MetalMind: A knowledge graph-driven human-centric knowledge system for metal additive manufacturing

Abstract

In the Industry 5.0 era, increasing manufacturing complexity and fragmented knowledge pose challenges for decision-making and workforce development. To tackle this, we present a human-centric knowledge system that integrates explicit knowledge from formal sources and implicit knowledge from expert insights. The system features three core innovations: (1) an automated KG construction pipeline leveraging large language models (LLMs) with collaborative verification to enhance knowledge extraction accuracy and minimize hallucinations; (2) a hybrid retrieval framework that combines vector-based, graph-based, and hybrid retrieval strategies for comprehensive knowledge access, achieving a 336.61% improvement over vector-based retrieval and a 68.04% improvement over graph-based retrieval in global understanding; and (3) an MR-enhanced interface that supports immersive, real-time interaction and continuous knowledge capture. Demonstrated through a metal additive manufacturing (AM) case study, this approach enriches domain expertise, improves knowledge representation and retrieval, and fosters enhanced human-machine collaboration, ultimately supporting adaptive upskilling in smart manufacturing.