基于声发射和支持向量机的滑动轴承润滑状态识别实验  

Experimental Study of Lubrication States Identification for Journal Bearings with Acoustic Emission Technique and Support Vector Machine Method

作　　者：王琳[1,2,3] 王星[1] 滕金磊[1] 张宝文 鲁如烨 WANG Lin;WANG Xing;TENG Jinlei;ZHANG Baowen;LU Ruye(School of Mechanical Engineering,Northwestern Polytechnical University,Xi’an 710072,China;National Center for Virtual Simulation Experimental Education of Mechanical-Fundamentals and Aeronautical Manufacturing,Northwestern Polytechnical University,Xi’an 710072,China;National Virtual Simulation Experimental Teaching Conter for Mechanical Fundamentals and Aviation Manufacturing,Northwestern Polytechnical University,Xi’an 710072,China)

机构地区：[1]西北工业大学机电学院,西安710072 [2]西北工业大学机械基础国家级实验教学示范中心,西安710072 [3]西北工业大学机械基础与航空制造国家级虚拟仿真实验教学中心,西安710072

出　　处：《实验室研究与探索》2024年第12期6-10,共5页Research and Exploration In Laboratory

基　　金：国家自然科学基金项目(51975475,52375265);教育部实验教学和教学实验室建设研究项目(SYJX2024-201);陕西本科和高等继续教育教学改革研究项目重点攻关项目(23BG006);教育部产学合作协同育人项目(220905248273407);西北工业大学教育教学改革研究项目(2023JGY11);西北工业大学高阶项目式课程建设项目(PX59245184)。

摘　　要：为了对滑动轴承润滑状态进行高效快速准确地在线识别,设计搭建了滑动轴承润滑状态实验台及声发射测量系统,使用声发射技术和遗传算法优化支持向量机的方法对滑动轴承润滑状态进行了实验测试和分析识别。通过对声发射信号进行预处理,从时域、频域、信息熵等多方面提取和选择有效特征参数,将提取到的有效特征参数组合成特征向量用作支持向量机的输入并得到支持向量机分类器的识别结果;通过遗传算法对惩罚因子和核函数参数组合进行优化,获得最佳的润滑状态识别结果,总体准确率达到了93.3%。In order to online identify the lubrication state of sliding bearing with an efficient,fast and accurate way,an experimental system and a relevant acoustic emission measurement platform were constructed.The lubrication states of sliding bearings were tested and analyzed by the acoustic emission technique and the genetic algorithm-support vector machine method.The acoustic emission signals were firstly pre-processed.The effective feature parameters were extracted and selected from time domain,frequency domain,information entropy,etc.Then,the extracted effective feature parameters were combined into a feature vector as the input of support vector machine.The multi-lubrication states were identified by the support vector machine classifier,and furtherly optimized by the combination of the penalty factor and the kernel function parameter through the genetic algorithm.The optimal lubrication states identification was obtained,and the overall accuracy rate reached 93.3%.

关 键 词：风电齿轮箱 滑动轴承 润滑状态 声发射 支持向量机 

分 类 号：TH117.1[机械工程—机械设计及理论]