monoclonal antibodies manufacturing in a fed-batch bioreactor using economic nonlinear model predictive control with extended kalman filter

Monoclonal antibodies (mab) are used in various medical applications including targeted treatment of autoimmune diseases and cancer. since mab are typically produced in fed-batch reactors, it is economically desirable to maximize the amount of mab produced in each batch, thereby reducing the number of batches or the batch volume required to meet a certain market demand. in this work, an economic nonlinear model predictive control (nmpc) strategy for online steering of the fed-batch operation toward maximum mab production is developed. the nmpc features final-time constraints on the reactor volume and bound constraints on the inlet glucose and glutamine flow rates as the manipulated variables. unlike previous work by the authors, the process states (e.g., component concentrations) are assumed to be unmeasured in the current work, as is the case in real applications. the only measured variable is the reactor volume. therefore, an extended kalman filter (ekf) is designed to estimate the process states and initialize the nmpc prediction model at each sample time. the bioreactor model is implemented in openmodelica. simulation results show the economic nmpc yields, depending on the ekf initialization, 9% to 21% increase in the mab production at the end of the batch compared with the nominal operation at constant feed flow rates.