adaptability needed to model all aspects of bioprocesses, as they were originally
designed for modelling of chemical processes. With the increasing focus on
bioprocess development, significant effort has been invested in the development of
model libraries for bioprocess unit operations in recent years. Software systems for
parameter estimation and computation of algebraic and differential equations provide a user-friendly and adaptable environment for model development and implementation of Digital Twins [9–11].
For the design of Digital Twins or “early-stage” Digital Twins that can be used for
the development, optimisation and realisation of control strategies, there are already
a variety of software packages available. Table 1 lists a selection of vendors and
associated software products and summarises the most important features of the
respective software packages. Most of the Digital Twin development tools listed are
designed for the chemical industry (e.g. UniSim Competency Suite [18] or
IndissPlus [19]), but some are also suitable for the development of bioprocess Digital
Twins (e.g. WinErs/C-eStIM [20, 21], PerceptiveAPC [22] or TMODS [23]).
2.3 Control Strategies for Bioprocesses
The multi-phase system in a bioprocess sets highest demands on measurement and
control technology [32–34]. To maintain optimal conditions for the entire process,
the composition of the liquid phase (e.g. medium), the suspended gas phase
(e.g. oxygen, carbon dioxide) and the dispersed solid phase (e.g. cells, cell assemblies, enzymes) must be monitored continuously [32]. Furthermore, complex
dynamics showing a wide range of time constants make it difficult to control the
process without sufficient process knowledge [32]. For example, the induction of a
gene through a temperature shift or the addition of a chemical inducer affects the
process several minutes after the expression of the desired protein because the
formation of a metabolically active protein will cause a time delay. This kind of
knowledge must be available and utilised for successful bioprocess control based on
detailed process analytics [32–35].
The choice of control strategies mainly depends on the selected bioprocess and
the available reactor type [33, 34]. In general, controllers are divided according to
continuous (e.g. PID control, soft sensor control) and discontinuous behaviour
(e.g. model predictive control (MPC) or nonlinear model predictive control
(NMPC)) [34]. Controllers with continuous behaviour calculate and transmit continuous control signals based on the current process characteristics [34]. Among the
best-known continuous controllers are the “conventional” controllers like two-point-,
three-point-, proportional- (P-), proportional-integral- (PI-) or PID-controllers. Controllers with discontinuous behaviour only calculate control signals or profiles at
specific process points [34].
As an example, conventional control strategies such as PI or PID control are
generally used to control temperature [34]. In many cases, the control system should
be able to maintain the desired setpoint, due to the rather weak influence of
Digital Twins for Bioprocess Control Strategy Development and Realisation
69
designed for modelling of chemical processes. With the increasing focus on
bioprocess development, significant effort has been invested in the development of
model libraries for bioprocess unit operations in recent years. Software systems for
parameter estimation and computation of algebraic and differential equations provide a user-friendly and adaptable environment for model development and implementation of Digital Twins [9–11].
For the design of Digital Twins or “early-stage” Digital Twins that can be used for
the development, optimisation and realisation of control strategies, there are already
a variety of software packages available. Table 1 lists a selection of vendors and
associated software products and summarises the most important features of the
respective software packages. Most of the Digital Twin development tools listed are
designed for the chemical industry (e.g. UniSim Competency Suite [18] or
IndissPlus [19]), but some are also suitable for the development of bioprocess Digital
Twins (e.g. WinErs/C-eStIM [20, 21], PerceptiveAPC [22] or TMODS [23]).
2.3 Control Strategies for Bioprocesses
The multi-phase system in a bioprocess sets highest demands on measurement and
control technology [32–34]. To maintain optimal conditions for the entire process,
the composition of the liquid phase (e.g. medium), the suspended gas phase
(e.g. oxygen, carbon dioxide) and the dispersed solid phase (e.g. cells, cell assemblies, enzymes) must be monitored continuously [32]. Furthermore, complex
dynamics showing a wide range of time constants make it difficult to control the
process without sufficient process knowledge [32]. For example, the induction of a
gene through a temperature shift or the addition of a chemical inducer affects the
process several minutes after the expression of the desired protein because the
formation of a metabolically active protein will cause a time delay. This kind of
knowledge must be available and utilised for successful bioprocess control based on
detailed process analytics [32–35].
The choice of control strategies mainly depends on the selected bioprocess and
the available reactor type [33, 34]. In general, controllers are divided according to
continuous (e.g. PID control, soft sensor control) and discontinuous behaviour
(e.g. model predictive control (MPC) or nonlinear model predictive control
(NMPC)) [34]. Controllers with continuous behaviour calculate and transmit continuous control signals based on the current process characteristics [34]. Among the
best-known continuous controllers are the “conventional” controllers like two-point-,
three-point-, proportional- (P-), proportional-integral- (PI-) or PID-controllers. Controllers with discontinuous behaviour only calculate control signals or profiles at
specific process points [34].
As an example, conventional control strategies such as PI or PID control are
generally used to control temperature [34]. In many cases, the control system should
be able to maintain the desired setpoint, due to the rather weak influence of
Digital Twins for Bioprocess Control Strategy Development and Realisation
69