一种基于群体智能聚类的设备性能横向比较算法.pdfVIP

上海交通大学学报 Vo l. 40 No. 3 第 40 卷第 3 期 Mar. 2006 2006 年 3 月 JOURNAL OF SHANGHAI ]IAOTONG UNIVERSI丁Y 文2在蝙寄:1006-2467(2006)03-0439ω05 种基于群体智能聚类的设备性能横向比较算法 张蕾1 ，曹其新1 ，李杰2 (1.上海交通大学机器人研究所，上海 200030; 2. 戚斯康星大学智能维护中心，密尔沃基 WI 53224 ) 摘 要:提出了一种感于群体智能的设备性能横向比校算法，该算法将设备模式投影于二维平面 上，然后依据群体智能聚痪，实现设备性能的自组织聚类分析.为了提高群体智能聚类算法的运行 效卒，提出了利用.i.成分分析改总模式投影时的随机性.此外，为了减小参数选取对算法的影响，提 出了一种史简单的相似皮衡量E公式.将改进后的算法与原算法进行比较，结果显示，改进算法的运 行效率更高.利用改进算法对某中央空调的数据进行分析，验证了算法能满足设备性能自组织聚类 的妥求. 关键询:智能维护;聚类算法;主成分分析 中固分棠号:丁P 18 文献析、识码 :A An Algorithm for Comparing Machine Performance 8ased on Swarm Intell igence Clustering ZHANG Lei 1 , CAO Qi-xiη 1 ， LEE]αy2 O. Research Inst. of Robotics , Shanghai Jìaotong Univ. , Shanghai 200030 , China; 2. Research Center of Intelligent Maintenance Systems , The Univ. of Wisconsin , Milwaukee WI 53224 , USA) Abstract: An algorithm for assessing the machine performance based on swarm intelligence was proposed. The patterns representing the machine performance are first mapped into a plane , then analyzed by the clustering algorithm based on the 日warm intelligence. 丁。 improve the running efficiency of the algorithm. the stochastic mapping of the patterns was modified based on principal component analysis. Moreover , to reduce the influence on the algorithm caused by the parameters s伫lection ， a simple way to calculate the similarities of the patterns was presented. Finally , the modified algorithm was compa