Bayesian reliability assessment for discrete data: a case study

Abstract The ever-increasing reliability requirement and reduced programme schedules and budgets, particularly in the field of space research, make very real the problem of producing reliability assessments without a large volume of objective test data. In this paper, an attempt is made to utilize Bayesian statistics for reliability measurement of a complex system by combining prior information actual test data for each subsystem. Bayesian statistics, as applied herein, codes the information available from past experience into a prior probability distribution which is updated with test results as they became available. The study includes: 1. (i) Bayesian models for discrete data analysis; 2. (ii) application of the Monte Carlo simulation technique for convolving subsystem reliability distributions together in order to arrive at the overall system reliability and 3. (iii) implementation of the Bayesian reliability measurement programme for the evaluation of the reliability estimate of a typical system of the satellite launch vehicle.