Weak Bearing Fault Diagnosis via Channel-Adaptive Convolution and Multisensor Feature Enhancement

Bearings serve as critical mechanical components in rotating machinery, making their intelligent fault diagnosis essential for operational reliability and maintenance efficiency. However, weak faults pose significant challenges in bearing fault diagnosis. To solve this challenge, a multi-sensor feature enhancement network for intelligent bearing fault diagnosis is proposed in this paper. Firstly, a channel adaptive convolution strategy is proposed, which automatically selects the optimal kernel size for each channel, thereby enhancing the expression of advantageous characteristics. Secondly, a multi-sensor feature complementary learning strategy is designed, which enhances the robustness and stability of the model by leveraging redundant information from multiple sensors. Thirdly, a multi-sensor feature enhancement strategy is proposed, which employs an attention mechanism to merge and amplify subtle features from individual sensors, thereby improving the representation of fault characteristics. To verify the effectiveness and reliability of the proposed method, validation experiments with a series of state-of-the-art methods are conducted on two datasets. The experimental results demonstrate that the proposed method achieves high accuracy in diagnosing bearing faults, even in cases of weak fault conditions.