韩国推进制造数据库和Physical AI项目，POSCO与韩华海洋披露试点动作

As the aging of skilled workers puts the tacit knowledge of manufacturing floors at risk of disappearing, the government and companies are moving quickly to prepare countermeasures. The approach involves designating skilled technicians as master craftsmen to encourage their participation in transferring skills, and turning workers' movements and judgment into data to train artificial intelligence (AI) and robots. However, concerns remain that most projects are stuck at the data-building or early pilot stage, making them insufficient to keep pace with the speed of skills loss.
According to relevant ministries and industry sources on the 19th, the Ministry of Trade, Industry and Energy is pushing to build a "manufacturing data library" that can safely store and manage manufacturing data provided by companies and retrieve it for use when needed. It is a kind of "national manufacturing data library" that standardizes and stores process data provided by companies, allowing it to be reused for developing manufacturing AI models.
The library is also being considered to contain not only quantified data such as equipment operation records, temperature, pressure and product quality, but also field know-how such as work sequences, judgment of abnormal signs and responses to unexpected situations. The aim is to reduce the time and cost each company spends building data separately, and to establish a foundation for AI learning that can be commonly used by each industry. Considering concerns that companies' core process information could leak externally, methods such as jointly using AI models trained on the data instead of the original data are also being discussed.
The Ministry of Science and ICT is responsible for securing physical AI source technology that will connect the collected field data to the actual judgment and actions of robots. In June this year, the ministry launched the 34 billion won "Physical AI Leading Technology Development Project," pushing technology development centered on 10 domestic industry-academia-research teams. The core is to localize a "world model" that implements real-world physical laws and work environments so that robots can repeatedly learn tasks in virtual space before being deployed to actual manufacturing sites.
At the same time, as the retirement of skilled baby-boomer workers who have formed the foundation of Korea's manufacturing competitiveness accelerates, the Ministry of Trade, Industry and Energy has also taken action after the technology and know-how of small and medium-sized enterprise (SME) manufacturing floors faced the risk of being severed. Through this year's "Manufacturing Tacit Knowledge-Based AI Model Development Project," the ministry selected 30 tasks and is supporting the development of technology to convert the experience and judgment of skilled workers into AI training data before master craftsmen with tacit knowledge retire.
Through this year's "Smart Manufacturing Innovation Support Project," the Ministry of SMEs and Startups is supporting a "rapid commercialization of AI application products" program to solve SMEs' chronic labor shortages and common challenges by industry. In particular, for small enterprises lacking their own capacity for smart transformation, the ministry plans to help introduce manufacturing robots and further expand support for automating labor-intensive manual processes. In addition, it will promote AI transformation (AX) and create a sustainable manufacturing innovation ecosystem through support for developing "SME manufacturing-specialized multi-AI agents" that strengthen SME manufacturing's ability to handle non-standardized tasks such as craftsmen's know-how, and optimize processes and quality.
Earlier, in March this year, the Ministry of Science and ICT established a "Physical AI Validation Lab" at the Korea Advanced Institute of Science and Technology (KAIST) and unveiled an integrated physical AI platform developed with domestic technology. It connects factory equipment, robots, AI models and digital twins into a single platform so that technology can be verified on manufacturing sites. The government plans to build "data factories" for each of the 10 major industries, including semiconductors, automobiles and shipbuilding, to secure the task and behavior data needed for robot learning, and to develop a domestic robot foundation model based on it.
Major manufacturing companies are also preparing measures to draw skilled workers' participation. POSCO is designating field technicians with high technical proficiency and company contributions as "POSCO Master Craftsmen," providing them with the highest level of treatment and pushing to convert the tacit knowledge they have accumulated into digital assets. Rather than simply demanding that skilled workers provide know-how, the aim is to officially recognize the value of their skills and motivate transfer activities. Lee Sang-hwi, a group part leader in the refractory group at Gwangyang Steel Works who was selected as a POSCO Master Craftsman this year, is a representative example. Having joined the refractory department at Gwangyang Steel Works in 1988, he was in charge of maintenance, repair and technical improvement of refractory equipment for 38 years. In particular, he was recognized for improving the design and management technology of refractories that protect the interior of high-temperature steelmaking equipment such as blast furnaces and converters, more than doubling their lifespan compared to before, and improving the efficiency of repair work to cut costs. POSCO plans to pursue "tacit knowledge digital assetization" going forward, turning such skilled technicians' field know-how into data and linking it to the company's AX.
Attempts to transfer skilled workers' "gestures" to robots have also begun. Hanwha Ocean is pushing a pilot project to apply physical AI-based humanoids to shipbuilding sites together with Ndot Light and Airobot. The method involves training robots on motion data such as workers' movements and tool operation, then repeatedly training them on rough-terrain walking, obstacle avoidance and item transport in a virtual shipyard.
However, the compensation and rights system for skilled workers who provide tacit knowledge remains a challenge to be solved. Unless anxiety that AI will replace their jobs is resolved, it will be difficult to draw participation from the field. "Skilled workers should be recognized not as simple data providers but as field experts who train and validate AI," an industry official said. "Compensation and job security measures must go hand in hand to gain cooperation from the manufacturing floor."