IMIBSE与ISOMAP在旋转机械故障诊断中的应用  

Application of IMIBSE and ISOMAP in fault diagnosis of rotating machinery

作　　者：周继彦[1] 柳金峰 胡义华 ZHOU Jiyan;LIU Jinfeng;HU Yihua(College of Robotics,Guangdong Provincial Science and Technology Cadre College,Zhuhai 519000,China;The Institute of International Education,Guangxi University of Science and Technology,Liuzhou 545006,China)

机构地区：[1]广东省科技干部学院机器人学院,广东珠海519000 [2]广西科技大学国际教育学院,广西柳州545006

出　　处：《机电工程》2024年第6期1027-1038,1067,共13页Journal of Mechanical & Electrical Engineering

基　　金：广东省教育厅重点领域专项(2022ZDZX1055);广东省教育厅科研项目(2021KTSCX218);广东省普通高校重点科研项目(2022ZDZX4075);广东省智能装备制造工程研究中心项目(2021GCZX018);广东省普通高校特色创新项目(2022KTSCX251)。

摘　　要：针对基本熵的区域划分标准不理想,导致无法有效测量振动信号的复杂度,使故障诊断的准确率不佳这一问题,提出了一种基于改进多尺度改进基本熵(IMIBSE)、等距特征映射(ISOMAP)和随机森林(RF)的旋转机械故障诊断方法。首先,采用基于方差的区域划分准则对基本熵进行了改进,结合改进的粗粒化处理,提出了IMIBSE,并将其用于提取故障特征;随后,利用ISOMAP对原始故障特征进行了特征降维,选择了对分类贡献最大的一组特征作为故障敏感特征;最后,基于RF建立了多故障分类器,将故障敏感特征输入至RF模型进行了训练和测试,实现了旋转机械的故障识别,利用齿轮箱和离心泵两种故障数据集将IMIBSE方法与复合多尺度基本熵、多尺度改进基本熵、多尺度基本熵进行了比较和分析。研究结果表明:IMIBSE不仅具有最佳的可视化效果,而且取得的识别准确率最高,二者均达到了100%,而二者的平均分类准确率分别为100%和99.8%;相较于其他故障诊断方法,IMIBSE方法的准确率更高,而且适用于小样本的故障识别问题。Aiming at the problem that the region division standard of basic entropy method was not accurate,which could not effectively measure the complexity of vibration signal of rotating machinery,and the accuracy of rotating machinery fault diagnosis was poor,a rotating machinery fault diagnosis method based on improved multi-scale improved basic entropy(IMIBSE),isometric feature mapping(ISOMAP) and random forest(RF) was proposed.Firstly,the regional division criterion of variance was used to improve the basic entropy,and combining with the improved coarse-grained processing,the IMIBSE method was proposed and used to extract the fault characteristics of rotating machinery.Then,ISOMAP method was used to reduce the feature dimension of the original fault features,and a group of features that contribute the most to classification was selected as the fault sensitive features.Finally,a multi-fault classifier was built based on RF,and the fault sensitive features were input to RF model for training and testing,so as to realize the fault identification of rotating machinery.The IMIBSE method was compared with composite multi-scale basic entropy,multi-scale improved basic entropy and multi-scale basic entropy by using two fault data sets of rolling bearing and centrifugal pump.The experimental results show that IMIBSE method not only have the best visualization effect,but also have the highest recognition accuracy,both reaching 100%,and the average classification accuracy of each is 100% and 99.8%,respectively.Comparing with other fault diagnosis methods,IMIBSE method has higher accuracy and is suitable for small sample fault identification.

关 键 词：齿轮箱 离心泵 故障诊断 改进多尺度改进基本熵 等距特征映射 随机森林 改进的粗粒化处理 

分 类 号：TH133.3[机械工程—机械制造及自动化]