Failure Prediction for an Exothermic Semi-batch Reactor via A combined EKF with Statistical Method

Abstract

Early failure detection for an exothermic semi-batch polymerization reactor is investigated in this paper. The extended Kalman filter (EKF) is used to estimate the system state from reactor nonlinear dynamics via input/output data. Then, a statistical method is employed to detect early system fault. The decision-making is made by a hypothesis testing through a generated innovation sequence. The reactor is a multivariable nonlinear dynamic process and is subjected to several major disturbances. A mathematical model is developed for the reactor with some model parameters identified from the input/output data, and then the developed continuous model is discretized into a discrete model. Being detected in this work are three faults on three sensors and one on the actuator. These fault are simulated on the reactor and are detected using the developed method. Simulation results are given.