Acoustics as a guide to welding quality

Abstract

Welding, the process of melting materials with high heat and fusing them during cool-down, is a fabrication technique common across a wide range of industries including construction, automotive, aerospace. Effective monitoring of the welding process is still an open problem. Weld testing and quality control methods require trained experts and specialized equipment and - in the majority - rely on post-welding inspection for defects or irregularities. In this work, we explore the acoustic qualities of welding sounds and their reliability as an automated quality indicator of the welding process. We investigate the most reliable acoustic features for identifying the quality of welding in the presence of environmental interference and factory floor noise, and we show that the sound picked up by a microphone next to the welding station can be sufficient for identifying a bad weld.