Statistical sampling techniques for improving probabilistic risk assessment analyses

Unified Stratified Sampling (USS) is introduced as a newly-developed sampling technique, and its application to a problem in Probabilistic Risk Assessment is illustrated. Comparative analyses of two commonly-used stratified sampling techniques, Latin Hypercube Sampling (LHS) and Factorial Stratified Sampling (FSS) are also presented, together with their advantages and disadvantages and the domains of their appropriate application. Hypothetical examples that demonstrate the superiority of USS over other stratified sampling techniques are given. In addition, a simple practical analysis for determining the failure probability of a backflow prevention system is included. The advantage of USS is its ability to achieve a desired precision in failure probability estimates using substantially less computer time than other sampling techniques. USS can be used effectively for any type of statistical analysis which requires time-consuming simulation runs to achieve the objective of the analysis