基于磨耗回归的地铁小半径曲线钢轨廓型优化

Wear is the main rail damage pattern on most sharp curves in metro lines, which causes rail life reduces to half or even worse. Alleviation of rail wear on sharp curves can be achieved by optimizing rail profile to improve rail/wheel contact conditions. However, current rail profile design methods using one averaged wheel profile when considering the variety of wheel profile. Furthermore, the current methods assume unimodality of the objective function but its sufficient conditions are not satisfied due to the highly nonlinear behavior of the rail/wheel contact on curves. These problems cause the designed rail profile is not the best rail profile. This project proposes a rail profile design optimization method based on a rail wear regression algorithm and a genetic algorithm to overcome the above limitations, which aims at controlling the rail wear on sharp curves in metro lines. Rail wear is integrated into the optimization process directly and the problem of rail wear characterization under various wheel profiles was solved using regression theory. The combined impact from factors such as flange contact, non-uniform friction coefficient on rail head and the safety boundary on the objective function of rail profiles optimization is considered. Designed rail profile will be subjected to application in metro lines to verify its effects. Finally a more accurate rail profile design method will be established to help alleviating rail wear on metro sharp curves.

磨耗是地铁小半径曲线钢轨的主要损伤形式，导致钢轨寿命在小半径曲线上减少将近一半。通过优化钢轨廓型，改善轮轨匹配关系，能够减缓曲线钢轨磨耗。但现有钢轨廓型设计方法在考虑多种车轮廓型时实际采用的是匹配一种合成廓型；且现有方法隐含假设目标函数具有单峰性，而小半径曲线轮轨接触的高度非线性使得目标函数不具备单峰性的充分条件。上述问题导致现有方法设计出的并非最优钢轨廓型。本项目提出钢轨磨耗回归算法，采用遗传算法优化理论建立一种以控制曲线钢轨磨耗为目标的钢轨廓型设计优化方法。优化方法将钢轨磨耗直接整合到优化过程中，用回归理论解决实际线路上全部车轮廓型影响的钢轨磨耗表征问题；考虑轮缘接触、非均匀轨面摩擦系数和车辆运行安全指标边界等综合因素对钢轨廓型优化目标函数的共同影响；并通过地铁曲线钢轨修型现场实验进行验证。通过本项目研究，为小半径曲线钢轨廓型设计提供一种更准确方法，达到减缓曲线钢轨磨耗的目的。

磨耗作为地铁小半径曲线的钢轨主要损伤形式，是本项目开展钢轨廓型优化解决的实际问题。运用车辆轨道耦合系统动力学建立车辆-轨道动态相互作用模型，模拟车辆曲线的轮轨接触状态。采用能量耗散理论，建立钢轨磨耗累积分布模型，模拟轮轨多次接触下的钢轨累积磨耗损伤，实现钢轨磨耗的快速预测。运用了回归理论，对同一条地铁线路上运营的车轮廓型进行统计学上的划分，得到了不同服役里程的车轮廓型分类，在子类中选出廓型代表，模拟这类车轮廓型几何状态对钢轨磨耗的贡献，实现线路上全部车轮廓型对钢轨磨耗贡献的理论预测。在允许钢轨廓型优化目标多峰性的条件下，运用遗传算法构建了地铁小半径曲线钢轨廓型数值优化算法。在优化算法中，考虑了曲线钢轨高轨和低轨的几何形状非对称性，还发展了利用现场实测廓型的NURBS参数化描述方法，将钢轨廓型设计点的控制量区分为位置控制量和形状控制量，增加了廓型几何描述的多样性，一方面在算法上强化了获得全局最优解的可能性，另一方面还可以继承现场测试廓型的优良特性。与标准钢轨廓型相比，用所建立的小半径曲线钢轨廓型优化方法设计的非对称钢轨廓型，可以显著改善轮对曲线通过性能和降低轮轨磨耗。