A Novel Ensemble Deep Ridgelet Auto-Encoders for Intelligent Fault Diagnosis of Bearing

The traditional fault diagnosis methods of rolling bearings have such shortcomings as largely dependent on expert prior knowledge, difficulty in fault feature extraction. How to automatically and accurately identify bearings faults has attracted more and more attention. For this purpose, a method based on ensemble deep ridgelet auto-encoders, i.e. EDRAE, is proposed. Firstly, different ridgelet auto-encoders with different ridgelet functions are designed. Secondly, different DRAEs are constructed with different ridgelet auto-encoders to realize unsupervised feature learning of raw data. Finally, the top features of each DRAE are fused as the input of the Softmax to realize accurate fault identification. The experimental results show that the proposed method can get rid of the dependence on artificial feature extraction and overcome the limitations of individual deep learning model, which is more effective than other intelligent diagnosis methods.