A New Concept for the Rapid Development of Digital Twin Core Models for Bioprocesses in Various Reactor Designs
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors
The presented manuscript deals with a modelling approach to bioreactors by subdividing them into stirred tank reactors, within the SRT models biokinetic and physico-chemical submodels are nested, which can be adjusted to fit the modelling needs of multiple different bioreactors. A software was developed to simplify the model building process and reduce friction for the application of digital twins in the industry and academia. This can be an important step towards efficient manufacturing and resilient process control.
Line 58-59: Can the pH not also be a function of the enzymatic reaction and therefore a part of the biokinetic submodel?
Line 233: How are fluid dynamics for non-ideal flow conditions and STR with different reactor zones considered in the model?
Line 305-307: How was the model parameterized? Was the parameterization carried out independently from the validation experiment?
Line 355: Please discuss reasons for the underestimation of the pH value from the model.
Line 358: How is the number of STR models determined?
Line 458: How is the reduction of 90% calculated? What is the reference? The article covers very little on the parameterization process of the models. There is a speed advantage in the model formation with the presented compartment model and software. But it is unclear how parameterization is carried out and how it compares to conventional mechanistic or other models.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors place great emphasis on creating a digital twin without sufficiently explaining why that is needed, what is lacking currently and how they are addressing it. The article can be better organized with more details on the models used, parameters used, how the constraints of solving multiple DEs without severely affecting simulation time. Some additional comments are below:
· Abstract: “The combination of multiple reactor models allows for the simulation of different reactor designs.” Please add where are these combined?
· Abstract”: Lines 23-26: The claim of “considerably reduced” is subjective. Would recommend focusing on the benefit on having a digital twin instead of focusing on faster model development.
· Introduction: “Due to the increasing demand for bioprocess DTs” authors seem to be focusing on the development of DTs. However, recommend aiming for an improvement in bioprocessing whereby DT is a just a tool to achieve that.
· Line 43: “must be able to map”. Do the authors mean predict?
· Line 93: “This feature also makes mechanistic models valid beyond the calibration space [8].” This is a tall claim. No model is valid beyond the calibration space unless proven scientifically. The model may be able to provide useful insights but the validity needs to be proven.
· Introduction: What is missing in the current state of work. How are the authors solving that problem
· Line 120: Recommend expanding STRs in first use
· Introduction: The introduction is very broad. It is unclear if the authors are providing a review of the state-of-the-art or explaining the approach they have taken. Recommend dividing the Introduction into specific sections dealing with each of the areas.
· How is this addressed? “The numerical solution of large mechanistic models, consisting of systems of a high number of nonlinear coupled differential equations requires high computational effort. It is particularly important to keep the computation times of DTs, especially for their parameterisation and application in process optimisation, as short as possible.”
· How is this addressed? “Non-ideal flow patterns in bioreactors may have an impact on the kinetics, performance, and dynamics of the process under consideration. Thus, for a realistic representation of these effects in DTs, possibilities should be created to represent non-ideal reactor behaviour and/or different reactor types with the reactor submodel;”
· Section 3.1: Recommend better organization of Section 3.1. The models and equations are not sufficiently explained. What are each of the terms in the model. Why are double sigmoidal functions used? If the authors aim to maintain a library of models, why is one model structure chosen over the other?
· In addition, the models can be discussed in separate sections per the different areas in the DT core model.
· Line 198: “The biokinetic submodel and the physico-chemical submodel are embedded in the reactor submodel.” What does embedded mean in mathematical terms?
· Line 265: “The temperature model and the DO model have the fastest time constants and are thus decisive for the number of necessary calculation steps.” There is no discussion on what models are used for this and the structure of the model
· Line 315: Authors use the phrase “through model reduction”. Please elaborate the reduction being referred to here.
· Line 313: “Without the possibility of selecting models and methods for model reduction, 378 differential equations would 314 be necessary in the model to perform the desired calculations” Was computation time/effort measured using the original configuration in order to compare improved performance?
· Line 317: “over 90% faster simulation times.” What were the simulation times?
· Line 316: How was the step-size chosen?
· Line 328: “Only the base strength of the initial cultivation medium was adjusted.” How was this chosen as the parameter that needs to be adjusted?
· Line 338-355: What are the model equations used to simulate this process and generate the plots in Figure 6. What were the parameters in this model?
· Line 357 and 360: “The definition of the reactor submodel structure using six networked STR models and the specification of the biokinetic and physico-chemical submodels was accomplished within one hour”; “Using the new software tool concept, the source code creation of the DT core model took less than one minute, instead of several days if this had to be done manually. This fast code 362 implementation also enables the testing of model compositions and sequential parameterisation of submodels.” This claim is very subjective- since it depends on the expertise of the user, if the developer was using the tool or another user was and several other factors. In adiditon, no information if provided in the article on the user interface and the steps required to perform this update.
· Section 3.2.2: What are the model equations used to simulate this process and generate the plots in Figure 6. What were the parameters in this model?
· Line 432: What is the Dilution term in the differential equations
Comments on the Quality of English Languageminor editing needed
Author Response
Author Response File: Author Response.docx