Aggregating Probabilistic Models of Complex Multi-State Systems

This paper proposes an aggregating method based on a bayesian approach to improve the continuity of service for complex multi-state systems into an industrial context. In the study of our “equivalent” model of a given machines configuration, each machine is considered as one bayesian multi-state sub-model. A simulation is made with or no learning on the node of maintenance policy for each configuration of machines in order to study, master its process of degradation and suggest the best actions of maintenance policy over an given exploitation phase with regard to the chosen performance indicators.