Associations of Knowledge Graph Reasoning and Semantic Grounding with Fault Diagnosis in Industrial Maintenance Logs

Abstract

The rapid advancement of artificial intelligence in industrial contexts has significantly improved automated fault diagnosis. However, the inherent opacity of deep learning models remains a critical barrier to their widespread adoption in high-stakes industrial maintenance environments. This paper presents a novel framework for explaining fault diagnosis by integrating knowledge graph reasoning with semantic grounding techniques applied to unstructured industrial maintenance logs. Maintenance logs contain a wealth of historical diagnostic reasoning and semantic relationships, yet their unstructured, jargon-heavy nature complicates automated analysis. By semantically grounding these logs into a structured industrial knowledge graph, our approach bridges the gap between raw textual data and formal diagnostic ontologies. The proposed framework extracts entities such as equipment components, failure modes, and maintenance actions, mapping them to a domain-specific knowledge graph. Graph-based reasoning algorithms are then employed to trace diagnostic paths, thereby generating interpretable explanations for inferred faults. Through comprehensive experimentation on a large-scale industrial dataset, this study demonstrates that the proposed method not only achieves diagnostic accuracy comparable to state-of-the-art black-box models but also provides verifiable, human-readable explanations. The integration of semantic grounding ensures that the diagnostic reasoning remains faithful to the practical realities documented by maintenance personnel. This research contributes to the broader field of explainable artificial intelligence in industrial systems, offering a reliable, transparent, and highly effective methodology for intelligent fault diagnosis.