Next Article in Journal
Design and Development of an Electronic Interface for Acquiring Signals from a Piezoelectric Sensor for Ultrasound Imaging Applications
Previous Article in Journal
BREAST-CAD: A Computer-Aided Diagnosis System for Breast Cancer Detection Using Machine Learning
 
 
Article
Peer-Review Record

Sustainable Mineral Processing Technologies Using Hybrid Intelligent Algorithms

Technologies 2025, 13(7), 269; https://doi.org/10.3390/technologies13070269
by Olga Shiryayeva *, Batyrbek Suleimenov and Yelena Kulakova
Reviewer 1:
Reviewer 2:
Reviewer 3: Anonymous
Technologies 2025, 13(7), 269; https://doi.org/10.3390/technologies13070269
Submission received: 14 May 2025 / Revised: 8 June 2025 / Accepted: 19 June 2025 / Published: 24 June 2025

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors present a hybrid intelligent control system for the beneficiation of fine chromite in a jigging machine, which is a crucial process for enabling sustainable mineral processing. The document needs to be fully reconstructed and major revisions should be addressed.

Comments:

  1. The introduction doesn't effectively introduce the main topic of the essay. It fails to engage the reader, provide necessary context, or establish a clear paper statement, in relation to hybrid models developed for data training.
  2. The paper structure should be fully revised, by describing the methodology followed, the parameters studied, data training mechanism and validation methodology.
  3. There is a big lack of information related to the simulation/modelling results,  like lines 263-265
  4. It is not clear how the model was trained (data introduced, system boundaries) etc.
  5. The quality of the English language should be improved.

 

Author Response

Comments 1: The introduction doesn't effectively introduce the main topic of the essay. It fails to engage the reader, provide necessary context, or establish a clear paper statement, in relation to hybrid models developed for data training.

Response 1: Thank you for pointing this out. We agree with this comment. Therefore, we have, accordingly, revised the Introduction (Page #3, Lines #96-100) to better explain the motivation and clarify the relevance of hybrid intelligent models. We have added a paragraph describing the specific limitations of traditional control in chromite ore beneficiation (Page #3, Lines #96-99). Table 1 (Page #3, Line #100) has also been added to compare traditional and hybrid control systems.

Comments 2: The paper structure should be fully revised, by describing the methodology followed, the parameters studied, data training mechanism, and validation methodology.

Response 2: Agree. We have, accordingly, revised Sections 3 and 4 to emphasize this point. These sections now have clear subheadings (3.1–3.3 and 4.1–4.4) and a clearer and more logical structure. The paragraphs are shorter and more focused, which improves readability. Section 3.3 (Page #9, Lines #334–340) now provides a comprehensive description of the data training mechanism.  The parameters studied - bed level (L), pulsation frequency (n) and ore composition (Cr) - are now clearly introduced in Sections 3.1-3.2, along with their respective roles as inputs, disturbances or optimization targets. Section 4 has been revised to clarify the choice of weighting factors (α = 0.6, β = 0.4) and a brief sensitivity analysis has been added to support this (Page #10, Lines #376–378). In Section 4, we have added information that the Interior-Point method was selected based on a comparative analysis with the Genetic Algorithm and Gradient Descent (Page #12, Lines #425–434). The comparative analysis showed that the fmincon (Interior-Point) method provides the fastest decrease in the value of the objective function. 

Comments 3: There is a big lack of information related to the simulation/modelling results, like lines 263-265.

Response 3: We have, accordingly, revised Sections 3 and 4 to emphasize this point. Additional commentary related to the simulation results has been included (Page #8, Lines #278–279, Page #12, Lines #425-432). We have accordingly checked all the figures to emphasize this point. Figure captions 3, 4, 8 and 9 have been revised to be more self-explanatory (Page #6, Line #222), (Page #8, Line #273), (Page #13, Line #446). An analysis of all the simulation results, which are depicted in the figures, is provided.

Comments 4: It is not clear how the model was trained (data introduced, system boundaries) etc.

Response 4: Agree. We have, accordingly, revised Section 3.3 (Pages #9-10, Lines #334-341) to emphasize this point. Specifically, we have added the number of data points used (1500 total: 70% for training, 15% for validation, and 15% for testing), the use of both real and synthetic data, and their respective sources. Measures to prevent overfitting, such as early stopping and L2 regularization, are also described in the revised text. The system boundaries are defined by the physical limitations of the jigging machine (see Page #6, Lines 223-230: bed level range: ; pulsation frequency range:  min⁻¹), and normalization was applied accordingly during both training and inference.

4. Response to Comments on the Quality of English Language

Point 1: The quality of the English language should be improved.

Response 1: The manuscript has been revised to improve the clarity, grammar, and technical accuracy of the English language. The revised version has also been proofread by a native English speaker with expertise in scientific writing.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The intelligent hybrid system presented here could be a promising solution for sustainable real-time control of gravity separation processes. However, below are a set of suggestions for improvement and/or questions that could have been answered throughout the article, or if this were not possible, referred to future research work. Since the system is hybrid and is simulated and not implemented in practical terms, the question arises as to whether in practice it will actually behave as in the simulation. A hybrid system of this type will require advanced automation/computational resources to perform the control, which will increase the cost. These costs are not mentioned or quantified anywhere in the article. Furthermore, will a system of this type lead to energy optimization (cost reduction) or to an increase in costs (by requiring more demanding computational resources and a variation in the acceleration and deceleration of the system's motors)? How much could a system of this type cost a company? Is the investment recoverable? In how many years?

Author Response

Comments 1: The intelligent hybrid system presented here could be a promising solution for sustainable real-time control of gravity separation processes. However, below are a set of suggestions for improvement and/or questions that could have been answered throughout the article, or if this were not possible, referred to future research work. Since the system is hybrid and is simulated and not implemented in practical terms, the question arises as to whether in practice it will actually behave as in the simulation.

Response 1: Thank you for pointing this out. We agree with this comment. Although the system is currently implemented in a simulated environment, its architecture is aligned with industrial control systems. As discussed in Section 5, the control logic has been incorporated into a human-machine interface (HMI) developed using SIMATIC WinCC software, replicating the operating conditions of the actual beneficiation plants. Therefore, the proposed methodology is suitable for implementation within existing automation infrastructures, and a pilot deployment is planned as part of future efforts.

Comments 2: A hybrid system of this type will require advanced automation/computational resources to perform the control, which will increase the cost. These costs are not mentioned or quantified anywhere in the article. Furthermore, will a system of this type lead to energy optimization (cost reduction) or to an increase in costs (by requiring more demanding computational resources and a variation in the acceleration and deceleration of the system's motors)?

Response 2: Agree. We have, accordingly, revised the Conclusion to emphasize this point. A structured Sustainability Impact Analysis has been added (Page #15, Lines # 522-531), summarizing environmental, economic, and resource efficiency effects. The proposed control system is not directly aimed at reducing water or energy consumption, as the jigging process already operates within a stable technological framework. Instead, the focus is on minimizing the chromium content in tailings and improving concentrate quality. This leads to increased economic efficiency and reduced environmental impact by reducing metal losses. Production costs remain unchanged, but product quality and resource utilization are improving significantly.

The proposed system can be integrated into existing automation platforms using MATLAB and standard Siemens controllers. The only additional cost is a MATLAB license (~USD 2500). No new hardware is required if the existing PLC and HMI infrastructure is in place. The computational load is minimal and does not significantly affect overall energy consumption.

Regarding the issues related to the increased energy consumption caused by acceleration and deceleration of the engine, the proposed system does not create any additional mechanical load and does not change the physical cycle of actuation of the jigging machine. Instead, the optimization algorithm adjusts the control settings (layer level and ripple frequency) within the normal operating range of the equipment, thus maintaining the dynamics of the engine within the nominal design parameters. As a result, no additional energy requirements or increased wear are expected due to the adaptation of the control.

Comments 3: How much could a system of this type cost a company? Is the investment recoverable? In how many years?

Response 3: We thank the reviewer for this practical and relevant question. Assuming that the MATLAB license is the only required additional investment (~USD 2500), and that existing infrastructure (sensors, PLCs, HMIs) is already in place, the estimated payback period would be within 1–2 production shifts. Operational data analysis indicates that suboptimal control may lead to losses equivalent to 2-2.5t of saleable concentrate with 53% grade per shift. These findings highlight that the proposed system is not only technically feasible but also economically attractive for industrial deployment. We have, accordingly, revised the Discussion (Page #14, Lines #488-491), to emphasize this point.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors, your raised an interesting research points, However, it is better to consider the following comments, questions and suggestions. 

  1. Abstract/Introduction
  • The abstract effectively summarizes the work but is somewhat dense. Consider breaking it into shorter sentences for clarity.
  • Could the authors briefly mention in the abstract how the proposed system could be generalized to other mineral processing systems?
  1. Research Motivation/Gap
  • The manuscript highlights the limitations of existing models , but a more explicit comparison with traditional process control methods would be beneficial.
  • What specific challenges in current industrial chromite ore beneficiation motivated the need for hybrid control? Better to provide a schematic or flowchart comparing traditional vs. hybrid systems for context.
  1. Methodology 
  • The integration of regression, particle motion, and neural network models is well-described. However, the neural network's training process lacks detail.
  • How many data points were used for training and validation? Were any overfitting issues encountered?
  • Better to clarify whether real or synthetic data were used and detail the sources.
  1. Optimization 
  • The real-time control system and optimization approach using MATLAB's fmincon are clearly presented.
  • How was the objective function weighting (α = 0.6, β = 0.4) selected? Were sensitivity analyses performed?
  • Explain the rationale behind using the Interior-Point method over others.
  1. Results 
  • Figures and performance metrics are well-integrated. However, actual implementation in an industrial or pilot setup is not discussed.
  • Has this system been tested beyond simulations? If not, are there plans to do so?
  • Include confidence intervals or statistical validation of MSE and performance improvements.
  1. Environmental/ Economic Impacts
  • The work emphasizes sustainability but lacks detailed environmental impact metrics (e.g., water savings, energy reductions).
  • How does the system contribute to energy or water use reduction compared to baseline methods?
  • Include a cost-benefit or sustainability impact analysis.
  1. Figures and Tables
  • Some figures (e.g., simulation surfaces and neural network architecture) are informative, but clarity could be improved.
  • : Ensure all axes are labeled with units, and provide figure captions that are self-explanatory.
  1. Related Work 
  • The literature review is thorough, but several recent advances in AI for mineral processing are not mentioned.
  • Have the authors considered incorporating federated learning or reinforcement learning for future real-time adaptability?
  • Discuss how this approach compares to emerging techniques in the literature.
  1. Writing Style and Organization
  • Overall, the manuscript is well-organized, but several long paragraphs would benefit from better transitions and clearer sub sections. Revisit sections 3 and 4 for better logical flow and separation of ideas.

Author Response

Comments 1: Abstract/Introduction.

The abstract effectively summarizes the work but is somewhat dense. Consider breaking it into shorter sentences for clarity.

Could the authors briefly mention in the abstract how the proposed system could be generalized to other mineral processing systems?

Response 1: Thank you for pointing this out. We agree with this comment. Therefore, we have revised the abstract by breaking down long and complex sentences into shorter ones (Page #1, Lines #8-10, Lines #12-13). Additionally, a sentence has been added at the end of the abstract clarifying how the proposed system can be generalized to other mineral processing applications (Page #1, Lines #22-24).

Comments 2: Research Motivation/Gap.

The manuscript highlights the limitations of existing models, but a more explicit comparison with traditional process control methods would be beneficial.

What specific challenges in current industrial chromite ore beneficiation motivated the need for hybrid control? Better to provide a schematic or flowchart comparing traditional vs. hybrid systems for context.

Response 2: Agree. We have, accordingly, revised the Introduction (Page #3, Lines #96-100), to emphasize this point. We have added a paragraph describing the specific limitations of traditional control in chromite ore beneficiation (Page #3, Lines #96-99). Table 1 (Page #3, Line #100) has also been added to compare traditional and hybrid control systems.

Comments 3: Methodology.

The integration of regression, particle motion, and neural network models is well-described. However, the neural network's training process lacks detail.

How many data points were used for training and validation? Were any overfitting issues encountered?

Response 3: Agree. We have, accordingly, revised Section 3.3 (Pages #9-10, Lines #334-341) to clarify the neural network training process. Specifically, we have added the number of data points used (1500 total: 70% training, 15% validation, 15% testing), the use of both real and synthetic data, and their respective sources. Measures to prevent overfitting, such as early stopping and regularization, are also described in the revised text.

Comments 4: Optimization.

The real-time control system and optimization approach using MATLAB's fmincon are clearly presented.

How was the objective function weighting (α = 0.6, β = 0.4) selected? Were sensitivity analyses performed?

Explain the rationale behind using the Interior-Point method over others.

Response 4: Agree. We have, accordingly, revised Section 4 (Page #10, Lines #376-378), to clarify the choice of weighting factors (α = 0.6, β = 0.4), and added a brief sensitivity analysis to support it. In Section 4, we added that the Interior-Point method was selected based on a comparative analysis with the Genetic Algorithm and Gradient Descent (Page #12, Lines #425-433). Comparative analysis showed that the fmincon (Interior-Point) method provides the fastest decrease in the value of the objective function.

Comments 5: Results.

Figures and performance metrics are well-integrated. However, actual implementation in an industrial or pilot setup is not discussed.

Has this system been tested beyond simulations? If not, are there plans to do so?

Include confidence intervals or statistical validation of MSE and performance improvements.

Response 5: Agree. We have, accordingly, revised Section 5 (Page #14, Lines #483-488), to emphasize this point. In the results of Section 5, confidence intervals and statistical validation of MSE and performance improvement were added to the revised version. The system was implemented in a digital twin environment using TIA Portal and SIMATIC WinCC for real-time visualization and testing. While full-scale industrial deployment has not yet been performed, pilot implementation is planned at a chromite beneficiation facility.

Comments 6: Environmental/ Economic Impacts

The work emphasizes sustainability but lacks detailed environmental impact metrics (e.g., water savings, energy reductions). How does the system contribute to energy or water use reduction compared to baseline methods? Include a cost-benefit or sustainability impact analysis.

Response 6: Agree. We have, accordingly, revised the Conclusion to emphasize this point. We have added a structured Sustainability Impact Analysis to the Conclusion (Page #15, Lines # 522-531), summarizing environmental, economic, and resource efficiency effects. The proposed control system is not directly aimed at reducing water or energy consumption, since the sedimentation process is already operating within a stable technological framework. Instead, the focus is on minimizing the chromium content in tailings and improving concentrate quality. This leads to increased economic efficiency and reduced environmental impact by reducing metal losses. Production costs remain unchanged, but product quality and resource utilization are improving significantly.

Comments 7: Figures and Tables

Some figures (e.g., simulation surfaces and neural network architecture) are informative, but clarity could be improved.

Ensure all axes are labeled with units, and provide figure captions that are self-explanatory.

Response 7: Agree. We have, accordingly, checked all the figures to emphasize this point. Figure captions 3, 4, 9 have been revised to be more self-explanatory (Page #6, Line #222), (Page #8, Line #273), (Page #13, Line #446). Additional commentary related to the simulation results has been included (Page #8, lines #278–279), (Page #12, Lines #425-432).

Comments 8: Related Work

The literature review is thorough, but several recent advances in AI for mineral processing are not mentioned.

Have the authors considered incorporating federated learning or reinforcement learning for future real-time adaptability?

Discuss how this approach compares to emerging techniques in the literature.

Response 8: Agree. We have, accordingly, revised the text of Section 5 “Discussion” to emphasize this point (Page #15, Lines #504-511). At the end of Section 5, a paragraph was added discussing the potential of integrating reinforcement learning, federated learning, transfer learning, and explainable AI as future extensions of the current hybrid system. Relevant references were included [33-36].

Comments 9: Writing Style and Organization

Overall, the manuscript is well-organized, but several long paragraphs would benefit from better transitions and clearer sub sections. Revisit Sections 3 and 4 for better logical flow and separation of ideas.

Response 9: Agree. We have, accordingly, revised Sections 3 and 4 to improve the logical flow and readability. Structural improvements have been made to the sections:

- Several long paragraphs were split into smaller, more focused units (Sections 3).

-  Subheadings were added to clarify the structure and improve the separation of ideas 4.1 (Page #10, Lines #353-354), 4.2 (Page #11, Line #386), 4.3 (Page #12, Line #418), 4.4 (Page #12, Line #439).

- Transition phrases were added at the beginning of Sections 3.1 (Page #6, Lines #208-209), 3.2 (Page #7, Lines #243-244), 4.1 (Page #10, Lines #353-354), 4.3 (Page #12, Lines #419-420).

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for taking into account our comments and for the document revisions

Comments on the Quality of English Language

A final proof reading is required

Reviewer 3 Report

Comments and Suggestions for Authors

All of my concerns have been addressed. 

Back to TopTop