An Intelligent Fault Diagnosis Method of Rolling Bearings Based on Short-Time Fourier Transform and Convolutional Neural Network

The rolling bearing is the key component of rotating machinery, and fault diagnosis for rolling bearings can ensure the safe operation of rotating machinery. Fault diagnosis technology based on deep learning has been largely studied for bearing fault diagnosis. However, for the deep learning model based on convolutional neural network, there are some intrinsic problems of producing inconspicuous features and useful feature information loss in the process of feature extraction of the raw fault vibration signals. In this work, an intelligent fault diagnosis method of rolling bearings based on short-time Fourier transform and convolutional neural network (STFT-CNN) is proposed. The one-dimensional vibration signals are converted into time–frequency images by STFT. Then, time–frequency images are inputted into STFT-CNN model for fault feature learning and fault identification. For the STFT of the vibration signals, the window type, window width and translation overlap width of the five typical window functions are studied and optimal one is obtained. And in the STFT-CNN model, the stacked double convolutional layers are adopted to improve the nonlinear expression capability of the model. To verify the effectiveness of the proposed method, experiments are carried out on the Case Western Reserve University (CWRU) and the Machine Failure Prevention Technology (MFPT) Society bearing datasets. The results show that the proposed method outperforms other comparative methods and reaches the identification accuracy of 100% and 99.96% for CWRU and MFPT, respectively.