are needed. In our opinion, there is a need to look for alternative approaches
which are easily applicable and better reflect the large-scale situation.
3. Due to the very high computing power, which is needed to solve these Euler–
Lagrangian models with a reasonable resolution, the current models cannot
describe the process dynamics over time, but only represent a very narrow time
point of the cultivation. Nevertheless, this kind of simulations, e.g. if they are
performed at different time points of a cultivation can provide important information to plan scale-down experiments.
4. Current models only consider the liquid phase. The implementation of the gas
phase would additionally need much computing power and in our opinion it
would be very laborious to validate these models.
5. An important characteristic of living systems in connection with their adaptation
to the environment is the heterogeneity in a population. Physiological (i.e.,
metabolic) and genetic heterogeneity ensure the survival of a population of
cells if environmental changes occur (stress phenomenon) and has been described
historically as the survival of the fittest. As growth in a bioreactor is related to
different stresses in different phases of a process and as additional perturbations
that occur in large-scale bioreactors add a further stress layer, the population
heterogeneity in a bioprocess is an important feature, which needs to be monitored and can be used for a validation of similarity between different process
scales.
As a consequence, every scale-down approach needs to start with a good understanding of the large scale especially in view of the cellular response dynamics.
Furthermore, these cellular response dynamics must be reduced in mathematical
models, the so-called digital twins, to the characteristics essential for the process
scale-down. Finally, methods need to be implemented which help to validate the
quality of the scale-down – and this is only possible by measurements.
4.2 Execution of Scale-Down Experiments
4.2.1 Combination of Scale-Down Experiments with Model-Based
Approaches
As shown in the scale-down scheme in Fig. 1, the characterization of the larger-scale
bioreactor environment is followed by transferring the environmental blueprint to
the laboratory-scale simulator in which the actual scale-down experiments are
executed. To achieve this transfer, the digital twin of the bioprocess should contain
model units that adequately describe the physiology of the cells, as well as the
cultivation process and geometric analysis of the bioreactors involved. These details
should be digitally embedded into the definition of the scale-down model (Fig. 1).
Thereafter, the application of the modeling framework (digital twin) in the context of
scale-down experimentation can take two forms: (1) the design of the scale-down
Potential of Integrating Model-Based Design of Experiments Approaches and. . .
17
which are easily applicable and better reflect the large-scale situation.
3. Due to the very high computing power, which is needed to solve these Euler–
Lagrangian models with a reasonable resolution, the current models cannot
describe the process dynamics over time, but only represent a very narrow time
point of the cultivation. Nevertheless, this kind of simulations, e.g. if they are
performed at different time points of a cultivation can provide important information to plan scale-down experiments.
4. Current models only consider the liquid phase. The implementation of the gas
phase would additionally need much computing power and in our opinion it
would be very laborious to validate these models.
5. An important characteristic of living systems in connection with their adaptation
to the environment is the heterogeneity in a population. Physiological (i.e.,
metabolic) and genetic heterogeneity ensure the survival of a population of
cells if environmental changes occur (stress phenomenon) and has been described
historically as the survival of the fittest. As growth in a bioreactor is related to
different stresses in different phases of a process and as additional perturbations
that occur in large-scale bioreactors add a further stress layer, the population
heterogeneity in a bioprocess is an important feature, which needs to be monitored and can be used for a validation of similarity between different process
scales.
As a consequence, every scale-down approach needs to start with a good understanding of the large scale especially in view of the cellular response dynamics.
Furthermore, these cellular response dynamics must be reduced in mathematical
models, the so-called digital twins, to the characteristics essential for the process
scale-down. Finally, methods need to be implemented which help to validate the
quality of the scale-down – and this is only possible by measurements.
4.2 Execution of Scale-Down Experiments
4.2.1 Combination of Scale-Down Experiments with Model-Based
Approaches
As shown in the scale-down scheme in Fig. 1, the characterization of the larger-scale
bioreactor environment is followed by transferring the environmental blueprint to
the laboratory-scale simulator in which the actual scale-down experiments are
executed. To achieve this transfer, the digital twin of the bioprocess should contain
model units that adequately describe the physiology of the cells, as well as the
cultivation process and geometric analysis of the bioreactors involved. These details
should be digitally embedded into the definition of the scale-down model (Fig. 1).
Thereafter, the application of the modeling framework (digital twin) in the context of
scale-down experimentation can take two forms: (1) the design of the scale-down
Potential of Integrating Model-Based Design of Experiments Approaches and. . .
17