DeepX and Doosan Robotics Joint Development of AI Welding Solution for Nuclear Power Plants
Published2026.10.06 14:26The Korea Institute of Industrial Technology Planning and Evaluation (KEIT) is supporting the project, and staged demonstrations will be conducted at Doosan Enerbility's nuclear power plant equipment manufacturing sites.
Key tasks include development of collaborative robot welding cells, micro-displacement compensation AI, and real-time welding machine control technology.
DeepX will develop domestic NPU-based on-device edge AI devices and software development kits (SDK), and will be responsible for building three-channel vision data pipelines, developing thermal imaging analysis AI models, and validating environmental durability.
SAIGE will develop 3D vision-based welding target recognition and welding path correction, as well as welding bead anomaly detection AI models, while Changwon National University's Industry-Academic Cooperation Division will handle AI-based welding path generation and physics-based digital twin technology.
DeepX stated that this project "will contribute to establishing an AI computing environment capable of efficiently processing AI operations required at manufacturing sites based on NPU and on-device AI technology and reliably operating."
A Doosan Robotics official said, "We will expand automation and intelligence of welding processes by combining collaborative robot technology with the AI technology of participating institutions."
Comments
Related news
DeepX Expands Integrations with YOLO, PaddlePaddle, and Raspberry Pi… Pushing to Build an Open Physical AI Ecosystem
DeepX, an on-device AI semiconductor company, has unveiled an open physical AI platform strategy that connects its NPU with global developer ecosystems including YOLO, PaddlePaddle, and Raspberry Pi. By integrating a unified execution environment spanning from the development phase through industrial mass production in a single architecture, the platform enables developers and enterprises to more easily implement AI in robotics, intelligent cameras, smart factories, and other applications.
2026.07.14by 배종인 기자
.jpg)
DeepX, KT, and Cesol Join Hands for NPU-Based On-Device AIoT Business
DeepX is collaborating with KT and Sesol to co-develop an AI Edge Box equipped with the ultra-low-power NPU chip DX-M1. As an on-device AI solution that significantly reduces power consumption and heat generation compared to GPUs, it will provide real-time video analysis services for EV charging stations, mobile CCTVs, and traffic infrastructure. The initiative aims to expand the edge AI ecosystem capable of real-time processing without cloud connectivity in industrial environments.
2026.07.31by 배종인 기자
DEEPX NPU Applied to Air Force AI Perimeter Surveillance System Construction Project
DEEPX's self-developed NPU (Neural Processing Unit) is being applied to the Air Force base AI perimeter surveillance system construction project. The V-NPU technology processes video decoding and AI inference in an integrated manner using only domestic NPU without foreign GPUs, enabling real-time analysis of CCTV footage and detection of anomalous behavior. As the first commercialization case of on-device AI semiconductors in the defense sector, it is evaluated as an opportunity to expand the domestic AI semiconductor ecosystem.
2026.08.05by 배종인 기자
DeepX Breaks Through 77 Global Orders in First Year of Domestic NPU Mass Production…Secures Cumulative Orders of $13 Million
Domestic NPU company DeepX has secured 77 global orders worth over $13 million from 8 industries across 9 countries within just one year of mass-producing its on-device AI semiconductor 'DX-M1'. Commercialization is expanding across physical and edge AI applications including robotics, smart factories, medical, and aviation, with more than half of total sales being generated overseas.
2026.08.14by 배종인 기자
DEEPX NPU Designated as 'Excellence in Research and Development Innovation Product' by Ministry of Science and ICT
DEEPX's NPU-based AI semiconductor 'DX-M1·DX-H1 series' has been designated as an Excellence in Research and Development Innovation Product by the Ministry of Science and ICT. As a low-power edge AI chip that reduces power consumption by up to 93% compared to NVIDIA GPUs, it will be eligible for government procurement contracts for three years. Within one year of mass production, the company has secured purchase orders exceeding 13 million dollars from over 10 countries including the United States, China, Taiwan, and Europe, and has delivery records for the Defense Ministry Air Force base perimeter surveillance system.
2026.08.19by 배종인 기자
