基于加权有向图的零件聚类模型研究  

Weighted directed graph based part clustering model

作　　者：赵刚[1] 江平宇[1] 

机构地区：[1]西安交通大学机械制造系统工程国家重点实验室

出　　处：《计算机集成制造系统》2006年第7期1007-1012,1017,共7页Computer Integrated Manufacturing Systems

基　　金：教育部新世纪优秀人才计划资助项目(NCET-04-0928)。~~

摘　　要：为解决e-制造单元规划及逻辑重构问题,对产品族零件的聚类问题进行研究。在深入分析产品族零件生产中所涉及的工序序列、生产准备时间、内部运输成本率等要素的基础上,提出了一种基于加权有向图的工艺描述数学模型,并以工艺合并后零件加工成本变化最小为标准,提出了一种基于关联矩阵二阶合并方法的零件相似距离模型,进而设计出相应的蚁群算法模型。最后,利用Java技术开发了一套软件系统,通过一个实例分析,验证了上述模型和算法的合理性和有效性。To solve problems of planning and reconfiguration for e--manufacturing cells, the parts clustering problems of Product Family （PF） were studied. Through analyzing some key elements in the production of PF parts, i. e. process sequence, preparation time of production and internal transportation cost etc. , a part process description model based on Weighted Directed Graph （WDG） was presented. More importantly, according to the criterion of lowest fluctuation of manufacturing cost, a part similarity distance model based on a method of two--grade combination of incidence matrixes was put forward. Referring to these mathematical models, a clustering system for manufacturing of PF parts based on ant colony algorithm was designed and developed. Finally, a software system was developed using Java technology. Analysis of clustering case has demonstrated that this system could rationally and effectively solve the PF parts clustering problems in e--manufacturing system for Mass Customization （MC） mechanism.

关 键 词：E-制造 大规模定制 加权有向图 聚类分析 蚁群算法 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]