Construction and application of a multimodal knowledge graph for machining deformation uncertainty analysis
针对加工变形不确定性分析中机理复杂、信息分散等问题，构建了融合文本、图表和有限元仿真信息的多模态知识图谱，并设计了基于大模型增强的智能系统，为全生命周期决策提供支持。
Uncertainty analysis of machining deformation is crucial for improving product quality, reliability, and manufacturing efficiency. Current machining deformation uncertainty problems still face many challenges, such as complex mechanism, scattered information, and high dependence on expert experience. In order to solve the aforementioned problems, the paper constructs a multi-modal knowledge graph for machining deformation uncertainty analysis, and designs a large model-augmented intelligent system driven by a multimodal knowledge graph. The constructed multi-modal knowledge graph realises the multi-integration of text information, chart information and finite element simulation information, and covers key links such as ontology construction, knowledge extraction, knowledge fusion, knowledge inference, and knowledge storage. Aiming at the knowledge inference link, the principal component analysis (PCA) and multi-head self-attention mechanism (MHSA) are innovatively introduced to enhance feature extraction, proposing a new knowledge inference method. Additionally, through multi-stage large model enhancement strategies and expert large model training approaches, the generalisation ability and multimodal interaction capability of the intelligent system have been effectively improved. Finally, a multimodal knowledge graph of machining deformation uncertainty is constructed, and an intelligent system is developed to provide multimodal decision support for machining deformation uncertainty analysis from a full life-cycle perspective, with its performance validated.