基于报警时序信息挖掘的报警关联分析方法  

Alarm Correlation Analysis Approach Based on Alarm Time Information Mining

作　　者：李俊杰 杨博[1] 李宏光[1] LI Junjie;YANG Bo;LI Hongguang(College of Information Science&Technology,Beijing University of Chemical Technology,Beijing 100029,China)

机构地区：[1]北京化工大学信息科学与技术学院,北京100029

出　　处：《信息与控制》2019年第1期29-34,共6页Information and Control

摘　　要：面对现代流程工业监控系统报警泛滥问题,为了寻找报警根源以减少无效报警,并针对传统方法在面对大量的报警数据时计算效率低下的问题,提出了一种基于报警数据时序信息挖掘的报警关联分析方法.通过将报警时序信息进行区块化处理,将报警时间序列转换为报警时间的节点序列,然后将区块之间的匹配度作为报警关联度的评价标准,减少了关联分析的运算量;采用滑动窗口比对计算相关报警时间序列的时间关联信息;采用TE过程(Tennessee Eastman process)数据,验证了所提方法的有效性.Alarm flooding is recognized as a serious problem in modern process industry.To find root alarms and reduce invalid alarms so as to tackle the low efficiency of the traditional methods,we propose an alarm correlation analysis approach based on alarm time information mining.Taking advantage of the block method of alarm time series,we convert the alarm data sequences to the time node sequences,and then we utilize the matching degree between the temporal blocks as the evaluation criterion of the alarm correlations,which can reduce the calculation burden of the correlation analysis.Additionally,we calculate the temporal information associated with the correlated alarm time series through contrasting sliding windows.The Tennessee Eastman(TE)process are employed to illustrate the benefits of the proposed method.

关 键 词：过程报警 时间序列 关联分析 数据挖掘 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]