Twins is a dynamic mathematical model, which can map the biological, chemical
and physical phenomena of the real process in detail [9]. This dynamic mathematical
process model should be coupled to a graphical user interface (GUI) [9]. Users can
monitor and make changes to the virtual process using graphical icons in the GUI.
From the author’s point of view, it is advantageous, if the structure of the GUI
corresponds to the process control system (PCS) on the physical counterpart. The
Digital Twin GUI is a functional image, derived from the P&ID (piping and
instrumentation diagram) flow chart of the real bioprocess and thus, also serves as
a realistic replica of important parts of the control and automation model. A realistic
GUI of a Digital Twin can, therefore, be used to check the usability (including
typical operating errors), as well as the control and automation of the real bioprocess.
The model of a Digital Twin is parameterised based on real process data to represent
the behaviour of the physical process [10]. Another possibility to keep Digital Twin
and the real process as identical as possible is an online and at-line data connection
between the “twins”. This enables the adaption of the Digital Twin using online and
at-line data, which is particularly useful if the real process frequently changes its
characteristics.
During process development or optimisation, Digital Twins can be used for the
following applications:
1. Determination of suitable controller types.
2. Improvement of controller performance.
3. Improvement of the overall process performance through appropriate process
control strategies.
If, for example, suitable controllers (e.g. for temperature, dissolved oxygen or
product concentration) should be designed, the controller type can be selected based
on simulations with the Digital Twin. An early step in controller selection should be
the definition of appropriate control targets [8]. When controlling the temperature of
a bioreactor, such control targets are, e.g., a short rise time, a high control accuracy
(especially important for temperature-sensitive organisms, particularly mammalian
cells) or a low overshoot. For example, the conventional proportional integral
derivative (PID) control can be compared to a more complex nonlinear model
predictive control (NMPC) by applying them to a Digital Twin. If both control
strategies yield equally good control results, PID control would be preferred,
because it is cheaper and easier to handle.
Once a control strategy has been able to control the virtual process satisfactorily,
the results are transferred to the real process. The transfer of the developed control
strategy from the Digital Twin to the real process may be further simplified if the
Digital Twin and the real process are linked to the identical PCS [8].
To illustrate the general approach of process control design utilising a Digital
Twin, the case study in Sect. 4 presents the selection and optimisation of suitable
control strategies for the cultivation of S. cerevisiae.
Digital Twins for Bioprocess Control Strategy Development and Realisation
67
and physical phenomena of the real process in detail [9]. This dynamic mathematical
process model should be coupled to a graphical user interface (GUI) [9]. Users can
monitor and make changes to the virtual process using graphical icons in the GUI.
From the author’s point of view, it is advantageous, if the structure of the GUI
corresponds to the process control system (PCS) on the physical counterpart. The
Digital Twin GUI is a functional image, derived from the P&ID (piping and
instrumentation diagram) flow chart of the real bioprocess and thus, also serves as
a realistic replica of important parts of the control and automation model. A realistic
GUI of a Digital Twin can, therefore, be used to check the usability (including
typical operating errors), as well as the control and automation of the real bioprocess.
The model of a Digital Twin is parameterised based on real process data to represent
the behaviour of the physical process [10]. Another possibility to keep Digital Twin
and the real process as identical as possible is an online and at-line data connection
between the “twins”. This enables the adaption of the Digital Twin using online and
at-line data, which is particularly useful if the real process frequently changes its
characteristics.
During process development or optimisation, Digital Twins can be used for the
following applications:
1. Determination of suitable controller types.
2. Improvement of controller performance.
3. Improvement of the overall process performance through appropriate process
control strategies.
If, for example, suitable controllers (e.g. for temperature, dissolved oxygen or
product concentration) should be designed, the controller type can be selected based
on simulations with the Digital Twin. An early step in controller selection should be
the definition of appropriate control targets [8]. When controlling the temperature of
a bioreactor, such control targets are, e.g., a short rise time, a high control accuracy
(especially important for temperature-sensitive organisms, particularly mammalian
cells) or a low overshoot. For example, the conventional proportional integral
derivative (PID) control can be compared to a more complex nonlinear model
predictive control (NMPC) by applying them to a Digital Twin. If both control
strategies yield equally good control results, PID control would be preferred,
because it is cheaper and easier to handle.
Once a control strategy has been able to control the virtual process satisfactorily,
the results are transferred to the real process. The transfer of the developed control
strategy from the Digital Twin to the real process may be further simplified if the
Digital Twin and the real process are linked to the identical PCS [8].
To illustrate the general approach of process control design utilising a Digital
Twin, the case study in Sect. 4 presents the selection and optimisation of suitable
control strategies for the cultivation of S. cerevisiae.
Digital Twins for Bioprocess Control Strategy Development and Realisation
67