China Mechanical Engineering ›› 2024, Vol. 35 ›› Issue (12): 2114-2121.DOI: 10.3969/j.issn.1004-132X.2024.12.003

Previous Articles     Next Articles

Machine Learning and Finite Element Simulation and Experimentation for Springback Prediction of Al-Li Alloys

HUI Shengmeng1;MAO Xiaobo4;ZHAN Lihua1,2,3   

  1. 1.Light Alloys Research Institute,Central South University,Changsha,410083
    2.School of Mechanical and Electrical Engineering,Central South University,Changsha,410083
    3.State Key Laboratory of Precision Manufacturing for Extreme Service Performance,Central South
    University,Changsha,410083
    4.AVIC Xian Aircraft Industry Group Company Ltd.,Xian,710089

  • Online:2024-12-25 Published:2025-01-13

铝锂合金回弹预测的机器学习及有限元仿真与实验

惠生猛1;毛晓博4;湛利华1,2,3   

  1. 1.中南大学轻合金研究院,长沙,410083
    2.中南大学机电工程学院,长沙,410083
    3.极端服役性能精准制造全国重点实验室,长沙,410083
    4.中航西安飞机工业集团股份有限公司,西安,710089

  • 作者简介:惠生猛,男,1997年生,博士研究生。研究方向为铝锂合金电脉冲蠕变时效成形。E-mail:223801010@csu.edu.cn。
  • 基金资助:
    国家自然科学基金(U22A20190,52175373,52005516);湖南省科技创新计划(2020RC4001)

Abstract: Creep aging tests were conducted on the 2195 Al-Li alloys under various stress conditions at temperatures of 180 ℃, 190 ℃, and 200 ℃ respectively. Constitutive equations were derived using MATLAB software and incorporated into the nonlinear finite element software MSC.Marc to build a finite element model for the creep aging forming of 2195 Al-Li alloy spade segments. The model utilized time, stress, and temperature as input parameters, with the springback radius being the critical output parameter. To enhance the accuracy and efficiency of predictions, a comparative analysis of various machine learning regression models was conducted, leading to the selection of the ridge regression model as the predictive tool, which facilitated the rapid and precise prediction of the springback radius under diverse processing conditions. The high predictive accuracy and practical utility of the model were validated through 1∶1 experimental verification, demonstrating a relative error of 0.9% between the experimental components springback profile and the target profile. 

Key words: Al-Li alloy, creep aging forming, machine learning, finite element simulation

摘要: 分别在180 ℃、190 ℃和200 ℃温度的不同应力条件下对2195铝锂合金进行蠕变时效试验,利用MATLAB软件拟合得到本构方程,并将本构方程整合到非线性有限元软件MSC.Marc中,构建了2195铝锂合金瓜瓣蠕变时效成形的有限元模型,模型以时间、应力和温度为输入参数,回弹半径为关键输出参数。为提高预测精度与效率,对比分析了多种机器学习回归模型,最终选定岭回归模型作为预测工具,实现了对不同工艺条件下回弹半径的快速准确预测。通过1∶1实验验证,实验构件回弹型面与目标型面的相对误差为0.9%,证明了模型的高预测精度和实用价值。

关键词: 铝锂合金, 蠕变时效成形, 机器学习, 有限元仿真

CLC Number: