Applying a Neural Network to Predict Surface Roughness and Machining Accuracy in the Milling of SUS304

Round 1

Reviewer 1 Report

Though the authors have tried to bring out the importance of using soft computing in predicting the roughness parameters, but the manuscript is too simple to be considered as a technical article. So, at present I am rejecting the article. The scientific soundness of the article is poor and it doesnot create a feeling of interest for the reader. 

The manuscript is too simple and the authors have only used the ANN as a
tool.
The scope of the paper needs to be broaden.
i. They should bring out the importance of using ANN and predicting the
roughness. It would be better if they can relate to the which roughness
parameter would be more influential. This can be predicted using ANN.
ii. They should relate their findings with tribological studies.

Author Response

1. Thank you for your opinion. The experiment had found that the feed per tooth is the most powerful parameter affecting the roughness. The description of this result is also revised in Chapter 3, 4.

2. Thank you for your advice. We will add the discussion of tribology in our future research. We will continue to improve research methods and content to achieve better papers.

Reviewer 2 Report

Abstract : Compare the prediction accuracy and analysis efficiency of the three models on different training data. Kindly check the language

Accordingly, we input the same machining signals into different models for analysis to compare their prediction accuracy. What is the outcome of this statement? 

On what basis cutting speed was selected for the experiment?

Ra seems to be very high for the cases above 20. Whether authors used same tool or different tool for each experiment?

How tool wear is considered for 120 cases?

Fig 10, Y axis doesn't contain unit

Discussion part is very weak.

No validation of results. Also, what is the outcome of the study?

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This paper uses the deep learning–based artificial neural networks to predict the surface roughness and machining accuracy. The authors used three models, namely CNN, DNN, and LSTM. The performances of these models were also analyzed and compared. This manuscript needs a minor revision in my opinion.

1. Lines 24 and 25. The full names of CNN, DNN and LSTM needs to be given.

2. Line 191 Table 4, according to this Table, we may get 81 groups of data. However, the authors claimed that 162 experimental data were obtained. Would you mind explaining it?

 

3. Line 285, “The number of nodes in the LSTM layer was 32”. Could you please clarify how you got the number of the nodes?

  3. At the end of Part 1, please add the motivation of this manuscript as well as summarize the shortcomings of previous studies on this topic. 4. At the end of Part 1, please provide a brief introduction to the paper’s arrangement for each chapter. 5. Line 161, please add the chemical compositions of SUS304 stainless steel. 6. Figure 5, please change “時間“ into “time”. 7. Part 5, please give more quantitative conclusions if possible.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

Tsai et al investigated the prediction capability and reliability of three different deep learning-based artificial neural networks (ANN) on surface roughness of milled SUS304 stainless steel hexahedrons. The topic is rather interesting. The experimental processes are described clearly. They also introduce the ANN models well. The obtained results might be of use for people working in the related fields. The manuscript is written in a proper way. The text is in the scope of this Journal. Thus, I’d like to suggest acceptance of this manuscript for publication in Electronics in the present form.

Author Response

Thank you for your affirmation and encouragement.

Round 2

Reviewer 1 Report

In a scientific manuscript "chapters" are not written. At present, there has not been much information on the tribology of the work. Sorry to point out that mere presenting the Soft computing methodologies is no longer remains an interest to the researchers working in the area of surface engineering.

The manuscript is very simple ..only mere soft computing has been done. This is not so much critical . the area of the research is good but not much surface characterization has been indicated. The papers such kind should have good surface characterizations however, the authors have deinied to use them as I can see in their response. So, I do not feel that just comparing two soft computing techniques seems to be a great work that can be published in reputed publications like MDPI.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Nil

Author Response

Thank you for review our manuscript.

Round 3

Reviewer 1 Report

Thank you for your detailed explanation. Hope to see the follow up paper soon.