Publications Details

An iterative Bayes procedure for reliability assessment

Prairie, R.R.

In component reliability assessment programs, three major sources of data are available for reliability assessment: a development program, production data, and field test data. In practice, reliability assessments are required at the end of each stages, and a common method of assessment is to simply combine the test data and provide a pooled estimate. The method suggested in this paper is Bayesian in that the uncertainty about the unreliability is expressed by means of a prior distribution with a specified upper limit. The method is hierarchical Bayes in that the uncertainty about the limit of that prior distribution is also expressed by means of a prior distribution. The data from the development program are incorporated with the prior on the unreliability and with the prior on the upper limit of the prior to obtain a new prior on unreliability. The production data are then used to obtain a revised estimate of the unreliability as well as a modified value for the limit of the prior distribution. This same concept will be carried through when the field data are obtained. The result is a final Bayesian reliability assessment that is iterative in nature and incorporates in a sequential fashion data from each of the three stages common to a component development, production, and surveillance program. 4 refs., 2 tabs.