关键词: 表面贴装, 知识图谱, 加权余弦相似度, XGBoost

Abstract: In the multi-variety and small-batch production mode of the surface mount industry, process management faces prominent challenges including scattered process knowledge, difficult reuse of practical experience, and low efficiency of parameter matching. This paper proposes an intelligent SMT parameter recommendation method based on knowledge graph. Firstly, a PCB feature-oriented process knowledge graph is constructed to realize the structured modeling and logical correlation of fragmented process knowledge. On this basis, a high-similarity process sample set is screened via weighted cosine similarity. Combined with the XGBoost regression model and knowledge graph threshold constraints, the mapping relationship between PCB features and SMT process parameters is established to achieve accurate recommendation of process parameters. Experimental results show that the debugging and verification cycle of process parameters is shortened from 4 hours to 0.25 h, and the qualification rate of process parameters is improved from 88.5% to 99%. The proposed method effectively addresses the strong experience dependence and poor reusability of traditional process design. It is applicable to flexible electronic product manufacturing scenarios and provides a technical support for the intelligent design of SMT processes.

Key words: surface mount technology, knowledge graph, weighted cosine similarity, XGBoost