ROBUST STATE ESTIMATION UNDER UNCERTAINTY IN A COPOLYMERIZATION PROCESS

ASTEASUAIN MARIANO; SANCHEZ MABEL CRISTINA; TUPAZ PANTOJA JHOVANY

Bahia Blanca

Congreso; IX Congreso Argentino de Ingeniería Química - CAIQ2017; 2017

Asociación Argentina de Ingeniería Química - Planta Piloto de Ingeniería Química

Polymerization processes are highly nonlinear systems that require a strict control of their dynamic operation in order to be competitive. Nonlinear controllers based upon rigorous models make use of the on-line state estimates obtained from the available measurements. The Unscented Kalman Filter has shown a rewarding performance for nonlinear state estimation. This filtering approach commonly assumes that the uncertainty is an additive noise. However, this is not a good approximation in some cases. Besides, noise may contain outliers that distort the quality of state estimates. Methods based on Robust Statistics have been proposed to deal with the effect of outliers. Until now, only Huber type M-estimators have been used as the loss function of the estimation problem. This work analyses the ability of other types of M-estimators to improve the robustness of the estimate. The performances of the M-estimators within the framework of a filtering technique based on the Unscented Transformation are compared for a copolymerization process. The filtering technique uses an augmented state vector formed by the state variables and both process and measurement noises.