Top-Down Hierarchical Construction and Application of a Domain Knowledge Graph Based on Multimodal Design Information

Market demands for diversified and personalized customer needs drive the need for continuous high-speed iteration of products. Among them, product conceptual design is the core process for meeting consumer needs and expectations. However, in the product conceptual design stage, data are distributed across sketches, 2D drawings, natural language texts, images, and other files, presenting characteristics of fragmentation, potential inconsistency, and multimodality. This complex data landscape can lead to challenges in accessing comprehensive and up-to-date knowledge for designers. To effectively improve designers' personal proficiency and abilities and break the knowledge isolation phenomenon, we constructed a top-down hierarchical structure of the product conceptual design domain knowledge model, achieving the formal expression of knowledge in the product conceptual design stage. First, a top-down product design domain knowledge graph (DDKG) construction method, including design feature analysis, relationship extraction, and concept definition, was proposed, resulting in the hierarchical ontology. Second, using techniques such as data analysis, knowledge extraction, and semantic similarity calculation, the automatic instantiation of the graph was realized and the DDKG was constructed. The DDKG provides three application modes: domain knowledge search, knowledge recommendation, and exploratory analysis. Finally, the proposed top-down hierarchical DDKG construction method was applied and verified in the high-end integrated home furnishing enterprise product conceptual design. We formed a unified core concept and terminology for the home furnishing conceptual design domain and constructed a knowledge service model, providing support for the intelligent transformation of the high-end integrated home furnishing design field.