Abstract: Aiming at the existing process planning methods for parts mainly relying on process engineers,and the generated process schemes difficult to satisfy the manufacturing logics of the part in an industry,an effective Numerical Control (NC) machining process planning method via integrating machining processes of machining features was proposed.The Process Knowledge And-Or Graph (PK-AOG) was constructed by introducing a context-free grammars for its compositional properties,which formed the solution space for the process scheme of the part,and the working step sequence of a part was a parse graph of PK-AOG essentially.Then,the implicit mapping modes between machining features and feature process labels were learned by using the deep learning method based on attention mechanism for the structured process data,and the probability distributions of different feature process labels were calculated.Finally,using the PK-AOG as a guide,the optimal process scheme of a part that was composed of working step sequences,its compatible feature processes and the machining process of each machining feature was jointly searched via integrating Ant Colony Algorithm (ACO) and Genetic Algorithm (GA),and thus a process scheme of a part with correct semantics and logics was generated.A prototype system based on CATIA was developed to verify the effectiveness of the proposed method.

Key words: data-driven, knowledge-guided, process knowledge and-or graph, integration, machining process joint inference

摘要： 针对已有的零件宏观工艺设计方法主要依赖于工艺设计人员,生成的数控加工工艺方案难以与零件的实际工艺过程相符等问题,提出了一种融合加工特征工艺的零件数控加工工艺生成方法。该方法首先引入具有复合性的上下文无关语法构建工艺知识与或图。每个零件的工步序列均是工艺知识与或图中的一个解析图。因此,工艺知识与或图本质上构成了零件数控加工工艺的搜索解空间。然后,从结构化工艺数据中,通过基于注意力机制的深度学习零件的特征与特征工艺标签的映射模式,计算不同特征工艺标签的概率分布。最后,以工艺知识与或图为引导,融合蚁群算法与遗传算法从工艺知识与或图中联合推理迭代搜寻零件的工步序列,与其相融的特征工艺,以及每个加工特征的特征工艺,从而获得符合逻辑、语义准确的数控加工工艺方案。以三轴数控铣削加工零件为研究对象,开发了一个基于CATIA的原型系统,通过实验验证所提方法的有效性。

关键词: 数据驱动, 知识引导, 工艺知识与或图, 融合, 工艺方案联合推理