AI-Driven Catalyst Engineering for Sustainable and Efficient Chemical Processes

Abstract

The development of sustainable, energy-efficient chemical processes remains a significant challenge in modern catalysis, due to high energy consumption, limited catalyst stability, and inefficient trial-and-error design strategies. This study aims to develop and validate an AI-driven catalyst engineering framework that integrates machine learning optimisation with systematic experimental evaluation to enhance catalytic performance and sustainability. Machine learning algorithms were employed to optimise catalyst composition and structure, followed by experimental validation across multiple performance indicators, including efficiency, product yield, energy consumption, stability, and prediction accuracy. The results demonstrate that the AI-designed catalysts achieved catalyst efficiencies exceeding 80%, with a maximum value of approximately 83%, and delivered product yields of up to ~92% across optimised compositions. Energy consumption was significantly reduced by nearly 50%, decreasing from about 90 kWh at 200 °C to approximately 45 kWh at 500 °C, indicating substantial improvements in process energy efficiency. Long-term stability tests showed that the catalysts retained around 76% of their initial activity after 30 reaction cycles, confirming strong resistance to deactivation. In parallel, the AI model demonstrated continuous learning, achieving a prediction accuracy of ~90% through iterative experimental feedback. Overall, this study confirms that AI-driven catalyst engineering enables simultaneous improvements in performance, energy efficiency, stability, and predictive reliability, providing a robust, transferable framework for developing sustainable, efficient chemical processes.