Exploitation Of Minimum Entropy Loss Based Cyclohexane Oxidization Process Monitoring Technique

Chemical industry process are usually carried out under extreme conditions, such as highpressure and high temperature. The chemical materials are usually explosive, toxic andcorrosive. Fault or abnormal conditions in the process may lead to great loss. It is necessary toemploy process monitoring method to keep process safety, since the process monitoring canensuring the success of the planed operations by recognizing anomalies of the behavior.Cyclohexane uncatalyzed oxidization to produce cyclohexanone is a high risk process becausethe materials in the process are toxic and explosive, explosion will occur when the concentrationof oxygen in the exhaust emissions reach explosion range. In this paper, a real industrycyclohexane oxidization process are used as research object, the aim of this work is to developa method can detect abnormal conditions in the process effectively. Compare with other study,this research focus on to solve information loss problem in the data transformation process, wehave developed a novel principle component select strategy based minimum informationentropy loss.In this paper, there are three part. The first part introduced the statistics control limit, whichis computed using kernel density estimate method in this work. Statistic process monitoringtechnique determine fault by comparing statistics of current process data from DCS with thecontrol limit determined by normal process data. In order to use more information in thestatistics, a two dimension control limit is developed, this control can determine fault earlierthan traditional one. The second part introduced the minimum information entropy loss basedprincipal component select strategy, according to this novel principal component select strategyand character of process data we developed kernel entropy component analysis based, kernellocality preserving projection based and MEL-KLPP based process monitoring method. All thetechniques developed are applied to simulated TE process and an industry oil process,monitoring result shows the validity and characteristics of monitoring method. In the last part,the cyclohexane oxidization process is used to examine the monitoring performance ofdeveloped method, comparing different monitoring technique by fault detect rate, CPU timeand sensitivity of algorithm parameter. KECA is determined as the monitoring method of cyclohexane oxidization process due to its high fault detect rate and insensitive of kernelparameter.