Expulsion of Resistance Spot Welding Recognition and Diagnosis Based on PSO-SVM-SHAP

Expulsion is a common defect in resistance spot welding, and expulsion occurring along the interface of joined plates can significantly compromise the quality of the weld spot. To address welding expulsion issues caused by factors such as fluctuating network voltage, unstable air pressure, and non-standard worker operations during production, this paper proposes a solution based on the Shapley additive explanation (SHAP) and particle swarm optimization support vector machine (PSO-SVM) classification method. The aim is to construct expulsion cause identification system for resistance spot welding using electrode displacement signals. By simulating three typical scenarios that lead to expulsion—excessive current, insufficient pressure, and unclean surfaces—the corresponding electrode displacement data were collected. Based on the electrode displacement data under abnormal conditions, seven electrode displacement signal features were manually extracted, and a feature set of electrode displacement signals was constructed. Subsequently, SHAP values were employed to select the optimal feature subset, and the model was trained and tested using a SVM optimized by the PSO algorithm. Experimental results indicate that the classification accuracy of the test set reaches 92.85%. Furthermore, this paper employs macro-average and micro-average precision, recall, and F1 scores as model evaluation criteria and conducts a comparative analysis with other common classification models. The findings reveal that only the PSO-SVM model surpasses 90% in both macro-average and micro-average metrics, with minimal differences in various indicators between the training and test sets, thus demonstrating superior generalization ability.