Data-Driven Monitoring of Process Stability in Resistance Spot Welding

Ensuring high process stability during Resistance Spot Welding (RSW) in the automotive industry is crucial for maintaining high-quality welds without destructive testing. This work proposes a data-driven approach to monitor process stability during RSW. Nine different methods for measuring the distance between two electric resistance curves are evaluated in terms of runtime and performance. Among these, measuring the distance between two curves in the latent space is introduced as a novel approach to assess process stability. Our findings show that deviations calculated using Time2Vec and wavelet transform are well-suited for distinguishing unstable from stable processes, enabling proactive quality control and targeted maintenance actions in automotive production lines. We evaluate our approach on a real-world data set containing over 3.7 million welds from 1 074 welding systems and 1567 material-thickness combinations collected from automotive production.