Developing Control Strategies to Regulate Dissolved Oxygen Levels in a Biological Fermenter

Abstract

Vaccine development comprises multiple stages, the first of which typically involves cultivating an organism in a microbial fermenter to produce a vaccine product. In order to ensure the optimal synthesis of the vaccine product, it is necessary to maintain adequate control of the dissolved oxygen (DO), which is required for the organism to grow and survive. Our work is concerned with controlling the dissolved oxygen in a biological fermenter using a PID (Proportional-Integral-Derivative) Controller.

The product from this fermenter is used to create the immunization for a medical illness. However, the present configuration of the PID Controller is inadequate for maintaining the dissolved oxygen at the desired level of 35% relative to saturation. This inadequacy results in violent DO oscillations which compromise the quality of the product. To solve this issue, we use open-loop experimental data to develop empirical transfer-function models of the control process for dissolved oxygen. Then, we create an optimization algorithm for the PID Controller and apply it to obtain the proportional, integral, and derivative gains that would best regulate the dissolved oxygen in the fermenter.

The parameters obtained from this algorithm are applied experimentally to the biological fermenter set-up and the results are used to demonstrate that the PID optimization algorithm provides controller settings which successfully regulate the dissolved oxygen. In addition, we employ our transfer-function models of the DO control process to design and configure a set of Internal Model Controllers and Model Predictive Controllers.

The Internal Model and Model Predictive Controllers are subsequently optimized to handle external disturbances and robustly regulate the dissolved oxygen levels. This optimization is performed by varying the tuning parameters of the controllers and selecting the parameters which best maintain the dissolved oxygen levels at their desired values, minimize the effect of external disturbances, and minimize the effect of errors in process modelling. Finally, a non-linear model of the DO control process is developed and utilized to successfully obtain a set of gain-scheduled PID tuning parameters.