Research on Denoising Algorithm for Acoustic Fault Signals of Bearings Based on FxLMS and RTH-FMD

In the field of bearing fault diagnosis, acoustic-based methods have significant advantages over vibration-based methods in non-contact measurement. However, strong noise poses a challenge to the accuracy of diagnostic results. This article proposes a bearing acoustic fault signal denoising method that combines active denoising technology and signal decomposition principles. Firstly, a parabolic acoustic mirror and a free field microphone were used to collect acoustic signals of bearing faults and environmental noise signals, respectively. The environmental noise was eliminated based on the active denoising algorithm using a dual microphone (Filter-x LMS, FxLMS). Then, based on the RTH-FMD (Feature Mode Decomposition, FMD) algorithm, further extract signal components closely related to fault information from the perspective of signal structure. Using the red-tailed hawk algorithm (RTH) with minimum information entropy as the fitness function, the filter length L and the number of modes n in the FMD algorithm are optimized, and the optimal modal components are selected based on the maximum kurtosis principle. The experimental results show that compared with the traditional LMS algorithm, the FxLMS algorithm has a significant denoising effect, and the signal features after RTH-FMD processing are more prominent than those before processing.