Professional Documents
Culture Documents
ON STRUCTURAL RELIABILITY
By Bruce Ellirtgwood, 1 M . ASCE
due to errors in planning, design, construction, and utilization, rather than sto-
chastic variability in construction material strengths and structural loads. Re-
cently developed probability-based load and resistance factor design methods
treat uncertainties due to inherent randomness and modeling more rationally
than is possible with current design methods, but still do not take errors in
design and construction into account. Any meaningful attempts to reduce fail-
ure rates will need to deal with this issue. This paper reviews the status of
design and construction errors in structural safety studies, summarizes some
simple mathematical tools for their analysis, and demonstrates why errors are
so difficult to treat in codified design.
INTRODUCTION
409
such errors. An error is not simply an extreme value of one of the load
or resistance variables used in the reliability analysis or in design. Rather,
the error corresponds to a different event entirely, one that may change
the probabilistic models that are applicable and the relevant limit state
as well. Thus, calculations of probability of failure that take into account
only chance occurrences of higher-than-design loads in combination with
lower-than-design resistances are not realistic. Moreover, attempts to in-
clude error by adjusting the distributions of the basic variables or factors
of safety generally are ineffective because they do not attack the root of
the problem (22).
This paper reviews the current state of knowledge concerning errors
that may lead to failures of structural systems, and considers how such
errors might be taken into account in evaluating structural reliability.
Although systematic procedures to mitigate errors ultimately should be
included in the management of structural safety, these have not been
developed within the limited scope of this review.
411
412
Ignorance, Reliance
negligence, Insufficient on Other
carelessness knowledge Forgetfulness, others sources Total
Reference (%) (%) mistakes (%) (%) (%) (%)
(1) (2) (3) (4) (5) (6) (7)
Downloaded from ascelibrary.org by University of Newcastle on 09/15/14. Copyright ASCE. For personal use only; all rights reserved.
developed from the event tree shown in Fig. 1. Let F = event of failure
of a structure or structural component; F = Fs U Fe, in which Fs = failure
due to stochastic variability in loads and capacities, and Fe = failure due
to error (we assume, naively perhaps, that these are the only alterna-
tives). Let E = event that error occurs and is undetected. Invoking the
theorem of total probability, the probability of failure, P(F), can be ob-
tained as
P(F) = [P(Fe\E) + F(FS\E)]P(E) + P(FS\E)P(E) (1)
in which E = event that a gross error does not occur; and P(F\E) = prob-
ability of failure on the condition that event E occurs. Eq. 1 emphasizes
that failures due to stochastic variability are possible even when errors
occur and are undetected; moreover, failure modes might differ for fail-
ures due to stochastic variability and for those due to error. However,
P(E)-P(F | E )
/ P(E)-P(F s E)
X
^ Failure due to
J~\. stochastic variability
E /Error ^ ^ No failure
/ occurs \.
\. P(E)-P(F|E)
P(E)'P(FSI'B)
Failure due to
stochastic variability
No failure
N . P(E)-P(F|E)
414
in which O, = event that the ;'th error occurs; and D, = event that the
r'th error is detected (and repaired). The probability of failure due to er-
ror is then
r
P FE
( ) = E P(F|AO,)P(A|0,)P(0,) (4)
1=1
Note the equivalence of event E and the event (D, fl Of) in Eqs. 2
and 3.
These simple models all contain the important notions of error con-
sequence, detectability (and correction), and resulting consequence. Eqs.
1-4 show that structural safety can be managed by controlling the in-
cidence of errors [limiting P(E)], by controlling the impact and conse-
quences of the errors on structural performance (limiting P(F\E) by de-
signing the structure to tolerate and absorb the effect of errors), or by a
combination of these strategies.
The probability P(F\E) in Eq. 2 is the classical failure probability treated
in reliability analyses conducted to date. Recently developed probability-
based load and resistance criteria (11,13,17) have been selected so as to
be consistent with a prescribed target value of P(F\E). The terms P(F\E)
and P(E) remain to be evaluated. Error survey results summarized pre-
viously indicate that these latter terms dominate P(F).
Evaluation of P(E).—The probability of undetected error, P(E), is a
function of the level of inspection and checking. If the errors occur only
rarely in time, the error occurrence can be modeled as a Poisson random
point process (30,44) and (dropping subscript i for convenience, cf. Eqs.
3 and 4)
P(0) = XT (5)
in which X = mean rate of occurrence of error during time period T. If
the probability that the error is not detected is q = 1 - P(D), then
P(E) = P(D 0 0) = q\T (6)
If quality assurance programs can be modeled as a series of n statis-
tically independent control stops (35), the error probability P(E) is
415
CFS 2 f ,/ P2\ ./ P1
{Vl + V2a)1/2 I (15)
It has been suggested (9) that the failure probability due to stochastic
variability [P(FE) in Eq. 1] and failure probabilities due to design /con-
struction errors [P(FE) in Eq. 1] should not be combined into a single
measure of safety, P(F), Rather, the measure of safety should be ex-
pressed as a pair of probabilities, [P(FE), P(FE)], that can be considered
417
Accordingly, while P(F\E) can be used as a basis for selecting load and
resistance factor criteria for design, P(FE) should be used in choosing
among design concepts, construction procedures, and quality assurance
plans.
CONCLUDING REMARKS
Structural codes do not address the error problem, other than through
general quality assurance statements. Current or proposed safety checks
such as LRFD (24) are not developed with errors in mind. Considering
the almost infinite number of ways for error to occur in the building
process, it may be unreasonable to expect that codes will ever be able
to give prescriptive criteria aimed at error reduction.
Nonetheless, public and professional reaction to recent structural fail-
ures has made it apparent that reductions in the rate of occurrence and
the effects of error in the building process would be desirable. Engineers
should recognize that things can and do go wrong. They also should
think in terms of hazard scenarios and how the structural system could
fail, as well as in terms of design loads and resistances. Such perfor-
mance-oriented thinking cannot be mandated by code, but it seems es-
sential if significant reductions in error incidence are to be achieved.
ACKNOWLEDGMENTS
APPENDIX.—REFERENCES
1. Allen, D. E., "Criteria for Design Safety Factors a n d Quality Assurance Ex-
Downloaded from ascelibrary.org by University of Newcastle on 09/15/14. Copyright ASCE. For personal use only; all rights reserved.
420
22. Knoll, F., '"Human Error in the Building Process—A Research Proposal,"
IABSE Periodica 4/1982, International Assocation for Bridge and Structural
Engineering, Zurich, Switzerland, 1982.
23. Lind, N., and Davenport, A. G., "Towards Practical Application of Structural
Reliability Theory," Probabilistic Design of Reinforced Concrete Buildings, ACI SP
31, Detroit, Mich., 1972, pp. 63-110.
24. Load and Resistance Factor Design Specification for Structural Steel Buildings (Pro-
posed), American Institute of Steel Construction, Chicago, 111., 1983.
25. Matousek, M., "Measures Against Errors in the Building Process," Institute
of Structural Engineering, Swiss Federal Institute of Technology, Zurich,
Switzerland, 1982 (Translation No. 2067, National Research Council of Can-
ada, Ottawa, Canada/1983).
26. Matousek, M., "A System for a Detailed Analysis of Structural Failures,"
Structural Safety and Reliability, Elsevier, Amsterdam, 1981, pp. 535-544.
27. Melchers, R. E., Baker, M. J., and Moses, F., "Evaluation of Experience,"
Proceedings, IABSE Workshop on Quality Assurance Within the Building Pro-
cess, Rigi, Switzerland, 1983, pp. 9-30.
28. Melchers, R. E., and Harrington, M. V., "Structural Reliability as Affected
by Human Error," Proceedings, Fourth International Conference on Appli-
cations of Probability and Statistics in Soil and Structural Engineering, Flor-
ence, Italy, June, 1983, pp. 683-694.
29. Melchers, R. E.,, and Stewart, M. B., "Data-Based Models for Human Error
in Design," Proceedings, International Conference on Structural Safety and
Reliability, Kobe, Japan, 1985, pp. 51-60.
30. Nessim, M. A., and Jordaan, I. J., "Models for Human Error in Structural
Reliability," Journal of Structural Engineering, ASCE, Vol. I l l , No. 6, June,
1985, pp. 1358-1376.
31. Nowak, A., "Effect of Human Error on Structural Safety," Journal of the Con-
crete Institute, Vol. 76, No. 9, Sept., 1979, pp. 959-972.
32. Nowak, A., and Carr, R., "Sensitivity Analysis of Structural Errors," Journal
of Structural Engineering, ASCE, Vol. I l l , No. 8, Aug., 1985, pp. 1734-1746.
33. Ofverbeck, P., "Control of Gross Errors," Report TVBK-3005, Department of
Structural Engineering, Lund Institute of Technology, Lund, Sweden, 1979.
34. "Quality Control and Quality Assurance for Concrete Structures," Bulletin D'
Information No. 157, Comite Euro-international de Beton, Paris, France, 1983.
35. Rackwitz, R., and Hillemeier, B., "Planning for Quality," Proceedings, IABSE
Workshop on Quality Assurance within the Building Process, Rigi, Switz-
erland, 1983, pp. 31-52.
36. Ravindra, M. K., and Campbell, R. D., "Treatment of Design and Construc-
tion Errors in PRA," Structural Engineering in Nuclear Facilities, Vol. 1, J. J.
Ucciferro, Ed., ASCE, New York, N.Y., 1984, pp. 312-331.
37. Ravindra, M. K., Banon, H., Sves, R., and Thrasher, R. D., "Sensitivity Studies
of Seismic Risk Models," Research Report No. NP-3562, Electric Power Re-
search Institute, Palo Alto, Calif., 1984, pp. 6-1-6.6.
38. Schneider, J., "Organization and Management of Structural Safety During
Design, Construction and Operation of Structures," Structural Safety and Re-
liability, Elsevier, Amsterdam, The Netherlands, 1981, pp. 467-482.
39. Sibly, P. G., and Walker, A. G., "Structural Accidents and Their Causes,"
Proceedings, Part 1, Institution of Civil Engineers, Vol. 62, May, 1977, pp.
191-208.
40. Stevenson, J. D., "An Estimate and Evaluation of Design Error Effects on
Nuclear Power Plant Design Adequacy," Structural Engineering in Nuclear Fa-
421
42^