Modelling Approaches for Bio-Manufacturing Operations - Measurement, Monitoring, Modelling, and Control of Bioprocesses - page 95

expressions become more complicated, e.g. in chromatography, the mass transport
phenomena are often represented by a complex series of non-linear partial dif-
ferential equations that may require access to powerful PCs equipped with
appropriate numerical solving software. Hence simplifying assumptions can often
be useful in reducing model complexity. This becomes especially true if one seeks
to link models of different unit operations in a process flowsheet, since the com-
putational time can otherwise become significant.
The current generation of bioprocess simulations produced by academic
research groups or end-users themselves have focussed predominantly on indi-
vidual unit operations alone [ 16 , 34 ], often involving the solution of complex
equations to evaluate material balances such as for fermentation [ 7 , 13 , 20 ],
homogenisation [ 38 , 44 ], primary recovery such as filtration or centrifugation
[ 6 , 28 , 36 , 37 ] and chromatography [ 12 , 24 ]. By definition, models of individual
unit operations fail to consider the likely interactions between process steps, and
although they constitute a useful first step toward the production of robust and
predictive modelling packages, it is the simultaneous consideration of all opera-
tions and replication of the interactions between them that is needed for the
assessment of whole process feasibility. Unit operations do not operate in isola-
tion, and even small changes at any given stage can affect operation there and
further downstream dramatically. Optimising individual operations separately with
respect to objectives defined for one stage alone runs the risk that performance of
the overall process may well be suboptimal [ 16 ]. Thus linkage is vital for quan-
tifying trade-offs between successive process steps, and this can be done most
easily by simulations. Examples of integrated process modelling are provided by
Groep et al. [ 16 ] and Varga et al. [ 41 ], who used models to investigate how, for
example, varying the number of homogeniser cycles affected technical and eco-
nomic outcomes during the recovery of an intracellular protein (alcohol dehy-
drogenase) produced by yeast fermentations.
3.2 Computational Fluid Dynamics
Computational fluid dynamics (CFD) modelling provides the capability to produce
an in silico imitation of the hydrodynamic environment within large-scale bio-
processing equipment. This can enable one to quantify phenomena such as shear
stresses at solid-liquid or gas-liquid interfaces. Such information may be neces-
sary in the context of design space mapping because small-scale devices on their
own may be inadequate for mimicking the hydrodynamic large-scale environment,
and hence any predictions made may need to be adjusted. A classic example of
this is in centrifuge bowls operated in a non-flooded manner. Small-scale rotors
tend to invoke far less powerful shear than at commercial scale, where forces
exerted as fluids impinge upon solid surfaces such as in the feed zone can damage
delicate materials including cells or shear-sensitive precipitates. Thus some form
of regime analysis is necessary to quantify large-scale shear, e.g. by using CFD to