Optimization and Control of Fed-Batch Bioreactors
 

R. Mahadevan and F.J. Doyle III
University of Delaware
Department of Chemical Engineering
Newark, DE 19716
mahadeva@che.udel.edu
(302) 831-0726
 

Biochemical processes are becoming increasingly important not only to 
manufacture traditional fermentation products such as pharmaceuticals 
but also a variety of other products. Typical biochemical processes in 
the industry are carried out in the fed-batch mode. Thus the area of 
dynamic optimization of fed-batch fermentation process is an active 
area of research. One of the main difficulties associated with 
fed-batch bioreactor optimization and control is that significant 
deviations from the nominal operating conditions occur frequently. This 
causes the open-loop feed policy to be suboptimal. Thus it is necessary 
to estimate the key model parameters for maintaining model accuracy. 
Lack of sensors in these processes has also lead to the use of  'soft 
sensors' based on Kalman filters to estimate the unmeasurable states 
and the parameters which are used in the optimization. Once the 
parameters are updated, the optimal trajectory has to be recomputed 
on-line. Efficient algorithms for on-line optimization of bioreactors 
will be presented. These are then used as part of a nonlinear model 
predictive controller to maximize the desired product.