基于GWO-SVM的石化旋转机械轴承故障诊断  

Fault Diagnosis of Petrochemical Rotating Machinery Bearings Based on GWO-SVM

作　　者：莫常春 刘美 费继友[2] 张清华 张斐 吴斌鑫 周正南 MO Changchun;LIU Mei;FEI Jiyou;ZHANG Qinghua;ZHANG Fei;WU Binxin;ZHOU Zhengnan(Guangdong University of Petrochemical Technology, Maoming 525000, China;Dalian Jiaotong University, Da Lian116028, China;Jilin University of Chemical Technology, Changchun 132022, China;Dongguan University of Technology, Donguan523419, China)

机构地区：[1]广东石油化工学院自动化学院,广东茂名525000 [2]大连交通大学机车车辆工程学院,辽宁大连116028 [3]吉林化工学院信息与控制工程学院,吉林132022 [4]东莞理工学院机械工程学院,广东东莞523419

出　　处：《广东石油化工学院学报》2022年第3期41-45,共5页Journal of Guangdong University of Petrochemical Technology

基　　金：国家自然科学基金面上项目(6207024001,62073091);国家自然科学基金重点项目(61933013);广东省高校重点领域(新一代信息技术)专项(2020ZDZX3042);广东省普通高校特色创新项目(2017KTSCX176);机械设备健康维护湖南省重点实验室开放基金项目(21903);东莞理工学院机器人与智能装备创新中心项目(KCYCXPT2017006);广东省普通高校机器人与智能装备重点实验室项目(2017KSYS009)。

摘　　要：针对传统的无量纲指标在不同故障之间存在数据重叠,导致故障诊断准确率低的问题,提出了一种基于互无量纲指标和灰狼算法优化支持向量机(GWO-SVM)的故障识别模型。利用灰狼算法强大的全局搜索能力,对支持向量机关键参数惩罚因子和核函数参数进行寻优,并采用广东省石化装备故障诊断重点实验室多级离心风机轴承数据进行验证。实验证明:GWO-SVM能精确地对轴承故障类型进行分类,同时GWO-SVM相对于单纯的SVM模型和遗传算法优化SVM模型具有更短的算法运行时间和更高准确率,平均准确率高达90%。In view of the problem that the dimensionless index has the phenomenon of coupling between different fault data,a dimensionless index based on each other is proposed.Indicators and gray wolf algorithm optimize the fault recognition model of GWO Support Vector Machine(GWO-SVM).The gray wolf algorithm is added on the basis of the traditional SVM,which can quickly and effectively search for the optimal combination of the penalty factor of the support vector machine and the nuclear parameter,and adopts the 11KW 5-level centrifugal compressor of the Guangdong Provincial Petrochemical Equipment Fault Diagnosis Key Laboratory Bearing data to verify.Experiments have proved that GWO-SVM can accurately classify it,and compared with pure SVM model and genetic algorithm optimized SVM model,GWO-SVM has shorter algorithm running time and higher accuracy.The accuracy rate is as high as 90%.

关 键 词：互无量纲指标 灰狼算法 支持向量机 参数优化 故障诊断 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]