An improved monte carlo method in fault tree analysis

Abstract

The Monte Carlo (MC) method is one of the most general ones in system reliability analysis, because it reflects the statistical nature of the problem. It is not restricted by type of failure models of system components, allows to capture the dynamic relationship between events and estimate the accuracy of obtained results by calculating standard error. However, it is rarely used in Fault Tree (FT) software, because a huge number of trials are required to reach a tolerable precision if the value of system probability is relatively small. Regrettably, this is the most important practical case, because nowadays highly reliable systems are ubiquitous. In the present paper we study several enhancements of the raw simulation method: variance reduction, parallel computing, and improvements based on simple preliminary information about FT structure. They are efficiently developed both for static and dynamic FTs. The effectiveness and accuracy of the improved MC method is confirmed by numerous calculations of complex industrial benchmarks.