基于多目标优化的扁挤压筒结构设计  

Structural Design of Flat Extrusion Container Based on Multi-objective Optimization

作　　者：冯秋红[1] 刘全坤[1] 胡龙飞[1] 

机构地区：[1]合肥工业大学,合肥230009

出　　处：《中国机械工程》2006年第17期1850-1853,1857,共5页China Mechanical Engineering

基　　金：高等学校博士学科点专项科研基金资助项目(20030359002)

摘　　要：在组合式扁挤压筒的结构尺寸设计中,为确保挤压筒最佳工作性能的同时,最大程度地减少过盈装配后内腔产生的变形,提出了多目标优化的概念。结合有限元模拟技术和BP神经网络方法,建立了变过盈量下三层组合式扁挤压筒结构尺寸与各层等效应力分布、内腔位移之间的非线性映射模型,采用多目标遗传算法对其进行优化。优化时,采用了向量评价法、最佳个体保存策略和小生境技术,得到了均匀分布的Pareto最优解,根据定义的满意度函数,选出了最终的满意解。结果表明,在该满意解下,扁挤压筒既实现了等强度设计,又保证了内腔的尺寸精度。In the design of structural sizes for a flat extrusion container, the concept of multi-objective optimization was developed in order to make full use of the potential of die material and greatly reducing the deformation of inner layer cavity. Finite-element simulation and BP neural network were combined together to build the nonlinear mapping relation among structural sizes of three-layer combined flat extrusion container and equivalent stress of each layer and displacement of inner cavity. Then multi-objective genetic algorithm was applied to optimize flat extrusion container sizes. Vector evaluated method, optimal solutions-retention strategy and niche technique were adopted in the optimizition process to obtain the uniformly distributed Pareto-optimal solutions. According to a defined satisfactory degree function, each solution was evaluated, and the satisfactory solution was selected. The results show that under the solution, the best performance of flat extrusion container and dimensional accuracy of inner layer cavity are both satisfied.

关 键 词：扁挤压筒 变过盈量 多目标优化 BP神经网络 遗传算法 

分 类 号：TG376.2[金属学及工艺—金属压力加工]