Predictive condition-based maintenance for continuously deteriorating systems

This paper addresses a predictive condition-based maintenance approach based on monitoring, modeling, and predicting a system's deterioration. The system's deterioration is considered as a stochastic dynamic process with continuous degrading. Structural time series, coupled with state-space modeling and Kalman filtering methods, is adopted for recursively modeling and forecasting the deterioration state at a future time. The probability of a failure is then predicted based on the forecasted deterioration state and a threshold of a failure. Finally, maintenance decisions are made according to the predicted failure probabilities, associated preventive and corrective maintenance cost, and the profit loss due to system performance deterioration. The approach can be applied on-line to provide economic and preventive maintenance solutions in order to maximize the profit of the ownership of a system.