基于非参数化模型的高铁钢轨振动系统实时识别研究

Structural abnormality of High-speed railway (HSR) track is an important influencing factor for train running safety. It is an effective way to realize real-time identification technology of track vibration system using nonparametric model. However, the identification difficulty, which is increased for reasons of variable load, fast damping vibration, longitudinal temperature force and so on, becomes the biggest bottleneck that constrains the technology application. The quantitative expression of the nonparametric model in the track vibration system identification is taken as the research object, and the basic scientific research is carried out on the mechanical characteristics of the multidimensional coupling field of the wheel-track-fastener system and the vibration system identification theory. The main contents include: interaction forces of the components of the track-fastener system and the influence of the vibration conduction and mode distribution of the tracks under different excitations and temperatures are calculated by laboratory vibration test and theoretical analysis; the problems about state space construction, model training and over-fitting in nonparametric model are studied to establish the theory of track vibration system identification; the model predictive evaluation method based on statistical probability and the classification method of evaluation result are studied. The purpose of this project is to reveal the behavior characteristics of track vibration, establish the theory about track vibration system identification, realize rapid and accurate identification of track safety status, and lay the theoretical and technical foundation for track vibration system identification in HSR safety monitoring.

高铁钢轨结构异常是影响行车安全的重要因素，基于非参数化模型的振动系统实时识别技术是解决该问题的有效途径。然而，钢轨所受激励复杂多变、振动衰减较快、内部存有温度力等问题，成为制约该项技术应用的主要瓶颈。本项目以钢轨振动系统辨识中非参数化模型各组成部分的定量表达为研究对象，对车轮-钢轨-扣件系统多维耦合场力学特征及振动系统辨识理论进行基础科学研究。具体内容包括：通过实验室振动试验和理论分析计算出不同激励和温度下钢轨-扣件系统各组成部件相互作用力及对钢轨振动传导与模态分布的影响规律；研究解决非参数化模型中状态空间构建、模型训练及过拟合问题，建立基于非参数化模型的钢轨振动系统识别理论；研究基于统计概率的模型预测评价方法和评价结果的分级方法。本项目旨在揭示钢轨振动的行为特征，建立钢轨振动系统识别理论，实现快速精准识别钢轨安全状态，为高铁安全监测中的钢轨振动系统识别和振动监测平台的完善奠定理论基础。

由于列车行驶速度、载重和运输密度的提高，使得轮轨动力学作用增强、冲击频率提高，增大了钢轨损伤概率，对于轨道的主动性监测是轨道“运营维护”中亟需解决的问题。为此，本项目进行了适用于钢轨振动的基于贝叶斯小波包降噪算法研究、温度场对钢轨振动模态影响规律研究、钢轨中超声导波传播特性研究、基于钢轨振动特性测量温度应力方法研究以及上述理论方法与系统的集成与实验验证研究等。这些项目研究成果将有助于建立国家铁路“运营维护”体系标准。