Optimisation of Operator Support Systems through Artificial Intelligence for the Cast Steel Industry: A Case for Optimisation of the Oxygen Blowing Process Based on Machine Learning Algorithms

Round 1

Reviewer 1 Report

The proposed manuscript depicts the creating agent of blowing O₂ in EAF process based on RL model. Followings are some suggestions to make this manuscript clearer.

 

1. The sub-title “2.1 Use Case” is confusing.
2. There are some many AI methods. Why did the authors choose RL model as implementation model for creating oxygen blowing agent? It is not clear in the proposed manuscript.
3. Table Appendix A “blowing O2” should be “blowing O₂”. “2” is subscript. Table Appendix A should be revised because there are a lot of empty sign “-“. And as mentioned in Page 4 Line 121-122, the temperature raise is due to blowing O₂, why did the authors depict two temperatures before blowing O₂ rather than temperature, blowing O₂, temperature?
4. M3 subscript error can be seen in the manuscript.
5. Please re-consider the necessity of Appendix B.
6. Description of BFI is missed before Line 173. Abbreviation of MDP should be added in Line 201.
7. Why cannot directly use “a digital twin model developed by BFI” but to use them as training of the RL agent?
8. There is only one sub-title “2.3.1. RL Control Method Selection” and no more sub-title 2.3.2, therefore, it seems no need to use this 2.3.1 sub-title.
9. As depicted in Figure 12, can the system operate directly from 1 to 3 without 2. If yes, why? If no, why?
10. Why there is no conclusion in the proposed manuscript? It seems that the original discussion should be integrated as results and discussion. And add another section as Conclusion.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The research topic of this manuscript is meaningful, but the writing of the paper does not reach the level of an academic paper.

(1) Some sentences in this manuscript are redundant, and cutting to the topic should be straightforward. For example, mentioning AlphaGo Zero (lines 33-43) is unnecessary; lines 75-78 can be deleted; the data in Table 1 has too many digits after the decimal point, etc. It is recommended that the entire manuscript be shortened by 10-20%.
(2) This manuscript has no conclusions. A conclusion is an integral part of a research paper.
(3) Judging from the structure of the manuscript, the results and discussions account for less than 3/29 of its length, which is not normal. More discussion and analysis of the results are needed to demonstrate their accuracy and inadequacies.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper proposes an application of reinforcement learning in the metallurgical sector. In particular, a reinforcement learning-based agent is proposed, which is the core of an operator support system applied for improving the efficiency of the oxygen blowing process of a real cast steel foundry.

The topic of the paper is interesting and well aligned to the aims and scope of the journal.

The paper is overall well written and nice to read, as it depicts a relevant industrial application. The lements of novelty mainly reside in such application, as the exploited approaches and paradigms are well known and consolidated. Nonetheless, the paper has value, as it describes a practical field implementation of reinforcement learning within a production process where deployment of Artificial Intelligence-based approaches is generally difficult, as the IT environment is often not favorable. However, some amendments and deeper insights are needed to make the paper suitable to publication.

The introduction provides a good overview of the state of the art related to the application of AI in the steel sector, and in particular in the EAF process. However, as the proposed application is introduced as the development of an intelligent “agent”, an overview should be provided also concerning application of agent-based approaches in the same context and, in particular, in the EAF-based steelmaking route. For instance, the following works could be mentioned:

• Marchiori, et al. Integrated Dynamic Energy Management for Steel Production, (2017) Energy Procedia, 105, pp. 2772-2777. https://doi.org/10.1016/j.egypro.2017.03.597
• Aksyonov, K., et al. Analysis of the electric arc furnace workshop logistic processes using multiagent simulation (2018) Advances in Intelligent Systems and Computing, 678, pp. 390-397. https://doi.org/10.1007/978-3-319-67934-1_35
• Zarandi, M.H.F., Ahmadpour, P. Fuzzy agent-based expert system for steel making process, (2009) Expert Systems with Applications, 36 (5), pp. 9539-9547. https://doi.org/10.1016/j.eswa.2008.10.084

A final paragraph should be added to the introductory section, which provides an overview of the organization of the paper and the main contents of the different sections.

Section 2 provide a comprehensive description of the considered case study. However, more details should be provided on the exploited agent-based architecture.

Section 3 is overall clear and well readable.

In Section 4 overall results are nicely presented. However, such section lacks of an extensive discussion on the advantages with respct to the traditional operating practice, which would help in emphasizing the practical value of the developed work. The value is present, but it must be better and more extensively highlighted.

Section 5 actually provides some concluding remarks, therefore it needs to be renamed “conclusions”.

As a minor formal remark, the readability of the paper would benefit from the addition of a list of acronyms and abbreviations.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors have made modifications according to the comments. The manuscript is significantly improved.

Reviewer 3 Report

The Authors thoroughly revised the paper according to the suggestions provided by the Reviewers. The main weaknesses of the previous version of the paper have been overcome. Now the paper is suitable to publication.