- Developing the 'Smart Solution for Coating Weight Control Based on AI'
- A representative successful case of POSCO-style industry-university cooperation
POSCO became the first manufacturer to introduce
artificial intelligence(AI) to production processes in an effort to transform
itself into a smart steel mill.
Recently, POSCO joined hands with the Steel Mills,
POSCO Technical Research Laboratories, and the Department of Systems Management
Engineering in Sungkyunkwan University(Prof. Jong-seok Lee) to develop the 'Smart
Solution for Coating Weight Control Based on AI,' and applied it in the field
This is the tremendous achievement of the industry-university
cooperation system in which POSCO has carried out thousands of entrusted projects
together with domestic universities and research institutes, such as POSCO Technical
Research Laboratories, POSTECH and the Research Institute of Industrial Science
The 'Smart Solution for Coating Weight Control
Based on AI' is a technology that can drastically reduce the deviation of coating
weight by using artificial intelligence to precisely control the Continuous
Galvanizing Line(CGL), i.e. the core technology for automotive steel sheet production.
In particular, this technology is an automation control technology that predicts
the coating weight in real time and accurately meets the target coating weight
by combining the coating weight production model of the artificial intelligence
technique with the control model of the optimization technique.
▶ POSCO succeeded in developing the 'Smart Solution for Coating Weight Control Based on AI' through a joint industry-university cooperation project, and applied it to POSCO 3CGL last January. Accordingly, as it is now possible to utilize artificial intelligence, hundreds of different types of data can be collected in real time, and target coating weights can be accurately predicted and controlled. A developer and a worker are monitoring the optimal coating weight predicted by artificial intelligence in the operating room.
Coating weight control is a highly-sophisticated technology for keeping the
thickness of the coating layer even while operating conditions change from time
to time at the request of customers, i.e. automakers. As the coating weight
used to be controlled manually, there must be deviation of quality depending
on the skill levels of workers, and a lot of expensive zinc was wasted inevitably.
As the plating process is now automatically controlled by artificial intelligence,
the quality of automotive coated steel can be improved, and the production cost
can be reduced thanks to the reduction of overcoating weight. In addition, as
automated operation reduced workload, it is expected that work efficiency and
productivity can be increased.
The automotive coated steel, POSCO's representative high-margin high-grade
steel, i.e. a WP(World Premium) product, is a high-grade product that requires
a highly-sophisticated technology. Only 20 out of 800 steel companies in the
world can produce it at this time. Last year, POSCO sold approximately 9 million
tons of automotive steel sheets, accounting for 10% of the global automotive
steel sheet market.
The development of the 'Smart Solution for Coating Weight Control Based on
AI' started when artificial intelligence was given increased attention with
the emergence of AlphaGo, POSCO, which had been preparing to take the lead in
smart solutions, e.g. construction of a model smart factory, since the inauguration
of POSCO CEO Ohjoon Kwon, came up with the idea of actively applying AI to industrial
Last June, POSCO Technical Research Laboratories identified the need for
coating weight control automation, and then heeded to the opinions of several
departments, such as operations, maintenance and EIC, and collected related
data. Then, it entrusted the development of the artificial intelligence coating
weight prediction model algorithm to Prof. Jong-seok Lee, Department of Systems
Management Engineering, Sungkyunkwan University, who is an expert in statistics,
data mining, machine learning and optimization methodologies, while Prof. Lee
collaborated with POSCO researchers to develop the coating weight prediction
program based on an understanding of the plating process.
The basic program was developed from last July over a period of about 2 months.
POSCO Technical Research Laboratories applied its operation know-how to the
developed artificial intelligence program, and added a program customized to
the field based on the control technology for smooth operation even if field
equipment and operating conditions are changed, completing the 'Smart Solution
for Coating Weight Control Based on AI.'
The core AI technology applied to coating weight control automation is the
self-learning method utilizing the big data deep learning* technique. This method
keeps the accuracy of control up-to-date as the AI program learns hundreds of
different types of data, generated in the plating process, in real time. It
can accurately predict and control coating weight through real-time self-learning
even in case of equipment replacement or change of operating conditions.
The completed coating weight control automation solution was applied to Gwangyang
Steel Works 3CGL on a trial basis for about 2 months from last November to December,
enhancing accuracy and stability. As a result, the deviation of coating weight
per m² was up to 7g in case of manual operation, but it was drastically
reduced to 0.5g in case of AI-based automated operation. Technical verification
of this solution was finished, and it has been actively applied to Gwangyang
3CGL since January 5.
POSCO is planning to apply this proven coating weight control automation
solution to other CGLs at home and abroad, and become a leader of automotive
coated steel technology in the global market, and actively introduce the artificial
intelligence technology to the production processes of other steel products,
while building smart factories that can lead the age of the 4th industrial revolution.