Yaskawa's MOTOMAN NEXT uses AI to mimic skilled welders

The welding industry is facing a worsening shortage of skilled workers, and the article notes that tasks like shipbuilding, bridge work, and heavy construction equipment have large weld beads and have depended on the sensory judgment of experienced welders. Conventionally, human welders adjust parameters while checking the weld pool according to the shape of the joint. The demonstration is positioned as an attempt to reproduce that kind of optimal control, which typically requires human judgment, with a robot.

The robot used in the demo, MOTOMAN NEXT-NHC12, has a camera on its arm that captures the weld pool in real time. The captured footage is sent to a welding AI model in the controller unit, and the system identifies the weld pool to correct the robot's movement and speed so it follows the designated line. Yaskawa says this allowed the robot to reproduce, in places, the subtle adjustments that skilled workers make by hand according to the state of the weld pool.

Yaskawa is not stopping at the welding itself; it is also working to connect pre-weld inspection data and other upstream processes, and to feed the obtained welding results back to the AI model. Whether that feedback loop leads to a more stable welding system will depend on how well the data links together and how reliably the AI model handles variations, but if it works it could reduce the reliance on experienced workers that the article highlights as a growing challenge.