부분최소제곱법 모델의 파라미터 모니터링을 이용한 화학 공정의 이상진단 모델 개발에 관한 연구

Abstract

It’s not easy to build the proper process models for monitoring, controlling the process conditions and getting results by existing data, information of the process. So there were many and various statistical approaches which could monitor the faults of process as well as process conditions. But the statistical models would have critical drawback that models could provide wrong results due to the fault magnitude even though same faults. This study suggests that parameter monitoring approach not the outputs of the process model. Once some faults occur in the chemical process, these lead changes of each outputs of process unit and operation conditions. Then, these also make new variables which are distorted with the model of normal conditions. These changes can be summarized to see the pattern of distorted parameter, called fault monitoring map. The predictive models are constructed by Partial Least Square method with selected variables as most influent to the process. In this study, the Tennessee Eastman (TE) process is used to verify the efficiency of the theory. And the cumulative summation (CUSUM) is applied to improve the performance of fault monitoring.