Quality prediction method for automotive body resistance spot welding based on digital twin technology

In the resistance spot welding (RSW) process of automotive bodies, accurately predicting the welding quality is of vital importance for ensuring the safety and reliability of vehicles. However, traditional prediction methods are limited by the constraints of on-site data collection, which poses challenges to the accuracy of predictions. Therefore, a new RSW quality prediction method based on digital twin (DT) is proposed. Firstly, virtual data of dynamic resistance is generated through the DT model. Then, the generative adversarial network (GAN) method is employed to expand the virtual data and physical data. Finally, the prediction model of RSW quality is constructed through a back propagation (BP) neural network and Bayesian optimization method. Finally, a case study is conducted to verify the effect of virtual data on the performance of the prediction model and demonstrate the effectiveness of the proposed method. In addition, the SHAP (Shapley additive explanations) method is used to investigate the impact of various input variables on RSW quality, thereby establishing a foundation for optimizing the RSW process.