基于GLCT及CPA-SVM的变转速齿轮泵健康状态分类研究  

Study on Health Status Classification of Variable Gear Pump Based on GLCT and CPA-SVM

作　　者：郭锐[1,2,3] 张印浩 牛雯雯 骆雄帅 蔡伟 王建伟 王岳峰[1] 赵静一 GUO Rui;ZHANG Yinhao;NIU Wenwen;LUO Xiongshuai;CAI Wei;WANG Jianwei;WANG Yuefeng;ZHAO Jingyi(Hebei Provincial Key Laboratory of Heavy Machinery Fluid Power Transmission and Control,Yanshan University,Qinhuangdao 066004;Key Laboratory of Space Launching Site Reliability Technology,Xichang Satellite Launch Center,Haikou 571126;Key Laboratory of Advanced Forging&Stamping Technology and Science,Yanshan University,Qinhuangdao 066004;Hebei Key Laboratory of Special Delivery Equipment,Yanshan University,Qinhuangdao 066004)

机构地区：[1]燕山大学河北省重型机械流体动力传输与控制重点实验室,秦皇岛066004 [2]西昌卫星发射中心航天发射场可靠性技术重点实验室,海口571126 [3]燕山大学先进锻压成形技术与科学教育部重点实验室,秦皇岛066004 [4]燕山大学河北省特种运载装备重点实验室,秦皇岛066004

出　　处：《机械工程学报》2023年第14期310-319,共10页Journal of Mechanical Engineering

基　　金：国家自然科学基金资助项目(52075469,12173054)。

摘　　要：针对齿轮泵变转速工况,提出广义线性调频小波变换(General linear chirplet transform,GLCT)和食肉植物算法(Carnivorous plant algorithm,CPA)优化支持向量机(Support vector machines,SVM)相结合的齿轮泵健康状态分类识别方法。首先选取4组磨损量不同的轴套,利用改造试验台采集不同状态下齿轮泵的振动信号;然后,引入GLCT时频分析方法消除转速变化的影响,提取瞬时旋转频率,进行角度域重采样,提取角度域中峰值指标、脉冲指标、峭度指标,与阶次谱方均根值、阶次域阶次幅值作为特征参数;最后,引入CPA对SVM两个参数c和g优化的分类方法,进行齿轮泵的健康状态进行分类识别,为了进一步验证算法有效性将其与SVM和极限学习机(Extreme learning machine,ELM)两种方法进行比较。结果表明,提出的分类方法平均准确率可达93.75%以上,能有效提高分类识别准确率。Aiming at the variable speed condition of gear pump,a gear pump health state classification and recognition method based on general linear chirplet transform(GLCT)and carnivorous plant algorithm(CPA)optimized support vector machines(SVM)is proposed.Firstly,four groups of shaft bushing with different wear amount are selected,and vibration signals of gear pump under different states are collected by the modified test bed.Then,the time-frequency analysis method of GLCT is introduced to eliminate the influence of speed change.The instantaneous rotation frequency is extracted,and the angle domain resampling is carried out.The peak index,pulse index,kurtosis index in Angle domain are extracted,and the root mean square value of order spectrum and the amplitude of order domain are taken as the characteristic parameters.Finally,CPA is introduced to optimize the c and g two parameters of SVM to classify and identify the health status of gear pump.In order to further verify the validity of the algorithm,it is compared with SVM and ELM.The results show that the average accuracy of the classification method proposed can reach more than 93.75%,which can effectively improve the accuracy of classification and recognition.

关 键 词：齿轮泵 变转速 健康状态评估 广义线性调频小波变换 支持向量机 

分 类 号：TG156[金属学及工艺—热处理]