Classification of Partial Discharge Sources in Ultra-High Frequency Using Signal Conditioning Circuit Phase-Resolved Partial Discharges and Machine Learning

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is an interesting and useful article that discusses the use of machine learning to classify partial discharge sources in UHF. The article suitably combines both theoretical and practical experimentation to form a useful information resource for work in this area. The paper is well-written and logically structured.

This work continues the research of the authors and a number of their previous publications are cited. These citations are suitable for the publication.

Specific comments and minor typos to consider:

1. Page 1, line 45, "partially circuit". Can this phrase be elaborated.

2. Page 2, line 51 "evolution ,". Unexpected space.

3. Page 4, line 1. "Figure 2" is bold text.

4. Page 4, lines 161, 163, and 165. There a missing spaces between words.

5. Page 7. "SelectKBest". Can the authors provide more information on why this was the best selection for the algorithm and what other algorithms were considered?

6. Page 7. The authors mention that not all data were valid and the impact that this would have. Were the data values preconditioned in any way to reduce effects such as outlier values?

7. Page 7, Figure 9. "Choosing" might be better worded as "selection".

8. Page 7, Figure 9. "Database" was mentioned but is this the "dataset"?

9. Page 9, line 282. Could additional information on the "most energetic" components be provided as this is an interesting statement to understand in more detail.

9. It would be interesting and useful to include additional discussion into the algorithm/model developed for the machine learning and the complexity in terms of development, training, and inference.

10. Page 12, Tables 3 and 4 captions. Should "database" be "dataset"?

 

Author Response

Reviewer comments

Reviewer 1

Q1 - Page 1, line 45, "partially circuit". Can this phrase be elaborated.

Response: Thank you. The sentence was re-elaborated.

Q2 - Page 2, line 51 "evolution ,". Unexpected space.

Response: Thank you. The space was deleted.

Q3. Page 4, line 1. "Figure 2" is bold text.

Response: Thank you. The bold text was corrected.

Q4. Page 4, lines 161, 163, and 165. There a missing spaces between words.

Response: Thank you. The bold the spaces were added

Q5. Page 7. "SelectKBest". Can the authors provide more information on why this was the best selection for the algorithm and what other algorithms were considered?

Response: Thank you for the pertinent question. The mutual information is quite interesting because it considers data on entropy and degree of proximity of the K-near neighbors. Important references such as https://0-doi-org.brum.beds.ac.uk/10.1016/j.procs.2024.04.326 and https://0-doi-org.brum.beds.ac.uk/10.3390/s23156664 use this Python KBest methodology to reduce the size of the dataset to be evaluated. In our work, we use KBest to sort the features that would serve as input to the database. In addition, we used this algorithm due to its wide use in the literature and its proven efficacy.

Q6. Page 7. The authors mention that not all data were valid and the impact that this would have. Were the data values preconditioned in any way to reduce effects such as outlier values?

Response: Thank you very much for the very pertinent question. In this work, the preconditioning of the signals that was used was filtering. We talk about data processing in the computational procedures and Threshold filtering sections. There are sequences of acquisitions in which no PD signal is acquired, due to the stochastic characteristic of the phenomenon, and adding data that are only noise would not help in the characterization of the phenomenon. In the document, the excerpt has been improved to be clearer.

Q7. Page 7, Figure 9. "Choosing" might be better worded as "selection".

Response: Figure changed.

Q8. Page 7, Figure 9. "Database" was mentioned but is this the "dataset"?

Response: Exactly. The right word is dataset. Figure changed.

Q9. Page 9, line 282. Could additional information on the "most energetic" components be provided as this is an interesting statement to understand in more detail.

Response: The focus given in the article was to highlight that there was a reduction in the sampling frequency necessary to acquire the signals with the conditioning system, which reduces the costs inherent to this acquisition. In this case, the term energy is used as a synonym for amplitude. Thus, the components with significant amplitudes, or more energetic, are those up to 50 MHz. More detailed analyses in terms of frequency were not the main objective of this article, however they can be evaluated in future works.

Q10. It would be interesting and useful to include additional discussion into the algorithm/model developed for the machine learning and the complexity in terms of development, training, and inference.

Response: All algorithms used were from Python's Scikit-learn library. The text has been improved so that this information is more clearly stated.

Q10. Page 12, Tables 3 and 4 captions. Should "database" be "dataset"?

Response: Yes. The words were changed.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This paper investigates detection and classification of phase resolved partial discharge patterns, using a printed UHF monopole antenna and a signal conditioning circuit which reduces some hardware requirements.  

For this, the authors proposed a new methodology using neural networks, support vector machines, and decision trees. In their experiments, they evaluated 108 different configurations in total, and the experimental results support the usefulness of the methodology proposed by the authors. 

 

Overall, this paper contains some meaningful results which are worth publishing, so I would recommend this paper for publication in Electronics. 

 

However, the paper is written in poor English and therefore needs a major revision on its English writing and presentation.

 

For instance,

 

p. 3, line 120: "and its photograph is show in Figure 2" 

→ "and its photograph is shown in Figure 2" 

 

p. 4, line 134: (past/present tense mismatch) "the measurement is carried out with the UHF sensor that was a printed monopole antenna" 

→ "the measurement is carried out with the UHF sensor which is a printed monopole antenna"

 

p. 15, line 426: "Threshold filtering technique was performed on the signals, and it was found that the amount of valid PRPD per database was above 90%."  

→ "Threshold filtering technique was performed on the signals, and the amount of valid PRPD per database was above 90%."  

 

p. 15, line 428: "The feature extraction considered the phase information of the filtered signals"

→ "The feature extraction takes into account the phase information of the filtered signals"

 

p. 16, line 444: "As summary, some contributions of this work can be highlighted"

→ "In summary, some contributions of this work can be highlighted" 

 

Author Response

Reviewer comments

Reviewer 2

Q1. p. 3, line 120: "and its photograph is show in Figure 2" → "and its photograph is shown in Figure 2".

Response: Thank you for the opportunity to improve the English quality. This sentence was changed.

Q2. p. 4, line 134: (past/present tense mismatch) "the measurement is carried out with the UHF sensor that was a printed monopole antenna"  → "the measurement is carried out with the UHF sensor which is a printed monopole antenna".

Response: This sentence was changed. And the verbs were inflected to the past tense. A thorough revision of the English and verb tenses was carried out to maintain the uniformity of the text. Thank you.

Q3. p. 15, line 426: "Threshold filtering technique was performed on the signals, and it was found that the amount of valid PRPD per database was above 90%."→ "Threshold filtering technique was performed on the signals, and the amount of valid PRPD per database was above 90%.".

Response: This sentence was changed to your suggestion. Thank you.

Q4. p. 15, line 428: "The feature extraction considered the phase information of the filtered signals"→ "The feature extraction takes into account the phase information of the filtered signals"

Response: This sentence was changed to: The feature extraction considers the phase information of the filtered signals.

Q5. p. 16, line 444: "As summary, some contributions of this work can be highlighted"→ "In summary, some contributions of this work can be highlighted"

Response: This sentence was changed to your suggestion. Thank you.

Author Response File: Author Response.pdf