Authors: Shuyuan Zhang, Yong Ren Tan, Cuong Manh Nguyen, Dogancan Karan, Srishti Ganguly, Nicholas Jose, Mei Qi Lim, Shuqiao Guo, Lianlian Jiang, Markus Kraft*, and Alexei A. Lapkin

• A full knowledge-graph based implementation of a chemical process digital twin is demonstrated
• Novel process ontologies were developed for continuous plant and communication with digital twins
• Anomaly detection workflow was demonstrated in the knowledge-graph based digital twin

Digitalisation holds promise for the transformation of modern chemical processes into intelligent manufacturing systems that are responsive to dynamic market demands and sustainability goals. To achieve interconnection of physical installations with their digital twins, we developed a semantic framework which enables systematic knowledge management and effective utilisation of plant data, serving as a backbone of intelligent manufacturing. Our approach leverages ontologies and agents built on a knowledge graph infrastructure to host the deployment of both first-principles and artificial intelligence (AI) models of chemical processes. Additional ontologies were designed for connecting models with industrial plants. We present an end-to-end demonstration of the knowledge graph-based digital twin system and illustrate this technology with a use case of anomaly detection in an industrial pilot plant for nanomaterials synthesis. The implementation includes data acquisition, secure communication protocols, cloud-hosted data storage, dedicated AI workflows, and a user interface with version control for both ontologies and models. The results demonstrate the effectiveness of the semantic framework in managing chemical process knowledge, linking plants to first-principles and data-driven models, and enabling the execution of complex AI-driven workflows.

• Access the article at the publisher: DOI: 10.1016/j.cej.2026.174361

Associated Themes: