基于EMD-HT-SVM的磨床振动故障监测方法研究  

Research on vibration fault monitoring method of grinder based on EMD-HT-SVM

作　　者：吴兵[1] Wu Bing(School of Aeronautical Engineering,Shaanxi Polytechnic Institute,Shaanxi Xianyang,712000,China)

机构地区：[1]陕西工业职业技术学院航空工程学院,陕西咸阳712000

出　　处：《机械设计与制造工程》2020年第9期107-111,共5页Machine Design and Manufacturing Engineering

基　　金：陕西省高等教育教改项目(15Z07)。

摘　　要：针对磨床工件加工产生的剧烈颤振及噪声会导致磨床其他零部件出现故障损坏等问题,提出了一种基于经验模态分解(EMD)、Hilbert变换(HT)以及支持向量机(SVM)的磨床振动故障监测方法。首先,利用传感器采集磨床振动信号,对信号进行降噪预处理;然后,将处理后的信号进行经验模态分解,并计算出有效的固有模态分量函数(IMF);再利用Hilbert变换计算出分解信号的能量分布和实时方差,并用信号的主频率带组成特征向量;最后,采用支持向量机算法进行样本分类识别训练并与BP神经网络识别方法进行对比。试验结果表明,该故障监测方法对磨床振动故障监测信号具有很好的判别效果。In order to solve that the violent flutter and noise caused by the machining of grinder workpiece can lead to fault damage of other parts of grinder,a research on vibration fault monitoring method based on empirical modal decomposition(EMD),Hilbert transform(HT)and support vector machine(SVM)is proposed.Firstly,it uses sensor to collect the vibration signal of the grinder to preprocess the signal noise reduction.Secondly,it decomposes the processed signal into empirical modes and calculates the effective intrinsic modal component function(IMF).Then,it calculates energy distribution and real-time variance of the decomposition signal by using the Hilbert transform,and forms the eigenvectors by using the main frequency band of the signal.Finally,it uses support vector machine(SVM)method for sample classification and recognition training and compared with BP neural network recognition method.The test results show that the fault monitoring method has a good judgement effect on the vibration fault monitoring signal of grinder.

关 键 词：磨床振动 经验模态分解 HILBERT变换 支持向量机 故障监测 

分 类 号：TH16[机械工程—机械制造及自动化]