Bayesian Reliability Assessment for Systems Program Decisions

摘要

Bayesian statistics provide the necessary mathematical techniques to pool all available subjective and experimental information when estimating reliability. The uncertainties associated with analytical predictions or limited test data considered separately are significantly reduced when these two sources of information are combined. The introduction of judgment and pertinent enginering theory and experience to qualify point estimates is the key to realistic and practical solutions to decision problems in which reliability is a primary consideration. A method for periodic reliability assessment is presented. A hypothetical example is used to show how iterative inference on system reliability can be drawn from initial estimates of unit/subsystem reliability and heterogeneous time and failure data accumulated during various stages of design verification, electrical performance, environmental, etc. testing. (Author)