Dagger-Sampling Monte Carlo For System Unavailability Evaluation

Reliability problems usually result in rare-event simulations, and hence direct Monte Carlo methods are extremely wasteful of computer time. This paper presents a new application of ``dagger-sampling'', for calculating the system unavailability of a large complicated system represented by a coherent fault tree. Since a small number of uniform random numbers generate a number of trials, dagger-sampling appreciably reduces computation time, and hence a large number of trials become possible for the rare-event problems. Further, dagger-sampling decreases the variance of the Monte Carlo estimator because it generates negatively correlated samples.