Vision-driven adaptive welding solutions for the top three challenges in welding fabrication

Abstract

With experience in more than over 100 robotic deployments in pipe prefabrication and a decade-long dedication to welding automation, we have pinpointed the key challenges, notably fit-up variation, tack adaptation, and live seam tracking. We engineered an innovative adaptive welding solution that integrates the perceptual and cognitive abilities of welders into articulated robots. This system dynamically responds to real-time welding scenarios, effectively tackling associated challenges. Unlike existing methods reliant on pre-scanning or laser readings before welding, our vision-based adaptive welding technology operates instantaneously, replicating the expertise of proficient human welders. The outcome is a consistent delivery of high-quality welds. Given the widespread advancement of AI, the heart of the adaptive welding system must skillfully manage diverse welding conditions, covering different joint preparations, types, positions, thicknesses, materials, and beyond. Addressing the necessity of training the AI core requires navigating through diverse practical challenges in deployments. Leveraging our expertise in deploying various methodologies, we ultimately provide an efficient solution for training the welding AI, primed for widespread deployment across high-mix low-volume applications. This solution incorporates a data tracing and monitoring platform across deployments, enhancing ERP (Enterprise Resource Planning) functionality, and providing insights into welding operations, historical performance analytics, and problem tracking with proactive improvements.