ChatP&ID

Piping and Instrumentation Diagrams (P&IDs) are the backbone of process engineering, yet they often feel impenetrable due to their intricate details and vast scope. While process engineers are trained to work with flowsheets, retrieving information can often be tough, time-consuming, and error-prone especially when dealing with bundles of flowsheets.

We develop ChatP&ID, a tools that bridges large language models (LLMs) and process engineering diagrams by digitizing and contextualizing P&IDs. This tool built upon and integrated into our digitization framework, DigiCo, which convert the legacy P&ID in image or pdf format into machine-readable smart P&ID format like DEXPI. The contextualization part is done by transforming DEXPI P&ID into semantically rich knowledge graphs with the help of in house build library pyDEXPI. To optimize information density and provide intuitive representation, the knowledge graph is condensed into a high-level knowledge graph abstraction. Finally, we integrate the graph database with graph-based retrieval augmented generation (GraphRAG) and chat interface for seamless interaction with the engineer or user.

This approach allows LLMs to retrieve and infer over engineering flowsheet effectively, while minimizing hallucinations by anchoring responses in structured flowsheet information. The high-level graphs preserve key process information while reducing token usage, enabling efficient and scalable AI interaction. We envision ChatP&ID serves as a foundation for various applications in process engineering, such as knowledge transfer, contextual learning, efficient information retrieval, and eventually AI-assisted flowsheet reasoning and modification. Following clip to see ChatP&ID in action!

As AI continues to evolve rapidly, we can expect to see an influx of new ideas, models, and techniques. Even these days, there are numerous information retrieval methods and an ever-growing list of language models. While models will continuously evolve, databases remain the foundation and this is where our expertise as chemical engineers comes into play to ensure the implementation is at its finest and safest.

Key Publication

• Alimin, A. A., Goldstein, D. P., Schulze Balhorn, L., & Schweidtmann, A. M. (2025). Talking like Piping and Instrumentation Diagrams (P&IDs). arXiv preprint, In: Proceedings of ESCAPE35, Ghent, Belgium.
• Goldstein, D. P., Alimin, A. A., Schulze Balhorn, L., & Schweidtmann, A. M. (2025). pyDEXPI: A Python framework for piping and instrumentation diagrams using the DEXPI information model. In: Proceedings of ESCAPE35, Ghent, Belgium.