STRATEGIES FOR OPTIMAL
DISSOLVED OXYGEN (DO) CONTROL
R. Mikler, W. Kramer, O. Doblhoff - Dier, K. Bayer
Institut für Angewandte Mikrobiologie
Universität für Bodenkultur
Nußdorfer Lände 11 A- 1190 Wien
Abstract
The exponential growth of biomass and the frequent changes of environmental conditions (fluid characteristics, addition of antifoam, etc.) create problems in DO control of fermentations in lab and pilot scale. Traditionally DO is controlled by application of conventioanl control algorithms, as PID, connected to a control cascade. The performance of the control loop depends on the oxygen consumption. For DO loop optimization of E. coli fermentation empirical adaption of tuning parameters and fuzzy tuning was applied.
Keywords: cascade control, fuzzy control, loop tuning
INTRODUCTION
Batch culture of single cell microorganisms can typically be divided into growth phases referred to as lag phase, exponential phase, declination (late log) phase, stationary phase and death phase. A number of aberrations, such as the diauxic growth of organisms (sequential utilisation of two different substrates) can be observed, but in principle biomass will increase during batch culture in a more or less exponential way and growth will slow down and stop on substrate limitation. The control of dissolved oxygen in microbial fermentations creates problems on both laboratory and pilot/production scale. This is most frequently due to:
- The increasing over all oxygen demand with increasing biomass
- Changing mass transfer coefficient due to changes of fluid characteristics by metabolic products
- Foaming at high aeration rates and high stirrer speed
The control of oxygen is even more complicated if it is controlled by both airflow or oxygen concentration in the aeration gas and stirrer speed. In this case traditional control strategies would normally employ proportional integral differential (PID) algorithms connected to a control cascade (fig. 1). As the control loop is dependent on the overall oxygen consumption, which increases dramatically with exponentially proliferating biomass, optimal tuning of the controller will be linked to oxygen uptake rate.
Software strategies for process dependent tuning of DO controller
Empirical PID tuning In a typical fed batch fermentation (10 l working volume) the oxygen transfer rate is about 0,5g O2/h at the beginning and increases to about 30-40g O2/h at high density of biomass at the end of fed batch fermentation. Due to this wide range of OTR the parameters of the PID controller (KP, TV, TN) have to be adapted to actual conditions to attain optimum control of DO during the whole fermentation. PID parameter adaption was achieved by optimization at an average oxygen transfer rate of about 5-10 g O2/h according to Ziegler and Nichols (1942). On the basis of these optimized parameters adaption of KP valid for other ranges of OTR was achieved by multiplication of optimum Kp with the ratio of flowact to flowopt" (flowopt represents the flow at which optimum parameters were evaluated). In order to avoid instability of the controller the actual flow (flowav) is averaged during a period of one to several minutes. This strategy has been used successfully in many cases.
Fig. 2: DO control loop applying optimized PID parameters
Fuzzy PID tuning
The second approach to achieve more stable control of dissolved oxygen, was based on the fuzzy tuning of a conventional PID loop. Expert knowledge was used to adjust the PID tuning parameters in dependance to the oxygen consumption, estimated by the average air flow. Figure 3 shows the structure of the control loop. In the figures 4-7 the membership functions of the input and output parameters are shown. Oxygen input was additionally controlled by increasing rpm in relation to average flow (rpm = % average flow * 1.64). 1% of flow corresponds to 0.05 l air /min or 0.01 vvm at a fermentation volume of 5 l.
Fig. 3: Fuzzy tuning of KP, TN, TV
Fig. 4: Fuzzification of air flow (averaged flow du- ring 10 min)
The following set of rules is based on the fact, that the tuning parameters for pO2 PID control change with increasing aeration (average) flow rate
RESULTS AND DISCUSSION
Fig. 5: Evaluation of KP
Fig. 6: Evaluation of TN (sec.)
Fig. 7: Evaluation of TV (sec)
Fig. 8a: DO Control during batch fermentation of E.Coli in exponential growth
Fig. 8b: Performance of fuzzy controller at set-point shift
in exponential growth phase
With this approach acceptable dissolved oxygen control could be achieved. A number of different strategies, such as the fuzzy control of dissolved oxygen without the use of conventional PID algoritms has been tested and will be published elsewhere. The big advantage of the PID fuzzy tuning is the relative simplicity of rules and membership functions. Especially for industrial fermentations this approach may be used for a relatively safe performance enhancement, as it can be performed by an external computer and tuned parameters can be downloaded to the industrial controller. Parameters can be double checked to stay within oparational limits to increase operational safety.
Acknowledgments: The technical support by Omron Austria was very much appreciated
REFERENCES
Kramer, W., G Elmecker, R. Weik, D. Mattanovich and K. Bayer (1994). Kinetic studies for the optimization of recombinant protein formation, Annals of the New York Academy of Sciences, in press
Ziegler, J.G. and N.B. Nichols (1942). Optimum Settings for Automatic Controllers. Trans. ASME 64, 759
GO TOP
04/07/95 by IAM
