Supervisory evolutionary optimization strategy for adaptive maintenance schedules

Abstract

A supervisory strategy is proposed for improving the performance of an evolutionary-algorithm-based system-maintenance optimizer developed in our previous work for offshore power systems. The system-maintenance optimizer generates a set of initial maintenance plans, and exports them to an intelligent maintenance advisor connected to it for implementation. The proposed supervisory strategy uses a set of intelligent rules for adjusting the crossover and mutation rates of the present evolutionary algorithm. A mechanism is developed for refining and generalizing the supervisory rules according to the user's experience. The proposed supervisory strategy aims to improve the search ability and efficiency of the present evolutionary algorithm. Merits of the proposed supervisory strategy are demonstrated in case studies using our system-maintenance optimizer. © 2011 IEEE.