Fault Diagnosis for Rolling Bearing Based on Deep Residual Neural Network

As the machine tool becomes more and more complex, the shallow model represented by machine learning and SVM is difficult to characterize the complex mapping relationship between the measured signal and the health status of the equipment, and it faced with the problem of dimensionality disaster. In view of the complex feature extraction process and the uncertainty of the traditional intelligent recognition, a method of fault feature extraction and recognition based on deep residual neural network is proposed in this paper. This method uses the original time domain signal to train the deep residual neural network and complete the intelligent classification of the fault type without periodic request to the time domain signal. Accordingly, it has strong applicability to effectively extract and identify features from multiple conditions, multiple fault locations and various fault levels. Compared with traditional fault diagnosis models, the deep residual neural network model improves the fault recognition rate and shows strong generalization performance.