Expulsion Prediction in Resistance Spot Welding – Process Optimization with Explainability Methods

Resistance Spot Welding (RSW) is a key joining technology in large automotive body shops. During the process, undesired expulsion (the eruption of molten material) can occur. We detect expulsion based on changes in the Dynamic Electric Resistance Curve (DERC) and train Machine Learning (ML) models to predict expulsion. In this work, we uncover case-dependent influencing factors that led to expulsion by leveraging explainability methods. This helps domain experts to optimize the process. We evaluate our approach with a real-world data set from a dynamic and complex environment out of a series production line.