Perturbation-based Control of Industrial Fed-batch Bioprocesses

Detta är en avhandling från Department of Automatic Control, Lund Institute of Technology, Lund University

Sammanfattning: The topic of this thesis is bioprocess control, more specifically control of industrial-scale microbial fed-batch bioprocesses. Its focus is therefore on methods which are easy to implement in an industrial setting, which gives certain limitations on sensors, actuators and control systems.

The main part of the work in the thesis concerns control of the microbial
substrate uptake rate by manipulation of the feed rate of liquid substrate to the bioprocess. This is an important parameter for improving process yields, as too low feed rates cause starvation of the microorganisms while too high rates lead to production of undesirable by-products. By-product formation decreases metabolic efficiency and the by-products have inhibiting effects on microbial growth and production. At high concentrations these can even halt growth completely, leading to process failure.

Due to large batch-to-batch variations and the complexity of the pro-
cesses, model-based control can be difficult to use in this type of system. The approach used in this thesis circumvents this problem by utilizing perturbations in the feed rate. It has previously been shown that the metabolic state with regard to substrate uptake rate can be determined by analysing the perturbation response in the dissolved oxygen level of a microbial process.

In this thesis, the concept is developed through the use of perturbations at a predefined frequency. This provides a number of advantages and allows for estimation of the metabolic state through observing the perturbation frequency in the measured signal.

The concept has been tested experimentally in industrial pilot and pro-
duction scale. It has been demonstrated that a controller based on this
concept can be used to compensate for batch-to-batch variations in feed demand and can rapidly compensate for changes in the demand. It has also
been shown that the method can be used for monitoring and control in
bioprocesses with a volume over 100 m3, using a low-complexity estimation
algorithm suited for industrial use.

The thesis also concerns mid-ranging control in non-stationary processes.
A modified mid-ranging controller suited for such processes is proposed, which allows control signals to increase in unison during the course of a
fed-batch process while maintaining the advantages of classical mid-ranging control. The concept can for instance be used for control of dissolved oxygen, an important process parameter in many bioprocesses. It has been successfully used for this purpose in pilot scale alongside the type of perturbationbased feed rate controller which is the main topic of this thesis, also showing how the latter can be used in conjunction with other control systems.