Influence Factor Analysis and Prediction Model of End-Point Carbon Content Based on Artificial Neural Network in Electric Arc Furnace Steelmaking Process

Round 1

Reviewer 1 Report

1- Illustrate the application of the proposed model.

2- Please define all variables after their first appearance in the article.

3- Please refer all relations to valid references after the first appearance in the article.

4- What is assumption and limitation of the proposed materials and model?

5- What can we learn from the presented manuscript? What is the difference of the current work against other published articles?

6- The results and discussion are very briefly organized. An extended result and discussion are needed.

Author Response

Dear referee

I will give you a precise answer to your question

1- Illustrate the application of the proposed model.

1-Thank you for your comments, I have explained the application of the model in Chapter 4. Due to the confidentiality of steel products, the specific steel type information is inconvenient to show, so the application effect of the model can only be reflected by the relative value.

 

2- Please define all variables after their first appearance in the article.

2- Thank you for your comments, The variables in the formula have been defined in the article, and the input variables have been defined in Table 1

 

3- Please refer all relations to valid references after the first appearance in the article.

3- Thank you for your comments, I have changed it to a valid references in the article.

 

4- What is assumption and limitation of the proposed materials and model?

4- Thank you for your comments, I didn't understand the question you raised, so I'll answer it the way I understand it. The hypothesis of the model is mainly reflected in the selection of activation function and hidden layer. The limitation of the model is mainly reflected in that it only predicts carbon, but in the actual smelting process, carbon element is only used as one of the evaluation criteria. Later, the model can be applied to various end-point control.

 

5- What can we learn from the presented manuscript? What is the difference of the current work against other published articles?

5- Thank you for your comments, In my opinion, this manuscript relates the end-point composition control of EAF steelmaking to initial conditions, explores the quantitative relationship between input variables and output variables, and develops a model for predicting end-point carbon composition, which can optimize the field smelting operation and improve production efficiency. At the same time, the end-point control technology for converter steelmaking is relatively mature, but it is not common to study the end-point control of EAF, and the model has certain self-learning ability, which can adapt to different conditions of different steel plants and has a wide range of adaptability.

 

6- The results and discussion are very briefly organized. An extended result and discussion are needed.

6- Thank you for your comments, I have made corresponding additions in the discussion and conclusion, please check

Author Response File: Author Response.docx

Reviewer 2 Report

Title: Influence factor analysis and prediction model of end-point carbon content based on artificial neural network in electric arc furnace steelmaking process

·       The abstract can mention the structure of the model to indicate how the carbon composition is predicted and how mean square error is improved.     

·       Reference should be accurately cited, for example Zhu er al.[18].

·       The English can be accurate, for example, ‘using simply using…’.

·       Introduction:  Please describe the mean square error and the factors governing the mean square error.      

·       Please make clear why addition of three extra variables changes the prediction accuracy, Figures 4 and 6 in section 3.4.      

·       Figure 10 can be replotted to show the difference in the predicted and actual end point carbon content with the heat number. 

·       The influence of carbon content of the steel on the accuracy of prediction should be mentioned (not as scope for further work) for improving the quality of the manuscript.      

 

Author Response

Dear referee

I will give you a precise answer to your question

 

1.The abstract can mention the structure of the model to indicate how the carbon composition is predicted and how mean square error is improved.

1. Thank you for your comments and I have revised them in the abstract

 

2. Reference should be accurately cited, for example Zhu er al.[18].
3. Thank you for your comments and all references in this paper are adjusted

 

3. The English can be accurate, for example, ‘using simply using…’.
4. Thank you for your comments and I have made changes in the article

 

4. Please describe the mean square error and the factors governing the mean square error.
5. Thank you for your comments and I have made changes in the Introduction

 

5. Please make clear why addition of three extra variables changes the prediction accuracy, Figures 4 and 6 in section 3.4
6. Thank you for your comments. As shown in picture figures 4 and 6, In Figure 4, R is 0.7632, and in Figure 8, R is 0.8274. Because this model is a regression model, the larger R is, the stronger the correlation between the predicted value and the real value, and the higher the accuracy of the model. This is the comparison of two experimental data under the same structure of this model. At the same time, I added the analysis from the principle in Section 3.4, which can more verify that dividing oxygen into phases is conducive to improving the accuracy of the model

 

6. Figure 10 can be replotted to show the difference in the predicted and actual end point carbon content with the heat number.
7. Thank you for your comments. In Figure 10, the left side shows the results when the model is run automatically, and the right side shows the results when the model is run manually. The orange part indicates that the predicted result is greater than the actual result, and the blue part indicates that the actual result is greater than the predicted result

 

7. The influence of carbon content of the steel on the accuracy of prediction should be mentioned (not as scope for further work) for improving the quality of the manuscript.
8. Thank you for your comments and I have added in the first paragraph of the Introduction.

  

Author Response File: Author Response.docx

Reviewer 3 Report

 

- English language of the manuscript is acceptable in general. However, it would be much better to improve. Please avoid unnecessary long sentences. Also, some grammatical and typos mistakes can be observed. For example Zhu er al.[18], CO partial pressure (pa), lg([%C][%O])

- Please respond to the following questions and make necessary revisions.

* What advantage does the proposed ANN model offer as compared to existing design codes (if there are any), equations, and models?

* Does the proposed ANN model have generalization capability i.e.is the robustness of the ANN model tested by sensitivity analysis? 

* Is it stated that the proposed ANN models will be valid within the ranges of variables used for ANN training? No. It should be stated in a single sentence that ANN models will be valid within the ranges of variables

- Acronyms, should all be defined at their first occurrence in the manuscript.

- All parameters used in the equations should be described. It is suggested to add a section for the acronyms and parameters at the end of the manuscript.

- Flowcharts of tests and characterizations should be provided along with sample coding.

- The novelty of your work should be presented better at the end of the introduction part. This should be presented with more details.

- References 21, 22, and 23 should be referenced in the text.

- The two axes X and Y in Figure 3 are not readable and should be corrected.

- In line 253, the authors have stated in the text that: ”The main decarburization reactions are shown as formula (3)-(6)”; While the number of reference formulas is (4) to (7).

- In line 262, the authors have stated in the text that: ”The carbon content in the bath can be expressed as formula (7)-(8)”; While the number of reference formulas is (8) and (9).

- Reference writing should be the same and according to the format of the journal

- Literature review is not sufficient and authors must review and cite more papers in the field of ANN and especially newly published ones. Doing this, reviewing the following refs could be helpful:

[a] Neural Computing and Applications 25(7), 2014, 1993-1999.

[b] Neural Network World, 23(2), 2013, 117.

Author Response

Dear referee

I will give you a precise answer to your question

1. English language of the manuscript is acceptable in general. However, it would be much better to improve. Please avoid unnecessary long sentences. Also, some grammatical and typos mistakes can be observed. For example Zhu er al.[18], CO partial pressure (pa), lg([%C][%O])
2. Thank you for your comments, I have revised the article accordingly

 

2. What advantage does the proposed ANN model offer as compared to existing design codes (if there are any), equations, and models?
3. Thank you for your comments, In this paper, the proposed ANN model has certain representativeness, can be seen from the results in the regulation of input variables within the scope of the model can achieve the threshold value automatic adjustment, at the same time, the model can use on the other end mills composition control, have a certain extension, combined with the actual production status of iron and steel, models can choose to manually input and automatic input two modes, Have certain self-learning ability

 

3. Does the proposed ANN model have generalization capability i.e.is the robustness of the ANN model tested by sensitivity analysis? 
4. Thank you for your comments, In reality, the most common method is to evaluate the generalization ability of the learning method through the test error. However, this evaluation depends on the test data set. This paper verifies that the model has certain generalization ability through data. The next research direction will compare the advantages and disadvantages of different learning methods by comparing their generalization error upper bounds. At present, sensitivity analysis is not considered, because raw materials in the steel plant will not change greatly at the beginning, only need to improve the detection means, and can get accurate input conditions, and the model also has two ways of automatic and manual, when the input conditions cannot be read, can be filled in by the field operator

 

4. Is it stated that the proposed ANN models will be valid within the ranges of variables used for ANN training? No. It should be stated in a single sentence that ANN models will be valid within the ranges of variables.
5. Thank you for your comments, I have added your proposal to Section 3.2

 

5. Acronyms, should all be defined at their first occurrence in the manuscript.
6. Thank you for your comments, I have revised the article accordingly

 

6. All parameters used in the equations should be described. It is suggested to add a section for the acronyms and parameters at the end of the manuscript.
7. Thank you for your comments, All parameters have been stated below the formula, and the units are specified. Because of the length of this article, it is not intended to list the parameters separately in the last section

 

7. Flowcharts of tests and characterizations should be provided along with sample coding.
8. Thank you for your comments, I have added this in Article 3.2

 

8. The novelty of your work should be presented better at the end of the introduction part. This should be presented with more details.
9. Thank you for your comments, I have revised it in the introduction

 

9. References 21, 22, and 23 should be referenced in the text.
10. Thank you for your comments, I have revised the article accordingly

 

10, The two axes X and Y in Figure 3 are not readable and should be corrected.

10. Thank you for your comments, I have revised the article accordingly

 

11. In line 253, the authors have stated in the text that: ”The main decarburization reactions are shown as formula (3)-(6)”; While the number of reference formulas is (4) to (7).

 In line 262, the authors have stated in the text that: ”The carbon content in the bath can be expressed as formula (7)-(8)”; While the number of reference formulas is (8) and (9).

11. Thank you for your comments, I have revised the article accordingly

 

12. Reference writing should be the same and according to the format of the journal
13. Thank you for your comments, I have revised the article accordingly

 

13. - Literature review is not sufficient and authors must review and cite more papers in the field of ANN and especially newly published ones. Doing this, reviewing the following refs could be helpful:

[a] Neural Computing and Applications 25(7), 2014, 1993-1999.

[b] Neural Network World, 23(2), 2013, 117.

13. Thank you for your comments, I have revised the article accordingly

 

 

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

I checked the revised manuscript and have not been satisfied with the responses. The responses are too short and the authors have not taken care of the given comments carefully. I should reject the paper and recommend strong modifications based on the previous review comments as well as new comments as below: 

- Authors should be careful in writing the units, for example, the unit of pressure is Pa and not pa (line 295); Also, the logarithm should be written with Log and not lg (equation 11)

- It is necessary to follow the writing tips; for example:

“and the activation function was logsig+logsig The neural network model...”

It should be modified as follows; 

“and the activation function was logsig+logsig. The neural network model...”

- The authors have mentioned in the revised manuscript that:

“When carbon content < 0.77%, the intensity increases with the increase of carbon content, and when the carbon content > 1.0%, the intensity decreases.”

What does intensity mean? Is it hardness or strength?

- Avoid abbreviations in the ABSTRACT. In addition, please mention your findings. This part should be written specifically not general.

- Abbreviations should all be defined at their first occurrence in the manuscript; mean squared error (MSE)

- Referencing is not accepted as a group and each reference must be checked and referred to separately. for example, [4-8] or [9-11]

-The authors have mentioned in the revised text that:

“From the whole process of EAF steelmaking, 13 factors affecting the carbon content at the end of EAF were selected: scrap weight, hot metal weight, hot metal [C] content, hot metal [Si] content, hot metal [Mn] content, hot metal [P] content, hot metal [S] content, hot metal temperature, smelting time, carbon weight, lime weight, electric energy, and oxygen volume.”

While in Table 1, 16 input data are defined. But in Table 2, 13 process variables are defined. Authors should provide a complete explanation of how to use oxygen first, oxygen second, and oxygen last as end-point variables.

-  “Figure 4. Model flow chart” is unclear and full explanations should be provided in the text and subtitles

-The authors have stated in the text that:

“ In the improved ANN model, the model adds 3 input variables, which are the oxygen consumption volume of 0-5 minutes, 5-30 minutes, and more than 30 minutes.”

These details should be given in the abstract and conclusion

- The following text should be moved from the conclusion to the results and discussion or deleted:

“In future work, we will try to connect computer technology and automation equipment directly to the electric arc furnace control system to achieve one-button steelmaking; Steelworks should also adapt to the era of intelligence, improve the level of factory automation, to provide a good foundation for the model. At the same time, we try to fit different algorithms, expand the training set, increase the process variables, increase the number of bank layers, and accurately control the threshold to improve the prediction accuracy of the model. It provides a new direction for the intelligent transformation of the steel industry.”

The literature review is not thorough enough. It is crucial to stress the novelty of this study. Therefore I strongly recommend expanding the literature review. The paper has some new original investigation and deserves to emphasize its originality. 

Author Response

Dear referee

I will give you a precise answer to your question

1.Authors should be careful in writing the units, for example, the unit of pressure is Pa and not pa (line 295); Also, the logarithm should be written with Log and not lg (equation 11)

1.Thank you for your comments and I have revised them.

 

2. - It is necessary to follow the writing tips; for example:

“and the activation function was logsig+logsig The neural network model...”

It should be modified as follows;

“and the activation function was logsig+logsig. The neural network model...”

2. Thank you for your comments and I have revised them.

 

3. The authors have mentioned in the revised manuscript that:

“When carbon content < 0.77%, the intensity increases with the increase of carbon content, and when the carbon content > 1.0%, the intensity decreases.”

What does intensity mean? Is it hardness or strength?

3. Thank you for your comments and I have revised them

 

4. Avoid abbreviations in the ABSTRACT. In addition, please mention your findings. This part should be written specifically not general.
5. Thank you for your comments and I have added to that in the abstract.

 

5. Abbreviations should all be defined at their first occurrence in the manuscript; mean squared error (MSE)
6. Thank you for your comments and I have added to that in the abstract.

 

6. Referencing is not accepted as a group and each reference must be checked and referred to separately. for example, [4-8] or [9-11]
7. Thanks for your comments. This is the article I found in Coatings website, and some parts are not quoted separately, please check
8. The authors have mentioned in the revised text that:

“From the whole process of EAF steelmaking, 13 factors affecting the carbon content at the end of EAF were selected: scrap weight, hot metal weight, hot metal [C] content, hot metal [Si] content, hot metal [Mn] content, hot metal [P] content, hot metal [S] content, hot metal temperature, smelting time, carbon weight, lime weight, electric energy, and oxygen volume.”

While in Table 1, 16 input data are defined. But in Table 2, 13 process variables are defined. Authors should provide a complete explanation of how to use oxygen first, oxygen second, and oxygen last as end-point variables.

7. Thank you for your comments, Table I is the data statistics of all variables. There are only 13 variables in Table II as you mentioned, because I only selected the total oxygen as the most important condition for PCA, so there are no other three oxygen quantities. In Section 3 and 3, it has been explained that oxygen is processed in three stages and these three variables are replaced by the total oxygen quantity.

 

8. “Figure 4. Model flow chart” is unclear and full explanations should be provided in the text and subtitles
9. Thank you for your comments and I have revised them

 

9. The authors have stated in the text that:

“ In the improved ANN model, the model adds 3 input variables, which are the oxygen consumption volume of 0-5 minutes, 5-30 minutes, and more than 30 minutes.”

These details should be given in the abstract and conclusion.

9. Thank you for your comments and I have revised them

 

10.The following text should be moved from the conclusion to the results and discussion or deleted:

“In future work, we will try to connect computer technology and automation equipment directly to the electric arc furnace control system to achieve one-button steelmaking; Steelworks should also adapt to the era of intelligence, improve the level of factory automation, to provide a good foundation for the model. At the same time, we try to fit different algorithms, expand the training set, increase the process variables, increase the number of bank layers, and accurately control the threshold to improve the prediction accuracy of the model. It provides a new direction for the intelligent transformation of the steel industry.”

10. Thank you for your comments and I have revised them

 

11.The literature review is not thorough enough. It is crucial to stress the novelty of this study. Therefore I strongly recommend expanding the literature review. The paper has some new original investigation and deserves to emphasize its originality.

11. Thank you for your comments and I have revised them

 

Author Response File: Author Response.docx