基于VMD-MDE的柱塞泵磨损故障诊断研究  

Fault diagnosis of plunger pump wear based on VMD-MDE method

作　　者：曲全鹏 曲海军[2] 张强 QU Quan-peng;QU Hai-jun;ZHANG Qiang(Engineering Training Center,Henan Institute of Engineering,Zhengzhou 451191,China;School of Mechanical and Power Engineering,Henan Polytechnic University,Jiaozuo 454003,China;Henan Ruizhi Hydraulic Equipment Co.,Ltd.,Zhengzhou 451191,China)

机构地区：[1]河南工程学院工程训练中心,河南郑州451191 [2]河南理工大学机械与动力工程学院,河南焦作454003 [3]河南睿智液压设备有限公司,河南郑州451191

出　　处：《机电工程》2021年第9期1202-1206,共5页Journal of Mechanical & Electrical Engineering

基　　金：河南省重点研发与推广科技攻关项目(192102210224);河南工程学院工程基础训练实验教学示范中心资助项目(508905)。

摘　　要：通过变分模态分解特征能量重构法(VMD)来实现对故障进行分析时,存在准确性不高的问题,针对这一问题,提出了一种通过变分模态分解特征能量重构法(VMD)和多尺度散布熵实现的柱塞泵滑靴磨损故障诊断方法。首先,对原始信号先进行了VMD分解,获得了能量余量;然后,设计了一种建立在特征能量占比(FER)基础上的变分模态分解特征能量重构法(VMD)和多尺度散布熵(MDE)的方法;最后,以柱塞泵故障诊断为研究对象,通过仿真分析方法,依次对柱塞泵在正常状态与滑靴端面磨损为0.1 mm、0.2 mm、0.3 mm状态下的情况进行了分析。仿真及研究结果表明:在逐渐增加时间尺度的过程中,粗粒化序列的随机性和复杂性都明显下降;故障程度增大后,形成了更加规律的变化过程;与DE、MSE和MFE相比,该方法的计算速度更快,分离效果更好;ELM相对SVM的训练时间缩短了12.5%,同时测试精度提升了17%;相对于其他方法,采用该方法诊断柱塞泵滑靴磨损故障时获得了更快的分类速率与更高的准确性,提高了故障诊断效率。The variational mode decomposition characteristic energy reconstruction(VMD)method has a problem of low accuracy in fault realization.Aiming at this problem,a fault diagnosis method of piston pump slipper wear based on variational mode decomposition(VMD)and multi-scale entropy was proposed.Firstly,the energy allowance of original signal was obtained by VMD decomposition,a VMD-MDE(multiscale distribute entropy)method based on characteristics of energy ratio(FER)was designed.Taking the fault diagnosis of the piston pump as the research object,the condition of the piston pump under the normal condition and the slipper end wear of 0.1 mm,0.2 mm and 0.3 mm was analyzed in turn.The simulation results show that the randomness and complexity of the coarse-grained sequence decrease obviously with the increase of time scale.After the degree of failure increases,a more regular change process is formed.Comparing with DE,MSE and MFE,the calculation speed is faster and the separation effect is better.The training time of ELM compared to SVM is reduced by 12.5%,and the test accuracy is increased by 17%.The efficiency of fault diagnosis is improved.Comparing with other methods,the proposed method can obtain faster classification rate and higher accuracy when diagnosing slipper wear fault of the plunger pump.

关 键 词：柱塞泵 磨损振动 信号提取 变分模态分解特征能量重构法 特征能量占比 多尺度散布熵 

分 类 号：TH322[机械工程—机械制造及自动化]