Towards the Automated Verification of Weibull Distributions for System Failure Rates

Abstract

Weibull distributions can be used to accurately model failure behaviours of a wide range of critical systems such as on-orbit satellite subsystems. Markov chains have been used extensively to model reliability and performance of engineering systems or applications. However, the exponentially distributed sojourn time of Continuous-Time Markov Chains (CTMCs) can sometimes be unrealistic for satellite systems that exhibit Weibull failures. In this paper, we develop novel semi-Markov models that characterise failure behaviours, based on Weibull failure modes inferred from realistic data sources. We approximate and encode these new models with CTMCs and use the PRISM probabilistic model checker. The key bene t of this integration is that CTMC-based model checking tools allow us to automatically and e ciently verify reliability properties relevant to industrial critical systems.