DeepX and Doosan Robotics said on Monday they will jointly develop an intelligent welding robot solution to be used in manufacturing nuclear power plant equipment. The project focuses on technology to make welding processes smarter by combining vision AI and collaborative robots. AI recognises welding targets and paths and corrects them, and it analyses vision and thermal imaging data collected during welding to judge weld conditions and whether abnormalities are present.
The two companies plan to implement these functions on a collaborative robot to deliver a solution that can be used at manufacturing sites. Doosan Robotics will lead the consortium and design the system integration architecture. It will also develop a collaborative-robot welding cell, micro-displacement correction AI and real-time welding machine control technology.
DeepX will build an on-device edge AI device based on a domestically made neural processing unit (NPU) and develop a software development kit (SDK). DeepX will also handle a three-channel vision data pipeline, linkage with robot controllers, development of thermal imaging analysis AI models, verification of durability in welding environments and support for reliability certification.
Sage will handle the work of recognising welding objects and correcting paths using 3D vision. The two companies and four organisations including Sage and the industry-academia cooperation foundation at Changwon National University signed a memorandum of understanding on July 21.
Sage will also develop an AI model to detect abnormalities in weld beads. The industry-academia cooperation foundation at Changwon National University will handle AI-based welding path generation and physics-based digital twin technology. The four organisations plan to integrate their technologies into a single welding system. During the project period, they will conduct step-by-step verification at Doosan Enerbility nuclear equipment manufacturing sites.
The schedule calls for completion of quality verification for nuclear equipment in 2029 and commercialisation in 2030. A DeepX official said, “This project will be a concrete application case of physical AI used at real manufacturing sites as ultra-low-power AI semiconductors combine with robots.” The official added, “Based on DeepX’s NPU and on-device AI technologies, we will contribute to building an AI computing environment that can efficiently handle AI computing required at manufacturing sites and operate stably.”
A Doosan Robotics official said, “Through this cooperation, we plan to develop an intelligent welding solution that can be used in real manufacturing sites by combining Doosan Robotics’ collaborative robot technology with AI technologies from participating organisations.” The official added, “We will contribute to solving problems faced by manufacturing sites such as shortages of skilled workers and to expanding automation and intelligence in welding processes.”
Leave a comment