Modelling Approaches for Bio-Manufacturing Operations - Measurement, Monitoring, Modelling, and Control of Bioprocesses - page 89

noted above, their acceptance in the bioprocessing industry has been notably
slower [ 30 , 37 ] and the technique remains underdeveloped for the description of
industrial-scale bio-manufacturing processes due to a number of factors. One key
challenge lies in performing process optimisation in the face of complex and often
multi-variable design problems [ 45 ]. Models aimed at assisting with these must
cope with the use of mixed operating modes (e.g. batch and semi-batch), strong
process interactions and run-to-run process variability with complex feeds [ 47 ].
Although the basic theory and fundamental equations share at least some com-
monality with the chemical engineering literature for a range of unit operations,
naturally one must account for the specific nature of bioprocess problems. Com-
plex and uncertain biophysical phenomena can mean additionally that unit oper-
ations may be only poorly characterised, and the use of different modes i.e. both
discrete pass and continuous processing [ 26 ] can complicate modelling efforts.
Often, the complex nature of bioprocess feedstocks can mean that physical
property data of adequate quantity and quality may be absent and default
assumptions need to be made. This can reduce the predictive power of a model. In
such situations, empirical or semi-empirical models may be needed, with experi-
mental data used to calibrate the equations. Even with such approaches, in many
cases, the properties of such material can vary within and between studies, com-
plicating attempts to quantify process behaviour. Process materials can sometimes
be highly sensitivity to even small changes in the design of a manufacturing unit or
facility. Added to this complexity is the wide variety of product types that are now
considered for bio-manufacturing processes, including proteins, nucleic acids and
whole cells, for which properties may be unknown or be highly variable from one
batch to the next. Where default values are used for selected mass transport
properties, this can make it difficult to construct valid, robust simulations. This
lack of available process data is one of the key stumbling blocks when starting to
create a model, and in such cases, it is necessary to acquire information from as
many sources as possible, including known expertise in the literature, accepted
industrial best practices, pre-existing development studies and full-scale manu-
facturing experience. In the latter case, a key problem is that data are often highly
correlated or co-linear, meaning that it can fail to cover a wide search space and
will instead focus upon selected combinations of operating conditions alone. In
such cases, it may be necessary to develop qualified scale-down process mimics in
order to acquire the additional information ( Chap. 7 ).
1.4 Technical Versus Business Modelling
Historically, the task of modelling the technical performance of manufacturing
processes has been treated as separate from examining the business aspects. This
has resulted in a gap between models that use sometimes quite sophisticated