4.1.2 Digital Twin “SSF-BC-Simulator” for the Development of Control
Strategies
To ensure that the Digital Twin is suitable for the development of control strategies
for the cultivation of S. cerevisiae, it must be able to represent the time courses of the
experimental data described in Figs. 3 and 4. These time courses do not have to be
simulated exactly, but the associated effects must be reproduced. For the development of the RQ feedback control strategy utilising the Digital Twin, it is important
that exhaust gas measurements, RQ value time course and associated effects can be
mapped. For the development of the OLFO controller with the Digital Twin, it is
necessary to simulate the course of the concentrations of substrate, product and
biomass and the corresponding effects.
Figure 3 shows that the time course of the measured variables can be mapped by
the Digital Twin with a high agreement. Also, ethanol formation due to the Crabtree
effect can be represented by the Digital Twin (0–3 h and 25–33 h). It is also clearly
recognisable that high ethanol concentrations inhibit the growth of the cultivated
S. cerevisiae strain in the simulation (30–52 h).
Figure 4 illustrates that the time courses of the measured exhaust gas values can
almost be exactly reproduced by the Digital Twin. Also, in the simulation, an
increase in the RQ value occurs if ethanol is formed due to the Crabtree effect
(0–3 h and 25–33 h). Furthermore, at the end of the simulated cultivation, almost no
CO 2 is formed or O 2 is consumed, corresponding to a low growth rate.
The results presented in Figs. 3 and 4 illustrate the high potential of the Digital
Twin for the development of an RQ feedback control strategy and an OLFO strategy
for the cultivation of S. cerevisiae. In the presented study, the control target was to
maximise the dry biomass concentration (S. cerevisiae). To achieve this target, it is
important to dose the substrate feed in such a way that the cells are sufficiently
supplied with glucose. However, overdosing substrate may lead to ethanol formation
(Crabtree effect), which then might cause growth inhibition.
4.2 Digital Twin Based Development of Control Strategies
for the Cultivation of S. cerevisiae
During process control strategy development, the different strategies were first
applied to the “SSF-BC-Simulator”. Simulations with varying controller designs
and tunings were then carried out on the Digital Twin until the desired controller
performance was achieved. Afterwards, the experimental validation of the control
strategies on the real plant took place. If the control result was still unsatisfactory,
further controller improvements were tested using the Digital Twin, before validating the controllers on a real cultivation process. By using the Digital Twin, many
complex experiments in the STR with elaborate preparation, execution and analysis
could be avoided in the development of the control strategies, which resulted in a
82
C. Appl et al.
Strategies
To ensure that the Digital Twin is suitable for the development of control strategies
for the cultivation of S. cerevisiae, it must be able to represent the time courses of the
experimental data described in Figs. 3 and 4. These time courses do not have to be
simulated exactly, but the associated effects must be reproduced. For the development of the RQ feedback control strategy utilising the Digital Twin, it is important
that exhaust gas measurements, RQ value time course and associated effects can be
mapped. For the development of the OLFO controller with the Digital Twin, it is
necessary to simulate the course of the concentrations of substrate, product and
biomass and the corresponding effects.
Figure 3 shows that the time course of the measured variables can be mapped by
the Digital Twin with a high agreement. Also, ethanol formation due to the Crabtree
effect can be represented by the Digital Twin (0–3 h and 25–33 h). It is also clearly
recognisable that high ethanol concentrations inhibit the growth of the cultivated
S. cerevisiae strain in the simulation (30–52 h).
Figure 4 illustrates that the time courses of the measured exhaust gas values can
almost be exactly reproduced by the Digital Twin. Also, in the simulation, an
increase in the RQ value occurs if ethanol is formed due to the Crabtree effect
(0–3 h and 25–33 h). Furthermore, at the end of the simulated cultivation, almost no
CO 2 is formed or O 2 is consumed, corresponding to a low growth rate.
The results presented in Figs. 3 and 4 illustrate the high potential of the Digital
Twin for the development of an RQ feedback control strategy and an OLFO strategy
for the cultivation of S. cerevisiae. In the presented study, the control target was to
maximise the dry biomass concentration (S. cerevisiae). To achieve this target, it is
important to dose the substrate feed in such a way that the cells are sufficiently
supplied with glucose. However, overdosing substrate may lead to ethanol formation
(Crabtree effect), which then might cause growth inhibition.
4.2 Digital Twin Based Development of Control Strategies
for the Cultivation of S. cerevisiae
During process control strategy development, the different strategies were first
applied to the “SSF-BC-Simulator”. Simulations with varying controller designs
and tunings were then carried out on the Digital Twin until the desired controller
performance was achieved. Afterwards, the experimental validation of the control
strategies on the real plant took place. If the control result was still unsatisfactory,
further controller improvements were tested using the Digital Twin, before validating the controllers on a real cultivation process. By using the Digital Twin, many
complex experiments in the STR with elaborate preparation, execution and analysis
could be avoided in the development of the control strategies, which resulted in a
82
C. Appl et al.