Publications Details
A hierarchical Bayesian approach to passive system reliability analysis
One source of concern in the nuclear power community is associated with performing PRAs on the passive systems used in Advanced Light Water Reactors. Passive systems rely on physical phenomena in order to perform safety actions. This leads to questions about how one should model the reliability of the system, such as how one should model the uncertainty in physical parameters that define the operational characteristics of the passive system and how to determine the degradation and failure characteristics of a system. Hierarchical Bayesian techniques provide a means for assessing the types of problems presented by passive systems. They allow the analyst to collect multiple types of data, including expert judgment and historical data from different sources and then combine them in one analysis. The importance of this feature is that it allows an analyst to perform a mathematically consistent PRA without large amounts of data for the specific system under scrutiny. As data become available, they are incorporated into the analysis using Bayes' rule. As the dataset becomes large, the data dominate the analysis. A study is performed whereby data are collected from a set of resistors in a corrosive environment. A model is created that related the environmental conditions of the sensors being used to the performance of the sensors. Prior distributions are then proposed for the uncertain parameters. Both longitudinal and failure data are recorded for the sensors. These data are then used to update the model and obtain the posterior distributions related to the uncertain parameters.