Rolling Bearing Fault Diagnosis Based on CEEMDAN with CNN-Transformer

Bearings are one of the easily damaged parts of rotating machinery and play an important role in industrial production. Therefore, it is important to monitor their condition. However, due to the complexity of the operating environment of bearings, the signals collected during condition monitoring contain complex noise signals that drown out many important fault characteristics. How to accurately extract features for accurate fault identification has become an urgent problem. Therefore, in this paper, we design a Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) with Convolutional Neural Network and Transformer Neural Network model based on Attention Mechanism. rolling bearing fault diagnosis method. First, the vibration signals of rolling bearings are decomposed by CEEMDAN to obtain multiple Intrinsic Mode Functions (IMFs). Second, since not every one of the IMFs contains fault information, the information entropy value of the IMFs is evaluated, and the IMFs that contain rich fault information are selected to be retained and the feature vector matrix is constructed. Finally, the CNN-Transformer neural network model is used to input the feature vector matrix into the neural network model to realize the fault recognition of rolling bearings. The results show that the method proposed in this paper has faster convergence speed with higher accuracy and can realize accurate classification of rolling bearing faults.