Prediction of Oil Sorption Capacity on Carbonized Mixtures of Shungite Using Artificial Neural Networks

Round 1

Reviewer 1 Report

Cristea, Baigulbayeva and Co-Workers described a « Prediction of Oil Sorption Capacity on Carbonized Mixture of Shungite Using Artificial Neural Networks». The article is well written, all methods are relevant and necessary to describe the results and conclusions obtained in this work.

 

 

While reading, there were a few questions and comments:

1) Need to improve the quality of Figure 2.

2) In Table 1, for the sample Sh:RH=6:1 15% for a duration of 5 days, 0.4 should be replaced by 0.40.

3) In figures 4 and 5 it is necessary:

- make drawings 4c and 5c according to the dimensions a and b;

 

- axis labels and legends should be enlarged or made bold.

4) Why is there a big difference between the experimental data and those calculated up to 30 days when predicting the results (Figure 8)?

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

1. The article spends a considerable amount of time writing about the literature review, mentioning the excellent adsorption capacity of shungite rock and also writing about the low cost, high quantity and good adsorption properties of risk husk biomaterials. So, I would like to know what is innovative about a new hybrid adsorbent made from a mixture of these two known excellent substances. The article mentions this question at literature 33, but this question was asked because the information was not available.

2.  The target of the application according to the article is to remove crude oil contaminants from the Karazhanbas field by means of a new hybrid sorbent. However, the sample prepared for testing is only 50g of sample soil; please explain how the sample was prepared in relation to the target of the article? And how were the oil contents of 10% and 15% considered?

3. The sample size for the article is a total of 60pairs, including 5 mixing ratios and 12 time intervals, for a total of 2 sample soil oil contents. This amount of data (i.e. too small to reveal a answer) does not seem to me to be the basis for a neuronal network algorithm. Please explain the need to use neuronal networks? Have any other machine learning algorithms been tried that are more suitable for this amount of data, such as decision trees? Or even linear regression? (Figure 6 shows a very strong linear relationship)? Please consult the basic requirements for the data volume of different big data analysis algorithms.

4. Please explain why at 15% oil content the performance of this mixed sorbent is significantly reduced compared to 10% oil content (see Table 1)? Could it be that the heavy oil content affects the adsorbent performance? Do we need to do a sensitivity analysis of the oil content?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The manuscript examines the sorption prediction of rice husk and shungite by the development of ANN and the results are validated experimentally. In general the manuscript is well written although some details can be omitted as they overload the manuscript. The novelty mainly lies in the ANN for the specific application. The manuscript can be published considering some recommendations below:

Lin 53. What is the type of oxidation products the authors refer to?

Lines 221-224. Please provide more information on how each mass was experimentally measured.

Section 2.2. The ANN is well described and perhaps some info can be condensed.

Line 301. Is the size of the training dataset big enough for the current ANN as there is a rule of thumb based on the input to consider the ANN trustworthy.

Line 324-327. For the utilisation of other algorithms with regard to deep learing which they might worth to consult see i.e. https://doi.org/10.1016/j.conbuildmat.2021.125481

 

Figures 6 & 7. Is this not a rather limited training/validation of points?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The authors have replied all comments proposed in revised version. Hence, Reviewer approve the acceptance of this manuscript.