基于小波包和独立分量分析的微弱多源故障声发射信号分离

发布时间：2018-06-29 11:35

  本文选题：多源分离 + 小波包分析　； 参考：《上海交通大学学报》2016年05期

【摘要】：针对旋转机械设备中同时存在的裂纹、摩擦等多故障源信号难以检测和分离的问题,提出了一种基于小波包分析(WPA)与独立分量分析(ICA)的多源故障信号提取方法,即首先用WPA对含噪线性混合信号降噪预处理,由db2小波基函数进行5层分解后保留62.5~187.5kHz频段信号,然后采用ICA中的FastICA算法对降噪后的混合信号分离,最后对各通道分离出的信号用收缩函数进行频段内去噪处理.对不同输入信噪比的含噪微弱裂纹和摩擦信号进行提取和分析的结果表明,该方法能有效提取出输入信噪比大于-15dB的裂纹和摩擦信号.当混合信号信噪比为-15dB时,裂纹和摩擦信号的输出信噪比分别为-1.31和-1.36dB,相关系数分别为0.62和0.63,提取效果好于结合小波包和FastICA分离方法(信噪比分别为-1.74和-2.06dB,相关系数分别为0.59和0.59)以及单独采用FastICA算法(信噪比分别为-4.57和-4.31dB,相关系数分别为0.17和0.19).因此,所提出的综合WPA和ICA的方法是一种较好的多源微弱信号提取方法.
[Abstract]:In order to solve the problem that it is difficult to detect and separate the signals of multi-fault sources such as cracks and friction in rotating machinery, a multi-source fault signal extraction method based on wavelet packet analysis (WPA) and independent component analysis (ICA) is proposed. Firstly, the noise reduction of the noisy linear mixed signal is pretreated, the db2 wavelet basis function is decomposed into five layers and the 62.5 ~ 187.5 kHz band signal is retained, then the noise reduction mixed signal is separated by the fast ICA algorithm. Finally, the signal separated from each channel is de-noised by shrinkage function in frequency band. The results of extraction and analysis of weak cracks and frictional signals with different input signal-to-noise ratio show that the proposed method can effectively extract the crack and friction signals with input signal-to-noise ratio greater than -15dB. When the signal-to-noise ratio of mixed signal is -15dB, The output SNR of crack and friction signals were -1.31 and -1.36 dB, respectively, and the correlation coefficients were 0.62 and 0.63, respectively. The extraction effect was better than that of wavelet packet and FastICA separation method (SNR = -1.74 and -2.06 dB, the correlation coefficients were 0.59 and 0.59, respectively) and the single method was better than the method of combining wavelet packet and FastICA (SNR = -1.74 and -2.06 dB, respectively). The FastICA algorithm is used (signal-to-noise ratio (SNR) is -4.57 and -4.31 dB, the correlation coefficient is 0.17 and 0.19 respectively). Therefore, the proposed method of synthesizing WPA and ICA is a better multi-source weak signal extraction method.
【作者单位】： 长沙理工大学工程车辆安全性设计与可靠性技术湖南省重点实验室;长沙理工大学道路灾变防治及交通安全教育部工程研究中心;广西大学机械工程学院;
【基金】：国家自然科学基金(51105045,51205031,51365006) 湖南省教育厅重点项目(15A008) 长沙理工大学工程车辆安全性设计与可靠性技术湖南省重点实验室基金(KF1507);长沙理工大学道路灾变防治及交通安全教育部工程研究中心项目(kfj140407)资助
【分类号】：TH17
，

本文编号：2081983