Rectifier Fault Diagnosis Using LTSA Optimization High-Dimensional Energy Entropy Feature

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Aiming at the fault diagnosis problem of electric locomotive rectifiers, the authors propose a fault diagnosis method that combines optimal wavelet transform, energy entropy features, LTSA dimensionality reduction and SVM. Since the methods used are commonly used in the field of signal analysis and pattern recognition, the overall novelty is average; however, the paper has a clear structure and a full amount of work.

The main problems are as follows:

(1) Some of the cited literature is rather dated, and the literature review should be focus on the latest research results in the past three years.

(2) The contribution part of the introduction mainly describes the implementation steps of the research method. It does not reflect the innovation and academic contribution of the paper, and needs to be further refined and modified.

 

(3) Feature dimensionality reduction is actually the compression of high dimensional features into low dimensional features, a certain amount of feature information is also lost in the process of dimensionality reduction. However, in Table 5, the fault detection accuracy of the original energy entropy features, PCA, KPCA, and ISOMAP methods are less than 60%. Why does LTST dimensionality reduction significantly improve the accuracy of recognition models? You are required to verify the accuracy of the results or give the root cause of such results.

(4)A comparative analysis with the existing methods in the literature should be added to illustrate the superiority of the method proposed in this paper.

Comments on the Quality of English Language

English expression can make readers read smoothly.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

1. The state of the art is not sufficient presented. I suggest to extend number of cited paper and present the background of each one.
2. The methodology section should provide a more detailed explanation of why LTSA was chosen for dimensionality reduction instead of other algorithms, such as PCA or t-SNE.
3. There is no explanation of how the neighborhood size (k) in LTSA was selected. Was this value determined experimentally, or are there theoretical foundations for its choice?
4. It is unclear why a specific number of wavelet packet decomposition levels (e.g., five levels) was chosen. Was the impact of this number on the results analyzed?
5. The SVM classification algorithm could be described in more detail, especially regarding the kernel parameters and the training procedure.
6. There is no information on the diagnostic accuracy for individual types of faults. The article reports an average diagnostic accuracy of 99.06%, but these values should be more specifically detailed for each fault class.
7. The section comparing the proposed method with others should include a chart illustrating LTSA's efficiency relative to other dimensionality reduction techniques.

Comments on the Quality of English Language

The article contains numerous grammatical errors and awkward phrasing in English, e.g. "In the during the operation" is incorrect or "are determined are selected" in line 19 or "corresponding to k is 400 is selected" in line 410. I suggest a major linguistic revision of the whole paper.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I have no further comments. 

Comments on the Quality of English Language

I have no further comments. 

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript has been properly revised in accordance with the Reviewer's comments.