基于网格变形与代理模型的横梁参数化建模与轻量化优化
Parametric modeling and lightweight optimization for cross beam of punch based on grid deformation and agent model

建立了包含导轨结合面的横梁有限元模型,并通过模态实验验证了有限元模型的正确性,在多种极限工况下对原横梁进行了静动态特性分析与评价;针对现有优化方法的不足,采用了一种新的优化设计方法:通过网格变形技术实现结构尺寸的参数化建模,基于拉丁超立方采样、相对灵敏度分析和Kriging近似模型建立横梁结构优化的代理模型以代替现有的有限元仿真模型;以体积最小为优化目标,以刚度、强度和动态性能不变为约束条件,采用遗传算法进行优化设计.优化结果表明:在保持横梁最大变形、静刚度以及一阶模态频率基本不变的情况下,横梁总质量减少了58 kg(减少了5.9%).
The finite element model for cross beam including guide joint surface is established, and the reliability of the model is validated through modal experiments. The static and dynamic analyses for the cross beam under different operating conditions are carried out. Due to the shortcomings of existing optimization methods, a new method is proposed. A parameterized model is set up through the mesh deformation technology. An agent model is built through Latin hypercube sampling, relative sensitivity analysis and the Kriging approximation model. Taking the volume as the goal, the stiffness, strength and the first order natural frequency as the constraint conditions, the lightweight design is produced by GA(genetic algorithms). The optimization results show that the mass of the cross beam can be reduced by 58 kg(5.9%), while the maximum deformation, strength and the first order natural frequency keep basically unchanged