South KoreaManufacturing2019 – ongoingPOSCO
AI-controlled smart blast furnace
POSCO's deep-learning control of a blast furnace added 240 tonnes of daily output
Problem
Conditions inside a blast furnace cannot be observed directly, and operators adjusted it largely from experience built up over decades. Fuel use and daily output depended on how well those judgements were made.
Approach
POSCO built a deep-learning system that automatically controls the charging of materials into the No. 2 blast furnace at Pohang Works, running on PosFrame, its in-house smart factory platform. The World Economic Forum named POSCO a "lighthouse" factory in July 2019, the first in Korea. In 2020 its research institute began applying explainable AI to the No. 2 and No. 3 furnaces with a KAIST team, so that operators could see which factors drove each prediction.
Results
Takeaway
All published gains come from POSCO itself, and the 2024 figures do not say which furnaces or period they cover. The company treated operators' trust as a separate problem from accuracy and added explainable AI after the control system was already running.
Categories
- Size
- Large enterprise
- Industry
- Manufacturing
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
Dailian reported on the Korean president's visit to POSCO's Pohang works, then Korea's only WEF lighthouse factory. It describes the PosFrame platform and the deep-learning No. 2 blast furnace, which cut fuel input by 4 kg per tonne and raised daily output by 240 tonnes. It also lists POSCO's wider smart factory savings and its support programme for smaller suppliers.
The Asia Business Daily reported that deep-learning AI automatically controls the No. 2 furnace's burden, cutting fuel use by 4 kg per tonne of hot metal and raising output by 240 tonnes a day. PosFrame structures decades of plant data to set process conditions in real time.
Kyongbuk Maeil reported that POSCO's research institute, working with a KAIST team led by Professor Choi Jae-sik, was applying explainable AI to the No. 2 and No. 3 blast furnaces at Pohang. The stated aim was to show the main causes behind AI results and so raise trust in the system. No results for the XAI work were reported.
The Korea Times reported that the World Economic Forum added POSCO to its network of lighthouse factories, the first Korean company named. The WEF described the plant as using AI to raise productivity and quality, and POSCO submitted its AI furnace technology.
A POSCO Holdings AI researcher said at a forum that blast furnace operation guided by five predictive models and AI raised productivity by about 5% and cut cost and CO2 emissions by about 1% each, compared with before. He did not state the period or the furnaces covered.
Full details are in the original sources.