混合多目标遗传算法及其在热挤压模具型腔形状优化设计中的应用  

Hybrid Multi-objective Genetic Algorithm for an Optimal Die Profile in the Hot Extrusion Process

作　　者：邹琳[1] 夏巨谌[2] 

机构地区：[1]华南理工大学,广州510460 [2]华中科技大学塑性成形数值模拟及模具技术国家重点实验室,武汉430074

出　　处：《中国机械工程》2006年第21期2261-2265,共5页China Mechanical Engineering

基　　金：广东省自然科学基金资助项目--博士启动基金资助项目(05300256)

摘　　要：将复合形法引入遗传算法来反映决策者对多目标问题中各目标函数的偏好信息,提出一种新的结合复合形法的混合多目标遗传算法。针对热挤压模具型腔轮廓形状优化问题,结合刚-粘塑性有限元模拟和神经网络技术,利用三次样条函数插值来表达凹模型腔轮廓形状,以表面载荷沿凹模型腔轮廓表面均匀分布和挤压力最小为目标,建立了多目标优化的数学模型,对挤压模具型腔轮廓形状进行多目标优化设计,得到了最优的凹模形状。对几种不同模具凹模型腔采用MARC/AutoForge有限元软件进行数值模拟对比研究,结果表明,结合复合形法的多目标遗传优化算法是一种较好的模具型腔形状多目标优化设计方法,其优化结果是有效的和显著的。研究结果说明,通过优化型腔形状来提高模具寿命的效果十分显著。An optimal die profile via employing several technologies coupling a hybrid multi--objective genetic algorithm, a neural network, and the rigid--viscoplastic finite element method （FEM） had been obtained in the hot extrusion processing. The die profile of the hot extrusion was represented by a cubic--spline curve. The aim of the present study was at reducing deformation load, yielding more uniform surface--load distribution on die profile surface and increasing the die life by means of an optimum die profile. Furthermore, using the finite--element software, MARC/AutoForge, the numerical simulations of the several different optimum die profiles under the same conditions were carried out. The results demonstrate the validity and effectiveness of hybrid multi--objective geneticalgorithm technique for the optimization of extrusion--die. The die with above optimal die profile can improve the die life in hot extrusion processes.

关 键 词：混合多目标遗传算法 热挤压 模具型腔形状 神经网络 刚-粘塑性有限元模拟 

分 类 号：TG375.4[金属学及工艺—金属压力加工] O241[理学—计算数学]