Acoustic monitoring of drawn arc stud welding

Drawn arc stud welding is widely used in structural and industrial applications. Arc blow, the magnetic deflection of the arc, can disrupt bead geometry and stud penetration. Current quality assurance methods, such as bending tests, X-ray inspection, or ultrasonic scanning, are performed after welding and therefore cannot provide in-situ feedback. This work examines whether microphones can be utilized to inspect the quality of arc stud welding in real-time. Experiments were conducted with the mass clamp (ground) in varying positions to systematically induce different arc blow conditions. The intensity and direction of arc blow were inferred from post-weld bead geometry measurements, establishing classification labels for model training. Acoustic data were captured using microphones covering the audible to ultrasonic frequency range. Preliminary results using convolutional neural networks (CNNs) to classify arc blow intensity levels based on spectrograms indicate that acoustic monitoring could provide a non-invasive diagnostic tool for arc blow detection and quality assessment during stud welding.