Use of probabilistic methods in fire hazards analysis

An overview of probabilistic methods developed for evaluating the risks associated with fires in nuclear power plants is presented. A two stage process is found to be most efficient. It involves simplified, but conservative, screening of all plant locations, followed up by detailed analysis of those fire scenarios found to be potentially risk significant. The analysis uses both statistical data and engineering judgement to evaluate fire initiation frequencies, suppression system response times and barrier fragilities, in combination with physical models to predict the rate of fire growth and the resulting thermal hazards. The loss of equipment operability coincident with, but unrelated to, the fire is considered, thus providing a more complex and realistic picture of the fire impact. The completed fire risk analysis forms an integral part of a full scope plant risk model, which can be used as a management tool for identifying weaknesses in design and mode of operation, and for prioritizing improvements. The results and insights gained from previous applications of the approach are summarized. (author). 14 refs, 4 figs, 3 tabs