基于传递熵与改进K2算法生成报警传播网络①柯永琦(南京邮电大学通信与信息工程学院,南京210003)通信作者:柯永琦,E-mail:1020010127@njupt.edu.cn摘要:工业报警变量数量增多所导致的“报警泛滥”问题,严重影响了报警系统的应有功能.针对此问题,提出一种从过程报警数据集中学习报警变量传递关系的方法.首先,利用传递熵具有准确衡量一阶或多阶自相关性变量间因果关系的特点,识别变量间的因果关系.其次以变量间熵大小为准则,保留传递信息量较大的节点,最后考虑变量在不同状态所占时间比重对K2算法进行改进,学习得到最终的报警传播网络.通过在田纳西伊斯曼过程数据集上的验证发现,该方法能够对报警发生的根本原因做出判断并较好地学习得到报警传播路径.关键词:传递熵;改进K2算法;报警泛滥;报警传播网络引用格式:柯永琦.基于传递熵与改进K2算法生成报警传播网络.计算机系统应用,2023,32(2):258–265.http://www.c-s-a.org.cn/1003-3254/8982.htmlGenerationofAlarmPropagationNetworkBasedonTransferEntropyandImprovedK2AlgorithmKEYong-Qi(SchoolofTelecommunicationsandInformationEngineering,NanjingUniversityofPostsandTelecommunications,Nanjing210003,China)Abstract:Theproblemof“alarmflooding”causedbytheincreaseinthenumberofindustrialalarmvariableshasseriouslyaffectedtheproperfunctionsofalarmsystems.Inresponse,thisstudyproposesamethodtolearnthetransferrelationshipofalarmvariablesfromtheprocessalarmdataset.First,leveragingtheabilityoftransferentropytoaccuratelymeasurethecausalrelationshipbetweenvariableswithfirst-orderormulti-orderautocorrelation,thestudyidentifiesthecausalrelationshipbetweenvariables.Second,dependingontheentropyvaluebetweenvariables,thenodesthattransmitalargeamountofinformationarereserved.Finally,theK2algorithmisimprovedbyconsideringthetimeproportionsofvariablesindifferentstates,andthefinalalarmpropagationnetworkisobtainedbylearning.TheverificationontheTennesseeEastmanprocessdatasetrevealsthatthemethodcanjudgetherootcauseofthealarmandwellachievealarmpropagationpathsthroughlearning.Keywords:transferentropy;improvedK2algorithm;alarmflooding;alarmpropagationnetwork工业报警的产生与报警阈值、过程测量信息和操作生效时间等因素相关,其提供...
