SK Innovation's Ulsan Complex (UlsanCLX) is leveraging artificial intelligence (AI) to predict and optimize processes. The facility employs a model known as 'Dual Brain AI,' which combines AI capabilities with the experience and expertise of on-site personnel.
Founded in 1964 as South Korea's first oil refinery, UlsanCLX has expanded its facilities and processes over the past 60 years. It now boasts the largest crude oil processing capacity in the country, handling 840,000 barrels per day, and ranks third globally in crude oil refining capacity for a single petrochemical plant. The site spans 2.5 million pyeong, including storage, production, and shipping facilities.
With over six decades of operation, UlsanCLX faces distinct challenges. Jeong Chang-hoon, head of the Manufacturing AI team at UlsanCLX, noted, 'There are generational differences among the equipment built in the 1960s, 70s, and 90s. These multi-generational facilities have 98 interconnected processes, making optimal operation difficult.'
To address this complexity, SK Innovation has introduced the 'Dual Brain' approach. In this model, AI analyzes and predicts vast amounts of data around the clock, while experienced engineers make final decisions based on their expertise. This structure does not fully rely on AI but rather complements human limitations in time and physical capacity.
The necessity of the Dual Brain approach became evident in the control room of the Heavy Oil Upgrading Facility (HOU) during a recent visit. Numerous process data points were displayed in real-time on monitors, with the number of screens across UlsanCLX reaching hundreds to over a thousand. The plant operates continuously, 24/7, with staff managing operations in four shifts.
Much of the process is already automated, with computers controlling equipment. Operators typically do not need to intervene unless unexpected situations arise, at which point workers quickly restore normal operations. A site representative explained, 'The experience and know-how of skilled workers accumulate during these unexpected situations.'
Ultimately, UlsanCLX aims to encapsulate this know-how within AI. The data accumulated over years of operation will be used to train AI, enabling it to identify anomalies and suggest responses in similar situations. However, the final judgment will still rest with humans. Jeong emphasized, 'I don't think we can replace people. Our unique experiences and knowledge must continue to be integrated.'
To implement the Dual Brain concept, UlsanCLX is pursuing five key manufacturing AI initiatives across three areas: optimization, equipment reliability, and safety, health, and environment (SHE). The optimization area includes production optimization AI and utility optimization AI, while the equipment reliability area features predictive maintenance AI and maintenance planning AI. The SHE area includes the integrated safety monitoring system 'SHIELD.'
The 'Production and Process Optimization AI' was demonstrated at the Ulsan World Future Industry Expo (WAVE 2026) on September 10. Participants could adjust cooling and heating conditions to find optimal operating points. While humans adjusted values one by one, AI quickly derived optimal values. AI analyzes variables that are difficult for humans to calculate, such as crude oil type, processing volume, operating conditions, cooling and heating conditions, product quality, and energy efficiency, and suggests optimal operating conditions for engineers to decide on final operations.
Beyond optimizing individual processes, efforts are underway to connect the entire plant. The utility optimization AI adjusts the usage of steam, electricity, and water needed for process operations, with the long-term goal of linking multiple processes to enhance overall value chain margins.
In areas that are difficult for humans to access, drones and robotic dogs are also being utilized. These mobile devices check equipment status and assess on-site hazards, collecting data. Currently, UlsanCLX operates one robotic dog, with plans to expand to about ten in the future. Drones are also expected to be used more extensively for equipment inspections and site patrols.
The generational transition of skilled personnel is another driving factor behind the AX initiative. Jeong noted, 'By 2030, many of the veteran workers who have been with us since 2025 are expected to retire.' It is crucial to capture the judgment criteria and experiences built over decades of responding to equipment issues and unexpected situations in data and AI to minimize technological gaps.
UlsanCLX aims to achieve an annual productivity increase of approximately 100 billion won, zero equipment downtime, and zero major accidents by 2030. This year, the facility will begin developing and validating predictive models and agents, with plans to enhance data infrastructure starting in 2027. By 2028, the goal is to ensure the reliability of Brain AI, and after 2030, to continue expanding AX applications to establish a Dual Brain operational system.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.