Can AI agents automate industrial diagram design?
Accelerating Manufacturing Scale-Up from Material Discovery Using Agentic Web Navigation and Retrieval-Augmented AI for Process Engineering Schematics Design
This paper introduces a two-stage framework using a multi-agent system and a Graph Retrieval-Augmented Generation (Graph RAG) approach to automate the creation of Process Flow Diagrams (PFDs) and Process & Instrumentation Diagrams (PIDs) for chemical processes. This addresses the challenge of scaling material discoveries to industrial production.
Key points for LLM-based multi-agent systems: Specialized sub-agents autonomously gather information from online sources (images, scholarly articles, patents, wikis, etc.) using SerpAPI and LLMs. A meta-agent coordinates these sub-agents and uses feedback from human experts and a "Gold" LLM to refine the collected knowledge. This information is then structured into an ontological knowledge graph using Graph RAG, which enables efficient retrieval and complex question-answering about PFD/PID generation, overcoming limitations of traditional RAG. This framework allows for automated generation of diagrams and precise answers to technical questions, bridging the gap between computational design and real-world industrial implementation.