Application Research On Bayesian Approach Of Small-sample Data In Nuclear Power Plant

Reliability data processing of systems, structures and components (SSCs) in nuclear power plant is the foundation of RCM (reliability-centered maintenance) and LCM (life cycle management). An appropriate processing method gives accurate results, which determines the scientificalness of the maintenance decision-making. For the components in the condenser, the main power plant generator and other plant systems, as high security is required and long life time is designed, the problems such as small-sample, non-independent data caused by maintenance and equipment aging are put forward, and the traditional statistical methods no longer apply.In this paper, the complex spot data are formed into the format of fixed time censored test without replacement. Then the repairable system model and data-trend analysis method are introduced. Two parameters Weibull distribution is used as the life distribution form. Using Bayes method, there gives a general process to solve the small sample problems. Based on the practical operation failure data in domestic nuclear power plants, the results show that this process provided better accuracy and wider applicability.Under the Weibull distribution, Bayes formula hardly gets analytical results. A series of Bootstrap sub-samples are used as the prior distribution of the parameters to process the numerical calculation. A brief discussion about robustness and applicability is presented.In the end, Bayes method is used to solve the zero-failure problems. The packet data, pollution data and missing data problems are briefly discussed.