Abstract: Tough working environment, expensive maintenance cost and production losses have put forward significant challenge to the maintenance of wind turbine and its components. The key components in the wind turbine such as gearbox and blade are selected as the object of study, the degradation process is divided into finite degradation states. With the objective of minimizing long-term discount cost, the critical factors including wind speed, spare parts logistics and production losses are taken into account, and a condition-based maintenance optimization model is established which is based on the semi-Markov decision process. The maintenance policy, the inspection interval under each state, and the transition probabilities between different states are analyzed, and the model is solved with the method of policy iteration. The proposed approach is applied to a wind turbine gearbox, by way of analyzing the inspection interval and maintenance cost under periodical inspection and non-periodical inspection, the optimal maintenance policies are obtained. The result shows the presented model can describe the degradation process effectively, and can be used to optimize the maintenance policies.

Key words: Condition-based maintenance, Policy iteration, Semi-Markov decision process, Wind turbine, Envelope scalogram , Incipient fault, Instantaneous envelope scalogram entropy , Optimal fault characterization phase, Rolling bearings