Summarization of Remaining Life Prediction Methods for Special Power Plants

Round 1

Reviewer 1 Report

This paper mainly summarizes the research status of the residual life prediction method of the special power plant and analyzes the problems at the present stage. Distinct categories of methods are presented. A more detailed comparison among methods with evaluation parameters may enrich the content. In addition, the writing and formatting may be improved.

In the dynamic model-driven RUL prediction methods part, it would be beneficial if the author compares the material deformation characteristic, contact and impact characteristic, friction and wear form and feature structure optimization characteristic in more details, i.e., the computation complexity, applicable working condition, and limit of each one.

In the dynamic data-driven RUL prediction methods part, convolutional neural network, deep belief network, recurrent neural network, transfer learning are listed and the fundamental working principles are explained. It would be more beneficial to the reader if a more detailed introduction of how these methods are utilized by different works in the RUL prediction and a comparison of the advantages and disadvantages of these methods. Also, rearrange these works as a consequence of method improvement will be more clear in logic.

In the background introduction, a more detailed review of how the degradation of a special power plant becomes more complex compared to the common one will be helpful to address the importance of this work.

Minors:

1.     References for physical health factors need to be cited on Line 227

2.     References for the third fusion need to be cited on Line 284

3.     Indices in Fig. 5 and Fig. 11 are hard to read

Typos:

1.     Line 52. “Is not ..”

2.     Line 209. “domain, the time..”

3.     Line 231. “[74],”

4.     Line 311. “[25、101-103]”

5.     Line 460. “dynamic data-driven”

The writing need to be improved.

Author Response

Dear Reviewer:

Thank you for your decision and constructive comments on my manuscript. We have carefully considered the suggestion of Reviewer and make some changes. We have tried our best to improve and made some changes in the manuscript.

The yellow part that has been revised according to your comments. Revision notes, point-to-point, are given as follows:

(1)In the dynamic model-driven RUL prediction methods part, it would be beneficial if the author compares the material deformation characteristic, contact and impact characteristic, friction and wear form and feature structure optimization characteristic in more details, i.e., the computation complexity, applicable working condition, and limit of each one.

The author’s answer：

Based on the reviewer's review comments, certain modifications have been made to this section of the manuscript and relevant descriptions have been added, making the content of the manuscript more rich and logical.

(2)In the dynamic data-driven RUL prediction methods part, convolutional neural network, deep belief network, recurrent neural network, transfer learning are listed and the fundamental working principles are explained.It would be more beneficial to the reader if a more detailed introduction of how these methods are utilized by different works in the RUL prediction and a comparison of the advantages and disadvantages of these methods. Also, rearrange these works as a consequence of method improvement will be more clear in logic.

The author’s answer：

Based on the reviewer's review comments, certain modifications have been made to this section of the manuscript and relevant descriptions have been added, making the content of the manuscript more rich and logical.

(3)In the background introduction, a more detailed review of how the degradation of a special power plant becomes more complex compared to the common one will be helpful to address the importance of this work.

The author’s answer：

Based on the reviewer's review comments, certain modifications have been made to this section of the manuscript, emphasizing the differences in mechanical performance degradation between special power units and other structures, highlighting the complexity of their degradation process analysis research, and further emphasizing the significance of the manuscript's research work.

(4)Minors:

1. References for physical health factors need to be cited on Line 227
2. References for the third fusion need to be cited on Line 284
3. Indices in Fig. 5 and Fig. 11 are hard to read

The author’s answer：

Based on the reviewer's revision comments, the above two parts have been cited and the index of Figures 5 to 11 has been revised.

(5)Typos:

1. Line 52. “Is not ..”
2. Line 209. “domain, the time..”
3. Line 231. “[74],”
4. Line 311. “[25、101-103]”
5. Line 460. “dynamic data-driven”

The author’s answer：

Based on the reviewer's modification comments, the above five content sections have been modified one by one.

(6)Comments on the Quality of English Language&The writing need to be improved.

The author’s answer：

We have contacted the editorial department of academic journals and selected a professional English editing agency to revise the overall language quality of the article.

Once again, thank you to the reviewers for their hard work during the manuscript review process. Due to the limitations of the author's academic level and ability, new issues may arise during the manuscript revision process. Please contact us at any time, and we will humbly accept your opinions and make timely revisions.

 

 

Yours Sincerely

Mr Li Chi

The second author of the manuscript

 

Author Response File: Author Response.docx

Reviewer 2 Report

1. The work explores various approaches, such as deep learning, principal component analysis, and recurrent neural networks, to predict RUL based on feature extraction and health factor construction. These methods show promising results in achieving accurate RUL prediction. However, the work acknowledges certain challenges, such as difficulty in hyperparameter tuning and network design, limited generalization of models, and the need for large training datasets.

2. The paper emphasizes the importance of developing RUL prediction methods with interpretability and generalization. It would be better to combine the domain knowledge in order to get and lead to more understandable and robust models.

3.  For more complex and diverse mechanical equipment, please explore methods that improve model generalization. Hopefully, this could reduce overfitting, and use the regularization techniques to improve the model's ability to handle different operating conditions.

4. Please explore the integration of sensor data into the models to enhance the accuracy of RUL predictions. It would provide valuable insights into the actual operating conditions of mechanical equipment and improve the models' ability to adapt to dynamic environments.

5. The conclusion emphasizes the need to combine dynamic degradation factors and stochastic degradation models to construct a multi-factor coupled dynamic model for RUL prediction. This approach aims to capture the complex and dynamic nature of degradation in special power plants, which is not explicitly mentioned in this article.

5. This paper emphasizes on the importance of developing RUL prediction methods with interpretability and generalization. This aligns with the earlier suggestion to focus on improving model generalization and incorporating domain knowledge, which can lead to more understandable and robust models.

It is suggested that this review would be continue to your research work in the future.

 

 

 

Author Response

Dear Reviewer:

Thank you for your decision and constructive comments on my manuscript. We have carefully considered the suggestion of Reviewer and make some changes. We have tried our best to improve and made some changes in the manuscript.

The yellow part that has been revised according to your comments. Revision notes, point-to-point, are given as follows:

(1)The work explores various approaches, such as deep learning, principal component analysis, and recurrent neural networks, to predict RUL based on feature extraction and health factor construction. These methods show promising results in achieving accurate RUL prediction. However, the work acknowledges certain challenges, such as difficulty in hyperparameter tuning and network design, limited generalization of models, and the need for large training datasets.

The author’s answer：

Thank you to the reviewers for their recognition and affirmation of the article's presentation work.

• The paper emphasizes the importance of developing RUL prediction methods with interpretability and generalization.It would be better to combine the domain knowledge in order to get and lead to more understandable and robust models.

The author’s answer：

Thank you to the reviewers for their recognition and affirmation of the article's presentation work.

• For more complex and diverse mechanical equipment, please explore methods that improve model generalization. Hopefully, this could reduce overfitting, and use the regularization techniques to improve the model's ability to handle different operating conditions.

The author’s answer：

Thank you to the reviewers for their recognition and affirmation of the article's presentation work.

• Please explore the integration of sensor data into the models to enhance the accuracy of RUL predictions. It would provide valuable insights into the actual operating conditions of mechanical equipment and improve the models' ability to adapt to dynamic environments.

The author’s answer：

All the authors strongly agree with the comments made by the reviewers. The reviewer emphasized the need to integrate the data collected by sensors into the established analytical model in order to improve the accuracy of RUL prediction on this issue. Section 2.3 of the manuscript provides a certain overview of this part of the research, but for the use of data collected by sensors, the data is still used as a medium to contain the health status information of mechanical equipment, extracted through existing analytical models for health management and remaining life prediction. In the process of model establishment, there is indeed little consideration. However, directly treating data as a part of the model, iterating and updating its model in real time, and predicting the RUL of mechanical equipment in real time from both data and model aspects is also one of the next research directions of the author of this article. However, due to the author's insufficient academic research and analysis abilities and limitations in academic perspective, no good research results have been extracted in this direction, and there has been no emphasis on analysis and discussion in the manuscript. We deeply apologize to the reviewers. If the reviewer has good research results in this area, can they provide some guidance to the author of this article? We would greatly appreciate your guidance.

• The conclusion emphasizes the need to combine dynamic degradation factors and stochastic degradation models to construct a multi-factor coupled dynamic model for RUL prediction. This approach aims to capture the complex and dynamic nature of degradation in special power plants, which is not explicitly mentioned in this article.

The author’s answer：

After carefully reviewing the manuscript, the author analyzed the main reasons for the revision suggestions proposed by the reviewers as follows: 1. During the writing process, the author lacked sufficient understanding of the overall content and structure of the article, resulting in partial ambiguity in the conclusion section and incomplete correspondence with the previous content, resulting in some lack of rigor. We have made modifications to the corresponding parts of the article and highlighted them in yellow. Please continue to provide valuable feedback from the reviewers.

• This paper emphasizes on the importance of developing RUL prediction methods with interpretability and generalization. This aligns with the earlier suggestion to focus on improving model generalization and incorporating domain knowledge, which can lead to more understandable and robust models.

The author’s answer：

Thank you to the reviewers for their recognition and affirmation of the article's presentation work.

 

Once again, thank you to the reviewers for their hard work during the manuscript review process. Due to the limitations of the author's academic level and ability, new issues may arise during the manuscript revision process. Please contact us at any time, and we will humbly accept your opinions and make timely revisions.

 

Yours Sincerely

Mr Li Chi

The second author of the manuscript

Author Response File: Author Response.docx