Advanced process characterization and machine learning-based correlations between interdiffusion layer and expulsion in spot welding

Over the past decades, substantial endeavors have been dedicated to unraveling the intricacies inherent to Resistance Spot Welding (RSW). However, a comprehensive and consensual understanding of the RSW process physics is still lacking, including the exact number of physical phases behind the RSW process. For example, a widely accepted model indicates that metal only starts melting after the peak of dynamic resistance, while the latest research on welding uncoated materials challenges this by suggesting that melting begins around the resistance peak. Furthermore, most of existing physical models only consider welding materials without coatings in a controlled lab setting, whereas coated sheet metal is the norm in real production. Addressing these challenges, this paper introduces an enhanced model for RSW that considers the melting phase of the coating's InterDiffusion Layer (IDL) in Press Hardening Steels (PHS). This phase is believed to influence both welding quality and the occurrence of expulsions. Additionally, the timing at which each phase starts has been determined by analyzing real-time, multi-variable sensing data from various welding scenarios, and a signal processing technique has been devised to automatically identify when these phases begin. Leveraging this refined process understanding and characterization, meaningful explainable features are extracted, and a data-driven multilayer perceptron model is constructed for 1) predicting IDL thickness and 2) detecting expulsions upon predicted IDL thickness. The experimental results validate that the proposed IDL-inclusive model advances existing physical models for RSW and the IDL prediction improves the RSW defect detection and process monitoring.