A Moving Window Double Locally Weighted Extreme Learning Machine on an Improved Sparrow Searching Algorithm and Its Case Study on a Hematite Grinding Process

Round 1

Reviewer 1 Report

The paper deals with the grinding of iron ore, specifically the use of a double weighted model of an extreme learning machine based on a moving window to predict on-line the size of the crushed ore. The authors proposed an improved sparrow search algorithm to optimize the model parameters. In order to verify the effectiveness of the proposed model, the results of the hematite grinding process were used.

The work is very extensive and valuable. Searching for numerical solutions in cases where experimental results are difficult or impossible to obtain can bring huge benefits.

Notes for paper:

In the last paragraph of chapter „1. Introduction”, a summary of the structure of the paper seems superfluous:

“…The rest of the paper is structured as follows. In the second part, the hematite grinding process and the grinding particle size are analyzed. In Section 3, the modeling 132 method of grinding particle size based on MW-DLW-ELM is introduced. In Section 4, the  original SSA is briefly reviewed at first. Then, the proposed ISSA is explained in detail. The fifth part shows the experimental results of the proposed model and optimization 135 algorithm. At last, conclusions are drawn in Section 6.”

The article is not its characterization. I suggest removing this part.

Author Response

 Thank you for your kind comments and suggestions for our manuscript. With regard to your comments and suggestions, we would like to reply as follows:

1^st Comment

In the last paragraph of chapter „1. Introduction”, a summary of the structure of the paper seems superfluous:

“…The rest of the paper is structured as follows. In the second part, the hematite grinding process and the grinding particle size are analyzed. In Section 3, the modeling 132 method of grinding particle size based on MW-DLW-ELM is introduced. In Section 4, the  original SSA is briefly reviewed at first. Then, the proposed ISSA is explained in detail. The fifth part shows the experimental results of the proposed model and optimization 135 algorithm. At last, conclusions are drawn in Section 6.”

The article is not its characterization. I suggest removing this part.

Response

Thanks for your comments.We have removed this part in Section 1.

Reviewer 2 Report

The paper tries to model a grinding process with different AI techniques. Conversely to what the title suggests and what the abstract affirms, the hematite grinding process is just considered as set of data to test different algorithms and the results are don’t used for process modelling purpose. This is due to the absence of mineral processing skill of the authors. This lack of knowledge is well visible in the section 2 devoted to the hematite grinding process itself for which inappropriate vocabulary is used and the performance parameters are wrongly defined.

Concerning the main part of the paper devoted to machine learning methods and used optimisation algorithms, it is divided into four parts:

-         A (too) long introduction (section 1) with a bibliography of a large spectrum of techniques in a too wide field of applications with too few examples in mineral processing.

-       A description of the machine learning type (MW-DLW-ELM) used for this study (section 3) which is not clear and without link with the final objective.

-        The section 4 is devoted to the algorithms used for the model parameters optimisation, where different algorithms are presented.

-        Finally (section 5), these different modelling and optimisation methods are tested on a set of benchmark functions. A small part is using the set of data coming from the grinding process and no conclusions are done about the use of such a model.

Concerning the form of this paper, English level is acceptable except the observed inappropriate vocabulary. The way to list and to refer to the references is not totally in accordance with the recommendations of the editors.

Author Response

Responses to reviewer 2#

  Thank you for your kind comments and suggestions for our manuscript. With regard to your comments and suggestions, we would like to reply as follows:

1^st Comment

 A (too) long introduction (section 1) with a bibliography of a large spectrum of techniques in a too wide field of applications with too few examples in mineral processing.

Response

Thanks. We have added some examples in mineral processing in the first paragraph of Section I. The examples of modeling and optimization algorithms have been simplified in page 2-4.

2^nd Comment

This lack of knowledge is well visible in the section 2 devoted to the hematite grinding process itself for which inappropriate vocabulary is used and the performance parameters are wrongly defined.

Response

Thanks for your suggestions. We have deleted the wrong definition of the performance parameters. The grinding particle size is re-described in the second paragraph of section II on page 4 of the revised manuscript.

3^rd Comment

A description of the machine learning type (MW-DLW-ELM) used for this study (section 3) which is not clear and without link with the final objective.

Response

Thanks for your comment. We have added an explanation of the relationship between MW-DLW-ELM and the proposed optimization algorithm in the last paragraph of section III on page 7 of the revised manuscript.

4^th Comment

Finally (section 5), these different modelling and optimisation methods are tested on a set of benchmark functions. A small part is using the set of data coming from the grinding process and no conclusions are done about the use of such a model.

Response.

Thanks for your comment. In the last two paragraphs of section 5 of the revised paper, we have added conclusions about the MW-DLW-ELM model.

5^th Comment

Concerning the form of this paper, English level is acceptable except the observed inappropriate vocabulary. The way to list and to refer to the references is not totally in accordance with the recommendations of the editors.

Response.

Thanks for your comment. We have revised the list of references according to the recommendations of the editors.

Round 2

Reviewer 2 Report

General point of view for the second version of the paper:

Most of the remarks made for the version 1 can be done for the version2. There are not real improvements concerning the weaknesses of the study itself and the paper presenting it.

The proposed title does not reflect the paper subject. Indeed, the main topic is not mineral processing and not specifically the grinding circuit for hematite-based iron ore. The main topic is to find a machine learning method and the most appropriate optimisation algorithm for the learning stage. Consequently, the title must be change.

Similarly, the abstract does not reflect the paper. It must be changed also by mentioning the grinding circuit as a case study to verify the advantages of the modelling technique. A researcher in the field of mineral processing and of its control, will not be interested by such a study. He is interested by the application.

Maybe two papers are preferable:

-        one presenting the modelling techniques with the set of data from the grinding circuit presented as a test of that model;

-        a second (written with the peoples which are involved in the use this model) devoted to the grinding process and how the model can be used day after day by the plant staff.

Other important remark:

-        The version 2 contains a lot of mistakes in the references. All the references must be carefully checked and updated, in the final list and in the text.

-        Many mathematical formulae seem wrong and must be verified.

-        Some values in tables are wrong.

 

Despite some improvements, the same weaknesses remain in this second version which cannot be published as it is. Major changes must be considered.

See attached file for remarks in the text.

Comments for author File: Comments.pdf

Author Response

Point-to-point Replies to Reviewers’ Comments

We are thankful to all reviewers for their valuable comments and suggestions, which are useful to improve the manuscript. We have made point-to-point responses to all reviewers’ comments. In the revised manuscript, we have marked the revised parts in yellow.

Responses to reviewer 2#

  Thank you for your kind comments and suggestions for our manuscript. With regard to your comments and suggestions, we would like to reply as follows:

1^st Comment

The proposed title does not reflect the paper subject. Indeed, the main topic is not mineral processing and not specifically the grinding circuit for hematite-based iron ore. The main topic is to find a machine learning method and the most appropriate optimisation algorithm for the learning stage. Consequently, the title must be change.

Response

Thanks. We have changed the title as “A moving window double locally weighted extreme learning machine on an improved sparrow searching algorithm and its case study on a hematite grinding process” in the revised manuscript.

2^nd Comment

Similarly, the abstract does not reflect the paper. It must be changed also by mentioning the grinding circuit as a case study to verify the advantages of the modelling technique. A researcher in the field of mineral processing and of its control, will not be interested by such a study. He is interested by the application.

Maybe two papers are preferable:

-        one presenting the modelling techniques with the set of data from the grinding circuit presented as a test of that model;

-        a second (written with the peoples which are involved in the use this model) devoted to the grinding process and how the model can be used day after day by the plant staff.

Response

Thanks for your suggestions. We have rewritten the abstract. The revised abstract can be found on the first page of the revised manuscript.

3^rd Comment

Other important remark:

-        The version 2 contains a lot of mistakes in the references. All the references must be carefully checked and updated, in the final list and in the text.

-        Many mathematical formulae seem wrong and must be verified.

-        Some values in tables are wrong.

Response

Thanks for your comment. We have verified the remarks in the text. (1) The references in the final list and in the text are checked and updated. (2) The formulae in section 3 and 4 are verified and modified. (3) The wrong values in table 1 and table 3 are updated. (4) The section 2 and conclusions are modified according to your suggestions. The revised parts are marked in yellow in the revised manuscript.

Round 3

Reviewer 2 Report

Improvements have been done following recommendations. Some minor revisions must be done to have an acceptable paper (see comments in the attached file).

Comments for author File: Comments.pdf

Author Response

Responses to reviewer 2# Thank you for your kind comments and suggestions for our manuscript. With regard to your comments and suggestions, we would like to reply as follows:

1st Comment

Improvements have been done following recommendations. Some minor revisions must be done to have an acceptable paper (see comments in the attached file).

Response

Thanks for your comment. We have checked the remarks in the text. (1) The marked texts in page 3 and page 4 are checked and updated. (2) The formulae in page 5 are verified and modified. (3) The descriptions about the samples for the experiments in page 14 are modified according to your suggestions. All the revised parts are marked in yellow in the revised manuscript.