A RippleNet-Based Recommended Diagnostics Method Based on Knowledge Graph Integrated Human-Machine-Environment

Efficient and intelligent equipment fault diagnosis is crucial in industrial applications. However, traditional knowledge graphs for fault diagnosis typically focus solely on analyzing machine factors and primarily rely on simple embedding or path-matching techniques to recommend diagnostic methods. This approach struggles to address complex diagnostic requirements. To address these challenges, this paper proposes a novel equipment fault diagnosis method that leverages historical diagnostic information and a fault knowledge graph. In the ontology construction phase, three key factors—human, machines, and the environment—are comprehensively considered, leading to the development of a multilevel equipment fault knowledge graph that integrates these elements. Subsequently, in the recommendation and diagnosis phase, a model based on RippleNet-based recommended diagnostics algorithm (RRDA) is developed. This model integrates the strengths of embedding-based and path-based recommendation approaches, achieving precise fault diagnosis and recommendation through an iterative propagation mechanism. This mechanism effectively fuses historical diagnostic data with the structural relationships within the knowledge graph, disseminating and aggregating relevant information in a layered manner. Experimental results indicate that the proposed method significantly outperforms in deep-level fault mode mining, diagnostic accuracy, and interpretability, highlighting its potential for broad application in equipment fault diagnosis.