Opportunistic predictive maintenance for complex multi-component systems based on DBN-HAZOP model

Abstract Predictive maintenance (PdM) focuses on failure prediction in order to prevent failure in advance and offer sufficient information to improve inherent safety and maintenance planning. A novel opportunistic predictive maintenance-decision (OPM) method integrating of machinery prognostic and opportunistic maintenance model is proposed in this paper to indicate the optimal maintenance time with minimal cost and safety constrains. DBN-HAZOP model quantifies hazard and operability analysis by dynamic Bayesian networks to provide prospective degradation trends of each component and the overall system for maintenance decision making. It is developed by integrating the prior knowledge of the interactions and dependencies among components and also the external environment, while the online condition monitoring data which is further to update the parameters of the model. Based on the future degradation trends given by DBN-HAZOP model, a local optimal proactive maintenance practice can be determined for each component by minimizing the expected maintenance cost per time unit. Understanding that for a complex system, whenever one of the components stops to perform a predictive maintenance action, the whole complex system must be stopped, at this moment, PdM opportunities arise for the other degraded components in the system at a reduced additional cost. Therefore, this paper further proposes an opportunistic PdM strategy for global cost optimization of predictive maintenance for the whole system, which considers failure probabilities, repair costs, down time cost and set-up cost. Case studies are given throughout to show how this approach works, and the sensitivity of the results to some of the driving cost parameters has also been examined.