An integrated hybrid methodology of time series forecast and case-based reasoning for fault prognosis

This paper presents a methodology for system prognosis based on indicative parameter time series of the equipment condition. The time series is divided in different candidate scenarios according to modifications on exogenous variables that represent external environmental conditions. Each valid scenario is associated with a specific progression model built based on ARIMA time series analysis approach. The forecast model is determined by merging the current scenario progression model with the progression model associated with most similar past scenario. The feasibility and effectiveness of the approach proposed is demonstrated through the prediction of the degradation characteristics provided by DC machine benchmark fault simulator.