A Novel Time-Varying Structural Element for Morphological Filtering-Based Bearing Fault Diagnosis

To enhance the robustness of morphological filtering-based bearing fault detection methods under complex noise interference, a performance-enhanced multi-order weighted time-varying morphological filtering is proposed. Firstly, to address the issue of traditional structural elements (SEs) being unable to accurately extract fault-related components from noisy vibration signals, a novel SE design strategy named multi-order time-varying SE is introduced, combining the advantages of multi-scale SEs and time-varying SEs, which can more accurately match and extract periodic transient features hidden in noisy signals. Subsequently, an information threshold is introduced into the filtered signals under time-varying SEs with different orders to construct a weighted function to enhance fault-related information and eliminate interference components. Finally, autocorrelation is applied to the weighted signal to further highlight fault-related features. The experimental results demonstrate that the proposed method can effectively extract bearing fault-related features and diagnose railway wheelset bearing faults, and its superiority is validated through comparison with existing methods.