Hybrid Method for Endpoint Prediction in a Basic Oxygen Furnace
Round 1
Reviewer 1 Report
For complex processes like, BOF, the reviewer agrees with author’s approach of using the principles based on science to the extent possible coupled with a data analytics approach.
However, the authors have not been able to bring out the novelty of their work. Authors have used mass and heat balance models which are widely available in literature. In BOF process parameters such as blowing strategy, idle time, etc, have large influence on the process performance. Ignoring these important factors, if not in theoretical frame work, but in the Machine Learning Techniques makes the model inadequate. Authors seems to have decoupled the mass balance and heat balance though it is not clear from the manuscript, the need for such a strategy. Authors have used Normalized Root Mean Square Error to claim the better performance of their model over theoretical models. However, there are not additional process insights which authors have been to come out through the hybrid strategy. Possibly, plotting the predicted data vs plant data can give better pictorial representation of the predictability of the model.
Considering all these points, reviewer recommends for a major revision of the manuscript so that the novelty factor can be brought out clearly. Further, additional process parameters which plant records and has significant influence on the process performance needs to be considered in the machine learning models.
Specific comments on the manuscript is as follows
1) Line 68 – Most theoretical models used ….. – Reviewer urges the authors to reconsider statement. Recently, there are published papers by Rahul Sarkar et al , Sngidha Ghosh et al which considers the thermodynamics, emulsion phenomena as well as operating parameters such as lance practice etc for simulating BOF Process. Authors are urged to look these references.
2) Line 125 - In general, theoretical models by using thermodynamic principles are beneficial to 125 understand the kinetic reactions taking place in BOF steelmaking – why the word kinetic has been used in this sentence? It may sound contradiction to the word thermodynamics principles.
3) Why dissolved oxygen in the liquid steel not considered in the mass balance model?
4) All the lime added to the process may not get into the slag (solution). Therefore, calculating the basicity based on the input CaO may not be true representation of slag basicity.
5) CO/CO2 has a large effect on the Oxygen and Carbon balance. In the plant how is the CO/CO2 ratio is measured? Depending on where the measurement has been carried out, the measured value can be quite different from the actual value at the mouth of the converter, due to ingress of air at the mouth of the converter.
6) Independent mass and heat balance model has been considered in the paper, however, ideally these models can be integrated together
7) It is not clear what input has been used in the Neural network. Is it the computed values from the theoretical model? What is the number inputs of considered for the input? Can the number of fitting parameters (Reviewer presumes, 2 per node) in the neural network be more than the number of input variables to the ANN?
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The paper is interesting, topical and fits into the theme of the magazine.
The issue addressed is one of the possible methods of streamlining the costs of producing ferrous materials. The models made can serve as a basis for computer-assisted steering of steelmaking.
The presentation does not rise to the level corresponding to the publication, only after a few small corrections.
45 bibliographic references are presented, but not all of them can be found in the text, and in certain situations, a paragraph refers to 25 references!
Does the notation of relations (equations) must be revised because after relations (1) (12) it is continued with (7) followed by (84)?
In the analysis of similarity with the TURNITIN software, the result was 15%, but the usual names, chemical formulas, percentage values, which do not represent plagiarism, are also marked.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
This paper deals with the hybrid algorithm based on heat and mass balance and neural network modelling monitoring and prediction of endpoints in Basic Oxygen Furnace.
Abstract structure: the content and structure of the abstract is convincingly presented.
Introduction of the article is clear to me.
Theory and Methodology (calculations are convincingly described and help to understand the research and the results).
Results and Conclusions corresponds to the conceptual architecture of the paper.
Pictures: adequate number, presentation clearly and very good quality.
Adequate list of the literature cited in the references.
In my opinion, revisions are not required - recommendation: Accept.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Authors have addressed all review comments
