Enhancing Fault Troubleshooting through Human-Machine Collaboration: A Multi-Stage Reasoning Approach

Ensuring the stable operation of critical industrial equipment is pivotal for maintaining production efficiency and economic gains. The complexity of modern industrial machinery, however, places a substantial cognitive load on maintenance personnel. To alleviate this, a Diagnostic Semantic-Enhanced Fault Causality Knowledge Graph (DSFCKG) is proposed to formalize fault information for computational analysis. Additionally, a Large Language Model (LLM)-based Knowledge Graph Construction (KGC) method is introduced for the automated assembly of DSFCKG. Building upon this, a multi-stage reasoning approach is designed for human-machine collaborative fault Troubleshooting. Experiments on real-world fault tickets demonstrate that our proposed method significantly enhances fault diagnosis and troubleshooting accuracy, especially in complex scenarios with long fault causal chains, which bring insights into futuristic smart maintenance.