基于小波分析与SVD的机械故障诊断

Mechanical Fault Diagnosis Using Wavelet Analysis and SVD

对小波变换后相邻两频带内信号存在频率混叠和因采样点数过多导致奇异值分解过程中无法计算或计算时间较长等问题,提出了小波-滑移时间序列奇异值分解方法。通过小波变换,将原始采样信号分解到不同频内。利用滑移时间序列奇异值分解法进一步处理所需频段。通过将所需频段内信号分成若干等长度的子时间序列,依次对其进行奇异值分解,根据奇异值提取需要的特征信息,然后将其拼接成等采样点数的时域信号。利用该方法,对滚动轴承内圈故障的运行声音信号做了处理。从而,验证了方法的可行性。

To solve frequency mixing problem of the adjacent two frequency band signal after wavelet transform and unable to calculating or calculate time too long because of the excessive number of sampling points in singular value decomposition process,wavelet transform-sliding time series of singular value decomposition method was proposed. The original sample signal is decomposed into different frequency by wavelet transform,processing the required frequency band signal by sliding time series of singular value. Then,the signal is divided into a number of equal length sub time series by using the signal in the desired frequency band,extracting feature information according to the singular values. Operation the sound signal of inner ring fault of rolling bearing by this method and proved to be feasible.