An Approach for Expulsion Predicting in Resistance Spot Welding

Abstract

Resistance Spot Welding (RSW) plays a pivotal role in the assembly of automotive body components. During the process, undesirable expulsions can occur, which compromise the quality of the welds and lead to cost-intensive manual rework. In the presented approach, we train Machine Learning (ML) and Deep Learning (DL) models to predict a probability for the occurrence of expulsions for future spot welds. Our approach is based on a real-world data set that stems from the dynamic and complex environment of a series production line. This, in contrast to laboratory data, ensures the applicability of the proposed method in an industrial setting. Our best-performing model is able to predict expulsion with an accuracy of 95.41%. This allows an adjustment of the process before the expulsion occurs, reducing rework, production costs, and time.