基于SOM的多粒度云制造资源组合推荐  

Multi-Granularity Cloud Manufacturing Resource Combination Recommendation Based on SOM

作　　者：邹元昊 赵晓东[1] ZOU Yuanhao;ZHAO Xiaodong(College of Electronic and information Engineering,Tongji University,Shanghai 201804,China)

机构地区：[1]同济大学电子与信息工程学院,上海201804

出　　处：《武汉大学学报（理学版）》2021年第6期555-560,共6页Journal of Wuhan University:Natural Science Edition

基　　金：国家重点研究计划(2019YFB1706401)。

摘　　要：在云制造过程中,细粒度的资源会使设计任务的匹配变得困难,难以满足云制造请求方对于制造任务的要求。目前实现方法中,多粒度资源组合推荐主要着重于分析资源性质,并从架构设计的角度解决多粒度资源组合的问题,而未从算法角度提供解决方案,泛化性能较差。针对这样的问题,提出基于自组织映射(self-organizing map,SOM)聚类分析的多粒度云制造资源组合推荐方式。该方法首先通过对请求方制造资源调度日志进行聚类分析,将制造资源按照QoS指标分为不同的类型;然后采用滑动窗口分析统计各种类型的资源调度方式,计算不同资源调度方式在整个资源调度过程中所占的比例,进而得出请求方在制造过程中常用的调度组合,以此作为向请求方推荐的资源组合;最后通过模拟实验的方式验证了本文方法时间消耗少于架构的固定组合。In the process of cloud manufacturing, fine-grained resources will make it difficult to match design tasks and meet the requirements of cloud manufacturing requestors for manufacturing tasks. In the current implementation methods, multi-granularity resource combination recommendation mainly focuses on analyzing the attribute of resources and solving the problem of multigranularity resource combination from the perspective of architecture design, but does not provide a solution from the perspective of algorithm, so the generalization performance is poor. To solve this problem, this paper proposes a multi-granularity cloud manufacturing resource combination recommendation method based on self-organizing mapping(SOM) cluster analysis. Firstly,through cluster analysis of the requester’s manufacturing resource scheduling log, the manufacturing resources are divided into different types according to the QoS index. Then the sliding window is used to analyze and count various types of resource scheduling methods, calculate the proportion of different resource scheduling methods in the whole resource scheduling process,and then obtain the scheduling combination commonly used by the requester in the manufacturing process, which is used as the resource combination recommended to the requester. Finally, simulation experiments show that the time consumption of the method is reduced compared with the fixed combination of the architecture.

关 键 词：云制造 SOM聚类 资源组合推荐 多粒度资源 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]