The strengths and limitations of PSA: where we stand

Abstract

Abstract Despite increasing maturity in probabilistic safety assessment (PSA) methods, there are several problems that create discomfort among decision makers. These include the use of expert opinions, the assessment of human reliability, and the impact of organizational factors. These problems are all manifestations of the larger problem of uncertainty in the real world and how that uncertainty is presented within the context of the PSA. Examples are presented to show how these issues have been addressed in a few existing PSA studies. The expert opinion issue is reframed as a representation of the gathering and evaluating of expert evidence. Uncertainty in physical models is approached by applying expert evidence to judging the likelihood of alternative models. Confidence in human reliability modeling can only be achieved if substantial effort is dedicated to qualitatively describing the scenario for analysis: the history of plant states and human intervention, as well as the ensemble of cues, training, and applicable experience. Organizational factors are structured as the key underlying dependency linking human reliability and equipment availability. Examples of other major sources of uncertainty—equipment aging, design and construction errors, common cause failures, plant area heatup, and shutdown events—are presented. The continuous thread through all the examples is that identifying and treating uncertainty explicity is the key to generating confidence in PSA among decision makers. PSA is shown to provide a language for quantifying uncertainty, a clear exposition of plant safety, and a flexible tool for managing safety. The framework that is PSA is, by its very nature, robust and able to provide the ability to address the identified problems.