Abstract This paper suggests some improvement of reliability estimation in the framework of the Stress-Strength Inference Theory, essentially based on a limit state model. The general purpose is to develop the reliability estimation models more closely corresponding to the real state in nature. Two different methods of improvement are studied. The first method assumes that the limit state model has an error which reflects actual imperfection of an applied physical theory. The error is considered to be uncertain. According to the second method, test data used for the estimation are studied to be vague. Two approaches to describing uncertain variables and non-precise data are used and compared in this paper. The first approach is a combination of the so-called fuzzy numbers method and the Bayesian one, using the former to describe the data vagueness and the latter as a basis for estimation under uncertain parameters. The second approach is exclusively based on the Bayesian approach. Both prior procedures and posterior ones are worked out using these approaches. Application and numerical examples are given to illustrate the results. This representation shows clearly how the two approaches relate to each other.
[1]
G. Schuëller,et al.
A critical appraisal of methods to determine failure probabilities
,
1987
.
[2]
Kailash C. Kapur,et al.
Reliability in engineering design
,
1977
.
[3]
Granino A. Korn,et al.
Mathematical handbook for scientists and engineers
,
1961
.
[4]
J. Berger.
Statistical Decision Theory and Bayesian Analysis
,
1988
.
[5]
Charles O. Smith,et al.
Introduction to reliability in design
,
1976
.
[6]
H. Hule,et al.
On Bayes' theorem for fuzzy data
,
1991
.
[7]
Walter Ledermann,et al.
Handbook of applicable mathematics
,
1980
.
[8]
H. Jeffreys,et al.
The Theory of Probability
,
1896
.
[9]
C. Bucher,et al.
On Efficient Computational Schemes to Calculate Structural Failure Probabilities
,
1989
.