Gear Early Fault Diagnosis Based on Multi-resolution Subband Adaptive Filtering

The early fault signals of gears in the gearbox are weak, and the fault signals in different frequency bands are easily masked by other noise signals. Currently, the extraction of early failure signals of gears generally involves filtering and decomposition of signals in the full frequency band or a single frequency band, which has certain limitations. Therefore, a multi-resolution subband adaptive filtering method (MSAF) is proposed in this paper. This algorithm decomposes the signal through an analysis filter bank into different frequency domains and applies adaptive filtering with different-length weighting vectors. Then, a synthesis filter bank is used to reconstruct the signal and obtain the full-band signal, thereby achieving rapid extraction of gear or bearing fault signals. Experimental results demonstrate that this algorithm has fast convergence speed, low deviation, and more pronounced denoising effects compared to other filtering methods.