown in analysis was carried out as follows:
Figure 8.14, found a factor of safety of 1.03. For the condition of limiting equilibrium, FS = 1
These factors of safety indicated that the and tan φ/FS = tan φ. By reversing the pro-cedure
stabil-ity of the slope was inadequate under outlined in Figure 8.5, a range of friction angles
the assumed conditions, and steps were taken were used to find the values of the ratio c/(γ H tan
to deal with the problem. ◦
φ) for a face angle of 42 . The value of the cohesion
c which is mobilized at failure, for a given friction
8.5.2 Example 2—highway slope angle, can then be calculated. This analysis was
carried out for dry slopes using chart number 1 (line
◦ B, Figure 8.15), and for saturated slopes using chart
A highway plan called for a cut at an angle of 42 .
The total height of the cut would be 61 m and it number 5 (line A, Figure 8.15).
Circular failure 187

0 .01 .02 .03

2.0 .05 .06
1.8 .10
1.6 .14
.17 c
1.4 H tan

FS .30
Slope angle
80 .45
0.6 .65
50 .70
40 .80
0.4 30 .90
0.2 10 2.0
0 8
0 .02 .04 .06 .08 .10 .12 .14 .16 .18 .20 .22 .24 .26 .28 c .30 .32 .34


Figure 8.10 Circular failure chart number 5—fully saturated slope.

Figure 8.15 shows the range of friction angles ◦

ratio c/(γ H tan φ) for a flatter slope of 30 , in the
and cohesions that would be mobilized at failure. ◦
same way as it was found for the 42 slope. The
The shaded circle (D) included in Figure 8.15 dashed line (C) in Figure 8.15 indicates the shear
indicated the range of shear strengths that were strength, which is mobilized in a dry slope with a
considered probable for the material under ◦
face angle of 30 . Since the mobilized shear
consideration, based upon the data presented in strength C is less than the available shear
Figure 4.21. This figure shows that the available strength D, the dry slope is likely to be stable.
shear strength may not be adequate to maintain
stability in this cut, particularly when the cut is 8.6 Detailed stability analysis
saturated. Consequently, the face angle could be
of circular failures
reduced, or ground water conditions investigated
to establish actual ground water pressures and The circular failure charts presented earlier in this
the feasibility of drainage. chapter are based upon the assumption that the
The effect of reducing the slope angle can be material forming the slope has uniform proper-
checked very quickly by finding the value of the ties throughout the slope, and that failure occurs
188 Circular failure


Location of center of 0.4

critical circle
0.3 = 10°

Ratio b/H
Y 0.2
Tension = 20 °
= 30°
H crack 0.1
= 40°
Drained slope 0
0 10 20 30 40 50 60 70 80 90
Failure through Slope face angle (°)
toe of slope Location of critical tension crack position

Distance X
−3H −2H −H 0 H 2H 3H
4H 4H
20˚ Friction angle
=30° = = 10°10°
3H 20 3H
30 
Distance Y

= 
40 40°
= 
2H 50 50°
 2H

0 0
−3H −2H −H 0 H 2H 3H
Location of center of critical circle for failure through toe

Figure 8.11 Location of critical sliding surface and critical tension crack for drained slopes.

along a circular slide path passing through the toe of

(1955) and the Janbu’s modified method of slices
the slope. When these conditions are not satis-fied,
(1954) are given in Figures 8.16 and 8.17 respect-
it is necessary to use one of the methods of slices
ively. Bishop’s method assumes a circular slide
published by Bishop (1955), Janbu (1954),
surface and that the side forces are horizontal; the
Nonveiller (1965), Spencer (1967), Morgenstern and
analysis satisfies vertical forces and overall moment
Price (1965) or Sarma (1979). This section
describes in detail the simplified Bishop and Janbu equilibrium. The Janbu method allows a slide
methods of stability analysis for circular failure. surface of any shape, and assumes the side forces
are horizontal and equal on all slices; the analysis
satisfies vertical force equilibrium. As pointed out by
8.6.1 Bishop’s and Janbu’s method of slices Nonveiller (1965), Janbu’s method gives reasonable
factors of safety when applied to shallow slide
The slope and slide surface geometries, and the surfaces (which are typ-ical in rock with an angle of
equations for the determination of the factor of ◦
safety by the Bishop’s simplified method of slices friction in excess of 30 and rockfill), but it is
seriously in error
Circular failure 189

Location of center of = 10°
critical circle

Ratio b/H
= 20°
Y 0.2
Ground water = 30°
H 0.1
Tension crack = 40° = 50°
Slope with ground water = 60°
(chart number 3) 0 10 20 30 40 50 60 70 80 90 Angle of slope face (°)
Failure through
toe of slope Location of critical tension crack position

Distance X
−3H −2H −H 0 H 2H 3H

3H 
Friction angle

20° =
Distance Y

2H Slope angle 30°

70° 60° 50°
H 

−3H −2H −H 0 H 2H 3H
Location of center of critical circle for failure through toe

Figure 8.12 Location of critical sliding surface and critical tension crack for slopes with ground water present.

X = 0.2 H
b = 0.1 H
Y = 1.85 H

Figure 8.13 Location of critical slide surface and critical tension crack for a drained slope at an
◦ ◦
angle of 30 in a material with a friction angle of 20 .
190 Circular failure

Input data for analysis:

Unit weight = 21.5 kN/m3
Friction angle = 37
Cohesion c = 6.9 kPa

Measured water
level 31 76.8 m

Critical failure
Figure 8.14 Slope profile
circle for Janbu
of China clay pit slope
considered in Example 1.

A—Saturated 42 slope the charts given in Figures 8.11 and 8.12 can
B—Dry 42 slope be used to estimate the center of the circle with
250 C—Dry 30 slope
D—Probable shear strength range for
the lowest factor of safety. In the Janbu
material in which slope is cut analysis, the slide surface may be defined by
200 (see Figure 4.21). known structural features or weak zones within
A the rock or soil mass, or it may be estimated in
c (kPa)

the same way as that for the Bishop analysis.

In either case, the slide surface assumed for

the first analysis may not give the lowest factor

B of safety, and a series of analyses are required

100 D
with variations on this pos-ition to find the
C surface with the lowest factor of safety.
50 Step 2: Slice parameters. The sliding mass
assumed in step 1 is divided into a number of
0 slices. Generally, a minimum of five slices should
0 10 20 30 40 50 be used for simple cases. For complex slope pro-
Friction angle () files, or where there are different materials in the
rock or soil mass, a larger number of slices may
Figure 8.15 Comparison between shear be required in order to define adequately the
strength mobilized and shear strength
available for slope considered in Example 2. prob-lem. The parameters which have to be
defined for each slice are as given here:

and should not be used for deep slide surfaces

• base angle ψb;
in materials with low friction angles. • the weight of each slice W is given by the
The procedures for using Bishop’s and product of the vertical height h, the unit
Janbu’s methods of slices are very similar and weight γr of the rock or soil and the width of
it is con-venient to discuss them together.
the slice x : W = (h γr x); and
• uplift water pressure U on the base of each
Step 1: Slope and slide surface geometry. The slice is given by the product of the height hw
geometry of the slope is defined by the actual or to the phreatic surface, the unit weight γw of
the designed profile as seen in a vertical section water and the width of the slice x, that is,
through the slope. In the case of a circular failure, U = (hw γw x).
Circular failure 191

X Center of rotation (see Figure 8.12)

Tension crack

Ground water
½ wz2 z/3

Typical slice

Failure through toe Water force = whw(x /cos b)

of slope

Factor of safety:
x /(1 + Y/FS) (8.3)
FS = Z + Q

X = [c + ( rh – whw) tan ] (x /cos b) (8.4)
Y = tan b tan (8.5)
Z = rh  x sin b (8.6)
Slice weight Q = ½ wz 2 ( /R ) (8.7)
= rhx

Note: angle b is negative when sliding uphill

b Figure 8.16 Bishop’s
simplified method of slices for
The following conditions must be satisfied for each slice:
the analysis of non-circular
rh – whw – c (tan b /FS) failure in slopes cut into
(1) = (8.8) materials in which failure is
1 + Y/FS
defined by the Mohr–Coulomb
(2) cos b (1 + Y/FS) > 0.2 (8.9) failure criterion.

Step 3: Shear strength parameters. The shear defined by non-linear failure criterion as dis-
strength acting on the base of each slice is cussed in Section 4.5, it is necessary to
required for the stability calculation. In the case of determ-ine the cohesion and friction angle for
a uni-form material in which the failure criterion is each slice at the effective normal stress for that
assumed to be that of Mohr–Coulomb (equa-tion slice (see Figure 4.23).
(1.1) in Section 1.4), the shear strength Step 4: Factor of safety iteration. When the
parameters c and φ will be the same on the base slice and shear strength parameters have been
of each slice. When the slope is cut in a number defined, the values of X, Y and Z are calculated
of materials, the shear strength para-meters for for each slice. The water force Q is added to Z,
each slice must be chosen according to the
the sum of the components of the weight of each
material in which it lies. When the shear strengths
slice acting parallel to the slide surface. An initial
of the materials forming the slope are
estimate of FS = 1.00 for the factor of safety is
192 Circular failure

Tension crack

Ground water
Typical slice
H ½ wz2 z/3

Failure through
toe of slope

d Water force = whw(x /cos b)

Factor of safety:
f0 X/(1 + Y/FS)
FS = (8.10)
Z + Q
X = [c + ( rh – whw) tan ] (1 + tan2 b) x (8.11)
Y = tan b tan (8.12)
Z = rh x tan b (8.13)
Q = ½ wz 2 (8.14) Figure 8.17 Janbu’s
hw modified method of
Slice weight
= r hx slices for the analysis
Note: angle b is negative when sliding uphill
of non-circular failure
in slopes cut into
Approximate correction factor f0 materials in which
f0 = 1 + K(d /L – 1.4(d /L)2) (8.15) for c = 0; K = failure is defined by
0.31 the Mohr–Coulomb
c > 0, > 0; K = 0.50 failure criterion.

used, and a new factor of safety is calculated from of each slice is always positive. If this con-
equations (8.3) and (8.10) given in Figures 8.16 and dition is not met for any slice, the inclusion of a
8.17, respectively. If the difference between the tension crack into the analysis should be
calculated and the assumed factors of safety is considered. If it is impossible to satisfy this
greater than 0.001, the calculated factor of safety is condition by readjustment of the ground water
used as a second estimate of FS for a new factor of conditions or the introduction of a tension
safety calculation. This process is repeated until the crack, the analysis as presented in Figure 8.16
difference between successive factors of safety is should be abandoned and a more elaborate
less than 0.001. For both the Bishop and the Janbu form of analysis, to be described later, should
methods, approximately seven iteration cycles will be adopted.
be required to achieve this result for most slope and Condition 2 in Figure 8.16 was suggested by
slide surface geometries. Whitman and Bailey (1967) and it ensures that
Step 5: Conditions and corrections. Figure 8.16 the analysis is not invalidated by conditions which
lists two conditions (equations (8.8) and (8.9)) can sometimes occur near the toe of a slope in
that must be satisfied for each slice in the Bishop which a deep slide surface has been assumed. If
analysis. The first condition ensures that the this condition is not satisfied by all slices, the
effective normal stress on the base slice dimensions should be changed and, if this
Circular failure 193

fails to resolve the problem, the analysis this procedure until the difference between
should be abandoned. successive factors of safety is less than 0.001.
Figure 8.17 gives a correction factor f0, which
Generally, about ten iterations will be required
is used in calculating the factor of safety by
means of the Janbu method. This factor allows to achieve the required accuracy in the
for inter-slice forces resulting from the shape of calculated factor of safety.
the slide surface assumed in the Janbu analysis.
The equation for f0 given in Figure 8.17 has been 8.6.3 Example of Bishop’s and
derived by Hoek and Bray (1981) from the curves Janbu’s methods of analysis
published in Janbu (1954).
A slope is to be excavated in blocky sandstone
with very closely spaced and persistent discon-
8.6.2 Use of non-linear failure criterion tinuities. The slope will consist of three, 15-m
in Bishop stability analysis high benches with two 8-m wide berms, the
primary function of which are to collect surface
When the material in which the slope is cut obeys runoff and control erosion (Figure 8.19). The
the Hoek–Brown non-linear failure criterion dis- ◦
bench faces will be at 75 to the horizontal, and
cussed in Section 4.5, the Bishop’s simplified
the slope above the crest of the cut will be at
method of slices as outlined in Figure 8.18 can be ◦
used to calculate the factor of safety. The an angle of 45 . The assumed position of the
following procedure is used, once the slice para- water table is shown on the figure. It is required
meters have been defined as described earlier for to find the factor of safety of the overall slope,
assuming that a circular type stability analysis
the Bishop and Janbu analyses:
is appropriate for these conditions.
The shear strength of the jointed rock mass is
1 Calculate the effective normal stress σ act-
based on the Hoek–Brown strength criterion, as
ing on the base of each slice by means of
discussed in Section 4.5, which defines the strength
the Fellenius equation (equation (8.17) on
as a curved envelope. The cohesion and friction
Figure 8.18).
angle for this criterion are calculated using the pro-
2 Using these values of σ , calculate tan φ
gram ROCLAB 1.004 (RocScience, 2002a), for
and c for each slice from equations (4.24)
which the input parameters are as follows:
and (4.25).
3 Substitute these values of tan φ and c into the • Very poor quality rock mass, GSI = 20;
factor of safety equation in order to obtain the
• Uniaxial compressive strength of intact rock
first estimate of the factor of safety. (from point load testing) ≈ 150 MPa;
4 Use this estimate of FS to calculate a new value
• Rock material constant, mi = 15;
of σ on the base of each slice, using the Bishop 3
• Unit weight of rock mass, γr = 0.025 MN/m ;
equation (equation (8.18) on Figure 8.18). 3
• Unit weight of water, γw = 0.00981 MN/m ;
5 On the basis of these new values of σ ,
• For careful blasting used in excavation, dis-
calcu-late new values for tan φ and c.
turbance factor D = 0.7; and
6 Check that conditions defined by equa-tions • Average slice height = 24 m (this height
(8.8) and (8.9) on Figure 8.16 are sat-isfied together with the rock mass unit weight
for each slice. defines the average vertical stress on the
7 Calculate a new factor of safety for the new sliding surface).
values of tan φ and c.
8 If the difference between the first and second Using these parameters, ROCLAB calculates, at the
factors of safety is greater than 0.001, return appropriate vertical stress level, a best fit line to the
to step 4 and repeat the analysis, using the curved strength envelope to define a friction angle
second factor of safety as input. Repeat ◦
of 43 and a cohesion of 0.145 MPa. This is
194 Circular failure

X Center of rotation

b Tension crack

Ground water
Y z
H 2 z/3
½ wz

Typical slice

Note: angle b is negative when sliding uphill

Failure through toe
of slope

x Factor of safety:

(ci + tan i ) (x /cos b)

FS = (8.16)
 rh x sin b + ½ wz 2 /R
= rh cos b – whw (Fellenius solution) (8.17)
rh – whw – (ci tan b/FS)
hw rh = (Bishop solution) (8.18)
1+ (tan i tan b/FS)

a –1
6am b (s + m b 3n)
= sin
a –1 (4.24)
b 2(1 + a)(2 + a) + 6 am (s + m )
b b 3n

a –1
ci [(1 + 2a)s + ( 1– a)m b 3n] ( + mb 3n)
c = a –1 (4.25)
(1 + a)(2 + a) – 1+ [6am (s + m s) ][(1 + a)(2 + a)]
b b 3n
where = = /
3n 3 max ci

The conditions which must be satisfied for each slice are:

(1) > 0, where is calculated by Bishop’s method
(2) cos b [1 + (tan b tan i )/FS] > 0.2

Figure 8.18 Bishop’s simplified method of slices for the analysis of circular failure in slope in material
in which strength is defined by non-linear criterion given in Section 4.5.

the equivalent Mohr–Coulomb shear strength. Table 8.1 shows the input parameters for the
In addition, the program calculates the Bishop and Janbu stability analyses assuming a
instantan-eous friction angle and cohesion linear shear strength, and for the Janbu analysis
corresponding to the effective normal stress on assuming a non-linear shear strength. The slope is
the base of each slice. divided into eight slices, and for each slice the base
Circular failure 195

angle, weight, pore pressure and width are meas- the slices, and the corresponding instantaneous
ured. The program SLIDE (RocScience, 2002b) is friction angle and cohesion are also calculated,
then used to calculate the factors of safety for the as shown in the last three columns of Table 8.1.
linear and non-linear shear strengths. In the case The calculated factors of safety estimates are
of the non-linear strength analysis, the val-ues of
Bishop simplified method of slices for = 1.39
the effective normal stress on the base of
Mohr–Coulomb shear strength
Janbu modified method of slices for = 1.26
Mohr–Coulomb shear strength
Critical center for = 45
Bishop simplified method of slices for = 1.39
non-linear shear strength
Note that the stability analysis of this slope
using numerical analysis methods is shown in
Section 10.4.1.

1 8.6.4 Circular failure stability

0 10 20 analysis computer programs
Scale - m
The circular failure charts discussed in Section 8.3
provide a rapid means of carrying out stability
4 n/R = 0.401 analyses, but are limited to simple conditions as
7 6 d /L = 0.117 illustrated in the examples. More complex ana-lyses
8 can be carried out using the Bishop and Janbu
methods discussed in Section 8.6.3, and the
Figure 8.19 Section of sandstone slope showing
water table, slice boundaries, tension crack and the purpose of providing details on the procedures is to
expected circular sliding surface. show the principles of the analyses.

Table 8.1 Calculated shear strength values of slices using Mohr–Coulomb and Hoek–Brown failure criteria

Slice parameters for all cases Mohr–Coulomb Non-linear failure criterion

values for Bishop values for Bishop analysis
and Janbu analyses

Slice Angle Slice Pore Slice Friction Cohesion, Base Inst. Inst.
member of slice height pressure, width angle, c (MPa) effective friction cohesion,
base, (MN) γw h w (m) normal angle, ci (MPa)
(degrees) (MPa) (degrees) stress, φi
σn (degrees)

1 25 1.312 0.017 6.1 43 0.145 0.139 53.8 0.068

2 29 1.597 0.047 6.1 43 0.145 0.169 53.8 0.068
3 34 2.603 0.071 6.1 43 0.145 0.270 49.8 0.096
4 39 2.635 0.087 6.1 43 0.145 0.261 51.3 0.085
5 44 3.501 0.095 6.1 43 0.145 0.323 49.0 0.104
6 50 3.914 0.087 6.1 43 0.145 0.324 48.6 0.107
7 57 3.592 0.047 6.1 43 0.145 0.240 50.3 0.092
8 65 2.677 0 6.1 43 0.145 0.109 54.5 0.064
196 Circular failure

There are, of course, computer programs slope excavated in sandstone, shale and
available to carry out stability analyses of siltstone above a proposed highway. For these
slopes, where the circular failure charts are not conditions the shape of the failure surfaces are
applicable. Important features included in influenced by the position and thickness of the
these programs, which allows them to be used beds of the weaker material.
for a wide range of conditions, are as follows:
8.6.5 Three-dimensional circular
• Slope face can include benches and a
failure analysis
variety of slope angles;
• Boundaries between the materials can be posi- The program XSTABL examines the stability of
tioned to define layers of varying thickness and a unit width slice of the slope, which is a two-
inclination, or inclusions of any shape; dimensional analysis that ignores any shear
• Shear strength of the materials can be stresses on the sides of the slice (this is the
defined in terms of Mohr–Coulomb or same principle that is used in the plane fail-ure
Hoek–Brown criterion; analysis described in Chapter 6). While two-
• Ground water pressures can be defined on dimensional procedures have found to be a
single or multiple water tables, or as reliable method of analysis, there may be cir-
specified pressure distributions; cumstances where three-dimensional analysis
• External loads, in any direction within the is required to define the slide surface and
plane of the slope cross-section, can be slope geometry more precisely. One program
posi-tioned at their correct location on the that provides a three-dimensional analysis is
slope. Such loads can include bridge and CLARA (Hungr, 1987), which divides the
building foundations and bolting forces; sliding mass into columns, rather than slices as
• Earthquake acceleration which is applied used in the two-dimensional mode. Figure 8.21
as a horizontal force in order to carry out shows an example of the CLARA analysis for a
pseudo-static stability analysis; partially saturated slope in which the water
• The shape and position of the slide surface table is below the bottom of the tension crack.
can be defined as a circular arc or straight
line segments;
8.6.6 Numerical slope stability analysis
• A search routine finds the slide surface with
the minimum factor of safety; This chapter has been concerned solely with
• Deterministic and probabilistic analysis the limit equilibrium method of analysis in
meth-ods that calculate the factor of safety which the factor of safety is defined by the ratio
and probability of failure, respectively. The of the resisting to the displacing forces on the
prob-abilistic analysis requires that the slide surface.
design para-meters be defined as An alternative method of analysis is to exam-
distributions rather than single values; ine the stresses and strains within the slope as a
• Error messages which identify negative means of assessing stability conditions. If the
stresses along the slide surface; and slope is close to failure, then a zone of high strain
• Drawing of slope showing slope geometry, will develop within the slope with a shape that will
material boundaries, ground water table(s) be approximately coincident with the circu-lar
and slide surface(s). slide surface. If the shear strength properties are
progressively reduced, there will be a sudden
One program that contains all these functions increase in the movement along the shear zone
is the program XSTABL (Sharma, 1991). An indicating that the slope is on the point of failure.
example of the output (partial) produced by The approximate factor of safety of the slope can
XSTABL is shown in Figure 8.20 for a benched be calculated from the ratio of the actual shear
Circular failure 197

1085 10 most critical surfaces

Surface of minimum Janbu Overburden
FS = 1.101


Y-axis (m)

Siltstone and


725 Shale
Siltstone and sandstone

0 90 180 270 360 450 540 630 720
X-axis (m)

Figure 8.20 Two-dimensional stability analysis of a highway cut using XSTABL.

8.7 Example Problem 8.1:

circular failure analysis

Surcharge on
crest of slope ◦
A 22-m high rock cut with a face angle of 60 has
Retaining wall been excavated in a massive, very weak vol-
canic tuff. A tension crack has opened behind the
crest and it is likely that the slope is on the point
Failure surface of failure, that is, the factor of safety is approx-
imately 1.0. The friction angle of the material is
m ◦ 3
10 5 0 estimated to be 30 , its density is 25 kN/m , and
the position of the water table is shown on the
sketch of the slope (Figure 8.22). The rock con-
tains no continuous joints dipping out of the face,
Figure 8.21 Three-dimensional stability analysis of and the most likely type of failure mode is circular
a slope incorporating a retaining wall and a
surcharge at the crest.

strength of the rock to the shear strength at
which the sudden movement occurred. This (a) Carry out a back analysis of the failure to
method of stability analysis is described in determine the limiting value of the
more detail in Section 10.3. cohesion when the factor of safety is 1.0.
198 Circular failure

(a) Tension crack chart number 3 (Figure 8.8) is used in the

analysis. When φ = 30 and FS = 1.0,
tan φ/FS = 0.58

Ground water table 22 m The intersection of this value for tan φ/FS

and the curve for a slope angle of 60 gives
Estimated slide surface

γ H FS = 0.086
(b) b = 2.9 m
∴ c = 0.086 × 25 × 22 × 1.0

= 47.3 kPa
X = –7.7 m
(b) If the slope were completely drained, circular
Y = 22 m failure chart 1 could be used for analysis.
c 47.3
γ H tan φ 25 × 22 × tan(30)
= 0.15

Figure 8.22 Slope geometry for Example The intersection of this inclined line with the
Problem 8.1: (a) slope geometry for circular ◦
curved line for a slope angle of 60 gives
failure with ground water table corresponding to
circular failure chart number 3; (b) position of tan φ
critical slide surface and critical tension crack. 0.52 FS

∴ FS = tan 30
(b) Using the strength parameters calculated in 0.52
(a), determine the factor of safety for a com-
= 1.11
pletely drained slope. Would drainage of the
slope be a feasible method of stabilization?
This factor of safety is less than that usually
(c) Using the ground water level shown in accepted for a temporary slope, that is, FS =
Figure 8.22 and the strength parameters 1.2, so draining the slope would not be an
cal-culated in (a), calculate the reduction effective means of stabilization.
in slope height, that is, amount of ◦
(c) When FS = 1.3 and φ = 30 , then
unloading of the slope crest required to
tan φ/FS = 0.44.
increase the factor of safety to 1.3.
On circular failure chart 3, the inter-
(d) For the slope geometry and ground water
section of this horizontal line with the curved
level shown in Figure 8.22, find the coordin-

ates of the center of the critical circle and line for a slope angle of 60 gives
the position of the critical tension crack. c
γ H FS = 0.11
∴ H = 25
× ×
1.3 0.11
(a) The ground water level shown in Figure 8.22
corresponds to ground water condition 3 in = 13.2 m
Table 8.4 in the manual, so circular failure This shows that the slope height must be
reduced by 8.8 m to increase the factor of
Circular failure 199

safety from 1.0 to 1.3. Note that a factor of Y=H

safety of 1.3 would only be achieved if the
ground water level dropped by an amount = 22 m, that is, 22 m above the toe
equivalent to the unloading.
(d) The critical circle and critical tension crack The location of the tension crack behind
for a slope with ground water present are the crest is
located using the graphs in Figure 8.12.

For a slope angle of 60 and a friction b/H = 0.13

angle of 30 , the coordinates of the center
of the circle are: b = 2.9 m

X = −0.35 H This critical circle is shown in Figure

= −7.7 m, that is, 7.7 m 8.22(b).
horizontally beyond the toe
Chapter 9

Toppling failure

9.1 Introduction along strike, is useful in the field identification

of topples.
The failure modes discussed in the three
Papers concerning field studies of toppling fail-
previous chapters all relate to sliding of a rock or
ures include de Freitas and Waters (1973) who
soil mass along an existing or induced sliding
discuss slopes in Britain, and Wyllie (1980) who
surface. This chapter discusses a different failure
demonstrates stabilization measures for toppling
mode—that of toppling, which involves rotation of
failures related to railway operations.
columns or blocks of rock about a fixed base.
Most of the discussion that follows in this
Similar to the plane and wedge failures, the
chapter is based on a paper by Goodman and
stability analysis of toppling failures involves, first,
Bray (1976) in which a formal mathematical
carrying out a kinematic analysis of the structural
solu-tion to a simple toppling problem is
geology to identify potential toppling conditions,
shown. This solution, which is reproduced
and then, if this condition exists, performing a
here, represents a basis for designing rock
stability analysis specific to toppling failures.
slopes in which top-pling is present, and has
One of the earliest references to toppling fail-ures
been further developed into a more general
is by Muller (1968) who suggested that block
design tool (Zanbak, 1983; Adhikary et al.,
rotation or toppling may have been a contribut-ory
1997; Bobet, 1999; Sageseta et al., 2001).
factor in the failure of the north face of the Vaiont
slide (Figure 9.1). Hofmann (1972) car-ried out a
number of model studies under Muller’s direction to 9.2 Types of toppling failure
investigate block rotation. Similar model studies Goodman and Bray (1976) have described a
carried out by Ashby (1971), Soto (1974) and Whyte number of different types of toppling failures that
(1973), while Cundall (1971), Byrne (1974) and may be encountered in the field, and each is dis-
Hammett (1974) who incor-porated rotational failure cussed briefly on the following pages. The
modes into computer analysis of rock mass import-ance of distinguishing between types of
behavior. Figure 9.2 shows a computer model of a toppling is that there are two distinct methods of
toppling failure in which the solid blocks are fixed stability analysis for toppling failures as described
and the open blocks are free to move. When the in the following pages—block and flexural
fixed blocks at the face are removed, the tallest toppling— and it is necessary to use the
columns of blocks topple because their center of appropriate analysis in design.
gravity lies outside the base. The model illustrates a
typical feature of toppling failures in which the
9.2.1 Block toppling
tension cracks are wider at the top than at the base.
This con-dition, which can best be observed when As illustrated in Figure 9.3(a), block toppling
looking occurs when, in strong rock, individual columns
Toppling failure 201

Figure 9.1 Suggested toppling mechanism of the north face of Vaiont slide (Muller, 1968).

Figure 9.2 Computer generated model of toppling failure; solid blocks are fixed in space while open
blocks are free to move (Cundall, 1971).

are formed by a set of discontinuities dipping steeply of rock, separated by well developed, steeply
into the face, and a second set of widely spaced dipping discontinuities, breaking in flexure as
orthogonal joints defines the column height. The they bend forward. Typical geological condi-
short columns forming the toe of the slope are tions in which this type of failure may occur are
pushed forward by the loads from the longer thinly bedded shale and slate in which ortho-
overturning columns behind, and this slid-ing of the gonal jointing is not well developed. Generally,
toe allows further toppling to develop higher up the the basal plane of a flexural topple is not as
slope. The base of the failure generally consists of a well defined as a block topple.
stepped surface rising from one cross joint to the Sliding, excavation or erosion of the toe of the
next. Typical geological con-ditions in which this slope allows the toppling process to start and it
type of failure may occur are bedded sandstone and retrogresses back into the rock mass with the
columnar basalt in which orthogonal jointing is well formation of deep tension cracks that become
developed. narrower with depth. The lower portion of the
slope is covered with disordered fallen blocks and
it is sometimes difficult to recognize a toppling
9.2.2 Flexural toppling
failure from the bottom of the slope. Detailed
The process of flexural toppling is illustrated in examination of toppling slopes shows that the
Figure 9.3(b) that shows continuous columns outward movement of each cantilevered column
202 Toppling failure

(a) (b)


Figure 9.3 Common classes of toppling failures: (a) block toppling of columns of rock containing widely
spaced orthogonal joints; (b) flexural toppling of slabs of rock dipping steeply into face; (c) block flexure
toppling characterized by pseudo-continuous flexure of long columns through accumulated motions
along numerous cross-joints (Goodman and Bray 1976).

produces an interlayer slip and a portion of the 9.2.4 Secondary toppling modes
upper surface of each plane is exposed in a
Figure 9.4 illustrates a number of possible
series of back facing, or obsequent scarps,
secondary toppling mechanisms suggested by
such as those illustrated in Figure 9.3(a).
Goodman and Bray. In general, these failures are
initiated by some undercutting of the toe of the
9.2.3 Block-flexure toppling slope, either by natural agencies such as scour
As illustrated in Figure 9.3(c), block-flexure or weathering, or by human activities. In all
toppling is characterized by pseudo-continuous cases, the primary failure mode involves sliding
flexure along long columns that are divided by or physical breakdown of the rock, and toppling is
numerous cross joints. Instead of the flexural induced in the upper part of the slope as a result
failure of continuous columns resulting in flex- of this primary failure (Figure 9.4(a) and (b)).
ural toppling, toppling of columns in this case Figure 9.4(c) illustrates a common occurrence of
results from accumulated displacements on the toppling failure in horizontally bedded sand-stone
cross-joints. Because of the large number of and shale formations. The shale is usu-ally
small movements in this type of topple, there significantly weaker and more susceptible to
are fewer tension cracks than in flexural weathering than the sandstone, while the sand-
toppling, and fewer edge-to-face contacts and stone often contains vertical stress relief joints. As
voids than in block toppling. the shale weathers, it undermines support for the
Toppling failure 203

(a) (b)

(c) (d) Tension cracks

2050 Circular sliding
Elevation (m)

at pit crest

1850 Sandstone

Figure 9.4 Secondary toppling modes: (a) toppling at head of slide; (b) toppling at toe of slide with shear
movement of upper slope (Goodman and Bray, 1976); (c) toppling of columns in strong upper material due
to weathering of underlying weak material; (d) toppling at pit crest resulting in circular failure of upper slope
(Wyllie and Munn, 1978).

sandstone and columns of sandstone, with 30 m occurred on the slope above the pit, result-
their dimensions defined by the spacing of the ing in cracks opening in the crest of the mountain
vertical joints, topple from the face. At some that were several meters wide and up to 9 m
locations the overhangs can be as wide as 5 deep. Continuous movement monitoring was
m, and failures of substantial volumes of rock used to allow mining to proceed under the moving
occur with little warning. slope, and finally the slope was stabilized by
The example of the slide base toppling mode back-filling the pit (Wyllie and Munn, 1979).
shown in Figure 9.4(d) is the failure of a pit slope A further example of the toppling mechan-
in a coal mine where the beds at the crest of the ism is illustrated in Figure 9.5 (Sjöberg, 2000).

pit dipped at 70 into the face, and their strike was In open pit mines where the depth of the slope
parallel to the face. Mining of the pit slope at an progressively increases, minor toppling move-
◦ ment may eventually develop into a substantial
angle of 50 initiated a toppling failure at the crest
of the pit, which in turn resulted in a circular failure. Careful monitoring of the movement,
failure that extended to a height of 230 m above and recognition of the toppling mechanism, can
the base of the topple. Detailed monitoring of the be used to anticipate when hazardous
slope showed that a total movement of about conditions are developing.
204 Toppling failure

Elastic Joint slip fully
rebound developed
Joint slip (exaggerated
Joints displacements)
redistribution Toe
New mining

Compression and
bending of columns
Movement on
slide surface

Tensile bending failure Tensile bending failure Displacements

at base of rotation propagated to crest starting from toe

Figure 9.5 Failure stages for large-scale toppling failure in a slope (Sjöberg, 2000).

9.3 Kinematics of block toppling failure that is, when

The potential for toppling can be assessed
ψp < φp (Stable) (9.1)
from two kinematic tests described in this
section. These tests examine first the shape of
but will topple when the center of gravity of the
the block, and second the relationship between
block lies outside the base, that is, when
the dip of the planes forming the slabs and the
face angle. It is emphasized that these two
x/y < tan ψp (Topple) (9.2)
tests are useful for identifying potential toppling
conditions, but the tests cannot be used alone For example, for a 3 m wide block on a base
as a method of stability analysis. ◦
plane dipping at 10 , toppling will occur if the
height exceeds 17 m.
9.3.1 Block shape test
9.3.2 Inter-layer slip test
The basic mechanics of the stability of a block on
a plane are illustrated in Figure 9.6(a) (see also A requirement for toppling to occur in the mech-
Figure 1.10). This diagram shows the conditions anisms shown in Figures 9.3 and 9.5 is shear
that differentiate stable, sliding or toppling blocks displacement on the face-to-face contacts on the
with height y and width x on a plane dipping at an
top and bottom faces of the blocks. Sliding on
angle ψp. If the friction angle between the base these faces will occur if the following conditions
of the block and the plane is φp, then the block are met (Figure 9.6(b)). The state of stress close
will be stable against sliding when the dip of the to the slope face is uniaxial with the direction of
base plane is less than the friction angle, the normal stress σ aligned parallel to the slope
Toppling failure 205

(a) (b)
x f


(c) (d) 20

(90 – f)

f d

(180 – f – d)

Figure 9.6 Kinematic conditions for flexural slip preceding toppling: (a) block height/width test for toppling;
(b) directions of stress and slip directions in rock slope; (c) condition for interlayer slip; (d) kinematic
test defined on lower hemisphere stereographic projection.

face. When the layers slip past each other, σ the field shows that instability is possible where
must be inclined at an angle φd with the normal the dip direction of the planes forming sides of

to the layers, where φd is the friction angle of the blocks, αd is within about 10 of the dip
the sides of the blocks. If ψf is the dip of slope direction of the slope face αf , or
face and ψd is the dip of the planes forming the
sides of the blocks, then the condition for ◦
interlayer slip is given by (Figure 9.6(c)): |(αf − αd)| < 10 (9.5)
The two conditions defining kinematic stability of
topples given by equations (9.4) and (9.5) can be
(180 − ψf − ψd) ≥ (90 − φd) (9.3)
depicted on the stereonet (Figure 9.6(d)). On the
or stereonet, toppling is possible for planes for
which the poles lie within the shaded area,
provided also that the base friction properties and
ψd ≥ (90 − ψf ) + φd (9.4) shape of the blocks meet the conditions given by
9.3.3 Block alignment test equations (9.1) and (9.2), respectively.

The other kinematic condition for toppling is that

9.4 Limit equilibrium analysis of
the planes forming the blocks should strike
toppling on a stepped base
approximately parallel to the slope face so that
each layer is free to topple with little constraint The method of toppling analysis described in this
from adjacent layers. Observations of topples in section utilizes the same principles of limiting
206 Toppling failure

a2 b
x s
n p


H ( f – p) d

f 2
a Figure 9.7 Model for
1 Topple
1 limiting equilibrium
Slide analysis of toppling on a
stepped base (Goodman
and Bray, 1976).

equilibrium that have been used throughout this a pseudo-static force acting on each block
book. While this method of analysis is limited to a (see Section 6.5.4), water forces can act on
few simple cases of toppling failure, it provides a the base and sides of each block, and loads
basic understanding of the factors that are produced by bridge foundations can be added
important in toppling, and allows stabilization to any specified block (Wyllie, 1999).
options to be evaluated. The stability analysis As an alternative to the detailed analysis
involves an iterative process in which the dimen- described in this section, Zanbak (1983)
sions of all the blocks and the forces acting on developed a series of design charts that can
them are calculated, and then stability of each is be used to identify unstable toppling slopes,
examined, starting at the uppermost block. Each and to estimate the support force required for
block will either be stable, toppling or sliding, and limiting equilibrium.
the overall slope is considered unstable if the
lowermost block is either sliding or toppling. A 9.4.1 Block geometry
basic requirement of this analysis is that the fric-
tion angle on the base of each block is greater The first step in toppling analysis is to calculate
the dimensions of each block. Consider the
than the dip angle of the base so that sliding on regu-lar system of blocks shown in Figure 9.7
the base plane does not occur in the absence of in which the blocks are rectangular with width x
any external force acting on the block (see and height yn. The dip of the base of the
equation (9.1)).
blocks is ψp and the dip of the orthogonal
The limit equilibrium method of analysis is planes forming the faces of the blocks is
ideally suited to incorporating external forces
ψd(ψd = 90 − ψp). The slope height is H , and
acting on the slope to simulate a wide variety of
actual conditions that may exist in the field. For the face is excavated at angle ψf while the
example, if the lower block or blocks are upper slope above the crest is at angle ψs.
unstable, then tensioned anchors with a specified
tensile strength and plunge can be installed in Angle of base plane (ψb). The base of the top-
these blocks to prevent movement. Also, ground pling blocks is a stepped surface with an overall
motion due to earthquakes can be simulated by dip of ψb (Figure 9.7). Note that there is no
Toppling failure 207

= x (1 – cos )
Upslope x
p Figure 9.8 Dilatancy of toppling
blocks with base plane
movement, coincident with normal to dip of
blocks (Zanbak, 1983).

explicit means of determining a value of the para- in the range

meter ψb. However, it is necessary to use an
appropriate value for ψb in the analysis because this ◦ ◦
has a significant effect on the stability of the slope. ψb ≈ (ψp + 10 ) to (ψp + 30 ) (9.6)
That is, as the base angle become flatter, the
lengths of the blocks increase and there is a greater
It is considered that an appropriate stability
tendency of the taller blocks to topple resulting in
analysis procedure for situations where the
decreased stability of the slope. If the base angle is value of ψb is unknown is to carry out a
coincident with the base of the blocks (i.e. ψb = ψp), sensitivity analysis within the range given by
then the geometry of top-pling requires dilatancy δ
equation (9.6) and find the value that gives the
of the blocks along the base plane and shearing on
least stable condition.
the faces of the blocks (Figure 9.8). However, if the Based on the slope geometry shown in
Figure 9.7, the number of blocks n making up
base is stepped (i.e. ψb > ψp), then each block can the system is given by
topple without dilatancy, provided there is
displacement on the face-to-face contacts (Figure
9.7). It is expected that more energy is required to n H cosec(ψ ) cot(ψb) − cot(ψf ) sin(ψ )
dilate the rock mass than to develop shear along = x b + sin(ψb − ψf ) s
existing dis-continuities, and so a stepped base is (9.7)
more likely than a planar base. Examination of base
friction, centrifugal and numerical models (Goodman
and Bray, 1976; Pritchard and Savigny, 1990, 1991; The blocks are numbered from the toe of the
Adhikary et al., 1997) show that base planes tend to slope upwards, with the lowest block being 1 and
be stepped, and the approximate dip angle is the upper block being n. In this idealized model,
the height yn of the nth block in a position below
208 Toppling failure

the crest of the slope is Pn are Mn and Ln on the upper and lower
faces respectively of the block, and are given
by the following.
yn = n (a1 − b) (9.8) If the nth block is below the slope crest, then
while above the crest
Mn = yn (9.13)
yn = yn−1 − a2 − b (9.9)
Ln = yn − a1 (9.14)
The three constants a1, a2 and b that are defined
by the block and slope geometry and are given by If the nth block is the crest block, then

1 = x tan(ψf − ψp) (9.10) Mn = yn − a2 (9.15)
a2 = x tan(ψp − ψs) (9.11)
Ln = yn − a1 (9.16)
b = x tan(ψb − ψp) (9.12)
If the nth block is above the slope crest, then
9.4.2 Block stability
Figure 9.7 shows the stability of a system of Mn = yn − a2 (9.17)
blocks subject to toppling, in which it is
possible to distinguish three separate groups Ln = yn (9.18)
of blocks according to their mode of behavior:
For an irregular array of blocks, yn, Ln and Mn
(a) A set of stable blocks in the upper part of can be determined graphically.
the slope, where the friction angle of the When sliding and toppling occurs, frictional
base of the blocks is greater than the dip forces are generated on the bases and sides of
of this plane (i.e. φp > ψp), and the height the blocks. In many geological environments, the
is limited so the center of gravity lies friction angles on these two surfaces are likely to
be different. For example, in a steeply dipping
inside the base (y/ x < cot ψp). sedimentary sequence comprising sandstone
(b) An intermediate set of toppling blocks where beds separated by thin seams of shale, the shale
the center of gravity lies outside the base. will form the sides of the blocks, while joints in
(c) A set of blocks in the toe region, which are the sandstone will form the bases of the blocks.
pushed by the toppling blocks above. For these conditions, the friction angle of the
Depending on the slope and block geometries,
sides of the blocks (φd) will be lower than friction
the toe blocks may be stable, topple or slide.
angle on the bases (φp). These two friction
Figure 9.9 demonstrates the terms used to define angles can be incorporated into the limit
the dimensions of the blocks, and the posi-tion equilibrium analysis as follows.
and direction of all the forces acting on the blocks For limiting friction on the sides of the block:
during both toppling and sliding. Figure 9.9(a)
shows a typical block (n) with the normal and
shear forces developed on the base (Rn, Sn), Qn = Pn tan φd (9.19)
and on the interfaces with adjacent blocks (Pn, Q
n−1 = Pn−1 tan φd (9.20)
Qn, Pn−1, Qn−1). When the block is one of the
toppling set, the points of application of all forces By resolving perpendicular and parallel to the
are known, as shown in Figure 9.9(b). The points base of a block with weight Wn, the normal and
of application of the normal forces shear forces acting on the base of block n are,
Toppling failure 209

(a) (b)
Qn Pn tan d
Pn Pn

P yn
Ln n

S tan b n Pn – 1 tan d


(c) Sn


n Wn

T Rn tan b
n–1 n–1

L Rn

Figure 9.9 Limiting equilibrium conditions for toppling and sliding of nth block: (a) forces acting on nth block;
(b) toppling of nth block; (c) sliding of nth block (Goodman and Bray, 1976).

respectively, When the block under consideration is one of

the sliding set (Figure 9.9(c)),

Rn = Wn cos ψp + (Pn − Pn−1) tan φd (9.21)

Sn = Rn tan φp (9.24)
Sn = Wn sin ψp + (Pn − Pn−1) (9.22)
However, the magnitudes of the forces Qn−1,
Considering rotational equilibrium, it is found Pn−1 and Rn applied to the sides and base of the
that the force Pn−1 that is just sufficient to
block, and their points of application Ln and Kn,
prevent toppling has the value are unknown. Although the problem is
indeterminate, the force Pn−1 required to pre-
Pn−1,t = [Pn(Mn − x tan φd) + (Wn/2) vent sliding of block n can be determined if it is
assumed that Qn−1 = (tan φd · Pn−1). Then the
(yn sin ψp − x cos ψp)]/Ln (9.23) shear force just sufficient to prevent sliding has
210 Toppling failure

the value the same procedure. It may be found that a

relatively short block that does not satisfy
P P Wn(cos ψp tan φp − sin ψp) equation (9.2) for toppling, may still topple
(1 − tan φp tan φd) if the moment applied by the thrust force on
n−1,s = n−
(9.25) the upper face is great enough to satisfy
the condition stated in (v) above. If the
9.4.3 Calculation procedure for toppling condi-tion Pn−1,t > Pn−1,s is met for all
stability of a system of blocks blocks, then toppling extends down to
block 1 and sliding does not occur.
The calculation procedure for examining toppling (vii) Eventually a block may be reached for
stability of a slope comprising a system of blocks which Pn−1,s > Pn−1,t . This establishes
dipping steeply into the faces is as follows:
block n2, and for this and all lower blocks,
the critical state is one of sliding. The
(i) The dimensions of each block and the stability of the sliding blocks is checked
number of blocks are defined using using equation (9.24), with the block being
equations (9.7)–(9.12). unstable if (Sn = Rn tan φb). If block 1 is
(ii) Values for the friction angles on the sides stable against both sliding and toppling (i.e.
and base of the blocks (φd and φp) are P0 < 0), then the overall slope is
assigned based on laboratory testing, or considered to be stable. If block 1 either
inspection. The friction angle on the base topples or slides (i.e. P0 > 0), then the
should be greater than the dip of the overall slope is considered to be unstable.
base to prevent sliding (i.e. φp > ψp).
(iii) Starting with the top block, equation (9.2)
is used to identify if toppling will occur, 9.4.4 Cable force required to stabilize a slope
that is, when y/ x > cot ψp. For the upper
toppling block, equations (9.23) and If the calculation process described in Section
(9.25) are used to calculate the lateral 9.4.3 shows that block 1 is unstable, then a
forces required to prevent toppling and tensioned cable can be installed through this
sliding, respectively. block and anchored in stable rock beneath the
(iv) Let n1 be the uppermost block of the zone of toppling to prevent movement. The
toppling set. design parameters for anchoring are the bolt
(v) Starting with block n1, determine the lat-eral tension, the plunge of the anchor and its
forces Pn−1,t required to prevent top-pling, position on block 1 (Figure 9.9(c)).
and Pn−1,s to prevent sliding. If Pn−1,t > Suppose that an anchor is installed at a
Pn−1,s, the block is on the point of toppling plunge angle ψT through block 1 at a distance
and Pn−1 is set equal to Pn−1,t , or if Pn−1,s L1 above its base. The anchor tension required
> Pn−1,t , the block is on the point of sliding to prevent toppling of block 1 is
and Pn−1 is set equal to Pn−1,s.
In addition, a check is made that there W1/2(y1 sin ψp − x cos ψp) + P1(y1 − x tan φd )
Tt = L1 cos(ψp + ψT )
is a normal force R on the base of the
block, and that sliding does not occur on (9.26)
the base, that is while the required anchor tension to prevent
sliding of block 1 is
Rn > 0 and (|Sn| > Rn tan φp)
P1(1 − tan φp tan φd) − W1(tan φp cos ψp − sin ψp)
(vi) The next lower block (n1 − 1) and all the Ts = tan φp sin(ψp + ψT ) + cos(ψp + ψT )

lower blocks are treated in succession using (9.27)

Toppling failure 211

When the force T is applied to block 1, the overturns by a small amount, there are edge-to-
normal and shear force on the base of the face contacts between the blocks, and the friction
block are, respectively, required to prevent further rotation increases.
Hence, a slope just at limiting equilibrium is meta-
R1 = P1 tan φd + T sin(ψp + ψT) + W1 cos ψp stable. However, rotation equal to 2(ψb − ψp) will
(9.28) convert the edge-to-face contacts along the sides
S1 = P1 − T cos(ψp + ψT) + W1 sin ψp of the columns into continuous face contacts and
the friction angle required to prevent further
rotation will drop sharply, possibly even below
that required for initial equilibrium. The choice of
The stability analysis for a slope with a tensioned factor of safety, therefore, depends on whether or
anchor in block 1 is identical to that described in not some deformation can be tolerated.
Section 9.4.3, apart from the calculations relating
The restoration of continuous face-to-face
to block 1. The required tension is the greater of
contact of toppled columns of rock is probably an
Tt and Ts defined by equations (9.26) and (9.27). important arrest mechanism in large-scale top-
pling failures. In many cases in the field, large
9.4.5 Factor of safety for limiting equilibrium surface displacements and tension crack forma-
analysis of toppling failures tion can be observed and yet the volumes of rock
that fall from the face are small.
For both reinforced and unreinforced slopes, the
factor of safety can be calculated by finding the
friction angle for limiting equilibrium. The pro-
cedure is first to carry out the limiting equilibrium 9.4.6 Example of limit equilibrium
stability analysis as described in Section 9.4.3 analysis of toppling
using the estimated values for the friction angles.
If block 1 is unstable, then one or both of the fric- The following is an example of the application
tion angles are increased by increments until the of the Goodman and Bray limit equilibrium ana-
value of P0 is very small. Conversely, if block 1 is lysis to calculate the factor of safety and
stable, then the friction angles are reduced until required bolting force of the toppling failure
illustrated in Figure 9.10(a).
P0 is very small. These values of the friction A rock face 92.5 m high (H ) is cut at an angle
angles are those required for limiting equilibrium. ◦
of 56.6 (ψf ) in a layered rock mass dipping at
The limiting equilibrium friction angles are ◦ ◦
termed the required friction angles, while the 60 into the face (ψd = 60 ); the width of each
block is 10 m ( x). The angle of the slope above
actual friction angles of the block surfaces are ◦
termed the available friction angles. The factor of the crest of the cut is 4 (ψs), and the base of the
safety for toppling can be defined by dividing the blocks is stepped 1 m at every block (atn (1/10) =
◦ ◦
tangent of the friction angle believed to apply 5.7 , and ψb = (5.7+ψp) = 35.7 ). Based on this
geometry, there are 16 blocks formed between
to the rock layers (tan φavailable), by the tangent the toe and crest of the slope (equation 9.7);
of the friction angle required for equilibrium block 10 is at the crest. Using equations (9.10)–
(tan φ ). (9.12), the constants are a1 = 5.0 m, a2 = 5.2 m
tan φ and b = 1.0 m. These constants are used to
calcu-late the height yn of each block, and the
FS = height to width ratio yn/ x as shown on the table
tan φrequired (9.30) in Figure 9.10(b).
The actual factor of safety of a toppling slope The friction angles on the faces and bases of
depends on the details of the geometry of the top- the blocks are equal and have a value of
pling blocks. Figure 9.7 shows that once a column ◦
38.15 (φavailable). The unit weight of the rock is
212 Toppling failure

(a) 5.8
13 14
11 12


6 n
4 n
1 T


n y P P
n yn /x Mn Ln n·t n·s Pn Rn Sn Sn /Rn Mode
16 4.0 0.4 0 0 0 866 500 0.577
15 10.0 1.0 0 0 0 2165 1250 0.577 STABLE
14 16.0 1.6 0 0 0 3463 2000 0.577 ________
13 22.0 2.2 17 22 0 0 0 4533.4 2457.5 0.542
12 28.0 2.8 23 28 292.5 –2588.7 292.5 5643.3 2966.8 0.526 T
11 34.0 3.4 29 34 825.7 –3003.2 825.7 6787.6 3520.0 0.519 O
10 35
40.0 4.0 35 1556.0 –3175.0 1556.0 7662.1 3729.3 0.487 P
9 36
36.0 3.6 31 2826.7 –3150.8 2826.7 6933.8 3404.6 0.491 P
32.0 3.2 32 27 3922.1 –1409.4 3922.1 6399.8 3327.3 0.520 L
28.0 2.8 28 23 4594.8 156.8 4594.8 5872.0 3257.8 0.555 I
5 24.0 2.4 24 19 4837.0 1300.1 4837.0 5352.9 3199.5 0.598 N
4 20.0 2.0 20 15 4637.5 2013.0 4637.5 4848.1 3159.4 0.652 G
3 16.0 1.6 16 11 3978.1 2284.1 3978.1 4369.4 3152.5 0.722
2 12.0 1.2 12 7 2825.6 2095.4 2825.6 3707.3 2912.1 0.7855 ________
1 8.0 0.8 8 3 1103.1 1413.5 1413.5 2471.4 1941.3 0.7855 SLIDING
4.0 0.4 4 – –1485.1 472.2 472.2 1237.1 971.8 0.7855

(c) Block
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

3 n



Figure 9.10 Limited equilibrium analysis of a toppling slope: (a) slope geometry; (b) table listing block
dimensions, calculated forces and stability mode; (c) distribution of normal (R) and shear (S) forces on
base of blocks (Goodman and Bray, 1976).

3 ◦
25 kN/m . It is assumed that the slope is dry, of the blocks is 38.15 and the dip of the base is
and that there are no external forces acting. ◦
30 , the upper blocks are stable against sliding.
The stability analysis is started by examining Equation (9.2) is then used to assess the toppling
the toppling/sliding mode of each block, starting mode. Since cot ψp = 1.73, blocks 16, 15 and 14
at the crest. Since the friction angle on the base are stable, because for each the ratio yn/ x is
Toppling failure 213

less than 1.73. That is, these three blocks are short defined by Rn > 0 and |Sn| < Rn tan φp are
and their center of gravity lies inside the base. satisfied everywhere.
For block 13, the ratio yn/ x has the value 2.2,
which is greater than 1.73 and the block topples. 9.4.7 Application of external
Therefore, P13 is equal to 0 and P12 is calcu- forces to toppling slopes
lated as the greater of P12,t and P12,s given by
There may be circumstances where there are
equations (9.23) and (9.25) respectively. This
calculation procedure is used to examine the external forces acting on the slope and it is
stability of each block in turn progressing down necessary to investigate their effect on stability.
the slope. As shown in the table of forces in Examples of external forces include water forces
Figure 9.10(b), Pn−1,t is the larger of the two acting on the sides and bases of the blocks, earth-
quake ground motion simulated as a horizontal
forces until a value of n = 3, whereupon Pn−1,s force acting on each block (see Section 6.5.4), and
is larger. Thus blocks 4 to 13 constitute the
point loads produced by bridge piers located on a
poten-tial toppling zone, and blocks 1 to 3
constitute a sliding zone. specific block(s). Another external force that has
The factor of safety of this slope can be been considered in the worked example in Section
found by increasing the friction angles until the 9.4.6 are rock anchors that are secured in stable
base blocks are just stable. It is found that the ground beneath the toppling mass and then
required friction angle for limit equilibrium con- tensioned against the face.
◦ A feature of limit equilibrium analysis is that
ditions is 39 , so the factor of safety as given any number of forces can be added to the
by equation (9.30) is 0.97 (tan 38.15/ tan 39). analysis, provided that their magnitude, direction
The analysis also shows that the required and point of application are known. Figure 9.11
tension in an anchor installed horizontally in shows a portion of a toppling slope in which there
block 1 to just stabilize the toe blocks is 500 kN is a sloping water table. The forces act-ing on
per meter length of slope. This compares with block n include the force Q inclined at an angle
the maximum value of P (in block 5) equal to ψQ below the horizontal, and three water forces
4837 kN/m.
V1, V2 and V3, as well as the forces Pn and
If tan φ is reduced to 0.650, it will be found
that blocks 1 to 4 in the toe region will slide Pn−1 produced by the blocks above and below.
while blocks 5 to 13 will topple. The tension in By resolving all these forces normal and parallel
an anchor installed horizontally through block to the base of the blocks, it is possible to modify
1, required to restore equilibrium, is found to be equations (9.23) and (9.25) as follows. Consider-
2013 kN/m of slope crest. This is not a large ing rotational equilibrium, the force Pn−1,t that is
number, demonstrating that support of the just sufficient to prevent toppling of block n has
“keystone” is remarkably effective in increas- the value:
ing the degree of stability. Conversely, remov-
Pn−1,t = {Pn(Mn − x tan φd)
ing or weakening the keystone of a toppling
slope that is near failure can have serious
+ Wn/2(yn sin ψp − x cos ψp)
When the distribution of P forces has been + V1yw /3 + γw x /6
defined in the toppling region, the forces Rn and
× cos ψp(zw + 2yw ) − V3zw /3
Sn on the base of the blocks can be calcu-lated
using equations (9.21) and (9.22). Assuming + Q[− sin(ψQ − ψp) x/2
[Qn−1 = Pn−1 tan φs], the forces Rn and Sn −1
can also be calculated for the sliding region. + cos(ψQ − ψp)yn]}Ln (9.31)
Figure 9.10(c) shows the distribution of these Assuming that the blocks are in a state of limiting
forces throughout the slope. The conditions equilibrium, the force just sufficient to prevent
214 Toppling failure

n d
M n


w d
n –

Wn 1

V n
1 P

b n
n Ln



Figure 9.11 Toppling block with external forces.

sliding of block n has the value The limit equilibrium stability analysis then
proceeds as before using the modified versions
Pn−1,s = Pn + {−W (cos ψp tan φp − sin ψp) of the equations for Pn−1,t and Pn−1,s.
+ V1 − V2 tan φp − V3 9.5 Stability analysis of flexural toppling
+ Q[− sin(ψQ − ψp) tan φp
Figure 9.3(b) shows a typical flexural toppling
+ cos(ψQ − ψp)]} failure in which the slabs of rock flex and
maintain fact-to-face contact. The mechanism of
× (1 − tan φp tan φd) (9.32) flexural toppling is different from the block top-
where pling mechanism described in Section 9.4. There-
fore, it is not appropriate to use limit equilibrium
stability analysis for design of toppling slopes.
1 2
V1 = 2 γw cos ψp · yw ; Techniques that have been used to study the sta-
V 1 bility of flexural toppling include base friction
2 = 2 γw cos ψp(yw + zw ) x
models (Goodman, 1976), centrifuges (Adhikary
1 2 et al., 1997) and numerical modeling (Pritchard
V3 = 2 γw cos ψpzw (9.33)
Toppling failure 215

(a) (b)

50 m

50 m


50 m

Figure 9.12 UDEC model of toppling pit slope: (a) pure flexural toppling deformation with grid point velocity
vectors; (b) contours of horizontal displacement for toppling slopes; (c) area of failed nodes due to flexure
(Pritchard and Savigny, 1990).

and Savigny, 1990, 1991). All these models information, for example, on the allowable face
show the common features of this failure angle for specific geological conditions and
mechanism including interlayer shearing, slope height.
obsequent scarps at the crest, opening of
Alternatively, computer simulations of block
tension cracks that decrease in width with
movement provide a means of studying a wide
depth, and a limiting dip angle ψb for the base range of geometric and material properties. One
of toppling. As discussed in Section 9.4.1, dip of the most suitable computer programs for these
angle ψb is steeper than the plane ψp (normal simulations is UDEC (Universal Distinct Element
◦ Code), which has been developed by Itasca
to the dip of the slabs) by about 10–30 .
The centrifuge modeling by Adhikary et al. has Consulting Group in Minnesota (Itasca, 2000).
been used to develop a series of design charts that Figure 9.12 shows the results of an analysis
relate stability to the slope face angle, the dip of the carried out on an open pit mine slope (Pritchard
blocks into the face and the ratio of the slope height and Savigny, 1990). The main features of UDEC
to the width of the slabs. Another input parameter is analysis for studying toppling slopes are that it
the tensile strength of the rock, because bending of
the slabs induces tensile crack-ing in their upper • incorporates a number of materials each
face. The design charts provide with differing strength properties;
216 Toppling failure

• recognizes the existence of contacts or failure is described by Davies and Smith (1993).
interfaces between discrete bodies, such as The toppling occurred in siltstones in which the
slabs of rocks formed by discontinuities dip- beds were very closely spaced and dipped at
ping steeply into the slope face; ◦ ◦
between 90 and 70 into the face. Excavation for
• calculates the motion along contacts by a bridge abutment resulted in a series of ten-sion
assigning a finite normal stiffness along the cracks along the crest, and stabilization of the
discontinuities that separate the columns of slope required the installation of tensioned rock
rock. The normal stiffness of a discontinuity bolts and excavation to reduce the slope angle.
is defined as the normal closure that occurs
on the application of a normal stress and 9.6 Example Problem 9.1:
can be measured from a direct shear test toppling failure analysis
(see Figure 4.17);
• assumes deformable blocks that undergo Statement
bending and tensile failure;
Consider a 6 m high slope with an overhanging
• allows finite displacements and rotations of ◦
the toppling blocks, including complete face at an angle of 75 There is a fault, dipping

detachment, and recognizes new contacts at an angle of 15 out of the face, at the toe of
automatically as the calculation progresses; the slope that is weathering and undercutting
• uses an explicit “time”-marching scheme to the face. A tension crack, which is wider at the
solve the equation of motion directly. This top than at the bottom, has developed 1.8 m
allows modeling of progressive failure, or behind the crest of the slope indicating that the
face is marginally stable (Figure 9.13). The
the amount of creep exhibited by a series ◦
of top-pling blocks for a chosen slope friction angle φ of the fault is 20 and the
condition, such as excavation at the toe of cohesion c is 25 kPa. The slope is dry.
the slope. Note that the time step in the
analysis is not actual time but a simulation Required
of progressive movement; and (a) Calculate the factor of safety of the block
• allows the user to investigate different against sliding if the density of the rock is
stabilization measures, such as installing 3
23.5 kN/m .
rock bolts or installing drain holes, to
(b) Is the block stable against toppling as
determine which scenario has the most
defined by the relation:
effect on block movements.
x/y > tan ψp—stable?
Because of the large number of input paramet-
ers that are used in UDEC and the power of the
analysis, the most reliable results are obtained if
the model can be calibrated against an existing
toppling failure in similar geological conditions to
those in the design slope. The ideal situation is in
mining operations where the development of the Y=6m
topple can be simulated by UDEC as the pit is
deepened and movement is monitored. This =15
allows the model to be progressively updated
with new data. Chapter 10 discusses numerical
x = 1.8 m
modeling of slopes in more detail.
The application of kinematic stability tests Figure 9.13 Toppling block illustrating Example
and reinforcement design for a flexural toppling Problem 9.1.
Toppling failure 217

(c) How much more undercutting of the fault (b) From the dimensions given on Figure 9.13,
must occur before toppling failure takes the following values are obtained to test
place? stability conditions:
(d) What stabilization measures would be
appropriate for this slope?
x/y = 0.3
Solution tan 15 = 0.27
(a) The factor of safety against sliding is
determined by the methods described in
The block is stable against toppling because
Chapter 6; the equation for a dry slope is
0.3 > 0.27.
cA + W cos ψp · tan φp (c) If weathering results in the width of the
FS base of the block being reduced by a
W sin ψp
further 0.2 m, toppling is likely to occur
= ((25 × 1.8) + (254 × cos 15 because: x/y = (1.8 − 0.2)/6 ≈ 0.27.
× tan 20))/(254 × sin 15) (d) Stabilization measures which could be
used on this slope include the following:
≈ 2.0

where • Prevention of ground water infiltration to

limit build-up of water pressure both in
A = base area of block the tension crack and on the fault at the
2 toe.
= 1.8 m /m • Application of reinforced shotcrete to the
W = weight of block/m fault to prevent further weathering.
• Trim blasting to reduce the slope angle
= 23.5 × 1.8 × 6
and the dimension y, if this can be
= 254 kN/m achieved without destabilizing the block.
Chapter 10

Numerical analysis
Dr Loren Lorig and Pedro Varona

10.1 Introduction (Young’s modulus and Poisson’s ratio) of the

material. Elastic–plastic models use strength
The previous four chapters discussed limit equi- parameters to limit the shear stress that a zone
librium methods of slope stability analysis for rock 2
bounded by specified slide planes. In con-trast, may sustain.
this chapter discusses numerical analysis The zones may be connected together,
methods to calculate the factor of safety without termed a continuum model, or separated by
pre-defining slide planes. These methods are discontinuities, termed a discontinuum model.
Discontinuum models allow slip and separation
more recent developments than limit equilibrium
at explicitly located surfaces within the model.
meth-ods and, at present (2003), are used
Numerical models tend to be general purpose in
predomin-ately in open pit mining and landslides
nature—that is, they are capable of solving a wide
studies, where interest often focuses on slope
variety of problems. While it is often desirable to
displace-ments rather than on the relative
have a general-purpose tool available, it requires
magnitude of resisting and displacing forces.
that each problem be constructed individually. The
Numerical models are computer programs that
zones must be arranged by the user to fit the limits
attempt to represent the mechanical response of
of the geomechanical units and/or the slope
a rock mass subjected to a set of initial conditions
geometry. Hence, numerical models often require
such as in situ stresses and water levels, bound-
more time to set up and run than special-purpose
ary conditions and induced changes such as
tools such as limit equilibrium methods.
slope excavation. The result of a numerical
There are several reasons why numerical
model sim-ulation typically is either equilibrium or
models are used for slope stability studies.
collapse. If an equilibrium result is obtained, the
resultant stresses and displacements at any point
in the rock mass can be compared with • Numerical models can be extrapolated
measured values. If a collapse result is obtained, confid-ently outside their databases in
the predicted mode of failure is demonstrated. comparison to empirical methods in which
Numerical models divide the rock mass into the failure mode is explicitly defined.
zones. Each zone is assigned a material model • Numerical analysis can incorporate key geo-
and properties. The material models are idealized logic features such as faults and ground water
stress/strain relations that describe how the providing more realistic approximations of
material behaves. The simplest model is a linear behavior of real slopes than analytic models.
elastic model, which uses the elastic properties
2 In numerical analysis the terms “elements” and “zones”
are used interchangeably. However, the term element is
1 Itasca Consulting Group, Inc., Minneapolis, Minnesota used more commonly in finite element analysis, and the
55401 USA. term zone in finite difference analysis.
Numerical analysis 219

In comparison, non-numerical analysis meth- strength to the reduced shear strength at failure.
ods such as analytic, physical or limit equi- This shear-strength reduction technique was
librium may be unsuitable for some sites or used first with finite elements by Zienkiewicz et al.
tend to oversimplify the conditions, possibly (1975) to compute the safety factor of a slope
leading to overly conservative solutions. composed of multiple materials.
• Numerical analysis can help to explain To perform slope stability analysis with the
observed physical behavior. shear strength reduction technique, simulations
• Numerical analysis can evaluate multiple are run for a series of increasing trial factors of
pos-sibilities of geological models, failure safety (f ). Actual shear strength properties,
modes and design options. cohesion (c) and friction angle (φ), are reduced
for each trial according to the equations
Many limit equilibrium programs exist to
determine factors of safety for slopes. These
execute very rapidly, and in the case of the method 1
of slices for circular failure, use an approximate ctrial = f c (10.1)
scheme in which a number of assumptions are 1
made, including the location and angle of inter-slice
forces (see Section 8.2). Several assumed slide φtrial = arctan f tan φ (10.2)
surfaces are tested, and the one giving the lowest
factor of safety is chosen. Equilibrium is satis-fied If multiple materials and/or joints are present, the
only on an idealized set of surfaces. With numerical reduction is made simultaneously for all materi-
models, a “full” solution of the coupled als. The trial factor of safety is increased
stress/displacement, equilibrium and constitutive gradually until the slope fails. At failure, the factor
equations is made. Given a set of properties, the of safety equals the trial factor of safety (i.e. f =
system is found either to be stable or unstable. By FS). Dawson et al. (1999) show that the shear
performing a series of simulations with vari-ous strength reduction factors of safety are generally
properties, the factor of safety can be found within a few percent of limit analysis solutions
corresponding to the point of stability. when an associated flow rule, in which the friction
The numerical analysis is much slower, but angle and dilation angle are equal, is used.
much more general. Only since the late 1990s, The shear strength reduction technique has two
with the advent of faster computers, has it main advantages over limit equilibrium slope sta-
become a practical alternative to the limit bility analyses. First, the critical slide surface is
equilibrium method. Even so, while limit found automatically, and it is not necessary to
equilibrium solu-tions may require just a few specify the shape of the slide surface (e.g. circular,
seconds, numerical solutions to large complex log spiral, piecewise linear) in advance. In general,
problems can take an hour or more, particularly the failure surface geometry for slopes is more
when discontinuum behavior is involved. The complex than simple circles or segmented sur-
third section of this chapter presents typical faces. Second, numerical methods automatically
safety factor analyses for the most common satisfy translational and rotational equilibrium,
discontinuum failure modes in rock slopes. whereas not all limit equilibrium methods do sat-isfy
For slopes, the factor of safety often is defined as equilibrium. Consequently, the shear strength
the ratio of the actual shear strength to the min- reduction technique usually will determine a safety
imum shear strength required to prevent failure. A factor equal to or slightly less than limit equilibrium
logical way to compute the factor of safety with a methods. Itasca Consulting Group (2002) gives a
finite element or finite difference program is to detailed comparison of four limit equilibrium
reduce the shear strength until collapse occurs. The methods and one numerical method for six different
factor of safety is the ratio of the rock’s actual slope stability cases.
220 Numerical analysis

Table 10.1 Comparison of numerical and limit equilibrium analysis methods

Analysis result Numerical solution Limit equilibrium

Equilibrium Satisfied everywhere Satisfied only for specific objects,

such as slices
Stresses Computed everywhere using field Computed approximately on certain
equations surfaces
Deformation Part of the solution Not considered
Failure Yield condition satisfied everywhere; Failure allowed only on certain
slide surfaces develop “automatically” pre-defined surfaces; no check on
as conditions dictate yield condition elsewhere
Kinematics The “mechanisms” that develop satisfy A single kinematic condition is
kinematic constraints specified according to the particular
geologic conditions

A summary of the differences between a assumed within deformable blocks. The most
numerical solution and the limit equilibrium widely used discrete element codes for slope
method is shown in the Table 10.1. stability studies are UDEC (Universal Distinct
Element Code; Itasca Consulting Group, 2000)
and 3DEC (3-Dimensional Distinct Element
Code; Itasca Consulting Group, 2003). Of fun-
10.2 Numerical models
damental importance in discontinuum codes is
All rock slopes involve discontinuities. Represent- the representation of joint or discontinuity
ation of these discontinuities in numerical models beha-vior. Commonly used relations for
differs depending on the type of model. There are represent-ing joint behavior are discussed later
two basic types of models: discontinuum models in this section.
and continuum models. Discontinuities in discon- Continuum codes assume material is con-
tinuum models are represented explicitly—that is, tinuous throughout the body. Discontinuities
the discontinuities have a specific orientation and are treated as special cases by introducing
location. Discontinuities in continuum mod-els are interfaces between continuum bodies. Finite
represented implicitly, with the intention that the 2
element codes such as PHASE (Rocscience,
behavior of the continuum model is sub-stantially 2002c) and its predecessor PHASES (P lastic
equivalent to the real jointed rock mass being H ybrid Analysis of Stress for Estimation of
represented. Support) and finite difference codes such as FLAC
Discontinuum codes start with a method designed (F ast Lagrangian Analysis of Continua; Itasca
specifically to model discontinua and treat Consulting Group, 2001) cannot handle gen-eral
continuum behavior as a special case. Dis- interaction geometry (e.g. many intersecting joints).
continuum codes generally are referred to as Their efficiency may degenerate drastic-ally when
Discrete Element codes. A Discrete Element code connections are broken repeatedly. Typical
will typically embody an efficient algorithm for continuum-based models may have less than ten
detecting and classifying contacts, and maintain a non-intersecting discontinuities. Of fun-damental
data structure and memory allocation scheme that importance to continuum codes and deformable
can handle many hundreds or thousands of blocks in discrete element codes is representation
discontinuities. The discontinuities divide the of the rock mass behavior. Continuum relations
problem domain into blocks that may be either rigid used to represent rock mass behavior are discussed
or deformable; continuum behavior is later in this section.
Numerical analysis 221

Finite element programs are probably more 10.2.1 Joint material models
familiar, but the finite difference method is
The material model used most commonly to
perhaps the oldest numerical technique used to
represent joints is a linear-elastic–perfectly-plastic
solve sets of differential equations. Both finite
model. The limiting shear strength is defined by
element and finite difference methods produce a
the usual Mohr–Coulomb parameters of friction
set of algebraic equations to solve. While the
angle and cohesion (see Section 4.2). A peak
methods used to derive the equations are differ-
and residual shear strength relation can also be
ent, the resulting equations are the same. Finite
spe-cified for the joints. The residual strength is
difference programs generally use an “explicit”
used after the joint has failed in shear at the peak
time-marching scheme to solve the equations,
strength. The elastic behavior of the joints is spe-
whereas finite element methods usually solve
cified by joint normal and shear stiffnesses, which
systems of equations in matrix form.
may be linear or piece-wise linear.
Although a static solution to a problem is usu-ally
of interest, the dynamic equations of motion are
10.2.2 Rock mass material models
typically included in the formulation of finite
difference programs. One reason for doing this is to It is impossible to model all discontinuities in a
ensure that the numerical scheme is stable when large slope, although it may be possible to model
the physical system being modeled is unstable. With the discontinuities for a limited number of
non-linear materials, there is always the possibility benches. Therefore, in large slopes much of the
of physical instability—for example, the failure of a rock mass must be represented by an equivalent
slope. In real life, some of the strain energy is continuum in which the effect of the discontinu-
converted to kinetic energy. Expli-cit finite-difference ities is to reduce the intact rock elastic properties
programs model this process directly, because and strength to those of the rock mass. This is
inertial terms are included. In contrast, programs true whether or not a discontinuum model is
that do not include inertial terms must use some used. As mentioned in the introduction to this
numerical procedure to treat physical instabilities. chapter, numerical models divide the rock mass
Even if the procedure is successful at preventing into zones. Each zone is assigned a mater-ial
numerical instability, the path taken may not be model and material properties. The material
realistic. The con-sequence of including the full law models are stress/strain relations that describe
of motion in finite difference programs is that the how the material behaves. The simplest model is
user must have some physical feel for what is a linear elastic model that uses only the elastic
happening. Explicit finite-difference programs are properties (Young’s modulus and Poisson’s ratio)
not black boxes that will “give the solution.” The of the material. Linear elastic–perfectly plastic
behavior of the numerical system must be stress–strain relations are the most commonly
interpreted. used rock mass material models. These models
FLAC and UDEC are two-dimensional finite- typically use Mohr–Coulomb strength paramet-
difference programs developed specifically for ers to limit the shear stress that a zone may
geomechanical analysis. These codes can simu- sustain. The tensile strength is limited by the spe-
late varying loading and water conditions, and cified tensile strength, which in many analyses is
have several pre-defined material models for rep- taken to be 10% of the rock mass cohe-sion.
resenting rock mass continuum behavior. Both Using this model, the rock mass behaves in an
codes are unique in their ability to handle highly isotropic manner. Strength anisotropy can be
non-linear and unstable problems. The three- introduced through a ubiquitous joint model,
dimensional equivalents of these codes are which limits the shear strength according to a
FLAC3D (F ast Lagrangian Analysis of Continua Mohr–Coulomb criterion in a specified direction.
in 3 Dimensions; Itasca Consulting Group, 2002) The direction often corresponds to a predominant
and 3DEC (Itasca Consulting Group, 2003). jointing orientation.
222 Numerical analysis

A more complete equivalent-continuum model to the failure criterion and that the flow rule is
that includes the effects of joint orientation and isotropic, whereas the Hoek–Brown criterion is not.
spacing is a micropolar (Cosserat) plasticity Recently, Cundall et al. (2003) has proposed a
model. The Cosserat theory incorporates a local scheme that does not use a fixed form of the flow
rotation of material points as an independent rule, but rather one that depends on the stress level,
parameter, in addition to the translation assumed and possibly some measure of damage.
in the classical continuum, and couple stresses Real rock masses often appear to exhibit pro-
(moments per unit area) in addition to the clas- gressive failure—that is, the failure appears to
sical stresses (forces per unit area). This model, progress over time. Progressive failure is a com-
as implemented in FLAC, is described in the plex process that is understood poorly and diffi-
context of slope stability by Dawson and Cundall cult to model. It may involve one or more of the
(1996). The approach has the advantage of using following component mechanisms:
a con-tinuum model while still preserving the
ability to consider realistic joint spacing explicitly. • Gradual accumulation of strain on principal
The model has not yet (as of 2003) been structures and/or within the rock mass;
incorporated into any publicly available code. • Increases in pore pressure with time; and
The most common failure criterion for rock • Creep, which is time-dependent deformation
masses is the Hoek–Brown failure criterion (see of material under constant load.
Section 4.5). The Hoek–Brown failure criterion is an
empirical relation that characterizes the stress Each of these components is discussed briefly
conditions that lead to failure in intact rock and rock later in the context of slope behavior.
masses. It has been used successfully in design Gradual accumulation of strain on principal
approaches that use limit equilibrium solutions. It structures within the rock mass usually results from
also has been used indirectly in numerical models excavation, and “time” is related to the excavation
by finding equivalent Mohr– Coulomb shear strength sequence. In order to study the pro-gressive failure
parameters that provide a failure surface tangent to effects due to excavation, one must either introduce
the Hoek–Brown fail-ure criterion for specific characteristics of the post-peak or post-failure
confining stresses, or ranges of confining stresses. behavior of the rock mass into a strain-softening
The tangent Mohr– Coulomb parameters are then model or introduce similar characteristics into the
used in traditional Mohr–Coulomb type constitutive explicit discontinuities. In practice, there are at least
relations and the parameters may or may not be two difficulties asso-ciated with strain-softening rock
updated during analyses. The procedure is awkward mass models. The first is estimating the post-peak
and time-consuming, and consequently there has strength and the strain over which the strength
been little direct use of the Hoek–Brown failure reduces. There appear to be no empirical guidelines
criterion in numerical solution schemes that require for estimat-ing the required parameters. This means
full constitutive models. Such models solve for dis- that the properties must be estimated through
placements, as well as stresses, and can continue calibration. The second difficulty is that, for a
the solution after failure has occurred in some loc- simulation in which the response depends on shear
ations. In particular, it is necessary to develop a localization and in which material softening is used,
“flow rule,” which supplies a relation between the the res-ults will depend on the zone sizes. However,
components of strain rate at failure. There have it is quite straightforward to compensate for this
been several attempts to develop a full con-stitutive form of mesh-dependence. In order to do this,
model from the Hoek–Brown criterion: for example, consider a displacement applied to the boundary of
Pan and Hudson (1988), Carter et al. (1993) and a body. If the strain localizes inside the body, the
Shah (1992). These formulations assume that the applied displacement appears as a jump across the
flow rule has some fixed relation localized band. The thickness of the band
Numerical analysis 223

contracts until it is equal to the minimum strength reduction process, these zones should
allowed by the grid, that is a fixed number of be considered as a new material with lower
zone widths. Thus, the strain in the band is strength, but no further softening should be
allowed due to the plastic strains associated to
the gradual reduction of strength.
ε = u/n z (10.3) Increases in pore pressure with time are not
where n is a fixed number, u is the common in rock slopes for mines. More com-
displacement jump, and z is the zone width. monly, the pore pressures reduce due to deepen-
If the softening slope is linear, the change in ing of the pit and/or drainage. However, there are
a property value p is proportional to strain, the cases in which the pore pressures do increase
change in property value with displacement is: with time. In such cases, the slope may appear to
fail progressively.
p s Creep, which is time-dependent deformation of
u=n z (10.4) material under constant load, is not commonly
· considered in the context of slope stability. It is
where s is the softening slope. much more common in underground excav-ations.
In order to obtain mesh-independent results, a Several material models are available to study creep
scaled softening slope s can be input, such that behavior in rock slopes. These include classical
viscoelastic models, power law models, and the
Burger-creep viscoplastic model. Applic-ation of a
s=s z (10.5) creep model to the study of slope beha-vior at
where s is constant. Chuquicamata mine in Chile is discussed later in
In this case, ( p/ u) is independent of z. If the this chapter (see Section 10.5.2).
softening slope is defined by the critical strain,
ε , then
crit 10.3 Modeling issues
1 Modeling requires that the real problem be ideal-
εcrit ∝ z (10.6) ized, or simplified, in order to fit the constraints
imposed by factors such as available material
For example, if the zone size is doubled, the models and computer capacity. Analysis of rock
critical strain must be halved for comparable mass response involves different scales. It is
results. impossible—and undesirable—to include all fea-
Strain-softening models for discontinuities are tures, and details of rock mass response
much more common than similar relations for mechan-isms, into one model. In addition, many
rock masses. Strain-softening relations for dis- of the details of rock mass behavior are unknown
continuities are built into UDEC and 3DEC, and and unknowable; therefore, the approach to
can be incorporated into interfaces in FLAC and model-ing is not as straightforward as it is, say, in
FLAC3D via a built-in programming language other branches of mechanics. This section
such as FISH functions. Strain-softening models discusses the basic issues that must be resolved
require special attention when computing safety when setting up a numerical model.
factors. If a strain-softening constitutive model is
used, the softening logic should be turned off
10.3.1 Two-dimensional analysis versus
during the shear strength reduction process or
three-dimensional analysis
the factor of safety will be underestimated. When
the slope is excavated, some zones will have The first step in creating a model is to
exceeded their peak strength, and some amount decide whether to perform two-dimensional
of softening will have taken place. During the or three-dimensional analyses. Prior to 2003,
224 Numerical analysis

three-dimensional analyses were uncommon, of 1. The average slope height was 100 m. Piteau
but advances in personal computers have and Jennings (1970) found that the average slope
permitted three-dimensional analyses to be angle for slopes with radius of curvature of 60 m
◦ ◦
performed routinely. Strictly speaking, three- was 39.5 as compared to 27.3 for slopes with a
dimensional analyses are radius of curvature of 300 m.
recommended/required in the following: Hoek and Bray (1981) summarize their experi-
ence with the stabilizing effects of slope
1 The direction of principal geologic structures curvature as follows. When the radius of
◦ curvature of a concave slope is less than the
does not strike within 20–30 of the strike of
the slope. ◦
height of the slope, the slope angle can be 10
2 The axis of material anisotropy does not steeper than the angle suggested by

strike within 20–30 of the slope. conventional stability ana-lysis. As the radius of
3 The directions of principal stresses are not curvature increases to a value greater than the
neither parallel nor perpendicular to the slope height, the correc-tion should be
slope. decreased. For radii of curvature in excess of
4 The distribution of geomechanical units twice the slope height, the slope angle given by a
varies along the strike of the slope. conventional stability analysis should be used.
5 The slope geometry in plan cannot be rep- To better quantify the effects of slope curvature
resented by two-dimensional analysis, which on stability, a series of generic analyses were
assumes axisymmetric or plain strain. per-formed. All analyses assumed a 500 m high

dry slope with a 45 face angle excavated in an
Despite the forgoing, most design analysis for iso-tropic homogeneous material with a density of
slopes assumes a two-dimensional geometry 3
2600 kg/m . Initial in situ stresses are assumed
com-prising a unit slice through an infinitely
to be lithostatic, and the excavation was made in
long slope, under plane strain conditions. In
40 m decrements beginning from the ground sur-
other words, the radii of both the toe and crest
face. For these conditions, pairs of friction angle
are assumed to be infinite. This is not the
and cohesion values were selected to produce a
condition encountered in practice—particularly factor of safety of 1.3 using circular failure chart
in open pit mining where the radii of curvature number 1 (see Section 8.3). A factor of safety of
can have an important effect on safe slope 1.3 is a value that is frequently used in the design
angles. Concave slopes are believed to be of slopes for open pit mines. The actual values
more stable than plain strain slopes due to the used are shown in Table 10.2.
lateral restraint provided by material on either A series of analyses was performed using FLAC
side of a potential failure in a concave slope. for different radius of curvature for both concave
Despite its potential importance in slope sta- and convex slopes. For concave slopes, the radius
bility, very little has been done to quantify this of curvature is defined as the distance between
effect. Jenike and Yen (1961) presented the
results of limit theory analysis of axisymmetric
slopes in a rigid, perfectly plastic material. Table 10.2 Janbu’s Lambda coefficient for various
However, Hoek and Brown (1981) concluded combinations of friction angle and cohesion
that the analysis assumptions were not
Friction angle Cohesion λ = γ H tan φ/c
applicable to rock slope design. (MPa)
Piteau and Jennings (1970) studied the influ-
ence of curvature in plan on the stability of slopes 45 0.22 59
in four diamond mines in South Africa. As a result 35 0.66 14
25 1.18 5
of caving from below the surface, slopes were all
15 1.8 2
at incipient failure with a safety factor
Numerical analysis 225

= 15 =

= 35
FS plane strain
FS slopes

= 45

Convex slopes Concave slopes

–10 –8 –6 –4 –2 0 2 4 6 8 10
Slope height/radius of curvature

Figure 10.1 Results of FLAC axisymmetric analyses showing effect on factor of safety of slope curvature.

the axis of revolution and the toe of the slope. beneficial effects of slope curvature should not
For the convex slopes, the radius of curvature be ignored—particularly in open pit mines,
is defined as the distance between the axis of where the economic benefits of steepening
revolu-tion and the crest of the slope—not the slopes can be significant. The same is true for
toe. Under both definitions, cones have a convex slopes, which are also more stable
radius of curvature of zero. than plane strain slopes. This goes against
Figure 10.1 shows the results with observed experience in rock slopes. If the slide
FS/FSplane strain versus height/radius of curvature surface is defined in terms of active (top) and
(H /Rc), which is positive for concave and negat- passive (bottom) wedges, the ratio of the
ive for convex slopes. The figure shows that the surface (and weight) of the passive wedge to
factor of safety always increases as the radius of the active wedge in a convex slope is greater
curvature decreases, but the increase is faster for than the plane strain condition. However, this
concave slopes. One unexpected result is that as only applies to a homo-geneous Mohr–
the friction angle increases, the effect of Coulomb material that might be found, for
curvature decreases. One possible explanation is example, in waste dumps. The reason why
that as Janbu’s lambda coefficient (λ = γ H tan “noses” in rock slopes are usu-ally less stable
φ/c) increases, the slide surface is shallower with may be related to the fact that they are more
only a skin for purely frictional material. This, exposed to structurally controlled failures.
makes the slope less sensitive to the confining
effect in concave slopes, and to the ratio of
10.3.2 Continuum versus
active/passive wedges for the convex ones.
discontinuum models
One reason that designers are reluctant to
take advantage of the beneficial effects of con- The next step is to decide whether to use a
cave slope curvature is that the presence of continuum code or a discontinuum code. This
discontinuities can often negate the effects. decision is seldom straightforward. There appear
However, for massive rock slopes, or slopes to be no ready-made rules for determining which
with relatively short joint trace lengths, the type of analysis to perform. All slope stability
226 Numerical analysis

problems involve discontinuities at one scale or 10.3.4 Initial conditions

another. However, useful analyses, particularly of
Initial conditions are those conditions that
global stability, have been made by assum-ing
existed prior to mining. The initial conditions of
that the rock mass can be represented as an
importance at mine sites are the in situ stress
equivalent continuum. Therefore, many analyses
field and the ground water conditions. The role
begin with continuum models. If the slope under
of stresses has been traditionally ignored in
consideration is unstable without structure, there
slope analyses. There are several possible
is no point in going to discontinuum models. If, on
reasons for this:
the other hand, a continuum model appears to be
reasonably stable, explicit incorporation of
principal structures should give a more accurate • Limit equilibrium analyses, which are widely
estimation of slope behavior. used for stability analyses, cannot include the
Selection of joint geometry for input to a model effect of stresses in their analyses. Neverthe-
is a crucial step in discontinuum analyses. Typic- less, limit equilibrium analyses are thought to
ally, only a very small percentage of joints can provide reasonable estimates of stability in
actually be included in a model in order to create many cases, particularly where structure is
models of reasonable size for practical analysis. absent, such as soil slopes.
Thus, the joint geometry data must be filtered to • Most stability analyses have traditionally
select only those joints that are most crit-ical to been performed for soils, where the range
the mechanical response. This is done by of pos-sible in situ stresses is more limited
identifying those that are most susceptible to slip than for rocks. Furthermore, many soil
and/or separation for the prescribed loading con- analyses have been performed for
dition. This may involve determining whether suf- constructed embankments such as dams,
ficient kinematic freedom is provided, especially where in situ stresses do not exist.
in the case of toppling, and calibrating the ana- • Most slope failures are gravity driven, and
lysis by comparing observed behavior to model the effects of in situ stress are thought to be
response. minimal.
• In situ stresses in rock masses are not
routinely measured for slopes, and their
10.3.3 Selecting appropriate zone size effects are largely unknown.
The next step in the process is to select an
appropriate zone size. The finite difference zones One particular advantage of stress analysis pro-
assume that the stresses and strains within each grams such as numerical models is their ability to
zone do not differ with position within the zone— include pre-mining initial stress states in stability
in other words, the zones are the lowest-order analyses and to evaluate their importance.
elements possible. In order to capture stress and In order to evaluate the effects of in situ
strain gradients within the slope adequately, it is stress state on stability, five cases were run
necessary to use relatively fine discretizations. using a simple model similar to the model
shown in Figure 10.2. For each of the five
By experience, the authors have found that at ◦
least 20 (and preferably 30) zones are required cases, the slope angle was 60 , the slope
over the slope height of interest. As discussed height was 400 m, the material density was
later, if flexural toppling is involved, a minimum of 3 ◦
2450 kg/m , the friction angle was 32 , and the
four zones across the rock column are required. cohesion was 0.92 MPa. The results of FLAC
Finite element programs using higher-order analyses are shown in Table 10.3.
elements likely would require less zones than the In general, it is impossible to say what effect
constant strain/constant stress elements common the initial stress state will have on any partic-
in finite difference codes. ular problem, as behavior depends on factors
Numerical analysis 227

FLAC (Version 3.40)


–138.9 < x < 2639 45
–1264 < y < 1514 500

Grid plot

0 500
Water table 0

axis (m)

Density = 2600 kg/m3

Friction angle = 35
Cohesion = 660 kPa

Horizontal axis (m)

500 1000 1500 2000

Figure 10.2 Problem geometry used to determine the effect of in situ stresses on slope stability.

Table 10.3 Effect of in situ stress on slope stability are not particularly important in slope
[x—horizontal in-plane direction; y—vertical in-plane studies.
direction; z—out-of-plane direction] • Initial horizontal stresses in the plane of ana-
In-plane Out-of-plane Factor of lysis that are less than the vertical stresses
horizontal stress horizontal stress safety tend to slightly decrease stability and reduce
σ σ σ σ the depth of significant shearing with respect
xx = yy zz = yy 1.30
to a hydrostatic stress state. This observa-tion
σxx = 2.0 σyy σzz = 2.0 σyy 1.30
σxx = 0.5 σyy σzz = 0.5 σyy 1.28 may seem counter-intuitive; smaller hori-
σxx = 2.0 σyy σzz = 0.5 σyy 1.30 zontal stresses would be expected to increase
σxx = 0.5 σyy σzz = 2.0 σyy 1.28 stability. The explanation lies in the fact that
the lower horizontal stresses actually provide
slightly decreased normal stress on potential
such as orientation of major structures, rock shearing surfaces and/or joints within the
mass strength and water conditions. However, slope. This observation was confirmed in a
some observations on the effects of in situ UDEC analysis of a slope in Peru where in
stress on stability can be made: situ horizontal stresses lower than the ver-tical
stress led to deeper levels of joint shear-ing in
• The larger the initial horizontal stresses, the toppling structures compared to cases
larger the horizontal elastic displacements. This involving horizontal stresses that were equal
is not much help, as elastic displacements to or greater than the vertical stress.
228 Numerical analysis

• It is important to note that the regional can be made regarding the displacement bound-
topography may limit the possible stress aries near the toe of any slope. One assumption
states, particularly at elevations above is that the displacements near the toe are
regional valley floors. Three-dimensional inhibited only in the horizontal direction. This is
models have been very useful in the past in the mech-anically correct condition for a problem
addressing some regional stress issues. that is perfectly symmetric with respect to the
plane or axis representing the toe boundary.
Strictly speak-ing, this condition only occurs in
10.3.5 Boundary conditions slopes of infinite length, which are modeled in
Boundaries are either real or artificial. Real two-dimensions assuming plane strain, or in
boundaries in slope stability problems corres-pond slopes that are axi-ally symmetric in which the pit
to the natural or excavated ground surface that is is a perfect cone. In reality, these conditions are
usually stress free. Artificial boundaries do not exist rarely satisfied. Therefore, some models are
in reality. All problems in geo-mechanics, including extended laterally to avoid the need to specify
slope stability problems, require that the infinite any boundary condi-tion at the toe of the slope. It
extent of a real prob-lem domain be artificially is important to note that difficulties with the
truncated to include only the immediate area of boundary condition near the slope toe are usually
interest. Figure 10.3 shows typical a result of the two-dimensional assumption. In
recommendations for locations of the artificial far- three-dimensional models, this difficulty generally
field boundaries in slope sta-bility problems. Artificial does not exist.
boundaries can be of two types: prescribed The far-field boundary location and condi-tion
displacement, or prescribed stress. Prescribed must also be specified in any numerical model for
displacement boundaries inhibit displacement in slope stability analyses. The general notion is to
either the vertical direction or horizontal direction, or select the far-field location so that it does not
both. Prescribed displace-ment boundaries are used significantly influence the results. If this criterion is
to represent the condi-tion at the base of the model met, whether the boundary is prescribed-
and toe of the slope. displacement or prescribed-stress is not important.
Displacement at the base of the model is always In most slope stability studies, a prescribed-
displacement boundary is used. The authors have
fixed in both the vertical and horizontal directions to
used a prescribed-stress bound-ary in a few cases
inhibit rotation of the model. Two assumptions
and found no significant differences with respect to
the results from a prescribed-displacement
boundary. The mag-nitude of the horizontal stress
W >W for the prescribed-stress boundary must match the
assumptions regarding the initial stress state in
order for the model to be in equilibrium. However,
following any change in the model, such as an
H excavation increment, the prescribed-stress
boundary causes the far-field boundary to displace
toward the excavation while maintaining its original
stress value. For this reason, a prescribed-stress
>H /2 bound-ary is also referred to as a “following” stress,
or constant stress boundary, because the stress
does not change and follows the displacement of
Figure 10.3 Typical recommendations for the boundary. However, following stresses are most
locations of artificial far-field boundaries in slope likely where slopes are cut into areas where the
stability analyses.
Numerical analysis 229

topography rises behind the slope. Even where concentrations near the toe of a slope, and slightly
slopes are excavated into an inclined over-predicts the pore pressure behind the toe by
topography, the stresses would flow around the ignoring the inclination of equipotential lines.
excavation to some extent, depending on the Seepage forces must also be considered in the
effective width of the excavation perpendicular analysis. The hydraulic gradient is the difference
to the downhill topographic direction. in water pressure that exists between two points
A summary of the effects of boundary condi- at the same elevation, and results from seepage
tions on analysis results is as follows: forces (or drag) as water moves through a porous
medium. Flow analysis automatically accounts for
• A fixed boundary causes both stresses and seepage forces.
displacements to be underestimated, whereas To evaluate the error resulting from specifying a
a stress boundary does the opposite. water table without doing a flow analysis, two
• The two types of boundary condition identical problems were run. In one case, a flow
“bracket” the true solution, so that it is pos- analysis was performed to determine the pore
sible to conduct tests with smaller models pressures. In the second case, the pressures were
to obtain a reasonable estimate of the true determined using only a piezometric surface that
solution by averaging the two results. was assumed to be the phreatic surface taken from
the flow analysis. The material properties and
A final point to be kept in mind is that all open pit geometry for both cases are shown in Figure 10.2.
slope stability problems are three-dimensional in The right-hand boundary was extended to allow the
reality. This means that the stresses acting in and far-field phreatic surface to coincide with the ground
around the pit are free to flow both beneath and surface at a horizontal distance of 2 km behind the
around the sides of the pit. Therefore, it is likely toe. Hydraulic conductivity within the model was
that, unless there are very low strength faults assumed to be homogeneous and iso-tropic. The
parallel to the analysis plane, a constant stress or error caused by specifying the water table can be
following stress boundary will over-predict the seen in Figure 10.4. The largest errors, under
stresses acting horizontally. prediction of up to 45%, are found just below the
toe, while over prediction errors in pore pressure
10.3.6 Incorporating water pressure
values behind the slope are gen-erally less than 5%.
The effect of water pressure in reducing effective The errors near the phreatic surface are
stresses and, hence, slope stability is well under- insignificant, as they result from the relatively small
stood. However, the effect of various assump- pore pressures just below the phreatic surface
tions regarding specification of pore pressure where small errors in small values result in large
distributions in slopes is not as well understood. relative errors.
Two methods are commonly used to specify pore For a phreatic surface at the ground surface at a
pressure distributions within slopes. The most distance of 2 km, a factor of safety of 1.1 is
rigorous method is to perform a complete flow predicted using circular failure chart number 3 (refer
analysis, and use the resultant pore pres-sures in to Section 8.3). The factor of safety determined by
the stability analyses. A less rigorous, but more FLAC was approximately 1.15 for both cases. The
common method is to specify a water table, and FLAC analyses give similar safety factors because
the resulting pore pressures are given by the the distribution of pore pressures in the area behind
product of the vertical depth below the water the slope where failure occurs is very similar for the
table, the water density and gravity. In this sense, two cases. The conclusion drawn here is that there
the water table approach is equivalent to is no significant differ-ence in predicted stability
specifying a piezometric surface. Both methods between a complete flow analysis and simply
use similar phreatic surfaces. However, the water specifying a piezometric surface. However, it is not
table method under-predicts actual pore pressure clear if this conclusion
230 Numerical analysis

FLAC (Version 4.00)

Legend 800
Pore pressures
underpredicted by
Step 20000 water table in this
Cons. time 1.3380E + 11 area
Pore pressure error
–1.5 0

i m
x (

s )
Contour interval = 5.00E-02
(zero contour omitted)



Pore pressures
overpredicted by
water table in this

Horizontal axis (m)

200 600 1000 1400 1800

Figure 10.4 Error in pore pressure distribution caused by specifying water table compared to performing a
flow analysis.

can be extrapolated to other cases involving, A reasonable approach regarding the

for example, anisotropic flow. number of excavation stages has evolved over
the years. Using this approach, only one, two
10.3.7 Excavation sequence or three excavation stages are modeled. For
each stage, two calculation steps are taken. In
Simulating excavations in numerical models the first step, the model is run elastically to
poses no conceptual difficulties. However, the remove any iner-tial effects caused by sudden
amount of effort required to construct a model removal of a large amount of material. Second,
depends directly on the number of excavation the model is run allowing plastic behavior to
stages simulated. Therefore, most practical ana- develop. Following this approach, reasonable
lyses seek to reduce the number of excavation solutions to a large number of slope stability
stages. The most accurate solution is obtained problems have been obtained.
using the largest number of excavation steps,
because the real load path for any zone in the
10.3.8 Interpretation of results
slope will be followed closely. In theory, it is
impossible to prove that the final solution is inde- As noted in the introduction, finite difference
pendent of the load path followed. However, for programs are not black boxes that “give the
many slopes, stability seems to depend mostly solution.” The behavior of the numerical system
on slope conditions, such as geometry and pore and results from finite difference models must be
pressure distribution at the time of analysis, and interpreted in much the same way as slope
very little on the load path taken to get there. movement data are interpreted. Finite difference
Numerical analysis 231

programs record displacements and velocities at • Toppling failure—block and flexural; and
nominated points within the rock mass. During • Flexural buckling failure.
the analysis, the recorded values can be
examined to see if they are increasing, remaining The two-dimensional distinct element code,
steady, or decreasing. Increasing displacements UDEC, was used for most of the analyses, and
and velocit-ies indicate an unstable situation; the characteristics of the slopes that were
steady displace-ments and decreasing velocities modeled are as follows:
indicate a stable situation. In addition, velocity
and displacement vectors for every point in the Slope height 260 m

model can be plot-ted. Fields of constant velocity Slope angle 55
and displacement indicate failure. Water pressure none/dry
The authors have found that velocities below Density 2660 kg/m
1e−6 indicate stability in FLAC and FLAC3D; 3
(26.1 kN/m )
conversely, velocities above 1e−5 indicate ◦
Rock mass friction angle 43
instability. Note that no units are given for velocit- Rock mass cohesion 675 kPa
ies. This is because the velocities are not real, due Rock mass tension 0
to the damping and mass scaling used to achieve Rock mass bulk modulus 6.3 GPa
static solutions. While the displacements are real, Rock mass shear modulus 3.6 GPa
the velocities are not, and there is no information on ◦
Joint friction angle 40
the “time” the displacement occurs.
It is also possible to examine the failure (plas- ◦
Dilation angle 0
ticity) state of points within the model, where A maximum zone size of 15 m was used
failure is defined as failure in tension or shear. except where noted. In all cases, the factor of
Care must be used in examining the failure state safety was estimated by simultaneously reducing
indicators. For example, local overstressing at the the rock mass properties and the joint friction
base of toppling columns can appear to form a until failure occurs using the procedure shown in
deep-seated slip surface when, in reality, it is just equa-tions (10.1) and (10.2). The safety factor
compressive failure of the columns. There-fore, was assumed to be the reciprocal of the
the failure (plasticity) indicators must be reviewed reduction required to produce failure. For
in the context of overall behavior before any example, if the strengths must be reduced by
definitive conclusions can be drawn. 25% (i.e. 75% of their “real” strength) in order to
achieve failure, the safety factor is 1.33.
10.4 Typical stability analysis
In this section, typical stability analyses for a vari- 10.4.1 Rock mass failure
ety of failure modes are discussed. The objective Numerical analysis of slopes involving purely
of this section is to show how numerical models rock mass failure is studied most efficiently using
can be used to simulate slope behavior, and com- 2
continuum codes such as PHASE , FLAC or
pute safety factors for typical problems. Modeling
FLAC3D. As mentioned in the previous section,
issues important in each of the following failure
discontinuities are not considered explicitly in
modes are discussed:
continuum models; rather, they are assumed to
be smeared throughout the rock mass. Assuming
• Rock mass failure; that rock mass shear strength properties can be
• Plane failure—daylighting estimated reasonably, the analysis is straightfor-
and non-daylighting; ward. The process for initially estimating the rock
• Wedge failure—daylighting mass properties is often based on empirical rela-
and non-daylighting; tions as described, for example, by Hoek and
232 Numerical analysis

UDEC (Version 3.20)



Boundary plot 400
User defined grid value


Displacement contour



interval = 0.5 300


Horizontal axis (m)

0 100 200 300 400 500 600

Figure 10.5 Rock mass failure mode for slope determined with UDEC.

Brown (1997). These initial properties are then σc = 150 MPa

modified, as necessary, through the calibration Disturbance factor, D = 0.7
Failure modes involve mainly shearing The tensile strength is estimated to be 0.012
through the rock mass. For homogeneous MPa. For the Bishop’s analysis method, the
slopes where the slide surface is often Mohr– Coulomb strength is estimated by fitting
approximately circular, intersecting the toe of a straight line to the curved Hoek–Brown
the slope and becom-ing nearly vertical near failure envelope at the normal stress level
the ground surface. The failure mode for the estimated from the slope geometry. Using this
parameters listed earlier is shown in Figure procedure, friction angle and cohesion were
10.5. The calculated safety factor is 1.64.
The following is a comparison of a slope ◦
stability analysis carried out using limit equilib- φ = 43
rium circular failure analysis (Bishop method) and c = 0.145 MPa
numerical stability analysis. In Chapter 8, the
stability of a benched slope in strong, but closely The mass density of the rock mass and water
fractured, sandstone including a water table and 3 3
were 2550 kg/m (25.0 kN/m ) and 1000
tension crack is described (see Figure 8.19). The 3 3
kg/m (9.81 kN/m ) respectively. The phreatic
rock mass is classified as a Hoek–Brown material
surface is located as shown in Figure 8.19.
with strength parameters:
Based upon these parameters, the Bishop
method produces a location for the circular
mi = 0.13 slide surface and tension crack, as shown in
Figure 8.19, and a factor of safety of 1.39.
GSI = 20
Numerical analysis 233

In using FLAC to analyze the stability of the FLAC solution includes the effect of stress
slope in Figure 8.19, the slide surface can evolve redis-tribution and progressive failure after
during the calculation in a way that is represent- movement has been initiated. In this problem,
ative of the natural evolution of the physical slide tensile failure continues up the slope as a
surface in the slope. It is not necessary to make result of the tensile softening. The resulting
an estimate for the location of the circular slide factor of safety allows for this weakening effect.
surface when beginning an analysis, as it is with
limit equilibrium methods. FLAC will find the slide
10.4.2 Plane failure—daylighting
surface and failure mechanism by simulat-ing the
and non-daylighting
material behavior directly. A reasonably fine grid
should be selected to ensure that the slide Failure modes that involve rigid blocks sliding
surface will be well defined as it develops. It is on planar joints that daylight in the slope face
best to use the finest grid possible when studying are most efficiently solved using analytical
prob-lems involving localized failure. Here, a zone methods. For comparison purposes, a UDEC
size of 2 m was used. ◦
analysis is per-formed for blocks dipping at 35
The FLAC analysis showed a factor of safety of out of the slope. The joints are assumed to
1.26, with the slide surface closely resembling have a cohesion of 100 kPa and a friction
that produced from the Bishop solution (Figure ◦
angle of 40 . The resulting safety factor is 1.32,
10.6). However, the tensile failure extends farther which agrees with the ana-lytic value given by
up the slope in the FLAC solution. It is important equation (6.4) in Chapter 6, assuming that no
to recognize that the limit equilibrium solution only tension crack forms. The plane failure mode in
identifies the onset of failure, whereas the the UDEC analysis is shown in Figure 10.7.

FLAC (Version 4.00)

Legend 80


Boundary plot
Vertical axis (m)

0 20
Max. shear strain increment
0 20
0.012 0
0.024 –20
Contour interval = 0.002

– 40

Horizontal axis (m)

– 40 –20 0 20 40 60 80 100

Figure 10.6 Failure mode and tension crack location determined with FLAC for slope in closely
fractured sandstone slope (refer to Figure 8.19).
234 Numerical analysis

UDEC (Version 3.10)

Cycle 487840 400
Block plot

axis (m)
Velocity vectors
Maximum = 0.043 300

0 0.2 200



Horizontal axis (m)

0 100 200 300 400 500 600 700

Figure 10.7 Plane failure mode with rigid blocks determined with UDEC.

If a tension crack does form, then the factor of analyses must be performed in three dimensions.
safety is slightly reduced. Deformable blocks with As with plane failure, sliding analysis of day-
elastic–plastic behavior are required to form lighting rigid blocks is best solved using analytic
tension cracks within the UDEC analysis. When methods, as described in Chapter 7. Analyses
deformable zones are used, the resultant safety involving formation of tension cracks and/or non-
factor is 1.27, similar to the value of 1.3 given by daylighting wedges require numerical analysis.
the analytic solution. The difference may be that Candidate codes include FLAC3D and 3DEC.
the analytic solution assumes a vertical tension The plasticity formulation in FLAC3D uses a
crack, whereas the UDEC analysis indicates that mixed discretization technique and presently
the tension crack curves where it meets the provides a better solution than 3DEC in cases
sliding plane (see Figure 10.8). where rock mass failure dominates. On the other
Similar analyses can be performed for non- hand, setting up problems involving more than
daylighting failure planes. In this case, failure one sliding plane in FLAC3D is more difficult and
involves sliding on discontinuities and shearing time consuming than similar problems in 3DEC.
through the rock mass at the toe of the slope, as
shown in Figure 10.9. Here, the cohesionless

slid-ing planes dip at 70 and are spaced 20 m
apart. The resultant safety factor is about 1.5.
10.4.4 Toppling failure—block and flexural
Toppling failure modes involve rotation and thus
usually are difficult to solve using limit equilib-
10.4.3 Wedge failure—daylighting
rium methods. As the name implies, block top-
and non-daylighting
pling involves free rotation of individual blocks
Analyses involving wedge failures are similar to (Figure 9.3(a)), whereas flexural toppling involves
those involving plane failures, except that the bending of rock columns or plates (Figure 9.3(b)).
Numerical analysis 235

UDEC (Version 3.20)


Cycle 541651 400

X displacement contours

Vertical axis (m)

Contour interval = 0.2
(zero contour line omitted)
0.4 200
1.0 100
1.8 0

Block plot

Horizontal axis (m)

0 100 200 300 400 500 600 700

Figure 10.8 Plane failure mode with deformable blocks determined with UDEC.

UDEC (Version 3.20)


Cycle 1246860
axis (m)
Time 1.286E + 03 sec
Y displacement contours

Contour interval = 0.2 200


(zero contour line omitted)

Block plot


Horizontal axis (m)

0 100 200 300 400 500 600 700

Figure 10.9 Non-daylighting plane failure mode determined with UDEC.