Evonik Oxeno has taken a significant step towards AI-driven chemical manufacturing, deploying artificial intelligence-based process models in live plant operations for the first time to improve energy efficiency, reduce CO₂ emissions and accelerate operational decision-making.

 

Already operational at facilities in Marl and Antwerp, the technology enables plant operators to respond more quickly to changes in feedstock streams while optimizing production conditions and minimizing the energy required to manufacture products.

 

The breakthrough comes under COMPAS (Control and Optimization using AI Models for Process Advancement and Cost Saving), a project that combines advanced process engineering with artificial intelligence to make complex chemical plant operations more efficient.

 

At the heart of the technology is a library of more than one million pre-calculated simulations covering different plant operating conditions. These simulations form the foundation of so-called gray-box models, which use specially programmed and trained AI models to capture the thermodynamics and kinetics of chemical processes.

 

The approach addresses a longstanding challenge in chemical manufacturing: not all processes inside a production facility can be directly observed, while identifying the most efficient operating conditions requires accounting for numerous variables.

 

Although conventional process simulations offer valuable insights, they are often complex and time-consuming. COMPAS overcomes this limitation by drawing on its extensive database of simulated scenarios to generate realistic plant-control recommendations within seconds.

 

“We combine our extensive knowledge of chemical processes with the capabilities of artificial intelligence. This allows us to bring complex process simulations directly into plant operations and support efficient plant control even under changing operating conditions,” says Niklas Paul, Head of Engineering at Evonik Oxeno.

 

The initial results are encouraging. Applications running at the Marl and Antwerp facilities have demonstrated a very high level of agreement between AI model predictions and actual plant operating data, providing a reliable basis for further operational optimization.

 

By identifying more efficient operating conditions, the models help minimize energy consumption and directly improve the carbon footprint of production. The company is also advancing additional applications, including the automated transfer of AI-generated results directly into plant control systems.

 

Evonik sees COMPAS as a building block in its longer-term ambition to develop a self-optimizing C4 Verbund, leveraging accumulated expertise across plant operations, engineering and research.

 

“COMPAS demonstrates that the knowledge jointly built by operations, engineering and research over many years creates new value in process optimization through artificial intelligence. Based on this technology, we are working step by step toward a self-optimizing C4 Verbund,” says Robert Franke, Head of Simulation & Kinetics at Evonik Oxeno.

 

The project team has also taken the technology into the scientific arena, publishing its methodology and latest findings in the journals Chemie Ingenieur Technik and Chemical Engineering & Technology.

 

Beyond its current applications at Evonik Oxeno, the COMPAS methodology could potentially be extended to other chemical and petrochemical production processes, opening up wider opportunities for AI-driven plant optimization.

 

The team is continuing to expand the technology's application scope and is seeking patent protection, signalling Evonik's ambition to translate AI-powered process intelligence into a broader industrial advantage.