NC AI, the artificial intelligence (AI) subsidiary of NC, is joining hands with Hanwha Ocean to develop an advanced robot AI brain needed for shipbuilding.
NC AI said on the 4th that it won Hanwha Ocean's project to develop a "vision recognition-based welding-dedicated model and a collaborative robot-based autonomous welding model." The project fuses advanced AI vision recognition and precise robot control technologies for welding, a core process in shipbuilding that has relied on the know-how of skilled workers.
The two companies aim to implement an "autonomous welding physical AI solution" in which robots recognize and judge weld areas on their own and perform optimal welding in real time, going beyond the limits of existing automation that merely repeats predetermined trajectories.
An NC AI official said, "The shipyard welding process is a key task that determines ship manufacturing costs and quality, but due to the nature of the work, it has extremely harsh conditions for vision recognition AI to operate, including intense arc light and sparks, welding fumes generated in real time, and camera lens contamination caused by rough field environments that include outdoor settings," and added, "To overcome these limits, NC AI is conducting on-site, hands-on research that reflects Hanwha Ocean's real worksite data and engineers' feedback in real time."
The plan is to complete "shipbuilding-specialized vision recognition technology" that precisely extracts geometric weld lines and detects welding defects in real time amid intense noise and contamination. The finalized autonomous welding model and robot system developed in this way will be applied to the construction process of next-generation commercial ships that Hanwha Ocean will build, as well as special-purpose ships that require high precision and security.
NC AI plans to use the recently released next-generation industry-specialized vision-language model (VLM) "VAETKI Vision" as the core engine of this Hanwha Ocean project. VAETKI Vision is a model that understands visual environmental information and text instructions, and based on this, NC AI plans to expand it into a "VLA model" that simultaneously processes vision, language, and action to control a robot's physical movements.
When a worker gives instructions by voice or text without complex coding, a collaborative robot uses VAETKI Vision to analyze in real time the shape of the welding target and the condition of the weld line, and then independently derives behavior control commands such as precise torch angles and speeds accordingly. This goes beyond simple hardware automation and is expected to be a definitive example of true "physical AI autonomous welding," where robots perceive their surroundings and respond flexibly on their own.
Lee Yeon-su, CEO of NC AI, said, "We will successfully develop robust vision recognition technology and an autonomous control model that overcome on-site dust and contamination, and create a world-class physical AI model that can be deployed in actual shipbuilding processes."