Identification and control of dissolved oxygen in hybridoma cell culture in a shear sensitive environment

The productivity of mammalian cells can be enhanced by facilitating adequat e oxygen transfer into the cultivation medium. However, current methods of controlling dissolved oxygen (DO) fail to account for alterations in medium composition during the course of the fermentation. These changes, which di rectly affect gas solubility and overall mass transfer coefficient, may be significant and deteriorate controller's performance in the long run. In th is paper, the applications of Generalized Predictive Controllers (GPC) to D O control were investigated in a shear sensitive environment and compared t o PID and Model Predictive Controllers (MPC). Input and output data for sys tem identification were initially generated by varying the composition of o xygen fed into the bioreactor from 0 to 0.21 mol % while keeping the total inlet gas flow rate at 8.75 vvm. The process was identified using an AutoRe gressive model with eXogeneous inputs (ARX) model and tested on different d ata sets. The model parameters were then correlated with the overall mass t ransfer coefficients. In simulation tests, the output of the PID controller switched from minimum to maximum values while more continuous control sign als were obtained with the MPC and GPC controllers. When tested in a cell-f ree medium, all three controllers were able to track setpoint changes with some chattering observed in the control signals. The GPC outperformed the M PC and PID controllers when applied to the cultivation of hybridoma cells.