Planetary-Gearbox Fault Classification by Convolutional Neural Network and Recurrence Plot
Round 1
Reviewer 1 Report
The reviewed manuscript is aimed at the experimental investigation and diagnostics of a planetary gearbox using the combination of special signal processing methods (particularly mainly recurrence plots) and a convolution neural network. Basic theoretical description of all relevant methods is presented in the paper. Then the particular planetary gearbox with known damaged elements is introduced and measurement of the vibrational signal is explained. The measurements results are then processed using the proposed method and its results are successfully compared with the results obtained by other possible diagnostics methods.
In my opinion, the paper is quite well written. It is purely experimental work with a big amount of signal processing. The originality of the paper lies in the combination of recurrence plots, convolution neural networks and faulty gearbox vibration. I suggest to consider several minor improvements:
1) The list of abbreviations should be written in the beginning of the manuscript.
2) There isn't any m variable in equation (1), however it is explained in the following paragraph.
3) Which software tools are used for the implementation of the whole methodology?
4) How exactly were valuation loss and valuation accuracy evaluated?
5) Values and units should be separated by a space (e.g. 1 m, 10 N, 100 s etc.).
6) Is the methodology applicable (generalizable) also to another gearbox with similar structure but without the knowledge of its vibration signals with faults? Please, explain it in the manuscript.
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The reviewer feels that the paper is interesting, and it is within in the scope of the Journal, but an additional revision should be done. In fact, the subject of the paper is very specific, even for a mechanical engineer, and very difficult to understand, but the first reference explains all the main issues related with this subject.
Specific comments:
In section 1, page 1, line 30-32, this statement should be improved
In subsection 2.1.1, page 2 line 65, authors say that letter m represents the embedding dimension, but in equation 1 there is no letter m, please correct the mistake.
In subsection 2.1.2, page 3 line 69-70, once more after equation (4) there is a reference to index m, but in this equation, there is no letter m. Please correct.
In section 4, page 5 line 120-121, improve that statement.
In subsection 5.2.1, page 5 line 146-147, explain how you get the number of sampling points. It should be 32x32? The quality of images presented at figure 3 should be improved.
In section 5.3, the results of figure 4 should be described with more detail.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
Dear Authors,
Congratulations, very interesting paper indeed. It will definitely be used by other scientists involved in this field of engineering knowledge. I think this paper will gain a little more value when you consider the following small remarks.
line 63: please add at least one or two additional literature entrys that will show the method from a slightly different angle, e.g. :
(1) -"End-To-End Convolutional Neural Network Model for Gear Fault Diagnosis Based on Sound Signals" Appl. Sci. 2018, 8, 1584; doi: 10.3390 / app8091584;
(2) - “Modified convolutional neural network with global average pooling for intelligent fault diagnosis of industrial gearbox” - Maintenance and Reliability Vol. 22, No. 1, 2020;
(3) – “A multivariate encoder information based convolutional neural network for intelligent fault diagnosis of planetary gearboxes” Knowledge-Based Systems 160 (2018) 237–250;
line 116: the transmission ratio formula of the planetary gearbox (2KH) should be:
ics = (Ns + No.) / Ns (**),
because, according to ISO Standard, transmission ratio is equal to the quotient of the driving gear (or arm) speed to the speed of the driven wheel or arm and hence the formula (**). For reduction gearboxes, the transmission ratio is always in ranges
ics <= -1 or ics => 1;
line 141: please enter the value of the wheel module, because the reader cannot assess how big the tooth-root crack lengths are in relation to the thickness of the tooth-root (can the wheels still work?);
line 154: in scientific paper one should not use the minute unit (ISO), i.e. rotating speed of the output shaft is equal to 800/60 rev/s, etc.
line 160: as above;
line 160: the m marking is reserved for the module.
There is half a page left at the end of the paper, so with the description of the gearbox (chapter 5.1) please insert the kinematical schema of the planetary gear. Not everyone reader has to be able to identify the transmission type according to the formula of transmission ratio.
Sincerely Reviewer
Author Response
Please see the attachment.
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.
