Fault Prediction Modeling for High-Impact Recorders Based on IPSO-SVM

Round 1

Reviewer 1 Report (Previous Reviewer 1)

Comments and Suggestions for Authors

The authors have carried out research related to Fault Prediction Modeling for High-impact Recorders based on IPSO-SVM. The paper follows the standard format of a scientific publication: introduction, methodology, experiments, results, discussion, and conclusions. This structure facilitates its review and readability by the academic community. It proposes the IPSO-SVM model, which improves the optimization of SVM models through an enhanced particle swarm optimization algorithm. This has direct applicability in fault prediction under limited data conditions, which may interest the engineering community. Experiments are conducted using real impact acceleration data, supported by simulations with software like Ansys and Is-Dyna to validate the methodology. The article explicitly addresses limitations, such as the small sample size and the potential need for comparisons with other algorithms

The manuscript could be considered for publication if the authors make minor corrections:

Expand the size of the experimental dataset, if possible, to improve the model's generalization.

Broader Comparisons: Although IPSO-SVM is compared to PSO-SVM, it would benefit from comparisons with other advanced algorithms, such as deep neural networks or AI-based methods.

Improve the quality of figures (e.g., fitness curves, flow diagrams) to ensure they are clear and high-resolution.

Add specific examples of how this model could be applied in other industrial contexts or real-world scenarios, such as aerospace or military development.

Author Response

Please see the attachment. The following text is the same as the content in the attachment.

Comments 1: [Expand the size of the experimental dataset, if possible, to improve the model's generalization.]

Response 1: Agree. Thank you for your valuable suggestion.  Indeed, expanding the dataset would likely lead to better training performance.  However, the focus of this study is specifically on addressing small-sample problems, aiming to achieve reliable predictive performance even with limited data.  Additionally, the experimental data in this study are challenging to obtain, and we have already included all available data in the training process.  Any additional data that were filtered out during training were excluded to avoid excessively long training times or negligible impacts on the results.

Comments 2: [Broader Comparisons: Although IPSO-SVM is compared to PSO-SVM, it would benefit from comparisons with other advanced algorithms, such as deep neural networks or AI-based methods.]

Response 2: Agree. Thank you for your valuable suggestion, which made me realize the importance of broader comparisons. In response, I have added comparisons with other models, including LSTM and Random Forest, in Chapter 3. This enhancement has significantly improved the quality of the paper, and I sincerely appreciate your input.

Comments 3: [Improve the quality of figures (e.g., fitness curves, flow diagrams) to ensure they are clear and high-resolution.]

Response 3: Agree. I have re-uploaded most of the unclear figures and ensured they are of high resolution. Additionally, I have disabled compression in Word to maintain their quality. Thank you for your helpful suggestion.

Comments 4: [Add specific examples of how this model could be applied in other industrial contexts or real-world scenarios, such as aerospace or military development.]

Response 4: Agree. Thank you very much for your valuable feedback, which made me realize that my previous description of the model's practical applications was too general.  I have revised Chapter 4 to provide more specific explanations of the method's applications and included relevant examples.  I truly appreciate your insightful suggestion.

Author Response File: Author Response.pdf

Reviewer 2 Report (Previous Reviewer 3)

Comments and Suggestions for Authors

Comments to the Authors of the manuscript Applsci-3393910 are as follows:

1.     The following comment: “Conclusions are too general and poor in terms of scientific significance. In particular, the conclusion must be expanded and quantified based on the obtained results.”, from the previous round of review, was not addressed as requested by this reviewer.

2.     The list of references must be improved in terms of state-of-the-art, as well as expanded.

3.     In Section 2.1, there is a jump from the citation of Reference [12] to the citation of Reference [14]. Specifically, there is no citation of Reference [13].

4.     In Section 2.1, in the second paragraph, there is the following sentence “To achieve this, the SVM adopts a dual problem formulation, converting the optimization problem using the Lagrange multiplier method[Error! Reference source not found.].” which should be corrected.

5.   Other comments were addressed in an acceptable way.

Author Response

Please see the attachment. The following text is the same as the content in the attachment.

Comments 1: [The following comment: “Conclusions are too general and poor in terms of scientific significance. In particular, the conclusion must be expanded and quantified based on the obtained results.”, from the previous round of review, was not addressed as requested by this reviewer.]

Response 1: Agree.Thank you for your valuable feedback. I have revised the Conclusions section in Chapter 5, making it more specific and quantitatively detailed to enhance the overall quality of the paper. Your suggestion was greatly appreciated.

Comments 2: [The list of references must be improved in terms of state-of-the-art, as well as expanded.]

Response 2: Agree. Thank you for your suggestion.  I have reviewed and updated the reference list, incorporating several additional studies to better reflect recent advancements and enhance the relevance of the citations.  Your feedback has been very helpful in improving the overall quality of the paper.

Comments 3: [In Section 2.1, there is a jump from the citation of Reference [12] to the citation of Reference [14]. Specifically, there is no citation of Reference [13].]

Response 3: Agree. Thank you for pointing out this oversight. I have carefully reviewed Section 2.1 and confirmed the issue with the missing citation of Reference [13]. I have corrected the numbering to ensure that all references are properly cited and sequentially ordered. I appreciate your attention to detail in helping improve the accuracy of the paper.

Comments 4: [ In Section 2.1, in the second paragraph, there is the following sentence “To achieve this, the SVM adopts a dual problem formulation, converting the optimization problem using the Lagrange multiplier method[Error! Reference source not found.].” which should be corrected.]

Response 4: Agree. Thank you for bringing this to my attention. The issue with the reference error in Section 2.1 has been addressed. I appreciate your careful review and helpful feedback.

Author Response File: Author Response.pdf

Reviewer 3 Report (New Reviewer)

Comments and Suggestions for Authors

The paper focuses on developing an approach based on suggested IPSO and SVM  for failure prediction in high-impact recorders.

The idea of creating groups, dedicated for exploration and exploitation separately, is nice and seem to be based on intuition. The equations suggest that if we encourage the search around the global best we would expect the algorithm to produce the perform exploitation. and if we discourage it we would expect the algorithm to produce more exploratory solutions.

However, the PSO algorithm is very simple but produce complex behavior, in a sense that the global best and the personal best constantly change positions, and that each single solution is based on both of these points... the issue with this paper is that it does not demonstrate that the algorithm truly does what it's supposed to do, i.e. the solutions generated by the exploration group actually do exploration and the opposite is true. Other than the results of the benchmark functions,  which seem to suggest that the IPSO is generally better than PSO in these functions.

The authors are advised to demonstrate the two group's behavior by results in the section "Performance of IPSO" or maybe create a new section " Behavior of IPSO". This will make the paper well-rounded to fit the quality of the rest of the paper.

It is also advised to discuss other algorithms that create distinct exploration and exploitation behaviors such as the YUKI algorithm in the literature review, that has been used to solve multiple damage identification problems. it is also advised to report on other algorithm improvement effort that push toward such search behavior.

 

Author Response

Please see the attachment. The following text is the same as the content in the attachment.

Comments 1: [The authors are advised to demonstrate the two group's behavior by results in the section "Performance of IPSO" or maybe create a new section " Behavior of IPSO". This will make the paper well-rounded to fit the quality of the rest of the paper.]

Response 1: Agree. Thank you for your valuable suggestion. Your comment regarding the need to validate the functionality of the particle grouping made me realize this gap. As a result, I have added Section 2.3, titled "Behavior of IPSO," to demonstrate and validate the behavior of the particle groups. This addition enhances the completeness of the paper and better showcases the model's performance. Thank you once again for your helpful feedback.

Comments 2: [It is also advised to discuss other algorithms that create distinct exploration and exploitation behaviors such as the YUKI algorithm in the literature review, that has been used to solve multiple damage identification problems. it is also advised to report on other algorithm improvement effort that push toward such search behavior.]

Response 2: Agree. Thank you for your thoughtful suggestion. I completely understand the value of discussing other algorithms, such as the YUKI algorithm, that create distinct exploration and exploitation behaviors. However, due to time constraints, I was unable to incorporate a detailed comparison with these algorithms in this version of the paper. I appreciate your feedback and will consider this direction for future work, where a broader comparison with other relevant algorithms could be explored. Thank you once again for your understanding.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report (New Reviewer)

Comments and Suggestions for Authors

The reviewer thank the authors for understanding the raised concerns and introducing the suggested section of IPSO behavior.

The idea of demonstrating the behavior through the average variance distances across iterations is nice, but it did not solve the issue. Because the closeness between the exploitation group and the standard PSO throughout the iterations raises more questions. And the large variance in the exploration at the late iterations suggests search inefficiency...

Since the authors could not find time to improve the research background as suggested in the first review round. It's clear that it would not be possible to find time to demonstrate the behavior in a conclusive manner.

The reviewer does not have farther comments on the paper. If it is ok for the authors and the editor. The reviewer does not object to the publication of this paper in the current form.

Author Response

Please see the attachment. The following text is the same as the content in the attachment.

Comments 1: [The reviewer thank the authors for understanding the raised concerns and introducing the suggested section of IPSO behavior.

 

The idea of demonstrating the behavior through the average variance distances across iterations is nice, but it did not solve the issue. Because the closeness between the exploitation group and the standard PSO throughout the iterations raises more questions. And the large variance in the exploration at the late iterations suggests search inefficiency...

 

Since the authors could not find time to improve the research background as suggested in the first review round. It's clear that it would not be possible to find time to demonstrate the behavior in a conclusive manner.]

 

Response 1: Agree. First of all, I would like to express my sincere gratitude for your valuable feedback. Your comments have helped me recognize the shortcomings in my paper, and I have made several revisions based on your suggestions.

 

I have revised the introduction to include a discussion on current advanced optimization algorithms, thus enhancing the background and depth of the research.

 

Regarding the behavior of the IPSO algorithm, I have provided further explanations and additional details in Section 2.3. In particular, regarding the inefficiency of the exploration group in later iterations, I realized that my original explanation was insufficient. In fact, the IPSO algorithm balances this by dynamically adjusting the grouping ratio between the exploration and exploitation groups. I had not adequately explained this, and I truly appreciate you pointing this out. I have now added this clarification in the paper.

 

Additionally, I fully agree with your comment that comparing with other algorithms would improve the quality of the paper. I am currently conducting relevant experiments and will analyze the results accordingly. If I manage to compile valid data before the paper is published, I will promptly contact the editor for an update. If I am unable to gather the necessary data in time, I plan to pursue this comparison as part of future research.

 

Once again, I sincerely appreciate your valuable feedback. Your suggestions have greatly contributed to my research.

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

Comments and Suggestions for Authors

Development of IPSO-SVM Model: A failure prediction model based on the combination of Particle Swarm Optimization (IPSO) and Support Vector Machines (SVM) is presented. This model is optimized to improve the accuracy in failure prediction in high-impact recorders, especially under small sample conditions. The IPSO-SVM model achieved a prediction accuracy of 88.1%, indicating a significant improvement compared to other methods used in the study. The IPSO-SVM model showed superior convergence speed and accuracy compared to other optimization algorithms, such as Group Swarm Optimization Algorithm (GWO) and Whale Optimization Algorithm (WOA).

It is highlighted that data-driven models, such as SVM, are advantageous in identifying failure patterns and system behaviors, especially in complex and nonlinear systems.

These findings underline the effectiveness of the IPSO-SVM model in failure prediction and its potential for applications in critical systems development.

Based on the findings and the structure of the paper, it appears to have strong potential for publication.

Reviewer 2 Report

Comments and Suggestions for Authors

 

1. Please explain which type of accelerometers  you used for collecting data. (sensitivity...)

2. Is it possible use microphone and accelerometers together for collecting data.

3. Is it possible use this classification SVM analysis for another system or only on this geometry.

4. How is this system usable for artificial intelligence, for recognition another fault.

5. For improve quality please add this literature in paper:

Application of SVM models for classification of welded joints
D Marić, M Duspara, T Šolić, I Samardžić
Tehnički vjesnik 26 (2), 533-538

Research on the Effect of Load and Rotation Speed on Resistance to Combined Wear of Stainless Steels Using ANOVA Analysis
G Rozing, M Duspara, B Dudic, B Savkovic
Materials 16 (12), 4284

Comments on the Quality of English Language

Paper is good written, and I suggest that paper could be published

Reviewer 3 Report

Comments and Suggestions for Authors

Comments to the Authors of the manuscript Applsci-3282929 are as follows:

1.      According to the content of the manuscript, the Authors used experimental data. However, it is not clear how the experimental data were generated, how they were used, and there is no presentation of them in the manuscript (in figures or tables).

2.      In the introduction, the review of literature lacks the references representing the relevant state-of-the-art.

3.      Conclusions are too general and poor in terms of scientific significance. In particular, the conclusion must be expanded and quantified based on the obtained results.

4.      The list of references contains 14 references and it seems that 13 of them belong to authors of the same nationality. The list of references must be relevant, state-of-the-art, international and extensive.

5.      In Table 2, the functions marked with the numbers 1, 2, 7 and 9 appear. What about the other functions marked with the numbers 3, 4, 5, 6 and 8?

6.      No constraints are defined for the optimization mathematical model.

7.      The first affiliation should start with a capital letter.

8.      Keywords should be listed in alphabetical order and followed by the corresponding abbreviation (if any) in parentheses.

9.      At the end of the introduction, there is a sentence containing the following text: “…identifying the optimal experimental scheme for small samples.” What does the term "optimal experimental scheme" represent here?

1 .      On Page 3 of 10, in the second paragraph, there is the text “…zation capability[11-Error! Reference source not found.].” which should be corrected.

   .      On Page 3 of 10, there is a jump from the citation of Reference [11] to the citation of References [13,14]. Specifically, there is no citation of Reference [12].

1.      Section 3 begins with the following sentence “To obtain more efficient experimental data, simulations were first performed using Ansys and Is-Dyna prior to conducting high-impact experiments.” The software called "Ansys" is based on the finite element method and there is no mention of how it was applied. The same applies to the software “Is-Dyna”.

Reviewer 4 Report

Comments and Suggestions for Authors

The paper presents IPSO-SVM, an SVM model based on particle swarm optimization, which is to be applied in failure detection with small datasets.

Strong points:

• The proposed Model performs better compared to the previous version of PSO-based Model in terms of optimizing four benchmark functions. It is also shown that the fault prediction accuracy of the IPSO-SVM model reaches 88.1%, which is higher than other PSO based models (81.1%)
• The Model works for classification with a small sample size.
• Explained why SMV was chosen for the selected problem.

Weak Points:

• What is IPSO? It is never defined.
• There is a citation error in line 97
• The position updating is not shown (in equation 3 or 4)
• More explanation about damaged data in the context of data generation is needed.
• Just typing ‘Experimental Peak Acceleration’, I could generate Table 1 and the related descriptions. If the authors did the same thing or used the table from somewhere else, they had to cite or mention it.
• IPSO is once compared with regular PSO then again with GWO and WOA. Why? - not clear. Did you intend to use GWO-SVM and WOA-SVM for comparative analysis?
• Overall, there is a lack of experimental analysis and comparative analysis with other similar application models.
• Overall, technical contribution is limited.