SK Innovation Cuts Process Inspection Time by 98% Through AI Collaboration with Startup [AI Autonomous Manufacturing: Opening the Future]

AI reduces interpretation time by 98%
AI brings together large corporations and startups... Certified as "Korean-style New Technology"
AI collaboration trend spreads across the industry

"Abnormal detected. Please upload the data again."

Park Jaehan, Project Manager (PM) of the Mechanical Engineering Team at SK Innovation's Ulsan CLX plant, shouted this over the phone while staring at his laptop in the plant office. An abnormality had just been detected in the heat exchanger sensor data transmitted by a partner company. Artificial intelligence (AI) captured the abnormal pattern in real time, and the headquarters technical team immediately ordered a remeasurement. This is how the AI-based heat exchanger inspection system, which will be introduced at all SK Innovation plants starting in the second half of this year, works.

An oil refining company in the Ulsan region is demonstrating the 'AI-based Heat Exchanger Non-Destructive Inspection System' jointly developed by SK Innovation and DEEP AI. Partner company employees conduct the initial inspection on-site (left), and the headquarters' mechanical engineering team office makes the final determination on any heat exchanger abnormalities (right). DEEP AI

Managing heat exchangers, which control the temperature in refining and petrochemical processes, is crucial because high-temperature fluids flow through them. Even minor wear or cracks can cause fluid leaks, and in the worst case, explosions. This is why each heat exchanger undergoes 30 to 60 inspections during every regular maintenance period. There are about 7,000 units installed at SK Innovation's Ulsan complex alone, and about 30,000 units across the entire Ulsan petrochemical complex. The inspection process has been labor-intensive and heavily dependent on skilled workers, making it a risk factor in management. Now, however, AI will be responsible for real-time management at all plants.

AI reduces interpretation time by 98%

The most significant change brought by AI in manufacturing processes is the reduction in quality variation. Until now, heat exchanger inspections relied on the eyes of non-destructive testing inspectors. Each inspection required more than 10 minutes of focused attention as they checked the shape of the graphs one by one. Only when a skilled inspector's intuition caught a subtle 'fluctuation' in the signal would a remeasurement or defect determination be made. Much like a radiologist interpreting a CT scan, the judgment depended entirely on the inspector's experience. This led to differences in skill levels and a decline in interpretation quality.

The refining industry adopted AI to improve the precision of heat exchanger non-destructive testing and prevent major accidents in advance. SK Innovation assembled a core team called the 'Smart Plant Task Force,' and before moving to the Mechanical Engineering Team, PM Park was responsible for developing the AI-based inspection technology. With a background in mechanical engineering and coding skills, he collaborated with Ulsan-based startup DeepEye to create an AI solution tailored to on-site needs.

On the 11th, DEEP AI employees are holding a meeting at the research institute located in Gijang-gun, Busan, to improve the quality of the 'AI-based Heat Exchanger Non-Destructive Testing System.' Busan - Photo by Oh Jieun

PM Park explained, "Non-destructive testing is essential, but the interpretation requires a lot of time and concentration, so human error occurred frequently. By supplementing or replacing this process with AI, we can reduce the burden on inspectors while improving accuracy." After introducing AI, diagnostic accuracy increased to over 95%, and the interpretation time per inspection dropped to under one minute, reducing the total time required by more than 90%. The system also established a double-check process in which on-site inspectors conduct an initial AI-based interpretation, followed by a review by the headquarters technical team.

AI brings together large corporations and startups... Certified as 'Korean-style New Technology'

SK Innovation's AI-based automated evaluation system for non-destructive testing is also significant as a result of collaboration with a startup. It is the outcome of three years of joint development with DeepEye. This technology received New Technology (NET) certification from the Ministry of Trade, Industry and Energy this year, the first of its kind in Korea. Unlike the conventional 'conditional evaluation' method, which determines there is no abnormality if certain criteria are met, the AI learns thousands of signal patterns and classifies them precisely to determine defects, making it a world-first technology. DeepEye, which is leading technology sales, is currently in discussions with seven countries, including Japan, the United States, Spain, and Singapore.

Kim Kisu, CEO of DeepEye (first from the right), and other officials are posing for a commemorative photo at the "2025 New Technology and New Product Certification Ceremony" held last month. The "AI-based Heat Exchanger Non-destructive Inspection System," jointly developed by SK Innovation and DeepEye, received New Technology (NET) certification from the Ministry of Trade, Industry and Energy. SK Innovation

The partnership between the two companies began three years ago at an industrial exhibition. Since then, they have continued to advance the technology through a memorandum of understanding (MOU) and proof of concept (POC), and from the second half of this year, it will be rolled out across all SK Innovation plants. DeepEye was responsible for the software algorithms, while SK Innovation handled field validation and data. Kim Kisu, CEO of DeepEye, said, "At a time when interest in AI was low, SK Innovation remained committed to the collaboration, enabling us to develop a Korean-style technology that does not rely on overseas solutions."

The company culture, in which engineers familiar with the field take the lead and collaborate flexibly with startups, is also drawing attention. An SK Innovation official at the Ulsan plant commented, "The key to successful collaboration was that engineers who understand the technology worked directly with the startup to lead the project."

AI collaboration trend spreads across the industry

The SK Innovation case reflects broader changes in the refining and chemical industries. GS Caltex is also operating a system at its Yeosu plant in Jeonnam that analyzes pipe exterior corrosion using AI in partnership with the startup PlantScanning. Previously, workers had to manually interpret thousands of image data files, but with AI, defect detection can now be completed in just a few seconds.

AI is also being applied to process optimization, such as naphtha cracking. GS Caltex has introduced an AI-based profit optimization model that automatically adjusts production volumes in real time according to market prices and facility conditions. For example, if the price of propylene surges, the proportion of related production is increased. Recently, the company has maximized the production of high-value-added naphtha products (HSR) with its self-developed machine learning-based Real Time Optimizer (RTO) system. With these technologies, GS Caltex received the Ministry of Trade, Industry and Energy's 'AI Autonomous Manufacturing Contributor Award' last year.

Kim Changheon, head of GS Caltex's Digital Innovation Team, said in an interview with Asia Economy, "After introducing AI, profitability improved by about 2%, and the effect will increase as we further advance the technology. Initially, some field workers had vague fears about the new technology, but after experiencing how AI reduces the burden of repetitive tasks and improves efficiency, the atmosphere changed." He added, "In the past, digital transformation was considered the responsibility of a few departments, but now it has become a culture of collaboration across departments."