Diagnosis and Prediction of Bearing Fault Using EEMD and CNN

Rolling bearing is a very essential component of the industrial machinery. The bearing fault could cause a significant loss. Therefore, it is necessary to perform fault diagnosis and prediction on the bearing. This paper combines Ensemble Empirical Mode Decomposition (EEMD), Singular Value Decomposition (SVD) difference spectrum de-noising, and the convolutional neural network (CNN) to realize the diagnosis and prediction of bearing faults. EEMD is used to extract features, and SVD difference spectrum de-noising is used to denoise the decomposed signals. The reconstructed vibration signals are then fed into CNN to realize fault diagnosis. Further, by analyzing the output of the softmax layer after the input of testing sets, the prediction of bearing fault can be realized. The bearing vibration signals are used to perform diagnosis. And we use partial samples of bearing fault full-period data to retrain CNN for prediction. In this paper, these methods successfully lead to bearing fault diagnosis with high accuracy and early bearing fault prediction.