The design of optimum component test plans in the demonstration of system reliability

System Testing is often not feasible, and it is a common practice in these situations to attempt to draw inferences about the reliability of the system from the results of component tests. An acceptance procedure based on the sum of failures during component tests has been previously suggested in the literature for a series system, a parallel system, and a serial connection of redundant subsystems. Methods for computing the individual component test times which minimize total test cost and guarantee the probability requirements for this acceptance procedure and for such system structures have also been given; none of them considers a priori information on component reliabilities. In this paper we consider the question of whether these optimum solutions are modified if a priori deterministic information on individual component failure rates is available. Results have shown that it is profitable to use the given information. We also provide a procedure to compute optimum component test times for general systems. This procedure is based on the well known cutting plane idea and column generation technique, and eventually solves a semi-infinite linear program. We also discuss the conditions for the existence of optimum component test times.