Cost–Benefit Prediction of Asset Management Actions on Water Distribution Networks
Round 1
Reviewer 1 Report
p.p1 {margin: 0.0px 0.0px 0.0px 0.0px; font: 11.0px 'Helvetica Neue'; color: #000000} p.p2 {margin: 0.0px 0.0px 0.0px 0.0px; font: 11.0px 'Helvetica Neue'; color: #000000; min-height: 12.0px} span.s1 {font-kerning: none}Readability of the paper improved and it is now more clear. By carefully reading the paper again, still some minor errors have been found:
Abstract: Acronyms WER and OPEX are used without introducing them previously.
Line 65: there is a whitespace after the reference
Line 97: Whitespace between SISPEA and bracket is missing
Line 152: Based …
Line 155: Whitespace after years missing
Line 176: Figure 1 instead of Figure 11
Equation 11 and 12: Suggestion: Write equation 12 and use braces under the terms with text that belongs to pipe reparation, connection reparation and leak detection. Hence, you can save one equation and it will be clearer to the reader which terms belong to what.
Line 232: Empty line
Figure 5: Very similar to figure in Deb et al. 2002, just minimisation instead of maximisation. I think therefore a reference with amended from … is necessary.
Line 407: … management defines We aim … I think here is something missing.
Line 409: ANN instead of Ann
Line 415: delete each
Line 417: Why is relu boldface here and not in Line 210?
Figure 8: Caption missing dot and whitespace
Line 502: NSGA II instead of NAGA II
Line 506: delete ii)
Check all equations that the numbering is aligned on the right side with all other equation numbers.
Check all references that the formatting is consistent. There are many paragraph breaks in the references, e.g., between [11] and [12].
Reference 19: check the formatting of the names, í instead of i´
Author Response
RESPONSE TO THE REVIEWERS COMMENTS
We would like to thank reviewers for their work and the quality of reviewing. We answer to their comments in the following document. We hope that all unclear aspects will be addressed. Thank you for your help to improve our research.
Reviewer # 1
Readability of the paper improved and it is now more clear. By carefully reading the paper again, still some minor errors have been found:
Abstract: Acronyms WER and OPEX are used without introducing them previously.
Addressed
Line 65: there is a whitespace after the reference
Addressed
Line 97: Whitespace between SISPEA and bracket is missing
Addressed
Line 152: Based …
Addressed
Line 155: Whitespace after years missing
Addressed
Line 176: Figure 1 instead of Figure 11
Addressed
Equation 11 and 12: Suggestion: Write equation 12 and use braces under the terms with text that belongs to pipe reparation, connection reparation and leak detection. Hence, you can save one equation and it will be clearer to the reader which terms belong to what.
Thank you for the advice but this modification implies many modifications, link between equation 11 and 12 seems to be obvious and each term seems to be clear.
Line 232: Empty line
Addressed
Figure 5: Very similar to figure in Deb et al. 2002, just minimisation instead of maximisation. I think therefore a reference with amended from … is necessary.
Addressed
Line 407: … management defines We aim … I think here is something missing.
Addressed
Line 409: ANN instead of Ann
Addressed
Line 415: delete each
Addressed
Line 417: Why is relu boldface here and not in Line 210?
Modified
Figure 8: Caption missing dot and whitespace
Addressed
Line 502: NSGA II instead of NAGA II
Addressed
Line 506: delete ii)
Addressed
Check all equations that the numbering is aligned on the right side with all other equation numbers.
Done
Check all references that the formatting is consistent. There are many paragraph breaks in the references, e.g., between [11] and [12].
Addressed
Reference 19: check the formatting of the names, í instead of i´
Addressed
Author Response File: 
Author Response.docx
Reviewer 2 Report
This is a resubmission of a previous paper.
I consider that the article, which is well structured, is almost ready for publication except for (too many) formal questions that should be considered by the authors:
Review formal issues such as:
-Why some of the n’s in table 2 are not in italics? Typically, all the variables are in italics!
-In this table n_c means ‘number of connections’, while in table 3 means ‘number of leaks on connections per year’. Please, in general, check for consistency of notation.
-Why eq (1) is bold? Missing left parenthesis!
-Please align equations and their numbers uniformly.
-Do not indent lines after equation beginning with lowercase word such as in lines 130, 134, 152, 161, etc. (there are many instances)
-Check for lack of blanks, as in line 155. Also for repeated consecutive blanks in many places.
-Use the same fonts for variables both in-text and in equations. Check, for example, C_tot in lines 184 (in-text) and 186 (in eq 10). Also consider these observations for variables in the figures; for example, compare n_inv in fig 1 and in eq 6. By the way, is n_inv defined anywhere? Check for all the variables: n_p, eps_m, etc. look different in different places. One can guess they are the same variables, irrespective of their aspect. This is not good practice.
-Eq 9 does not maintain size proportionality with the rest.
-Quality of figures should be improved. Also consider their size to be balanced with the rest of the text.
-Numbers shouldn’t be in italics: check X_1 and X_2 in section 2.4.1 and compare with the same symbols in eqs 28 and 29. Don’t forget they also should be the same font.
-Variables in fig 5 should (once more) be in italics.
-Check last parenthesis in eq 32
-‘if’, ‘or’ and ‘and’ shouldn’t be in italics in eq 41, as they are not variables.
-Extra semicolon in L386
-Some equations have extra space up and down, while others do not. See, for example, eqs 42-45.
-L417, why 'relu' is bold?
-L418, N should be in italics.
-Figure 8, missing blank
-Why left column of table 8 is in bold?
-L507, missing blank, P in italics
-Please revise references for uniformity. For example, are Journal names abbreviated or not? Also consider other aspects. Check for Journal style and tightly adjust to it.
-References 18, 19, 24, 25, 28 and 31 are not suitably formatted.
All in all, it is this reviewer feeling that the authors should be much more careful with so many details!
Fortunately, English is much better than in the previous version.
I still think that the paper may be useful to WDS managers, and it is my opinion that it can be published after the above said minor revision. Please, be more scrupulous with your work.
Author Response
RESPONSE TO THE REVIEWERS COMMENTS
We would like to thank reviewers for their work and the quality of reviewing. We answer to their comments in the following document. We hope that all unclear aspects will be addressed. Thank you for your help to improve our research.
Reviewer # 2
Comments and Suggestions for Authors
This is a resubmission of a previous paper.
I consider that the article, which is well structured, is almost ready for publication except for (too many) formal questions that should be considered by the authors:
Review formal issues such as:
-Why some of the n’s in table 2 are not in italics? Typically, all the variables are in italics!
Addressed
-In this table n_c means ‘number of connections’, while in table 3 means ‘number of leaks on connections per year’. Please, in general, check for consistency of notation.
We agree, we modified as requested and consider: nc the number of connections, and nlc the number of annual leaks on connections. All equations where nc appears have been modified.
-Why eq (1) is bold? Missing left parenthesis!
Addressed
-Please align equations and their numbers uniformly.
Addressed
-Do not indent lines after equation beginning with lowercase word such as in lines 130, 134, 152, 161, etc. (there are many instances)
Addressed
-Check for lack of blanks, as in line 155. Also for repeated consecutive blanks in many places.
We checked all the paper in order to fix it
-Use the same fonts for variables both in-text and in equations. Check, for example, C_tot in lines 184 (in-text) and 186 (in eq 10). Also consider these observations for variables in the figures; for example, compare n_inv in fig 1 and in eq 6. By the way, is n_inv defined anywhere?
Addressed, we add the definition of ninv in Table 2 as : Number of invisible breaks/leaks.
Check for all the variables: n_p, eps_m, etc. look different in different places. One can guess they are the same variables, irrespective of their aspect. This is not good practice.
We checked the paper in order to address your comment
-Eq 9 does not maintain size proportionality with the rest.
Addressed
-Quality of figures should be improved. Also consider their size to be balanced with the rest of the text.
Figures are provided as an attachment in pdf format, their quality was improved.
-Numbers shouldn’t be in italics: check X_1 and X_2 in section 2.4.1 and compare with the same symbols in eqs 28 and 29. Don’t forget they also should be the same font.
Addressed
-Variables in fig 5 should (once more) be in italics.
Addressed
-Check last parenthesis in eq 32
Addressed
-‘if’, ‘or’ and ‘and’ shouldn’t be in italics in eq 41, as they are not variables.
Addressed
-Extra semicolon in L386
Addressed
-Some equations have extra space up and down, while others do not. See, for example, eqs 42-45.
Addressed
-L417, why 'relu' is bold?
Addressed
-L418, N should be in italics.
Addressed
-Figure 8, missing blank
Addressed
-Why left column of table 8 is in bold?
Error
-L507, missing blank, P in italics
Addressed
-Please revise references for uniformity. For example, are Journal names abbreviated or not? Also consider other aspects. Check for Journal style and tightly adjust to it.
-References 18, 19, 24, 25, 28 and 31 are not suitably formatted.
All in all, it is this reviewer feeling that the authors should be much more careful with so many details!
We agree and think you so much for your time and your help !
Fortunately, English is much better than in the previous version.
I still think that the paper may be useful to WDS managers, and it is my opinion that it can be published after the above said minor revision. Please, be more scrupulous with your work.
Author Response File: Author Response.docx
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
A very interesting article. As for the methodology itself, I have no objections. But there are numerous editorial shortcomings:
- verse 15 our model => author's model
- line 38, double space
- line 40 operation. [2] discuss -?
- line 46 we => Authors
- line 50 55, 56, 62, 66, 68, 71, - citation is not clear
- line 125 - equation is not clear - 2 (t)?
- all equations numbers should be aligned to the right
- line 154 - underlining one character in the word annual'-
- line 202 - description of the sum symbol should be under and above the sum mark
- line 287 - the sentence starts with: k Elements - it does not look good
- line 319-323 - different distance between lines
- table 8. OPEX - the number should fit in one line - this way it is illegible
- line 526, 530 - different edition after the citation number
- line 566 - smaller font
-
Reviewer 2 Report
p.p1 {margin: 0.0px 0.0px 8.0px 0.0px; font: 12.0px 'Helvetica Neue'; color: #000000; -webkit-text-stroke: #000000} span.s1 {font-kerning: none}Brief summary
This paper aims to predict costs and benefits of water distribution system asset management. It proposes a combination of (1) an artificial neural network (ANN) as a surrogate model for predicting the effect of different asset management actions on water efficiency rate; and (2) optimising the efficiency rate and the cost with a multi-objective optimisation algorithm utilising this surrogate. Furthermore, if data is lacking for a specific water distribution system, the ANN can be fitted with multi-set of data from several water utilities or a national database in combination with measurements from the current investigated system.
Broad comments (C)
C1: The research gap is not clearly defined and it is not possible for the reader to judge after the literature review, if the current paper is original nor where the novelty lies. I recommend the authors to emphasise more on the novelty and the differences of the current paper to former literature.
C2: In my opinion, the current manuscript can be tremendously improved by an overview figure at the beginning of the methods section that explains, how the different parts of the methods are connected and brought together.
C3: The methodology section of the paper focusses broadly on standard methods like ANNs and the NSGA2 algorithm, methods that can be easily referred to in past literature. Parts that are specifically interesting for the readers of this paper, e.g., the estimation of the water losses based on expert opinion and Monte Carlo analysis is only mentioned, but never explained in more detail. I recommend to introduce additional sections in the methodology section describing this methods.
C4: Parts of the methodology section are missing crucial citations (more details in the specific comments).
C5: Only tournament selection is explained in the methods section and the text, but an explanation of random and roulette wheel selection (see Table 6) is missing. It also seems that the selection methods got unintentionally swapped, since section 3.3. uses roulette wheel selection and section 3.4. uses tournament selection, but shouldn’t that be similar calculations with the same NSGA2 selection methods?
C6: The English of this paper is partially poor and can be improved. Throughout the paper there are several places, where sentences are too long and the structure of the sentence sounds strange. Recommendation to the authors: Let a native speaker check the paper to improve the readability.
C7: There is no reference to Table 3 in the text.
C8: The quality of the Figures is mostly insufficient. The resolution of figures is not high enough, so that the text within the figures is very blurry. Furthermore, the font sizes for axis labels, legends or annotations in the figures are most of the time very small and, hence, unreadable. This is true for Figures 1, 2, 3, 4, 5, 6, 7. Furthermore, are Figures 1 and 4 taken from literature? Then a reference to the literature is missing.
C9: Equations within the manuscript are sometimes inconsistent or erroneous. The notation and naming of some variables is sometimes very confusing (more details in the specific comments). In my opinion, the comprehensibility of the equations would benefit from shorter and more different variable names and make vectors boldface.
C10: The formatting of the references is not consistent, especially concerning the volume and page numbers of journals (sometimes xx (x) xxx-xxx as in [2], sometimes xx(x):xxx-xxx as in [3] and sometimes xx(x) xxx-xxx as in [5]). Please correct this.
C11: Some interesting questions are not addressed throughout the manuscript, e.g., (1) How the material mix of water distributions system with different failure rates can be incorporated within this method or (2) How does the asset management effect input parameters? For example, increasing l_det will decrease the number of leaks in the systems, once repaired and will make the leakage rate smaller of the system. Is this incorporated in the model?
C12: I am missing a comparison of the proposed method to costs and WER compared to commonly used strategies of water utilities. This can be specifically done for the investigated network.
Specific comments
13: The sentence “Maintenance actions concern both …” is not clear to the me. Maybe there is a word or a part of the sentence missing?
30: Explain in more detail what you mean by ex-post and ex-ante? What is the difference in using other more commonly word pairs, e.g. a priori - posteriori or even more simple before - after?
37: “This shortfall tends to be solved” Is this really the case for water utilities nowadays?
43: Split into two sentences “… decision-making process. Data gathering …”
43: “… a mine that should be exploited …” The language sounds a little bit flowery.
46: We aim to answer the following question:
46: “How the existing data collection of IS can be exploited for prospecting …” Does this paper really answer this question? Because this point needs then further discussion and elaboration in the following chapters throughout the paper.
60: “Authors assume that the optimal time for pipe renewal is reached when expected benefits are greater than costs“ This is quite logical.
62: [10] discuss …
63: Authors assume that there exist two …
66: “The use of machine learning seems relevant to tackle…” Why? Explain further.
73: Pipe deterioration ratio, …
76: get rid of “in order”
83: ..for calibrating a decision model that is able to predict …
85: Explain what you mean by low-level and high-level data.
88: “smart water systems” not “water smart systems”
92: Looks like SISPEA is the shortcut of “…mandatory KPIs”
94: current section instead of first section
102: delete “As a contribution,”
103: … functions: (1) the water efficiency rate considered as benefit and (2) the total cost obtained …
110: Missing reference to decree of May 2007. What is the difference to IWA definitions?
Table 2: Is debit the right word? What about cost?
Equation (1): Why boldfaced? The 2 is overlapping with (t). I think it is better to use variable names instead of VP.232
Equation (2): What does the line over Age mean?
Equation (3): r_m is suddenly introduced without explanation, because Table 3 is never introduced within the text of the paper. That makes it for the reader very hard to understand the equations.
134: How is the number of leaks estimated?
138: Explain in more detail how the terms of equation (6) are derived.
Equation (8) Equal sign is missing. Furthermore, a better representation would be (latex code) $\frac{1}{5} \sum_{i=0}^{4} rp(t-i)$
151: Which are the remaining explanatory variables?
170: References to ANN literature are missing in this section
176: delete “by the equation ”
180: delete “By generalising (16), the vector neuron^k grouping all values …”
Equation (17): Maybe use other letter than W do not mix it up with leak rates.
192: Delete Therefore,
196: input variables are known
199: Rearrange sentence “The function …”
200: Rearrange sentence “A function …”
203: where without capital letter, maybe end equation above with comma
204-206: get rid of bullet points
Equation (19): L in equation (18) is not a function of w_i, so the derivative is zero, which means that G_i(t) will be zero for all t in equation (20).
Equation (23): l_det_min and l_det_max, etc. are not defined
245: “The population size is set in advance” - To what and what is the reasoning behind that?
258: What is the considered set P
259: Crowding distance section is missing a reference
268: Switch order of equation (19) and (20), first defined d and the explain the boundaries
286: Selection method section is missing a reference
299-301: “It appears that some …” This sentence is very vague. I recommend the authors to have a look at the No-Free-Lunch Theorem (Wolpert, D.H., Macready, W.G. (1997), "No Free Lunch Theorems for Optimization," IEEE Transactions on Evolutionary Computation 1, 67. )
303: Explain in more detail why flat crossover has been chosen.
305-315: The first variable in the vector is a continuous variable, the other three variables are integers since the rate of annual meter renewal is discrete. Hence, the problem is a mixed-integer optimisation problem. Flat crossover is an operator working in continuous space, which may lead to problems when applied on the discrete dimensions (r_p, r_c, r_m). How is this problem circumvented?
320: “The mutation is an operator …to escape from local optima thanks to a big jump …” In general, this is not true. The purpose of the mutation operator is usually to introduce small changes and not big jumps to avoid local minima by preventing the population of chromosomes from becoming too similar to each other
321: The reference [25] is wrong. Polynomial mutation is not introduced in this paper.
327: “This must be reduced to P”. Why?
343: Delete “and calculated by equation (41)”
346-350: Restructure this paragraph to avoid forward references to equations.
361: delete “a pipes”
362: “…detecting 14 leaks on average.” Does this number decrease when leak a water utility is actively searching for leaks more often? How is this considered in the presented approach?
363: Delete “We notice low-level renewal rates. In fact,” and “the following”
363: We aim at improving WER by conducting alternative asset management actions at lower costs that commonly used strategies.
366: Ann is fitted on the SISPEA data
367: Delete “As a reminder”
368: Delete “years”
368: The data is split …
372: Missing comma between [16] and many
374: delete “It appears that”
374-375: three hidden layers with 144, 36 respectively 9 neurons at each layer.
375: activation function is relu …
376: Delete the two sentences, because this is already mentioned multiple times.
378: I suggest to emphasise a little bit more on explaining the expert opinion and Monte Carlo analysis, since this is very interesting for the reader. Maybe it is a good idea to shorten the descriptions of ANN and NSGA2 and introduce subsections focused on this topic.
Table 5: Comparison of the observed and the predicted WER between 2010 and 2016. Don’t break the table over multiple pages.
387: random selection and roulette wheel selection is not explained nor is there a reference to literature where you can find it.
389: Explain why the specific values for the population size and the number of generations is chosen.
398-399: It depends on the form of the Pareto front, if the population moves forward to the true front. If there is a bit valley directly in front of the Pareto front, that the optimisation procedure can’t overcome, then it won’t move in the direction of the Pareto optimum.
401: “The front moves to optimise objective functions.” What does that mean?
404: P=200
407: front stabilises during the last generations.
407-408: “The obtained front confirms the relevance of using the roulette wheel selection” Why?
411: “tournament selection” The former section says that Roulette wheel selection is used.
416: delete “equal to”
Table 8: Use mathematical symbols as column heads. Length of leak detection km instead of m?
Discussion: In my opinion the discussion section has to be rewritten, since the methods and results are not really discussed.
438: “seems to be coherent with the practice”. Why?
439: “seems to be accurate and consistent”. Why?
443: “This is really appreciated by the water utility manages.” Why?
459: “…render the model difficult to implement by the water utility because it requires specific skills.” Why should water utilities implement this? Shouldn’t it be made by experts or consultants? Is it possible to find general parameters for all water networks that may not work optimal, but lead to better results than commonly used asset management strategies?
463: “We demonstrate to water utility managers the usefulness of …” This has to be emphasised more, since the interested reader is not able to find this in the paper.
470: “The same if assets are …” I think there is a word missing
478: “The test on WER seems to be positive and successful”. Explain this in more detail.
Reviewer 3 Report
In this article, the authors address the problem of anticipating maintenance actions in water distribution systems, instead of carrying out corrective actions in the event that certain indicators so advise. It is an excellent idea that, as the authors acknowledge, is difficult to implement in the absence of an adequate information system (in real time, if possible). In their proposal, the authors provide possibilities to replace / complement such an information system by / with other mechanisms derived from an exploitation of compulsory databases for water distribution systems. Consequently, the theme of the article is perfectly relevant to WATER and, in particular, to the URBAN WATER MANAGEMENT section. The solution provided by the authors is well developed and, through a case study, they prove its suitability. They also provide guidelines for the extension of the proposed methodology to contexts in which the information system exhibits greater power, including the case of the existence of a true smart water grid.
I consider that the article, which is well structured, is almost ready for publication except for formal questions that should be considered by the authors:
Review formal issues such as:
- lack of definition of MRA in line 82;
- awkward use of "[n] recognize ..." or "[8] introduce ..." at the beginning of several sentences on page 2;
- indentation and use of capital letters after some formulas (as in lines 126, 149, 185 ... among others);
- consistent use of fonts and italics for the variables (for example, in section 2.2, the Xs in 2.4.1, some variables in Figures, etc.);
- other formatting issues (for example: around line 237; line 176; equation (1));
- fix problems in references [11], [12] and [13].
And last but not least,
-have the manuscript reviewed by a native English, as it contains many misprints and deficiencies in language style.
I have read the article with interest, I think it may be useful to WDS managers, and it is my opinion that it can be published after the above said minor revision.
