测力钢轨轮轨力连续输出算法  

Wheel/rail force continuous exporting algorithm of instrumented rail

作　　者：李奕璠 刘建新[1] 王开云[1] 林建辉[1] 王超峰[1] 

机构地区：[1]西南交通大学牵引动力国家重点实验室,四川成都610031

出　　处：《交通运输工程学报》2011年第4期36-40,共5页Journal of Traffic and Transportation Engineering

基　　金：国家863计划项目(2009AA11Z202);牵引动力国家重点实验室自主研究课题(2009TPL_T07)

摘　　要：根据轮轨相互作用特点,利用地面测试数据的信息,采用阈值判断法提取有效的轮轨力数据。设计基于径向基函数神经网络的算法,用以处理不同测试单元处的轮轨力的非线性关系,用不同车轮不同作用点位置作用下的横、垂向力训练神经网络,实现了测力钢轨轮轨力的连续测试,并对3种工况进行了仿真试验。分析结果表明:既存在干扰又存在应变片损坏时的连续轮轨力精度较只存在干扰及部分应变片损坏的低;算法能对采样频率低于8 720.9 Hz的轮轨力信号进行实时处理,算法具有很好的适用性。According to wheel/rail interaction characteristics,ground test data were used,and the available data of wheel/rail force were extracted by using threshold value judgmental method.An algorithm based on radial basic function neural network（RBFNN） was designed to deal with the nonlinear relationship of wheel/rail forces at different measure points,and the neural network was trained by lateral and vertical forces at different contact points of different wheels.The wheel/rail force continuous measurement of instrumented rail was achieved,and the simulation experiments under three working conditions were conducted.Analysis result indicates that the accuracy of continuous wheel/rail force when both interference and strain gage damage exist is lower than that when interference or part strain gage damage exists respectively.The algorithm can real-timely process wheel/rail force signal when sampling frequency is less than 8 720.9 Hz,and has good applicability.5 figs,14 refs.

关 键 词：轮轨力 测力钢轨 连续测试 径向基函数神经网络 安全监测 

分 类 号：U211.5[交通运输工程—道路与铁道工程]