多点成形各向异性回弹预测方法  

Prediction method of anisotropic springback for multi-point forming

作　　者：李明 陈传东 常祥 LI Ming;CHEN Chuandong;CHANG Xiang(Roll Forging Research Institute,Jilin University(Nanling Campus),Changchun 130025,China)

机构地区：[1]吉林大学南岭校区辊锻工艺研究所,吉林长春130025

出　　处：《模具工业》2024年第11期1-8,共8页Die & Mould Industry

摘　　要：铝合金板材具有较强的各向异性,其在多点成形工艺中回弹难以预测,模具调型准确度不高,Yld2004-18p屈服准则可以对铝合金的各向异性进行准确描述,但是参数难以确定。以Al2024板材为例,对比基于BP神经网络模型和LM寻优算法,对Yld2004-18p进行参数识别。在回弹预测模型中,分别使用这2种方法获取参数并与多点成形试验结果进行比较。结果表明,使用BP神经网络模型确定的参数具有更好的精度。Due to the strong anisotropy of Alalloy sheet,its springback in the multi-point forming process was difficult to predict and the profiles of multi-point die was hard to adjust,the Yld2004-18p yield criterion could accurately describe the anisotropy of aluminum alloys,but the parame-ters were difficult to determine.Taking Al2024 sheet as an example,the parameter identification of Yld2004-18p based on BP neural network model and LM optimization algorithm was compared.In the springback prediction model,the parameters were obtained by using these two methods re-spectively and compared with the results of multi-point forming test.The results showed that the parameters determined by BP neural network model had better accuracy.

关 键 词：BP神经网络模型 参数识别 屈服准则 多点成形 

分 类 号：TG76[金属学及工艺—刀具与模具] O242.21[理学—计算数学]