Application of Data Mining Techniques to Predict Luminance of Pavement Aggregate

Round 1

Reviewer 1 Report

This paper proposed an approach to assessing the aggregate luminance coefficient by data processing, and finally recommended the optimal method and the type of aggregate. Some of the conclusions are interesting, while discussions and revisions suggested are listed below:

1. Pg.1 Para.1, it seems that the conclusions of this manuscript would be applied in surface layers with aggravates coated with asphalt or “light-colored” materials rather than macadam surface, the reflectance and refractive index of road surface light would be affected by coated binders and other factors, please illustrate the coupled effects of binders on luminance as well as highlight the role of aggregates.

2. Pg.1 Ln.137, the statement “there are no known articles in which an attempt has been made to …” is absolute.

3. In section 2.1, considering that this research attempted to create “a large learning dataset” (Line.102), which the DM requires, so whether the sources of aggregates were wide and the selections of aggregates representative? The reasons and considerations need to be further illustrated.

4. Pg.6 Ln.245, whether the statement refers to Fig. 2(a)?

5. Pg.6 Ln.249, “The measurements are performed when the ambient temperature is between 0 and 30 °C, and the temperature of the test sample/surface is between 5 and 40 °C”, whether the temperature span is wide? How to guarantee the stability of results?

6. Pg.6 Ln.252, the statement “The research plan took into account the effect of aggregate conditioning by taking into account the compaction of the sample” is confusing.

7. Please check the semantic errors and the accuracy of sentences in the whole manuscript.

8. Pg.7 Ln.254, what are the dimensions of the samples? The preparation of samples should also be illustrated.

9. Pg.7 Ln.254, what does “adequate repeatability” means?

10. Pg.7 Ln.255, please explain the reason for choosing vibrating compaction. How the test operation “with a compaction time of 3 min at an amplitude of 3 mm” was determined?

11. Pg.7 Ln.262, what does “Each aggregate was evaluated using 5 series” means?

12. Section 3.2 is not convincing, since the necessary test procedures and calculation principles are not mentioned in the manuscript. It is doubted that the results are actual and correct. Please add necessary contents.

13. Pg.10 Ln.361, the abbreviation of “Random forests (RT)” is wrong.

14. Please add necessary legends and notes in Fig.5 and Fig.11, the quality of the figures needs to be further improved.

15. Please improve the legends and notes in Fig.8.

16. Whether the results (RMSE and the R²) in Tab.7 meet the requirements of statistical analysis?

17. Please check whether the content about the influence of compaction on the sample is missing, which is one of the purposes stated in line 253.

Author Response

(in text corrections were marked red color)

1. Pg.1 Para.1, it seems that the conclusions of this manuscript would be applied in surface layers with aggravates coated with asphalt or “light-colored” materials rather than macadam surface, the reflectance and refractive index of road surface light would be affected by coated binders and other factors, please illustrate the coupled effects of binders on luminance as well as highlight the role of aggregates.

A crucial information was added to the article on the basis of  suggestions of reviewer. In this moment, investigations included composition factors are in progress. It is done to expand the DM model, taking into account the composition of selected bituminous mixtures made in the laboratory as well as the results assigned to composites placed in the pavement with a known history of use. Therefore, in the introduction, we added literature in which the need for further research was signaled by other authors, where the influence of the aggregate luminance on a composition of the mineral-asphalt mixture should be taken into account. Nevertheless, in the article we focused only on the construction of the model for the aggregate luminance forecasting.

 

2. Pg.1 Ln.137, the statement “there are no known articles in which an attempt has been made to …” is absolute.

Thank you for this remark. This sentence was modified

3. In section 2.1, considering that this research attempted to create “a large learning dataset” (Line.102), which the DM requires, so whether the sources of aggregates were wide and the selections of aggregates representative? The reasons and considerations need to be further illustrated.

Aggregates of various petrographic origin and mineral composition in Poland are available. Aggregates used in tests are frequently used to manufacture bituminous mixtures. We tried to choose  these cases of aggregates that are representative and often indicated by customers. In machine learning, the largest possible number of boundary conditions should be taken into account so that the statistical model can predict fuzzy cases as accurately as possible. In order to increase the effectiveness of modeling, attention was paid to aggregates obtained from sedimentary, metamorphic and igneous rocks, i.e. those that dominate in use in Poland and the EU.

4. Pg.6 Ln.245, whether the statement refers to Fig. 2(a)?

This sentence was corrected.

5. Pg.6 Ln.249, “The measurements are performed when the ambient temperature is between 0 and 30 °C, and the temperature of the test sample/surface is between 5 and 40 °C”, whether the temperature span is wide? How to guarantee the stability of results?

Thank you for this insightful remark. These conditions were taken from polish obligatory document WT-2/2014 that specifies such temperature ranges. This documents is in line with similar German requirements. We assumed that such temperature ranges was chose, in this document, to avoid condensation of water on the aggregate surface that could influence on Qd results. We didn’t have at our disposal other document so we accepted it for tests.

6. Pg.6 Ln.252, the statement “The research plan took into account the effect of aggregate conditioning by taking into account the compaction of the sample” is confusing.

This sentence was modified. We agree that it was not a accurate expression. The authors' intention was to point out that samples surface preparation was also important.

7. Please check the semantic errors and the accuracy of sentences in the whole manuscript.

This manuscript was revised again to eliminate potential mistakes.

8. Pg.7 Ln.254, what are the dimensions of the samples? The preparation of samples should also be illustrated.

Thank you for this remark. It was added the scheme of test stand and all figures were connected with text.

9. Pg.7 Ln.254, what does “adequate repeatability” means?

Thank you for this remark. For an adequate measure of homogeneity, we established a 10% range of mean value for results in accordance with the document WT2/214 (in Poland). This information was added manuscript.

10. Pg.7 Ln.255, please explain the reason for choosing vibrating compaction. How the test operation “with a compaction time of 3 min at an amplitude of 3 mm” was determined?

Thank you for this remark. We performed several preliminary studies which showed that subjecting the aggregate to vibration caused a decrease in the variance of Qd. This was observed for all cases of aggregates. Therefore, in order to homogenize the research process and take into account the fact that the aggregate is compacted in the pavement, we left this method of sample preparation for further research. The increase in homogeneity was approx. Qd=3 mcd⋅m^-2⋅lx^-1 .

Pg.7 Ln.262, what does “Each aggregate was evaluated using 5 series” means?

Thank you for this remark. The measuring regarded to 5 measurement of each series that were 5. Thus, luminance coefficient of each aggregate (for each gradation) covered 25 measurement including variance of a sampling site. This inaccuracy was expanded in text of manuscript

12. Section 3.2 is not convincing, since the necessary test procedures and calculation principles are not mentioned in the manuscript. It is doubted that the results are actual and correct. Please add necessary contents.

Thank you for this valuable remark. It was hard to make simply because results are implemented in XML files (they consist of huge data). Therefore, we added many information on conditions and parameters that were utilized during building DM models on the basis on selected algorithms. Using this conditions set it will be easier to reproduce results in article. Additionally in text were added information needed for enhancing a description of tests procedures.

13. Pg.10 Ln.361, the abbreviation of “Random forests (RT)” is wrong.

We did it by mistake. This error was provided in text.

14. Please add necessary legends and notes in Fig.5 and Fig.11, the quality of the figures needs to be further improved.

We improved the resolution and some formats in a figure to be more readable.

15. Please improve the legends and notes in Fig.8.

We tried to improve the resolution and some formats in a figure to be more readable.

16. Whether the results (RMSE and the R²) in Tab.7 meet the requirements of statistical analysis?

Thank you for this remark. Of course we forgot to emphasis more pronounced that the error increase was not greater than 10% (especially MAE). That’s why taking in consideration the WT-2/2014 recommendation this result we considered as  satisfactory.

17. Please check whether the content about the influence of compaction on the sample is missing, which is one of the purposes stated in line 253

Thank you for this remark. We agree with reviewer that the need of  compaction was not emphasized enough. The explanation was done by mean of additional Fig. 3  and detailed explanation was inserted in text.

Reviewer 2 Report

Overall, this is a well written paper with an interesting study. However I found that Figure 9 need a little improvement, please remove the line connected between agregate types (both solid and dashed lines) considering no relationship betweem them

Author Response

(in text corrections were marked green color)

Overall, this is a well written paper with an interesting study. However I found that Figure 9 need a little improvement, please remove the line connected between agregate types (both solid and dashed lines) considering no relationship betweem them

Thank you for your review. According to this we modified Fig 9 in line with reviewer recommendation.

Reviewer 3 Report

This is an interesting paper. The review comments are as follows：

1）In the last part of the introduction, the purpose and main research contents of the paper are explained, and the research flow chart is supplemented.

2）Supplement aggregate performance information in Section 2.1.

3）The conclusion part is too much to be simplified and clearly clarified.

Author Response

(in text corrections were marked  light blue color)

1）In the last part of the introduction, the purpose and main research contents of the paper are explained, and the research flow chart is supplemented.

2）Supplement aggregate performance information in Section 2.1.

3）The conclusion part is too much to be simplified and clearly clarified.

Thank you for reading and commenting on this manuscript. We add  some data for section 2.1 and modified conclusions to logically simplified their reception