Addressing Uncertainty by Designing an Intelligent Fuzzy System to Help Decision Support Systems for Winter Road Maintenance
Round 1
Reviewer 1 Report
This is a novel work of design of the efficient and effective winter road maintenance based on an accurate prediction of the road surface friction coefficient. To do so the authors take advantage of adaptive neuro-fuzzy networks to be able of reduce the uncertainties in historical data about the weather and road conditions. The objective is to better determine the quantity of salt to use to reduce the accidents.
Based on this very useful information and on the very novel approach I strongly recommend the article for publication in the present state.
Only some small changes and questions, listed below, need to be addressed
Line 62, when you make a citation it is not necessary to put the name of the authors with the style citation chosen by you.
Line 106 and 107 explain better how “the main contribution of this study is to design an ANFIS model to predict RSFC using real data measured by optical and road-mounted sensors”, because previously in the text you say in line 88 that “optical and road mounted sensors are mostly used to measure the data related road surface conditions. Sometimes, numerical data derived from sensors can be associated with uncertainty due to imprecision, 9vagueness, or ambiguity”.
Line 128, change: “As you can see in Figure 1, the framework of this study includes different steps that 127 are explained in this section.”
By the observation of the Figure 1 is possible to realise that the framework of this study includes different steps that are explained in this section.
In table 1 please uniformize the number of significative numbers for each quantity.
Line 176 More information is required for a better undertsanting of the section: Dividing the Dataset into Training and Testing Sets. As an example, from where came the values in the section 3.5Training Using ANFIS?? No references were given for the initial step size, the decrease rate and increase rate.
Comparing Figure 7 and figure 8 is possible to realise that in some cases the quantities in study and the errors are of the same order of magnitude, what kind of explanation you can provide to these?
Author Response
Deep thanks to the editor and reviewers for their valuable comments. Attached files are my response to the reviewers comments.
Best regards,
Mahshid Hatamzad
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Line 24: What this number mean? R?
Line 131: Though these input variables seem rational, some documentation and reference must be provided.
In figure 1:
- into what? I guess training and testing datasets?
- What does the ANFIS?Predict accurately I guess once again. What is the criterion for a prediction to considered "accurate" or not? Is there a certain number? A MAPE% or something else?
- There are some grammar mistakes. Please pay attention to this.
Table 1: Some explanation of this table must be provided.
Line 163: Does the training-testing percent ratio affects the accuracy of the system? Is there any rule or a formula indicating that a 70-30 arrangement is optimal? If yes, please refer.
Line 172: Why are you so confident that MATLAB's tool default values are the optimal for your problem?
Table 7: In the abstract, the training and testing procedures considered to demonstrate "excellent capability" to learn (0.965) and to predict (0.962). Apart from the fact that you don't acknowledge this numbers (I just assume that it is R or R2), you aren't mention them in the whole script as findings/results.
Legend in figure 7, page 9: what exactly "special" means? Why is this interval special?
Line 286: Some explanation of the graphs must be given.
Author Response
Deep thanks to the editor and reviewers for their valuable comments. Attached files are my response to the comments.
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors investigate the estimating model of RSFC. This paper is well organized and easy to understand. Please note the following specific comments.
Page 3, line 135-137: What is the pavement type of the test site? A brief introduction of the testing conditions of air temperature, the thickness of snow layer and ice layer is need, since the measuring methods and accuracy are important for the evaluating the validation of the modeling work.
By the way, did authors consider testing the thickness data and the surface temperature of the water film on the pavement surface? Furthermore, the effect of the surface temperature of the
Page 8, line 244-245: Please analyze the affecting factors of the obvious errors that ’Only a few numbers of them show obvious errors’ in the manuscript.
Page 11, line300: How to define ‘the amount of ice layer’? Furthermore, how to define the thickness of ice layer in this manuscript?
Author Response
Deep thanks to the editor and reviewers for their valuable comments. Attached files are my response to the comments.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The authors replies are satisfactory.
Author Response
The authors deeply appreciate your comment.
Author Response File: Author Response.pdf
Reviewer 3 Report
It seems that the testing details and conditions at Testsite 18 are insufficient in the webpage provideded by the authors. The key testing conditions influencing the validation and accuracy of testing data should be supplemented in the revised manuscript to facilitate readers.
Author Response
The authors deeply appreciate your comment.
Author Response File: Author Response.pdf
