基于ACO-BP NN的飞机零件橡皮囊成形起皱预测（英文）  

Wrinkling prediction of rubber forming for aircraft parts based on ACO-BP NN

作　　者：张凌云 刘国庆 朴小东 范作鹏 Ling-yun ZHANG;Guo-qing LIU;Xiao-dong PIAO;Zuo-peng FAN(Key Laboratory of Fundamental Science for National Defense of Aeronautical Digital Manufacturing Process, Shenyang Aerospace University,Shenyang 110136,China;Air Force University Aviation Training Base,Jinzhou 121000,China;AVIC Shenyang Aircraft Industrial (Group)Co.,Ltd.,Shenyang 110034,China)

机构地区：[1]沈阳航空航天大学航空制造工艺数字化国防重点学科实验室,沈阳110136 [2]空军航空大学飞行训练基地,辽宁锦州121000 [3]中航工业沈阳飞机工业(集团)有限责任公司,沈阳110034

出　　处：《机床与液压》2018年第24期48-55,共8页Machine Tool & Hydraulics

摘　　要：利用ABAQUS/Explicit有限元分析软件对2A12-O态铝合金橡皮囊成形凸曲线弯边零件起皱现象进行分析,借助正交试验工具,进行了橡皮囊成形数值模拟试验。通过分析材料性能参数和工艺参数对凸弯边橡皮囊成形零件的影响,获取了影响凸曲线弯边零件起皱的主效应参数,并在此试验基础上建立了基于蚁群神经网络的起皱预测模型。在利用大量试验数据对其训练之后,使用该模型预测橡皮囊成形零件的起皱情况,同时对2A12-O态铝合金板材橡皮囊成形凸弯边零件进行工艺试验验证,结果表明:该模型能够在研究新零件成形过程中快速获得最佳成形参数,并且可以提高效率的工业,且预测误差控制在5%以内,且满足工业应用标准。The ABAQUS/Explicit is used to analyze the wrinkling phenomenon of convex flange parts of 2A12-O aluminum alloy in rubber forming.Numerical simulation tests of rubber forming were carried out with orthogonal test tools.By analyzing the effect of material'performance parameters and process parameters on the forming parts of the convex flange,the main effect parameters affecting the wrinkling of the convex flange parts are obtained. Based on this experiment,a prediction model of wrinkling in BP neural network (BP NN)by the ant colony optimization (ACO)is established.After using many experimental data to train it,this model is applied by utilizing new data to predict the occur-rence of wrinkling in convex flange parts,and this model was applied to predict the wrinkles of parts during the rubber forming.At the same time,the process test of 2A12-O aluminum alloy sheet rubber forming convex flange parts is carried out.The results show that using this model can quickly obtain the optimal forming parameters when studying a new part forming,and it can be boosted the efficiency of its industrial.Last but not least,predictive errors are controlled within 5% and meets industrial application standards.

关 键 词：2A12铝合金 正交试验 蚁群神经网络 橡皮成形 起皱预测 

分 类 号：TG386[金属学及工艺—金属压力加工]