Abstract: To monitor process industry’s production, the large history data is analyzed by clustering algorithm. The equipment’s parameters clustering models are built by Feature Weight’s K-means algorithm. The proportion between quantity under different similarity factor and the whole cluster is calculated by different similarity methods, and then a new factor of scatter is defined. Based on the conception of CBLOF(t), a new definition of outlier is brought forward to study the real-time outlier when the equipments circulate. Based on the models, equipments process and monitor faults can be optimized.

Key words: Clustering analysis, Feature weight’s K-means algorithm, Factor of scatter, Factor of outlier, Process industry

摘要： 采用数据挖掘中的聚类算法对流程企业的大量的历史数据进行分析，采用基于欧几里德距离的加权K-means算法建立了参数的聚类模型，分析簇团内不同相似度时的参数个数比例，得到参数点离核指数的定义。针对实时检测出的异常点，结合CBLOF(t)的概念，提出了一种新的离群指数的定义。以此为基础，有效地对设备的运行状况进行监控，从而起到设备运行优化和故障预警的作用。

关键词: 聚类分析, 加权K-means算法, 离核指数, 离群指数, 流程企业

CLC Number: 

• TP312