Approximating Posterior Distributions of System Reliability Derived from Component Test Data.
Abstract
The determination of exact interval estimates for system reliability, when only the components of the system have been tested, is difficult both from a theoretical and a computational point of view. This work develops approximations to exact Bayesian intervals using asymptotic expansions and other techniques. Several approximations prove to be very accurate, especially in the region where computation of exact results becomes formidable. The assessment of accuracy is by comparison with exact results, when these are available. For systems with complex structure comparisons are with results obtained by simulating the posterior distribution of system reliability using Monte Carlo methods. Test data considered can be either pass/fail or data obtained from procedures for which failure times are exponentially distributed. Mixed test data and repeated components are also considered. (Author).
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1984
- Accession Number
- ADA152057
Entities
People
- A. Winterbottom
Organizations
- City, University of London