Synergizing melt pool geometry metrics and acoustic analysis for enhanced defect detection in GMAW and WAAM
This research investigates the integration of melt pool geometry metrics with acoustic analysis to improve defect detection in Gas Metal Arc Welding (GMAW) and Wire Arc Additive Manufacturing (WAAM). By employing the near-infrared spectrum through the Xiris XVC-750 weld camera, we enhance the visibility of the GMAW weld pool, facilitating the use of advanced machine learning tools such as Xiris MeltPool AIô for real-time weld pool recognition, segmentation, and measurement. The resulting segmentation mask provides multiple geometric features that are sensitive to variations in weld bead geometry, contamination, and shielding gas deficiencies. Simultaneously, changes in weld sound, captured using Xiris WeldMicô and analyzed with Audio AIô software, are examined through audio feature extraction methods, including spectrum analysis and Mel-frequency cepstral coefficients (MFCCs). This study offers a comparative evaluation of the effectiveness of these individual methods and their combined use in detecting weld defects and geometric anomalies. The fusion of melt pool geometry metrics with acoustic analysis shows significant promise for advancing defect detection in welding and additive manufacturing processes.