A Human-AI Collaborative Product Conceptual Design Method Driven by LLMs and Knowledge Graphs for Design Knowledge Reasoning

Product conceptual design is a critical stage in product development, where user requirements, design knowledge, functional reasoning, and structural solution generation need to be integrated into a coherent design process. In this stage, user requirements are often expressed in an abstract and incomplete manner, while relevant design knowledge is distributed across patents, technical literature, and prior engineering cases. This makes it difficult to directly transform user intentions into reusable functional knowledge and feasible structural solutions. Existing studies have established a solid foundation in design knowledge organization, graph reasoning, and intelligent knowledge extraction. Building on these studies, a Human-AI collaborative conceptual design workflow still needs to be further refined to connect requirement understanding, knowledge mapping, graph-based reasoning, and solution generation. To address this issue, a Human-AI collaborative product conceptual design method driven by large language models (LLMs) and knowledge graphs is proposed. The method constructs a multi-level knowledge graph from patents and technical literature, uses LLMs-based semantic parsing and semantic similarity-based mapping to transform natural language requirements into graph-compatible functional representations, and applies graph reasoning to generate requirement-oriented Function-Structure solution combinations. In this process, users provide design requirements, while designers participate in requirement judgment, semantic mapping review, and solution evaluation, forming a Human-AI collaborative conceptual design workflow. A case study on a resilient pneumatic control valve demonstrates the feasibility and effectiveness of the proposed method.