基于VMD-MMPE的轧机轴承滚动体与保持架故障诊断  

Failure-Diagnosis Approaches of VMD-MMPE‑Based Rolling Elements and Cages in Rolling-Mill Bearings

作　　者：计江 赵琛[3] 王勇勤 JI Jiang;ZHAO Chen;WANG Yongqin(College of Mechanical Engineering,Chongqing University Chongqing,400030,China;China National Heavy Machinery Research Institute Co.,Ltd.Xi′an,710032,China;National Cold Rolling Strip Equipment and Process Engineering Technology Research Center,Yanshan University Qinhuangdao,066004,China)

机构地区：[1]重庆大学机械与运载工程学院,重庆400030 [2]中国重型机械研究院股份公司,西安710032 [3]燕山大学国家冷轧板带装备及工艺工程技术研究中心,秦皇岛066004

出　　处：《振动．测试与诊断》2023年第2期290-297,409,共9页Journal of Vibration,Measurement & Diagnosis

基　　金：国机集团重大科技专项基金资助项目(SINOMAST-ZDZX-2018-06)。

摘　　要：针对板带轧机轴承工作环境恶劣、保持架与滚动体极易损坏、信号噪声大、识别困难以及实际工况对诊断速度要求高等问题,首先,提出粒子群优化变分模态分解(particle swarm optimization-variational mode decomposition,简称PSO-VMD)和多元多尺度排列熵(multivariate multiscale permutation entropy,简称MMPE)的故障诊断方法,并结合粒子群优化支持向量机(particle swarm optimization-support vector machine,简称PSO-SVM)实现故障分类;其次,轴承振动信号经VMD处理为若干模态分量(intrinsic mode functions,简称IMF),选最优分量进行包络分析;然后,针对轧机轴承垂直水平轴向振动差别较大且受较大径向力与轴向力的特点,采用MMPE并考虑3维振动信号的4个分量的MMPE值与时域指标组成特征向量;最后,基于PSO-SVM模型对方法的有效性进行验证。计算和实验结果与集合经验模态分解(ensemble empirical mode decomposition,简称EEMD)与局部均值分解(local mean decomposition,简称LMD)方法对比表明,VMD-MMPE可以优化模型的输入,提高模型的诊断正确率和速度,实现轴承保持架与滚动体不同部位和不同损伤程度的故障诊断,具有重要的工程意义。Concerning the poor working conditions existing in rolling mill bearings,there have many problems,such as larger signal-noise interference,difficult identification,easy to damage cages and rolling elements,as well as higher diagnosis-speed requirements.This paper proposes the two feasible failure diagnosis approaches,namely particle swarm optimization-variational mode decomposition(PSO-VMD)and multivariate multiscale permutation entropy(MMPE).And with combination of particle swarm optimization-support vector machine(PSO-SVM),it successfully realizes identifications for multiple failures and corresponding classifications.The bearing vibration signal is processed by VMD into several intrinsic mode functions(IMF),and the optimal component envelope is selected for analysis.For the mill bearing vertical horizontal axial vibration difference,subject to large radial force and axial force characteristics,using MMPE to consider the four components of the three-dimensional vibration signal of the MMPE value and time domain indicators to form a feature vector.Finally,based on the model of PSO-SVM,it further verifies these methods in effectiveness.And through a series of comparisons with the achieved calculations and experimental results with ensemble empirical mode decomposition(EEMD)and local mean decomposition(LMD),it indicates that VMD-MMPE can better optimize model-inputting work,as well as enhance corresponding diagnostic accuracies and speeds.Finally,it realized effective diagnoses upon failures with different damage degrees in different and same parts of bearing cages and rolling elements,achieving important engineering significance.

关 键 词：轧机轴承 变分模态分解 包络谱 多元多尺度排列熵 粒子群优化支持向量机 故障诊断 

分 类 号：TH133.33[机械工程—机械制造及自动化] TG333.17[金属学及工艺—金属压力加工]