AI-Based Signal Analysis and Quality Prediction in Gas Metal Arc Welding

Gas Metal Arc Welding (GMAW) is widely used in various industrial sectors and plays a key role in automated mass production. Ensuring consistently high weld quality is crucial, especially when welding robots are used. One essential aspect of this is effective quality assurance. Conventional quality control methods—such as visual inspection and destructive testing—are still commonly used. However, they are time-consuming, costly, and difficult to integrate into continuous production lines. These methods do not allow for real-time monitoring, making them unsuitable for modern, highly automated manufacturing environments. A new approach to weld quality assurance involves the use of advanced technologies based on process data. This includes artificial intelligence (AI) and machine learning (ML), which offer innovative solutions for non-destructive, real-time quality assessment. This paper presents an AI-based method that captures and analyzes process data such as current and voltage signals during the GMAW process. By collecting time-series signals, feature extraction can be performed using a Convolutional Neural Network (CNN). These extracted features provide valuable insights into weld quality and enable predictions of impending instabilities in the welding process. The results presented in this paper demonstrate that the developed method is capable of both detecting welding defects and anticipating emerging process instabilities. This opens up a range of possibilities for automated welding. On the one hand, real-time assessment of weld quality becomes possible, allowing adaptive adjustments to welding parameters during the process to stabilize it in case of detected irregularities.