摘要：AbstractThis paper addresses the problem of maximizing biomass productivity and, indirectly, monoclonal antibody production, of a fed-batch hybridoma cell culture. A dynamic model of the process is used by a model predictive algorithm to determine the optimal feed rate. As only a few component concentrations can be measured on line, an unknown input observer is used to estimate the reaction rates. This observer is based on a multimodel decomposition of the original model, and uses a minimum variance (Kalman-like) approach. The overall control scheme exhibits satisfactory performance, in face of model uncertainties.
