Machine learning approach for integrated maintenance and spare parts management strategies

Abstract

We targeted a data mining and machine learning approach for integrated maintenance/production and spare parts management problems for components of a wind farm where the level of degradation is noticeable. The degradation is established as a function of the actual functioning mode. Spare parts are stored in a local inventory. The costs associated with the supply of spare parts and their renewal are related to the functioning mode. Our goal is to use a data mining and a machine learning approach in order to optimize the total current cost of maintenance, production and spare parts related costs over a fixed planning horizon. We formulate the problem of the optimal policy structure, which turns out to be a three-threshold policy in all operating modes. Our numerical results show that the cost reductions achieved by the integrated maintenance, production and spare parts optimization are significant.