A Novel Remaining Useful Life Prediction Method of Rolling Bearings Based on Deep Transfer Auto-Encoder

Deep learning (DL) has shown effectiveness in the field of Prognostic and Health Management (PHM) of rolling bearings and successfully solved many problems with its powerful feature learning and function fitting capabilities. However, it is still very challenging to transfer a remaining useful life (RUL) prediction model trained by historical data to hearings running under new operating conditions. Domain adaptation (DA) is an effective solution to obtain crass-domain features of bearings. This article presents an RUL prediction method for rolling bearings based on a deep transfer auto-encoder. Two strategies, parameter transfer and feature representation transfer, are used to transfer the feature extraction model and RUL regression model trained in the labeled source domain to the unlabeled target domain. In this way, the full utilization of the historical model and the rapid adaptation to the new operating conditions are realized. A case study on the IEEE PHM Challenge 2012 bearing data set verifies the effectiveness of the proposed method. The comparison with other nontransfer and transfer machine learning models shows the advantages of this method in precision and accuracy.