基于聚类方法的CAPP零件知识库构建

A Clustering Based Method for the Knowledge Base Construction of CAPP

传统的零件分类一般根据零件编码从特征矩阵中得到分类结果,未能很好地表达各个零件之间的相似关系,对工艺设计也不能提供启发性的推理策略.此外,零件分类矩阵本身的相似性标准也难以确定,给零件的工艺制作带来了很多困难.为此,提出了一种利用聚类技术构造树型结构表达零件相似性的方法,并根据零件之间的相似性建立层次结构以进行动态分类,进而构建一种有自学习能力的零件知识库.考虑到机器智能的局限性,分类结果可能不尽合理,分类树又能够在自动压缩优化的基础上进行手工优化,并将优化结果记录于分类树中.该知识库能及时反映零件信息的动态更新,并对零件进行多层次、细粒度的动态分类,使零件分类粒度不受数据规模的限制,从而可以通过建立索引结构,实现自适应的工艺设计自动化.

Traditional parts classification results are acquired from feature matrices according to their coding. However, the method cannot express well the degree of similarity among the parts or provide a heuristic reasoning scheme for process design; furthermore, the similarity criteria themselves of the matrices are hard to determine. These bring many difficulties to the manufacturing process of the parts. This article proposes a method of describing similarity using clustering analysis, builds a hierarchical structure for the feature similarities among the parts to facilitate dynamic clustering and automatic process making, and makes a thorough study of the problem of parts clustering. Thus a knowledge base with self-learning ability is constructed, which may arrange parts more suitably, represents their inherent relationship of features and functions more accurately, and reflect immediate updating of parts.