Abstract: In order to extract the process knowledge from the process data in the Computer Aided Process Planning (CAPP), a knowledge discovery method for typical process sequence based on clustering analysis was presented. A mathematical model using the data matrix was built to describe the process sequence. Based on the operation order coding, the similarity between two operations was measured by the Manhattan distance. The similarity between two process sequences was calculated by the Euclidean distance and the dissimilarity matrix was used to indicate dissimilarity between process sequences. The similarity between two clusters based on the dissimilarity matrix was evaluated by the average distance method,and the process sequence clusters were eventually merged by the agglomerative hierarchical clustering method. Finally，the clustering result was determined by the clustering granularity of process sequence. This method has been applied successfully to discover the typical process sequence of a kind of axle sleeves.

Key words: computer aided process planning, typical process sequence, clustering analysis, knowledge discovery

摘要： 为解决计算机辅助工艺设计系统从工艺数据中提取工艺知识的问题，提出了应用聚类分析法获取典型工艺路线的方法，建立了以矩阵表达工艺路线数据的数学模型。在工序编码的基础上，应用曼哈坦距离计算工序间的相似度。应用欧氏距离计算工艺路线间的相似度，并以相异度矩阵表示工艺路线的相异度。通过平均距离法计算工艺路线簇间的距离，并应用凝聚的层次聚类法进行工艺路线聚类。最后，通过工艺路线聚类粒度的确定方法确定聚类结果，并以轴套类零件典型工艺路线发现为例，验证了该方法的有效性。

关键词: 计算机辅助工艺设计, 典型工艺路线, 聚类分析, 知识发现