A combined risk-based and condition monitoring approach: developing a dynamic model for the case of marine engine lubrication

Abstract

Abstract This paper focuses on the creation of a dynamic probabilistic model which simulates deterioration trends of a marine engine lubrication system. The approach is based on risk and the implementation is achieved through a dynamic Bayesian network (dBN). Risk can be useful for decision making, while dBNs are a powerful tool for risk modelling and prediction models. The model takes into account deterioration of engine components, oil degradation and the off-line condition monitoring technique of oil analysis, in the context of predictive maintenance. The paper aims to efficiently predict probability evolution for main engine lubrication failure and to decide upon the most beneficial schemes from a variety of lubrication oil analysis interval schemes by introducing monetary costs and producing the risk model. Real data and respective analysis, along with expert elicitation, are utilized for achieving model quantification, while the model is materialized through a code in the Matlab environment. Results from the probabilistic model show a realistic simulation for the system and indicate the obvious, that with more frequent oil analyses and respective maintenance or repairs, the probability of failure drops significantly. However, the results from the risk model highlight that the costs can redefine scheme suggestions, as they can correspond to low probabilities of failure but also to higher costs. A two-month interval scheme is suggested, in contrast to the most preferred practice among shipping companies of a three-month interval. The developed model is in general identified as a failure prediction tool focusing on marine engine lubrication failure.