Artificial Neural Network-Based Automated Finite Element Model Updating with an Integrated Graphical User Interface for Operational Modal Analysis of Structures
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe paper introduces a denoising approach and an ANN-based GUI for FEM inputs, which is interesting. However, the novelty may be somewhat limited as it builds on existing methodologies without a highly distinctive breakthrough. The paper raises some potential inconsistencies, such as the reliability of the ANN under uncertain data conditions, which other studies (e.g., Naranjo-Pérez et al., 2020) have flagged as a concern. The authors mention the use of probabilistic approaches to address these uncertainties but do not fully explore this avenue in their methodology.
1-Data Generalization: How well does the ANN model generalize to other structural types not included in the training data? Have tests been conducted on structures with significantly different properties?
2-User Interaction with GUI: Can the authors provide more details on the user interaction with the GUI? Specifically, how intuitive is the GUI for non-experts in FEM and ANN?
3-Denoising Process: What specific criteria were used to select the optimal wavelet type and decomposition level in the DWT approach?
Upon reviewing the document, I identified the following sentences with grammar or spelling issues:
Sentence: "The approach targets to reduce the uncertainties inherent in both the experimental data and the computational model."
Issue: The phrase "targets to" is awkward.
Suggested Correction: "The approach aims to reduce the uncertainties inherent in both the experimental data and the computational model."
Sentence: "This paper focuses on presenting a denoising approach for data of the OMA approach..."
Issue: Repetition of the word "approach."
Suggested Correction: "This paper focuses on presenting a denoising method for data from the OMA approach..."
Sentence: "This step is followed by the creation of a database to train the ANN through iterations over dynamic analysis of all combinations, using the optimal input parameters of the FEM calculated from the sensitivity analysis step."
Issue: The sentence is too long and could be more concise.
Suggested Correction: "This step is followed by the creation of a database to train the ANN, iterating over all dynamic analysis combinations and using the optimal FEM input parameters identified in the sensitivity analysis."
Author Response
The authors appreciate the constructive feedback from the reviewer. Please find the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe manuscript presents an investigation of an ANN-based graphical user interface designed to automate the process of finite element model updating using data from operational modal analysis. The topic is in the scope of the journal and the writing of the paper is generally in good style. The current version of the manuscript requires some improvements before it can be reconsidered for publication in Buildings. The following are the detailed comments for consideration.
1. The authors should include a concise statement that defines the research gap their study fills and how their approach or findings differ from those in existing literatures, especially the listed ones [13]-[18]. This should be placed early in Section “1. Introduction”.
2. Literature review is not adequate as too many relevant references were not reasonably included, particularly those related to the state-of-the-art neural networks I will recommended the following:
“Convolutional Neural Network based fault detection for rotating machinery, J. Sound Vib. 377 (2016) 331–345.”
“A deep learning approach for electromechanical impedance based concrete structural damage quantification using two-dimensional convolutional neural network. Mech. Syst. Signal Pro. 183 (2023) 1-16.”
“Study of factors affecting the magnetic sensing capability of shape memory alloys for non-destructive evaluation of cracks in concrete: Using response surface methodology (RSM) and artificial neural network (ANN) approaches, Heliyon 10(2024) e35772”
“Convolutional Neural Networks for Local Component Number Estimation from Time–Frequency Distributions of Multicomponent Nonstationary Signals, 2024 Mathematics 12(11),1661”, etc.
3. The flowchart presented in Figure 1 does not clearly represent the research process described in the manuscript. It is unclear how the different steps in the process relate to each other and what the exact sequence of operations is. To improve the clarity and usefulness of this figure, I recommend revising the flowchart to ensure that the flow of the process is logical and follows a clear path from start to end.
4. The authors declare that the proposed model was developed using a Feed-Forward architecture and trained with the Adam optimizer via backpropagation, consisted of a simple three-layer network. Would a more complex network structure, such as adding more layers or using different activation functions, improve the model’s performance or robustness? More discussions and explanations are recommended to be given.
5. The font style and size used in figures are suggested to be consistent with the journal’s guidelines.
Comments on the Quality of English LanguageThe manuscript presents an investigation of an ANN-based graphical user interface designed to automate the process of finite element model updating using data from operational modal analysis. The topic is in the scope of the journal and the writing of the paper is generally in good style. The current version of the manuscript requires some improvements before it can be reconsidered for publication in Buildings. The following are the detailed comments for consideration.
1. The authors should include a concise statement that defines the research gap their study fills and how their approach or findings differ from those in existing literatures, especially the listed ones [13]-[18]. This should be placed early in Section “1. Introduction”.
2. Literature review is not adequate as too many relevant references were not reasonably included, particularly those related to the state-of-the-art neural networks I will recommended the following:
“Convolutional Neural Network based fault detection for rotating machinery, J. Sound Vib. 377 (2016) 331–345.”
“A deep learning approach for electromechanical impedance based concrete structural damage quantification using two-dimensional convolutional neural network. Mech. Syst. Signal Pro. 183 (2023) 1-16.”
“Study of factors affecting the magnetic sensing capability of shape memory alloys for non-destructive evaluation of cracks in concrete: Using response surface methodology (RSM) and artificial neural network (ANN) approaches, Heliyon 10(2024) e35772”
“Convolutional Neural Networks for Local Component Number Estimation from Time–Frequency Distributions of Multicomponent Nonstationary Signals, 2024 Mathematics 12(11),1661”, etc.
3. The flowchart presented in Figure 1 does not clearly represent the research process described in the manuscript. It is unclear how the different steps in the process relate to each other and what the exact sequence of operations is. To improve the clarity and usefulness of this figure, I recommend revising the flowchart to ensure that the flow of the process is logical and follows a clear path from start to end.
4. The authors declare that the proposed model was developed using a Feed-Forward architecture and trained with the Adam optimizer via backpropagation, consisted of a simple three-layer network. Would a more complex network structure, such as adding more layers or using different activation functions, improve the model’s performance or robustness? More discussions and explanations are recommended to be given.
5. The font style and size used in figures are suggested to be consistent with the journal’s guidelines.
Author Response
The authors appreciate the constructive feedback from the reviewer. Please find the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsTo build a data-driven model between inputs and outputs with multiple design variables:
How can the number of simulation cases be determined using Design of Experiments (DOE)?
This is a critical but often overlooked aspect when dealing with many design variables.
Are there any minimum and maximum bounds for design variables?
How can the sensitivity with respect to design parameters be obtained theoretically or through artificial neural networks (ANNs)?
While this approach is relatively straightforward, it tends to work well only for small finite element (FE) models with a few design variables.
In conclusion, it would be valuable to include some comments on the potential future applications for larger models with many design parameters.
Comments on the Quality of English LanguageNot applicable.
Author Response
The authors appreciate the constructive feedback from the reviewer. Please find the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe authors have effectively addressed my comments. I suggest accepting the paper.
Author Response
Thank you for your time.
Reviewer 3 Report
Comments and Suggestions for AuthorsContrary to the author's claim, "The text has been revised accordingly, and the minimum and maximum bounds for the design variables are clearly mentioned (Lines 441-443)," the minimum and maximum bounds of the design variables are not described.
Furthermore, some of the lines in the author's revision do not align well with the rest of the text.
Additionally, the assumption that perturbations in the material properties range from 0% to 50% is unreasonable, as the material in question is metal.
Author Response
Please find the attachment.
Author Response File: Author Response.pdf
Round 3
Reviewer 3 Report
Comments and Suggestions for AuthorsThe definitions of the natural frequency sensitivity factor and the nodeshape sensitivity factor seem to be relatively new concepts, making them difficult to find in other literature. Therefore, they should be referenced according to other sources. If the authors have devised these concepts for the first time, please clearly specify this point.
Comments on the Quality of English LanguageNot applicable.
Author Response
In response to the reviewer's feedback, the '5.4. Sensitivity Analysis' section has been revised.
