Abstract: In response to the intelligent development demands of railway scientific and technological intelligence and knowledge services, this paper developed a large model application system named Tiekezhiwen for the railway industry. Adopting a hierarchical architecture design, the system targeted practical railway scientific research scenarios and provided functions including professional knowledge question answering, multilingual text polishing, rapid literature reading, automatic research review generation and safety accident bulletin compilation. By constructing a railway professional knowledge base Rail Knowledge Base (Rail-KB) and integrating Retrieval-Augmented Generation (RAG) technology, the system improved the professional knowledge comprehension and content generation capability of large models in railway-specific scenarios. Application results show that the system remarkably boosts the acquisition efficiency of railway scientific research information and ensures the professionalism and reliability of generated professional contents. It also offers practical references for the implementation and popularization of large models in the railway industry.