Abstract
This paper proposes an optimized DenseNet-Transformer model based on FFT-VMD processing for bearing fault diagnosis. First, the original bearing vibration signal is decomposed into frequency-domain and time-frequency-domain components using FFT and VMD methods, extracting key signal features. To enhance the model's feature extraction capability, the CBAM (Convolutional Block Attention Module) is integrated into the Dense Block, dynamically adjusting channel and spatial attention to focus on crucial features. The alternating stacking strategy of channel and spatial attention further improves the feature extraction ability at different scales. This optimized structure increases the diversity and discriminative power of feature representations, enhancing the model's performance in fault diagnosis tasks. Furthermore, the Transformer module, replacing the LSTM, is employed to model long-term and short-term dependencies in the time series. Through its Self-Attention mechanism, Transformer efficiently captures the global relationships within the sequence, improving the model's ability to handle complex temporal dependencies. Experimental results show that the proposed model achieves outstanding performance in bearing fault classification, with a classification accuracy of 99.68%, demonstrating excellent generalization ability. This model provides an effective and reliable solution for the health monitoring of chemical equipment.