Rolling bearing fault diagnosis based on EMD-CNN-Transformer

With the rapid development of industrial equipment toward high-speed and heavy-load operations, real-time monitoring and accurate diagnosis of bearing conditions have become critical for ensuring operational safety and reliability. However, the nonlinear and non-stationary characteristics of bearing vibration signals, coupled with the limitations of traditional methods in adaptability and global feature extraction, hinder diagnostic accuracy and efficiency. To address these challenges, this paper proposes a hybrid model integrating Empirical Mode Decomposition technique (EMD), Convolutional Neural Network (CNN), and Transformer. The model first applies EMD to adaptively decompose raw vibration signals into multi-scale Intrinsic Mode Function (IMF), effectively isolating high-frequency details and low-frequency trends. Subsequently, CNN extracts local fault features from each IMF component to enhance the characterization of critical information. The Transformer module is then introduced to dynamically fuse global features, overcoming the limitation of traditional CNN models constrained by local receptive fields. Experiments on the CWRU bearing dataset demonstrate that the proposed model outperforms existing methods such as standalone Transformer, CNN-LSTM, and CNN-Transformer in diagnosing 10 fault states, achieving superior accuracy, recall, precision, and F1-score. These results validate the effectiveness and robustness of the model in improving bearing fault diagnosis.