Rolling Bearing Fault Diagnosis with Time-Frequency Image Based on Deep Learning

To accurately describe the trend of rolling bearing fault diagnosis as well as to improve the accuracy and efficiency of fault type discrimination. In this paper, a rolling bearing test dataset under 10 different operating conditions is established, and the AlexNet model multi-feature fusion technique as well as Dropout algorithm and ReLU activation function are used to enhance the performance of the classifier. The test data were obtained from the Case Western Reserve University (CWRU) bearing failure dataset. Firstly, the original data set is divided into training and validation sets in the ratio of 3:7; secondly, the training and validation sets are input into the AlexNet model for training to fully exploit the feature data of time-domain images in faulty bearings; finally, the weight file generated from the feature data realizes the classification analysis of rolling bearing fault types. The test results show that the training accuracy reaches 99% and the fault diagnosis type prediction accuracy reaches 98%. It is proved that the method can effectively provide efficient classification efficiency for the time domain signal images of bearing fault diagnosis and provide a research basis for the rolling bearing fault diagnosis classification method.