Abstract
Advanced process control theories are usually tested in a simulated environment. This article shows how a non-classical control algorithm can be implemented in a discrete fashion on a real biological process with satisfactory results. Kalman filters and model predictive controllers were implemented to delay starvation-induced apoptosis in CHO cell cultures. Apoptosis was mostly caused by deprivation of glutamine and asparagine in the medium. The concentrations of the state variables were estimated every 15 hours by forward integration of the system equations. The off-line measured concentrations of viable cells, lactate, and glucose were used to update the state estimates. Neural Network and Kalman filter techniques were then used to approximate the concentration of apoptotic cells in the bioreactor based on the concentrations of viable cells, glutamine and asparagine. This information was then fed to a model based predictive controller that was activated when the apoptotic cell estimate reached a concentration of 0.1 million cells per ml. This resulted in improved protein productivity and a reduced final apoptotic cell concentration.
Original language | English (US) |
---|---|
Pages (from-to) | 1579-1584 |
Number of pages | 6 |
Journal | Proceedings of the American Control Conference |
Volume | 2 |
DOIs | |
State | Published - 2002 |
Event | 2002 American Control Conference - Anchorage, AK, United States Duration: May 8 2002 → May 10 2002 |
All Science Journal Classification (ASJC) codes
- Electrical and Electronic Engineering