Professional Documents
Culture Documents
www.elsevier.com/locate/compgeo
Received 4 January 1999; received in revised form 18 May 1999; accepted 24 May 1999
Abstract
This paper presents one of the geotechnical initiatives in reliability-based code development
that has been sponsored by the Electric Power Research Institute for transmission line struc-
ture foundations. The framework for the development of a practical reliability-based design
approach is illustrated using the design of drilled shafts (bored piles) for uplift under
undrained loading. A target reliability index of 3.2 is selected based on the reliability indices
implied by existing working stress designs. Two simple design formats (load and resistance
factor design and multiple resistance factor design) are rigorously calibrated using the ®rst-
order reliability method to produce designs that achieve a known level of reliability con-
sistently. # 2000 Elsevier Science Ltd. All rights reserved.
1. Introduction
During the past 40 years, signi®cant progress has been made in the ®eld of struc-
tural safety. The primary theme in structural safety is reliability analysis, which can
be de®ned as the consistent evaluation of design risk using probability theory. Any
design methodology that incorporates the principles of reliability analysis, either
explicitly or otherwise, may be classi®ed as reliability-based design (RBD). Much of
the impetus for this innovation arose from the widespread rethinking of the whole
* Corresponding author.
0266-352X/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S0266-352X(99)00037-3
170 K.-K. Phoon et al. / Computers and Geotechnics 26 (2000) 169±185
design process that was brought about by the proliferation of new technology and
new materials accompanying the boom in construction following World War II. In
the ensuing years, several structural RBD codes were developed and implemented in
short succession around the world, for example, in the UK in 1972 [1], in Canada in
1974 [2], in Denmark in 1978 [3], and in the US in 1983 for concrete [4] and in 1986
for steel [5]. More recent developments include the introduction of API RP 2A-
LRFD [6], Eurocode 1 [7], and ISO 2394 [8]. In the ®eld of transmission line design,
the Electric Power Research Institute (EPRI) has sponsored research studies direc-
ted toward the implementation of these new safety concepts for the design of trans-
mission line structures [9]. Parallel research and development eorts in this ®eld also
have been undertaken by the ASCE Task Committee on Structural Loadings [10]
and the IEC Technical Committee 11 [11].
Most of the impetus in the development of this new design methodology (RBD)
arises from the structural engineering community. In recent years, however, there is
a growing awareness and interest among geotechnical engineers in these new safety
concepts and design methodologies [12±14]. It is anticipated that the geotechnical
design community will become more involved in this code evolution process as the
need arises to maintain compatibility between structural and geotechnical design
codes and the trends towards code standardization between dierent countries
become more prevalent [15,16]. To date, only a few major geotechnical initiatives in
reliability-based code development have been undertaken [17, 18], but the verdict on
their usefulness is still out in view of their relative immaturity. This paper presents
one of these major initiatives [18], sponsored by the Electric Power Research Insti-
tute for transmission line structure foundations, with the aim of encouraging more
discussion in this arena. The framework for the development of a practical relia-
bility-based design approach for transmission line structure foundations will be
illustrated using the design of drilled shafts (bored piles) for uplift under undrained
loading.
The presence of uncertainties and their signi®cance in relation to design has long
been appreciated [19]. The engineer recognizes, explicitly or otherwise, that there is
always a chance of not achieving the design objective, which is to ensure that the
system performs satisfactorily within a speci®ed period of time. Traditionally, the
engineer relies primarily on factors of safety to reduce the risk of adverse perfor-
mance (collapse, excessive deformations, etc.) at the design stage. Factors of safety
between 2 and 3 generally are considered to be adequate in foundation design [20].
Important considerations that aect the factor of safety include variations in the
loads and material strengths, inaccuracies in design equations, errors arising from
construction, and the consequences of failure. Traditionally, the engineer does not
actually go through the process of considering each of these factors separately and
K.-K. Phoon et al. / Computers and Geotechnics 26 (2000) 169±185 171
Qn
Fn <
1
FS
clearer formulation of the traditional design approach that would help facilitate the
explicit recognition and treatment of engineering risks. In recent years, the rapid
development of RBD in the structural engineering community has overshadowed
considerably the fundamental role of limit state design. Much attention has been
focused on the consistent evaluation of safety margins using advanced probabilistic
techniques [22]. Although the achievement of consistent safety margins is a highly
desirable goal, it should not be overemphasized to the extent that the importance of
the principles underlying limit state design become diminished. A detailed exposition
of this important issue and its impact on foundation design is given elsewhere [23].
Foundations must be designed to: (a) ensure sucient safety against ultimate
failure, and (b) limit foundation deformations to allowable limits for the super-
structure in question. The ®rst condition relates to the ultimate limit state, while the
second relates to the serviceability limit state. Following the principles of limit state
design, the occurrence of these two limit states must be checked and shown sepa-
rately to be suciently improbable.
At the ultimate limit state, the uplift capacity of drilled shafts under undrained
loading conditions is governed by the following general equation [24]:
in which Qu =uplift capacity, Qsu =side resistance, Qtu =tip resistance, and W=
weight of foundation. The side resistance (Qsu ) can be calculated using the total
stress method as follows:
D
Qsu B su
zdz
3
0
in which ui =initial pore water stress at the foundation tip or base, u=change in
pore water stress caused by undrained loading, and Atip =tip or base area. Note that
no suction force develops unless ÿu exceeds ui . Extensive comparisons with full-
scale load tests have shown that this overall model provides realistic predictions of
the uplift capacity for drilled shafts under undrained loading conditions [25,26]. The
criterion of realism is crucial for reliability-based design. If the model is con-
servative, it is obvious that the probabilities of failure calculated subsequently will
be biased, because those design situations that belong to the safe domain will be
assigned incorrectly to the failure domain, as a result of the built-in conservatism.
Detailed discussions of this predictive model are given elsewhere [24±26].
The ultimate limit state is de®ned as that in which the undrained uplift capacity is
equal to the ultimate applied load. Clearly, the drilled shaft will fail if the undrained
K.-K. Phoon et al. / Computers and Geotechnics 26 (2000) 169±185 173
uplift capacity is less than this applied load. Conversely, the drilled shaft should
perform satisfactorily if the applied load is less than the undrained uplift capacity.
These situations can be described concisely by a single performance function P, as
follows:
P Qu ÿ F 5
in which Qu =uplift capacity determined from Eqs. (2)±(4) and F=applied founda-
tion load. Mathematically, the above situations simply correspond to the three pos-
sible conditions of P 0, P < 0, and P > 0.
Finally, the occurrence of each limit state must be shown to be suciently
improbable. The philosophy of limit state design does not entail a preferred method
of ensuring safety. Since all engineering quantities (e.g. loads, strengths) are inher-
ently uncertain to some extent, a logical approach is to model the above performance
function using probabilistic means. The mathematical formalization of this aspect of
limit state design constitutes the main thrust of RBD. Aside from probabilistic
methods, less formal methods of ensuring safety, such as the partial factors of safety
method [27,28], have also been used within the framework of limit state design.
3. Reliability-based design
The basic objective of RBD is to ensure that the probability of failure of a com-
ponent does not exceed an acceptable threshold level. While the above objective
clearly is satis®ed if the probability of failure of a component lies far below the
threshold, it is equally clear that the design is not economical. Therefore, a realistic
interpretation of the design objective would include the implicit requirement that the
probability of failure does not depart signi®cantly from the threshold. For the
design of drilled shafts under undrained uplift loading, the RBD objective can be
formally stated as follows:
pf Prob
Qu < F4pT
or
pf Prob
P < 04pT
6
Reliability analysis attempts to resolve the problem associated with the traditional
method of ensuring safety by rendering broad, general concepts, such as uncertainties
174 K.-K. Phoon et al. / Computers and Geotechnics 26 (2000) 169±185
and risks, into precise mathematical terms that can be operated upon consistently.
Uncertain engineering quantities (e.g. load and soil strength) are modeled by ran-
dom variables, while design risk is quanti®ed by the probability of failure. Note that
the term ``failure'' used herein is synonymous with ``adverse performance'' and does
not necessarily refer to a collapse event.
From the de®nition of the limit state performance function given in Eq. (5), it can
be seen that the main uncertain parameters are: (a) adhesion factor (), (b) undrained
shear strength (su ), (c) tip suction (Qtu ), and (d) foundation load (F). In this study,
the adhesion factor was determined by the following regression equation [26]:
pa
0:31 0:17 "
7
su
in which pa =atmospheric pressure 100 kN/m2, su =undrained shear strength
determined from consolidated isotropic undrained triaxial compression tests, and
"=uncertainty about the regression equation. It is important to note that the de®-
nition of su in Eqs. (3) and (7) should be consistent, because su is not a fundamental
material property and depends on many factors, such as the mode of testing. A
detailed discussion on the importance of evaluating soil properties speci®cally within
a particular design context has been given elsewhere [29].
in which W=total foundation weight. By substituting Eq. (8) into Eq. (4), the tip or
base suction force can be estimated. It is important to note that the suction stress
acting over the tip area can not exceed one atmosphere, which is about 100 kN/m2
[25]. This physical constraint must be imposed on the above tip suction calculation.
Some measurements of tip suction stress have been performed on ten model drilled
shafts subjected to undrained uplift loading [30]. All ten of the shafts had a diameter
of 89 mm and depth to diameter ratios equal to 6.7. The measured tip suction forces
for these nominally identical drilled shafts were found to vary erratically from 0 to
424 N. For comparison, the maximum tip force based on one atmosphere of suction
stress is 622 N, while the solution obtained from Eqs. (4) and (8) is 53 N.
Tip suction apparently is dicult to predict accurately because it is sensitive to the
drainage conditions in the vicinity of the tip. The erratic tip suction measurements
noted above might be attributed to variable concrete porosity and the presence of
small air voids at the shaft tip and possibly within the soil [30]. It also should be
noted that tip suction could be eectively eliminated if there is a signi®cant thickness
of cohesionless material beneath the foundation tip [25]. In the absence of a rigorous
model that can account for these important construction eects and the lack of sta-
tistical data, it is reasonable to assume that the tip suction stress is uniformly dis-
tributed between zero and one atmosphere. A less diuse probability distribution
can be assumed when additional information is available.
F kV2 9
ÿÿ1 pf 10
In principle, the most economical target probability of failure (pT ) can be deter-
mined by conducting a cost±bene®t analysis, as shown in Fig. 1. By studying the
variation of the initial cost, maintenance costs, and the expected failure costs with
pf , it is possible theoretically to arrive at the most economical target probability of
failure for design [32]. At present, such an approach is not yet practical because of
the diculties in evaluating failure costs (e.g. cost of human lives) and the eect of
component failure on the system. Another approach is to set the value of pT at a
level that is comparable with the failure rates estimated from actual case histories
(Fig. 2). However, comparing the theoretical probability of failure derived from
reliability computations with a value established by actual case histories is not
straightforward. It has been noted that the theoretical probability of failure usually
is signi®cantly smaller than the actual failure rate [33]. This result is not surprising,
because the safety of a design is not aected by uncertainties underlying design
Table 1
Relationship between the reliability index () and the probability of failure (pf )
1.5 0.0668
2.0 0.0228
2.5 0.00621
3.0 0.00135
3.5 0.000233
4.0 0.0000316
a
( . )=Standard normal probability distribution.
K.-K. Phoon et al. / Computers and Geotechnics 26 (2000) 169±185 177
Fig. 2. Empirical rates of failure for civil engineering facilities. Source: Baecher, 1987 [35].
major advantage of keeping the new design methodology compatible with the exist-
ing experience base. This approach is consistent with the evolutionary nature of
codes and standards that require changes to be made cautiously and deliberately.
The general approach for the calibration of pT involves the following steps [18]:
in which Qsun =nominal uplift side resistance and Qtun =nominal uplift tip resis-
tance. The nominal uplift side resistance was de®ned as follows:
Qsun BDn msu
13
in which n =nominal adhesion factor determined from the mean undrained shear
strength (msu ) using Eq. (7). The nominal tip resistance (Qtun ) was evaluated using
Eqs. (4) and (8) as follows:
W
Qtun ÿ ui Atip
14
Atip
The global factor of safety (FS) used in the working stress design approach [Eq.
(11)] generally lies between 2 and 3.
The reliability levels implicit in these designs are highly variable, as illustrated in
Fig. 3. It can be seen that the reliability indices lie in an approximate range of 2.8±
3.6, and the average reliability index is about 3.2. For reliability-based design, a
target reliability index of 3.2 was chosen based on the following important con-
siderations:
Fig. 3. Reliability levels implicit in existing ultimate limit state design of drilled shafts in undrained uplift.
b. It is applicable to the ultimate limit state of all the loading modes considered
throughout this study [18].
c. The target reliability index for foundations should exceed that of the structures
in transmission line structure design. This approach is consistent with the
design philosophy that foundations should be safer than the structures,
because foundation repairs are more dicult and costly [9,11].
d. It is consistent with the empirical rates of failure shown in Fig. 2 after an
appropriate adjustment has been made to account for the one order of mag-
nitude dierence between the actual and theoretical rates of failure.
180 K.-K. Phoon et al. / Computers and Geotechnics 26 (2000) 169±185
The target reliability index of 3.2 has been chosen based on extensive reliability
studies of existing ultimate limit state foundation designs. The reliability index of 3.2
also has been chosen to be consistent with the empirical evidence on actual founda-
tion failure rate and to conform with the widely accepted transmission line design
philosophy of making foundations safer than structures. Therefore, there are strong
theoretical, empirical, and philosophical reasons for using a target reliability index
of 3.2. This target can be considered to be optimal within the scope of this study,
and it is recommended for future transmission line structure foundation design. This
target reliability index of 3.2 is incorporated into all the ultimate limit state relia-
bility-based design equations presented in the EPRI study [18].
Reliability-based design in the form of Eq. (6) involves the repeated use of fairly
complicated reliability assessment routines, such as FORM, to evaluate the prob-
abilities of failure of trial designs until the computed probability of failure is rea-
sonably close to the chosen threshold level. While the approach is rigorous, it may
not be suitable for designs that are carried out on a routine basis. Routine designs
would include conventional types of structures and foundations with no abnormal
risks or unusual or exceptionally dicult ground or loading conditions. In the EPRI
study [18], a simpli®ed RBD approach was developed that involves the use of con-
ventional lumped or multiple-factor formats for checking foundation designs, as
given below for uplift loading of a drilled shaft:
can not freely change the predictive model, the underlying probability distributions
for loads and strengths, and the target probability of failure. However, such a dis-
advantage is more apparent than real. As noted earlier, the ¯exibility associated with
the traditional factor of safety approach produces inconsistent safety margins.
In principle, any format can be used for reliability calibration. The selection of an
appropriate format is unrelated to reliability analysis. However, practical issues such
as simplicity, familiarity, and compatibility with the existing design approach, are
important considerations that will determine if the simpli®ed RBD design approach
can gain ready acceptance among practicing foundation engineers. The formats
shown in Eqs. (15a) and (15b) clearly satisfy these considerations. For example, the
resistance factor u in Eq. (15a) corresponds to the reciprocal of the factor of safety
(FS) in the traditional working stress design approach [compare with Eq. (1)]. Eq.
(15a) also is known as the load and resistance factor design or LRFD format. It
already has been adopted widely in the structural community for reliability-based
design [4,5,34], and it appears in a number of the recently proposed RBD codes for
foundations [14,17]. Eq. (15b) is a broad generalization of Eq. (15a) (called multiple
resistance factor design or MRFD) that involves the application of one resistance
factor to each component of the capacity rather than the overall capacity, and it is
the preferred and more physically meaningful format for foundation engineering.
The results of an extensive reliability calibration study for ultimate limit state
design of drilled shafts under undrained uplift loading are presented in Tables 2 and
3 and are to be used with Eqs. (15a) and (15b), respectively. All other limit states,
foundation types, loading modes, and drainage conditions addressed in the EPRI
study [18] have similar types of results, with simple RBD equations and corre-
sponding tables of resistance factors. Note that the resistance factors for undrained
Table 2
Undrained ultimate uplift resistance factors for drilled shafts designed using F50 u Qun a,e
Table 3
Undrained uplift resistance factors for drilled shafts designed using F50 su Qsu tu Qtun w Wa,e
uplift of drilled shafts depend on the clay consistency and the coecient of variation
(COV) of the undrained shear strength (su ). The clay consistency is classi®ed broadly
as medium, sti, and very sti, with corresponding mean su values of 25±50, 50±100
and 100±200 kN/m2, respectively.
Some of the trends in the resistance factors presented in Tables 2 and 3 can be
explained readily. For example, both u and su decrease with increasing COVsu
because an increase in the uncertainty of the undrained shear strength produces a
less reliable side resistance estimate, which makes the overall capacity estimate less
reliable as well. The eect of msu on u and su is caused by the decrease in the
mean adhesion factor as the mean undrained shear strength increases. This resulted
in an increase in the uncertainty of the adhesion factor relative to its mean value
because the standard deviation of the adhesion factor does not change. The increase
in tu and w as msu and COVsu increase can be explained partially by noting that
the relative contribution of the uncertainty in the side resistance to the overall
uncertainty in the capacity increases as the above undrained shear strength statistics
increase. As a result, it becomes less eective to use tip resistance and weight factors
to control the reliability of the design.
Foundations are designed using these new RBD formats in the same way as in the
traditional approach, with the exception that the rigorously-determined resistance
factors shown in Tables 2 and 3 are used in place of an empirically-determined fac-
tor of safety. The performance of the two simpli®ed RBD formats is examined by re-
designing the examples shown in Fig. 3 using the resistance factors given in Tables 2
and 3. The results of this performance study are shown in Fig. 4. A comparison
between Figs. 3 and 4 illustrates clearly the improvement in the uniformity of the
reliability levels resulting from use of the new RBD formats.
K.-K. Phoon et al. / Computers and Geotechnics 26 (2000) 169±185 183
Fig. 4. Performance of ultimate limit state RBD formats for drilled shafts in undrained uplift.
5. Summary
loading. A target reliability index of 3.2 is selected based on the reliability indices
implied by existing working stress designs. Two simple design formats (load and
resistance factor design and multiple resistance factor design) are rigorously cali-
brated using the ®rst-order reliability method to produce designs that achieve a
known level of reliability consistently. The resulting resistance factors are summar-
ized in a simple tabular form that can be easily applied to practice.
Acknowledgements
This paper is based, in part, on research sponsored by the Electric Power Research
Institute. The EPRI project manager is A. Hirany.
References
[1] Code of practice for structural use of concrete. London: British Standards Institution, 1972 [CP110
(Pt 1)].
[2] Cold-formed steel structural members. Committee on Design of Light Gauge Steel Structural Mem-
bers (N. C. Lind, Chair), Standard S136, Canadian Standards Association, Rexdale, 1974.
[3] Recommendations for loading & safety regulations for structural design. Report 36, Nordic Com-
mittee on Building Regulations (NKB), Copenhagen, 1978.
[4] ACI Committee 318, Building code requirements for reinforced concrete. ACI 318±83, American
Concrete Institute, Detroit, 1983.
[5] Load & resistance factor design manual steel construction. 1st ed. Chicago, American Institute of
Steel Construction, 1986.
[6] Recommended practice for planning, designing and constructing ®xed oshore platforms Ð load
and resistance factor design. API RP 2A-LRFD, American Petroleum Institute, 1993.
[7] CEN/TC250. Basis of design and actions on structures Ð Part 1, basis of design. Eurocode 1, ENV-
1991-1, European Committee for Standardisation (CEN), 1993.
[8] General principles on reliability for structures. ISO 2394, International Standard Organisation, 1998.
[9] Criswell, M.E., Vanderbilt, M. Reliability-based design of transmission line structures: methods.
Report EL-4793 (1), Electric Power Research Institute, Palo Alto, 1987.
[10] Task Committee on Structural Loadings. Guidelines for electrical transmission line structural load-
ing. Manual & report on engineering practice 74, New York: ASCE, 1991.
[11] IEC Technical Committee 11. Loading & strength of overhead transmission lines. Technical Report
826, International Electrotechnical Commission, Geneva, 1991.
[12] Meyerhof GG. Development of geotechnical limit state design. Canadian Geotechnical Journal
1995;32(1):128±36.
[13] Becker DE. Limit states design for foundations. Part I. An overview of the foundation design pro-
cess. Canadian Geotechnical Journal 1996;33(6):956±83.
[14] Becker DE. Limit states design for foundations. Part II. Development for the national building code
of Canada. Canadian Geotechnical Journal 1996;33(6):984±1007.
[15] CEN/TC250. Geotechnical design Ð Part 1. General rules. Eurocode 7, ENV-1997-1, European
Committee for Standardisation (CEN), 1994.
[16] Research Committee on Present & Future of Japanese Foundation Design and Soil Investigation in
View of International Equivalency (O. Kusakabe, Chair). Foundation design standards in the
world Ð toward performance-based design. Japanese Geotechnical Society (1998).
[17] Barker RM, Duncan JM, Rojiani KB, Ooi PSK, Tan CK, Kim S.G. Manuals for design of bridge
foundations. NCHRP Report 343, Transportation Research Board, Washington, 1991.
K.-K. Phoon et al. / Computers and Geotechnics 26 (2000) 169±185 185
[18] Phoon K-K, Kulhawy FH, Grigoriu MD. Reliability-based design of foundations for transmission
line structures. Report TR-105000, Electric Power Research Institute, Palo Alto, 1995.
[19] Casagrande A. Role of calculated risk in earthwork and foundation engineering. Journal of Soil
Mechanics & Foundations Division (ASCE) 1965;91(SM4):1±40.
[20] Focht Jr JA, O'Neill MW. Piles & other deep foundations. Proceedings, Eleventh International
Conference on Soil Mechanics & Foundation Engineering, San Francisco, 1985;1:187±209.
[21] Vick SG. Risk in geotechnical practice. Geotechnical News 1992;10(1):55±7.
[22] MacGregor JG. Problem from viewpoint of structural engineer. Symposium on Limit States Design
in Foundation Engineering, Canadian Geotechnical Society, Toronto, 1989, 1±16.
[23] Kulhawy FH, Phoon KK. Engineering judgment in the evolution from deterministic to reliability-
based foundation design. Uncertainty in the geologic environment Ð from theory to practice (GSP
58), ASCE, New York, 1996. p. 29±48.
[24] Kulhawy FH, Trautmann CH, Beech JF, O'Rourke TD, McGuire W, Wood WA, et al. Transmis-
sion line structure foundations for uplift-compression loading. Report EL-2870, Electric Power
Research Institute, Palo Alto, 1983.
[25] Stas CV, Kulhawy FH. Critical evaluation of design methods for foundation under axial uplift &
compression loading. Report EL-3771, Electric Power Research Institute, Palo Alto, 1984.
[26] Chen Y-J, Kulhawy FH. Case history evaluation of behavior of drilled shafts under axial & lateral
loading, Report TR-104601, Electric Power Research Institute, Palo Alto,1994.
[27] Code of Practice for Foundation Engineering. Bulletin 36. Copenhagen: Danish Geotechnical Insti-
tute, 1985.
[28] Technical Committee on Foundations. Canadian foundation engineering manual. Richmond:
Canadian Geotechnical Society, 1992.
[29] Kulhawy FH. On evaluation of static soil properties. Stability & performance of slopes and
embankments II (GSP 31), ASCE, New York, 1992. p. 95±115.
[30] McManus KJ, Kulhawy FH. Cyclic axial loading of drilled shaft foundations in cohesive soil for
transmission line structures. Report EL-7161, Electric Power Research Institute, Palo Alto, 1991.
[31] Melchers RE. Structural reliability, analysis, & prediction. New York: Wiley, 1987.
[32] Baecher GB, Pate ME, De Neufville R. Risk of dam failure in bene®t-cost analysis. Water Resources
Research 1980;16(3):449 Ð 456.
[33] Construction Industry Research & Information Association. Rationalization of safety & service-
ability factors in structural codes. Report 63, CIRIA, London, 1977.
[34] Ellingwood B, Galambos TV, MacGregor JG, Cornell CA. Development of probability-based load
criterion for American national standard A58. Special Publication 577, National Bureau of Stan-
dards, Washington, 1980.
[35] Baecher GB. Geotechnical risk analysis user's guide. FHWA/RD-87-011, Federal Highway Admin-
istration, McLean, 1987.