Research on noise reduction method of vibration signal based on CEEMDAN-PE-Wavelet Packet

The operational health status of the gearbox determines the performance efficiency and precision of industrial robots, significantly impacting long-term stability. Therefore, it is highly significant to analyze the impact of various component vibration factors on the dynamic response generated by the gearbox transmission. This paper focuses on the non-stationary vibration signals of RV reducers. Based on empirical mode decomposition, it proposes a denoising technique based on CEEMDAN-PE-Wavelet Packet vibration signals. This technique aims to address the noise issue of RV reducer vibration signals that interfere with subsequent studies. Through comprehensive comparisons of various denoising algorithms via simulation experiments, this paper presents a vibration signal denoising method with a Signal-to-Noise Ratio (SNR) of 9.0037 and a Root Mean Square Error (RMSE) of 0.38, both of which outperform other denoising techniques. The analysis demonstrates that this method can enhance both the SNR and computational accuracy when applied to the vibration signals of RV reducers. The feasibility of the technique applied to RV reducer vibration signals is proved by building a vibration test bench, collecting vibration acceleration signals, and de-manipulating them using the noise reduction method proposed in the paper.