弯扭耦合激励下基于代理模型的高速列车车轴裂纹在线定量识别方法研究

Axle is one of the important components in wheelset system of the train bogie, and its quality status directly affects the transportation safety of high-speed railway. Cracks in the axles may lead to very serious accidents especially in high speed trains, therefore, it is extremely important to monitor the health condition of the wheelset system axles. Taking into account the bending-torsional loads of the axle suffered in high speed operation and the non-linearity problems in axle crack identification, this project intends to investigate the online quantitative identification method of the crack in high speed train axle. Firstly, the dynamic model of the wheelsets system with a cracked axle under bending-torsional coupling excitation is established, and the dynamic response of which under high speed operation is studied. Secondly, the surrogate model of the crack parameters and the fault characteristics of the cracked axle is built, which will be used to identify the crack locations, depths, angles and other parameters quantificationally combined with an optimized algorithm. Finally, the residual life of the high speed train axle is predicted efficiently and accurately, and experiments are carried out to verify the above problems. The research results can not only provide an effective online monitoring method of axle crack, improve crack fault diagnosis theory and promote its practical progress, but also have great significance on preventing the accidents resulted from high-speed train axle fracture and ensuring the running safety of high-speed train.

车轴是列车转向架轮对系统中重要的部件之一，其质量状态直接影响高速铁路的运输安全。尤其在高速运行的高铁系统中，车轴裂纹可能导致非常严重的事故，因此轮对系统车轴的结构健康监测极为重要。本项目拟充分考虑车轴高速运行过程中承受的弯曲和扭转载荷以及车轴裂纹识别的非线性问题，对高速列车车轴裂纹进行在线定量识别研究。首先，建立弯扭耦合激励下含裂纹车轴的轮对系统动力学模型，研究高速运行条件下含裂纹车轴的轮对系统动力学响应特性；其次，通过建立车轴裂纹参数和裂纹故障特征的代理模型，将代理模型与优化算法相结合实现对车轴裂纹位置、深度、角度等参数的实时定量识别；最后，对高速列车车轴剩余寿命进行动态预测，并对上述问题进行实验验证。本项目的研究成果不仅可以提供一种有效的车轴裂纹状态在线监测方法，完善轴系裂纹故障诊断理论并推动其实用化进程，而且对预防高速列车车轴断裂事故，确保高速列车行车安全具有重大意义。

车轴作为列车转向架轮对系统的核心部件之一，在高速列车运行过程中承受着复杂的弯扭耦合激励和轮轨随机激励。随着列车速度的提高，车轴应力循环频率加快，极易导致车轴疲劳裂纹的萌生，严重威胁高速列车运行的安全性和稳定性。因此，对高速列车车轴裂纹进行在线定量识别具有重要的研究意义。. 针对上述需求，本项目首先将高速列车轮对系统简化为双盘转子，建立了弯扭耦合激励下多裂纹耦合的转子系统动力学模型；构建了含裂纹-扁疤-不圆顺耦合故障的高速列车轮对系统有限元模型。研究了裂纹参数及故障类型对含裂纹转子和轮对系统动力学特性的影响，分析了系统的振动响应、轴心轨迹和频域特征，并结合全谱与短时全谱方法总结了不同裂纹参数及故障下转子和轮对系统动力学响应规律。基于VMD、VMD-ICA等方法提取了不同故障下的系统振动特征，结合LightGBM、CNN等网络模型，实现了转子系统12种故障的分类。构建了裂纹参数和谐波响应特征间的Kriging代理模型，将裂纹参数识别问题转化为多目标函数的寻优问题，最后基于改进的NSGA-III完成了转子和轮对车轴裂纹位置、深度、角度参数的定量识别。设计了含裂纹转轴的加速寿命实验，构建了多尺度可分离卷积网络（MSSCN）模型和移动融合门控循环单元（MFGRU）模型，实现了含裂纹转轴剩余寿命的动态预测。. 研究结果表明，本项目建立的多裂纹耦合的转子系统模型和含裂纹-扁疤-不圆顺耦合故障的高速列车轮对系统模型能准确反映真实系统的动力学规律。在信噪比为20dB~-20dB的环境下，结合轴心轨迹相位特征、参数优化的VMD及LightGBM模型的多故障分类方法对转子系统12种故障的平均分类精度大于93.0%。基于Kriging代理模型与改进的NSGA-III算法对转子和轮对裂纹参数识别的平均误差均低于3.16%。建立的MSSCN模型和MFGRU模型对含裂纹转轴剩余寿命预测的均方根误差分别小于4.74和3.17。本项目研究成果可为轮对系统动力学特性分析、多故障分类、裂纹参数识别和剩余寿命预测提供理论依据，具有重要的工程适用价值。