A knowledge graph framework for digital twins of chemical processes

Area of Science:

• Chemical Engineering
• Computational Chemistry
• Data Science

Background:

• Digital twins are crucial for chemical process development and sustainability.
• Integrating data, domain knowledge, and predictive models for chemical processes is challenging.
• A systematic approach is needed to represent chemical processes effectively within digital twins.

Purpose of the Study:

• To propose a knowledge graph framework with autonomous functional agents for developing digital twins of chemical processes.
• To enable seamless integration of chemical databases, AI models, and large language models.
• To provide a scalable method for managing and utilizing process models as chemical knowledge.

Main Methods:

• Developed a knowledge graph framework with autonomous functional agents.
• Created ontologies for physical models of chemical processes.
• Integrated chemical databases, AI models, and large language models.
• Demonstrated with case studies on model assembly, search, and optimization.

Main Results:

• The framework facilitates the integration of diverse data and models for digital twins.
• Ontologies enable scalable construction and calibration of physical process models.
• Case studies show successful application in model assembly, search, and reaction optimization.
• Established a method to manage process models as a knowledge repository.

Conclusions:

• The proposed framework provides a systematic approach to digital twin development for chemical processes.
• It enhances the integration of data, domain knowledge, and predictive models.
• This foundation supports future advancements in chemical process development and manufacturing.