基于并行NSGA-Ⅲ算法的高速列车车体侧墙结构高维多目标优化

Parallel NSGA-Ⅲ Based Multi-objective Optimization for Side Wall Section Size of High-speed Train Car-body

针对传统多目标优化算法在高维优化问题中多样性不足、收敛速度慢,联合仿真优化的计算效率难以满足实际工程需求等问题,提出基于SPMD并行的NSGA-Ⅲ联合仿真优化方法,由MATLAB优化器与FEM求解器构建联合仿真平台,通过调用多个求解器并行计算优化过程中的适应度函数,以提高面向复杂结构高维优化的计算效率。简单型材算例分析表明,在合理利用计算资源的情况下,提出的并行联合优化方法与传统串行方法相比,优化效率提高近一倍。以大型复杂机械结构高速列车车体侧墙为对象,建立多工况高维多目标优化问题,解析车体强度、刚度对侧墙结构不同区域型材厚度敏感性;对比分析Pareto解集,在侧墙质量减少0.96%的前提下,弯曲和扭转刚度分别增加了7.51%、5.39%,最大应力减少了12.18%。同时,相较于NSGA-Ⅱ算法,NSGA-Ⅲ算法可为车体结构优化设计提供更多符合期望的优化解集。

For many-objective optimization problems,traditional multi-objective optimization methods are limited due to insufficient diversity,slow convergence and computational cost in co-simulation,which are all critical issues that need to be taken into account in practical engineering.This study presents a SPMD parallel-based NSGA-Ⅲ co-simulation optimization method,for which the framework is constructed by MATLAB optimizer and FEM solver,respectively.To improve the efficiency for high-dimensional optimization problems,multiple solvers are employed to calculate the fitness function in parallel optimization.Analysis on a simple profile example shows that,the parallel optimization presented in this study improves approximately twice as efficiency as that of the traditional serial approach under the rational utilization of computing resources.Taking the side wall of a high-speed train car-body as an objective,high-dimensional optimization under multi-conditions is implemented to analyze the sensitivity of structural strength and stiffness to the profiles’thickness in different areas.Considering the Pareto solution set,the total mass of the side wall is reduced by 0.96%,while the bending and torsional stiffness are increased by 7.51%and 5.39%respectively,and the maximum stress is reduced by 12.18%.Compared with NSGA-II,NSGA-Ⅲ can provide more solution sets that can meet the expectations for the optimization design of vehicle car-body structure.