变分模态分解和K-L散度在振动筛轴承故障诊断中的应用  

Application of Variational Modal Decomposition and K-L Divergence to Bearing Fault Diagnosis of Vibrating Screens

作　　者：徐元博[1] 蔡宗琰[1] XU Yuan-bo;CAI Zong-yan(Key Laboratory of Road Construction Technology and Equipment, Chang’an University,Xi’an 710064, China)

机构地区：[1]长安大学道路施工技术与装备教育部重点实验室,西安710064

出　　处：《噪声与振动控制》2017年第4期160-165,共6页Noise and Vibration Control

基　　金：中央高校教育教学改革专项经费建设项目资助(jgy16049;0012-310600161000);中央高校基本科研业务费专项资金资助(310825153313)

摘　　要：振动筛属于振动机械设备中的筛分设备,其结构特点与运行原理与一般的旋转机械有很大不同,因此提取出的振动信号同从旋转机械提取出的振动信号也同样有较大区别,主要体现在信号中不仅存在大量背景噪声,而且成分也较为复杂。对于此类信号,模态分解算法是个行之有效的方法,模态分解算法在去除大量高频噪声的同时,还能将振动信号分解成一系列具有单一成分的模态分量,从而能更好发现振动信号的物理意义。基于此,引入一种新的故障诊断方法,首先利用变分模态分解将故障信号分解为若干个窄带模态分量,然后根据K-L散度值选定最佳的分量,最后进行包络运算得出故障频率。通过仿真模拟实验与振动筛轴承故障诊断的实际应用,并与之前的经典模态分解算法——经验模态分解和集成经验模态分解进行对比,发现该算法更具有优越性和实用性。Vibrating screen is a kind of vibrating equipment in the field of vibrating machines,whose characteristics of structure and principle of operation are quite different from those of rotating machinery.Therefore,there is a large difference between the vibration signals extracted from the two different kind of machines.The vibration signals emanating from the vibrating screens contain a great deal of background noise as well as complex components.The mode decomposition algorithm is an effective method for this type of signals.The mode decomposition method can remove the background noise a great deal.Simultaneously,it can decompose the given signal into a series of mono-components to find physical meanings of the vibration signal.On this basis,a new non-recursive Variational Mode Decomposition(VMD)method is presented in this paper.This method can avoid the mode mixing and has a better robustness.This method firstly decomposes a fault signal into several different unknown modes,and then K-L divergence method is employed to select the sensitive modes.Eventually,the fault frequency of the selected modes is detected by envelope demodulation.The results of simulation and bearing fault experiments of the vibrating screen indicate that the proposed method can effectively extract fault features.In comparison with previous mode decomposition methods,such as Empirical Mode Decomposition(EMD)and Ensemble Empirical Mode Decomposition(EEMD),the feasibility and superiority of this method is verified.

关 键 词：振动与波 振动筛 轴承故障诊断 变分模态分解 经验模态分解 集成经验模态分解 

分 类 号：TH133.33[机械工程—机械制造及自动化] TG806[金属学及工艺—公差测量技术]