A Hybrid Physics-Informed and Data-Driven Approach for Predicting the Fatigue Life of Concrete Using an Energy-Based Fatigue Model and Machine Learning

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Please see the attachment

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This study proposes a hybrid approach combining an energy-based fatigue model and machine learning to predict the fatigue life of concrete under cyclic loading, leveraging physical principles and data-driven techniques for enhanced accuracy and efficiency. The energy-based model is validated against experimental data, while machine learning models, including KNN and DNN, are trained and compared, with the DNN showing superior performance, particularly for out-of-range predictions. While the results of the research are noteworthy and hold significant value, certain weaknesses in the current document must be addressed to enhance its overall rigor and quality to be commensurate with the publication's value.

1. Data collection could be improved by incorporating more references and experimental data. The statistics in Table 1 are insufficient; adding box plots and distribution charts would better illustrate data patterns and enhance the dataset's reliability for model training and evaluation.

2. In addition to the existing KNN and DNN models, other machine learning models, such as Random Forest, Support Vector Machines (SVM), and XGBoost, can be further explored for comparative analysis to evaluate their performance in terms of prediction accuracy and computational efficiency. To reduce the reliance on highly complex networks, feature selection or dimensionality reduction techniques can be introduced to optimize the number of input variables and further enhance computational efficiency.

3. The current research focuses more on model accuracy but has not deeply explored its implications for practical engineering design guidance. It is recommended to provide specific case studies, such as fatigue life prediction and optimization design for bridge beams or construction materials, to enhance the model's applicability in engineering practice.

4. The paper does not specify the materials used in the experiments or their sources, which may affect result reproducibility. It is recommended to include details on the experimental materials, their origins, and key properties, along with the experimental setup, data collection, and processing methods. Adding seismic testing data for columns and beam-column joints could further enhance the model's relevance to seismic design applications.

5. Fig. 14 shows that the experimental data presented are limited, with the energy-based model achieving the best fit, while the performance of other models is relatively poor. It is recommended to improve the results by incorporating more experimental data for comparison and refining the other models to enhance their predictive accuracy and consistency with the experimental observations.

6. For references, you may add more ML articles (Exploring temperature-resilient recycled aggregate concrete with          waste rubber: An experimental and multi-objective optimization analysis. Reviews on Advanced Materials Science.).

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

SUMMARY

The presented article formulates a hybrid approach based on physics and data for predicting the fatigue life of concrete using an energy-based fatigue model and machine learning.

The relevance of the study is due to the need to study and predict the fatigue of concrete structures. The safety and durability of concrete and reinforced concrete buildings and structures directly depend on this.

The authors proposed an interesting model for predicting fatigue life and presented a hybrid approach based on physics and data. This article has new fundamental scientific knowledge, as well as prospects for engineering application.

The reviewer believes that this article has scientific novelty and practical significance, but the comments that were made should be corrected. They are given below.

 

COMMENTS

1. I would like the authors to emphasize in the abstract that the hybrid approach based on physics and data is fundamentally new.

2. At the end of the abstract, the authors should show how much better the forecasts for the obtained models are than similar ones. A percentage quantitative expression of these results is needed.

3. Only 4 keywords are presented. At least 6 keywords should be presented. Searching for an article in search algorithms is an important aspect. The authors should take care of future readers.

4. In the Introduction section, I would like to see more clearly formulated research objectives. It is necessary to describe the course of the study, starting with the analysis of the literature and ending with its implementation in practice.

5. It is proposed to add a Materials and Methods paragraph. I would like the authors to reflect in it the data that were used, as well as describe the methodology in detail. It is necessary to reflect both the physical methods of the study and the machine learning methods.

6. The Discussion section is missing, allocated in a separate paragraph. A detailed comparison of the obtained results with the results of other authors should be provided. In this case, the authors should compare their hybrid approach with single-stage studies where only physical methods or only machine learning methods were used.

7. The scientific novelty and practical significance of the study should be formulated more clearly. What is the novelty of the obtained scientific results? What is their engineering significance? The results need to be assessed for both the construction industry and the field of artificial intelligence. This should be added to the Discussion.

8. It would be nice if the authors suggested specific engineering objects where this approach can be applied now.

9. The Conclusions section needs to be rewritten. It should be presented in the format of numbered conclusions. Numbered conclusions should include the applied result, the scientific result, the scientific prospects for the development of the study, and the prospects for engineering application with an indication of real objects.

10. The list of references includes 41 titles, but most of the studies are older than the last 5 years. The authors need to work with the scientific literature and provide more recent studies over the last 5 years. Machine learning and fatigue life issues have been discussed in great detail on the MDPI platform. At least 15-20 references over the last 5 years should be added.

After correcting the comments, the reviewer would like to take another look at the article. Overall conclusion – Major Revisions.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors could not address my comments

Author Response

Comment 1: The authors could not address my comments
Response 1: Thank you for your response. The authors would like to clarify that, in the present study, temperature effects are ignored since the model is formulated for a brittle material (concrete) subjected to static loading with low-frequency strain loading, where energy dissipation due to heat can be neglected. The sources of dissipation in this case arise from microcrack formation and dislocation movements, leading to material damage and plasticity, respectively. This point has been addressed in the revised manuscript on page 3, section 1, and line 109–114.
Additionally, the purpose of this study is to couple physics-based models with data-driven machine learning models, which represents a fundamentally new approach. However, the authors appreciate the reviewer’s comments and would like to mention that these aspects will be addressed in the future scope of the study, as referenced in the revised manuscript on page 18, section 5, and line 489–493. We hope this clarification addresses your concern.

Reviewer 3 Report

Comments and Suggestions for Authors

The authors did a great job. All comments have been corrected. The article is ready for publication.

Author Response

Comment 1: The authors did a great job. All comments have been corrected. The article is ready for publication.
Response 1: Thank you so much for your positive feedback.