Experimental procedure

Yeast strain, Saccharomyces cerevisiae, produced by Goodman Fielder Milling & Baking N.Z. Ltd. was grown in Yeast Extract, Peptone and Dextrose (YEPD) medium [97] with the following composition: Dextrose, 20g/L; Yeast extract, 10g/L; Peptone, 20g/L and commercial anti-foam, 10 drops/L. The starter culture was created in a shaker at 30° C and 200RPM for 60 to 90 minutes.

The reactor vessel and initial medium were prepared prior to fed-batch fermentation. 1.4 liter YEPD of liquid was added to the vessel as an initial medium. The complete assembly of vessel, head-plate, pH probe and DO probe were sterilized together with the initial medium by autoclave at 110° C for 20

minutes. The pH probe (Ingold Electrodes Inc.) was calibrated before the sterilization; while the DO probe (Mettler-Toledo Process Analytical, Inc.) was calibrated after the sterilization.

After the starter culture, 100 mL of pre-culture was inoculated from flasks into the vessel of bench-top fermentor. The temperature, agitation speed, airflow and pH were controlled at the nominal values of 30°C , 800RPM, 3L/min and 4.5 respectively. Nutritive substrate was automatically added into the bioreactor according to the predetermined feed rate trajectory. Silicon tubes (HV-96400-16 Precision silicone (peroxide) Tubing, Masterflex, USA) with inside diameter 3.1mm were used to feed the nutrients to the reactor. The feed rate was pre-calibrated before the fermentation was started. Three different types of trajectories were used in the experiments: constant feed rate, square-wave feed rate and stair-shape feed rate. The total feed volume was 1L.

Biomass concentration was measured off-line, which involved measuring the wet weight of yeast after centrifuging (Eppendorf centrifuge, Germany) 10mL broth samples for 10 minutes at 4500rpm and decanting the supernatant liquid. DO in the bioreactor was monitored by the oxygen electrode and the data was stored in a database in the computer. The fermentation volume was obtained by solving Equation B.6 as shown in Appendix B, and no measurement was needed.

The fermentor was operated for five to eight hours. Medium samples were taken approximately every 18 minutes to determine biomass concentrations, while DO values were monitored every minute. Three sets of data were collected. For the cultivation with constant feed rate, 14 biomass samples were obtained during a five-hour fermentation run; For the cultivations with square-wave and stair-shape feed rates, 27 biomass samples were measured during an eight-hour fermentation run. The experimental data are shown from Figure 4.7 to Figure 4.9. These three data sets were used for re-training the neural network, validating the network being trained and testing the prediction ability of the trained network respectively. Due to the infrequent sampling of biomass and unequal sampling time between DO and biomass, an interpolation method was needed to process the biomass data. To preserve the monotonicity and the shape of the data, a piecewise cubic interpolation method [98] was adopted here.

0 50 100 150 200 250 300 350 400 450 500

0 50 100 150 200 250 300 350 400 450 500

0 50 100 150 200 250 300 350 400 450 500

0 50 100 150 200 250 300 350 400 450 500

O o o OOqOOOO

0 50 100 150 200 250 300 350 400 450 500 Time (min)

Fig. 4.7. Square-wave feed.

Fig. 4.8. Constant feed.

0 50 100 150 200 250 300 350 400 450 500

0 50 100 150 200 250 300 350 400 450 500

0 50 100 150 200 250 300 350 400 450 500

0 50 100 150 200 250 300 350 400 450 500

0 50 100 150 200 250 300 350 400 450 500

Time (min)

Fig. 4.9. Stair-shape feed.

0 0

Post a comment