ChatP&ID: GraphRAG for engineering diagrams in AIChE Journal

Our paper "GraphRAG for engineering diagrams: ChatP&ID enables LLM interaction with P&IDs" by Achmad Anggawirya Alimin and Artur M. Schweidtmann was published in the AIChE Journal.

Extracting information from Piping and Instrumentation Diagrams (P&IDs) is a tedious part of many process engineering workflows. ChatP&ID lets engineers query smart P&IDs in natural language. It uses Graph Retrieval-Augmented Generation (GraphRAG): DEXPI-encoded P&IDs are converted into knowledge graphs, which large language model (LLM) agents query through retrieval tools. This grounds the answers in the actual diagram data.

Key results from benchmarking commercial and open-source LLMs:

• Graph representations improve answer accuracy by 18% compared with raw image input.
• Token costs fall by 85% compared with ingesting smart P&ID files directly.
• The ContextRAG retrieval strategy reaches 91% accuracy at about $0.004 per query with GPT-5-mini.
• For smaller open-source models, vector-based retrieval improves accuracy by up to 40%.

The work lays the groundwork for AI-assisted HAZOP studies, automated P&ID checking and multi-agent engineering workflows. More on the ChatP&ID research page.

Paper: A. A. Alimin, A. M. Schweidtmann (2026). GraphRAG for engineering diagrams: ChatP&ID enables LLM interaction with P&IDs. AIChE Journal, 72(10), e70540. https://doi.org/10.1002/aic.70540