Sensitivity analysis approach for robust probabilistic risk assessment
The main objective of this investigation is to develop a robust and simplified Probabilistic Risk Assessment (PRA) approach specifically oriented to produce results for risk management decisions of high technology systems. The techniques are based on defining a set of three Significant Indices which quantify the importance of each component, and hence develop a ranking of the components, both with respect to the mean and variance, in the fault tree/event tree structure. The variations in the Top Event probability distribution upon the variations in the component input probability distributions are also evaluated, as well as the first and second moments of the Top event;A method is devised to analyze the sensitivity of the component-ranking upon the basic event probability distribution model. If the new set of Significance Indices, developed as a result of the sensitivity analysis, do not warrant a substantial change in the ranking of the components, then the PRA results are defined here as robust. An important feature of the Significance Indices, and hence the robustness of the PRA results, is the convenience of finding the mean and variance of the Top event due to single or multiple variations in the mean and variance of one or more basic events in the system model.