ANFIS MODELING IN PROJECTION WELDING OF NUTS TO SHEETS

Projection welding Adaptive neuro-fuzzy inference system, Genetic algorithm.ojection welding Adaptive neuro-fuzzy inference system, Genetic algorithm. this study, the maximum weld strength in projection welding of nuts to sheets (DD13 sheet metal part and AISI 1010 nut) was investigated using optimum input parameters (weld current, weld time and hold time) in order not to damage weld joint during the car assembly. The experimental strength tests were carried out for each one hundred and forty actual samples. The projection welding process was modeled using an Adaptive Neuro-Fuzzy Inference System (ANFIS). One hundred of these experimental data were used for training in order to model the physical system. Forty experimental data were randomly selected over one hundred and forty actual experimental data. The randomly selected experimental data were used to validate the developed model. The proposed model exhibits good agreements with test data. The developed model was also checked based on squared coefficient of correlation. After all good agreements and validations, Genetic Algorithm (GA) was used to obtain some intermediate values not found on the actual experimental data. Maximum output parameter (weld strength) was determined using these optimum intermediate input parameters. Maximum strength of the projection welding was increased by five percentages using these optimum intermediate input parameters. Finally, all open source Matlab codes and a Google drive links for all Matlab files related to modeling and optimization of the physical system were added to Appendices in order to make other researchers’ works easy. Besides, these Matlab codes and files are applicable for any physical system with very small changes in codding.

本研究围绕螺母与板材（DD13钣金件及AISI 1010螺母）的投影焊接工艺展开，以避免汽车装配过程中损伤焊接接头为目标，通过优化焊接电流、焊接时间与保压时间三类输入参数，对最大焊缝强度进行了研究。本研究共完成140组实际试样的实验强度测试。采用自适应神经模糊推理系统（Adaptive Neuro-Fuzzy Inference System, ANFIS）对投影焊接过程进行建模：从140组实际实验数据中随机选取100组用于训练以构建物理系统模型，剩余40组数据用于验证所开发的模型。所提出的模型与实验测试数据展现出良好的拟合一致性。随后通过平方相关系数对所建立的模型进行了校验。在验证模型具备良好拟合效果后，采用遗传算法（Genetic Algorithm, GA）获取实际实验数据中未涵盖的中间参数值，并基于优化后的中间输入参数确定了投影焊的最大焊缝强度。相较于原工艺，采用该优化中间输入参数可使投影焊的最大焊缝强度提升5%。最后，本研究将所有开源Matlab代码及相关建模与优化的Matlab文件的Google云端硬盘链接整理至附录中，以方便其他研究人员复用；此外，仅需对代码进行少量修改，该套代码即可适用于其他物理系统。