并行遗传算法在模具型腔形状优化设计中的应用  

Extrusion Die Shape Design Optimization with Multipopulation Parallel Genetic Algorithm

作　　者：邹琳[1] 夏巨谌[1] 胡国安[1] 

机构地区：[1]华中科技大学

出　　处：《中国机械工程》2003年第24期2077-2080,共4页China Mechanical Engineering

基　　金：湖北省科技厅资助项目(991P0201)

摘　　要：针对挤压模具型腔形状的优化设计,提出了并行微观遗传优化方法,该方法采用多种群的并行微观遗传算法进行优化计算,利用BP神经网络的预测功能获得目标函数值,采用三次样条插值函数表达挤压模具型腔形状。训练BP神经网络模型的导师信号利用刚塑性有限元数值计算获得。以表面载荷沿凹模型腔轮廓表面均匀分布为目标,建立了优化数学模型,对挤压模具型腔轮廓形状进行了优化设计。采用有限元软件MARC/AutoForge对优化结果进行了有限元仿真,仿真结果验证了优化结果的有效性。A new approach to optimal design of thedie shape in extrusion was presented. Optimization of thedesign variables was conducted by a multi -- populationparallel genetic algorithm(RPGA), where the fitness val-ues were obtained on basis of a BP neural network. Thedie profile of extrusion was represented by a cubic -spline curve. The data which was utilized to train the BPneural network was evaluated by a FEM (Finite ElementMethod) analysis model. The object is to yield more uni-form surface -- load distribution on die profile Surface. The result is satisfactory and the approach is approved tobe effective. Moreover, this approach can be applied tothe determination of the die shapes that are optimal withregard to various objective functions.

关 键 词：并行微观遗传算法 BP神经网络 挤压模具 模具型腔形状 优化设计 人工智能 

分 类 号：TG375.4[金属学及工艺—金属压力加工] TP18[自动化与计算机技术—控制理论与控制工程]