OPTIMAL EXPERIMENTAL-DESIGN FOR PRACTICAL IDENTIFICATION OF UNSTRUCTURED GROWTH-MODELS

In this paper, optimal experimental design for parameter estimation of unstructured microbial growth models during growth of biomass on a si ngle limiting substrate in a fed-batch bioreactor is considered. The r atio of the largest to the smallest eigenvalue of the Fisher informati on matrix (i.e,, the modified E-criterion for optimal experimental des ign) is used to evaluate the information content of several simulation fed-batch experiments, each with a different volumetric feed rate pro file. The construction of optimal feed rate profiles is based on the f ollowing Conjecture: A feed rate strategy which is optimal in the sens e of process performance, is an excellent starting point for feed rate optimization with respect to estimation of those parameters with a la rge influence upon process performance. The optimal value of 1 for the modified E-criterion is obtained for several feed rate profiles with different structures, after optimization of their corresponding degree s of freedom with respect to the information content of the experiment . For each profile a criterion evaluating the violation of model valid ity is calculated. (C) 1998 IMACS/Elsevier Science B.V.