Gross Calorific Value Estimation in Coal Using Multi-Model FTIR and Machine Learning Approach

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper developed and validated an efficient and reliable technique for estimating the gross calorific value of coal by utilizing mid-infrared Fourier transform infrared (FTIR) spectroscopy and multiple machine learning algorithms. This is an interesting and worth exploring issue, but I believe there are still some areas that need to be modified in this article.

Here are some of my suggested modifications.

1. The sample source of the paper is single, and the types of coal are limited (only covering sub-bituminous coal). It is suggested to supplement external sample tests across regions and coal grades, or clarify the applicable boundaries of the model in the discussion.
2. The study employs the average of "PLSR + RFR + XGB" as the multi-model fusion result, but does not explain the rationale for selecting this specific combination. It is recommended to supplement the analysis with comparative experiments of different model combinations and justify the rationality of the optimal fusion strategy.
3. The study achieves accurate prediction of GCV but fails to conduct interpretability analysis of the machine learning models, making it impossible to identify which functional groups contribute most significantly to GCV. It is recommended to supplement the analysis with feature importance ranking and correlation analysis between functional group content and GCV, and integrate the predictive model with coal chemistry mechanisms to enhance the scientific depth of the research.
4. The discussion mentions that the FTIR method is "more efficient and lower-cost" compared to the traditional bomb calorimetry method, but fails to provide specific quantitative comparative data. It is recommended to supplement the analysis with a comparative table that clearly demonstrates the advantages of the FTIR method across the three dimensions of "efficiency-cost-accuracy" to enhance persuasiveness.
5. The study lacks interpretability analysis for models such as XGB and RFR, making it impossible to clarify the contribution mechanisms of functional groups to GCV. It is recommended to supplement with SHAP value visualization analysis and generate feature dependence plots to validate the coal chemistry theory that "higher aromaticity and lower oxygen content lead to higher GCV."
6. The reference list is predominantly composed of literature published prior to 2023. It is recommended to incorporate highly-cited publications from the last two years for comparative analysis, thereby clearly delineating the innovative aspects of this study in feature engineering and model architecture.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

A brief summary

The paper presents the results of mid-infrared FTIR spectroscopic analysis of coal to identify fifty-six selective absorption bands (controlled input variables) sensitive to the content of organic functional groups in coal, combined with several machine learning methods to model the gross calorific value of coal samples from Johilla coal basin, India.

 

General concept comments

The manuscript fully complies with the journal's profile and areas of interest. This work is original in nature and makes a significant contribution to the advancement of modern scientific knowledge. The article is written at a high level: the presentation is clear, logical, and meets academic standards, as is the presentation of scientific results. The methodological basis of the study is carefully thought out, based on reliable data, and accompanied by a detailed and clear description of the applied methodology. The obtained results possess both theoretical and practical innovation. The text is well structured, logically constructed, and precisely corresponds to the stated topic, ensuring its easy comprehension. The literature review reflects key aspects of the consensus in the field and is progressive and relevant. Gaps in understanding the current state of the issue were not identified. The majority of cited sources published over the past five years confirm the scientific novelty of the work. The hypothesis is soundly substantiated, and the chosen research methods are adequate to the objectives and ensure the reproducibility of the results. Visual support (figures, tables, diagrams) is appropriately selected and effectively illustrates the key points of the work. The conclusions formulated follow logically from the presented data and arguments and will undoubtedly be of interest to the journal's readers, as they aim to advance scientific knowledge in this field. After minor revisions, I propose publishing this article.

 

Specific comments

Comment 1: Section "1. Introduction" should ideally conclude with a summary of the literature review, describing the purpose of the work and how it differs from previously obtained results described in the review.

Comment 2: How were the coal composition values ​​presented in Table 1 obtained (equipment and methods)? How was the error determined?

Comment 3: How can the results of this study be applied in practice? Was the economic feasibility of this method compared to the classical method (bomb calorimeter) assessed, taking into account the time required for analysis?

Comment 4: Is it possible to use this method to determine the gross calorific value of biomass?

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

In this manuscript, a multi-model method was proposed to estimate the GCV of coal using mid-infrared FTIR spectroscopy coupled with machine learning algorithms, including piecewise linear regression (PLR), partial least squares regression (PLSR), support vector regression (SVR), random forest regression (RFR), artificial neural networks (ANN), and extreme gradient boosting regression (XGB). The work is relatively interesting, and the structure and language are well prepared. Some comments below need to be addressed or considered to improve the technical depth of the manuscript:

(1) The purpose of this manuscript should be added in the introduction.

(2) In section 2.3-2.9, the widely known knowledge should be simplified, and it is recommended to add analysis flowcharts for each method and text research analysis flowcharts.

(3) The conclusion should describe the innovative results of the research findings. Currently, the conclusion only summarizes the methods and evaluates their effectiveness.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

accepted