Mapping relationship analysis of welding assembly properties for thin-walled parts with finite element and machine learning algorithm  

基于有限元和机器学习算法的薄壁件焊接装配特性映射关系分析

作　　者：Pan Minghui Liao Wenhe Xing Yan Tang Wencheng 潘明辉;廖文和;幸研;汤文成(南京理工大学机械工程学院,南京210094;南京理工大学数控成形技术与装备国家地方联合工程实验室,南京210094;东南大学机械工程学院,南京211189)

机构地区：[1]School of Mechanical Engineering, Nanjing University of Science and Technology, Nanjing 210094, China [2]Digital Forming Technology and Equipment National-Local United Engineering Laboratory, Nanjing University of Science and Technology, Nanjing 210094, China [3]School of Mechanical Engineering, Southeast University, Nanjing 211189, China

出　　处：《Journal of Southeast University(English Edition)》2022年第2期126-136,共11页东南大学学报（英文版）

基　　金：The Natural Science Foundation of Jiangsu Province,China(No.BK20200470);China Postdoctoral Science Foundation(No.2021M691595);Innovation and Entrepreneurship Plan Talent Program of Jiangsu Province(No.AD99002).

摘　　要：The finite element(FE)-based simulation of welding characteristics was carried out to explore the relationship among welding assembly properties for the parallel T-shaped thin-walled parts of an antenna structure.The effects of welding direction,clamping,fixture release time,fixed constraints,and welding sequences on these properties were analyzed,and the mapping relationship among welding characteristics was thoroughly examined.Different machine learning algorithms,including the generalized regression neural network(GRNN),wavelet neural network(WNN),and fuzzy neural network(FNN),are used to predict the multiple welding properties of thin-walled parts to mirror their variation trend and verify the correctness of the mapping relationship.Compared with those from GRNN and WNN,the maximum mean relative errors for the predicted values of deformation,temperature,and residual stress with FNN were less than 4.8%,1.4%,and 4.4%,respectively.These results indicate that FNN generated the best predicted welding characteristics.Analysis under various welding conditions also shows a mapping relationship among welding deformation,temperature,and residual stress over a period of time.This finding further provides a paramount basis for the control of welding assembly errors of an antenna structure in the future.为了深入挖掘天线平行T形薄壁件结构焊接装配特性之间的关联映射关系,考虑焊接方向、焊接夹具夹持和释放时间、固定约束和焊接顺序等因素,采用有限元仿真进行焊接特性分析,揭示焊接特性之间的映射关系.同时,采用广义回归神经网络(GRNN)、小波神经网络(WNN)和模糊神经网络(FNN)等机器学习算法,预测薄壁件焊接的多重特性,以反映其变化趋势和映射关系的正确性.与广义回归神经网络和小波神经网络的预测结果相比,采用模糊神经网络方法所预测的焊接变形、温度和残余应力值的相对误差最大的均值分别小于4.8%、1.4%和4.4%.结果表明,采用模糊神经网络方法预测的焊接特性结果优于其他2种方法.此外,针对不同焊接工况下的变化分析结果亦表明,焊接变形、温度和残余应力之间在某一时间段确实存在相应的关联映射关系.

关 键 词：parallel T-shaped thin-walled parts welding assembly property finite element analysis mapping relationship machine learning algorithm 

分 类 号：TG457[金属学及工艺—焊接]