NONLINEAR MONTE-CARLO RELIABILITY-ANALYSIS WITH BIASING TOWARDS TOP EVENT

This paper deals with the Monte Carlo evaluation of the reliability and availability of a complex system made up of a large number of components, each with many possible states. To make allowance for the fact that in reality the transition probabilities depend on the system configuration, so that the problem becomes highly nonlinear, a connection model in which all the transition probabilities may be suitably varied after each transition occurrence, has been developed. In the present work the model obeys the usual Markovian assumption of transitions dependent on the present system configuration, but removal of this assumption is easy and it would account for system aging. To drive the system towards the more interesting but highly improbable cut set configurations, a variance reduction technique, based on the introduction of distances between the present and the cut set configurations, is also proposed.