Stochastic Reliability Analysis – the Data Interface

In two previous papers [1, 2], the general model of stochastic reliability analysis (SRA) was developed. However, the problem of determining model parameters and their uncertainty distributions from operational experience on the failure of redundant components was not satisfactorily solved. This deficiency is removed by the present paper which establishes the data interface in a stochastically adequate way within a Bayesian approach.A single probability density junction for a heterogeneous population is constructed which contains both information about relevant operational experience and necessary modifications resulting from expert judgement. Sum rules and normalization factors are extensively used in order to keep the composite model simple, transparent and flexible at every stage of development. Occurrence frequencies for (single and) multiple dependent failure of redundancy groups with arbitrary degree of redundancy can be calculated in a unified way.Numerical results obtained from the SRA model, the Basic Parameter and the Binomial Failure Rate model are compared with example operational experience used as data input to these models before. Furthermore, a stochastic correction to the Basic Parameter model is proposed