Asymptotic observers as a tool for modelling process dynamics

Abstract

This paper deals with the dynamical modelling of processes with a special emphasis on wastewater treatment. Dynamical process modelling is usually derived from mass balance considerations and results in models which contain two types of terms representing the hydrodynamics and transport dynamics, and the conversion due to the involved (bio)chemical reactions involved, respectively. The conversion terms are indeed generally speaking the products of two terms: a yield coefficient and a reaction rate (strictly speaking, the kinetics). The latter usually concentrates a predominant part of the model uncertainty. The bioprocess model building exercise is usually global in the sense that it concerns all the terms involved in the model. In addition to the inherent bioprocess modelling difficulties (model complexity, data scarcity), this makes the interpretation of the model identification results difficult if not questionable. In this paper it is suggested how to use asymptotic observers to model and validate reaction schemes independently of the reaction kinetics. The results will be illustrated with two wastewater treatment processes: an activated sludge process and a detoxification reactor.

Observer design for differential-algebraic model of an aerobic culture of a recombinant yeast

“…The latter is a dynamical system that combines a nominal model of the process with online measurements of the input and the output of the process to estimate states and/or parameters. In bioprocess engineering, many algorithms have been developed in the literature [2,3,4,5,6,7].…”