Pressure Sensor Placement in Water Supply Network Based on Graph Neural Network Clustering Method
Round 1
Reviewer 1 Report
Please see attachment
Comments for author File: 
Comments.docx
Author Response
All replies are in the attachment, please check and read.
Author Response File: Author Response.docx
Reviewer 2 Report
Novel and relevant work. Generally, well presented. Structure is easy to follow and good example cases have been used to illustrate the approach.
Needs English language revision. The English language is generally good and understandable, however, there are issues throughout the paper with missing prepositions (the/a/an) and a few places with sentence construction (particularly in the discussion).
The term "perception" is important to the paper and used throughout. The paper lacks a verbal description of what is meant by "perception" and subsequently "perception rate", "perception node", "perception domain", "nodal perception domain", "partition perception domain/rate" and "perception results" in this context. The authors should strongly consider whether "detection" (and "detection rate" and so on) would not be more descriptive and therefore easier to understand. In the discussion the term "coverage rate" is suddenly introduced (and without any definition). If coverage rate is the same as
Similarly, consider using "zone" in place of "domain" and "partition" (authors should as a minimum ensure consistency in terminology between the abstract and the main text concerning partition/domain. Consider consistently using either "zone" or "partition".
In figure 10, it is not clear what is on the three axes. Either explain axes in the figure text or in the figure itself. Consider whether 3D graphs are actually the best way of illustrating the data shown. It might be easier to grasp the essence of the data with two dimensional graphs with "Coverage rate" (y-axis) against "burst level" (x-axis) with a line for each "number of sensers" (6, 8, 10, 12).
In the discussion, line 390-392, the authors state "However, when σ reaches to a certain high level, most of the nodes in the network can be sensed due to the high pressure drop, so the coverage rate of the monitoring system tends to be flat.". This argument does not tell the whole story. If most nodes in the network could be sensed at high levels then one would expect the blue curves in figure 9 (for 6 and 12 sensors respectively) to flatten out at the same level.
On a practical note. Avoid figures with red and green as contrast colours. (approx. 300 mil. people worldwide are red/green colour blind and will have difficulty seeing the difference). Particularly, figures 6 and 11 should be modified with respect to this.
Author Response
All replies are in the attachment, please check and read.
Author Response File: Author Response.docx
Reviewer 3 Report
The manuscript title: Pressure Sensor Placement in Water Supply Network Based on Graph Neural Network Clustering Method. The paper offers a new framework with robustness and applicability for the decision-making process of monitoring system establishment. Structural Deep Clustering Network algorithm is used for clustering analysis with the integration of complicated topological and hydraulic feature, and WSN was divided into several monitoring partitions.Remarks: What lessons should be drawn from this analysis?Did you examine the effects of model uncertainty? Add some more detailed information about future work, did you concern more complicated hydraulic feature with longer duration time? Some final considerations should be added. Conclusion is not sufficiently described. The main achievements of this paper should be emphasized in the Conclusion.
Author Response
All replies are in the attachment, please check and read.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors tried to address all comments from the reviewers. The revised version of the manuscript is significantly improved compared to its original version. I suggest to accept the paper for publication in Water journal.
