摘要
被动式钢轨打磨是利用列车拖拽打磨砂轮使其被动旋转产生切削力,从而对钢轨进行磨削,打磨砂轮在作业时一旦发生裂槽故障,会影响打磨效果。为识别出被动式打磨砂轮的裂槽故障,对其振动信号进行分析。通过被动式打磨试验台获取裂槽故障砂轮打磨作业时的振动信号,利用局部均值分解方法将振动信号分解到多个分量,计算分量的模糊熵对信号特征进行量化,选取部分分量模糊熵构建特征向量训练SVM模型,对打磨砂轮的裂槽故障进行识别。研究表明:LMD将被动式打磨砂轮振动信号分解为5个PF分量,分量的模糊熵依次降低;正常、一处裂槽和两处裂槽三种运行状态的打磨砂轮在其振动信号PF2和PF3分量的模糊熵中存在差异;利用砂轮的该故障特征信息训练诊断模型,SVM模型有99.47%的砂轮裂槽故障识别准确率。
Passive rail grinding is to grind the rail by dragging the grinding wheel and making it rotate passively to produce cutting force,once the grinding wheel cracks during operation,it will affect the grinding effect.The vibration signal is analyzed to identify the crack groove fault of passive grinding wheel.The vibration signal of grinding wheel with crack groove fault is obtained by passive grinding test bench,and decomposed into multiple components by local mean decomposition method.Then calculate the fuzzy entropy of components to quantify the signal features,and selects part of the fuzzy entropy to construct the eigenvector and train SVM model,and identify the crack groove fault of grinding wheel.Research shows that:LMD decompose the vibration signal of passive grinding wheel into 5 PF components,and the fuzzy entropy of components decrease successively.There are differences in the fuzzy entropy of PF2 and PF3 components of the vibration signals of grinding wheels in normal,crack groove and two crack grooves operating states.And this fault characteristics of grinding wheel were used to train the diagnosis model.SVM model has 99.47%accurate recognition rate of grinding wheel crack groove fault.
作者
白祥
温昊宇
刘晓龙
蒋晓光
赵鑫江
王衡禹
BAI Xiang;WEN Haoyu;LIU Xiaolong;JIANG Xiaoguang;ZHAO Xinjiang;WANG Hengyu(State Key Laboratory of Traction Power,Southwest Jiaotong University,Chengdu 610031,China;Hohai-Lille College,Hohai University,Changzhou 210098,China;China Railway Chengdu Group Co.,Ltd.,Chengdu 610082,China;Sichuan SWJTU Railway Development Co.,Ltd.,Chengdu 610091,China)
出处
《机械》
2025年第2期1-7,共7页
Machinery
基金
国家自然科学基金面上项目(51775454)
四川省科技计划项目(24NSFSC7243)。
关键词
钢轨被动式打磨
打磨砂轮
局部均值分解
模糊熵
支持向量机
passive rail grinding
grinding wheel
local mean decomposition method
fuzzy entropy
support vector machine
