머신러닝을 이용한 Al/Cu 레이저 용접부의 강도예측 모델링

Abstract

In these days, the importance of EV battery stability has increased due to strict environmental regulations. Al and Cu, which have high reflectivity and low absorption during welding, are applied to the battery cell of an electric vehicle. In addition, it is difficult to obtain a high-quality joint due to the intermetallic compounds (IMCs) generated in the Al/Cu welding joint. It is necessary to derive a process variable road map through strength prediction modeling using machine learning.

In this study, machine learning models and shallow neural network models were developed by using 135 experimental welding data which were constructed for fiber laser with a maximum output of 2.5 kW for Al1050H and Cu1020-HO materials with a thickness of 0.5 mm. The machine learning model Gaussian process regression (GPR) could be accurately predicted tensile-share load and penetration depth with the decision factor R2 of 0.97, 0.95, respectively. The shallow neural network model improved the accuracy to R2 of 0.99, 0.98. It was confirmed that accurate prediction was possible as a result of comparison and evaluation with decision factor of optimized machine learning and shallow neural network model.