ARMAX modelling and state estimation of an enzyme fermentation process

Abstract

The suitability of input/output models for state estimation of fermentation processes has been investigated. A batch glucoamylase fermentation provides an example and a relatively simple ARMAX model was used to estimate, on-line, both the enzyme activity and the biomass concentration from ammonia addition and carbon dioxide evolution measurements, respectively. The model parameters were estimated by the recursive least-squares method. Model fit and estimator performance were improved by signal conditioning. The estimator was capable of estimating the state of the process starting from the same initial parameter values and off-line measurements could be used readily for updating the estimator parameters thereby further improving the estimator performance.