Date of Award

5-2016

Document Type

Thesis

Degree Name

Master of Science (MS)

Legacy Department

Electrical Engineering

Committee Member

Dr. Richard Groff, Committee Chair

Committee Member

Dr. William Harrell

Committee Member

Dr. Sarah Harcum

Abstract

Most biopharmaceuticals today are focused on the production of one of three major cell types: the bacterium Escherichia coli, yeasts (Saccharomyces cerevisiae, Pichia pastoris) and mammalian cells (Chinese Hamster Ovary cells). Growth opti-mization is a major focus as this dictates the pace of advancements in drug manu-facturing. The process involved in producing these cells itself is very complex and modeling a system to accurately capture these characteristics can be difficult. The overall process is expensive to run and repeated testing of various control algorithms to optimize growth can prove to be very time consuming as well. In order to develop control strategies and improve the yield of protein, it is beneficial to model a system that captures the responses of the bioprocess. The model can be coupled with different controllers to test the yield output and determine the most effective control strategies without incurring additional costs or time delays. Model parameters are determined by the process of numerical minimization, making use of experimentally obtained data to ensure accurate simulation system behavior. Additionally, a separate system can be developed to switch between the simulation platform and the actual process, with the same control strategy being implemented to compare against results of the simulation and the actual process. This allows for further adjustments to be made to more effectively model the bioprocess. This thesis describes the implementation of the Xu model, found in literature as the simulation counterpart to an experimental hardware setup. A hardware-in-the-loop simulation is developed with the ability to accurately model system parameters against experimentally obtained results in order to carry out control strategy testing on the simulation side before switching to the experimental hardware side. Accurate parameter estimation is achieved by fitting simulation results to experimentally logged data to ensure the simulation replicates the behavior of the physical system, and is subsequently verified against non-training data.

Share

COinS