基于近似模型的高速列车头型多目标优化设计

为改善高速列车气动性能，建立一套高效的多目标气动优化设计方法，对流线型头型进行多目标气动优化设计。建立高速列车流线型头型三维参数化模型，并提取5个优化设计变量；为减少优化设计时间，利用最优拉丁超立方设计方法在优化设计空间中进行均匀采样，利用计算流体力学方法获得对应于各个采样点的气动载荷，利用Kriging代理模型构建优化设计变量和气动载荷之间的近似模型；利用多体系统动力学方法计算气动载荷作用下的高速列车轮重减载率；以气动阻力和轮重减载率为优化目标，利用多目标遗传算法NSGA-II对高速列车流线型头型进行多目标优化。优化设计变量和优化目标均呈现收敛的趋势，采用Kriging近似模型优化计算的Pareto前沿与采用CFD（Computational fluid dynamics，CFD）优化计算的Pareto前沿较为接近。优化后高速列车的气动阻力最多可降低3.27%，轮重减载率最多可降低1.44%，气动阻力最优的头型与轮重减载率最优的头型的主要差异在于中部辅助控制线的变化，前者向内凹，后者则向外凸。

Multi-objective Optimization Design of the High-speed Train Head Based on the Approximate Model

In order to improve the aerodynamic performance of the high-speed train, an efficient multi-objective aerodynamic optimization design method is set up to carry out the multi-objective aerodynamic optimization design of the streamlined head. The three-dimensional parametric model of the streamlined head of the high-speed train is set up, and five optimization design variables are extracted. To reduce the optimization time, the optimal Latin hypercube design method is used for the uniform sampling in the optimization design space, and the aerodynamic loads corresponding to each sampling point are obtained through the computational fluid dynamic method. The Kriging surrogate model is used to construct the approximate model between optimization design variables and aerodynamic loads. The load reduction factor of the high-speed train caused by the aerodynamic loads is computed by the multi-body system dynamic method. Then the aerodynamic drag force and load reduction factor are set as optimization objectives and the multi-objective optimization of the high-speed train head is conducted by the multi-objective genetic algorithm NSGA-II. The optimization design variables and optimization objectives show the tendency of convergence. The Pareto frontier computed by the Kriging approximate model is close to that computed by the computational fluid dynamics (CFD). After optimization, the aerodynamic drag of the optimized train is reduced by up to 3.27%, and the load reduction factor is reduced by up to 1.44%. As for the optimal head with minimum aerodynamic drag force and the optimal head with minimum load reduction factor, the main difference is the deformation of the central auxiliary control line, with the former concave and the latter convex.