基于AHP-BP铝机罩外板回弹分析应用  

Springback analysis and application of aluminum hood plate based on AHP-BP

作　　者：胡文治[1] 周海龙[1] 张亚岐[1] 穆传坤 阮楹妍[1] HU Wen-zhi, ZHOU Hai-long, ZHANG Ya-qi, MU Chuan-kun, RUAN Ying-yan(Technical Center, Dong Feng Motor Corporation, Wuhan 430058, China)

机构地区：[1]东风汽车公司技术中心

出　　处：《塑性工程学报》2018年第5期123-129,共7页Journal of Plasticity Engineering

摘　　要：为获取最优工艺参数组合,采用层次分析法并结合BP神经网络（AHP-BP）来确定铝板冲压工艺参数。首先按等深拉延、均匀变形原则,设计满足符合试验条件的工艺补充面;其次,采用层次分析法确定压边力、拉延筋直径以及摩擦系数的权值;通过正交试验获取多组分析结果,构建基于层次分析法BP网络模型,并依据网络模型预测值与CAE分析值的接近程度来确定最优的工艺参数组合,压边力、拉延筋直径以及摩擦系数分别为：135 t、20 mm、0. 2;最后,将AHP-BP网络输出的回弹值作为模具的修正的基准值,并加工实物进行验证。结果表明：将拉延工艺模型回弹修正量设为AHP-BP模型预测值的1. 2倍比较合适。In order to obtain the optimal process parameters,analytic hierarchy process and BP neural network（ AHP-BP） were adopted to determine the aluminum sheet stamping process parameters. Firstly,the addendum surface meeting the test conditions was designed according to the same drawing depth principle and uniform deformation principle. Secondly,the weight of blank-holder force,drawbead diameter and friction coefficient were determined by the analytic hierarchy process. Through orthogonal test,several groups of analysis result were obtained and the corresponding network model was established. According to the degree of proximity between the predicted value of the network model and the CAE analysis value,the optimal combination of process parameters is determined. The blank-holder force,drawbead diameter and friction coefficient are separately 135 t,20 mm,and 0. 2. Finally,the springback value of the AHP-BP network output was used as the base value of the mold correction,and verified through physical object. The results show that when the amendment of drawing springback process model is set to 1. 2 times,the predicted value of BP model is more appropriate.

关 键 词：铝板成形 回弹 层次分析法 BP神经网络 权值 修正 

分 类 号：TG386[金属学及工艺—金属压力加工]