Reliability Based Design for Slope Stabilization Using Drilled Shafts

Lin Li
1
and Robert Y. Liang
2



1
Research Assistant, Department of Civil Engineering, University of Akron, Akron, OH
44325-3905; Email: li10@zips.uakron.edu, Tel: 234-738-6084.
2
Distinguished Professor (Corresponding Author), Department of Civil Engineering, University
of Akron, Akron, OH 44325-3905; Email: rliang@uakron.edu; Tel: 330-972-7190; Fax:
330-972-6020.




Re-Submission date: 11/08/2012
Word count: 4750
Figure count: 10
Table count: 1


TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 1


ABSTRACT 1
In this paper, a reliability-based computational algorithm was developed and coded into a 2
computer program, P - UASLOPE, for design of a row of equally spaced drilled shafts to achieve 3
target reliability index for the drilled shaft-reinforced slope system. The Monte Carlo simulation 4
technique was used in the previously developed deterministic computational program, in which 5
the limiting equilibrium method of slices was modified to incorporate arching effects of the 6
drilled shafts in a slope. Uncertainties of soil parameters for each soil layer in the slope were 7
considered by statistical descriptors, including mean, coefficient of variance (c.o.v.), and 8
distribution function. Model errors involving the semi-empirical predictive equation for the load 9
transfer factor for characterizing the soil arching effects were considered by statistics of bias. A 10
total of 41 cases of 3-D finite element simulation results were used to determine the statistics of 11
bias. A design example was given to demonstrate the use of P-UASLOPE program for optimized 12
design of a drilled shaft-reinforced slope system for achieving the most economic combination of 13
design variables (i.e., location, spacing, diameter, and length of drilled shafts) while satisfying 14
the design requirements in terms of target reliability index of the drilled shafts/slope system and 15
the structural performance of the drilled shafts. Sensitivity analysis of the influence of bias of 16
model errors on the computed probability of failure for the design example indicates the need for 17
more cases of 3-D finite element simulation results for obtaining a more accurate semi-empirical 18
predictive equation for the load transfer factor. 19
20
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 2


INTRODUCTION 21
Slope stabilization of un-stable slopes or man-made embankments of highways has been an 22
important geotechnical issue that needs to be addressed to ensure operational safety of roadways. 23
A vast variety of slope stabilization methods have been presented by numerous researchers in the 24
past. Among these methods, a concept of using a row of drilled shafts to stabilize unstable slopes 25
has been used successfully by the Ohio DOT (Liang, 2010). Analysis and design of the drilled 26
shafts to stabilize an unstable slope has been a research topic since 1980s. In general, the analysis 27
methodologies for treating the effects of drilled shafts on slope could be categorized into two 28
approaches: a) an increase in the resistance due to the added shear strength of the reinforced 29
concrete drilled shafts (e.g., Ito, et al. 1981; Hassiotis, et al. 1997; Reese, et al. 1992; Poulos, 30
1995), and b) a decrease in the driving force due to soil arching from the inclusions of rigid 31
structural elements. The later approach was pioneered by Liang and his associates (Liang and 32
Zeng 2002; Yamin 2007; Al Bouder 2010; Liang, 2010 and Joorabchi 2011). 33
Soil arching phenomena in a drilled shaft-reinforced slope system has been studied in 34
Liang and Zeng (2002) by means of 2-D finite element simulations and in Yamin (2007) and Al 35
Bouder (2010) by detailed and comprehensive 3-D finite element simulation results. Liang and 36
Zeng (2002) and Liang (2010) have shown the procedure of incorporating the soil arching effects 37
in the drilled shaft-reinforced slope system in a framework of a limiting equilibrium method of 38
slices methodologies for determining the global factor of safety of the drilled shafts/slope system. 39
The resulting computer program, UA SLOPE was available from Ohio Department of 40
Transportation web site for open access. However, the UA SLOPE program is a deterministic 41
program which cannot systematically take into account of the uncertainties of soil parameters 42
and the semi-empirical equation for quantifying the soil arching effects. There is a need for a 43
probability based version of UA SLOPE program for reliability based analysis and design of a 44
row of drilled shafts to stabilize an unstable slope. Presented in this paper is the theoretical 45
basis of the probabilistic computational algorithms for determining the probability of failure (or 46
reliability) of a slope reinforced with a row of equally spaced drilled shafts, considering all 47
important sources of uncertainties of the soil parameters input and any bias introduced by the 48
semi-empirical equations for the soil arching effects. The details of finite element simulation 49
techniques for deriving the semi-empirical equations for soil arching are presented, together with 50
statistical analysis to derive mean and variance of bias of the semi-empirical equation. Finally, a 51
design example of a two-layer soil slope is presented to illustrate the use of the probabilistic 52
computing algorithm for optimization of the design of a row of drilled shafts on slope to achieve 53
the required target reliability of the global factor of safety as well as economy of construction 54
cost. 55
56
LIMIT EQUILIBRIUM METHOD WITH SOIL ARCHING 57
58
Global Factor of Safety 59
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 3


The method for incorporating the effects of drilled shafts in a drilled shafts/slope system within the 60
framework of limit equilibrium approach can be done in two ways. In the first approach, as given 61
in Equation (1), the resistance to the slope sliding is increased after installing a row of the drilled 62
shaft, thus enhancing the safety factor (Ito, et al., 1981; Reese et al, 1992 and Poulos, 1995). In 63
contrast, as given in Equation (2), the increase of factor of safety is due to soil arching induced 64
reduction on the driving force (Liang and Zeng, 2002; Yamin, 2007; Liang and Al Bouder, 2010). 65
D
shaft R R
F
F F
FS
) (A +
=
(1) 66
arching D D
R
F F
F
FS
) (A
=
(2) 67
where, FS is global factor of safety of a slope/shaft system, F
R
is the resistance Force, 68
(F
R
)
shaft
is additional resistance due to drilled shafts, F
D
is driving force, (F
D
)
arching
is drilled 69
shaft induced arching effect on the driving force. 70
71
3-Dimensional Finite Element Analysis on Arching Effect 72
To interpret the arching effects in a drilled shafts/slope system, a load transfer factor has been 73
introduced as shown in Figure 1, which is defined as the ratio between the horizontal force on the 74
down-slope side of the vertical plane at the interface between the drilled shaft and soil (i.e., 75
P
down-slope
) to the horizontal force on the up-slope side of the vertical plane at the interface 76
between the drilled shaft and soil (i.e., P
up-slope
). Mathematically, the load transfer factor is 77
expressed as: 78
=P
down-slope
/P
up-slope
(3)

79
80
81
82
FIGURE 1 Definition of the load transfer factor. 83
84
The resultant force in the soil upslope and downslope is estimated by integrating the 85
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 4


horizontal soil stresses from the top of the shaft down to the failure surface, as described in 86
equation (4) and (5). 87
/ 2
0 0
f
L
D
up xx
F dsdz
t
o =
} }

(4)

88
/ 2
0 0
'
f
L
D
down xx
F dsdz
t
o =
} }

(5) 89
where, D is the diameter of the drilled shaft; L
f
is the distance from the top of the shaft
90
down to the failure surface;
11 xx
S o =
, which is the horizontal soil stresses on the upslope side of
91
the shaft;
11
' '
xx
S o =
, which is the horizontal soil stresses on the downslope side of the shaft; ds is
92
the integration increment along the periphery of the shaft; dz is the depth of increment.
93
To analyze the slide force acting on the pile, Vessely et al. (2007) has provided a case 94
history to deal with a discrete, deep-seated shear zones. To quantify the soil arching effects and 95
the load transfer factor in a drilled shaft/slope system, Al Bouder (2010) constructed a 3-D finite 96
element model using ABAQUS program (version 6.6) for studying the soil structure interaction 97
behavior of the drilled shafts on a slope under the effect of shear strength reduction. The strength 98
reduction method in finite element simulation was first proposed by Zienkiewicz et al. (1975) and 99
then developed by Qianjun et al. (2009) to study the slope stability problems. The concept of the 100
strength reduction method in the finite element method for determining the FS of a slope is to 101
gradually decrease the soil strength parameters (C and ) until the condition of slope failure (FS=1) 102
is reached. The initial soil strength parameters [C and tan ()] are reduced incrementally by 103
dividing them with a Reduction Factor (RF). Therefore, the reduced cohesion, C
R
and internal 104
friction angle,
R
are given as: 105
R
C C RF = (6) 106
[tan( )] tan( )
R
RF = (7) 107
In Al Bouders (2010) 3-D model, soil is modeled as a linear elastic-perfectly plastic 108
material characterized by the angle of internal friction, cohesion, elastic modulus and Poissons 109
ratio. The mesh of the 3-D model is depicted in Figure 2, which consists of 7, 696 hexahedral 110
elements for the soil body, 23, 600 similar elements for the rock, and 420 similar elements for the 111
drilled shafts. The mesh of the shaft and the adjacent area was finer than that of the other zones 112
because this region was expected to be a high-stress zone. The mesh was refined on the basis of the 113
convergence of the numerically computed FS. At the beginning, a trial mesh was made, and the 114
corresponding FS was found; then the mesh was refined incrementally, and the FS was computed. 115
When the obtained value of the FS becomes stable (almost constant value), then the mesh with the 116
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 5


minimum number of elements corresponding to this safety factor is used. 117
Based on the arching effect (Liang, 2010), the single row of drilled shafts in a slope cause 118
the reduction of the driving stresses in the soil. The effect of the shaft is observed in the changes 119
occurring in the horizontal stresses in the soil mass on the up-slope and down-slope sides of the 120
drilled shafts. The variation of the horizontal soil stresses and the soil arching effects can be seen 121
in Figure 3, which represents a horizontal stress in a horizontal cross section of the drilled 122
shaft/slope system. 123
Joorabchi (2011) proposed a semi-empirical equation, given in Equation (8), to compute 124
the load transfer factor using the regression analysis techniques on a total of 41 cases of 3-D finite 125
element model simulation results conducted by Al Boulder (2010). Considering that sufficient 126
length of drilled shaft is socked into rock layer, the most important influencing factors consist of 127
six parameters: soil cohesion C, friction angle , drilled shafts diameter D, center to center shaft 128
spacing S
0
, shaft location on slope , and slope angle . The stiffness of the shaft will be evaluated 129
in the final design step using LPILE program. 130
)) 0.57( - 0.61 + (-0.252
0.876D) + )(0.065 )(e
D
S
1.114 (-1.17 ) (tan -0.272C =
2
) (-0.578tan 0 0.429 - 0.153
x x
| q

+
(8)
131
where,
x
= (x
i
-x
toe
)/(x
crest
-x
toe
), x
i
= the x-coordinate of the location of drilled shaft, x
toe
= 132
the x-coordinate of the location of the toe of the slope, x
crest
= the x-coordinate of the location of 133
the crest of the slope. 134
135
136
137
FIGURE 2 3-D finite element model developed by Al Bouder (2010). 138
139
To check the validity of the developed semi-empirical formula, Joorabchi (2011) 140
calculated the load transfer factor using Equation (8) for all cases of the FE parametric studies. The 141
comparisons between the results of Equation (8) and the FE results are shown in Figure 4, 142
indicating that there is bias of the semi-empirical equation predictions. 143
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 6


144
145
FIGURE 3 Soil arching as observed from the horizontal soil stresses in the direction of the 146
soil movement (after Al Bouder, 2010). 147
148
149
150
FIGURE 4 Comparison of load transfer factor computed by semi-empirical equation and 151
FEM (after Joorabchi 2011). 152
153
Limiting Equilibrium Method of Slices Incorporating Soil Arching Effects 154
The limiting equilibrium method incorporating the arching effects of the drilled shafts on the slope 155
was developed by Liang and Zeng (2002). Figure 5a shows a typical slice with all force 156
components acting on the slice. The load in the drilled shafts/slope system will be generated by 157
gravity and friction and then be transferred slice by slice through the interslice force P
i
. Without 158
y = 0.9783x
R = 0.7869
0.00
0.20
0.40
0.60
0.80
1.00
0.00 0.20 0.40 0.60 0.80 1.00

C
o
m
p
u
t
e
d

FEM
Series1
Linear (Series1)
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 7


drilled shafts on the slope, the interslice force P
i
can be expressed as an equilibrium equation 159
given in Equation (9). However, with the insertion of drilled shafts on the slope, the interslice 160
force on the down-slope side of the drilled shaft will experience a reduction due to arching, i.e., it 161
will be reduced by a multiplier called the load transfer factor () from the previous interslice 162
force P
i-1
, shown in Equation (10) and (11): 163

+(

)
tan()

] +

1
(9) 164

+(

)
tan()

] +

1
(10) 165

= cos(
1

) sin(
1

(11) 166
The net force applied to the drilled shaft due to the difference in the interslice forces on the 167
up-slope and down-slope sides of the drilled shaft can be calculated as follows: 168

= (1 )
1

0
(12) 169
where, P
i
is the interslice force acting on the down-slope side of slice, W
i
is weight of 170
slice i,
i
is inclination of slice i base, c
i
is soil cohesion at the base of slice i,
i
is soil friction 171
angle at the base of slice i, u
i
is the pore water pressure at slice i, is the load transfer factor, P
i-1
172
is the interslice force acting on the up-slope side of slice (Figure 5b). F
shaft
is the force on the 173
drilled shaft, and S
0
is the center-to-center spacing between two adjacent drilled shafts. Based on 174
Equations (9) to (11), factor of safety for a drilled shafts/slope system can be calculated in an 175
iterative computational algorithm by satisfying boundary load conditions and equilibrium 176
requirements, along with Mohr-Coulomb strength criterion. A PC based computer program, UA 177
SLOPE, has been developed (Liang 2010) based on the above computational algorithm to 178
calculate the factor of safety and the shaft net force for a drilled shaft- reinforced slope system. 179
180

181
182
FIGURE 5a A typical slice showing all FIGURE 5b Slice force change

183
force components. due to arching. 184
185
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 8


RELIABILITY-BASED DESIGN FOR A DRILLED SHAFTS/SLOPE SYSTEM 186
187
Uncertain Parameters in a Drilled Shafts/Slope System 188
The influencing parameters in the drilled shafts/slope system can be divided into two major 189
categories: soil properties (cohesion C, friction angle , and unit weight ) and drilled shaft 190
related parameters (shaft diameter D, clear spacing S between the adjacent drilled shafts, the 191
location of the shaft on the slope
x
). In this paper, the drilled shaft related parameters are treated 192
as certain, while the soil properties are considered as uncertain in the developed reliability based 193
analysis method for a drilled shaft-reinforced slope system. The independent lognormal 194
distributions are assumed for these soil parameters. The mean and c.o.v. of soil parameters can be 195
obtained from in-situ test data. In addition, the Ziggurat algorithm (Marsaglia and Tsang, 2000), 196
an algorithm for pseudo-random number sampling based on the Accept-Reject algorithm, is used 197
in the developed probability based computational program, P-UA SLOPE, as a pseudo-random 198
number generator. 199
200
Bias of Load Transfer Factor 201
The semi-empirical load transfer factor () function given in Equation (8) contains bias as 202
compared to the true value from the results of 41 cases of 3-D finite element simulations. The 203
load transfer factor bias (

) is considered as a random variable with the mean and variance 204

statistically analyzed by comparing the finite element simulation results and the predictions of 205
the semi-empirical equation. The mean and c.o.v. is 1.01 and 0.15, respectively. As indicated in 206
Equation (13), the load transfer factor is randomly generated through the randomly generated 207
bias: 208
=

( , , , ,

,
0
) (13) 209
where, is the randomly generated load transfer factor;

is the randomly generated 210

bias of load transfer factor; and are the randomly generated soil cohesion and friction 211
angle, respectively; , D,
x
and S
0
are considered as deterministic parameters. 212
213
Monte-Carlo Simulation (MCS) 214
The probability of failure for the drilled shafts/slope system is computed by means of Monte 215
Carlo simulation method, as expressed by Equations (14) to (15). A probabilistic version of the 216
UASLOPE computer program, P-UA SLOPE, based on the described algorithm was coded for 217
reliability analysis of a drilled shafts stabilized slope. 218

= ([])
1

[ < 1]

=1
=
1

+(

< 0)

=1
(14) 219
(

) =

+(

< 0)

)
2

=1

(15) 220
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 9


= 1 (

) (16) 221
where, P
f
is the probability of failure for the drilled shafts/slope system, (

) is the 222
coefficient of variance (c.o.v.) of P
f
, I[FS<1] is the indicator function, N is the sample numbers, 223

is the computed probability of failure. is the reliability index. can be negative when the 224

is bigger than 0.5. 225

The computational algorithms of the P-UASLOPE program include the following features 226
and steps. 227
Step 1: Specify the slope and drilled shafts geometry; 228
Step 2: Specify the probability distribution for soil properties, (i.e. cohesion C, friction 229
angle and unit weight ); currently, it is assumed that these variables follow log-normal 230
function with the specified mean and variance values; 231
Step 3: Perform statistical analysis of the bias of the load transfer factor, expressed in the 232
previous section; 233
Step 4: Perform Monte-Carlo simulations using the deterministic computing algorithm 234
for the drilled shafts/slope system ; 235
Step 5: Obtain the reliability index for a drilled shafts/slope system. 236
The flow chart of the P- UASLOPE program is depicted in in Figure 6. 237
238
239
240
FIGURE 6 Flow chart of MCS method using P-UASLOPE. 241
242
Specify the slope, drilled shafts geometry
Specify the probability distribution for soil properties: C,
, Log-Normal distribution
Reliability analysis on Eta function
Reliability index
Calculate the factor of safety using C++ code depending on
limiting equilibrium and arching conception
Monte-Carlo
Simulation
Adding into Indicator Function: I(FS<1)
Yes
No
Iteration
converge
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 10


A STEP BY STEP DESIGN EXAMPLE 243
The slope shown in Figure 7 was considered by Joorabchi (2011) as an example of the 244
deterministic method of computation. It consists of two soil layers with soil properties for each 245
layer summaried in Table 1, in which the range of coefficient of variance of each soil parameter 246
is taken from Phoon and Kulhawy (1999) and the soil parameters for the two soil layers follow 247
independent identical distribution. The critical slip surface for the slope without the drilled shafts 248
was determined by a conventional slope stability analysis program, STABLE, with the computed 249
FS equal to 0.78. The identified critical slip surface is represented by connecting 14 points. The 250
ground water table is assumed at elevation of (-7.93m). It is noted that effective stress approach 251
is used in the analysis. After obtaining the critical slip surface using STABLE program, the 252
reliability index for the drilled shaft-reinforced slope system will be calculated by P UASLOPE 253
program. 254
During Monte Carlo Simulation (MCS), the relationship between probability of failure (P
f
) 255
and the corresponding c.o.v. can be expressed as follows: 256
(

) = (1

)/(

)

1/(

) (17) 257
In the present design example, we generated 100,000 samples for each variable: cohesion 258
C, friction angle and unit weight as well as the bias of load transfer factor. 259
260
261
262
FIGURE 7 Slope geometry for two-layer slope (
x
= 0.42). 263
264
9.15m 9.15m 6.71m 8.54m
7.01m

7.93m

X
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 11


TABLE 1 Soil Properties of the Two-Layer Slope 265
266

Layer No.
Mean
c.o.v.
Layer 1 Layer 2
C (kN/M
2
) 0.958 2.394 0.2
(degree) 11 18 0.1
(kN/M
3
) 21.21 22.78 0.01
267
Step by Step Design Procedure 268
Step 1: Data concerning the geometry of the slope, soil paremeters, ground water table, 269
critical slip surface, etc. are collected and presented in Figure 7 and Table 1. 270
Step 2: Choose a target reliability index. As recommended by Abramson et al. (2002), the 271
target reliability index
Target
can be seleced as 3.0. 272
Step 3: Select different drilled shafts locations. The feasible locations for drilled shafts 273
are between 12.3m and 21.8m (
x
= 0.2~0.8) horizontally from the crest of the slope to the toe of 274
the slope (shown in Figure 7). In the current design, we analyze the location starting from 275
X=12.3m and ending at X=21.8m, with an increment equal to 1.9m. 276
Step 4: Select different pairs of clear spacing S and shaft diameter D combinations within 277
the permissible range. Usually, this may depend on the site situations, and local availability of 278
drilled shaft construciton equipment. In this example, the range for D is selected to be between 279
0.6m to 2.4m (2ft~8ft), and the range of S/D is selected to be between 1.0~3.0. The following 280
combinations for (S, D) were selected: (0.6, 0.6), (1.2, 0.6), (1.8, 0.6), (1.2, 1.2), (2.4, 1.2), (1.8, 281
1.8), (2.4, 2.4). All units in the parenthesis are in meter. 282
Step 5: Calculate reliability index using P - UASLOPE. For each (S, D) combination, plot 283
the relationship between computed reliability index and shaft location, as shown in Figure 8. 284
Step 6: Calculate shaft force using P - UASLOPE. For each (S, D) combination, plot the 285
relationship between the computed net shaft force and shaft location, as shown in Figure 9. 286
Step 7: Optimize the design for achieving the target reliability index, while requiring the 287
least amount of drilled shaft volume for the project. As can be seen in Figure 8, the reliability 288
index tends to increase with shaft location (measured by the distance from the crest of slope) 289
and then decrease after reaching the middle location of the slope. The location of 16m provides 290
the highest reliability index for the given shaft diameter and spacing. However, as shown in 291
Figure 9, the drilled shafts at the location equals to 16m are also subject to the largest net forces, 292
which would have resulted in higher internal moments and shears in the shaft and higher 293
required reinforcement ratio as well. It appears that the force on the shafts would decrease after 294
the location 18m, and the computed reliability index would still satisfy the target reliability 295
index. Furthermore, the required length of drilled shafts could be reduced if the location of the 296
drilled shafts is moved further downslope. With the objective of minimizing the net shaft force 297
while satisfying the target reliability index , three combinations are selected: (a) location 298
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 12


X=21.8m, D=0.6m, S/D=1.0; (b) X=19.9m, D=0.6, S/D=2.0; (c) X=19.9m, D=1.2m, S/D=1.0. 299
300
301
FIGURE 8 Reliability index of the shaft/slope system versus shaft location for different (S, D) 302
combinations. 303
304
305
306
FIGURE 9 Shaft force versus shaft location for different (S, D) combinations. 307
-2
-1
0
1
2
3
4
5
6
12 14 16 18 20 22
D=0.6m, S/D=1 D=0.6m, S/D=2
D=0.6m, S/D=3 D=1.2m, S/D=1
D=1.2m, S/D=2 D=1.8m, S/D=1
D=2.4m, S/D=1
R
e
l
i
a
b
i
l
i
t
y

I
n
d
e
x

(

)

Shaft Location (m)

Target
100
250
400
550
700
850
1000
12 14 16 18 20 22
D=0.6m, S/D=1 D=0.6m, S/D=2
D=0.6m, S/D=3 D=1.2m, S/D=1
D=1.2m, S/D=2 D=1.8m, S/D=1
D=2.4m, S/D=1
S
h
a
f
t

N
e
t

F
o
r
c
e

(
k
N
)

Shafts Location (m)
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 13


Step 8: The software LPILE (Reese and Wang, 1989) was used for the structural analysis 308
of the shaft. Assuming that at least 20% length of drilled shaft is embedded into rock layer below 309
slip surface, the three combinations obtained from step 7 (i.e., (a) D=0.6m, S/D=1.0, F=351KN, 310
X=21.8m; (b) D=0.6m, S/D=2.0, F=423KN, X=19.9m; (c) D=1.2m, S/D=1.0, F=583KN, 311
X=19.9m) with the corresponding net forces can be used as input in LPILE program to calculate 312
the lateral deflection and the internal forces and moments in the drilled shaft. The computer 313
program, LPILE (Reese et al., 2004), is used to analyze the three combinations. In LPILE 314
analysis c- soil is used to model the soil and weak rock-Reese model is used to represent the 315
rock. In LPILE simulation, zero shear and moment at the shaft head are set as the boundary 316
conditions. The net force is distributed as a triangularly distributed shear force acting on the 317
drilled shaft above the slip surface. The total net force is taken from step 7. After calculation with 318
LPILE, the lateral deflection on the top of the shaft is 0.0035m, 0.0052m, 0.0081m respectively 319
for the design combinations of (a), (b) and (c), and the total shaft lengths of the three 320
combinations are 2.5m, 3.5m, 3.5m respectively. All the lateral deflections on the top of the shaft 321
can be considered to be within the allowable deflection (say, half an inch, or equivalently 322
12.7mm). If we perform the quantitative analysis of the required drilled shafts volume, the 323
drilled shaft volume per unit width of a slope for three combinations are 0.588m
3
, 0.550m
3
, 324
1.649m
3
respectively. Thus, the combination (b) (D=0.6m, S/D=2.0, F=423KN, X=19.9m) can 325
be chosen as the most economical design with the least amount of drilled shaft volume required 326
to meet the required target reliability index. The computed reliability index is 3.4 for the 327
combination (b). 328
329
Sensitivity Analysis of Soil Arching Model Errors 330
This section presents a sensitivity analysis on the influence of model errors associated with the 331
empirical equation for the load transfer factor. This is because the statistics of bias of the predictive 332
equation for load transfer factor could be different from current values, once additional finite 333
element simulation results are available, or even high quality field monitoring data becomes 334
available. The combination b (D=0.6m, S/D=2.0, F=423KN, X=19.9m) in last section is chosen as 335
a model in the sensitivity analysis to demonstrate how the model errors will affect the design 336
results. Figure 10 presents the relationship between the computed probability of failure and the 337
mean value of bias of the load transfer factor, while keeping the other values to be the same as in 338
the example run. SD in Figure 10 is standard deviation of bias of the predictive equation for load 339
transfer factor. It can be seen that the computed probability of failure is sensitive to the statistics of 340
the bias of eta. Therefore, it is important to note that more 3-D finite element simulation results 341
would be desirable for improving the developed semi-empirical equation for load transfer factor. 342
343
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 14


344
345
FIGURE 10 Relationship between P
f
and mean of bias of for different standard deviation. 346
347
SUMMARY AND CONCLUSIONS 348
In this paper, a reliability-based computational algorithm for design of drilled shafts to stabilize an 349
unstable slope is presented. The computational algorithm is based on the deterministic approach 350
using a modified limiting equilibrium method of slices incorporating the drilled shafts induced soil 351
arching effects in calculating the global factor of safety of a drilled shaft-reinforced soil slope 352
system. The uncertainties of soil parameters and the semi-empirical equation for predicting the 353
load transfer factor in characterizing the soil arching effects are taken into consideration in the new 354
computer program, P - UASLOPE, by using Monte Carlo simulation techniques. The Ziggurat 355
algorithm is used for pseudo- random generation of random number for the uncertain variables of 356
the system, including soil cohesion, soil friction angle, soil unit weight, and bias of the load 357
transfer factor. The P - UASLOPE program is used in a design example to illustrate the design 358
procedure to achieve the required target reliability index while using the most economical drilled 359
shafts combinations in terms of location of the drilled shafts on the slope, the diameter, the length, 360
and the spacing of the drilled shafts. Finally, a sensitivity analysis of model error on the computed 361
probability of failure of the drilled shaft/slope system is presented to demonstrate the need to 362
continue to improve the predictive model of the semi-empirical equation for the load transfer 363
factor. Specific conclusions based on the design example and sensitivity study can be made as 364
follows. 365
366
1. Design of a single row of equally spaced drilled shafts for stabilizing an unstable slope 367
involves consideration of both geotechnical global factor of safety (or reliability index) as 368
well as structural performance of the drilled shafts in terms of meeting the requirements of 369
limiting the drilled shaft deflection and the necessary reinforcements for sustaining the 370
1.00E-05
1.00E-04
1.00E-03
0.7 0.8 0.9 1.0 1.1 1.2
SD=0.11
SD=0.13
SD=0.15
SD=0.17
P
r
o
b
a
b
i
l
i
t
y

o
f

F
a
i
l
u
r
e

Mean of Load Transfer Factor Bias
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 15


internal forces and moments. This paper has presented methodologies for addressing both 371
design issues. 372
2. The location, spacing, diameter, and length of the drilled shafts can be varied in different 373
combinations while achieving the target reliability index. The final selection of the design 374
combination should be based on economic analysis and consideration of constructability of 375
the drilled shafts. 376
3. The bias of the load transfer prediction equation, or the model error, can affect the 377
computed probability of failure of a drilled shaft-reinforced slope system. There is an 378
incentive for conducting more 3-D finite element simulations covering more simulation 379
conditions so that the bias of the semi-empirical predictive equation can be reduced. 380
381
REFERENCES 382
1. ABAQUS. ABAQUS Standard Users Manual, Version 6.6. Hibbitt, Karlsson, and Sorensen, 383
Inc., Palo Alto, Calif., 2006. 384
2. Abramson, L. W. Slope Stability and Stabilization Methods, 2nd Ed., Wiley, New York, N.Y., 385
2002, pp.427. 386
3. Al-Bodour, W. Development of Design Method for Slope Stabilization Using Drilled Shaft. 387
Ph.D. Dissertation, The University of Akron, 2010. 388
4. Marsaglia, George and Tsang, Wai Wan, The Ziggurat Method for Generating Random 389
Variables. Journal of Statistical Software, 5 (8), 2000, pp.1-7. 390
5. Hassiotis, S., Chameau, J. L. and Gunaratne, M. Design method for stabilization of slopes with 391
piles. J. Geotach. Geoenviron. Eng., 3(4), 1997, 314-323. 392
6. Ito, T., Matsui, T. and Hong, P. W. Design method for stabilizing piles against landslideone 393
row of piles. Soils and Foundations, Vol. 21, No. 1, 1981, pp. 21-37. 394
7. Joorabchi, A. E. Landslide stabilization using drilled shaft in static and dynamic condition. Ph. 395
D. Dissertation. The University of Akron, 2011. 396
8. Liang, R. Y. Field instrumentation, monitoring of drilled shafts for landslide stabilization and 397
development of pertinent design method. 2010 Report for Ohio Department of Transportation. 398
State Job Number: 134238, 2010. 399
9. Liang, R. Y., Al Bodour, W., Yamin, M. and Joorabchi, A. E. Analysis method for drilled 400
shaft-stabilized slopes using arching concept. Transportation Research Record: Journal of the 401
Transportation Research Board, 2010, pp. 38-46. 402
10. Liang, R. Y., Zeng, S. Numerical study of soil arching mechanism in drilled shafts for slope 403
stabilization. Soil and Foundations, Japan. Geotech. Socie., 42 (2), 2002, pp. 83-92. 404
11. Phoon, K. K., and Kulhawy, F. H. Characterization of geotechnical variability. J. Can. 405
Geotech. 36, 1999, pp. 612-624. 406
TRB 2013 Annual Meeting Paper revised from original submittal.
Lin Li and Robert Y. Liang 16


12. Poulos, H. G. Design of reinforcing piles to increase slope stability. Can. Geotech. J., Vol. 32, 407
pp. 808-818, 1995. 408
13. Qianjun, X., Y. Honglei, C. Xiangfeng, and L. Zhongkui. A temperature driven strength 409
reduction method for slope stability analysis. Mechanics Research Communications, Vol. 36, 410
2009, pp. 224-231. 411
14. Reese, L. C., Wang, S. T., Isenhower, W. M., and Arrellaga, J. A. LPILE Plus 5.0 for 412
Windows, Technical and User Manuals, ENSOFT, Inc., Austin, Texas, 2004. 413
15. Reese, L. C., Wang, S. T. and Fouse, J. L. Use of drilled shafts in stabilizing a slope. Proc. 414
Specialty Conference on Stability and Performance of Slopes and Embankment, Berkeley, 415
ASCE 2, 1992, pp. 1318-1332. 416
16. Robert, C. P.; Casella, G., Monte Carlo Statistical Methods (2nd ed.), New York: Springer, 417
ISBN 0-387-21239-6, 2004. 418
17. Vessely, A., Yamasaki, K., and Strom, R. Landslide stabilization using piles. Proc. of 1st N. 419
American Conf. on Landslides, AEG, v.23, 2007. 420
18. Yamin, M. M. Landslide stabilization using a single row of rock-socketed drilled shafts and 421
analysis of laterally loaded shafts using shaft deflection data. PHD Dissertation, Department 422
of Civil Engineering, University of Akron, Ohio, 2007. 423
19. Zienkiewicz, O. C., Humpheson, C., Lewis, R. W. Associated and non-associated 424
visco-plasticity and plasticity in soil mechanics. Geotechnique, 25(4), 1975, pp. 689-691. 425
TRB 2013 Annual Meeting Paper revised from original submittal.