Synergizing Reliability Modeling Languages: BDMPs without Repairs and DFTs

Abstract

Static Fault Trees (SFTs) are a key model in reliability and safety analysis. Various extensions have been developed to model, e.g., functional dependencies, state-dependent failures, and SPARE elements. This paper studies the expressive power of two important extensions of SFTs: Dynamic Fault Trees (DFTs) and Boolean Logic Driven Markov Processes (BDMPs). We outline a set of BDMP-to-DFT translation rules and apply them to thirty-three BDMP test cases modeling various scenarios of security, software and system reliability. The main contribution is a DFT modeling an industrial BDMP benchmark study of a Nuclear Power Plant (NPP). Although this DFT does not consider repairs, it is one of the largest industrial cases reported so far and is challenging for DFT analysis. We compare the performance and capabilities of analysis tools for BDMPs—the Monte-Carlo simulation tool YAMS, the proprietary Markovian analysis tool FigSeq—and the DFT analysis capability of the probabilistic model checker Storm. We also address how to do a system sensitivity analysis of the NPP benchmark using probabilistic model checking.