奇异值分解对轴承振动信号中调幅特征信息的提取

Amplitude Modulation Feature Extraction of Bearing Vibration Signal Using Singular Value Decomposition

为了从复杂的轴承振动信号中提取微弱的故障信息,提出了一种基于奇异值分解的特征提取方法.分析了基于奇异值分解的信号分解和特征提取原理,指出其信号分解的实质是一种线性叠加分解,并通过对轴承振动信号构造Hankel矩阵,利用奇异值分解处理后得到多个分量信号,并选择前面一定数目的分量信号进行叠加,准确地提取到了因滚道损伤引起的调幅特征,进而研究分析了不同数目分量所获得的调幅特征效果,并与小波变换进行比较.研究结果表明SVD对调幅特征的提取效果优于小波变换.

In order to extract the faint fault information from complicated vibration signal of bearing, a feature extraction method based on singular value decomposition （SVD） is proposed. The principle of SVD based signal decomposition and feature extraction is analyzed, and it is indicated that the essence of signal decomposition of SVD is linear addition decomposition. On this basis, a Hankel matrix is constructed for bearing vibration signal, and by SVD processing a group of component signals are obtained, then some of front component signals are selected to be added together, and the amplitude modulation feature caused by damaged roll track is clearly extracted. In addition, the amplitude modulation feature extracted by different numbers of component signals is also investigated. The comparative result shows that the amplitude modulation extraction effect of SVD is much better than that of wavelet transform.