Worst pattern analysis for Markov reliability models

This paper proposes a new approach to computing the reliability of a system. Worst pattern analysis finds the component misbehavior pattern that produces the greatest system unreliability. The motivation is that it is difficult and expensive to obtain information about the behavior of faulty components. It is also difficult and expensive to obtain information about a system's response to faulty components. For these reasons, this approach is based on minimum information about faulty components, and minimum requirements for system response. There are two minimum requirements. First, quantitative reliability requires that the failure rate of the components be known. Second, long term reliability requires that the system recognize when a redundant unit is misbehaving. In the worst pattern approach all other information is replaced by a computational effort. The procedure is illustrated by an example.