RUL prediction for rolling bearings based on Convolutional Autoencoder and status degradation model

Abstract

The remaining useful life (RUL) prediction of rolling bearings plays a key role in improving the safety and reliability assessment for rotating machinery. To accurately describe the degradation degree of bearings and perform RUL prediction, an RUL prediction method of rolling bearing combining Convolutional Autoencoder (CAE) networks and status degradation model is proposed. Firstly, the CAE is used to extract the features from the degraded bearing data; then the status degradation model is built, and the multi-dimensional health status mapping function is used to downscale the extracted features, and the reduced data points are fused with the Euclidean distance to establish the health status index that can characterize the degraded bearing. Finally, the status degradation function in the constructed model and the online update and prediction algorithm are used to adaptively estimate the RUL. The proposed method is validated with PHM datasets for RUL prediction, and its prediction performance is compared with eight prediction methods. The experimental results show that the proposed approach effectively predicts the RUL of rolling bearings and accurately evaluates the degradation degree of the bearing in a future stage.