Development and Utilization of Bridge Data of the United States for Predicting Deck Condition Rating Using Random Forest, XGBoost, and Artificial Neural Network
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis paper utilizes Random Forest, XGBoost, and ANN models to predict the rating of bridge deck, and a substantial amount of work has been carried out in data collection. However, to meet the publication requirements of the journal, there are some shortcomings needed to be corrected:
1. The first paragraph of the introduction focuses on the increasing number of bridges that are nearing the end of their useful life and need major repairs or decommissioning. These statements emphasize the importance of predicting the condition rating of bridge structures, rather than the condition rating prediction of bridge deck, which is the focus of the article.
2. The introduction only mentions the related research conducted by scholars using decision tree models, but this article mainly uses random forest models. Although the main body of random forest is still a decision tree, there are still differences between the two, so the introduction should still explain the related work conducted by scholars around random forest models.
3. The introduction section lacks a comprehensive summary of existing research. The author only mentions the simplest and most basic machine learning models. This resulting in the failure to properly sort out the problems in the existing research.
4. In constructing the ANN model in the paper, a trial-and-error approach was employed to determine the number of hidden layers and neurons. Due to the final model having 5 layers and hundreds of nodes, the workload of relying solely on trial and error was particularly substantial. Moreover, the accuracy of the ANN model developed in this paper is not good. Using a trial-and-error approach may not be an optimal choice. The author could consider employing the Grid SearchCV method to more accurately determine hyperparameters. This approach not only enhances the performance of the ANN but also adds credibility to the paper.
5. The author has consistently emphasized the substantial improvement in model predictive performance through the use of historical bridge data. However, the author has not explored why the utilization of historical bridge data enhances the model's predictive capabilities. Since the data volume differs significantly between the 2016-2020 dataset and the 2020-only dataset employed in this paper, and it is well known that data volume has a profound impact on machine learning, could this enhancement be solely attributed to the increase in data volume?
6. The feature analysis of the parameters in the text only presents the results without providing explanations.
7. The feature analysis of parameters is often a preparatory step to further streamline the model. If the author can further simplify the model based on the feature analysis of parameters and observe whether the simplified model exhibits improvement compared to the original, it would serve as strong evidence for the accuracy of the parameter feature analysis and further enhance the overall performance of the author's model.
8. The clarity of the pictures in the article needs to be improved.
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Comments and Suggestions for AuthorsThe paper is well-written and presents interesting research for the readers. Nonetheless, there are a few minor issues that could be addressed for further improvement
1. It is recommended to enhance the resolution of the figures to improve their clarity.
2. In Figure 2, the word "Political boundary" may be changed to "Administrative border."
3. Could you please elaborate on how the provided comparison between methods ensures fairness?
4. The activation functions used in Table 13 should be introduced.
5. On page 21, the citation to equation (3) should be revised.
6. As one of the main contributions of the paper is on data-driven SHM, the relevant review papers can be cited in the Introduction section. The following list might be helpful:
- Data-driven structural health monitoring and damage detection through deep learning: State-of-the-art review
- Artificial-Neural-Network-Based Surrogate Models for Structural Health Monitoring of Civil Structures: A Literature Review
- A review on non-destructive evaluation of construction materials and structures using magnetic sensors
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
Comments and Suggestions for AuthorsThe paper is well written and organized, some minor revisions regard the introduction and the conclusion. In particular the introduction should be more oriented on the scientific gap that the paper would fill and the conclusion should better highlight the results attaimed and the future developments in the research.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 4 Report
Comments and Suggestions for AuthorsThe paper presents an experimental activity on the evaluation of health condition of a bridge deck using Random Forest, XGBoost, and Artificial Neural Network. The study is methodologically valid and presents some valid results.
It is recommended that the authors dig deeper into the implications of the study in the discussion and conclusions and make more recommendations for future work targeting the use of such kind of techniques in order to address the health monitoring of infrastructures.
There are also some inconsistencies in the formatting of references. Please ensure that all references follow a consistent format, as per the journal/conference guidelines.
Moreover, some useful references related to the work can be added:
Gattulli, V., Franchi, F., Graziosi, F., Marotta, A., Rinaldi, C., Potenza, F. and Sabatino, U.D., 2022. Design and evaluation of 5G-based architecture supporting data-driven digital twins updating and matching in seismic monitoring. Bulletin of Earthquake Engineering, 20(9), pp.4345-4365.
Smarra, F., Di Girolamo, G.D., De Iuliis, V., Jain, A., Mangharam, R. and D’Innocenzo, A., 2020. Data-driven switching modeling for mpc using regression trees and random forests. Nonlinear Analysis: Hybrid Systems, 36, p.100882.
Other minor suggestions:
- Quality of images needs to be improved. All the Figures lacks in resolution.
- Plot shown In Figure 8 is difficult to read due a low quality.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe author has made modifications to the proposed questions.
Comments on the Quality of English LanguageEnglish Language needs minor revision
