Round 1

Reviewer 1 Report

Please find the attached PDF file. 

Comments for author File: Comments.pdf

Author Response

Response to Reviewer 1 Comments

 

Point 1: The introduction is not the right place to present equations1 to 7 and should be revised. It is better to first state the research problem in the introduction and then explain the necessity and importance of the research. In the following, the general literature review should be presented in detail. It is suggested to strengthen this section with updated and new papers. In the last paragraph of the introduction, the purpose of the research should be clearly stated, relying on the novelty of the present study.

Response 1: We modified the introductory section by moving the formula to Section 2. The introduction has also been revised to highlight the novelty of this paper study.

 

 

Point 2: In order to better describe the AI models used in this study, it is recommended to use the following references and similar papers:

- https://doi.org/10.1016/j.compstruct.2022.116599

- https://doi.org/10.12989/scs.2022.44.3.389

- Ensemble deep learning-based models to predict the resilient modulus of modified base materials subjected to wet-dry cycles. (2023). Geomechanics and Engineering, Volume 32, Number 6, March25 2023, pages 583-600.

Response 2: We thank the reviewers for their comments and we have added the appropriate references in the manuscript. This literature played an important role in the process of revising our paper.

 

Point 3: Fig. 6 should be presented separately for training and testing datasets.

Response 3: Figure 6 was modified to show the training set and test set separately. The ratio of data in the training set to the test set is 7:3

 

Point 4: The box plot curve should be added to fig. 6.

Response 4: The curves for the box plot already exist in Figure 6.

 

Point 5: PI and VAF performance evaluator indexes should also be added to Table 5.

Response 5: According to the reviewer's comments, we added metrics such as PI\VAF\SI in Table 5 to evaluate the performance of different models from multiple metrics.

 

Point 6: The models should first scored and then ranked based on all the performance evaluator indices used in Table 5.

Response 6: The performance of several metrics is ranked in Table 5 to comprehensively evaluate the performance of different models. Finally, the XGBoost model has the highest score and is the optimal model.

 

Point 7: For the results of each model, the frequency diagram and error histogram distribution should be presented as shown below.

Response 7: We supplemented each machine learning model with frequency diagrams and error histogram distributions.

 

Point 8: Taylor diagram for both dataset should be provided considering the training and testing dataset (overall 4 Taylor diagram is needed).

Response 8: We added four Taylor diagrams to compare the performance of different machine learning models on the test and training sets.

 

Point 9: For the developed model, sensitivity analysis and uncertainty assessment (U95 analysis) should be investigated.

Response 9: Thanks very much for the reviewers' comments. I wonder if we understand sensitivity analysis correctly, i.e., the effect of uncertainty in the input and model parameters on the predictive performance of the model, or "sensitivity". In this case, we have analyzed and validated the model using 10-fold cross-validation and parameter search. In addition, based on the reviewers' comments, we added uncertainty analysis (U95) for different machine learning models.

Reviewer 2 Report

Interesting article. The greatest advantage is the explanation in simple words of the meaning of the individual indicators, which can be useful for engineers trying to use the information presented.

The article, while the title refers to machine based learning, presents in a general way the matching of selected indicators (coefficients, algorithms) and is based on their matching in the conditions of shear, bending and compression. As indicated by the results, depending on the model, the fit is different, which directly affects the result of the assessment (negative or positive). Nevertheless, the matching factor is high (0.997), which makes this analysis worth showing wider. Nevertheless, it is worth paying attention to tables 2 and 3, where the standard deviation is very high. This is where the key question arises whether the initial assumptions for the analysis of algorithms are selected correctly - Fig. 6.

The conclusions were referred to the obtained results of the analysis

Author Response

Response to Reviewer 2 Comments

 

Point 1: Interesting article. The greatest advantage is the explanation in simple words of the meaning of the individual indicators, which can be useful for engineers trying to use the information presented.

The article, while the title refers to machine based learning, presents in a general way the matching of selected indicators (coefficients, algorithms) and is based on their matching in the conditions of shear, bending and compression. As indicated by the results, depending on the model, the fit is different, which directly affects the result of the assessment (negative or positive). Nevertheless, the matching factor is high (0.997), which makes this analysis worth showing wider. Nevertheless, it is worth paying attention to tables 2 and 3, where the standard deviation is very high. This is where the key question arises whether the initial assumptions for the analysis of algorithms are selected correctly - Fig. 6.

The conclusions were referred to the obtained results of the analysis

Response 1: The reviewer's comments are greatly appreciated. On the one hand, the data ranges differ, resulting in some large standard deviations. Moreover, although the dataset contains larger data, it still does not adequately cover all parameter ranges, resulting in a high standard deviation of the dataset. This is the drawback of data-driven machine-learning based algorithms, which require enough data for training to ensure good generalization performance. The next step is to conduct a more in-depth study with more advanced algorithms that can reduce the data requirements and improve the generalization performance of the model.

 

Reviewer 3 Report

The paper entitled " Machine learning-based prediction of elastic buckling coefficients on diagonally stiffened plate subjected to shear, bending and compression" is worth for publishing in this esteemed Journal. However some modifications are needed. I am herewith enclosing them in point wise. 

1. Page 1, line 18, abbreviation  SHapley Additive exPlanations and some other lines also 

2. Page 2, validity conditions for Kcr , r >1 and 2/3, should be mentioned in the text so that reader can understand importance of the work. 

3. Explanation is needed for given symbols in equation 6

4. page 3, line 73, is not clearly written and equation 8 and 9 should explain conceptually. 

5. Proper reason is not written for why r>1 and 0.5<r (gamma )

6. Fig 2 pictorial explanation may give more strength to paper. 

7. typo error for K-NN in line 143, 357, 358 etc. 

8, more conceptual discussion is needed for better understanding of the best fitting of K cr for Yonezewn and the author contribution. The same can be pointed out in the abstract part also. 

 

 

 

Author Response

Response to Reviewer 3 Comments

Point 1: Page 1, line 18, abbreviation  SHapley Additive exPlanations and some other lines also

Response 1: In the manuscript, "SHapley Additive exPlanations" has been abbreviated to "SHAP".

 

 

Point 2: Page 2, validity conditions for Kcr , r >1 and 2/3, should be mentioned in the text so that reader can understand importance of the work.

Response 2: The conditions for the use of shear buckling coefficient and bending buckling coefficient kcr under the separate action of shear and bending are added in the manuscript.

 

Point 3: Explanation is needed for given symbols in equation 6

Response 3: The descriptions of the parameters related to Eqs. 5~7 are added. Furthermore, for more information about these formulas, refer to Section 3.3 for more clarity.

 

Point 4: page 3, line 73, is not clearly written and equation 8 and 9 should explain conceptually.

Response 4: The description of the concepts of equations 8 and 9 is added in the manuscript. These equations are equations fitted by Mikami to calculate the buckling coefficients of diagonally stiffened plates in pure shear and pure bending based on the resultant curves of numerical analysis.

 

Point 5: Proper reason is not written for why r>1 and 0.5<r (gamma )

Response 5: Martins and Cardoso performed numerical simulations for stiffened plates of 0.5<γ<2.0, fitted the computational equations based on the finite element results, and chose a segmented fitting function, thus dividing the range of application of the formula into two intervals of 0.5 to 1.0 and 1.0 to 2.0.

 

Point 6: Fig 2 pictorial explanation may give more strength to paper.

Response 6: Additional explanations are provided for Figure 2.

 

Point 7: typo error for K-NN in line 143, 357, 358 etc.

Response 7: All "KNN" in the manuscript has been changed to "K-NN".

 

Point 8: more conceptual discussion is needed for better understanding of the best fitting of K cr for Yonezewn and the author contribution. The same can be pointed out in the abstract part also.

Response 8: The abstract has been modified to highlight the innovations in this paper. Yonezewn's formulation considers only a single load action and assumes that the stiffeners are ideal rigid bodies. In this paper, diagonally stiffened plates under the combined action of compression, bending, and shear are studied, and the flexural and torsional stiffnesses of stiffened ribs are considered. The proposed ML model helps to calculate the elastic buckling coefficient of the structure efficiently and accurately.

 

Reviewer 4 Report

The similarity index is more than 20% ; moreover the individual plagarism is more than 5% which is not acceptable. This article is a paraphased verison of the article titled: 

Numerical Study on Elastic Buckling Behavior of Diagonally Stiffened Steel Plate Walls under Combined Shear and Non-Uniform Compression

Author Response

Response to Reviewer 4 Comments

Point 1: The similarity index is more than 20% ; moreover the individual plagarism is more than 5% which is not acceptable. This article is a paraphased verison of the article titled: Numerical Study on Elastic Buckling Behavior of Diagonally Stiffened Steel Plate Walls under Combined Shear and Non-Uniform Compression

Response 1: Thank you for your comments. This paper is a follow-up study to "Numerical Study on Elastic Buckling Behavior of Diagonally Stiffened Steel Plate Walls under Combined Shear and Non-Uniform Compression". In the previous study, the buckling behavior of diagonally stiffened steel plate walls under combined compression-bending-shear was initially analyzed by finite elements. However, we were unable to propose an elastic buckling load calculation formula that could be used for the calculation. In this paper, we further investigate on this basis with a view to establishing a model for predicting the elastic buckling coefficient by machine learning algorithm and analyze in depth the contribution of relevant parameters to the buckling coefficient. The two papers have similar research backgrounds, but the main research methods and research focus are different. For the reader's convenience and a better understanding of the research background of this paper, a detailed research background, including references to previous research results, is presented in Section 2, and the main research content is not the same.

We have added references to cited sources in the manuscript and revised some of the content to avoid the high similarity problem in this manuscript.

Round 2

Reviewer 1 Report

The authors have improved the quality of the paper in terms of organization and writing. In the reviewer's opinion, the current manuscript can be accepted for publication on the Journal of sustainability. 

Reviewer 4 Report

I don't see any significant improvement in the article.