Synergising Reliability Modelling Languages: BDMPs and Repairable DFTs

Abstract

Adding repairs to dynamic fault trees (DFTs) is intricate and has given rise to several different, unfortunately inconsistent, interpretations. This is mainly due to many possible repair behaviours for each dynamic gate. This paper takes a pragmatic perspective and considers repair behaviours that have shown to be of long-standing industrial use in another, related, reliability formalism: Boolean logic-driven Markov processes (BDMPs). BDMPs are intensively used by the largest electrical energy producer and distributor in France to model and assess the reliability of repairable energy systems of different kinds. This paper takes the repair mechanisms of BDMPs as starting point and lifts them to repairable DFTs (rDFTs) by providing a set of BDMP-to-rDFT translation rules. The result is a repairable variant of DFTs in which repairs are interpreted consistently with BDMPs, in which repairs are a key asset. We empirically validate the correctness of this transformation by assessing the availability of a multiprocessor computing system and comparing the probabilistic model checking results of the obtained rDFTs against those for the original BDMPs.