Research on Yield Prediction Technology for Aerospace Engine Production Lines Based on Convolutional Neural Networks-Improved Support Vector Regression

Round 1

Reviewer 1 Report

This manuscript presents the yield prediction technology of an aerospace engine production line Based on CNN-ISVR.

I have several comments as follows,

- Please separate the related works, and present more about the most related work to yours.

- All tables and figures should be reformated.

- All figures should be explained in their captions.

- Please edit mathematics format

- The contribution should be presented more clearly and focus on your contribution more.

- Please do not include much of the fundamentals of ML methods such as SVM, CNN (fig 2)

- For easier attention, I recommend an overall structure should be added that includes your contribution to the block.

- Please add more comparisons to other regression ML algorithms

N/A

Author Response

Manuscript ID:Machines-2558211

Title: Research on Yield Prediction Technology of Aerospace Engine Production Line Based on CNN-ISVR

Author name: HongJun Liu

Reply to the comments

Dear editor

We thank the referees for their careful reading of our paper. We have carefully considered the comments and have revised the manuscript accordingly. Please find below our responses to the reviewers’ comments. All the revisions have been addressed in the Reply and highlighted in the manuscript with yellow background. We hope the revised manuscript can be considered acceptable.

Reply to the comments of Reviewer 1

Comments:

(1) Please separate the related works, and present more about the most related work to yours.

Reply:

We thank you for reminding us this important point. According to your suggestion, we have separated the relevant works and presented the most pertinent examples in our work. The modified content can be found on the first and second pages of the manuscript.

Comments:

(2) All tables and figures should be reformated.

Reply:

We thank you for your careful reading of our paper and providing us with some keen scientific insight. According to your suggestion, we have formatted all the tables and figures, accurately incorporating them into the manuscript.

Comments:

(3) All figures should be explained in their captions.

Reply:

We thank you for reminding us this useful point. According to your suggestion, we have provided explanations for all the figures in their respective captions, for instance, Table 1 on the third page of the manuscript.

Comments:

(4) Please edit mathematics format.

Reply:

We thank you for reminding us this useful point. According to your suggestion, in the manuscript, we have made modifications to the formatting of all mathematics format. As an example, the equation (1) on the fourth page of the manuscript has been meticulously edited.

Comments:

(5) The contribution should be presented more clearly and focus on your contribution more. and for easier attention, I recommend an overall structure should be added that includes your contribution to the block.

Reply:

We appreciate the meaningful and scientific opinions you have provided. According to your suggestion, in our analysis of the author's contributions, we diligently expounded upon my own contributions to this article, as elaborated on the nineteenth page of the manuscript.

Comments:

(6) Please do not include much of the fundamentals of ML methods such as SVM, CNN (fig 2)

Reply:

We thank you for raising this useful point. According to your suggestion, we have removed unnecessary machine learning fundamentals, and the Figure 2 that was originally present in the manuscript has been deleted.

Comments:

(7) Please add more comparisons to other regression ML algorithms

Reply:

We thank you for raising this useful point. According to your suggestion, we have added other machine learning algorithms including linear regression, ridge regression, and random forest. We compared these algorithms with the proposed CNN-ISVR model and the comparison of the prediction results between CNN-ISVR and other machine learning algorithms is presented in pages 15 to 19 of the manuscript.

Thank you so much for providing meaningful and scientific feedback on our manuscript. We truly appreciate your input. Your suggestions have been of great significance to our research and have helped us improve the manuscript further. If you have any additional comments or suggestions regarding the revised manuscript, please let us know. Thank you once again for your thorough review and valuable insights.

Best regards,

HongJun Liu

Author Response File: Author Response.pdf

Reviewer 2 Report

Aerospace engine manufacturing companies have always prioritized optimizing production line yield. The accuracy of predicting production line yield directly relies on how well the production line is designed and how product production is arranged. In response to this challenge, a novel method known as Convolutional Neural Networks-Improved Support Vector Regression (CNN-ISVR) has been introduced to predict the production line yield of aerospace engines. However the aerospace engine sector is not well addressed, for instance no mention to the works of people working with companies, as Sensors 2023, 23(12), 5694; https://doi.org/10.3390/s23125694 and others by Sastoque, in MDPI even, giving how to deal with projects and TRL. 2 or 3 works at least.

Yield prediction…..I think you have no special interest for engine makers, at least you must say with which companies do you work, or at least the source of your data. Many machine learning works are not really interesting because data used as source are meaningless. You can look for research centres about engines, like CFAA in Spain, or Nottingham university. They published many realistic works.

The CNN-ISVR method divides the factors influencing production line yield into two categories: production cycle and real-time status information of the production line. It then analyses the key feature factors within these categories. To address the problem of poor prediction performance in traditional SVR models due to subjective selection of kernel function parameters, an improved SVR model is proposed. This improved approach combines the elite strategy genetic algorithm with hyperparameter optimization methods based on grid search and cross-validation to determine the best penalty factor C and kernel function width (σ) for the RBF kernel function.

aerospace engine manufacturing: which company is the main soruce of data?

As China's aerospace industry continues to flourish. Eliminate this kind of siperflous sentences. Flourish? This is not a magazine to make ads of one country. I do not see a clear link with one specific industry.

CNCC and other machine learning techniques are Ok, however why did you use that specific algorithm? A.Bustillo explained the problem of data poor in some cases, in https://doi.org/10.1016/j.jmsy.2018.06.004 because the poor data are a main drawback, and boosting ensembles can help. Please discuss as A.Bustillo did.

Eliminate REF 8. Refs are of the same country and Asia, when engines are mostly produces in other countries.

The hybrid model used in this paper combines the advantages of CNN and SVR 57 models: but whar are they?. Conclsuions are no very well-written.

You can reduce a Little the model explanation and figures: go to the real new contributions.

Genetic algorithm(GA), conceived by John Holland in the 1970: this is not necessary, is it an open discusión.

Author Response

Manuscript ID ： Machines-2558211

Title: Research on Yield Prediction Technology of Aerospace Engine Production Line Based on CNN-ISVR

Author name: HongJun Liu

Reply to the comments

Dear editor

We thank the referees for their careful reading of our paper. We have carefully considered the comments and have revised the manuscript accordingly. Please find below our responses to the reviewers’ comments. All the revisions have been addressed in the Reply and highlighted in the manuscript with yellow background. We hope the revised manuscript can be considered acceptable.

We thank the referees for their careful reading of our paper. We have carefully considered the comments and have revised the manuscript accordingly. Please find below our responses to the reviewers’ comments. All the revisions have been addressed in the Reply and highlighted in the manuscript with yellow background. We hope the revised manuscript can be considered acceptable.

Reply to the comments of Reviewer 2

Comments:

(1) However the aerospace engine sector is not well addressed, for instance no mention to the works of people working with companies, as Sensors 2023, 23(12), 5694; https://doi.org/10.3390/s23125694 and others by Sastoque, in MDPI even, giving how to deal with projects and TRL. 2 or 3 works at least.

Reply:

We thank you for reminding us this important point. According to your suggestion, Upon reviewing the manuscript from the Sensors journal you sent us, we have made modifications to the abstract provided on the first page. The revised abstract is as follows: Improving the prediction accuracy of aerospace engine production line yield is of significant importance for enhancing production efficiency and optimizing production scheduling in enterprises. Within the engine manufacturing process, the proficiency of personnel operations affects the efficiency of the production line. However, quantifying and obtaining the proficiency of personnel operations is challenging. Therefore, this study adopts an artificial timing method to collect the working hours at each production workstation as one of the input features for the SVR model. This approach allows the influence of personnel to be reflected in the working hours at the production workstations.

Comments:

(2) I think you have no special interest for engine makers, at least you must say with which companies do you work, or at least the source of your data. Many machine learning works are not really interesting because data used as source are meaningless. You can look for research centres about engines, like CFAA in Spain, or Nottingham university. They published many realistic works.

Reply:

We thank you for your careful reading of our paper and providing us with some keen scientific insight. According to your suggestion, the production data used in this study originates from an assembly line of a certain engine model. Due to the sensitivity of the data, it cannot be displayed. The purpose of this research is to develop a predictive algorithm for assembly lines of similar products. Necessary data validation was conducted, and we collected production cycle times and other process data from various production stations on the engine assembly line using manual stopwatch methods. These data were utilized as input features for regression prediction.

Comments:

(3) As China's aerospace industry continues to flourish. Eliminate this kind of flourish sentences. Flourish? This is not a magazine to make ads of one country. I do not see a clear link with one specific industry.

Reply:

We thank you for reminding us this useful point. According to your suggestion, I apologize for any inconvenience caused. We will remove the sentence from the paper. Thank you for your correction. We are not advertising a specific country's magazine, so we have already deleted the sentence.

Comments:

(4) CNCC and other machine learning techniques are Ok, however why did you use that specific algorithm? A.Bustillo explained the problem of data poor in some cases, in https://doi.org/10.1016/j.jmsy.2018.06.004 because the poor data are a main drawback, and boosting ensembles can help. Please discuss as A.Bustillo did.

Reply:

We thank you for reminding us this useful point. According to your suggestion, we have addressed the issue in the second section of our manuscript. The prediction method based on neural networks for production forecasting aims to optimize the empirical risk and unfortunately cannot avoid the problem of converging to local minima [11]. On the other hand, the prediction method based on SVM can reduce structural risk [12]. By introducing a regularization term, it effectively solves the issue of overfitting. However, it requires quadratic programming to partition the separating hyperplane. Constructing a prediction model using this method on a large-scale aerospace engine production process data significantly increases computational complexity and consumes a considerable amount of processing time, thereby increasing training difficulty. This is not economically efficient in practical production. Based on the limitations of using traditional single machine learning algorithms for yield prediction, this paper proposes a method for engine yield prediction based on CNN and ISVR by combining the neural network algorithm with the support vector machine regression algorithm [15-17]. Additionally, an adaptive feature extraction is performed using the shallow non-dimensional reduction CNN designed in this paper, eliminating the need for feature pre-extraction on production data and effectively overcoming the limitations of SVR.

Comments:

(5) The hybrid model used in this paper combines the advantages of CNN and SVR 57 models: but whar are they?. Conclsuions are no very well-written, and You can reduce a Little the model explanation and figures: go to the real new contributions.

Reply:

We appreciate the meaningful and scientific opinions you have provided. According to your suggestion, we have improved the writing of our conclusion to properly reflect its contribution, which is now presented on page 19 of the manuscript.

Comments:

(6) Eliminate REF 8. Refs are of the same country and Asia, when engines are mostly produces in other countries.

Reply:

We thank you for raising this useful point. According to your suggestion, thank you for your feedback. We have already removed reference 8 as you suggested.

Comments:

  (7) Genetic algorithm(GA), conceived by John Holland in the 1970: this is not necessary, is it an open discusion.

Reply:

We thank you for raising this useful point. According to your suggestion, “Genetic algorithm(GA), conceived by John Holland in the 1970”，the sentence you provided is unnecessary, and we have deleted it.

Thank you so much for providing meaningful and scientific feedback on our manuscript. We truly appreciate your input. Your suggestions have been of great significance to our research and have helped us improve the manuscript further. If you have any additional comments or suggestions regarding the revised manuscript, please let us know. Thank you once again for your thorough review and valuable insights.

Best regards,

HongJun Liu

Author Response File: Author Response.pdf

Reviewer 3 Report

The article presents a novel approach for predicting aerospace engine assembly line production using a hybrid CNN-ISVR model. By combining convolutional neural networks (CNN) and support vector regression (SVR), the research seeks to overcome the limitations of traditional SVR models and offer a more accurate prediction mechanism. They present a hybrid model that merges the strengths of CNNs and SVRs. The model effectively quantifies factors affecting production volume, such as production pace and equipment status. Improved SVR's predictive performance by optimizing kernel function parameters using elite strategy genetic algorithms combined with grid search and cross-validation. The hybrid model exhibited superior generalization and prediction accuracy with a determination coefficient above 0.92. Given the critical nature of aerospace engine production and the need for accurate forecasting in manufacturing, the model's approach is highly relevant. Efficient production predictions can lead to cost savings, better resource allocation, and timely deliveries. The integration of CNN with ISVR for this specific application appears innovative. The methodologies employed are well-established, and the use of genetic algorithms for hyperparameter optimization ensures model fine-tuning. Addresses a real-world challenge in aerospace engine assembly line production forecasting.

Some inconsistencies and formatting issues in the content might make it slightly difficult for readers to follow. The article mentions that certain factors, like material supply efficiency and operator skill level, which influence production, were not explored. This indicates potential areas that could be enhanced in the research.

The article introduces a promising and innovative approach to aerospace engine production forecasting, showcasing significant scientific depth and practical application. While the presentation could be refined and some aspects expanded upon, the research's overall merit is high and offers valuable insights for the aerospace and manufacturing sectors.

Line 54 (numbering starts at page 18): "is proposed. This method combines elite strategy geneticvalgorithms with hyperparame-". I suppose it should be "genetic algorithms" ?

5.4.1 lines are not numbered: Instead of the word "flowchart" where the process is described in words, "outline" or "procedure" would be appropriate replacements.

Other minor changes could include: 
Original: "Selecting a production line of a certain type of aerospace engine as the research object..."
Suggestion: "Selecting a specific type of aerospace engine production line for our research..."
Original: "The selected iterative condition for this section is evolutionary generation..."
Suggestion: "For this section, the chosen criterion for iteration is the evolutionary generation..."
Original: "The experiment detailed in this text was based on the Windows 11 operating system."
Suggestion: "The experiment conducted for this study used the Windows 11 operating system."
Original: "PyCharm served as the development environment, and Python 3.6 was used to implement the experiment."
Suggestion: While the sentence is technically correct, it could be clearer: "We used PyCharm as our development environment and implemented the experiment using Python 3.6."

Author Response

Manuscript ID:Machines-2558211

Title: Research on Yield Prediction Technology of Aerospace Engine Production Line Based on CNN-ISVR

Author name: HongJun Liu

Reply to the comments

Dear editor

We thank the referees for their careful reading of our paper. We have carefully considered the comments and have revised the manuscript accordingly. Please find below our responses to the reviewers’ comments. All the revisions have been addressed in the Reply and highlighted in the manuscript with yellow background. We hope the revised manuscript can be considered acceptable.

Reply to the comments of Reviewer 3

Comments:

(1) Some inconsistencies and formatting issues in the content might make it slightly difficult for readers to follow.

Reply:

We thank you for reminding us this important point. According to your suggestion, we have already made amendments to the inconsistencies and formatting issues in the content of the article.

Comments:

(2) Comments on the Quality of English Language.

Reply:

We thank you for your careful reading of our paper and providing us with some keen scientific insight. According to your suggestion, we have implemented all of your suggested English revisions, with the designated pages – 11, 13, and 14 – marked using yellow highlighting. Thanks for rectifying my grammatical mistakes and inappropriate words in the thesis, I will keep improving my expressions with English, Try my best to erase the language barrier.

Thank you so much for providing meaningful and scientific feedback on our manuscript. We truly appreciate your input. Your suggestions have been of great significance to our research and have helped us improve the manuscript further. If you have any additional comments or suggestions regarding the revised manuscript, please let us know. Thank you once again for your thorough review and valuable insights.

Best regards,

HongJun Liu

Author Response File: Author Response.docx

Reviewer 4 Report

The authors have made ambitious and commendable efforts to propose a CNN-ISVR model with excellent potential applicability to the aerospace sector. However, the review is suggesting the following :

1. Abstract: what is the kernel function width symbol? Please remove all strange symbols and equations from the abstract. 

2. Abstract: Can authors write R² and not R^2?

3. Line 3, page 1: Authors should provide the reference or citation to the claim "Engine production yield is clearly related to factors...

4. Since authors used small letters before the abbreviation, please change the following: 

Support Vector Regression> support vector regression (SVR)

Convolutional Neural Network (CNN)>> convolutional neural network (CNN)

see sections 3.1 and 3.2 and the introduction. 

5. Figure 1 was mentioned on page 5 but the figure was located on page 7. Please locate figures on the same or near page.

6. Some abbreviations were created multiple times. Examples are:

Cross-Validation (CV) on page 10 and page 11. please use cross-validation (CV) on page 10 and subsequently use CV

7. If Table 5 is the parameters of the CNN, what is then the NOVELTY of the paper? 

8. Table 3: please change Mse to MSE

9. Please provide the F1-Score result in Table 7

10. Create a Table of abbreviations and place them after the conclusion. 

Minimal corrections regarding abbreviations and usage style are required. 

Author Response

Manuscript ID:Machines-2558211

Title: Research on Yield Prediction Technology of Aerospace Engine Production Line Based on CNN-ISVR

Author name: HongJun Liu

Reply to the comments

Dear editor

We thank the referees for their careful reading of our paper. We have carefully considered the comments and have revised the manuscript accordingly. Please find below our responses to the reviewers’ comments. All the revisions have been addressed in the Reply and highlighted in the manuscript with yellow background. We hope the revised manuscript can be considered acceptable.

Reply to the comments of Reviewer 4

Comments:

(1) Abstract: what is the kernel function width symbol? Please remove all strange symbols and equations from the abstract, and Can authors write R2 and not R^2?

Reply:

We thank you for reminding us this important point. According to your suggestion, we have removed all peculiar symbols from the abstract and replaced "goodness of fit" with "R^2" to enhance clarity.

Comments:

(2) Line 3, page 1: Authors should provide the reference or citation to the claim "Engine production yield is clearly related to factors...

Reply:

We thank you for your careful reading of our paper and providing us with some keen scientific insight. According to your suggestion, we have already provided and referenced the literature [15] on the third page of our manuscript, Kang, Z.; Catal, C.; Tekinerdogan, B. Machine learning applications in production lines: A systematic literature review. J. Computers & Industrial Engineering. 2020, 149, 106773.

Comments:

(3) Since authors used small letters before the abbreviation, please change the following: Support Vector Regression> support vector regression(SVR)Convolutional Neural Network (CNN)>> convolutional neural network (CNN)see sections 3.1 and 3.2 and the introduction. 

Reply:

We thank you for reminding us this useful point. According to your suggestion, we have already capitalized the lowercase letters before abbreviations in the manuscript, such as Convolutional Neural Networks (CNN) and Improved Support Vector Regression (ISVR) on page 2.

Comments:

(4) Figure 1 was mentioned on page 5 but the figure was located on page 7. Please locate figures on the same or near page.

Reply:

We thank you for reminding us this useful point. According to your suggestion, we have already moved Figure 1 to the position on the fifth page so that readers can find Figure 1 on the same page.

Comments:

(5) Some abbreviations were created multiple times. Examples are: Cross-Validation (CV) on page 10 and page 11. please use cross-validation (CV) on page 10 and subsequently use CV.

Reply: We appreciate the meaningful and scientific opinions you have provided. According to your suggestion, we have already used the abbreviation "CV" after using Cross-Validation (CV), for example, on page 9 of the manuscript, we use "CV" as a shorthand for cross-validation.

Comments:

(6) If Table 5 is the parameters of the CNN, what is then the NOVELTY of the paper? 

Reply:

We thank you for raising this useful point. According to your suggestion, the novelty of this paper lies in the combination of Convolutional Neural Networks (CNN) and Improved Support Vector Regression (ISVR) algorithms to address the limitations of using traditional single machine learning algorithms for yield prediction. Additionally, a shallow non-dimensional reduction CNN is designed in this paper to perform adaptive feature extraction, eliminating the need for feature pre-extraction on production data and overcoming the limitations of Support Vector Regression. The hybrid model used in this paper combines the advantages of both models. Compared to using CNN or Support Vector Regression alone, the complexity of the model does not increase significantly, but it effectively improves the accuracy of the prediction results.

Comments:

(7) Table 3: please change Mse to MSE.

Reply:

We thank you for raising this useful point. According to your suggestion, we have modified the Mse in Table 3 to MSE.

Comments:

(8) Please provide the F1-Score result in Table 7.

Reply:

We thank you for raising this useful point. According to your suggestion, the F1-Score is commonly used to evaluate the performance of classification models. However, the SVR, LR, RG, RF, and CNN-ISVR models used in this study are typically employed for regression tasks rather than classification tasks. Therefore, it would not be normal to use the F1-Score to assess the performance of the CNN-SVR model. For regression tasks, common evaluation metrics include Mean Squared Error (MSE), Mean Absolute Error (MAE), and R-squared. These metrics are utilized to measure the error and fit between predicted values and actual values.

Comments:

(9) Create a Table of abbreviations and place them after the conclusion. 

Reply:

We thank you for raising this useful point. According to your suggestion, thank you for your valuable feedback. We have incorporated your suggestions and have created an abbreviation table, which can be found on page 20 of the manuscript. We appreciate your contribution and support in enhancing the academic quality of our work.

Thank you so much for providing meaningful and scientific feedback on our manuscript. We truly appreciate your input. Your suggestions have been of great significance to our research and have helped us improve the manuscript further. If you have any additional comments or suggestions regarding the revised manuscript, please let us know. Thank you once again for your thorough review and valuable insights.

Best regards,

HongJun Liu

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Thanks for your efforts, this time your manuscript was improved.

However, all my suggestions were not edited satisfy enough, I recommend the authors edit your manuscript carefully and note your change with respect to your reply in the edited manuscript for easy review.

N/A

Author Response

Manuscript ID: Machines-2558211

Title: Research on Yield Prediction Technology of Aerospace Engine Production Line Based on CNN-ISVR

Author name: Hong Jun Liu

Reply to the comments

Dear editor

We thank the referees for their careful reading of our paper. We have carefully considered the comments and have revised the manuscript accordingly. Please find below our responses to the reviewers’ comments. All the revisions have been addressed in the Reply and highlighted in the manuscript with yellow background. We hope the revised manuscript can be considered acceptable.

Reply to the comments of Reviewer 1

Comments:

(1) Please separate the related works, and present more about the most related work to yours.

Reply:

We thank you for reminding us this important point. According to your suggestion, we have separated the relevant works and presented the most pertinent examples in our work. The highlighted section on the first page of the manuscript showcases the modified content.

Comments:

(2) All tables and figures should be reformated.

Reply:

We thank you for your careful reading of our paper and providing us with some keen scientific insight. According to your suggestion, we have formatted all the tables and figures, accurately incorporating them into the manuscript. To illustrate, let's consider Table 1 on page 3 of the manuscript and Figure 1 on page 5.

Comments:

(3) All figures should be explained in their captions.

Reply:

We thank you for reminding us this useful point. According to your suggestion, we have provided explanations for all the figures in their respective captions, to provide an example, let's showcase Table 2 on page 12 of the manuscript and Figure 3.

Comments:

(4) Please edit mathematics format.

Reply:

We thank you for reminding us this useful point. According to your suggestion, in the manuscript, we have made modifications to the formatting of all mathematics format. As an example, the mathematic (1) on the fourth page of the manuscript has been meticulously edited.

Comments:

(5) The contribution should be presented more clearly and focus on your contribution more.

Reply:

We appreciate the meaningful and scientific opinions you have provided. According to your suggestion, in our analysis of the author's contributions, we diligently expounded upon my own contributions to this article, as elaborated on the 21th page of the manuscript.

Comments:

(6) Please do not include much of the fundamentals of ML methods such as SVM, CNN (fig 2)

Reply:

We thank you for raising this useful point. According to your suggestion, we have removed unnecessary machine learning fundamentals, and I have deleted the figure that introduced the basic knowledge of CNN from the original manuscript.

Comments:

(7) Please add more comparisons to other regression ML algorithms

Reply:

We thank you for raising this useful point. According to your suggestion, we have added other machine learning algorithms including Linear Regression (LR), Ridge Regression (RG), and Random Forest (RF). The parameter table for these several machine learning algorithms is located on page 17 of the manuscript, specifically in Table 7. We compared these algorithms with the proposed CNN-ISVR model and the comparison of the prediction results between CNN-ISVR and other machine learning algorithms is presented in pages 17 to 20 of the manuscript.

Comments:

(8) For easier attention, I recommend an overall structure should be added that includes your contribution to the block.

Reply:

We thank you for raising this useful point. According to your suggestion, dear reviewer, I saw you commented about recommending an overall structure, I was so confused about that. What exactly did you referring to? I have incorporated a comprehensive structure, titled "Author Contributions," which is reflected on page 21 of the manuscript.

(Chang et al., 2013)

Thank you so much for providing meaningful and scientific feedback on our manuscript. We truly appreciate your input. Your suggestions have been of great significance to our research and have helped us improve the manuscript further. If you have any additional comments or suggestions regarding the revised manuscript, please let us know. Thank you once again for your thorough review and valuable insights.

Best regards,

HongJun Liu

Author Response File: Author Response.pdf

Reviewer 2 Report

Authors did not follow previous review, the source of data is still missed and data can be not realistic. We understand the confidentiality, but in some cases it is weird, data from production must be identified as realistic.

Is it important to say that data are from aero engines productions?

Please read carefully previous review and do point by point a complete improvement of the work.

A.Bustillo worked with data not perfect, with the approach of boosting ensembles. Can you applied them?. Connectivity is key in production, and in aero engines people are following the ideas of edge computing, see Sensors 2023, 23(12), 5694; https://doi.org/10.3390/s23125694 how to get information is more important than algorithms.

Please take you time to improve the text, main ideas are OK, but some aspects needs clarification.

Special issue is about. Electrical Machines and Drives  where are these kind of devices?

CNN-ISVR hybrid model is shown to outperform other models, which ones did you consider?

eliminate. "With the advent of advanced technologies such as big data, the Internet of Things, and cloud computing, aerospace engine companies are committed to digitalization and intelligent transformation and upgrading. " many sentences are commercial ones.

Author Response

Manuscript ID: Machines-2558211

Title: Research on Yield Prediction Technology of Aerospace Engine Production Line Based on CNN-ISVR

Author name: Hong Jun Liu

Reply to the comments

Dear editor

We thank the referees for their careful reading of our paper. We have carefully considered the comments and have revised the manuscript accordingly. Please find below our responses to the reviewers’ comments. All the revisions have been addressed in the Reply and highlighted in the manuscript with yellow background. We hope the revised manuscript can be considered acceptable.

Reply to the comments of Reviewer 2

Comments:

(1) A. Bustillo worked with data not perfect, with the approach of boosting ensembles. Can you applied them?. Connectivity is key in production, and in aero engines people are following the ideas of edge computing, see Sensors 2023, 23(12), 5694; https://doi.org/10.3390/s23125694 how to get information is more important than algorithms.

Reply:

We thank you for reminding us this important point. According to your suggestion, thank you for providing scientific modifications to our article. Your comments are accurate. Connectivity is crucial in production. We have referenced and cited A.Bustillo's paper and utilized integrated enhancement to handle the dataset. We have also read the article you recommended from Sensors, which has provided us with inspiration on how to obtain production data information, which is more important than algorithms. The modified content is highlighted in yellow on pages 13 and 14 of the manuscript.

Comments:

(2) Special issue is about. Electrical Machines and Drives  where are these kind of devices?

Reply:

We thank you for your careful reading of our paper and providing us with some keen scientific insight. According to your suggestion, I sincerely apologize for the confusion. Although the special issue is focused on Electrical Machines and Drives, I am unaware of the reason why your article was included in this special issue. From the initial submission, your manuscript has been categorized into this special issue. I recommend contacting the editor for clarification regarding the special issue.

Comments:

(3) CNN-ISVR hybrid model is shown to outperform other models, which ones did you consider?

Reply:

We thank you for reminding us this useful point. According to your suggestion, in our article, we first compared our proposed model with traditional regression algorithms such as linear regression(LR), ridge regression(RG), and random forest(RF). Then, we further compared the predictive performance of our model with CNN, SVR, and CNN-SVR to demonstrate its superiority and generalizability. The comparison between the proposed algorithm and other algorithms is presented in the content of section 5.7 on page 17 of the manuscript.

Comments:

(4) eliminate. "With the advent of advanced technologies such as big data, the Internet of Things, and cloud computing, aerospace engine companies are committed to digitalization and intelligent transformation and upgrading. " many sentences are commercial ones.

Reply:

We thank you for reminding us this useful point. According to your suggestion, thank you for your suggestions. Your advice is highly accurate. I have removed the commercial content you mentioned from the manuscript. Thank you for providing scientific insights to our article!

Thank you so much for providing meaningful and scientific feedback on our manuscript. We truly appreciate your input. Your suggestions have been of great significance to our research and have helped us improve the manuscript further. If you have any additional comments or suggestions regarding the revised manuscript, please let us know. Thank you once again for your thorough review and valuable insights.

Best regards,

Author Response File: Author Response.pdf

Round 3

Reviewer 1 Report

Thank you for your efforts,

Please read and smooth your manuscript, I have no further comments,

bests,

N/A

Reviewer 2 Report

Ok