基于CEEMDAN多尺度排列熵和SO-RELM的高压隔膜泵单向阀故障诊断  

Fault diagnosis of one-way valve of high-pressure diaphragm pump based on CEEMDAN multi-scale permutation entropy and SO-RELM

作　　者：李瑞[1,2] 范玉刚 LI Rui;FAN Yugang(College of Information Engineering&Automation,Kunming University of Science and Technology,Kunming 650500,China;Yunnan Provincial Key Lab of Artificial Intelligence,Kunming 650500,China)

机构地区：[1]昆明理工大学信息工程与自动化学院,昆明650500 [2]云南省人工智能重点实验室,昆明650500

出　　处：《振动与冲击》2023年第5期127-135,共9页Journal of Vibration and Shock

基　　金：国家自然科学基金(62163020);云南省基础研究计划项目(2019FD042);工业控制技术国家重点实验室(浙江大学)开放课题资助(ICT2022B06)。

摘　　要：高压隔膜泵单向阀受负载、摩擦和冲击等因素的影响,运行产生的振动信号具有非平稳、非线性的特点,为了从振动信号中提取设备的非线性动力学特征,将多尺度排列熵(multi-scale permutation entropy, MPE)引入高压隔膜泵单向阀故障诊断研究。提取振动信号多尺度排列熵特征,用于建立结构优化正则化极限学习机(structure optimization regularized extreme learning machine, SO-RELM)故障诊断模型,模型利用K-means优化RELM结构,提高模型识别精确度及稳定性。首先采用自适应噪声完备经验模态分解(complementary ensemble empirical mode decomposition with adaptive noise, CEEMDAN)将高压隔膜泵单向阀振动信号自适应分解为多个固有模态分量(intrinsic mode function, IMF),以相关系数为指标,优选包含故障特征信息丰富的分量;然后,计算IMFs的多尺度排列熵值,提取信号的非线性动力学特征;最后,基于多尺度排列熵,建立基于SO-RELM的故障诊断模型。试验结果表明,CEEMDAN多尺度排列熵能够准确表征高压隔膜泵单向阀运行状态的非线性动力学特征,基于CEEMDAN多尺度排列熵建立的SO-RELM故障模型,能够有效识别高压隔膜泵单向阀工况类型,准确率达98.89%。Due to effects of load, friction, impact and other factors on one-way valve of high-pressure diaphragm pump, its vibration signals generated in operation have non-stationary and nonlinear characteristics. Here, to extract nonlinear dynamic characteristics of equipment from vibration signals, multi-scale permutation entropy(MPE) was introduced into fault diagnosis of one-way valve of high-pressure diaphragm pump. MPE features of vibration signals were extracted to establish the fault diagnosis model of structure optimization regularized extreme learning machine(SO-RELM). The model used K-means to optimize RELM structure, and improve accuracy and stability of model identification. Firstly, vibration signals of one-way valve of high-pressure diaphragm pump were adaptively decomposed into multiple intrinsic mode functions(IMFs) by using the complementary ensemble empirical mode decomposition with adaptive noise(CEEMDAN). Taking correlation coefficient as the index, IMFs with rich fault feature information were selected. Then, MPEs of the selected IMFs were calculated to extract nonlinear dynamic characteristics of vibration signals. Finally, based on MPEs, a fault diagnosis model based on SO-RELM was established. Test results showed that CEEMDAN multi-scale permutation entropy can accurately characterize nonlinear dynamic characteristics of operating state of one-way valve of high-pressure diaphragm pump;SO-RELM fault diagnosis model established based on CEEMDAN multi-scale permutation entropy can effectively identify working condition type of one-way valve of high-pressure diaphragm pump with the accuracy of 98.89%.

关 键 词：自适应噪声完备经验模态分解 排列熵 结构优化正则化极限学习机 故障诊断 

分 类 号：TN710.1[电子电信—电路与系统] TH165.3[机械工程—机械制造及自动化]