The Digital Twin can map the starch hydrolysis, the cultivation of S. cerevisiae
and the whole-cell biocatalysis of ethyl (S)-3-hydroxybutyrate from ethyl acetate in a
small scale STR (Biostat C, 20 L, B. Braun). The development of the “SSF-BC
Simulator” was carried out using the procedure described in Sect. 2.2. The integrated
dynamic mathematical model was written in C++ and was implemented in WinErs
[20, 65]. Using the Digital Twin, it is possible to accelerate the simulation of the
bioprocesses up to 100-fold. The Digital Twin can be monitored and operated via the
GUI shown in Fig. 2.
The GUI in Fig. 2 presents the process equipment (e.g. reactor, feed tanks, etc.) as
well as all measured value displays (e.g. temperature, pH value, DO, etc.) and all
essential functions of the control system (e.g. temperature or DO control) to the user
of the Digital Twin. Behind each measured value display or control button,
sub-models represent the real measuring or control instrument. The reactor properties and the biological process are mapped in the dynamic mathematical model of the
Digital Twin. The GUI is part of the control and automation model within the Digital
Twin. To use the Digital Twin for the development, optimisation and realisation of
control strategies, it is therefore important that the GUI corresponds to the PCS of the
real process with high similarity.
4.1.1 Parameterisation of the Digital Twin “SSF-BC-Simulator”
For the parameterisation of the dynamic mathematical process model implemented
in the Digital Twin “SSF-BC-Simulator”, a variety of parameterisation experiments
were carried out, using batch and fed-batch cultivations.
The procedure of model parameterisation will be illustrated using a dataset from a
laboratory experiment where an aerobic fed-batch cultivation was carried out in a
small scale STR (Biostat C, 20 L, B. Braun). The temperature was controlled at
30
C, the pH value at 4.5 and the DO at 10%. At the beginning of the cultivation, a
nutrient medium was supplied in the STR (Batch medium). After the batch phase of
the cultivation, a fed-batch nutrient medium was fed to the STR (see Table 4).
During the cultivation process, the following state variables required for process
monitoring and process control were measured (see Table 5).
After the experiment was carried out, the model of the Digital Twin “SSF-BCSimulator” was parameterised using the Nelder-Mead simplex algorithm, written in
R [67], to adjust the values of selected parameters to match the simulated with the
measured data satisfactorily.
Figures 3 and 4 present the measured state variables of the fed-batch S. cerevisiae
cultivation in a small scale STR compared to the simulated time courses of the
Digital Twin (after parameterisation).
Figure 3 shows that in the batch phase of the experiment (0–7 h), glucose was
consumed. Ethanol (EtOH) was formed, which was subsequently metabolised again
(diauxic growth). The biomass density shows a slight increase during the batch
phase. After the substrate feed has been activated (7–25 h), the dry biomass
concentration increases to a value of more than 30 g L
À1 . At a processing time of
78
C. Appl et al.
and the whole-cell biocatalysis of ethyl (S)-3-hydroxybutyrate from ethyl acetate in a
small scale STR (Biostat C, 20 L, B. Braun). The development of the “SSF-BC
Simulator” was carried out using the procedure described in Sect. 2.2. The integrated
dynamic mathematical model was written in C++ and was implemented in WinErs
[20, 65]. Using the Digital Twin, it is possible to accelerate the simulation of the
bioprocesses up to 100-fold. The Digital Twin can be monitored and operated via the
GUI shown in Fig. 2.
The GUI in Fig. 2 presents the process equipment (e.g. reactor, feed tanks, etc.) as
well as all measured value displays (e.g. temperature, pH value, DO, etc.) and all
essential functions of the control system (e.g. temperature or DO control) to the user
of the Digital Twin. Behind each measured value display or control button,
sub-models represent the real measuring or control instrument. The reactor properties and the biological process are mapped in the dynamic mathematical model of the
Digital Twin. The GUI is part of the control and automation model within the Digital
Twin. To use the Digital Twin for the development, optimisation and realisation of
control strategies, it is therefore important that the GUI corresponds to the PCS of the
real process with high similarity.
4.1.1 Parameterisation of the Digital Twin “SSF-BC-Simulator”
For the parameterisation of the dynamic mathematical process model implemented
in the Digital Twin “SSF-BC-Simulator”, a variety of parameterisation experiments
were carried out, using batch and fed-batch cultivations.
The procedure of model parameterisation will be illustrated using a dataset from a
laboratory experiment where an aerobic fed-batch cultivation was carried out in a
small scale STR (Biostat C, 20 L, B. Braun). The temperature was controlled at
30
C, the pH value at 4.5 and the DO at 10%. At the beginning of the cultivation, a
nutrient medium was supplied in the STR (Batch medium). After the batch phase of
the cultivation, a fed-batch nutrient medium was fed to the STR (see Table 4).
During the cultivation process, the following state variables required for process
monitoring and process control were measured (see Table 5).
After the experiment was carried out, the model of the Digital Twin “SSF-BCSimulator” was parameterised using the Nelder-Mead simplex algorithm, written in
R [67], to adjust the values of selected parameters to match the simulated with the
measured data satisfactorily.
Figures 3 and 4 present the measured state variables of the fed-batch S. cerevisiae
cultivation in a small scale STR compared to the simulated time courses of the
Digital Twin (after parameterisation).
Figure 3 shows that in the batch phase of the experiment (0–7 h), glucose was
consumed. Ethanol (EtOH) was formed, which was subsequently metabolised again
(diauxic growth). The biomass density shows a slight increase during the batch
phase. After the substrate feed has been activated (7–25 h), the dry biomass
concentration increases to a value of more than 30 g L
À1 . At a processing time of
78
C. Appl et al.