Rolling Bearing Composite Fault Diagnosis Method Based on Convolutional Neural Network

Abstract

Rolling bearing feature extraction and fault identification techniques using deep learning algorithms have been widely adopted in recent years. We proposed a method for diagnosing composite faults in rolling bearings by employing multisensor decision fusion and convolutional neural networks. Different types of bearing faults and eccentricity faults have different fault eigenfrequencies in vibration signals. In the proposed method, vibration and acoustic signals are collected, their characteristics are analyzed, and multisensor data fusion processing is conducted. A neural network is then used to identify the signals containing bearing fault characteristics to diagnose bearing faults at different rotational speeds. We demonstrated the effectiveness of the proposed method by conducting comparative experiments on existing methods.