Fraunhofer Institute for Digital Media Technology IDMT
Acoustic AI for continuous inline monitoring for weld seam inspection
Initial situation for MAG welding
In industrial production using MAG robotic welding, the quality of the weld seam is a critical success factor. Irregularities such as weld spatter, pores, cracks or insufficient weld penetration lead to high reworking effort, rejects and increased costs. Currently, quality control is usually carried out on a random basis, visually or using destructive testing methods - but these are complex, expensive, time-consuming and carry the risk of unnoticed defects. Existing inline inspection systems (optical, structure-borne sound) are technically limited, expensive or not transferable to different applications. A fully automated, robust and intelligent solution for process monitoring in real time is currently not available.
Sizzle - sissle - siscle: non-destructive weld seam testing thanks to acoustic AI
Fraunhofer IDMT is developing an AI-based acoustic system for the inline monitoring of the welding process together with its project partners. The airborne sound signal generated during MAG robot welding - the distinctive hissing of the arc - is analyzed in near real time. Due to the constant hissing accompanied by crackling sounds, the acoustic fingerprint of the arc is so distinctive that it can be used for pattern matching. Machine learning methods are used to detect irregularities in the process from the recorded audio data, enabling qualitative conclusions to be drawn about the weld seam quality – contact-free and without a line of sight between the sensor and the area to be monitored.
Innovation in MAG welding: monitoring weld seam quality in the process
The system enables fully automated, 100% inline inspection of MAG robot welding processes. This allows quality deviations to be detected and corrected immediately, while the process is still running. This significantly reduces rejects and reworking, saving costs and freeing up production capacity. Non-destructive weld seam testing conserves valuable resources and prevents material losses. At the same time, continuous monitoring increases process reliability and the reproducibility of welding results, regardless of the operating personnel's experience.
Errors in the process can be identified at an early stage, contributing to higher system availability and reduced downtime. Therefore, the new technology for welding robots increases the overall efficiency of production while maintaining the same level of personnel costs. Overall, this solution provides a smart, economical and robust way to ensure consistently high welding quality.