Emerson, a U.S.-based industrial automation company, and Aramco, a Saudi energy company, have deployed an AI-driven refinery optimization system that integrates Aspen Hybrid Models into Aramco’s existing refinery planning environment.
Emerson claimed that the deployment as one of the world’s largest multi-site, multi-period optimization models and said it has achieved yield and quality prediction accuracy of up to 98.5% in key refinery units.
The models are already running in Continuous Catalyst Regeneration and Platformer units, where they are being used to improve feedstock blending, reduce gaps between planning assumptions and actual plant performance, and improve margin forecasting across Aramco’s global refining network.
Emerson said current work is focused on expanding the same modeling approach into hydrocracker units across Aramco assets.
Integrating first-principles with machine learning
Aspen Hybrid Models combine first-principles engineering models, domain expertise and AI. AspenTech’s product material says the approach is designed for process industries where conventional models alone may struggle with complex processes, changing plant behavior and sustained model accuracy over time.
Claudio Fayad, chief technology officer of Emerson’s Aspen Technology business, said the deployment optimizes “complex, multi-site, multi-period planning workflows” and shows “the tangible value of combining deep domain expertise with advanced AI.”
That architecture is central to the Aramco deployment. Emerson said Aramco used thousands of converged simulation cases built on first-principles models calibrated with actual plant data to create nonlinear optimizations for refinery planning.
Based on Emerson’s description, the deployment is closer to a refinery-planning integration than a standalone machine-learning overlay.
The approach also aligns with how AspenTech positions production planning software for refineries. Its Aspen Unified PIMS material says refineries and petrochemical plants often face production efficiency gaps that affect margins, and that Aspen Hybrid Models combine first-principles models with domain expertise and AI to improve planning-model accuracy.
Structural barriers to digital adoption
AI adoption is currently still constrained by data, skills and operational complexity. The International Energy Agency said lack of digital skills is the largest barrier to greater AI adoption in energy, with fragmented data, data protection, privacy, cybersecurity and inadequate equipment digitalization also limiting uptake.
Those barriers make the 98.5% figure material for refinery operators, but it remains a company-reported metric. Emerson said the models are already running in refinery units and were built from first-principles models calibrated with actual plant data.
The company said the deployment is also intended to reduce manual adjustments by engineers, automate model updates and maintain applicability across a wide range of refinery operations.
Aramco says it began a digital transformation program in 2017 to coordinate digital projects across areas including compliance, sustainability, supply chain, workforce and operations. Its operational program uses fourth industrial revolution (4IR) technologies to improve upstream and downstream operations, while its sustainability program uses big data, AI algorithms and analytics to make operations more productive and efficient.
Expanding the AI value-creation roadmap
Aramco has also disclosed in its 2025 results that it generated $5.3 billion in “Technology Realized Value” from AI, digital and other solutions in 2025, taking cumulative value to $11.3 billion since 2023. The same release said the company was progressing toward a minority stake in HUMAIN to unlock new AI value-creation opportunities.