K-SVD小波降噪在齿轮故障诊断中应用

摘要

For the large noise disturbance in the early stage gear fault diagnosis, it is difficult to extract the fault features. In this paper, a new method of wavelet denoising method based on K-SVD sparse representation is proposed. This method can overcome the disadvantage of the conventional threshold method that it only deals with the wavelet coefficients one by one but ignores the whole structure of the coefficients. In this method, the wavelet coefficient structure characteristic is sufficiently considered. It has good robustness even in strong noise background. Through the analysis of simulated signals and measured signals of an aero-engine gear hub, the correctness and validity in engineering application of the proposed method are verified.%针对早期齿轮故障诊断中噪声干扰大，故障特征难以提取的问题提出基于K-SVD稀疏表示小波降噪算法。该算法克服传统小波阈值降噪算法只对小波系数进行逐点处理，而忽略小波系数整体架构的缺点，充分考虑小波系数结构特点，在强噪声下仍具有很好稳健性。通过对模拟信号和实测发动机减速器齿轮毂信号分析，证明小波降噪算法正确性和在实际工程应用中的价值。