Application of Multi-Dimension Input Convolutional Neural Network in Fault Diagnosis of Rolling Bearings

Round 1

Reviewer 1 Report

Please give the reasons why CNN is applied for rolling bearing’s fault diagnosis. In general, CNN is specialized in Image analysis. However, the authors uses the CNN for identifying  fault patterns. The explanations are necessary.

The abstract is well-written.

Several mathematical terms (e.g. 1th, jth) in sentences have to be changed with relevant superscripts.

Section 2.1 requires the re-writing, as it explains a general NN (not CNN).

Section 2.2 and 2.3 require to be placed as a separated section.

Give some rationales for “Due to ~ CNN for raw signal”, in Line number 135, Section 2.2.

Figure 1 gives much ambiguities. It needs to be redrawn with more clarities. The integrated explanation with an example may help for clear understanding. In addition. It doesn’t use any mathematical term that are used in the following sections.

Similarly, section 2.2 needs to be expanded.

Equ. (9) has to be redefined or re-explained. It is not suitable currently.

Give a logic for “Step 3” in Section 2.2.1.

In Figure 2, explain more “Accuracy 1,2 and 3”.

In Section 4.1.1, give the reference at “CWRU dataset”.

Figure 4 has to be reformatted with time index, y axis and others.

Table 4 has to be explained more in detail.

In Figure 13, explanations or analogy why 2D-CNN is worse than 1D-CNN, is required.

Author Response

Dear  Reviewer:

Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “Application of Multi-Dimension Input Convolutional Neural Network in Fault Diagnosis of Rolling Bearings” (ID: applsci-525616).Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. Revised portion are marked with different colors (reviewer 1 is red, and reviewer 2 is blue) in the paper. The main corrections in the paper and the responds to the reviewer’s comments are in the PDF file.

The reviewer’ constructive comments are really helpful and invaluable for the authors to improve the quality of this research paper again. In addition, we have perfected English language and style of the manuscript by using a professional English editing service. The authors would be glad to express their sincere appreciation to the reviewers as well as to the editorial members of Applied Sciences, and hope that the correction will meet with approval.

Once again, thank you very much for your comments and suggestions.

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors proposed a data-driven approach based on CNN with multi-dimensional inputs available. Experiments and validations of the models were conducted. I read the paper with interest and provide several suggestions as below.

1) In abstract, “not only maximizes the advantages of the convolution…”, the term maximizes is unclear, best point out in what advantages is optimized.

2) Also, “…verified by experiment 2”, best avoid saying “experiment 2” in the abstract, abstract should enable reader of grasping the essentials with reading the main body.

3) “..and has good robustness and…”, so how good it was? Best also provide the improvement %

4) In the introduction, “3) The classification model is shallow…”, what’s the shallow mean here?

5) “2) Feature extraction process can result in loss of information in some way.”, please specify the way that will cause loss of information from feature extraction process.

6) “1) Signal processing can result in data loss…”, please give examples of how signal processing can result in data loss.

7) Please suggest the minimum recording required for MDI-CNN or the traditional CNN for the fault diagnosis.

8) In addition to the raw data, please list the types of data after signal processing for the multi-dimensional inputs in this work.

9) I suggest the authors indicate the computation environment, as well as the link for the scripts of the (traditional) CNN, especially for the parts which are publicly accessible (if).

Author Response

Dear  Reviewer:

Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “Application of Multi-Dimension Input Convolutional Neural Network in Fault Diagnosis of Rolling Bearings” (ID: applsci-525616).Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. Revised portion are marked with blue in the paper. The main corrections in the paper and the responds to the reviewer’s comments are in the PDF file.

The reviewer’ constructive comments are really helpful and invaluable for the authors to improve the quality of this research paper again. In addition, we have perfected English language and style of the manuscript by using a professional English editing service. The authors would be glad to express their sincere appreciation to the reviewers as well as to the editorial members of Applied Sciences, and hope that the correction will meet with approval.

Once again, thank you very much for your comments and suggestions.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The weak points in the original version are explained in this version.