06.2 - Flight Dynamics and Control (UAV related)
FLIGHT CONTROL SYSTEM KNOWLEDGE GRAPH CONSTRUCTION BASED ON AERONAUTICAL DOMAIN KNOWLEDGE AUGMENTED LARGE LANGUAGE MODEL
Y. Fan, China; Y. Sun, China; B. Mi, China
Continuous intelligent safety assurance of the flight control system (FCS) determines the possibility of mission execution of the aerospace vehicle, and the system state monitoring and assurance of massive data is a task that relies heavily on the empirical knowledge of experts. Therefore, this paper develops a set of FCS knowledge graph construction and application process based on the domain knowledge augmented large language model (LLM) for the problem of difficult knowledge discovery and organization of FCS. By integrating the aeronautics domain knowledge base, to form the FCS ontology, combining the entity relationship types defined in the ontology, adopting the retrieval augmented generation (RAG) e.g. Langchain method on the basis of the basic LLM e.g. ChatGPT, injecting the information in the domain knowledge base, and designing the simple content-type and sentence-type prompt information templates for different FCS knowledge extraction tasks, we can realize the FCS knowledge graph construction and application process. The proposed method can realize the zero-shot or one-shot intelligent extraction FCS knowledge, and display the FCS knowledge in the form of visual graph, which can be applied to FCS data and knowledge value mining and state monitoring.