Prediction of Uniaxial Compression Strength of Limestone Based on the Point Load Strength and SVM Model
Round 1
Reviewer 1 Report
None
Author Response
Thank you very much for your reading.
Reviewer 2 Report
Ref. No.: MINERALS - 1486907
Title: Prediction of uniaxial compression strength of limestone based on the point load strength and SVM model
Number of Authors: 3
Type of Article: Research article type
Authors are request to rewrite the abstract, which consist of a short introduction, objective of the study, experimental work and quantitative results. Avoid the sentence which are too long for example line page 1, line 9-12. Authors are requested to check the nomenclature of the article throughout the text. Results and discussion part only contain the experimental discussions, not scientific discussion of readings. Results should be compared with literatures and scientific findings need to be added.
Apart from above common comments, I had several questions below
- Page 1, Line 29-30: Rewrite the sentence
- Page 2, Line 54-56: Check the sentence
- Introduction about Support Vector Machine with recent literatures are need to be added
- Change heading 2 as Methods and Materials
- Page 8, Add reference for the assessment formula
More citations are required for discussion part, results and discussions. Hence, I accept the article with minor revision
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 3 Report
Dear Authors:
Review for
Journal Minerals
Title:
Prediction of uniaxial compression strength of limestone based on the point load strength and SVM model
Authors:
Shaoqian Li, Yu Wang, and Xuebin Xie
In this work were performed tests on cylindrical specimens of limestone under uniaxial stress to find the best fit between PLS and UCS, and the correlation of relationship between PLS and UCS is found by using four basic fitting functions, and a prediction of the models is established based on the nonlinear relationship by using SVM algorithm. Authors divided 90 specimens in three groups by 30 in each, and the specimens were separately prepared for axially loaded point load test, diametral loaded point test and uniaxial compression test.
This work is important for building basic for express analysis on engineering cite.
Some parts of manuscript have misprints or errors.
I suggest minor revision of the manuscript. Please find below my comments aimed at improving the presentation of the manuscript.
The manuscript is written in good language and can be read without difficulty. Unfortunately, there are some technical errors in the text that need to be corrected.
First group of mistakes relate to units of stress. Please carefully check and change wrong represented stress in “Mpa” to the correct “MPa” in the text. The same mistake is in Figure 2, Figure 3, Figure 4, Figure 6, and in Figure 7, and should be corrected.
Second group of errors is in the References. The authors did not take in account the requirements to the format of the References. The names of authors often typed in different style, so there is necessary to bring the references to the required format for the journal.
Table 1. The authors collected large number of the empirical equations of relation between PLS and UCS. In this Table are presented the names of authors in different style. Please correct as in the references. Second note: The Table 1 shows the equations in connection to the year of publication, and it is difficult to analyze why the equations are so different. However, it could be much more informative and useful for the readers if to add the column of rocks type opposite the equations and R2.
Week part of the rock specimen description in the manuscript is lack of physical properties such as density and porosity, as well as mineralogical composition. The availability of such data could improve the article.
Some correction needed in rows:
55 – “…common Greek sedimentary rocks”.
75 – “…certain rocks”.
Please be more specific.
118 – instead of “porous” use “pores”.
203, 204 – In my opinion here could be used the Past time in the sentences, otherwise it sounds as instruction but not as a summary what did you do or find.
The manuscript should be improved first and cannot publish now without improvement.
Best wishes
Author Response
Thank you very much for your reading and valuable suggestions. On the basis of
this paper, we make the following modifications. Please see the attachment.
Author Response File: Author Response.pdf
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
This paper used a SVM to predict the UCS of rock by using PLS results. According to the prediction results, it shows a more accurate predictive results by using the proposed method. This is interesting and meaningful. I think this paper can be accepted after revision.
- Change uniaxial compressive strength to uniaxial compression strength.
- The sentences of the manuscript should be carefully revised to show your meanings.
- In figure 4 and 5, the captions of the horizontal and longitudinal ordinates should be added.
- In rock mechanics, there are many empirical correlations between different index. As is shown in Table 1, selecting a proper correlation is very difficult. Although your prediction method is also based on parameter relationship, a progress is made in this paper. In rock materials, there are many defects such as pores and micro cracks. These influencing factors may have different impacts on UCS and PLS, which has been studied in “A multifrequency ultrasonic approach to extracting static modulus and damage characteristics of rock, IJRMMS” for reference.
- Finally, a major question should be asked to the authors. Although your method shows a large advantage over traditional empirical correlations. Can your method be applied to other rocks?
Reviewer 2 Report
I have major concerns about the relevance and validity of the work. For instance, in section 2.1 it is mentioned that the samples were collected from a limestone mine in China. As such, all samples are of the same rock type. From the text I can suppose that the authors collected 30 blocks with apparently different characteristics and prepared three samples from each, one for each test, correct? If that is the case, the work deals with correlations of unreplicated experiments and is an exercise of overfitting.
The use of such a powerful tool such as SVM to deal with a single-variable problem (y=f(x)) is certainly questionable, since it disregards the nature of the errors involved in both measurements.
In my opinion the authors should shorten significantly their manuscript and resubmit it for publication.
It is a case study, which is valid for the rock and mine in question, but is of no relevance when applied to other rocks.
It might be interesting to plot the expressions, instead of Table 1, in order to demonstrate their similarity and/or differences.
Since the authors used cylindrical specimens this should be emphasized right in the abstract.
I believe the title of the work should also characterize the fact that the work was only done with limestone samples, which certainly limits the potential application.
The title of 3.1 seems missing (subsection).
Line 167: it represents a load, not a pressure.
Lines 210: please use “loading rate of 0.6 MPa/s”.
Line 212: what is monoclinic failure? I suppose the authors meant catastrophic or monotonic failure.
Figure 3: histograms are not the best way to verify normality of the data when the number of data points is relatively limited. Strongly suggest using a normality plot, in which all sets of data can be included in a single graph.
Axes should be identified in Figures 4 and 5. By the way, there is no reason why four graphs should appear when a single figure with the four fits can be presented in a single graph.
Authors should acknowledge that some of the correlations in Figure 4 lead to unrealistic extrapolations, with negative predictions of UCS from PLT data.
Tables 2 to 4 should either be in the appendix or the supplementary material, not the body of the text.
What is the actual number of parameters fitted in the SVM model? I suspect many, as in other tools such as neural networks, given the very good agreement found.
