Insights on Weld Quality using Unsupervised Learning: Clustering of MIG/MAG Process Signals

With the increasing complexity of gas metal arc welding (GMAW) processes, data-driven approaches for monitoring and understanding automated GMAW production lines are gaining increased prominence. In this work, welding process data recorded in a production environment is analysed using unsupervised learning methods. We describe a data processing pipeline for feature engineering and apply a state-of-the-art clustering method to gain more insights into the welding production process. The clustering results are compared with the results from the application of dimensional reduction techniques and discussed based on human-interpretable characteristics of the welding process.