Computers and Geotechnics 32 (2005) 326–339

[Link]/locate/compgeo

Stability of earth–rockfill dams: Influence of geometry on the

three-dimensional effect
Yuzhen Yu *, Liquan Xie, Bingyin Zhang
Department of Hydraulic Engineering, Tsinghua University, Beijing 100084, China

Received 13 September 2004; received in revised form 22 March 2005; accepted 23 March 2005
Available online 24 May 2005

Abstract

A series of parametric studies were performed to investigate the three-dimensional (3D) effect on the stability of an earth–rockfill
dam, with respect to some major factors, such as geometrical characteristics of the dam and topography of the canyon. The length–
height ratio of the earth–rockfill dam clearly influences the 3D safety factor. A power function suitably expresses the relationship
between the increment ratio for the safety factor and the length–height ratio. The canyon shape, gradient of dam slope and height of
the dam significantly influence the 3D effect. The results may be used as a blueprint by combining the two-dimensional (2D) safety
factor in common use, and the relationship between 2D and 3D safety factors established in this study.
Ó 2005 Elsevier Ltd. All rights reserved.

Keywords: Earth–rockfill dam; Stability; Safety factor; Three-dimensional effect

1. Introduction of global stability is one of the key issues in the design of

a new earth–rockfill dam and the safety assessment of an
Most landslides display spatial but not cylindrical slip old one. However, slope stability has been mainly ana-
surfaces, and the three-dimensional (3D) effect on the lyzed using 2D methods, due to their simplicity. If the
stability of slopes has attracted much attention in recent 3D effect is ignored, costs may increase. The 3D effect
years [1–3]. Cai et al. [4] and Faheem et al. [5] have ana- is sometimes regarded as the safety reserve; nevertheless,
lyzed the 3D effect for excavations. 3D effects also exist it is difficult to estimate accurately, thereby obtaining a
in the stability of an earth–rockfill dam, which is one of reasonable balance between safety and cost. Therefore, a
the main types of artificial slope and the most popular quantitative and thorough investigation of the 3D effect
type of water-retaining structure. Lefebvre et al. [6] dis- on the stability of earth–rockfill dam is necessary.
cussed the effect of valley topography on the stresses Some theories and methods have been suggested for
within a dam, using 3D finite element method (FEM) calculating the 3D safety factor (Fs) of the slope. The
analyses, and concluded that two-dimensional (2D) 3D method of columns, based on the limit equilibrium
analyses were not applicable to the study of dams in val- theory, has been well established and developed, espe-
leys with steep wall slopes. cially over the last three decades [7–11]. However, these
Both safety and economic factors should be consid- methods need many assumptions and fail to give satis-
ered in the design of an earth–rockfill dam. Designers factory results under complicated working conditions
often work assiduously to decrease the inverse ratio of owing to limitations in describing geometry and material
slope (H:V) by as small as 0.1 to reduce costs. Analysis properties. On the other hand, with advances in com-
puter technology and numerical analysis, it is possible
*
Corresponding author. Fax: +86 10 6278 5593. to perform slope stability analysis using the FEM. In
E-mail address: yuyuzhen@[Link] (Y. Yu). particular, the shear strength reduction finite element

0266-352X/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/[Link].2005.03.003
 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339 327

method (SSR-FEM) makes it straightforward to con- loads, including seepage force and self-gravity, are
sider complex geometric and loading conditions, as well translated into nodal force. At this moment, the initial
as the elasto-plastic constitutive relationships of the value of F is assumed to be small enough to ensure al-
material without previously assuming the potential slid- most all the materials are in an elastic state. Then, the
ing surface [4,5,12–14]. However, the computation is rel- value of F is increased step-wise, in small increments
atively time-consuming and requires 3D mesh of, say 0.01, until, finally, global failure develops, i.e.,
discretization. Therefore, for stability analysis, neither the finite element calculation diverged under a physically
the 3D method of columns nor 3D FEM are, as yet, real convergence criterion. The global safety factor Fs is
popular in practice. considered to be the immediately previous value of the
2D stability analysis is easily performed and has the shear strength reduction factor F at which the iteration
benefit of practical experience. Therefore, an alternative limit is reached. For further details, see [4,12]. The solu-
measure to estimate 3D stability is to use the results of tion to the 2D problem is basically the same as the 3D
2D stability analysis combined with the 3D effect on problem, except that 2D equations are employed instead
the stability of an earth–rockfill dam. The key issue is of Eqs. (1) and (2). Home-coded software was employed
to establish the relationship between 2D and 3D to perform the analysis [19].
stability. Four test examples, denoted as TE1, TE2, TE3 and
Some researchers have examined the 3D effect of an TE4, were analyzed to evaluate the effectiveness of the
individual dam [15–18], but it has not been studied in de- SSR-FEM method used in the paper. TE1 and TE2 refer
tail, and, as yet, the relationship between 2D and 3D sta- to a homogenous sand slope with theoretical solution of
bility has not been established quantitatively. safety factor. TE3 and TE4 were extracted from [20],
This paper seeks a reasonable expression of the rela- with referee solutions.
tionship between 2D and 3D stability of an earth–rock- In TE1, a 2D homogenous sand slope, 10.0 m in
fill dam. A series of parametric studies were performed height with a slope angle of a = 45.0°, was taken as an
to ascertain the 3D effect on stability, with respect to example, as shown in Fig. 1. A series of analysis with
some major factors, such as the geotechnical properties different internal friction angles was performed. Mate-
of materials, geometrical characteristics of the dam and rial parameters are listed in Table 1. The comparison be-
topography of the canyon site. tween the calculated safety factors, obtained by the
SSR-FEM, and the theoretical solutions, determined
by Fs = tan //tan a, is shown in Fig. 2. As can be seen,
2. Calculation method and verification the calculated values agree very well with the theoretical
solutions.
FEM, based on the concept of shear strength reduc-
tion (SSR-FEM), was employed to analyze global stabil-
ity of the dam. The elasto-perfectly plastic constitutive
model was applied to the soil and the Mohr–Coulomb
principle was accepted as the yield criteria, as well as
the Drucker–Prager equation as the plastic potential
function. Under 3D conditions, the Mohr–Coulomb
principle can be expressed as
1h pffiffiffi i
f ðI 1 ; J 2 ; hÞ ¼ 3ð1
sin /Þ sin h þ 3ð3 þ sin /Þ cos h
2 pffiffiffiffiffi
 J 2
I 1 sin /
3c cos / ¼ 0; ð1Þ
where I1 is the first invariant of stress, J2 is the second
invariant of deviatory stress, h is LodeÕs angle, c is the
cohesion intercept and / is the internal angle of friction.
The Drucker–Prager equation is
pffiffiffiffiffi
gðI 1 ; J 2 Þ ¼
aI 1 þ J 2 ; ð2Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 Fig. 1. Configuration and mesh of the sand slope in TE1.
where a ¼ tan w= 9 þ 12 tan w and w is the dilation
angle.
In the SSR-FEM, the reduced shear strength param-
Table 1
eter cF and /F replace the corresponding values of c and Material properties for TE1 and TE2
/ in Eq. (1) by cF = c/F and /F = //F, where F is the
c (kN/m3) c (kN/m2) u (°) w (°) E (kN/m2) v
shear strength reduction factor. Firstly, to perform the
18 0.0 25–50 0 50 000 0.3
FEM analysis and obtain the initial stress state, all
328 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339

1.4 1.2

1.2 1.1

1.0
1.0
Fs

0.8
0.9
0.6 Theoretical

Fs -3D
Calculated 0.8
0.4

0.7
0.2
20 25 30 35 40 45 50 55
Line Fs-3D=Fs-2D
Internal angel of friction φ 0.6
Comparison
Fig. 2. Comparison between the calculated and theoretical safety
factors of TE1. 0.5

TE2 is a 3D problem and the slope body is con- 0.4

0.4 0.6 0.8 1.0 1.2
structed by expanding the 2D section of TE1 to 60 m
Fs -2D
in length, as shown in Fig. 3. At both ends, the bound-
ary is fixed in the direction perpendicular to the section Fig. 4. Comparison between the calculated 3D and 2D safety factors
and free in other directions. Therefore, if the same mate- of TE1 and TE2.
rial properties are used, the 3D safety factors should be
the same as 2D factors. Fig. 4 shows the comparison be-
tween the calculated 3D and 2D safety factors. It can be
seen that the calculated points are distributed very clo- 40
sely around the 1:1 line. (50, 35) (70, 35)
TE3 and TE4 correspond, respectively, to EX1a and
(m)

EX1c of the 10 standard test problems of the soil slope 30

stability programs review project [20] organized by the
Australian Association for Computer-Aided Design (30, 25)
(20, 25)
(ACADS). Many users and some authors of various
20
slope stability programs predicted the solutions to these 20 30 40 50 60 70
standard test problems; then, Donald and Giam [20] gave (m)

a comprehensive review and issued the Ôreferee answersÕ. Fig. 5. Configuration of the earth slope, with only one soil layer, of
The configurations for the two test examples are TE3 (EX1a).
shown in Figs. 5 and 6, and the material properties are
given in Tables 2 and 3, respectively. The safety factors
of the slope in TE3 and TE4 are calculated by the SSR-
FEM program and compared with results, collected by
40
(50, 35) (70, 35)

Material-1
30 (50, 29)
(54, 31) (70, 31)
(30, 25) Material-2
(40, 27)
(20, 25) Material-3 (52, 24) (70, 24)
20
20 30 40 50 60 70
(m)

Fig. 6. Configuration of the earth slope, with three soil layers, of TE4
(EX1c).

Table 2
Material properties for TE3 (EX1a)
c (kN/m2) u (°) c (kN/m3) E (kN/m2) m K0
3.0 19.6 20.0 1.0 E 4 0.25 0.65
Fig. 3. Mesh of the 3D sand slope in TE2.
 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339 329

Table 3 the dam and geotechnical properties of the materials,

Material properties for TE4 (EX1c) as well as load condition of the earth dam.
Material c (kN/m2) u (°) c (kN/m3) E (kN/m2) m K0 In this paper, only the final construction stage, without
1 0.0 38.0 19.5 1.0 E 4 0.25 0.65 water storage, and the stable seepage stage, with full stor-
2 5.3 23.0 19.5 1.0 E 4 0.25 0.65 age, are considered. The downstream slope is usually
3 7.2 20.0 19.5 1.0 E 4 0.25 0.65 slightly steeper than the upstream one, so the downstream
slope is the control side and will be emphasized hereafter.
For convenience, several parameters, relative to 3D
Table 4
Results of TE3 and TE4 and comparison with the recommended
effect, canyon topography and geometry of the dam,
results [20] are defined as follows.
Cases Values of safety factors
The geometrical characteristics of an idealized earth–
rockfill dam, as shown in Fig. 7, are described by three
Range of the Referee SSR-FEM
submitted values answers
parameters, i.e., maximum height of the dam H, length–
height ratio g, as expressed in Eq. (3), and the gradient
TE3 (EX1a) 0.94–1.08 1.00 1.00
TE4 (EX1c) 1.23–1.52 1.39 1.42
ratio 1:r (V:H) of the downstream slope.

Donald and Giam [20], as shown in Table 4. It can be

seen that the calculated values of Fs by the SSR-FEM
is in the middle range submitted by various researchers,
and equal or very close to the referee answers.
The validity and effectiveness of the program used in
this paper is verified by the four test examples.

3. Case analyses

The 3D stability of the slope depends on canyon

topography at the site, geometrical characteristics of Fig. 7. Sketch of an idealized dam.

Table 5
Combined cases
Groups Subgroups Numbers of g S H (m) r Materials Working conditions
Group 1 Subgroup 1 23 0.5 50 2.0 M1 FCS
Subgroup 2 23 M2
Subgroup 3 23 M3
Group 2 Subgroup 4 23 0.5 50 2.0 M1 SSS
Subgroup 5 23 M2
Subgroup 6 23 M3
Group 3 Subgroup 7 16 0.7 50 2.0 M1 FCS
Subgroup 8 16 M2
Subgroup 9 16 M3
Group 4 Subgroup 10 16 0.9 50 2.0 M1 FCS
Subgroup 11 16 M2
Subgroup 12 16 M3
Group 5 Subgroup 13 16 0.5 100 2.0 M1 FCS
Subgroup 14 16 M2
Subgroup 15 16 M3
Group 6 Subgroup 16 17 0.5 150 2.0 M1 FCS
Subgroup 17 17 M2
Subgroup 18 17 M3
Group 7 Subgroup 19 15 0.5 50 1.60 M1 FCS
Subgroup 20 15 1.70
Subgroup 21 15 1.75
Subgroup 22 15 1.80
Subgroup 23 15 1.90
FCS, the final construction stage; SSS, the stable seepage stage.
330 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339

L Table 6
g¼ ; ð3Þ Material parameters
H
Sets Materials c c u w E m k
where L is the crest length of the dam.
(kN/m3) (kPa) (°) (°) (kN/m2) (m/s)
The topography of the canyon can be characterized
M1 Mat-1 20 12 37 0.0 70,000 0.3 1.0 E
5
by the shape index S
Mat-2 18 20 29 0.0 30,000 0.3 1.0 E
9
Lm Mat-3 19 10 35 0.0 70,000 0.3 1.0 E
5
S¼ ; ð4Þ
L M2 Mat-1 21 12 34 0.0 60,000 0.3 1.0 E
5
where Lm is the average length of the dam, as shown in Mat-2 19 20 25 0.0 25,000 0.3 1.0 E
9
Mat-3 20 10 30 0.0 60,000 0.3 1.0 E
5
Fig. 7. When the shape of the canyon is simplified as a
trapezoid, Lm can be expressed as Lm = (L + Lb)/2, M3 Mat-1 20 10 33 0.0 70,000 0.3 1.0 E
5
Mat-2 18 18 23 0.0 30,000 0.3 1.0 E
9
where Lb is the toe length of the dam. It is obvious that
Mat-3 19 8 28 0.0 70,000 0.3 1.0 E
5
the shape of the canyon is a triangle or ÔV-shapedÕ, when
S = 0.5, and ÔU-shapedÕ when S is close to 1.0.
The increment ratio of the safety factor, denoted by k
in Eq. (5), is a parameter used to quantitatively reflect
the 3D effect on stability of earth–rockfill dams.
F s3
F s2
k¼ . ð5Þ
F s2
F s2 is the safety factor of the typical cross-section for the
2D condition and F s3 is the safety factor for the 3D
condition.
Several factors, which could potentially influence the
3D effect in the stability of an earth–rockfill dam, are
considered, including the topographic characteristics of
the canyon, i.e., shape index S, the working condition
of the dam (with or without seepage), and the geometri-
cal characteristics of the earth–rockfill dam, i.e., the
maximum height of the dam H, length–height ratio g
and the gradient ratio 1:r (V:H) of the downstream
slope. Moreover, the geotechnical properties of materi-
als were varied to provide a range of 2D safety factor
F s2 . All the analysis cases, combined from the above fac-
tors, are shown in Table 5, seven groups in total, includ-
ing 23 subgroups with 408 cases. As expected intuitively,
the length–height ratio g should be the most apparent
influencing factor on the 3D effect, so a series of values
of length–height ratio g, ranging from 1.2 to 6.0, is
adopted in each subgroup.
The dam includes three material zones, as shown in
Fig. 11. Three sets of material parameters, correspond-
ing to different F s2 , are used, as shown in Table 6. The
influence of F s2 on increment ratio of safety factor k
can be investigated by comparing the results of each
subgroup in groups 1–6. By contrasting the results of Fig. 8. Meshes of dams with different L–H ratios (S = 0.5, H = 50 m).
group 1 with those of group 2, the 3D effect under
two working conditions, i.e., the final construction
stage without water storage and the stable seepage 50, 100 and 150 m, are considered. By comparing the
stage with full storage, can be studied. The storage results of groups 1, 5 and 6, the impact of the dam
water levels are shown in Fig. 11. Three kinds of typ- height on the 3D effect can be studied. Six gradients
ical canyon topography are considered, i.e., S = 0.5, of the downstream slope of the dam are employed,
S = 0.7 and S = 0.9, as shown in Figs. 8–10, respec- i.e., r = 1.6, 1.7, 1.75, 1.8, 1.9 and 2.0. By comparing
tively. By comparing the results of groups 1, 3 and 4, the results of subgroup 1 in group 1 and each subgroup
the influence of canyon topography on the 3D effect in group 7, the influence of the ratio of slope on the 3D
can be made clear. Three maximum heights of dam, effect can be investigated.
 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339 331

Fig. 9. Meshes of dams with different L–H ratios (S = 0.7, H = 50 m). Fig. 10. Meshes of dams with different L–H ratios (S = 0.9, H = 50 m).

all the results of the increment ratios of safety factor k

4. Results and discussion versus g are shown in the following figures. As a basis
of the investigation, the calculated F s2 of each case is
4.1. Results also listed in Table 7. Each case in the same subgroup
shares an identical F s2 because they own the same typi-
Some typical results of the calculated safety factors cal cross-section and geotechnical parameters of
versus length–height ratio g are listed in Table 7 and materials.

Fig. 11. Material zoning over the typical cross-section (H = 50 m, r = 2.0).

332 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339

Table 7 ward direction. At the same time, the slip surface occurs
Results of the safety factors of the typical cases in the upper part of the downstream slope. Hence, in
Groups Subgroups F s3 F s2 such a condition, the seepage has little influence on the
Length–height ratio g stability of the earth–rockfill dam. Therefore, the 3D ef-
1.2 2.0 3.0 4.0 5.0 6.0 fect will be investigated mainly under the condition of
Group 1 Subgroup 1 1.90 1.80 1.77 1.72 1.71 1.70 1.62
the final construction stage without water storage.
Subgroup 2 1.61 1.53 1.48 1.47 1.44 1.42 1.36 Fig. 14 shows the contours of the displacement incre-
Subgroup 3 1.45 1.38 1.33 1.32 1.30 1.29 1.23 ment just before failure on the 3D condition and that of
Group 2 Subgroup 4 1.91 1.81 1.75 1.72 1.71 1.68 1.61 the maximum section of subgroup 1 at the final con-
Subgroup 5 1.61 1.53 1.48 1.45 1.43 1.42 1.36 struction stage without water storage. By contrasting
Subgroup 6 1.46 1.39 1.34 1.32 1.30 1.29 1.23 Fig. 14 with Fig. 12, it can be seen that the depth of
Group 3 Subgroup 7 1.80 1.73 1.69 1.67 1.66 1.66 1.62 the slip surface is shallower than that of the 2D surface
Subgroup 8 1.52 1.45 1.42 1.40 1.39 1.39 1.36 owing to the 3D effect.
Subgroup 9 1.37 1.32 1.29 1.27 1.27 1.25 1.23 Results in Table 7 show that, in each subgroup, the
Group 4 Subgroup 10 1.75 1.70 1.67 1.65 1.65 1.65 1.62 safety factor clearly increases with decreasing the
Subgroup 11 1.47 1.42 1.40 1.39 1.38 1.38 1.36 length–height ratio, which means the 3D effect is signifi-
Subgroup 12 1.33 1.29 1.27 1.26 1.26 1.25 1.23 cant, although varying from case to case. If the 3D effect
Group 5 Subgroup 13 1.69 1.63 1.61 1.59 1.58 1.57 1.52 is considered in the design, the cost will be greatly reduced
Subgroup 14 1.41 1.37 1.34 1.33 1.32 1.31 1.26 in some cases. An earth–rockfill dam 100 m in length and
Subgroup 15 1.29 1.24 1.22 1.21 1.20 1.20 1.15 50 m in maximum height, located in a ÔV-shapedÕ canyon
Group 6 Subgroup 16 1.61 1.58 1.55 1.54 1.53 1.53 1.48 (i.e., S = 0.5), in subgroup 1, is taken as an example. The
Subgroup 17 1.35 1.31 1.29 1.28 1.28 1.27 1.23 ratios of slope are 1:2.2 and 1:2.0 for upstream and down-
Subgroup 18 1.23 1.20 1.18 1.17 1.17 1.16 1.12 stream slopes, respectively, so the volume of earth–rock-
Group 7 Subgroup 19 1.57 1.49 1.45 1.42 1.41 1.40 1.28 fill is 213,106 m3. The equivalent 2D safety factor F s2 ,
Subgroup 20 1.65 1.57 1.52 1.50 1.49 1.48 1.36 which is usually used in the design, is 1.62. Taking the
Subgroup 21 1.69 1.61 1.56 1.54 1.53 1.51 1.41
same value of safety factor 1.62 as a criterion, considering
Subgroup 22 1.73 1.65 1.60 1.57 1.55 1.54 1.46
Subgroup 23 1.81 1.73 1.67 1.65 1.63 1.62 1.54 3D effect of the dam, i.e., by setting F s3 ¼ 1.62, the ratios
of slope for upstream and downstream slopes can be in-
Fig. 12 shows the contours of the displacement incre- creased to 1:1.93 and 1:1.75, respectively. As a result,
ment just before failure of the 2D section of subgroup 1 the volume of earth–rockfill will decrease to 186,468 m3,
at the final construction stage without water storage. i.e., a 12.5% reduction.
The area with densest isolines means the potential slip
surface. Fig. 13 shows the contours of the displacement
increment just before failure of the 2D section of sub-
group 4 at the stable seepage stage with full storage. Be-
cause the coefficient of permeability of the dam core is
much lower than in materials in other areas, almost all
the seepage force acts inside the core in the down–for-

Fig. 12. Contours of displacement increment just before failure for 2D

condition at the final construction stage without water storage
(H = 50 m, r = 2.0, subgroup 1).

Fig. 13. Contours of displacement increment just before failure for 2D Fig. 14. Contours of displacement increment just before failure on 3D
condition at the stable seepage stage with full storage (H = 50 m, condition at the final construction stage without water storage
r = 2.0, subgroup 4). (subgroup 1, L = 100 m).
 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339 333

4.2. Discussion factor F s3 along with decreasing length–height ratio is

higher for a greater F s2 . To eliminate the influence of
The influence of some factors on the 3D effect on the F s2 on F s3 , the safety factor increment F s3 can be nor-
safety of the earth–rockfill dam is discussed as follows. malized by F s2 to increment ratio k, as defined in Eq. (5).
The relationships between k and g, for each sub-
4.2.1. General tendency and influence of F s2 group, are plotted in Figs. 15–24. As seen from these fig-
As seen from Table 7, in groups with only F s2 or the ures, under any certain condition, k clearly decreases
geotechnical properties of material are different between with g. This confirms that the length–height ratio is cer-
subgroups, i.e., in groups 1–6, the increment of safety tainly an important factor in the 3D effect.

0.20 0.20

Increment Ratio of Fs ( =y)

Increment R atio of Fs ( =y)

H =50m, S =0.5, r =2.0 H =50m, S =0.5, r =2.0

0.15 0.15 M2,FCS
M1,FCS

0.10 0.10
-0.8022
y = 0.2013x y = 0.2057x-0.7821
0.05 2
0.05
R = 0.9812 R2 = 0.9635
0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(a) Length-Height Ratio ( =x) (b) Length-Height ratio ( =x)

0.20 0.20
Increment R atio of Fs ( =y)

Increment Ratio of Fs ( =y)

H =50m, S =0.5, r =2.0 H =50m, S =0.5, r=2.0

0.15 M3,FCS 0.15
FCS
0.10 0.10
-0.8028
0.05
y = 0.2118x y = 0.2062x-0.7957
2 0.05
R = 0.9902 R2 = 0.9746
0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(c) Length-Height Ratio ( =x) (d) Length-Height Ratio ( =x)

Fig. 15. Relationship between increment ratio of Fs and L–H ratio of dam for group 1.

0.20 0.20
Increment Ratio of Fs ( =y)
Increment R atio of Fs ( =y)

H =50m, S =0.5, r =2.0 H =50m, S =0.5, r =2.0

0.15 M1,SSS 0.15 M2,SSS

0.10 0.10
-0.8919
y = 0.2275x-0.8497 y = 0.2266x
0.05 2 0.05 2
R = 0.9884 R = 0.99

0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(a) (b) Length-Height Ratio ( =x)
Length-Height ratio ( =x)

0.20 0.20
Increment Ratio of Fs ( =y)

Increment Ratio of Fs ( =y)

H =50m, S =0.5, r =2.0 H =50m, S =0.5, r =2.0

0.15 M3,SSS 0.15
SSS

0.10 0.10
y = 0.2187x-0.8319 y = 0.2242x
-0.8578
0.05 2 0.05
R = 0.9873 2
R = 0.9848
0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(c) Length-Height ratio ( =x) (d) Length-Height Ratio ( =x)

Fig. 16. Relationship between increment ratio of Fs and L–H ratio of dam for group 2.
334 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339

0.20

Group 1: FCS Group 2: SSS

Increment Ratio o f Fs ( =y ).
0.15
S=0.5, H =50m, r =2.0

0.10

0.05

0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Length-Height Ratio ( =x)

Fig. 17. Comparison of relationship between increment ratio of Fs and L–H ratio of dam at different stages (groups 1 and 2).

0.12 0.12
Increment Ratio of Fs ( =y)

Increment R atio of Fs ( =y)

H =50m, S =0.7, r =2.0 H =50m, S =0.7, r =2.0
0.10 0.10
M1,FCS M2,FCS
0.08 0.08
0.06 0.06
0.04 -1.0356
0.04 -1.0638
y = 0.1335x y = 0.1357x
0.02 2 0.02 2
R = 0.9715 R = 0.9889
0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(a) Length-Height Ratio ( =x) (b) Length-Height Ratio ( =x)
0.12 0.12
Increment Ratio of Fs ( =y)

Increment Ratio of Fs ( =y)

H =50m, S =0.7, r =2.0 H =50m, S =0.7, r =2.0

0.10 0.10
M3,FCS FCS
0.08 0.08
0.06 0.06
0.04 0.04
y = 0.1478x-1.0603 y = 0.1388x
-1.0526
0.02 2 0.02
R = 0.9631 2
R = 0.9687
0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(c) Length-Height Ratio ( =x) (d) Length-Height Ratio ( =x)

Fig. 18. Relationship between increment ratio of Fs and L–H ratio of dam for group 3.

According to the calculated results, the relationship ted in Fig. 15(d). It can be seen that the data distribute in a
between k and g can be fitted by the following power very narrow band in Fig. 15(d), which can be expressed in
function with a high correlation coefficient: a single equation. This means F s2 has little influence on the
k–g relation, although it clearly affects the F s3 –g relation.
k ¼ Agn ; ð6Þ
The same phenomena exist in groups 2–6.
where A and n are the coefficient and index of the func-
tion, respectively, while n is always a minus value. When 4.2.2. Influence of working conditions
g tends to 0, the value of k tends to infinity, which means As discussed previously, under the conditions de-
the dam is fixed and is definitely stable; when g tends to scribed in this paper, seepage has little influence on
infinite, k tends to 0, which implies F s3 ¼ F s2 and the 3D dam stability. Fig. 16 shows the results of k–g relation
effect vanishes. This tendency agrees perfectly with the of group 2, including subgroups 4–6, at the stable seep-
stability mechanism. age stage with full storage. The data-fitting curves of re-
In Fig. 15 for group 1, (a), (b) and (c) represent the rela- sults for groups 1 and 2 are plotted together in Fig. 17. It
tionship between and g of each subgroup with different F s2 can be seen that these two curves are almost identical,
separately, and data from all the three subgroups are plot- which further confirms that the seepage has little
 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339 335

0.10 0.10
Increment Ratio of Fs ( =y)

Increment R atio of Fs ( =y)

H =50m, S =0.9, r =2.0 H =50m, S =0.9, r =2.0
0.08 0.08
M1, FCS M2, FCS
0.06 0.06

0.04 0.04
y = 0.0913x-1.0484 y = 0.0863x-0.9808
0.02 2 0.02
R = 0.95 R2 = 0.9407
0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(a) Length-Height Ratio ( =x) (b) Length-Height Ratio ( =x)

0.10 0.10
Increment R atio of Fs ( =y)

Increment R atio of Fs ( =y)

H =50m, S =0.9, r =2.0 H =50m, S =0.9, r =2.0
0.08 0.08 FCS
M3, FCS
0.06 0.06
0.04 0.04
y = 0.093x-0.8958 y = 0.0901x-0.9751
0.02 0.02
R2 = 0.9551 2
R = 0.9194
0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(c) Length-Height Ratio ( =x) (d) Length-Height Ratio ( =x)

Fig. 19. Relationship between increment ratio of Fs and L–H ratio of dam for group 4.

0.20
Goup 1: S=0.5
0.18
Goup 3: S=0.7
Increment Ratio of Fs ( =y ) .

0.16
Goup 4: S=0.9
0.14
H =50m, r =2.0
0.12
FCS
0.10
0.08
0.06
0.04
0.02
0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Length-Height Ratio ( =x)

Fig. 20. Comparison of relationship between increment ratio of Fs and L–H ratio of dam with different canyon hypsography (groups 1, 3 and 4).

influence on the stability analysis under conditions as- however, the value of k for S = 0.5 still reserves a va-
sumed in this paper. lue of 0.05.

4.2.3. Influence of shape index of canyon 4.2.4. Influence of height of dam

The results of groups 1, 3 and 4 with the shape in- Figs. 22 and 23 show the results of k
g relation of
dex 0.5, 0.7 and 0.9, are plotted in Figs. 15, 18 and 19, groups 5 and 6, with heights of 100 and 150 m, respec-
respectively. In order to investigate the influence of the tively. The data-fitting curves of the subgroups in groups
shape index of canyon S on the 3D effect, the data- 1, 6 and 7 are plotted together in Fig. 23. It can be seen
fitting curves for results of groups 1, 3 and 4 are plot- that the height of the dam is also an important influenc-
ted together in Fig. 20. It can be seen that the canyon ing factor on the 3D effect. The lower a dam is, the
shape index significantly influences the 3D effect, i.e., stronger the 3D effect becomes.
the lower the shape index of the canyon is, the greater
the 3D effect becomes. When the value of g is low, say 4.2.5. Influence of slope gradient
1.2, the value of k for S = 0.5 is about 2.3 times that The results of k–g relation of group 7, including sub-
for S = 0.9. When the value of g is high, say 6.0, the groups 19–23, with the value of the inverse slope gradi-
value of k for S = 0.9 decrease to as low as 0.016; ent r ranging from 1.6 to 1.9, and that of subgroup 1 in
336 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339

0.14 0.14
Increment Ratio of Fs ( =y)

Increment R atio of Fs ( =y)

0.12 H =100m, S =0.5, r =2.0 0.12 H =100m, S =0.5, r =2.0
M1,FCS M2,FCS
0.10 0.10
0.08 0.08
0.06 0.06
y = 0.1304x-0.7334 y = 0.1439x
-0.7194
0.04 0.04
R2 = 0.9858 2
R = 0.9875
0.02 0.02
0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(a) (b) Length-Height Ratio ( =x)
Length-Height Ratio ( =x)

0.14 0.14
Increment R atio of Fs ( =y)

Increment R atio of Fs ( =y)

H =100m, S =0.5, r =2.0 H =100m, S =0.5, r =2.0
0.12 0.12
M3,FCS FCS
0.10 0.10
0.08 0.08
0.06 0.06
y = 0.1308x-0.6553 y = 0.1348x-0.7023
0.04 2
0.04 2
R = 0.9768 R = 0.9627
0.02 0.02
0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(c) Length-Height Ratio ( =x) (d) Length-Height Ratio ( =x)

Fig. 21. Relationship between increment ratio of Fs and L–H ratio of dam for group 5.

0.10 0.10
Increment Ratio of Fs ( =y)

Increm ent Ratio o f Fs ( =y )

H =150m, S =0.5, r =2.0 H =150m, S =0.5, r =2.0

0.08 0.08
M1,FCS M2,FCS
0.06 0.06
y = 0.1008x-0.6475 y = 0.1083x-0.6616
0.04 2
0.04
R = 0.9803 R2 = 0.9634
0.02 0.02
0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(a) Length-Height Ratio ( =x) (b) Length-Height Ratio ( =x)

0.10 0.10
Increment Ratio of Fs ( =y)

Increment Ratio of Fs ( =y)

H =150m, S =0.5, r =2.0 H =150m, S =0.5, r =2.0

0.08 0.08
M3,FCS FCS
0.06 0.06
0.04 y = 0.1078x-0.6262 y = 0.1056x
-0.645
0.04
R2 = 0.9689 2
R = 0.9565
0.02 0.02
0.00
0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(c) Length-Height Ratio ( =x) (d) Length-Height Ratio ( =x)

Fig. 22. Relationship between increment ratio of Fs and L–H ratio of dam for group 6.

0.18
Group 1: H=50m
0.16
Increment Ratio o f Fs ( =y ) .

Group 5: H=100m
0.14
Group 6: H=150m
0.12
S =0.5, r =2.0
0.10
FCS
0.08
0.06
0.04
0.02
0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Length-Height Ratio ( =x)

Fig. 23. Comparison of relationship between increment ratio of Fs and L–H ratio of dam with different heights (groups 1, 5 and 6).
 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339 337

0.25 0.25
Increment R atio of Fs ( =y)

Increment R atio of Fs ( =y)

H =50m, S =0.5, r =1.6 H =50m, S =0.5, r =1.7
0.20 M1,FCS 0.20 M1,FCS
0.15 0.15
0.10 0.10
-0.5525
y = 0.2421x y = 0.2297x
-0.5749
0.05 0.05
R2 = 0.9879 2
R = 0.9811
0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(a) Length-Height Ratio ( =x) (b) Length-Height Ratio ( =x)

0.25 0.25
Increment Ratio of Fs ( =y)

Increment R atio of Fs ( =y)

H =50m, S =0.5, r =1.75 H =50m, S =0.5, r =1.8
0.20 0.20
M1,FCS M1,FCS
0.15 0.15
0.10 0.10
-0.6347
y = 0.218x y = 0.2149x
-0.749
0.05 2 0.05
R = 0.9841 R2 = 0.9941
0.00 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
(c) Length-Height Ratio ( =x) (d) Length-Height Ratio ( =x)
0.25 0.20
Increment R atio of Fs ( =y)

Increment R atio of Fs ( =y)

H =50m, S =0.5, r =1.9

0.20 H =50m, S =0.5, r =2.0
M1,FCS 0.15
M1,FCS
0.15
0.10
0.10
-0.7731 -0.8022
y = 0.2041x y = 0.2013x
0.05 0.05
R2 = 0.9948 R2 = 0.9812
0.00
0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
(e) Length-Height Ratio ( =x) (f) Length-Height Ratio ( =x)

0.25
slope r: 1.6 slope r: 1.7 slope r: 1.75
slope r: 1.8 slope r: 1.9 slope r: 2.0
Increment Ratio o f Fs ( =y )x

0.20

S=0.5, H =50m
0.15 FCS

0.10

0.05

0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0
(g) Length-Height Ratio ( =x)

Fig. 24. Comparison of relationship between increment ratio of Fs and L–H ratio of dam with different gradients (subgroup 1 and group 7).

group 1, with an r value of 2.0, are plotted in Fig. 24(a)– According to the results available, it is easy to con-
(f), respectively. The data-fitting curves for these clude the relative contribution to the 3D effect from
subgroups are plotted together in Fig. 24(g). It can be the factors considered above, as shown in Table 8. Com-
seen that the gradient of dam slope influences the 3D ef- bining the influence of g, r and H, it may be concluded
fect significantly. The steeper the slope is, the stronger that the larger the scale of the earth–rockfill dam is, the
the 3D effect is. weaker the 3D effect becomes.
338 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339

Table 8
Comparison of the contribution of each factor to the 3D effect
No. Factors Contribution to the 3D effect
1 Length–height ratio, g Under all cases this ratio causes the largest 3D effect
2 Shape index of canyon, S (0.5–0.9) 0.178
0.075 = 0.103 when g = 1.2
0.050
0.016 = 0.034 when g = 6.0
3 Height of dam, H (50–150 m) 0.178
0.094 = 0.084 when g = 1.2
0.050
0.033 = 0.017 when g = 6.0
4 Gradient of dam slope, r (1.6–2.0) 0.219
0.174 = 0.045 when g = 1.2
0.090
0.048 = 0.042 when g = 6.0
5 Pore pressure 0.192
0.178 = 0.014 when g = 1.2
0 when g = 3.0

5. Conclusions [2] Morgenstern NR. The evaluation of slope stability: a 25-year

perspective. In: Proceedings of ASCE specialty conference on
stability and performance of slopes and embankments II. Berke-
A series of parametric studies were performed to ley: University of California; 1992. p. 1–26.
investigate the 3D effect on stability of earth–rockfill [3] Stark TD, Eid HT. Performance of three-dimensional slope
dam, with respect to some major factors, such as the stability analysis method in practice. J. Geotech. Eng. ASCE
geometrical characteristics of the dam and the topogra- 1998;124:1049–60.
phy of the canyon site. The following conclusions can be [4] Cai F, Ugai K, Hagiwara T. Base stability of circular excavation
in soft clay. J. Geotech. Geoenviron. Eng. 2002;128(8):702–6.
drawn from the calculated results and discussion: [5] Faheem H, Cai F, Ugai K. Three-dimensional base stability of
rectangular excavations in soft soils using FEM. Comput.
(1) The length–height ratio of the earth–rockfill dam Geotech. 2004;31:67–74.
clearly influences the 3D safety factor, and a power [6] Lefebvre G, Duncan JM, Wilson EL. Three-Dimensional finite
function expresses the relation between the incre- element analysis of dams. J. Soil Mech. Found. Div. ASCE
1973;99(SM7):495–507.
ment ratio of safety factor and the length–height [7] Hovland HJ. Thee-dimensional slope stability analysis method. J.
ratio of the earth–rockfill dam. The 2D safety fac- Soil Mech. Found. Div. ASCE 1977;103(GT9):971–86.
tor F s2 has little influence on k–g relation, although [8] Hungr O. An extension of BishopÕs simplified method of slope
it evidently affects the F s3 –g relation. stability analysis to three dimensions. Geotechnique
(2) The shape index of canyon affects the 3D effect sig- 1987;37(1):113–7.
[9] Chen R, Chameau JL. Three-dimensional limit equilibrium
nificantly. The lower the shape index, the greater analysis of slopes. Geotechnique 1983;33(1):31–40.
the 3D effect becomes. [10] Lam L, Fredlund DG. A general limit-equilibrium model for
(3) The gradient of the dam slope and the height of three-dimensional slope stability analysis. Can. Geotech. J.
the dam greatly influence the 3D effect. The steeper 1993;30:905–19.
and lower the slope is, the stronger the 3D effect is. [11] Chen ZY, Wang XG, Haberfield C. A three-dimensional slope
stability analysis method using the upper bound theorem: Part I:
(4) The results of this paper may be used as a blue- theory and methods. Int. J. Rock Mech. Mining Sci.
print in dam design by combining the F s2 , which 2001;38:369–78.
is in common use, and the relationship between [12] Ugai K. A method of calculation of total factor of safety of slopes
F s2 and F s3 , established in this study. by elasto-plastic FEM. Soils Found. 1989;29(2):190–5.
[13] Yu Y, Ugai K, Ida H. The effect of retaining piles evaluated
by 3D FEM. In: Proceedings of the 31st National Conference
on Geotechnical Engineering, JGS; 1996. p. 396–7 [in
Japanese].
Acknowledgements [14] Dawson E, Motamed F, Nesarajah S, Roth W. Geotechnical
stability analysis by strength reduction. Geotechnical Special
This research is supported by Ô985Õ project of Tsing- Publication, GeoDenver Conference ÕSlope Stability 2000Õ, Den-
ver, CO, USA; 2000. p. 99–113.
hua University and the Scientific Research Foundation [15] Rasskazov LN, Belyakov AA. Stability of a pebble dam with an
for Returned Overseas Chinese Scholars of the State injected core. Hydrotech. Constr. 1984;18(5):226–33.
Education Ministry. [16] Belyakov AA. Three-dimensional behavior of an earth dam at a
wide site. Hydrotech. Constr. 1989;22(12):718–25.
[17] Leshchinsky D, Baker R. Three-dimensional slope stability: end
References effects. Soils Found. 1986;26(4):98–110.
[18] Mejia LH. Three-dimensional stability of an old dam. In:
[1] Seed RB, Mitchell JK, Seed HB. Kettlemen Hills waste landfill Proceedings of the 12th international conference on soil mechan-
slope failure. II. Stability analysis. J. Geotech. Eng. ASCE ics and foundation engineering, Rio de Janeiro, Brazil; 1989. p.
1990;116(4):669–90. 1579–82.
 Y. Yu et al. / Computers and Geotechnics 32 (2005) 326–339 339

[19] Yu Y, Zhang B, Hou W, Zhang J-M. FEM-based stability [20] Donald B, Giam PSK. The ACADS slope stability
analysis of Nuozhadu high earth–rockfill dam. In: Proceedings of programs review. In: Sixth international symposium on
the ninth chinese national conference of soil mechanics and landslides, Christchurch, New Zealand; 1992. p. 1665–
geotechnical engineering, Beijing; 2003. p. 1015–18 [in Chinese]. 70.