基于VMD-CWT和CNN数控机床刀具健康诊断方法的研究  

Research on the methods of tool health diagnosis for CNC machine tools based on VMD-CWT and CNN

作　　者：王寿元 李积元 张涛 WANG Shouyuan;LI Jiyuan;ZHANG Tao(College of Mechanical Engineering,Qinghai University,Xining 810016,China)

机构地区：[1]青海大学机械工程学院,青海西宁810016

出　　处：《青海大学学报》2023年第6期69-77,共9页Journal of Qinghai University

基　　金：青海省科学技术厅项目(2020-ZJ-740)。

摘　　要：在数控机床切削过程中,刀具的健康状况直接影响产品的加工质量。因此,为了对刀具的健康状态作出诊断评价,本文提出了一种基于变分模态分解(VMD)和连续小波变换(CWT)特征提取与卷积神经网络(CNN)的刀具健康诊断方法。该方法首先采集不同健康状态下的刀具在切削时的振动信号,然后经VMD分解成若干IMF分量,并求解每个IMF分量的相关系数,选取相关系数较大的分量进行信号重构;其次采用连续小波变换来构造重构信号的时频图;最后将得到的时频图输入构建的CNN模型中,通过多层卷积、池化处理得到信号特征与刀具健康状态之间的准确映射,进而实现刀具的健康诊断。经实验验证表明,本文所提方法的识别准确率达到98.9%,具有良好的状态识别能力和泛化性,可为刀具健康诊断方法提供一定的理论依据。The health state of CNC machine tools directly affects the processing quality of product in the cutting process.Therefore,a tool health diagnosis method based on the Variational mode decom-position(VMD),Continuous wavelet transform(CWT)feature extraction method and Convolutional neural network(CNN)isproposed in order to makeahealth diagnostic evaluation of the tool.Firstly,the vibration signals of the tool in different health states are collected during the cutting process.Then,these vibration signals are decomposed into several IMF components by using VMD,and then the correlation coefficients of each IMF component are calculated.The components with lar-ger correlation coefficients are selected for signal reconstruction.Secondly,CWT is used to construct the time-frequency diagram of the reconstructed signals.Finally,the obtained time-frequency dia-gram is input into the constructed CNN model.The accurate mapping between signal features and tool health states is obtained through multi-layer convolution and pooling,so as to realize the health diagnosis of the tool.The experiments show that the recognition accuracy of the proposed method rea-ches 98.9%with good state recognition ability and generalization,providing an important theoretical basis for the health diagnosis method of the tool.

关 键 词：刀具健康诊断 变分模态分解 连续小波变换 卷积神经网络 

分 类 号：TG659[金属学及工艺—金属切削加工及机床]