LMD支持向量机电机轴承故障诊断研究
2018-12-21尹召杰许同乐郑店坤
尹召杰 许同乐 郑店坤
摘要:针对支持向量机(SVM)对处理大样本数据和多分类问题以及核函数选择的局限性,提出LMD支持向量机电机轴承故障诊断方法。首先应用局域均值分解(LMD)算法对信号进行自适应分解,得到一系列PF分量,并利用相关分析剔除虚假分量,提取真实PF分量能量组成特征向量;其次应用新的核函数对SVM进行改进,实现自适应的训练,并针对大样本数据和多分类问题采用‘一对多的方法;最后以特征向量作为改進SVM的训练样本和测试样本,对电机轴承故障信息进行训练,预测。实验验证,该方法能有效的对电机轴承故障进行自适应的诊断。
关键词:
局域均值分解;支持向量机;故障诊断;电机轴承故障
DOI:1015938/jjhust201805007
中图分类号: TH165
文献标志码: A
文章编号: 1007-2683(2018)05-0035-05
Abstract:Aiming at the limitation of the support vector machine (SVM) to deal with the large sample data and the multi classification problem and the selection of kernel function, the fault diagnosis method of motor bearing based on LMD and support vector machine is proposed Firstly, the local mean decomposition (LMD) algorithm is used to adaptively decompose the signal to get a series of PF components , and the correlation analysis is used to eliminate false components Then, the energy feature vector is formed by extracting energy of the real PF component Secondly, the new kernel function is used to improve the SVM to complete the adaptive training, and the “one to many” method is used to solve the large sample data and multi classification problem Finally, the energy feature vector is used as the training sample and test sample of SVM, and the fault information of motor bearing is trained and predicted Experimental results show that the proposed method can effectively diagnose the fault of motor bearing
Keywords:local mean decomposition; support vector machine; fault diagnosis; bearing fault of motor
0引言
电机是生产中使用最频繁,也是最重要的工具。若其发生故障,将会影响与电机相关设备的运转及性能。在电机的故障诊断中,因轴承损坏而引起电机故障约占电机故障发生总数的30%[1-2]。因此对电机轴承故障诊断是非常有必要的。
在电机轴承故障诊断中,由传感器获得的轴承振动信号往往含有很强的噪声,为了保证诊断信息的质量,以及提高故障诊断的准确率,需要准确提取出包含主要故障信息的信号,并对实际测得的含噪信号进行故障信息提取。针对支持向量机在大样本数据、多分类问题及核函数选择上的局限性问题,提出了LMD支持向量机电机轴承故障诊断方法。该方法将LMD算……
