基于IMIE、MCFS和SSA-ELM的离心泵故障诊断方法  

Fault diagnosis method of centrifugal pump based on IMIE and SSA-ELM

作　　者：辜文娟 张扬 GU Wenjuan;ZHANG Yang(Jiangxi Business School,Nanchang 330038,China;School of Mechanical Engineering,Hubei University of Technology,Wuhan 430068,China)

机构地区：[1]江西省商务学校,江西南昌330038 [2]湖北工业大学机械工程学院,湖北武汉430068

出　　处：《机电工程》2023年第9期1456-1463,共8页Journal of Mechanical & Electrical Engineering

基　　金：江西省教育厅科技课题(GJJ151486)。

摘　　要：采用多尺度排列熵对离心泵振动信号进行分析时,存在忽略信号幅值信息以及粗粒化处理存在不足的问题,从而导致离心泵的故障识别准确率不高,为此,提出了一种基于改进多尺度增长熵(IMIE)、多聚类特征选择(MCFS)和麻雀搜索算法优化极限学习机(SSA-ELM)的离心泵故障诊断方法。首先,基于改进粗粒化处理,提出了改进多尺度增长熵(IMIE)方法,将其用于提取故障特征,构造了反映离心泵损伤属性的特征矩阵;随后,采用多聚类特征选择(MCFS),对原始故障特征进行了重要性排序,获得了对分类识别贡献度更高的故障特征,提高了故障特征的质量;最后,将低维的敏感特征输入至基于麻雀搜索算法(SSA)的极限学习机(ELM)中,进行了离心泵故障分类,完成了离心泵不同故障类型的识别任务;并采用离心泵故障数据集,对基于IMIE、MCFS和SSA-ELM的故障诊断方法的有效性进行了实验验证。研究结果表明:所提故障诊断方法的故障识别准确率达到了100%,多次实验的平均准确率和标准差也优于其他对比的故障诊断方法,即IMIE能够准确地提取信号中的故障信息,进而表征离心泵的健康状态;SSA-ELM能够准确地识别离心泵的故障类型,证明该方法具有一定的有效性和优越性。Aiming at the shortcomings of multi-scale permutation entropy in analyzing the vibration signal of centrifugal pump,which ignored signal amplitude information and had insufficient coarse granulation processing,resulting in low fault recognition accuracy,a fault diagnosis method of centrifugal pump based on improved multi-scale increment entropy(IMIE),multi cluster feature selection(MCFS)and sparrow search algorithm optimized extreme learning machine(SSA-ELM)was proposed.Firstly,based on improved coarse-grained processing,an improved multi-scale increment entropy(IMIE)method was proposed,which was used to extract the fault characteristics,and construct a characteristic matrix reflecting the damage properties of the centrifugal pump.Then,multi cluster feature selection(MCFS)was used to rank the importance of the original fault features,obtain fault features that contribute more to classification recognition,and improve the quality of fault features.Finally,the low-dimensional sensitive features were input into the extreme learning machine based on the sparrow search algorithm(SSA)for classification,and the identification of different fault types of centrifugal pumps was completed.The effectiveness of the proposed fault diagnosis method was experimentally tested by the fault data set of centrifugal pumps.The research results show that the fault identification accuracy of the proposed fault diagnosis method reaches 100%,and the average accuracy and standard deviation under multiple classification experiments are also better than other comparison fault diagnosis methods,which is superior to the comparative fault diagnosis method,that is,IMIE can accurately extract the fault information in the signal and characterize the health status of the centrifugal pump,and SSA-ELM can accurately identify the fault type of centrifugal pump,which proves that the method has certain effectiveness and advantages.

关 键 词：叶片式泵 改进粗粒化处理 改进多尺度增长熵 多聚类特征选择 麻雀搜索算法 极限学习机 特征矩阵 

分 类 号：TH311[机械工程—机械制造及自动化]