Abstract: Aiming at the influence of multistage machining processes on the finishing tool wear,the interrelated coupling between the process factors and the inevitable data loss problems,a data mining method of incomplete information system based on dynamic hierarchical clustering and similarity relation was proposed.The correlated coupling factors were decomposed considering the correlation between different processes,and the independent factors were obtained.The continuous attributes were discretized by using dynamic hierarchical clustering method,and the importance degree of each influencing factor was calculated by using the attribute reduction algorithm of incomplete information system based on similarity relation.The incompatibility threshold of the decision table was set,and the optimal reduction attribute such as the key factors that influence finishing tool wear were obtained by adjusting the discretization result according to the approximation quality of the decision table.The practicality and validity of the proposed method were verified by an illustrative example of slotting cutter.

Key words: incomplete information system, data mining, rough set, multistage machining processes, tool wear

摘要： 针对实际生产中多道工序对精刀磨损存在影响,工序间的影响因素相互关联耦合,且不可避免地发生数据缺失的问题,提出一种基于动态层次聚类和相似关系的不完备信息系统数据挖掘方法。通过工序间的关联性将多工序关联耦合因素进行分解,得到相互独立的影响因素。利用层次聚类法将连续属性离散化,并采用基于相似关系的不完备信息系统属性约简算法计算各影响因素的重要度。设定决策表的不相容度阈值,并根据决策表的分类质量动态调节离散化结果获得最优约简属性,进而挖掘影响刀具磨损的多工序关键工艺要素。以轮槽铣刀为例进行了实验分析,验证了所提方法的实用性和有效性。

关键词: 不完备信息系统, 数据挖掘, 粗糙集, 多工序, 刀具磨损