Coyan: Fault Tree Analysis – Exact and Scalable

We revisit one of the oldest problems in probabilistic safety analysis and present a breakthrough in scalability and precision. This paper develops a novel, scalable and exact method to compute the unreliability of a system modelled by a Fault Tree (FT). This unreliability corresponds to the top event probability (TEP) of the tree. Our method harvests recent advances in Weighted Model Counting (WMC), together with tailored encodings mapping each FT to an equisatisfiable Boolean formula and tailored weight assignments. We then resort to state-of-the-art WMC solvers to obtain an exact value of the TEP. This approach delivers precise results in high reliability scenarios and it scales to very large and complex trees, where so far one needed to resort to approximate analysis (e.g., based on minimal cut sets). We provide empirical evidence regarding superiority of our method by comparing to state-of-the-art tools for Fault Tree Analysis across a large set of Fault Tree benchmarks.