Analysis of decision strategies for supervising process safety

This paper discusses the safety analysis of complex systems, as well as those who are responsible for the specification, inspection and approval of such systems. The paper addresses the problem by considering safety-and reliability- failures as classification errors, and by using available mathematical tools for calculating the quality of a strategy in terms of both safety and reliability. In particular, we show that an optimal safety/reliability trade-off is achieved in strategies satisfying the Neyman-Pearson optimality criterion. In safety terminology, this suggests that in choosing a strategy, the maximum risk that is allowed in the process is a more informative factor than the total risk of safety failure. Adopting the idea of maximizing process reliability for a given maximum risk, instead of for the total probability of safety failure, we reach a class of strategies that we term optimal strategies, as opposed to strategies merely satisfying the Neyman-Pearson criterion, which we denote quasi-optimal strategies.