Smarter Industrial Fault Diagnosis with Knowledge Graphs and Large Language Models

Session Track: AI Engineering

Session Time: 07:30 am - 08:00 am November 6

Session description

Manufacturers worldwide struggle with unplanned downtime and knowledge loss due to the fragmented, multilingual nature of maintenance data. In this session, Wan will demonstrate how GraphRAG, an AI-powered approach combining RAG with knowledge graphs, can transform unstructured multilingual maintenance logs and machine manuals into actionable, structured insights using Neo4j as the core technology. Attendees will see how GraphRAG extracts and links fault locations, symptoms, causes, and measures, overcoming language barriers and document silos. The session will highlight a real-world deployment on complex industrial equipment, where engineers interact with a multilingual knowledge graph through natural language, enabling faster, more accurate fault diagnosis and capturing critical operational expertise. You will learn practical steps for building a similar system with Neo4j, leveraging LLMs and your own multilingual maintenance data. You will gain insights into architecting knowledge graphs for AI-driven troubleshooting, integrating LLMs for retrieval and generation, and designing scalable solutions to augment human expertise in maintenance.