Evaluation and Prediction of Pavement Deflection Parameters Based on Machine Learning Methods

Round 1

Reviewer 1 Report

Line 78: The citation style of [15] is not consistent. et al. can be used.

Line 80: The citation style of [8] is not consistent. et al. can be used.

Line 111: What kind of rehabilitation methods were selected in this study? Which state's LTPP sites were utilized to extract data? Please provide more details about the dataset.

Line 155: In Table 2, what's the definition of improvement thickness? For age, is it based on years? Why is the Min -13?

Line 166: Please add the axis title to Figure 2.

Figure 4 through 12 exhibits the feature importance of variables. Some variables, like Depth 1, show low importance. Why don't remove such variables with low importance to simplify the model?

Many similar studies utilized machine learning methods and FWD data to predict the deflection of AC pavement. The literature review can add more of these studies. The authors should emphasize the research significance of this study. What're the key improvements achieved in this study? What're the advantages identified compared to other similar studies?

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This paper presents study on evaluation and prediction of pavement deflection parameters, based on classification and regression tree (CART), random forest (RF), and gradient boosting decision tree (GBDT) machine learning methods. Authors provide the deflections and influence factors including falling weight deflectometer test conditions, pavement structural parameters, climatic factors, traffic level, rehabilitation level, and many others.

Although presented article is very well written and may be interesting to the readership of this journal, the paper may only be considered for publication after the following concerns have been addressed successfully in a minor revision:

Figures 4-12 in my opinion should be presented in more comparative form. Authors may use one summary graph or put all measurements together on one figure to reduce the space and increase the possibility to compare individual results. Please consider these possibilities and apply any of them.

Figures 13-15 should be in my opinion presented as one Figure with a), b) and c) graphs because they are all about one thing.

The "Conclusion" part needs some improvement. In my opinion this section is too general and does not relate essentially to the results. Which method allows to provide which factors? Any numbers? Percents? What should be done next? What are guidelines for pavement designers? What is the significance of this manuscript in the context of the results obtained? Unfortunately, Authors do not answer these questions and this is a great place to do so. Please consider that.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

1. Figure with the scheme of the FWD measurement principle is missing.

2. A legend should be added to Figure 1, i.e. what each curve refers to.

3. In the text before equations 4-8 should be added (as shown in Eq. (4-8))       as stated before equations 1-3.

4. It is necessary to indicate which computer program (for example ELMOD        or ...) is most used in Backcalculation Methodology within LTPP Data.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors well addressed my comments.