Maximum likelihood estimation of probabilistic non-suppression model for OECD NPP electrical fire applying non-negative continuous distribution

Abstract

The securement and improvement of realism in a probabilistic fire risk assessment (Fire PRA) are important in risk-informed performance-based regulations and decision support. In this study, a probabilistic fire brigade non-suppression model is developed for electrical fires using the Organization for Economic Cooperation and Development fire incident data on operating nuclear power plants collected from various countries by applying a non-negative continuous probability distribution with the maximum likelihood estimation method. The result of fitting 15 types of a non-negative continuous probability distributions shows that the log-normal probability model is the best fitting and most adequate model, and can best represent the actual fire suppression time by a fire brigade. The selected log-normal probability model was compared with the exponential probability model being used in an existing Fire PRA, which shows that the level of adequacy of the log-normal probability model is improved with a decrease in the bayesian information criteria by 7.9%, residual sum of squares by 100.0%, and mean squared error by 57.6%. The log-normal probability model selected from this study is expected to contribute to an enhancement of the Fire PRA realism in support of risk-informed decision making by reflecting actual fire suppression experience.