Design of optimal experiments for improved modeling of food processes
Résumé
The identification of model parameters in food processes is both an experimental and a numerical challenge. The aim of the design of experiments is to optimize (tune) the experiments at the lowest experimental cost in order to get best identification results. In this work, an original approach and implementation of the sequential experiment design is presented. It combines multistart optimization, systematic reconditioning by hand of all unknowns being optimized or integrated, direct-differentiation method to compute sensitivities. This approach is validated over four applications, and was able to: (i) find again the optimal pharmacokinetic experiment from Pronzato (2008) (ii) find again the optimal fermentation experiments from van Derlinden et al (2008) (iii) compute three optimal experiments drying rice with varying temperature Goujot et al. (2012)