Using Models to Design and Optimize an SSF Bioreactor

Figure 12.2 gives a more detailed view than Fig. 12.1 of how the design process should be carried out for production-scale SSF bioreactors, starting with the necessary laboratory-scale studies and ending with final optimization at large scale. It highlights the fact that it is ideally a process in which experimental and modeling work is undertaken simultaneously, with the mathematical model being refined constantly in the light of experimental evidence. The current section gives a broad overview of this bioreactor design process. It assumes that, after optimizing product formation by a particular organism on a particular solid substrate at laboratory scale, you have decided to develop a large-scale process.

12.2.1 Initial Studies in the Laboratory

Early studies will be needed in the laboratory to understand how the organism grows and how this depends on the environmental conditions that it experiences. On the basis of these studies, a growth kinetic model will be proposed (See Boxes 1 and 2 in Fig. 12.2). However, several questions must be asked before the experimental studies are planned. For example, what type of model will be used to model the growth kinetics? With what depth will it model the growth process? Will it simply describe biomass growth as a global value, such as g-biomass g-dry-solids-1, or g-biomass m-3? Or will it describe the spatial distribution of biomass at the particle level, for example, describing the biomass concentration as a function of height and depth above and below the particle surface? In answering these questions, it is important to consider that any decision that increases the complexity of the model may bring subsequent difficulties not only in solving it, but also in measuring all the necessary model parameters. These difficulties must be balanced against an evaluation of the potential advantages of improved predictive power that can be gained by describing the phenomena in greater detail. The appropriate level of detail for modeling growth kinetics within SSF bioreac-tors is discussed in greater depth in Chap. 13. It is only after these decisions have been made that the experimental program is planned. The experiments are planned

Fig. 12.2. Details of the strategy for using models as tools in the design and optimization of operation of SSF bioreactors

in such a way as to enable the development of mathematical expressions relating the growth rate to the various environmental variables. The way in which these experiments might be done and the types of mathematical expressions that might be used are described in Chaps. 14 to 17.

Ideally, various bioreactor types should be tested experimentally at laboratory scale, and, in fact, preferably at pilot scale, although few laboratories have sufficient resources to build laboratory-scale prototypes of all the possible bioreactor types, let alone pilot-scale prototypes (Fig. 12.2, Box 3). At the very least, experiments should be done in which some cultures are left static and others are submitted to various agitation regimes of different frequency, duration, and intensity. The results will be very useful in guiding bioreactor selection and determining the agitation regime to be used in the fermentation.

12.2.2 Current Bioreactor Models as Tools in Scale-up

Mathematical models have already been proposed for the various bioreactor types that are used in SSF. It makes sense to take advantage of these models, imperfect as they are (Fig. 12.2, Box 4). At this stage, it is quite likely that many of the parameters, such as transfer coefficients and substrate bed properties, will simply be based on literature values for similar systems. It may be appropriate to improve one or more of the models (Fig. 12.2, Box 5). Ideally laboratory-scale bioreactors should be operated in such a way as to mimic any limitations that will prevail at large scale, and the model predictions carefully validated against performance of these bioreactors. Disagreements between predicted and real performance should stimulate an investigation into the cause, which might be the mathematical form of the equations, but could also be the values used for some of the model parameters.

Simulations with the models will point to which bioreactor has the best potential to provide appropriate control of bed temperature and water content at large scale (Fig. 12.2, Box 6). Once a bioreactor has been selected, the appropriate model then represents a very useful tool for making decisions about design (e.g., geometric aspect) and operating conditions (e.g., air flow rate) (Fig. 12.2, Boxes 7 and 8). Careful attention must be given to the question as to whether the operating conditions necessary for good performance in the simulations are practical to achieve at large scale.

It is advisable to proceed to a scale that is intermediate between the laboratory scale and the final production scale, although this has not always been done. In any case, once a larger scale version of the selected bioreactor has been built, it is essential to validate the model again, since it is quite possible for the relative importance of the various heat and mass transfer phenomena to change with increase in scale (Fig. 12.2, Boxes 9 and 10). Phenomena that were not important at small scale and which were therefore not included in the model might suddenly become quite important at large scale. In this case the model will probably fail to describe large-scale performance with reasonable accuracy. If necessary, the model must be improved. Parameter values also must be determined with care. For example, it may be necessary to determine the bed-to-air mass transfer coefficient that is actu ally achieved within the production-scale bioreactor rather than to rely on estimates based on correlations given in the literature.

12.2.3 Use of the Model in Control Schemes

Once the bioreactor has been built with the help of the model, the model, improved in the light of data obtained at large scale, is still useful. It is highly likely that bioreactor performance will be significantly improved by implementing control strategies and the model can also play a useful role in the development of the control scheme (Fig. 12.2, Box 11). For example, the proposed control scheme can initially be tested and tuned with the model, which is obviously much cheaper than doing this initial testing and tuning with the bioreactor itself. The model may be embedded into the control system that is used to control the bioreactor.