High Level Representations for Markov Analysis of Complex Dynamic Systems

M. Bouissou and S. Muffat (France)

Keywords

Dependability, Markov process, BDMP, modeling, fault tree.

Abstract

We present a very innovative approach, fully supported by software developed at EDF, which allows to calculate the reliability and the availability of complex repairable systems, having reconfiguration capacities. The performances of this approach are well beyond what is classically possible, thanks to the automatic transformation of a high level representation (like a P&I diagram) of a system into a Markov model optimized so as to reduce considerably the combinatorial (number of states to be considered) of the problem. The markovian model is defined "locally", by the possibility of determining all the states that can be reached from a given state, with the corresponding transition rates. Such a definition lends itself to original calculation methods, based on the investigation of sequences in the state graph. This type of calculation has the advantage not to require the more or less exhaustive construction of the graph (with the resulting problems due to memory limitation) and to give very useful qualitative information to validate the models and interpret the results.

Important Links:



Go Back