Research On Fault Diagnosis Method Of Chemical Process Based On Information Entropy And Transfer Entropy

With the rapid development of computer technology and the large scale of modern chemical plants,a lot of process monitoring data are often obtained in chemical production.It has become a new research hotspot that how fault features are extracted quickly and effectively from these big data to monitor the process to ensure the safety of production.Therefore,a method of chemical process fault diagnosis based on information entropy and transfer entropy is presented in this paper.Firstly,the chemical process monitoring model using the relevant information entropy is established.The entropy residuals of the related information between the real-time condition and the normal condition are calculated,and when the residuals exceed the threshold,it is judged as abnormal condition.Because only the strong correlation variables need to be monitored and calculated,the monitoring efficiency is improved.Then,for abnormal condition,the characteristic of arbitrary data relation can be evaluated to extract the features of process data and remove redundant variables by using information entropy.The transfer entropy is introduced to analyze the transfer relationship between two variables,and the fault propagation path is obtained to describe the fault mechanism.Meanwhile,the sensitive variables in the fault propagation path are extracted as the characteristic variables of the fault and the alarm value is set.Because each of root cause of the fault will have corresponding a set of characteristic variables characteristic variables,the root cause database of the fault is established according to the sequence of the characteristic variable alarm.In online monitoring,when the abnormal condition is detected,the fault type is identified by comparing with the characteristic variable alarm set in the fault diagnosis database established above,that is,the root cause of the fault is diagnosed.Finally,the Tennessee Eastman(TE)process is taken as an example.The results show that the real-time abnormal conditions of complex chemical processes based on information entropy and transfer entropy can be quickly and effectively monitored by using the relevant information entropy.The database between root cause of the fault and fault characteristics variables established by the method can provides a reliable basis of making decision for both operators and managers.