Automated Measurement and Monitoring of Bioprocesses: Key Elements of the M3C Strategy - Measurement, Monitoring, Modelling, and Control of Bioprocesses - page 14

3.2.2 Gas Balancing Techniques
A very similar situation holds for the gas balancing techniques. Gas analysis is
quite straightforward (see Sect. 2.2.5 ), and the aeration rate is usually also well
known. A simple gas balance over the reactor yields quantitative gas consumption
and production rates (independent of the oxocaloric information). Knowing the
working volume of the reactor as well allows one to derive the gas transfer rates
that a reactor is able to achieve under given operating conditions very accurately,
precisely, and simply. Since these all represent rate information, one can also
deduce the specific growth rate of a growing culture from these signals (or soft
sensors): provided the CO 2 production is strictly growth associated—an assump-
tion that very often holds true—and chemisorption of CO 2 is not a great problem
(due to a nonalkaline and well-controlled pH value), a linear regression of the
natural logarithm of the rate versus time reveals the specific growth rate as the
slope. All these calculations are simple arithmetic and should be implemented on
process controllers or computers—for the time being, this information source is
left unexploited (exceptions may confirm this rule). We also recommend inte-
grating the CO 2 production rate over time (numerically) in order to calculate the
carbon recovery in every bioprocess, provided the amounts or concentrations of
other ''important'' carbon-containing reactants are known. If the carbon recovery
does not match 100 ± a few percent, then there are actually ''important'' con-
tributors (substrates or byproducts) that are simply not known; this fact—by
itself—is ''important,'' as it reveals a poor understanding of the process which,
according to the PAT initiative, should be urgently overcome.
Gas, redox, charge, elemental, and proton balancing methods have been
exploited for online data reconciliation, along with predictions derived from online
infrared and dielectric spectroscopic datasets and offline calibrated models [ 47 ].
The data reconciliation algorithm could be directly implemented into the predic-
tion algorithms of the online spectrometers.
3.2.3 In Situ Fluorescence Monitoring
Fluorescence monitoring at a distinct wavelength (couple: one specific excitation
and one specific emission) or with scanning extensions (so-called two-dimensional
fluorescence monitoring) is of course less expensive than FCM but yields a pop-
ulation average only. For many purposes, this may be sufficient and desirable, as
one can follow growth, product formation, or physiological changes of a popu-
lation extremely rapidly and noninvasively. The interpretation of the resulting
signals deserves some care, however, as illustrated by the disaster of one pio-
neering fluorescence sensor, built specifically for NAD(P)H (both components that
are essential for every living cell); however, the conclusion (and marketing
argument) that such a sensor would monitor the biomass concentration was false.
Indeed, every living cell contains NADH and NADPH, but depending on the
balance state of growth, the ratio between the reduced and oxidized forms can