Abstract: With the wide applications of advanced manufacturing technologies such as Computer Aided Design(CAD),Computer Aided Manufacturing(CAM)and Computer Aided Quality Inspection(CAQI)in the product manufacturing process,the companies have accumulated a large amount of process data(including the inspection quality results).How to effectively utilize these process data to identify the key processes in the process plan of the parts to be manufactured has become a key issue for ensuring the manufacturing quality of the parts.However,the existing key processes identification methods mainly depend on the process designers and have not effectively learned the correlations between the process plans and manufacturing quality of parts in the existing process data.For this problem,a key processes identification method that integrated process-quality grammar knowledge for complex parts was proposed.A multi-level structured process model based on machining features was constructed to represent the effective correlations between the three-dimensional geometry of the part,process plan and quality inspection results in the process data.Based on the structured process data,a working step-quality pair prediction model was constructed by using Bayes learning method to calculate the confidence of different working steps-quality pairs at different times in the process plan of the parts,and a compound Quality-Process Knowledge And-Or Graph model(Q-PK-AOG)was established.Given the process plan of the parts to be manufactured,according to the confidence of different working steps-quality pairs at different times,the working step quality of the parts was predicted based on grammar parsing guided by Q-PK-AOG to realize the identification of key processes in the process plan of the parts.Taking three-axis CNC milling parts as the research object,a prototype system based on CATIA was developed,and the effectiveness of the proposed method was verified through experiments.

Key words: key processes identification, Bayes learning, quality-process knowledge and-or graph, grammar parsing

摘要： 随着以计算机辅助设计(CAD)/计算机辅助制造(CAM)/计算机辅助质量检验(CAQI)为代表的先进制造技术在产品制造过程中的广泛应用,企业积累了大量的工艺数据(包括与之关联的质量数据),如何有效利用这些工艺数据实现待制零件工艺方案中关键工序的识别成为保证零件加工质量的一个关键问题。但是,已有的关键工序识别方法主要依赖工艺设计人员,未能有效学习已有工艺数据中零件的工艺方案与加工质量之间的关联关系。提出一种融合工艺质量语法知识的复杂零件关键工序识别方法。首先,构建基于加工特征的多层次结构化工艺模型,实现工艺数据中零件三维几何与工艺方案、质量检测数据之间的有效关联。其次,根据结构化工艺数据,采用统计学习方法,构建工步质量对预测模型,从而计算零件工艺方案中不同时刻的工步质量对的置信度,并且建立具有复合性的质量工艺知识与或图模型(Q-PK-AOG)。最后,给定待制零件的工艺方案,以Q-PK-AOG为引导,根据不同时刻工步质量对的置信度,基于语法解析实现待制零件工艺方案中工步质量的预测,从而实现零件工艺方案关键工序的识别。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,通过实验验证了所提方法的有效性。

关键词: 关键工序识别, 贝叶斯学习, 质量工艺知识与或图, 语法解析