Degradation state model-based prognosis for proactively maintaining product performance

Maintenance plays now a critical role in manufacturing for achieving important cost-savings and competitive advantage while preserving product conditions. It suggests moving from conventional maintenance practices to predictive strategy. Indeed the maintenance action has to be done at the right time according the component Remaining Useful Live (RUL) assessed by a prognosis process. The accuracy of the RUL is mainly depending on the relevance of the component degradation model used for prediction. In that way, this paper aims at discussing an efficient degradation model taking into account the operational conditions, the health monitoring and the maintenance actions. This model is based on discrete states associated with the degradation levels, and on a cumulative function modelling the transition time between successive states. The model is implemented by means of Stochastic Activity Networks (SAN). The feasibility and added value of such degradation models for prognosis is then highlighted through experimentations made on manufacturing TELMA platform.