Mining data in a dynamic PRA framework

Abstract

Computational, also known as Dynamic, Probabilistic Risk Assessment (PRA) methods employ system simulation codes coupled with stochastic analysis tools in order to determine probabilities of certain outcomes such as system failure. In contrast to Classical PRA methods (i.e., Event-Tree and Fault-Tree) in which timing and sequencing of events is set by the analyst, accident progression is dictated by the system control logic and its interaction with the system temporal evolution. Due to the nature of the problem, Dynamic PRA methods can be expensive form a computational point of view since a large number of accident scenarios is simulated. Consequently, they also generate a large amount of data (database storage may be on the order of gigabytes or higher). We investigate and apply several methods and algorithms to analyze these large time-dependent data sets. The objective is to present a broad overview of methods and algorithms that can be used to improve data quality and to analyze and extract information from large data sets containing time dependent data. In this context, “extracting information” means constructing input-output correlations, finding commonalities, and identifying outliers.