Improving Hardenability Modeling: A Bayesian Optimization Approach to Tuning Hyperparameters for Neural Network Regression

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

Thank you for this very interesting work, summarizing the attempt of modelling the levels of hardness that can be achieved from a specific steel. However, a list of comments must be addressed so as to improve the coherence and the soundness of the technical work.

Abstract

The scope and the outputs of the present work are very well given in the abstract section. However, a reader cannot identify which process is used to harden the material or for which application, as well as if there is a predictability limit for the model. Please improve this paragraph by addressing this comment.

Introduction

Page 2, lines 49-55: The authors could consider also to talk about the effect of material processing in the generated temperature field and along with the material properties to the effect of temperature field on the hardness. Please find hereafter a work where the effect of process parameters has affected the temperature field and the resulting hardness. Consider to mention different processes such as conventional machining and non-conventional processes and comment on how the different process inputs may affect the resulting hardness and the attempt for modelling.

Page 2, lines 67-86. The authors mention requirements of ANN related to data availability, labelling, clustering etc. However, it is recommended to quantify these requirements either for their case or for different cases in literature as per below:

·        A. Papacharalampopoulos, K. Tzimanis, K. Sabatakakis, P. Stavropoulos, "Deep Quality Assessment of a Solar Reflector Based on Synthetic Data: Detecting Surficial Defects from Manufacturing and Use Phase", Sensing Technology and Data Interpretation in Machine Diagnosis and Systems Condition Monitoring, Sensors, Volume 20, Issue 19, 5481, (2020)

 

Materials and methods

It is recommended to compare the requirement lists that have been created in the introduction section with the actual training process of the model 1 by 1, since the complexity of machine learning models is hidden on the data manipulation and the preparation for training purposes. Apart from the aforementioned literature sources, the authors may consult and the following source for Regression Neural Network training. It would be also interesting to elaborate on the key tunable parameters of the modelling attempt.

·        Mohammad Bataineh, Timothy Marler, Neural network for regression problems with reduced training sets, Neural Networks, Volume 95, 2017, Pages 1-9, ISSN 0893-6080, https://doi.org/10.1016/j.neunet.2017.07.018.

 

Additionally, it is advisable to introduce under which method the improved hardness parameters are succeeded.

Results

Can the authors elaborate more on the generalizability of the model and under which technical assumptions has been built?

On top of that, it seems that the authors should also include a physics-based interpretation of the results since it could be interesting why the obtained data are reasonable?

Discussion

The authors may include expected functionalities of the model in the future as well as other contribution of the model to the prediction of key indicators in the hardness process.

 

Thank you

Author Response

Dear Reviewer,

Responses to the review have been included in the attached file.

Authors

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The research paper demonstrates the developed model  for a specific steel grade and shows the effectiveness of the model in predicting and optimizing heat treatment results.This paper has certain research value, but there are the following problems.

1.In the Section of Materials and Methods , Please provide the maximum, minimum, average, and standard deviation of the percentage of mass fraction of concentration of the seven basic alloying elements in the modeling data and testing data, respectively.

2.In the section of Discussion, the discussion section of this paper is too short, which needs to be compared with the research of other scholars.

3.Please compare and analyze 10 fold cross validation and 5 fold cross validation, analyze the research of other scholars on the two types of validation, and list the references of other scholars.

4.What language did the author use for programming? If possible, please provide the relevant code in the attachment of the paper.

Comments on the Quality of English Language

Minor editing of English language required.

Author Response

Dear Reviewer,

Responses to the review have been included in the attached file.

Authors

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

There are still some problems in the paper that need to be modified, such as R2 in Table 6, which should be R². Please ask the author to read the entire article and review and revise the paper.

Comments on the Quality of English Language

 Moderate editing of English language required.

Author Response

Dear Reviewer,

Thank you for your detailed comments and for pointing out errors and inaccuracies.
We have re-read the article very carefully and corrected the errors.
The linguistic correction has also been made and we hope that the current version of the article is acceptable.

Authors