针对第3代非支配排序遗传算法 (NSGA-III)求解离散拓扑优化收敛性差的问题，提出一种改进的NSGA-III (mNSGA-III)，用于某白车身焊装生产平台的结构优化．提出新型极值点选择算法，以稳定种群中个体的归一化过程．建立该生产平台的有限元模型，并以其总质量、最大应力和多个节点的z向位移最小为目标，进行离散拓扑优化．利用MATLAB集成MSC.Nastran软件，二次开发离散拓扑优化程序．通过优化、筛选，获得合理的杆件布置设计方案，使得结构总质量降低了30.1%，且刚度和强度均符合标准．优化结果表明，mNSGA-III 在求解高维多目标离散拓扑优化方面，具有计算过程稳定、收敛速度较快等优点，为大型钢结构的优化设计提供了新方法，具备工程实际应用价值．

This paper proposes a modified third generation non-dominated sorting genetic algorithm (mNSGA-III) to overcome the poor convergence of third generation non-dominated sorting genetic algorithm (NSGA-III) in handling discrete topology optimization. It uses the mNSGA-III for the structural optimization of a body-in-white (BIW) welding production platform. It proposes an advanced extreme point selection to stabilize the normalization of populations. It constructs the finite element model of BIW welding production platform. Using discrete topology optimization, it treats the total mass, maximum stress and z-direction displacements of several nodes of platform as objective functions. It developed a discrete topology optimization program by using MATLAB interfaced the commercial finite element code MSC.Nastran. Finally, it selected the design with appropriate layout in view of stiffness and strength of the structure. The optimal design conforms to the design standards and the total mass reduces by 30.1%. The results show that mNSGA-III gets a more stable optimization process and easy to converge when solving the multi-objective discrete topology optimization problems. The proposed method provides an applicable method for the optimization of giant steel structures and has great values for practical engineering problem.