Knowledge graph construction technology and its application in aircraft power system fault diagnosis

The electrification degree of airborne equipment continues to increase with the development of science and technology. Owning to this, failure in aircraft power system is posing an increasing threat to flight safety, which granted necessity to fast and accurate health state assessment. The commonly used data -driven fault diagnosis method cannot make use of expert knowledge. Meanwhile, the result of data-driven method is lack of interpretability, and therefore, limits its application in real practice. Knowledge graph has the ability to normalize the storage of such unstructured data as expert knowledge and use it for diagnosis. Moreover, knowledge graph can utilize the unstructured knowledge and supply a reasonable explanation for the cause of the failure. However, in the field of fault diagnosis,there are still few studies on the application of knowledge graph technology. In this article, a knowledge graph construction and application technology for aircraft power system fault diagnosis is proposed. First, ontology of the knowledge graph, which specifies the entity and relation types in the knowledge graph, is constructed based on the priori expert knowledge. Then, Bi-Directional Long Short-Term Memory (Bi-LSTM) method is trained with BMEO-tagged corpus and utilized to extract entities from the unstructured texts. After that, an attention -based Bi-LSTM algorithm is trained with relation-tagged corpus and then utilized to realize relation extraction. Finally, the knowledge graph for aircraft power system fault diagnosis is constructed based on the extracted entities and relations.A fault isolation manual of aircraft power system is used as raw corpus data in the case study to verify the effectiveness of the proposed method by the indicators of precision and recall. Based on the knowledge graph, intelligent searching, recommending and Q & A are realized , which strongly support the application prospect of knowledge graph in the field of fault diagnosis. © 2022 AAAS Press of Chinese Society of Aeronautics and Astronautics. All rights reserved.