105 / 2023-09-19 19:07:18

Spacecraft Fault Knowledge Graph Completion Based on Knowledge Reasoning

knowledge graph,relation reasoning,knowledge representation

In this paper, based on the knowledge graph of spacecraft fault, the methods of semantic representation learning and graph embedding, which are commonly used in the field of knowledge graph embedding, are explored. Based on the data multi-source heterogeneity of the fault knowledge graph, this paper proposes a multi-embedding information fusion method to improve the embedding vector characterization capability. To address the problem of insufficient information in the adjacent subgraphs, a GraphSAGE graph neural network method is proposed for secondary aggregation to construct node embedding and feature extraction. These methods accomplish the transformation from graphs to computer-recognizable vectors, and furthermore, relational reasoning with multi-feature fusion is realized using the attention mechanism. In addition, for the problem of different training mechanisms with different embedding modes, the embedding effect is evaluated using labeled multiclassification assessment to target the embedding quality and improve the reasoning effect. Finally, the reasoning methods used in this paper are comparatively analyzed by means of experiments.