A CM&CP framework with a GIACC method and an ensemble model for remaining useful life prediction

Abstract

Data-driven methods based on health indicators (HIs) have been proven effective for remaining useful life (RUL) prediction. Significant differences may exist among HI degradation sequences due to various equipment failure behaviors. In order to solve this problem, the first clustering then modeling and first classifying then predicting (CM&CP) framework for RUL prediction is often used. Differently, for the processes of clustering and classifying in the framework, the grey incidence analysis (GIA) method is first proposed to measure the similarity level of two sequences; and the GIA-based clustering and classifying (GIACC) method is suggested. The ensemble model with multiple appropriate machine learning models as its base learners for RUL prediction is trained for each cluster, aiming to achieve good prediction accuracy and high computational efficiency. Since aero engines have various failure modes, the related datasets are used to verify the effectiveness of the proposed method. The case study shows that the CM&CP framework with the GIACC method and the ensemble model can achieve good performance in both prediction accuracy and computational time and provide stable support for equipment health management, especially under multiple failure modes.