宇航元器件选用是航天任务中的关键环节，传统选用方式存在效率低下、专业知识依赖度高等问题。本文设计并实现了一种基于大语言模型和检索增强生成技术的宇航元器件智能选用推荐系统。系统采用智能体架构，实现了从用户自然语言需求到元器件推荐方案的端到端智能处理。构建了包含468个宇航级元器件和60套系统级物料清单的专业知识库，设计了融合精确匹配与语义理解的多策略级联检索机制，并针对航天安全要求设计幻觉防控机制。系统可处理单一元器件推荐和系统级方案生成等应用场景。在包含65个测试用例的评估实验中，系统宏平均F1分数为0.829，较人工关键词检索提升10.4%，较纯大语言模型提升69.5%，验证了系统的有效性。本文设计的宇航元器件智能选用推荐系统，可有效支撑宇航元器件的智能化选用工作。

The selection of aerospace components is a critical link in space missions. Traditional selection methods are plagued by issues such as low efficiency and high reliance on professional expertise. This paper designs and implements an intelligent selection and recommendation system for aerospace components based on large language models (LLMs) and retrieval-augmented generation (RAG) technology. Adopting an agent-based architecture, the system achieves end-to-end intelligent processing, converting users' natural language requirements into component recommendation schemes seamlessly. A professional knowledge base is constructed, encompassing 468 aerospace-grade components and 60 system-level bills of materials (BOMs). A multi-strategy cascaded retrieval mechanism integrating exact matching and semantic understanding is designed, and an hallucination prevention and control mechanism is developed to meet aerospace safety requirements. The system is capable of handling application scenarios including single component recommendation and system-level scheme generation. In an evaluation experiment involving 65 test cases, the system achieves a macro-average F1-score of 0.829, representing a 10.4% improvement over manual keyword retrieval and a 69.5% improvement compared to pure LLM methods, thus verifying the system's effectiveness. The intelligent selection and recommendation system for aerospace components proposed in this paper can effectively support the intelligent selection of aerospace components.