针对双机器人钻铆工艺参数依赖人工经验、工艺知识碎片化导致的智能化决策难题，提出一种融合知识图谱与规则推理的双机器人钻铆智能参数推荐方法。首先，构建双机器人钻铆工艺多层知识图谱，融合工具、工件、工艺、特征、实例层5层的292个实体与711条关系，实现钻铆材料属性、工艺参数、质量指标的多模态语义关联；接着，提出一种两阶段推理方法：通过AHP法定权的加权余弦相似度算法匹配相似案例，结合基于优先级的规则引擎进行钻铆机器人参数约束与冲突消解，提升双机器人钻铆推荐参数的准确性与一致性；进而，提出一种基于数据与规则约束的动态置信度计算模型，构建双机器人钻铆的数据规模、质量、分布及规则冲突四维评估方法，基于动态权重置信度量化评估实现双机器人钻铆智能参数推荐结果的可解释性。最后，在中国商飞上海飞机制造有限公司的双机器人钻铆工作站的实验表明，该方法推荐参数综合置信度达0.93，且生成参数符合工艺物理约束，满足工艺要求。本文所提出的方法为航空制造机器人场景的工艺知识沉淀与智能化决策提供了可行的解决方案。

To address the challenge of intelligent decision-making in dual-robot drilling and riveting, which is often hin-dered by reliance on manual expertise and fragmented process knowledge, this study proposes an intelligent parameter recommendation method that integrates knowledge graphs with rule reasoning. A multi-layer knowledge graph for dual-robot drilling and riveting is first constructed, comprising 292 entities and 711 rela-tionships across five layers—tool, workpiece, process, feature, and instance—to enable multimodal semantic associations among material properties, process parameters, and quality indicators. Building on this, a two-stage reasoning framework is introduced: similar cases are retrieved through a weighted cosine similarity algo-rithm, where the weights are determined using the Analytic Hierarchy Process, and a priority-based rule engine is applied for parameter constraints and conflict resolution, thereby improving the accuracy and consistency of parameter recommendations. Furthermore, a dynamic confidence evaluation model is developed by incorporat-ing data- and rule-based constraints, establishing a four-dimensional assessment along the axes of data scale, data quality, data distribution, and rule conflicts. This dynamic weighting mechanism enhances the interpreta-bility of the recommendation results. Experimental validation on a dual-robot drilling and riveting work-station at COMAC Shanghai Aircraft Manufacturing Co., Ltd. demonstrates that the proposed method achieves a comprehensive confidence level of 0.93, with the recommended parameters satisfying physical process con-straints. Overall, this method provides an effective approach for capturing and reusing process knowledge while supporting intelligent decision-making in complex manufacturing scenarios.