Continuous cultures with cell retention by an internal membrane filter were carried out using tapioca hydrolysate as medium to produce ethanol. Tapioca hydrolysate is one of cheap industrial media used widely in domestic alcohol production factories. In this study, the resulting mixture after liquefaction and saccharification of tapioca, contained 210~240 g/L of reducing sugars, of which about 92% was glucose and about 8% was maltose. In a total cell retention culture operated at a dilution rate of 0.18 /$h^{-1}$, the yeast concentration, the residual reducing sugar concentration, the ethanol concentration, and the volumetric ethanol productivity were about 40 g/L, 15 g/L, 81.4 g/L, and 14.7 g/L-h, respectively. In another cell retention culture operated at a dilution rate and a bleed rate of 0.20 /$h^{-1}$ and 0.14, respectively, the yeast concentration increased to 22 g/L and the ethanol concentration oscillated around 68 g/L. The volumetric ethanol productivity was about 13.6 g/L-h and the residual reducing sugar concentration about 12 g/L containing glucose of about 4.5 g/L. It was observed that maltose could not be consumed at all due to the repression by the residual glucose in all of the continuous cultures carried out using tapioca hydrolysate.
Batch cultures were carried out for several initial concentrations of glucose and maltose to observe more carefully the repression by glucose of maltose utilization, a kind of catabolite repression, and to obtain necessary data sets for modelling. A model for ethanol production from a glucose-maltose mixture has been proposed, with a term representing the glucose repression effect on maltose consumption. The kinetics of glucose repression was represented most appropriately by a saturation type function with a single parameter called repression constant in this study. The model parameters were estimated from batch experimental data mentioned above. The maximum growth rates were determined by a smooth cubic spline approximation. The maximum growth rate on glucose (0.401 /$h^{-1}$) was found to be considerably higher than that on maltose (0.353 /$h^{-1}$), while the ethanol yield from maltose (0.451) was slightly higher than that from glucose (0.429). The cell yield from glucose (0.109) was almost same to that from maltose (0.104). Monod constants for glucose and maltose, and the repression constant were determined based on goodness-of-fitting. Results of sensitivity analysis on these three parameters showed that a change of 10% in these parameters had no significant effect on data fitting.
It was thought that the dilution rate should be kept very low to maintain the glucose concentration very low when tapioca hydrolysate was used, which contained both glucose and maltose. Otherwise, it was expected that maltose could not be consumed due to the repression by glucose, making the substrate utilization very low. However, a very low dilution rate meant a low volumetric productivity of the process. To overcome such a difficulty, three types of two-stage processes were considered. Guide lines to determine optimal operation condition were suggested for them. The first one was consisted of a continuous stirred tank bioreactor, a buffer tank, and a secondary fed-batch fermentor. The second one was consisted of a continuous stirred tank bioreactor and a secondary fed-batch fermentor. The third one was consisted of a continuous stirred tank bioreactor and a plug flow bioreactor. From simulation results on these three systems using the model previously established, we could see that the substrate utilization was improved by having the secondary fermentor, while the overall volumetric productivity of the whole fermentor system decreased significantly. The second process showed the best performance in overall. But, a more careful economic study on fermentor installation cost and operation cost must be followed for a detailed and accurate optimization of the two-stage processes.
Finally, a software for on-line cell concentration monitoring, one of important tools for monitoring and realization of optimal operations of fermentation processes under consideration, was developed. A laser turbidimeter was used. A third-order correlation between the sensor signal, that is, optical density, and the cell mass concentration was constructed. The parameters of this correlation equation were found strongly depended upon agitation speed. A rather simple algorithm was proposed to remove measurement noise and to compensate for the effect of agitation speed. A user-friendly software package out of this algorithm was developed, which had capabilities of data acquisition, noise filtering, estimation, data storage, graphic display, and interaction with the user.
