Development of systems performance standards based on probabilistic risk assessments

Probabilistic risk assessments (PRA) of nuclear reactors now represent a substantial data base which describes relative plant accident potentials, plant failure modes, and plant system failure probabilities. The PRA analyses can be used to make design related decisions. In the past, however, nuclear plant design relied primarily on traditional engineering approaches along with quality control and regulatory review. In this study, the authors try to see whether system and subsystem performance standards can be extracted from a set of existing PRAs using contribution to core melt as the primary attribute of interest. The approach taken uses system and plant failure data from six pressurized water reactor PRAs. Importance measures are developed and applied to gage the impact of various systems on plant core melt. These measures are also applied to modified plant configurations which meet a dominant sequence core melt contribution limit of 1.0 x 10-5 core melts per plant per reactor year. In addition, plant failure modes are put into functional groups, and sequences are placed into accident classes. Together, these approaches define the failure characteristics of the six plants and seek to discern failure patterns and the defacto existence of performance standards. A method of limiting the sensitivity of the plant (measured as an increase in core melt frequency) to hypothetical increases in event unavailabilities is proposed. Bounding estimates of the effect of the standard pairs on plant core melt frequency are derived