SK Innovation is moving quickly on its Dual Brain initiative at the Ulsan Complex, South Korea's largest crude processing facility. The idea is to combine two brains—AI's computing power and the experience of veteran engineers—to analyze and make decisions on the complex's processes around the clock. AI detects early warning signs in processes and equipment and proposes responses; on-site engineers review the situation comprehensively before making the final call. Those decisions are then accumulated as data and fed back into the AI's next analysis.
The scale of the site
The Ulsan Complex spans about 8.26 million square meters, roughly three times the size of Yeouido. It refines 840,000 barrels a day across 98 integrated processes, and some 3,000 workers a day carry out as many as 330,000 maintenance and repair jobs a year. Rather than relying on field staff alone to manage operations at that scale, the company aims to deploy AI to secure both productivity and safety.
Three pillars
Dual Brain rests on three pillars: process optimization, equipment reliability, and safety, health, and environment. The goal is to maximize integrated margins across the refining and chemicals value chain while maintaining uninterrupted operations 365 days a year and running the plant without casualties or fires.
The core tool built in-house for equipment diagnostics is SKADI, an equipment diagnostics platform that uses AI to monitor vibration and temperature changes in key rotating machinery such as pumps and compressors in real time, pinpointing the causes of abnormalities. The analysis is passed directly to a maintenance planning AI, which proposes the optimal timing and work plan for repairs. SK Innovation plans to add predictive AI to expand the scope of equipment diagnostics and sharply improve accuracy by 2028.
On the safety side, SHEILD is a control platform that implements safety, health, and environment functions. It shifts field inspections from people-centered patrols to AI-based, 24-hour monitoring, analyzing safety data such as the status of high-risk work from multiple angles. Robot dogs and drones have also been deployed to collect mobile data, and the company plans to increase the number of robot dogs on site from one at present to 10.
Productivity targets
SK Innovation expects Dual Brain to deliver productivity gains of about 100 billion won, roughly $72 million, a year. It is also considering commercializing the system externally once it has been proven. The Ulsan Complex started 60 years ago as South Korea's first oil refinery, and the company's stated aim is that 60 years from now it will be a plant operated on Dual Brain, where humans and AI work together.
What the model reveals
The Dual Brain approach is a useful example of how industrial AI is being deployed in practice: not as full autonomy, but as an always-on recommendation layer that augments the temporal and physical limitations of human workers. AI reads process data continuously and suggests optimal operating conditions; humans with field experience and contextual knowledge make the final decisions. That midpoint—AI as control tower and controller, humans as final arbiter—is becoming a common pattern in assets where downtime, safety, and accumulated operational knowledge all matter at once.