Enhancing Welding Quality Assessment in Challenging Environments: A BiAGAN-Based Approach With Operation Result Scoring

Abstract

In recent years, the welding industry has witnessed the integration of numerous automation and intelligent technologies, driven by the challenging working conditions inherent in the welding process. Despite these advancements, particularly in specialized environments such as high-altitude and underwater welding, or in the production of low-volume, high-quality products, reliance solely on automation and smart technologies remains insufficient. The quality of welding is profoundly impacted by the skill and expertise of the personnel involved in the operation. Addressing this pivotal issue, our paper introduces a novel approach aimed at enhancing welding quality assessment in such demanding environments. Central to our proposal is the introduction of a quality scoring module integrated into a comprehensive welded box quality scoring system. This innovative system combines BiAGAN’s advanced quality inference model with operation result scoring, providing a robust framework for evaluating welding quality. To validate the efficacy of our approach, extensive offline comparisons were conducted, demonstrating remarkable results with an accuracy rate of 95% and a recall rate of 92%. Moreover, to ensure the reliability and applicability of our findings in real-world scenarios, we conducted online tests involving multiple operators, thereby further corroborating the effectiveness of our proposed methodology. Our study represents a significant advancement in welding quality assessment practices, offering a practical solution tailored to the challenges posed by diverse and demanding environments. By harnessing the capabilities of BiAGAN-based methodologies alongside operation result scoring, our approach holds promise for revolutionizing welding quality assessment, ultimately leading to improved productivity and product quality.

Keywords

• Fusion welding
• non-destructive detection
• generative adversarial networks
• subsequent anomalies
• anomaly detection