基于支持向量机的随机聚焦搜索算法优化冲压成形工艺  

Application of Stochastic Focusing Search Algorithm Based on SVM in Optimization of Sheet Metal Forming Process

作　　者：龙玲[1,2] 殷国富[1] 宋超[2] 彭必友[3] 

机构地区：[1]四川大学制造科学与工程学院,四川成都610065 [2]成都纺织高等专科学校机械工程与自动化系,四川成都611731 [3]西华大学材料科学与工程学院,四川成都610039

出　　处：《四川大学学报（工程科学版）》2012年第5期220-225,共6页Journal of Sichuan University (Engineering Science Edition)

基　　金：国家科技重大专项资助项目(Sk201201A26-01);国家自然科学基金资助项目(51075287)

摘　　要：针对板料冲压成形工艺优化问题,研究了一种新的优化设计方法。采用支持向量机(support vector ma-chine,SVM)构建工艺参数与成形质量之间的多元非线性回归函数模型,在此基础上将一种新的群集智能算法,即随机聚焦搜索(stochastic focusing search,SFS)算法应用于冲压成形工艺参数寻优,以达到优化成形质量的目的。结合盒形件拉深实验证明,SVM在小样本条件下学习后所构建的非线性拟合精度比神经网络具有优势,表明了SVM具有更好的泛化性能。在SVM模型基础上应用SFS算法对板料冲压成形的工艺参数进行优化,将优化后的工艺参数进行实验验证,结果表明可获得较好的成形质量,说明了该优化方法具有较好的精确度和有效性,有一定的工程实用价值。In order to optimize the stamping process of sheet metal forming,a new optimal design method was studied.The multiple non-linear regression function model between process parameters and forming quality wsa constructed using support vector machine（SVM）.Based on the model,a kind of swarm intelligence algorithm called stochastic focusing search（SFS） was applied to search the optimization parameters of stamping process and optimize the forming quality.It was proved with an example of box-shaped deep drawing workpiece that the fitting accuracy of the non-linear function constructed by SVM with small samples of numerical simulation experiments has a better advantage over neural networks,which shows that SVM has better generalization performance.Then the SFS algorithm was applied to optimize the stamping process parameters based on the SVM model.And the process parameters after the optimization by SFS were validated by finite element simulation and verification that good forming quality can be obtained by this method,which suggested that this optimization method has higher precision and effectiveness,and provides a competitive project practical ability.

关 键 词：工艺参数优化 SVM 随机聚焦搜索算法 数值模拟 

分 类 号：TG386[金属学及工艺—金属压力加工]