Fault diagnosis of rolling bearing based on generalized S-transform and dropout CNN

Abstract

In order to solve the problems of feature extraction in the field of fault diagnosis, such as insufficient feature extraction and too complex classifier training, this paper takes rolling bearing, the key component of mechanical transmission device, as an example, and proposes to combine the feature extraction based on time-frequency analysis of generalized S-transform with dropout CNN to realize the fault detection of rolling bearing. In the diagnosis model, the time-frequency map of the original bearing data is obtained by the generalized S-transform, then the secondary feature is extracted by convolution neural network, and then the fault is classified by the classifier, so as to carry out the fault diagnosis of rolling bearing. The experimental results show that the accuracy of the diagnosis model can reach 99.6%, and the extracted features are highly differentiated. Compared with support vector machine (SVM) and convolutional neural network (CNN), this model has higher diagnostic accuracy and stability.