Deep Learning based End-to-End Rolling Bearing Fault Diagnosis

Abstract

Rolling bearings play an important part in rotating machinery. As they work in complex conditions, faults will occur sometimes. Therefore, it is necessary to detect the faults early. Traditional bearing fault diagnosis methods are often based on mechanism analysis and feature selection, and the process is relatively complicated. Deep learning methods, however, have the ability to extract and select features automatically, which greatly reduces the workload. In recent years, deep learning-based methods have been successfully used in many fields, such as computer vision, voice recognition, medical diagnosis. In this paper, the end-to-end fault methods based on deep learning are proposed. The Long Short-Term Memory (LSTM) network, Gated Recurrent Unit (GRU) network and One-Dimensional Convolutional Neural Network (1D CNN) are used to build the deep learning network architecture respectively. A methodology is proposed for rolling bearing fault diagnosis, including data preprocessing, network modeling, training, validation and testing. Test bench data is used for fault diagnosis and the results show that deep learning based end-to-end methods are effective for the fault diagnosis of rolling bearings and that the model based on 1D CNN has the best performance.