An optimal sequential preventive maintenance policy under stochastic maintenance quality

This paper presents a sequential imperfect preventive maintenance policy for a degradation system. Two kinds of activity, called continuous preventive maintenance (PM) and minimal repair, are simultaneously considered when arranging discrete imperfect preventive maintenance schedules. In order to obtain the maximum benefit in a finite lifetime, an expected benefit model is formulated based on maximal/equal cumulative-hazard rate constraints, and the optimal PM intervals are obtained using a genetic algorithm (GA). It is usually difficult to determine fixed maintenance quality after performing maintenance activities. This problem is addressed in the present paper by assuming that the reduction factor is a stochastic variable following probability distributions at fixed times. It is more rational to describe the fluctuation and trend of quality of discrete preventive maintenance during a lifetime; this makes optimisation results more robust and insensitive to the randomness of the crucial parameters in imperfect PM models. A numerical case is presented to illustrate the proposed model and some discussions are summarised.