Abstract
Today, the demands for sustainability, safety, productivity, and quality are pressing production systems, and the industry is sparing no effort to meet them. This paradigm shift towards digitalization, automation, and interconnectivity is crucial for developing smarter, more efficient manufacturing processes, in the scope of trend scenarios of Advanced Manufacturing or Industry 4.0. In this context, robotics and AI offer great solutions, that can benefit pipeline construction projects. The very viability of the new pipelines projects is affected by productivity and costs. Also, there is a strong correlation between pipeline construction and environmental issues, as well as with worker (welder) safety issues. This paper focuses on developing an AI-based laser tracking system for root pass welding in pipes, utilizing a controlled short-circuiting GMAW technology. The study employed a dedicated 7-axis anthropomorphic robot integrated with a laser sensor. The primary goal was to develop a robust integration between the robot and the laser sensor, enabling the creation of a weld seam tracking algorithm to locate and adjust the torch path along the welding of the joint. This algorithm utilized tracking points, which describe the inflection points of a previously set groove geometry. Additionally, the laser sensor system provides joint dimension data, such as bottom and top width, height, and area, which are input into an AI system to adjust welding parameters in real-time (Adaptive Welding). The AI algorithm was trained with welding parameters from experiments varying root openings and welding positions. For each condition, parameters like the controlled short-circuit current waveform, wire feed speed, travel speed, and waving parameters were optimized. Final tests were conducted on pipe sections with a V-joint to evaluate the performance and system’s robustness for both algorithms path and weld parameters correction. The results demonstrated satisfactory performance: the path correction algorithm was robust, the system was capable to overcome significant geometrical variations of the root opening (1 to 4.5 mm) and pipe mismatch, and the AI algorithm performed well even in out-of-position welding, providing robustness and adequate quality to the root pass.
Keywords: Orbital welding, Machine Learning, Online trajectory control, Smart Manufacturing, Adaptive Welding Systems
Referência:
GALEAZZI, D.; SILVA, R. H. G.. Development of a laser tracking system based AI for automatic root pass welding in pipes. Proceedings of the 2024 14th International Pipeline Conference, IPC 2024. Calgary, Alberta, Canadá. September, 2024