A Fault Diagnosis Method of Rolling Bearing of CNC Machine Tool Based on Improved Convolutional Neural Network

Abstract

In the industrial production process, the rolling bearing failures of huge mechanical equipment such as CNC machine tools frequently occur, which seriously affects the production performance and service life of the machine tools. In order to identify the types of faults in rolling bearings and improve the safety of the equipment, this paper presents a fault diagnosis method on account of an improved Convolution Neural Network (CNN). The improved CNN model is to add a convolutional layer before the fully connected layer, after several convolutional layers and several pooling layers, and use an improved stochastic gradient descent training algorithm with momentum to speed up the training speed to enhance the serviceability of the model. Traditional fault diagnosis methods are time-consuming, high in labor costs and low in work efficiency. The method in this paper improves the intelligence of the rolling bearing of CNC machine tools fault diagnosis process, improves the correctness of fault diagnosis, and adapts to the characteristics of big data fault diagnosis. Finally, the data set of Case Western Reserve University's rolling bearing database is used for experimental verification. The experimental results reveal that this method has a high recognition accuracy rate for various types and severity of rolling bearing faults, and has good practicability and application prospect.