Quality prediction of aluminum alloy resistance spot welding with shunting by introducing spacing factor

Abstract

Shunting is a very common phenomenon in the practical manufacturing of aluminum alloy resistance spot welding (RSW). Effective quality prediction method for RSW with shunting is still insufficient. In this study, after dimensionality reduction by principal component analysis (PCA), the ensemble learning model AdaBoostRegressor was used to predict the weld diameter under shunting conditions in aluminum alloy RSW. The prediction effects of various signals collected during the welding process on the welding quality were compared and discussed. After the introduction of the spacing factor, except for the welding voltage and dynamic resistance signals, the effects of other signals in predicting the welding quality have been significantly improved. The electrode displacement signal introduced with the spacing factor has the best effect on the quality prediction of welds among all the collected signals, and the coefficient R2 reaches 0.88. It is concluded that the electrode displacement signal is suitable for quality monitoring of aluminum alloy RSW under various conditions. Undersized welds can be quickly identified by calculating the work done by the weld spots to the upper electrode. Our study lays a foundation for the real-time quality monitoring in the process of aluminum alloy RSW.