【作者】 李奕璠；

【导师】 林建辉；

【作者基本信息】 西南交通大学 ， 载运工具运用工程， 2013， 博士

【摘要】 对列车关键部件进行实时的安全监测是确保列车运行安全性的重要手段,通过实时的记录、分析与诊断,及时发现存在严重安全隐患的车辆,并查找出故障的位置与原因,实施扣车检修,对保障铁路行车安全具有十分积极的作用,具有显著的社会与经济效益。本文正是在中国铁路跨越式发展背景下,以轮轨动态作用安全为主题,从列车和轨道线路整体大系统的角度,综合研究列车与轨道动态相互作用安全问题。主要进行了如下的研究工作：(1)轮轨力连续测试的实现根据轮轨相互作用的特点和轨道的约束条件,通过仿真计算研究了轮轨垂向力与横向力测试问题,详细分析了荷载在移动过程中轨道的响应的变化以及不同位置测点的应变输出,确定了连续测试方案中测点的布置位置,并通过室内试验进行了验证。以此为依据,在现场测试中提出了一种基于径向基函数神经网络的轮轨力连续测试方法,以及一种基于灰色理论、遗传算法与神经网络的组合模型的轮轨力连续测试方法。通过现场试验,证明了其高精度与可行性,并降低了传感器误差对测试结果的影响。(2)轮轨力解耦问题的研究通过仿真计算研究了横向力作用下的垂向力测点的应变输出以及垂向力作用下的横向力测点的应变输出,发现横向力和垂向力间存在串扰问题,同时在室内试验中也发现了此现象。针对此问题,本文提出使用一种改进的独立分量分析算法对轮轨力的串扰进行解耦,为高精度轮轨力测试提供了一种新的手段。(3)车辆轨道耦合动力学仿真分析模型的建立以车辆轨道耦合动力学理论为基础,以高速列车为研究对象,采用铁道车辆动力学仿真技术,系统地分析车辆轨道耦合系统的轮轨接触和相互作用问题,分别建立车辆和轨道结构的力学模型和动力学仿真分析模型,采用Hertz接触理论和Kalker滚动接触理论,研究轮轨滚动接触关系中车轮失圆时的轮轨相互作用的过程和轮轨作用力,为研究车轮失圆对轮轨系统动力特征的影响提供基础。(4)车轮失圆对轮轨动力作用的影响分析车轮失圆是影响行车安全的重大隐患。对车轮失圆形式进行定义分类,结合所建立的车辆轨道耦合动力学仿真分析模型,研究车轮扁疤、车轮周期性多边形磨耗等常见的车轮失圆类型所造成的轮轨冲击振动的特征,从而为轮轨力实时监测系统的建立提供理论基础。(5)基于轮轨力的车轮失圆的识别利用车辆轨道耦合动力学仿真模型,计算了80组不同长度、不同深度的车轮扁疤作用下的轮轨冲击力,根据轮轨力的大小将车轮分为3种类型：正常车轮、问题车轮和严重问题车轮。首先通过特征提取方法对原始数据进行压缩,避免维数灾难,再详细分析了3种分类器：BP神经网络、学习向量量化神经网络和支持向量机对车轮进行分类识别情况。(6)基于轴箱振动加速度的车轮失圆的识别车轮失圆将会引起轴箱的异常振动,通过安装在轴箱上的加速度传感器可能对车轮失圆情况进行识别。本文提出一种基于改进的EMD的Hilbert谱方法,利用实测数据及仿真计算,有效地对车轮失圆问题进行了定性的识别,该方法不仅能判断车轮是否存在失圆状况,还能判断失圆类型是车轮扁疤还是车轮多边形化,且不受车速的影响。同时,利用小波包分解还能对失圆程度进行定量判断。更多还原

【Abstract】 Real-time safety monitoring of train key component is an important means to ensure the security of train operation. Through the real-time recording, analysis and diagnosis, timely detect vehicles which exists serious safety hazard and find out the location of the fault and its reason, furthermore, implement of vehicle maintenance, which has a very positive role in the protection of railway traffic safety and significant social and economic benefits. This paper is in the context of the leaps and bounds development of the China railway, use the wheel-rail dynamic impact as the theme, from the point of view vehicle and track overall system, comprehensively research vehicle-track dynamic interaction security issue. This paper mainly carries out the following research works:(1) Realization of wheel-rail force continuous measurementAccording to wheel-rail interaction characteristics and the restriction condition of track, the single point measure of wheel-rail vertical force and lateral force are studied by means of simulation calculation, rail response change when load moves and stain outputs of measure points in different position are detailed analysed, finally the measure points arrangement position of continuous measurement is determined, and simulation results are verified by laboratory experiment. On this basis, a radial basis function neural network and a combination model based on the grey system, genetic algorithm and neural work based wheel-rail force continuous measure method is put forward. It has been proved their high precision and feasibility through field test and reduces the influence of sensor failures on measure results.(2) Research of wheel-rail force decoupingStudy strain outputs of wheel-rail vertical force measure point under the action of lateral force and strain outputs of lateral force measure point under the action of vertical force by simulation, the results show that wheel-rail vertical force and lateral force exist crosstalk, at the same time, this phenomenon is found in laboratory test. Aim at this problem, this paper proposes to use an improved independent component analysis algorithm to decouple the crosstalk of wheel-rail vertical force and lateral force, this method provides a new approach for high precision wheel-rail force measurement.(3) Establish of vehicle-track coupling dynamics modelOn the basis of the theory of vehicle-track coupling dynamics, with high speed train as research object, railway vehicle dynamics simulation technology is used to systematically analyse wheel-rail contact and interaction of vehicle-track coupling system, a mechanical model of vehicle and track structure and dynamics simulation model are established respectively, the Hertz contact theory and Kalker rolling contact theory are employed to research wheel-rail interaction process and wheel-rail force of out-of-round wheel, which provides a basis for the study the influence of out-of-round wheel on wheel-rail dynamic characteristics.(4) Analysis of the influence of out-of-round wheel on wheel-rail dynamic actionOut-of-round wheel is a serious hidden danger for vehicle running safety. This paper defines and classifies the forms of out-of-round wheel, combines with the established vehicle-track coupling dynamics model to study wheel-rail impact and vibration characteristic caused by different out-of-round wheel styles, such as wheel flats and wheel periodic polygon, which provides a theoretical basis for the establishment of wheel-rail force real-time monitoring system.(5) Identification of out-of-round wheel through wheel-rail forceVehicle-track coupling dynamics model is used to calculate wheel-rail impact force caused by80groups of wheel flat with different length and depth, and the wheels are divided into3styles according to the amplitude of wheel-rail force, there are normal wheels, fault wheels and severe fault wheels. Compress the original data through feature extraction firstly to avoid the curse of dimensionality, and then analyses three different classifiers:BP neural network, learning vector quantization neural network and support vector machine for wheel classification case in detail.(6) Identification of out-of-round wheel through axle box vibration accelerationOut-of-round wheel could cause abnormal vibration of the axle box, we could identify potential wheel defect through axle box vibration information. This paper uses measured data and simulation results, puts forward a Hilbert spectrum method based on improved EMD to qualitative identify out-of-round wheel effectively, the method can not only determine whether the wheel exists defect, but also judge the wheel defect is whether wheel flat or wheel polygon regardless of the speed. At the same time, quantitatively judge defect degree by means of wavelet package decomposition.更多还原