How Aramco Trained AI to Run Its Refineries

Emerson's Aspen Hybrid Models go live across Aramco's global refineries, hitting 98.5% yield accuracy and redefining AI in downstream planning

1 May 2026

Emerson announced on April 28, 2026, the full deployment of its Aspen Hybrid Models artificial intelligence system across Aramco's global refinery network, producing what company officials described as one of the largest multi-site, multi-period optimization models in the industry. The system has achieved yield and quality prediction accuracy of up to 98.5 percent in select units, a result that analysts said could redefine planning standards in downstream operations.

The technology fuses first-principles engineering models with purpose-built machine learning to capture the nonlinear behaviors that govern how feedstocks perform under varying processing conditions, a challenge that has long resisted straightforward computational approaches. In Aramco's Continuous Catalyst Regeneration and Platformer units, the gains have been immediate. Engineers are spending less time on manual corrections, according to company statements, and more time on higher-order decisions that affect margin and output.

Practical improvements extend across the full planning workflow. Feedstock blending decisions have grown more precise, officials said, and the gap between planned and actual plant performance has narrowed. Margin forecasting, historically one of the more uncertain variables in refinery economics, has become more reliable across a network serving markets on multiple continents.

Ahmad Alkudmani, director of the global optimizer department at Aramco, said the deployment represented a meaningful step in applying innovative technology to smarter refining optimization. Claudio Fayad, chief technology officer of Emerson's Aspen Technology business, pointed to the value of combining domain expertise with advanced AI to manage complex, multi-site planning workflows. Yet both executives acknowledged that the current phase is not the endpoint.

Expansion into hydrocracker units is now underway, a technically demanding step that will test the system's scalability across some of the most complex conversion processes in the industry. If that phase succeeds, the AI planning layer will cover the full breadth of Aramco's downstream portfolio. As refinery margins tighten globally and sustainability requirements grow more exacting, the results could shape how energy producers across the region approach optimization in the years ahead.