A dynamic approach for maintenance evaluation and optimization of multistate system

Abstract

This work presents a quantitative approach on the basis of Dynamic Bayesian Network to model and evaluate the maintenance of multi-state degraded systems and their functional dependencies. The reliability and the availability of system are evaluated taking into account the impact of maintenance repair strategies (perfect repair, imperfect repair and under condition-based maintenance (CBM)). According to transition relationships between the states modeled by the Markov process, a DBN model is established. Using the proposed approach, a DBN model for a separator Z1s system of Sour El-Ghozlane cement plant in Algeria is built and their performances are evaluated. Through the result of diagnostic, for improving the performances of separator, the components E, R and F should given more attention and the results of prediction evaluation show that in comparing with perfect repair strategy, the imperfect repair strategy cannot degrade the performances of separator, whereas the CBM strategy can improve the performances considerably. These results show the utility of this approach and its use in the context of a predictive evaluation process, which allows to offer the opportunity to evaluate the impact of the decisions made on the future performances measurement. In addition, the maintenance managers can optimize and improve maintenance decisions continuously.