Review Reports - Application of Generalized Regression Neural Network and Gaussian Process Regression for Modelling Hybrid Micro-Electric Discharge Machining: A Comparative Study

Round 1

Reviewer 1 Report

The manuscript" Comparison of GRNN & GPR for predicting responses in hybrid micro-electric discharge machining process" submitted for review. Author's have done good effort. However I have following comments:

Title should be revised and should not contain any abbreviations.
latest literature review should be added
Originality and novelty from this research study should be highlighted in a separate paragraph.
Authors have conducted an experiment and needs to add schematics and picture of original testbed in the manuscript.
Results and comparisons of results needs to be presented in tabular format.
Results discussion needs to edited for more clarity.
Future recommendations and extension of this research work needs to be improved and should be edited as separate section apart from conclusion

Author Response

General Comment:

Response:

Thank you for reviewing the manuscript and considering the research topic important and finding the paper rich in content. All the suggestions are incorporated in the revised manuscript and presented below for your reference.

Comment #1:

Title should be revised and should not contain any abbreviations.

Response:

The title has been revised as per suggestion. Following is the revised title:

“Application of Generalized Regression Neural Network and Gaussian Process Regression for Modelling Hybrid Micro-Electric Discharge Machining: A Comparative Study”

Comment #2:

Latest literature review should be added.

Response:

Authors have cited more of recently published papers relevant to current research work. Now, out of 27 references, 19 references are now in the last 5 years. The following is year wise breakup:

2018-3; 2019-2; 2020-4; 2021-7; 2022-3.

Comment #3:

Originality and novelty from this research study should be highlighted in a separate paragraph.

Response:

A separate paragraph has been added at the end of the Introduction section to highlight the originality & novelty of this study.

In light of the discussed literature, it is clear that GRNN and GPR techniques could be applied for modelling complex hybrid machining. Such studies are rarely available in the literature and application of GRNN and GPR for modelling of vibration assisted micro-EDM has been not reported previously. Thus, this article explores the efficacy of two different non-parametric modelling techniques, viz. GRNN & GPR for predicting the different significant responses in a hybrid -EDM process, which has not been attempted earlier. Therefore, in this work, blind microholes of various depths were initially fabricated using μ-EDM on different materials by varying the energy input and vibration frequency. Post fabrication, three key process outputs were measured- hole drilling rate, rate of tool wear and center line average surface roughness of the bottom surface. These outputs are then modelled using GRNN and GPR, and a comparison is made in terms of mean absolute percentage error (MAPE), root mean square error (RMSE) and coefficient of determination (R²).

Comment #4:

Authors have conducted an experiment and needs to add schematics and picture of original testbed in the manuscript.

Response:

The schematic of the experimental procedure is already provided with Figure 1. A photograph of the machine on which the experiment is done has been added as Figure 2.

Comment #5:

Results and comparisons of results needs to be presented in tabular format.

Response:

Kindly note that all the experimental results are already reported in Table 2. With regard to the comparison of the two techniques, a graphical approach is much better to help the readers visualise the difference such as in Figures 9, 14 and 19.

Comment #6:

Results discussion needs to edited for more clarity.

Response:

Results are now been edited carefully for more clarity. The conclusion section is rewritten focusing outcomes of this research work.

Comment #7:

Future recommendations and extension of this research work needs to be improved and should be edited as separate section apart from conclusion.

Response:

Future recommendations and extensions of this research work have been suitably modified and also presented as a separate section.

Considering the proven efficacy of GPR in this research work, GPR models can be implemented in future, particularly for vibration-assisted micro-EDM, to pre-determine with reasonable accuracy the time and tool requirements for performing a particular drilling operation. Also, in future, using more training data, the model may be extended for other operations performed using hybrid micro-EDM such as micro-millings to develop a single holistic model to predict the output of all operations performed with vibration-assisted hybrid -EDM. The extension of this work could also be in the direction of combining the approach of Big Data along with GPR modelling by extracting any possible real-time continuous data from different machining processes.

Reviewer 2 Report

The paper deals with the Comparison of GRNN & GPR for predicting responses in
hybrid micro-electric discharge machining process.
The paper, in the opinion of the reviewer, isn't worth publishing
at Processes Journal, and consequently the reviewer rejects the paper for publishing.
While the authors have made considerable research effort,
the presentation of the paper and the results are considered unsatisfactory.

Comment 1
No Line numbering. This is difficult to the reviewer.

Comment 2
Abstract
ARD squared exponential kernel

The authors should replace

Automatic Relevance Determination (ADR) squared exponential kernel

Comment 3
Page 2
Byiringiro etal. [10] demonstrated
The authors should replace (insert a space)
Byiringiro et al. [10] demonstrated

Comment 4
Page 3
The authors should give more details for cutting parameters.
The authors should insert a Figure with a real foto of the machine with a specimen and a micro electrode.
The authors should insert a Figure with a typical specimen (foto) after the cutting.
The authors should give more details what aspect ratio mean.
The authors should explain what they did not use the Material Removal Rate (MRR) for the results instead the DR.

Comment 5
Page 3
The authors should explain why they used a constant electrode RPM of 1500 and feed rate of 0.2 mm/min,
instead they can used as the cutting parameters and their levels (for investigation).

Comment 6
Page 4
The authors should give more details for their statement "the distance of tool tip from the reference top
surface of workpiece pre- and post-machining using the machine’s Z-axis that has a least
count of 0.1 μm."
How did they measured this distance?

Comment 7
The numbering of the equations should be at the end of the line on the right side of the text.

Comment 8
Page 7
The pioneering work by Rasmussen and Williams [18] may be referred to
The authors should check if the number 18 is right.

Comment 9
Pages 6, 7
It is not so good to use the word "we".
The authors should rephrase.

Comment 10
Page 6
The process be
comes a bit simplistic if we assume the prior mean function as 0. Thus
The authors should format the paper according to the journal's instructions.

Comment 11
The authors must format the Ref. according to the journal's instructions.
References - Format
1. Author 1, A.B.; Author 2, C.D. Title of the article. Abbreviated Journal Name Year, Volume, page range.

Comment 12
Page 11
in Figure 12 also highlights this difference

The authors should replace (insert a .)

in Figure 12 also highlights this difference.

Comment 13
1) the statistical model seems to be very much dependent on the value of the smoothing parameter σ.
Is there any guidance on how to choose σ in practice? Probably a method that would produce the optimal σ subject
to minimizing the model's MAPE or RMSE would be very beneficial for the application of the method.
2) One impact of the dependence of the model on the smoothing parameter and in general a subtle point on the applicability
of the model seems to be the fact that even though the input values might match e.g. any of the input values in the author's experiment,
there is no guarantee that the output would be even close to the replicated experiments, see for example last line of Table 2.
Is there any explanation on why can the model be so off and how to go about it in similar situations in practice?
3) From a mathematical perspective the Gaussian process regression employed by the authors seems to be best suited for continuous data.
However the input data in Table 2 are of almost binary type (-1, 0, 1). The authors should comment and justify the applicability
of the model under this situation as well as provide some assessment of its prediction ability.

Author Response

Comment #1:

No Line numbering. This is difficult to the reviewer.

Response:

Apologies for the inconvenience. Line numbering has been added as per the reviewer’s suggestion.

Comment #2:

The authors should replace Automatic Relevance Determination (ADR) squared exponential kernel.

Response:

ARD has been replaced as Automatic Relevance Determination (ARD) squared exponential kernel.

Comment #3:

The authors should replace (insert a space) Byiringiro etal. [10] demonstrated.

Response:

Space inserted at “Byiringiro et al.”

Comment #4:

The authors should give more details for cutting parameters.

Response:

Details regarding the input machining parameters have been included in the revision.

The discharge energy (DE) in RC circuit-based micro-EDM typically depends on the voltage and capacitance values. DE is a product of voltage and capacitance square. The vibrational frequency has been identified to be playing an important role in improving micro-EDM performance [4]. However, work material and aspect ratio have not been researched well in the available literature. Therefore, to conduct experimentation, energy input, vibrational frequency and microhole’s aspect ratio were selected as the control factors; in addition, the work material is considered as a categorical factor. The aspect ratio is taken as the ratio of intended hole depth to the electrode diameter.

Comment #5:

The authors should insert a Figure with a real photo of the machine with a specimen and a micro electrode.

Response:

A photo of the machine tool used for experimentation has been added as Figure 2 and a photo of a specimen has been added as Figure 3.

Comment #6:

The authors should insert a Figure with a typical specimen (photo) after the cutting.

Response:

A photo of a specimen has been added as Figure 3.

Comment #7:

The authors should give more details what aspect ratio mean.

Response:

Details the on meaning of aspect ratio added.

The aspect ratio is taken as the ratio of intended hole depth to the electrode diameter.

Comment #8:

The authors should explain what they did not use the Material Removal Rate (MRR) for the results instead the DR.

Response:

Typically, in micro-EDM operation, the volume of material removed is very small. Thus, measuring such small material weight loss could lead to inaccurate data. Thus, instead of measuring the weight loss, the depth of the hole could be measured with better accuracy. Overall, the MRR calculation is highly prone to error and hence Drilling Rate approach was used for higher accuracy for getting reliable data.

Comment #9:

The authors should explain why they used a constant electrode RPM of 1500 and feed rate of 0.2 mm/min, instead they can used as the cutting parameters and their levels (for investigation).

Response:

Following explanation has been added for addressing this comment.

The constant electrode RPM and feed rate were selected based on multiple previous pilot experiments conducted for different materials. The machine was found to function best at the selected constant parameters.

Comment #10:

The authors should give more details for their statement "the distance of tool tip from the reference top surface of workpiece pre- and post-machining using the machine’s Z-axis that has a least count of 0.1 μm." How did they measured this distance?

Response:

Kindly note that the least count axis movement is 0.1 μm which is actually a machine manufacturer’s data showing the resolution of the axis. This is not a variable parameter and is always fixed for a particular machine.

Comment #11:

The numbering of the equations should be at the end of the line on the right side of the text.

Response:

All the numbering of equations adjusted to the right side of the text at the end of the line.

Comment#12:

The pioneering work by Rasmussen and Williams [18] may be referred to. The authors should check if the number 18 is right.

Response:

It has been corrected and cited accordingly in the correct sequence of reference. Note, now all reference numbering has been changed and authors used Endnote citation manager.

Comment #13:

It is not so good to use the word "we". The authors should rephrase.

Response:

The used of the word “we” have been avoided throughout the revised manuscript.

Comment #14:

The process becomes a bit simplistic if we assume the prior mean function as 0. Thus

The authors should format the paper according to the journal's instructions.

Response:

The comment is not clear to authors. However, authors have checked and made changes according to the journal’s instructions.

Comment #15:

The authors must format the Ref. according to the journal's instructions.

References - Format

Author 1, A.B.; Author 2, C.D. Title of the article. Abbreviated Journal Name Year, Volume, page range.

Response:

Authors have now used ENDNOTE citation manager and used MDPI ACS endnote style to manage the references. Now, all references are as per the journal’s format requirement only.

Comment #16:

The authors should replace (insert a .) in Figure 12 also highlights this difference.

Response:

Thanks for highlighting the mistake. A full stop is inserted as required.

Comment #17:

The statistical model seems to be very much dependent on the value of the smoothing parameter σ. Is there any guidance on how to choose σ in practice? Probably a method that would produce the optimal σ subject to minimizing the model's MAPE or RMSE would be very beneficial for the application of the method.

Response:

The selection of σ is subject to minimizing the prediction error. For the current paper, RMSE has been selected as the parameter to be minimized. Exploring different optimization techniques for σ would require further substantial work and at the moment is not an objective for this article.

Comment #18:

One impact of the dependence of the model on the smoothing parameter and in general a subtle point on the applicability of the model seems to be the fact that even though the input values might match e.g. any of the input values in the author's experiment, there is no guarantee that the output would be even close to the replicated experiments, see for example last line of Table 2. Is there any explanation on why can the model be so off and how to go about it in similar situations in practice?

Response:

In table 2, all the experiments have been performed at distinct values of input variables. Variability is usually always present in experimentations which leads to different outputs at the same value of input parameters, however, when it comes to modelling, a model will always provide the same result for a set of fixed input variables. Thus, extreme variation between experimental and predicted results is actually an indication towards the variability in experimentation and not in modelling.

While conducting the experiments, if the constant parameters are rigidly kept constant and environmental variation is reduced, then ideally there will not be variation in replications.

Comment #19:

From a mathematical perspective the Gaussian process regression employed by the authors seems to be best suited for continuous data. However, the input data in Table 2 are of almost binary type (-1, 0, 1). The authors should comment and justify the applicability of the model under this situation as well as provide some assessment of its prediction ability.

Response:

The input data is actually a combination of both continuous and discrete variables, and none of them is binary in nature. Vibration frequency, Energy and Aspect Ratio are continuous variables, while workpiece material is a discrete variable. In Table 2, these factors are presented in the coded form which is the typical nature of CCD design of experiments. The actual values of process parameters and their coding are already presented in Table 1. Kindly, take note of the same. Therefore, from the results, it can be concluded that the model works well even when the input parameters are a combination of continuous and discrete.

Round 2

Reviewer 1 Report

I appreciate the authors for their significance in the revised version. Especially in the conclusion and future directions of this research study.

Author Response

Thank you for reviewing the manuscript and considering the research topic important and finding the paper rich in content.

Reviewer 2 Report

Comment 1
Line 59
Unune et al. [10]reported
The authors should replace
Unune et al. [10] reported

Comment 2
Line 114
them. As they are brought
The authors should replace (delete the extra space)
them. As they are brought

Comment 3
Line 125
Figure 1. Schematic of Micro-feature
The authors should replace (delete the extra space)
Figure 1. Schematic of Micro-feature

Comment 4
Lines 127? (no numbering)
The authors should format the text according to the journal's instructions (text limits different from the rest paper)

Comment 5
Line 129
The discharge eznergy
The authors should check the paper for typographical and spelling errors.

Comment 6
Lines 318 and 355
It's not so good to start the sub-section at the bottom of the page without using text.

Author Response

Comment #1:

Line 59
Unune et al. [10]reported
The authors should replace
Unune et al. [10] reported

Response:

Thanks for highlighting the mistake. Correction made as per comment.

Comment #2:

Line 114
them. As they are brought
The authors should replace (delete the extra space)
them. As they are brought

Response:

Thanks for highlighting the mistake. Correction made as per comment.

Comment #3:

Line 125
Figure 1. Schematic of Micro-feature
The authors should replace (delete the extra space)
Figure 1. Schematic of Micro-feature

Response:

Thanks for highlighting the mistake. Correction made as per comment.

Comment #4:

Lines 127? (no numbering)
The authors should format the text according to the journal's instructions (text limits different from the rest paper)

Response:

Some lines were not numbered due to some formatting errors. The same has been corrected and the text has been formatted as per the journal’s instructions.

As per author instructions provided on the Processes website, there is no limit on the length of the paper.

Following are the key points on the instruction page:

“Processes has no restrictions on the length of manuscripts, provided that the text is concise and comprehensive. Full experimental details must be provided so that the results can be reproduced.”
“We do not have strict formatting requirements, but all manuscripts must contain the required sections: Author Information, Abstract, Keywords, Introduction, Materials & Methods, Results, Conclusions, Figures and Tables with Captions, Funding Information, Author Contributions, Conflict of Interest and other Ethics Statements.”.

All other formatting requirements are followed.

Comment #5:

Line 129
The discharge eznergy
The authors should check the paper for typographical and spelling errors.

Response:

Thanks for highlighting the error. The manuscript has been thoroughly proofread and corrections have been made for English and Grammar mistakes that were present in the previous version of the paper. All such corrections are highlighted.

Comment #6:

Lines 318 and 355
It's not so good to start the sub-section at the bottom of the page without using text.

Response:

Changes made as per suggestion.