Figure 3.3 Main sources of knowledge
usually available for developing a
bioprocess model.
model. This knowledge spectrum (Figure 3.3) ranges from mechanistic knowledge to
information hidden in process data records. People in industry seem to be attracted by
this methodology for the reason that heuristic knowledge and rules of thumb may be
incorporated directly in the process model. It should however be stressed that these
concepts face a certain criticism, mainly in that the substitution of mechanistic kinetic
models by black-box models has obvious risks in affecting model robustness.
Knowledge and expression of knowledge
A variety of information sources is normally available on biotechnological cultivation
processes (Schubert et al., 1994a; Lübbert and Simutis, 1994). Three main types of
knowledge can be identified:
- Mechanistic (phenomenological) knowledge: this kind of knowledge is usually
represented by mathematical models. This is the classical approach followed by
chemical and biochemical engineers for developing their process models mentioned
previously. It has the highest level of sophistication, involving the understanding of
the basic transport mechanisms and kinetics. These mechanisms are often poorly
understood or even completely unknown. Therefore this kind of knowledge is usually
the one available in minor quantities. - Heuristic knowledge and common sense: this kind of knowledge is more qualitative
than the former, being usually available in larger quantities in the industrial
environment. The Fuzzy theory is used to manipulate this type of information. It
provides methods for quantifying qualitative knowledge. Heuristic knowledge is often
stated in terms of rules of thumb. These can be readily represented by the so-called
knowledge-based systems such as fuzzy inference systems and expert systems (e.g.
Sugeno, 1985; Kosko, 1992; Wang, 1994). - Knowledge hidden in the process data acquired during process operation: in many
situations the available mechanistic and/or heuristic knowledge is not sufficient to
New methodologies for multiphase bioreactors 3 71