利用相关系数图选择振动信号特征值的方法

Method of vibration signal eigenvalue selection by correlation coefficient diagram

振动信号常被用来监测机械设备工作状态,其特征值选择会直接影响监测效果。以振动信号识别发动机故障为工程背景,为了快速有效地提高识别率,提出构建相关系数图并利用其选择振动信号特征值的方法。首先,对发动机振动信号提取时域特征参数,计算各特征值之间的皮尔逊相关系数(PCC)和最大互信息系数(MIC);然后,选择不同阈值构建相关系数图,筛选特征值;最后,将特征值作为广义回归神经网络(GRNN)的输入,分析比对故障识别效果。实验结果表明,利用阈值为0.9的MIC图筛选特征值可以在仅提取少量时域特征值的前提下获得较高的振动信号对发动机故障的识别率。

Vibration signals are often used to monitor the working state of mechanical equipment,so the selection of eigenvalues will directly affect the monitoring effect.A method of selecting the eigenvalues of vibration signals by correlation coefficient diagrams is proposed to identify engine faults by vibration signals and improve the identification rate quickly and effectively.The time-domain characteristic parameters of engine vibration signals are extracted,and the Pearson correlation coefficient(PCC)and maximal information coefficient(MIC)between the eigenvalues are calculated respectively.Different threshold values are selected to construct correlation coefficient diagrams,and the eigenvalues are screened.The eigenvalues are used as the input of the generalized regression neural network(GRNN)to analyze and compare the effects of engine fault identification.The experiment results show that,on the premise of extracting only a small amount of time-domain eigenvalues,higher identification rates of engine faults can be obtained by vibration signals when the MIC diagram with the threshold value of0.9 is used to screen the eigenvalues.