You are on page 1of 7

Technical Note

Probabilistic Assessment of Serviceability Limit State of


Diaphragm Walls for Braced Excavation in Clays
Wengang Zhang 1; Anthony T. C. Goh 2; and Yanmei Zhang 3

Abstract: Braced excavation systems are commonly required to ensure stability in the construction of basements for shopping malls and
Downloaded from ascelibrary.org by San Diego State University on 01/26/16. Copyright ASCE. For personal use only; all rights reserved.

underground transportation facilities. For excavations in deposits of soft clays, stiff retaining wall systems such as diaphragm walls help to
restrain ground movements and wall deflections in order to prevent damage to nearby buildings and utilities. It is quite common for designers
to limit the maximum wall deflection to 0.5% times the maximum excavation depth. However, a review of measured diaphragm wall displace-
ments from various published case histories of successful deep excavations show that wall deflections can be up to more than 2% times the
excavation depth. Since the allowable threshold wall displacement depends on various influencing parameters, wall deflections should not be
limited to an arbitrary value. This paper presents a probabilistic framework combining a simplified wall deflection estimation model with
reliability methods to determine the probability of serviceability limit state failure. Consequently, using this approach, the paper presents a
methodology that allows engineers to determine the required threshold limiting ðδ hm =H e ÞT ratios to meet the different target serviceability
reliability indices. DOI: 10.1061/AJRUA6.0000827. © 2015 American Society of Civil Engineers.
Author keywords: Wall deflection; Braced excavation; Probabilistic assessment; Polynomial regression model; Inverse analysis.

Introduction lation method, and workmanship. Thus, limiting the ratio to a sin-
gle value δ hm =He may not be appropriate.
For excavations in ground that comprises thick soft clays overlying The soil parameters used for analysis by numerical methods
stiff clay, braced walls are usually constructed to minimize ground are not necessarily the “true” values of these parameters, since such
movements. The maximum wall deflection δ hm is an important data obtained are estimated based on limited field tests, laboratory
consideration in the design of the wall and in the assessment of data, or local experience, generally involving considerable uncer-
its impact on the adjacent structures. The response of the structures tainty. In addition, the wall and ground responses in a supported
to the wall displacement depends on many parameters such as the excavation may also be influenced by construction quality/
wall type, the location, the foundation systems, and the existing workmanship and other environmental factors. Thus, deterministic
condition of the structures. It is common to normalize the modeling using numerical methods such as the finite element or
maximum horizontal displacement of the retaining walls, δhm , finite difference method, even with advanced soil constitutive
by dividing it by the excavation depth, He . models, may not be able to take into account all the uncertainties
It is common to use a limiting ratio δ hm =He of 0.5% as a gov- and complexities mentioned above. To this end, the probabilistic
erning criterion for safety of the works and adjacent buildings. framework combining a simplified wall deflection estimation
However, published case histories of successful excavations from model with reliability methods that are able to take into account
various countries indicate that the measured wall deflections (δ hm ) the soil parameter uncertainties is desirable.
are in the range of 0.05% to more than 2% of the excavation depth In this paper, a recently developed polynomial regression (PR)
ðHe Þ (Yoo and Kim 1999, Long 2001, Moormann 2004). Fok et al. model (Zhang et al. 2015) for estimation of the excavation-induced
(2012) reported that in actual practice, the performance of retaining wall deflection is adopted for prediction of δ hm . Based on the PR
walls for deep excavation depends on various factors such as types model, design charts are proposed for preliminary design and
of soil, type of wall, ground water conditions, support systems, checking. This paper demonstrates that with this PR model, the
construction method (top down or bottom up method), wall instal- first-order reliability method (FORM) and Monte Carlo simulation
(MCS) can be applied to determine the probability of serviceability
limit state failure. Consequently, using this approach, the paper
1
Research Fellow, School of Civil and Environmental Engineering, presents a methodology that allows engineers to determine the
Nanyang Technological Univ., Singapore 639798, Singapore (correspond- required threshold limiting ðδ hm =He ÞT ratios to meet the different
ing author). E-mail: zhangwg@ntu.edu.sg target serviceability reliability indices.
2
Associate Professor, School of Civil and Environmental Engineering,
Nanyang Technological Univ., Singapore 639798, Singapore. E-mail:
ctcgoh@ntu.edu.sg Review of Polynomial Regression (PR) Model
3
Ph.D. Student, Nanyang Environmental and Water Research Institute,
Interdisciplinary Graduate School, Nanyang Technological Univ., The polynomial regression (PR) model is a semiempirical model
Singapore 639798, Singapore. E-mail: yzhang051@e.ntu.edu.sg
proposed by Zhang et al. (2015) for predicting the maximum wall
Note. This manuscript was submitted on November 15, 2014; approved
on March 25, 2015; published online on May 26, 2015. Discussion period
deflection induced by braced excavation in clays. It is based on more
open until October 26, 2015; separate discussions must be submitted for than 1,000 FEM simulations using the hardening small strain (HSS)
individual papers. This technical note is part of the ASCE-ASME Journal model which accounts for the increased stiffness of soils at small
of Risk and Uncertainty in Engineering Systems, Part A: Civil Engi- strains. The PR model consists of equations that can be used to
neering, © ASCE, 06015001(7)/$25.00. estimate wall deflection using the following parameters: excavation

© ASCE 06015001-1 ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng.

ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng., 2015, 1(3): 06015001
B/2
Eq. (1) is applicable for the case with the ground water table at
the ground surface, which is the most unfavorable condition. Addi-
strut tional analyses performed to investigate the influence of the ground
2m EA=3.8 × 106 kN
water table indicate that the maximum wall deflections decrease
5m almost linearly with decreasing ground water level and that the
Final h
excavation
8m water table correction factor μw can be approximated as μw ¼
E=2.0 × 107 kN/m2
depth He 11 m 1 − 0.1l, where l is the depth of the ground water table below
14 m T the ground surface.
17 m Soft Clay The developed PR model was validated through a total of 21
20 m well-documented excavation case histories in Zhang et al.
Wall (2015). The plot of predicted wall deflections versus the measured
values indicates that the PR model is able to predict reasonably well
penetration depth 3~5 m the excavation-induced wall deflections.
Downloaded from ascelibrary.org by San Diego State University on 01/26/16. Copyright ASCE. For personal use only; all rights reserved.

d
Stiff Clay c u =500 kPa
Eu =500c u 20 m
γ =20 kN/m3 Design Charts Based on PR Model

Based on Eq. (1) and Table 1, a series of charts relating δ h to


Fig. 1. Cross-sectional soil and wall profile B; T=B, and He =B for general excavation widths with common soil
relative shear strength ratio cu =σv0 and stiffness ratio E50 =cu have
been developed as shown in Figs. 2–7, assuming μw ¼ 1. The pro-
width B; excavation depth H e ; soft clay thickness T; soil unit weight posed charts are potentially useful for preliminary estimation of the
γ; the system stiffness lnðSÞ [S ¼ EI=w h4avg , as defined in Clough maximum wall deflections for given excavation geometries, soil
and O’Rourke (1990), where E is the Young’s modulus of wall parameters, and support conditions.
material, I is the moment of inertia of the wall section, γ w is the unit
weight of water, and havg is the average spacing of the struts]; the
relative soil shear strength ratio cu =σv0 , where cu is the undrained Example
shear strength and σv0 denotes the vertical effective stress; and the This example considers the case of Bugis MRT in Singapore, one of
relative soil stiffness ratio E50 =cu , where E50 is the secant stiffness the 21 case histories presented in Zhang et al. (2015). Based on the
in a standard drained triaxial test. For illustration, Fig. 1 presents the PR model, the probability density of the wall deflection can be
cross-sectional soil and wall profile considered. The maximum wall
deflection estimated by the PR model δ h takes the following form:

δ h ðmmÞ ¼ a0 þ a1 B þ a2 B2 þ a3 T þ a4 T 2 þ a5 He þ a6 H2e
þ a7 ðcu =σv0 Þ þ a8 ðcu =σv0 Þ2 þ a9 ðE50 =cu Þ
þ a10 ðE50 =cu Þ2 þ a11 lnðSÞ þ a12 lnðSÞ2 þ a13 γ
þ a14 γ 2 þ a15 γ lnðSÞ þ a16 ðcu =σv0 Þ lnðSÞ
þ a17 He lnðSÞ þ a18 He T ð1Þ

The values of the coefficients are shown in Table 1.

Table 1. Response Surface Coefficient for δh


Coefficients Values
a0 2,816.7
a1 3.2659
a2 −0.0255
a3 2.7198
a4 −0.1515
a5 11.634
a6 −0.2348
a7 −3,089.0
a8 1,691.0
a9 −0.7102
a10 9.8 × 10−4
a11 −259.91
a12 5.1584
a13 −136.76
a14 2.1079
a15 7.0204
a16 213.27
a17 −2.5553 Fig. 2. Charts relating δh to T=B and He =B for given soil parameters
a18 0.6817 E50 =cu ¼ 200, cu =σv0 ¼ 0.25, and lnðSÞ: (a) B ¼ 30 m; (b) B ¼ 40 m

© ASCE 06015001-2 ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng.

ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng., 2015, 1(3): 06015001
Downloaded from ascelibrary.org by San Diego State University on 01/26/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Charts relating δ h to T=B and H e =B for given soil parameters Fig. 4. Charts relating δh to T=B and He =B for given soil parameters
E50 =cu ¼ 300, cu =σv0 ¼ 0.3, and lnðSÞ: (a) B ¼ 30 m; (b) B ¼ 40 m E50 =cu ¼ 200, cu =σv0 ¼ 0.3 and lnðSÞ: (a) B ¼ 30 m; (b) B ¼ 40 m

determined through MCS. The analysis involved 1,000,000 realiza- strength properties have been summarized by Phoon and Kulhawy
tions assuming that parameters E50 =cu and cu =σv0 are random var- (1999) and Duncan (2000). The statistics of the random variables
iables while the other parameters are deterministic and that both and values used for the deterministic variables are listed in Table 2.
E50 =cu and cu =σv0 follow a lognormal distribution with a coefficient
of variation (COV) of 0.15. The values for deterministic parameters
Serviceability Criterion and Threshold δhm =H e T
are γ ¼ 16.5 kN=m3 ; B ¼ 21 m; H e ¼ 18 m, lnðSÞ ¼ 8.18, and
T ¼ 35 m. The mean values of cu =σv0 and E50 =cu are 0.25 and The ratio δ hm =H e is adopted as the serviceability limit state cri-
150, respectively. Fig. 8 shows the plot of the probability density terion and in assessing the serviceability limit state of diaphragm
of δ h based on the parameter statistics and the PR model. As can be walls for braced excavation, the limit state function based on the PR
seen from Fig. 8, the variation of δ h best follows the normal dis- model can be expressed as
tribution, ranging from 75 mm to 275 mm. The most probable δ h
gðxÞ ¼ R − S ¼ ðδhm =H e ÞT − δ h =He ð2Þ
value is 136 mm, very near the actual measured value of 150 mm.
This section shows that the combined use of the PR model and in which ðδ hm =He ÞT is the limiting threshold value while δ h =He is
MCS can be used to determine the distribution of the wall deflec- derived from the PR model. Serviceability limit state failure is
tions and to perform reliability analysis. deemed to occur if the predicted δ h =H e is greater than the limiting
ðδ hm =He ÞT .

Reliability Analysis on Serviceability Limit State


Reliability Assessment Methods: FORM and MCS

Statistical Information of Input Variables The probability that the serviceability limit state Pf is exceeded can
be determined using Eq. (2). Such calculation of Pf involves the
Among the seven input variables for the PR model, the soil unit determination of the joint probability distribution of the resistance
weight γ, the soft clay thickness T, the excavation width B and R and the loading S and the integration of the Probability Density
depth He , and the system stiffness lnðSÞ are simply treated as Function (PDF) over the failure domain gðxÞ < 0. Considering that
constants, as they have rather limited influence on the results of the PDFs of the random variables are not known in engineering
probabilistic analysis due to their low uncertainty (Hsiao et al. applications and the integration is computationally demanding
2008). The two key soil parameters E50 =cu and cu =σv0 are treated when multiple variables are involved, an approximate method
as random variables and are assumed as lognormal (LN) distribu- known as the first-order reliability method (Hasofer and Lind
tion. The COV values for the two soil parameters are assumed to be 1974), is commonly used to assess Pf . The approach involves
0.15. Useful guidelines on typical COVs of many common soil the transformation of the limit state surface into a space of standard

© ASCE 06015001-3 ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng.

ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng., 2015, 1(3): 06015001
Downloaded from ascelibrary.org by San Diego State University on 01/26/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Charts relating δ h to T=B and H e =B for given soil parameters


E50 =cu ¼ 150, cu =σv0 ¼ 0.25 and lnðSÞ: (a) B ¼ 30 m; (b) B ¼ 40 m
Fig. 7. Charts relating δh to T=B and He =B for given soil parameters
E50 =cu ¼ 250, cu =σv0 ¼ 0.3 and lnðSÞ: (a) B ¼ 30 m; (b) B ¼ 40 m

normal uncorrelated variables, wherein the shortest distance from


the transformed limit state surface to the origin of the reduced var-
iables is the reliability index β (Cornell 1969). For normal distrib-
uted random variables, Pf ≈ 1 − ΦðβÞ, in which Φ = cumulative
probability of a standard normal distribution. Mathematically, Low
and Tang (2004) have shown that β can be computed using
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
   ffi
xi − μi T −1 xi − μi
β ¼ minx∈F ½R ð3Þ
σi σi

in which xi = the set of n random variables; μi = the set of mean


values; σi = the standard deviation; R = the correlation matrix; and

Fig. 6. Charts relating δ h to T=B and H e =B for given soil parameters


E50 =cu ¼ 250, cu =σv0 ¼ 0.3 and lnðSÞ: (a) B ¼ 30 m; (b) B ¼ 40 m Fig. 8. Distribution of the predicted δ h values for Bugis MRT

© ASCE 06015001-4 ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng.

ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng., 2015, 1(3): 06015001
Table 2. Statistics of Random Variables and Values Used for Deterministic Table 3. Hypothetical Cases and Parameters Considered
Variables
Case number Parameters considered
Parameter Values considered
1 B ¼ 30 m; T ¼ 15 m; H e ¼ 18 m; lnðSÞ ¼ 7.30;
Random variables γ ¼ 16.7 kN=m3
cu =σv0 Mean ¼ 200, 300; COV ¼ 0.15 E50 =cu ¼ LNð200,30Þ; cu =σv0 ¼ LNð0.25; 0.04Þ
E50 =cu Mean ¼ 0.25, 0.29; COV ¼ 0.15 2 B ¼ 40 m; T ¼ 30 m; H e ¼ 24 m; lnðSÞ ¼ 8.18;
Deterministic variables γ ¼ 17.5 kN=m3
γðkN=m3 Þ 16.7, 17.5 E50 =cu ¼ LNð300,45Þ; cu =σv0 ¼ LNð0.29; 0.045Þ
B (m) 30, 40 3 B ¼ 30 m; T ¼ 15 m; H e ¼ 12 m; lnðSÞ ¼ 6.10;
T (m) 15 (for B ¼ 30 m), 30 (for B ¼ 40 m) γ ¼ 16.7 kN=m3
H e (m) 18 (for B ¼ 30 m), 24 (for B ¼ 40 m) E50 =cu ¼ LNð200,30Þ; cu =σv0 ¼ LNð0.25; 0.04Þ
lnðSÞ 7.30, 8.18
Downloaded from ascelibrary.org by San Diego State University on 01/26/16. Copyright ASCE. For personal use only; all rights reserved.

F = the failure region. Low (1996) has shown that a Microsoft and the calculation of probability failure Pf for the case
Excel spreadsheet can be used to perform the minimization and with B ¼ 30 m, γ ¼ 16.7 kN=m3 , H e ¼ 18 m, lnðSÞ ¼ 7.30,
determine β. T ¼ 15 m, mean value of E50 =cu ¼ 200, and average value of
As an alternative to the FORM method, MCS is adopted for cu =σv0 ¼ 0.25 are illustrated in Fig. 9(a). It can be observed that
comparison. In this study, the Microsoft Excel add-in program for a limiting threshold ðδ hm =He ÞT ¼ 0.5%, the probability that
@RISK (http://www.palisade.com) was used. this value is exceeded is 4.3%, while for ðδ hm =He ÞT ¼ 0.2%,
the Pf value can be as high as 56%. The typical FORM procedures
and the calculation of Pf are illustrated in Fig. 9(b). Cells D3:G4
The Developed FORM_PR and MCS_PR Frameworks are parameters which are set corresponding to the variable
For each MCS simulation, the number of iterations is 1,000,000 distribution types. For normal distributions, cells D3:D4 corre-
and Latin hypercube sampling is adopted. The MCS procedures spond to the mean values and cells E3:E4 correspond to the

Fig. 9. The developed reliability assessment frameworks: (a) MCS_PR; (b) FORM_PR

© ASCE 06015001-5 ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng.

ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng., 2015, 1(3): 06015001
Downloaded from ascelibrary.org by San Diego State University on 01/26/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. Influence of ðδ hm =H e ÞT on Pf and determination of


ðδhm =H e ÞT according to the target β SLS , Case 2
Fig. 10. Influence of ðδ hm =H e ÞT on Pf and determination of
ðδhm =H e ÞT according to the target β SLS , Case 1

standard deviations. For non-normal distributions, the non-normal


parameters are replaced by an equivalent normal ellipsoid, centered
at the equivalent normal mean. The correlation matrix R cells K3:
L4 are used to define the correlations between E50 =cu and cu =σv0 .
The dimensionless xi0 vector in cells M3:M4 contains equations for
ðxi − ui Þ=i . Cell I3 contains the estimations of the maximum wall
deflection based on the predictive PR model δ h . Cell N3 contains
the serviceability limit state function gðxÞ ¼ ðδ hm =He ÞT − δh =He .
The design point (xi values) was obtained by using the spread-
sheet’s built-in optimization routine SOLVER to minimize the cell,
by changing the xi0 values, under the constraint that gðxi Þ ¼ 0. Prior
to invoking the SOLVER search algorithm, the xi0 values in cells
M3:M4 were set equal to the mean values (0, 0), corresponding
to the mean values (200, 0.25) of the original two random variables.
Iterative numerical derivatives and directional search for design
point xi were automatically carried out in the spreadsheet. It is ob-
vious from Figs. 9(a–b) that both methods give the same Pf value
of 4.3%.

Probabilistic Analyses and Target Reliability Indices


Based on the MCS_PR and FORM_PR frameworks illustrated in
the section above, three hypothetical cases were adopted to show
the influence of the chosen threshold ðδ hm =H e ÞT values on the Pf
results and the determination of ðδhm =H e ÞT according to the target
serviceability reliability index β SLS . The assumed statistics of the
seven input variables for the three cases are listed in Table 3. The
lognormal distribution was assumed for the two random variables
since it avoids allowing negative values in modeling material prop-
Fig. 12. Influence of ðδ hm =H e ÞT on Pf and determination of
erties. Furthermore, the lognormal distribution has the advantage in
ðδhm =H e ÞT according to the target β SLS , Case 3
modeling the right-hand tail of the distribution. The right-hand tail

© ASCE 06015001-6 ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng.

ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng., 2015, 1(3): 06015001
is important because it governs the probability of excessive defor- soil deposits that can be modeled with the set of soil parameters
mations and the reliability of the serviceability limit state (SLS) used in this study. In addition, the error in the maximum wall de-
design. Typical values for reliability index for ultimate limit state flection measurements will also influence the reliability results.
(ULS) β ULS and the corresponding Pf have been proposed by the Furthermore, as the hypothetical cases to build the PR model as-
U.S. Army Corps of Engineers (1997). Wang and Kulhawy (2008) sume a single layer of soft clay for numerical simulations, conse-
stated that although ULS design codes apply reliability principals, quently, the PR model and the proposed framework are not
serviceability limit state designs are still evaluated using conven- applicable to excavations in heterogeneous or layered clay deposits.
tional approaches where the serviceability reliability index β SLS re-
mains unknown. That is, there are no guidelines with regard to the
target β SLS , though Wang and Kulhawy (2008) presented for the References
first time a relationship between β SLS and β ULS and applied it
to augered cast-in-place piles. In this study, β SLS values of 2.0, Clough, G. W., and O’Rourke, T. D. (1990). “Construction induced
2.3, 2.6, and 3.1 are selected as the target reliability indices since movements of in situ walls.” Proc., Design and Performance of Earth
Downloaded from ascelibrary.org by San Diego State University on 01/26/16. Copyright ASCE. For personal use only; all rights reserved.

their corresponding Pf values are 2%, 1%, 0.5%, and 0.1%, respec- Retaining Structures, ASCE, Reston, VA, 439–470.
tively, which are commonly used for serviceability considerations. Cornell, C. A. (1969). “A probability-based structural code.” ACI Struct. J.,
Figs. 10–12(a) show the significant influence of ðδ hm =He ÞT on 66(12), 974–985.
Pf . It is obvious that small limiting ðδhm =H e ÞT will result in higher Duncan, J. M. (2000). “Factor of safety and reliability in geotechnical
engineering.” J. Geotech. Geoenviron. Eng.,10.1061/(ASCE)1090
Pf since low thresholds are easily exceeded. The Pf decreases with
-0241(2000)126:4(307), 307–316.
the increase of the ðδhm =H e ÞT values. Figs. 10–12(b) plot the
Fok, P., Neo, B. H., Veeresh, C., Wen, D., and Goh, K. H. (2012). “Limiting
curves of the threshold ðδ hm =He ÞT values versus the target β SLS values of retaining wall displacements and impact to the adjacent
of 2.0, 2.3, 2.6 and 3.1. It can be observed that ðδ hm =He ÞT increases structures.” IES J. Part A Civ. Struct. Eng., 5(3), 134–139.
with the increase of the target β SLS , indicating that the choice of Hasofer, A. M., and Lind, N. (1974). “An exact & invariant first-order
target β SLS has a significant effect on determination of the threshold reliability format.” J. Eng. Mech., 100(1), 111–121.
ðδ hm =He ÞT values. Hsiao, E. C. L., Schuster, M., Juang, C. H., and Kung, T. C. (2008).
“Reliability analysis of excavation-induced ground settlement for build-
ing serviceability evaluation.” J. Geotech. Geoenviron. Eng., 10.1061/
Conclusions and Discussions (ASCE)1090-0241(2008)134:10(1448), 1448–1458.
Long, M. (2001). “Database for retaining wall and ground movements due
This paper presents a probabilistic framework combining a recently to deep excavations.” J. Geotech. Geoenviron. Eng.,10.1061/(ASCE)
proposed polynomial regression model used for estimating wall 1090-0241(2001)127:3(203), 203–224.
deflections induced by excavation with reliability methods MCS Low, B. K. (1996). “Practical probabilistic approach using spreadsheet.”
and FORM to determine the probability of serviceability limit state Proc., Uncertainty in the Geologic Environment—From Theory to
failure. The main work and findings include the following: Practice, ASCE, Madison, WI, 2, 1284–1302.
1. A series of design charts are proposed, which are potentially Low, B. K., and Tang, W. H. (2004). “Reliability analysis using object-
useful for preliminary estimation of the maximum wall deflec- oriented constrained optimization.” Struct. Saf., 26(1), 69–89.
tions for given excavation geometries, soil parameters, and Moormann, C. (2004). “Analysis of wall and ground movements due to
support conditions. In addition the example case of Bugis deep excavations in soft soil based on a new worldwide database.” Soils
MRT station shows that the combined use of the PR model Found., 44(1), 87–98.
and MCS can be used to determine the distribution of the wall Phoon, K. K., and Kulhawy, F. H. (1999). “Characterization of geotechnical
deflections and to perform reliability analysis. variability.” Can. Geotech. J., 36(4), 612–624.
2. Probabilistic frameworks FORM_PR and MCS_PR are devel- U.S. Army Corps of Engineers. (1997). “Engineering and design: Introduc-
tion to probability and reliability methods for use in geotechnical
oped for reliability assessment of the probability of service-
engineering.” Dept. of the Army, Washington, DC.
ability limit state failure.
Wang, Y., and Kulhawy, F. H. (2008). “Reliability index for serviceability
3. The influence of the threshold ðδhm =H e ÞT values on probabil- limit state of building foundations.” J. Geotech. Geoenviron. Eng.,
ity of serviceability limit state failure is significant. The choice 10.1061/(ASCE)1090-0241(2008)134:11(1587), 1587–1594.
of the threshold ðδ hm =H e ÞT values can be determined through Yoo, C. S., and Kim, Y. J. (1999). “Measured behaviour of in situ walls in
the target serviceability reliability index β SLS . Korea.” Proc., 5th Int. Symp. on Field Measurements in Geomechancis,
It should be noted that the focus of this paper is on the proposed A. A. Balkema, ed., Rotterdam, Netherlands, 211–216.
probabilistic framework combining a simplified empirical estima- Zhang, W. G., Goh, A. T. C., and Xuan, F. (2015). “A simple prediction
tion model with reliability methods to determine the probability of model for wall deflection caused by braced excavation in clays.”
serviceability limit state failure, and thus its application is limited to Comput. Geotech., 63, 67–72.

© ASCE 06015001-7 ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng.

ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng., 2015, 1(3): 06015001

You might also like