Predicting Mechanical Properties of Magnesium Matrix Composites with Regression Models by Machine Learning

Round 1

Reviewer 1 Report (Previous Reviewer 2)

It is noticeable that the authors have significantly expanded the article, including many additional explanations. The paper gives a prediction of the yield strength of nanocomposites with a reinforcement content of 0-4%; the authors state that the yield strength drops significantly when the reinforcement content is less than 2 and greater than 4% (line 471). However, from the data in Table 8 and Graph 9, it can be seen that the yield strength drops sharply around at 1%. You need to pay attention to this.

It is also not clear why Table 6 (which consists of a single row) is needed if the information from this table is repeated in Tables 7 and 8. Tables 6 and 7 could well be combined.

Author Response

Reviewer 1

It is noticeable that the authors have significantly expanded the article, including many additional explanations. The paper gives a prediction of the yield strength of nanocomposites with a reinforcement content of 0-4%; the authors state that the yield strength drops significantly when the reinforcement content is less than 2 and greater than 4% (line 471). However, from the data in Table 8 and Graph 9, it can be seen that the yield strength drops sharply around at 1%. You need to pay attention to this.

It is also not clear why Table 6 (which consists of a single row) is needed if the information from this table is repeated in Tables 7 and 8. Tables 6 and 7 could well be combined

Response to Reviewer :

Thank you for your comment. We have made some revisions in Section 4.3, combined Table 6 and Table 7 into a single table, and also modified a word in Line 471. In addition, we also show the result from coding to support the explanation in line 471.

 

• Table 6. Predictions with Optimization from the Best Model and Other Models using ML Algorithm

ML Algorithm	Matrix	Reinforcement	Reinforcement Particle Form	Variation of Reinforcement (wt%)	Heat Treatment	Mechanical Working	Yield Strength
Decision tree regression	AZ31	GNP	Nano	3 wt %	No heat treatment	Extrusion temperature 3500C	187.MPa
Extra tree regression	AZ31	GNP	Nano	3 wt %	No heat treatment	Extrusion temperature 3500C	187.MPa
Random forest regression	AZ31	GNP	Nano	3 wt %	No heat treatment	Extrusion temperature 3500C	154.353 MPa
XGBoost regression (Best Model)	AZ31	GNP	Nano	3 wt %	No heat treatment	Extrusion temperature 3500C	186.99731 MPa

 

 

• The data in the above table suggested that variations in the reinforcement percentage, especially those near the optimized parameter, only resulted in slight changes to the yield strength. When the reinforcement value was below 3 wt%, the optimized parameter, the yield strength results were marginally lower. Conversely, increases in the reinforcement percentage did not result in a significant change in yield strength compared to the optimized parameter. It's also crucial to acknowledge the role of Graphene Nanoplatelets (GNPs) in these composites. The incorporation of GNPs can significantly influence the microstructure of the composite, potentially causing changes in properties such as basal texture intensity [10]. Interestingly, a drop in yield strength can be observed when the GNP content is either less than 1.25% or greater than 4%. This drop is primarily attributed to weak interfacial bonding between GNPs due to the GNPs' poor wettability.

 

• Variation_of_Reinforcement (wt%) Predicted_Yield_Strength

0                               1.00                171.676147

1                               1.01                171.676147

2                               1.02                171.676147

3                               1.03                171.676147

4                               1.04                171.676147

5                               1.05                171.676147

6                               1.06                171.676147

7                               1.07                171.676147

8                               1.08                171.676147

9                               1.09                171.676147

10                              1.10                171.676147

11                              1.11                171.676147

12                              1.12                171.676147

13                              1.13                171.676147

14                              1.14                171.676147

15                              1.15                171.676147

16                              1.16                171.676147

17                              1.17                171.676147

18                              1.18                171.676147

19                              1.19                171.676147

20                              1.20                171.676147

21                              1.21                171.676147

22                              1.22                171.676147

23                              1.23                171.676147

24                              1.24                171.676147

25                              1.25                186.998749

26                              1.26                186.998749

27                              1.27                186.998749

28                              1.28                186.998749

29                              1.29                186.998749

30                              1.30                186.998749

31                              1.31                186.998749

32                              1.32                186.998749

33                              1.33                186.998749

34                              1.34                186.998749

35                              1.35                186.998749

36                              1.36                186.998749

37                              1.37                186.998749

38                              1.38                186.998749

39                              1.39                186.998749

40                              1.40                186.998749

41                              1.41                186.998749

42                              1.42                186.998749

43                              1.43                186.998749

44                              1.44                186.998749

45                              1.45                186.998749

46                              1.46                186.998749

47                              1.47                186.998749

48                              1.48                186.998749

49                              1.49                186.998749

50                              1.50                186.998749

 

Author Response File: Author Response.pdf

Reviewer 2 Report (Previous Reviewer 3)

The authors accepted my comments, and I am satisfied with the corrected version of the manuscript.

Author Response

Thank you for your comment and appreciation.

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

In my opinion, this material should not be published in any journal, as it does not contain useful information for readers. The title of the article uses the buzzwords “machine learning”, and the topic is magnesium-based alloys, the improvement of the properties of which is indeed an urgent task, but the value of the material is limited to this.

 The main drawback is that neither the initial data nor the equations, according to which those who wish can calculate the target parameter (yield strength), are given. Comparison of various data processing methods, to which the main part of the article is devoted, does not make any sense without demonstrating data, since it critically depends on how they are distributed for each of the input parameters (evenly over the entire range of values or for some data areas much more, there are outliers, etc.) and how many values there are for each input parameter.

 The criteria by which the input parameters were selected are not discussed by the authors, although this is very important for further analysis. For example, if we turn to the articles from which the authors took data, we can see that the compositions of the initial matrices in different studies differ (for example, AZ91 in [8] and [9]). Differences in concentrations and in the very presence of certain alloying elements seem to be small, but the values of tensile yield strength differ by a factor of two. Of course, the differences in TYS can also be due to the difference in the conditions for obtaining alloys (in the sequence of heat treatment, etc.), but there are no grounds for ignoring the effect of all alloying elements. By the way, the material contains neither the characteristics of the matrices, nor the modes of heat treatment, but there is a completely unnecessary Table 3, supposedly to explain the meaning of the coefficient of determination.

 In addition, the authors do not discuss possible differences in the values of the yield strength depending on the test mode: tension or compression. For those alloys for which they differ, what values did the authors use for analysis and why?

 The statement that the authors predicted the system with the maximum value of the yield strength is incorrect, since from Fig. 4 it follows that in their sample there was at least one example with a larger value of YS. Actually, the most reliable forecasts do not go beyond the boundaries of the range of values used for the analysis. The problem is that the value of one parameter for an alloy of arbitrary composition is of no practical interest. Strictly speaking, a combination of several properties is important for practical application. But if the authors believe that they can offer correlations for only one property, then at least for it, equations should be given that allow calculations to be carried out for an arbitrary set of input parameter values. As already mentioned above, they are absent in the material.

 In general, one gets the impression that the authors began to deal with this topic quite recently, since the list of references contains many articles published in the last two or three years, but there are no references to pioneering or fundamental works in the field of statistical treatment (or, if you like, in the field of "machine learning"). At the same time, as already noted, the authors explain very simple things that are clear with knowing basics (e. g., Table 1, Table 3), while more complex ones are left without comment or are illustrated very poorly (for example, it’s impossible to understand the meaning of Fig. 1d or make out the inscriptions on it).

Instead of "Aluminum-zinc elements" (line 33) better (in my opinion) "aluminum and zinc".

The sentence above (lines 30–32)  is best restructured so that the alloying elements are listed first and then the abbreviations for the corresponding alloys. 

Reviewer 2 Report

The article is devoted to the prediction of the mechanical properties of magnesium-based composites using machine learning methods. The work is likely to be relevant and trendy. Needs a few clarifications before publishing.

1. Is the training sample too small to make a convincing choice of a regression model or prediction?

2. In table 5, as far as I understand, the optimized parameters are obtained using the XGBoost regression model, and what optimized parameters will other models give, for which there was also a high R2?

3. In the article [10], to which you refer, the mechanical properties of composites with graphene nanopellets are investigated in the range of 1.5–3.0 wt. %. You get the optimum concentration of 3.0% (Table 5), but why does it follow that, for example, that a composition with more than 3.0% will have a lower yield strength?

 

Or you were looking for an optimum in the range of 1.5-3.0, but then what is the predictive power of the model if it repeats the known experimental data?

English is understandable enough 

Reviewer 3 Report

Interesting written scientific paper. The originality and novelty of the research are not well highlighted.

·   Page 1, Please, describe more deeply the state of the art in the Introduction chapter, or better to add a new section, „Literature review“. 

P   Page 8/Figure 4, Page 11/Figure 5- please check the visibility. The font in the drawings is too small and invisible.

Page 12/ Conclusion- describe the future direction of your research. Knowing how to help with this research in the next research work is always important.

 

 

 

 

 

 

Clear-written scientific manuscript. Minor editing of English language required.