Equipment fault detection by parametric identification - application to chemical reactor

摘要

Equipment faults in industrial chemical reactors lead to variations in product quality. Due to the highly energetic potential of the reaction mass, they can in addition cause runaways or major incidents. The methodology for detecting process anomalies in a chemical reactor, is based on a parametric estimation-prediction of the reaction mixture. The recursive least square method with forgetting factor has been retained to appreciate the evolution of parameters in real time. Thus allowing us, to point out a criterion of decision based on relative prediction error between two successive vector parameters.