Two-Leak Isolation in Water Distribution Networks Based on k-NN and Linear Discriminant Classifiers

Round 1

Reviewer 1 Report

REVIEW

 

General Comment

The paper presents a new approach to locate two simultaneous leakages in a WDN. This is a very interesting topic and the paper is well written, with sufficient references and a good presentation of the techniques used. However, a major drawback is the case study presented. The WDN network is too small to indicate the feasibility of the methodology for real cases. In addition, it is not clear what is the leakage flow rates effectively used in the results presented and how this technique is an improvement compared to other techniques that focuses in locate a single leakage. In real cases, two leakages occurring simultaneously are still far from the reality. Thus I strongly recommend the authors to provide a case study in a more complex network to effectively understand the accuracy and improvement of the proposed technique.

 

Specific Comments

1. Introduction

The introduction is well referenced, presenting several approaches for leak location. However, as it is written, it implies that the only problem of these techniques is detect multiple leaks occurring simultaneously. I suggest that the authors some comments about the several difficulties and uncertainties on the process of leak detection.

Line 56 – define FIR

 

2. Material and Methods

Line 105 - I suggest the authors to comment in section 2.1 how the number of sensors to be used can be chosen.

Line 218 – figure 2 should be cited and explained in text

 

3. Results

Line 236 – assuming a calibrated network is far from the reality. I suggest the authors to perform the study subject to uncertainties, calibrating the network with the noise pressure data described in the following sections. In addition, the Hanoi network is too small to provide any reliable result that could be feasible in real WDNs.

Line 238 – correct the Table number. Or even, remove the table, as figures 4 and 5 provide the same information

Line 260 – scenarios A to C could be presented in a single section. In addition, what is the main difference between them to justify their study? Finally, what are the leakage flow rates considered? These values could seriously bias the results

Comments for author File: Comments.pdf

Author Response

Please revise attached file, section "reviewer 1".

Author Response File: Author Response.pdf

Reviewer 2 Report

I'm very sorry, although the authors have done a lot of work, the presentation of the manuscript (whether it's text or graphics) is difficult for people to agree with.

I'm very sorry, there are many English expression errors in the manuscript.

Author Response

Please revise attached file, section "reviewer 2".

Author Response File: Author Response.pdf

Reviewer 3 Report

Brief summary

This paper is interesting for those who deal with water distribution network management, particularly to those dealing with water loss control.

The paper presents a methodology to locate two simultaneous leaks in water distribution networks, using two well-known classifiers: k-NN and discriminant analysis (DA). Results show that DA presents a better accuracy than k-NN.

The paper follows a traditional structure, it is well written, and it has a good set of references, most of them quite recent.

Although this is an interesting paper, there are some issues that should be better addressed.

The methodology was designed to locate two leaks in a water distribution network and, as far as I understood, both leak flows must be equal. This seems to be quite restrictive for real-world application. Is this true?

Nothing is mentioned about it, but I guess that the methodology assumes that the node demands are perfectly known, and we have a quite accurate (calibrated) model of the network. Is that so?

Authors mention that some uncertainties in the pressure measurements and in the leak flows were considered, but nothing is said about the level of these uncertainties.

The Hanoi network is used in the literature as an example for testing design methodologies and can be seen as a major skeleton from a water distribution network, made of pipes from 12” (300mm) to 40” (1,000mm), with huge flows (the peak flow is almost 20,000 m3/h – more than 5,500 l/s):

- Why using this example and not any other one more representative of a common real-world network?

- There are many different design solutions published. Which solution was used in this paper (pipe diameters)?

- The common data for this network is only the peak demands to be used in the design. Which average demands were considered, and which daily pattern was assumed?

- This network peak flow is quite huge (almost 20,000 m3/h) and so 0.27%-2.7% (8-80 l/s) of the “nominal flow” (2,890 l/s) is quite a huge flow for leaks, it seems more like a burst. Why not using smaller leak flows?

Authors should mention, not only the locations, but also the leak flows considered in the three leak scenarios presented (A, B and C).

From my experience, even assuming that the node demands are known and assuming that we have an accurate model of the network, it is quite difficult to accurately locate several leaks, with different flows, and considering common leak flow values (not huge values seeming like bursts). But unfortunately, this is what we find in the real-world. I wonder if authors could share their opinion about this subject and refer to the applicability of these methodologies (k-NN and DA) in real-world conditions.

To conclude, I think this is an interesting paper, and, in my opinion, authors should be encouraged to rewrite the paper taking into consideration the reviewer comments and resubmit it.

Specific comments

Line 56: Explicit the meaning of “FIR”.

Line 85: Suggest changing “water distribution network” to “WDN”.

Line 141: Suggest changing “an leak” to “a leak”.

Line 146: Suggest changing “water distribution networks” to “WDNs”.

Line 175: Suggest changing “matrix the (19)” to “matrix (19)”.

Line 207: Suggest changing “classes in as large” to “classes is as large”.

Line 232: Suggest changing “water distribution network (WDN)” to “WDN”.

Line 238: Suggest changing “Table ?? present” to “Table 1 presents”.

Line 240: Suggest changing “Fig. 4 and Fig. 5 presents” to “Fig. 4 and Fig. 5 present”.

Line 283: Suggest changing “Figure 14” to “Figure 15”.

Author Response

Please revise attached file, section "reviewer 3".

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors improved the paper sufficiently

Reviewer 2 Report

The modified content is quite complete, and there are currently no other issues.