Modeling analyses and data in human reliability

Abstract

The safety of nuclear power plants must be proved, certified and improved. Probabilistic safety assessments are used to estimate the core meltdown risk, by means of sequential analyses of accidents. In order to assess probabilities of the appearance of these sequences, it is necessary to specifically assess probabilities of operation failures accomplished by human operators in a degraded mode. For this purpose, EDF, the French producer of electricity, developed a method that models failures of human actions, by means of a systematic determination of scenarios corresponding to different failure modes.

This method, called MERMOS, has been used for several probabilistic safety assessments. In order to increase its reproducibility and to make it more robust, example missions and scenarios will be built. This set of example analyses will be used by experts assessing human reliability: they will develop studies and deduce results more easily.

The purpose of this study involves the creation of a methodology to model existing analyses and human reliability data used in MERMOS. This study consists of optimizing a second generation human reliability assessment method in order to overpass its current weaknesses in an operational context by means of the identification of a set of example analyses.