Method for Remaining Useful Life Prediction of Turbofan Engines Combining Adam Optimization-Based Self-Attention Mechanism with Temporal Convolutional Networks

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The abstract should be improved by including quantitive values to show the performance of the proposed method over literature approaches. Please, describe deeper the problem to be modeled and solved.

All contributions derived from this study should be included at the end of Section 1. The authors should include an additional section for Related Work, taking some material from Section 1.

 The captions in Figures and Tables are shallow, please include a more complete description of figures and tables. Besides, the police size used in the images should be similar in size and type as the current text.

The manuscript requires deep proofreading and the equations must be double-checked. Please, revise (13) to verify sign in the exponential functions.

The results should be discussed deeply in a subsection for "Discussions"

 

 

Comments on the Quality of English Language

The masnuscript should be carefully revised to improve clarity.

Author Response

Response to Reviewer 1 Comments

Dear Reviewer,

Thank you for reviewing our manuscript entitled "Method for Remaining Useful Life Prediction of Turbofan Engines Combining Adam Optimization Based Self-attention Mechanism with Temporal Convolutional Networks"[applsci-3093032]. We sincerely appreciate your valuable feedback and suggestions, as they have greatly contributed to enhancing the quality and impact of our research. We have thoroughly examined each of your comments and are pleased to inform you that we have successfully addressed the required revisions. In light of your feedback, we have implemented the following revisions:

Point 1:The abstract should be improved by including quantitive values to show the performance of the proposed method over literature approaches. Please, describe deeper the problem to be modeled and solved.

Response 1:Based on your suggestion, I have added quantitative values of experimental results in the abstract of the paper, along with a description of how these results compare with other literature. The added content is as follows (Line 12):

"In the experiments, the RMSE on four datasets are 11.50, 16.45, 11.62, and 15.47 respectively. These values indicate further reduction in errors compared to methods reported in other literature."

Point 2:All contributions derived from this study should be included at the end of Section 1. The authors should include an additional section for Related Work, taking some material from Section 1.

Response 2:Based on your suggestion, all contributions derived from this study have been supplemented at the end of Section 1. The added content is as follows:

"This approach accurately captures significant feature data to address the issue of low RUL prediction accuracy. The main contributions of this paper are as follows:” (Line 87)

"On the FD001~FD004 datasets, RMSE values obtained are 11.50, 16.45, 11.62, and 15.47 respectively. The corresponding Score values are 225.32, 1136.27, 259.79, and 1365.40. From these two evaluation metrics, it can be observed that the predictive accuracy of the proposed method in this paper is lower compared to results in other literature, indicating superior prediction performance." (Line 99)

Based on your suggestions, relevant sections and content have been added to Section 3 of the paper, where the information from Section 1 has been extracted and supplemented. The main additions are as follows: (Line 175)

“Research on RUL prediction of turbofan engines is a crucial topic in the field of flight safety. Accurately predicting the engine's remaining lifespan can assist airlines in optimizing maintenance schedules and reducing unscheduled maintenance incidents. Currently, machine learning and artificial intelligence technologies are predominantly used in the research of turbofan engine RUL prediction, supported by advanced data collection and sensor technologies.

In existing studies, Qiao et al. [26] discussed data and results from RUL prediction experiments on turbofan engines, finding that indirect mapping approaches yield better results, despite being more challenging and time-consuming to implement. Song et al. [27], considering the high dimensionality and volume of engine monitoring data, designed a bidirectional long short-term memory network for RUL prediction. Li et al. [28] respectively utilized CNN and LSTM networks to extract spatial features, perform data fusion, and employed the SAM method to obtain feature weights. Zhen et al. [29] proposed a TCN-attention model for oil well production prediction to overcome issues with traditional neural networks such as poor data processing effects and gradient vanishing. Although this method mitigates the shortcomings of traditional neural networks, its prediction accuracy and overall performance still require improvement.

To address these challenges effectively, this study introduces a self-attention mechanism into the temporal convolutional network to obtain different feature weights and utilizes the Adam optimization algorithm to enhance the overall performance of the prediction model, significantly improving prediction accuracy.”

 Point 3:The captions in Figures and Tables are shallow, please include a more complete description of figures and tables. Besides, the police size used in the images should be similar in size and type as the current text.

Response 3:According to your suggestions, the titles of Figures 1, 2, 3 and 12 have been supplemented for completeness.

The title of Figure 1 has been changed to "Structure diagram of self-attention mechanism principle."

The title of Figure 2 has been changed to "Temporal convolutional network structure diagram."

The title of Figure 3 has been changed to "Model structure diagram of self-attention mechanism and temporal convolutional network."

The title of Figure 12 has been changed to "Distribution diagram of prediction error for remaining useful life of engine."

Adjustments have been made to the font sizes, types, and other details in Figures 2, 3, 4, 5, and 6 to ensure consistency with the text of the paper.

Point 4:The manuscript requires deep proofreading and the equations must be double-checked. Please, revise (13) to verify sign in the exponential functions.。

Response 4:Based on your advice, careful proofreading of all equations in the manuscript has been conducted. Equation (13) has been standardized with correct parentheses, and the symbols in the exponential function have been verified for accuracy. (Line 261)

 

Point 5:The results should be discussed deeply in a subsection for "Discussions".

Response 5: Based on your suggestions, we have thoroughly discussed the results in Section 6.1 of the paper. The content is as follows: (Line 332)

“This paper proposes a hybrid network structure using Adam optimization and SAM-TCN for predicting the remaining useful life of engines. It utilizes a self-attention mechanism to measure the contributions of different features to the engine's remaining life. The results are evaluated using comprehensive performance metrics, with RMSE and Score on four datasets reported as follows: 11.50, 16.45, 11.62, and 15.47 for RMSE, and 225.32, 1136.27, 259.79, and 1365.40 for Score. From these metrics, it is evident that the proposed method performs well on the FD001 and FD003 datasets with smaller errors. This is primarily because the other two datasets contain more noise and complex operating conditions, making it challenging for the prediction model to capture these variations, resulting in less satisfactory performance. However, overall, the proposed model achieves good predictive accuracy. To comprehensively assess the prediction effectiveness of the proposed method, comparisons and analyses with methods from other literature are conducted in the next section.”

Comments on the Quality of English Language

Point 6:The masnuscript should be carefully revised to improve clarity.

Response 6: Based on your suggestions, we have carefully revised this paper, checking for correct English grammar and formatting. We have removed unnecessary punctuation in the paper (line 12) to enhance clarity and effectiveness of expression. (Line 12)

 

We express our sincere gratitude for your positive evaluation of the manuscript's organization and publication potential, as well as your recognition of the practical significance of our research. Taking into account your valuable suggestion, we have diligently revised the manuscript to enhance its robustness and suitability for publication.

We firmly believe that these revisions have significantly elevated the overall quality and comprehensiveness of our manuscript. Once again, we extend our appreciation for your valuable feedback and the considerable time and effort you invested in reviewing our work. Your guidance has provided us with the opportunity to strengthen our research further. We are hopeful that our revised manuscript meets your expectations and eagerly await your further guidance.

Sincerely,

The Authors

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper is well written & presents an interesting incremental advance on the field. I have only a few minor comments:

1. PHM is usually an abbreviation for "Performance and Health Management", on L21 you say "Prognostics.." and reference [1]. I do not find a reference to "Prognostics...". Do you want to change it?

2. your references are good in the introduction. Below I list 3 other references that you did not refer to. Do you want to include them?

3.  Figure 10 is very hard to understand & adds little value to the presentation. I suggest either deleting it, or showing the error between the RUL and RUL-estimate.

Potential additional references:

Qiao, Xianpeng, et al. "Advances and limitations in machine learning approaches applied to remaining useful life predictions: a critical review." The International Journal of Advanced Manufacturing Technology (2024): 1-18.

Song, Ya, et al. "Remaining useful life prediction of turbofan engine using hybrid model based on autoencoder and bidirectional long short-term memory." Journal of Shanghai Jiaotong University (Science) 23 (2018): 85-94.

Li, Jie, et al. "Remaining useful life prediction of turbofan engines using CNN-LSTM-SAM approach." IEEE Sensors Journal 23.9 (2023): 10241-10251.

Comments on the Quality of English Language

Some editorial mistakes in the language - Missing letters or spaces, extra hyphens, Capitalisation.  See, for example L60, L275

Author Response

Response to Reviewer 2 Comments

Dear Reviewer,

Thank you for reviewing our manuscript entitled "Method for Remaining Useful Life Prediction of Turbofan Engines Combining Adam Optimization Based Self-attention Mechanism with Temporal Convolutional Networks"[applsci-3093032]. We sincerely appreciate your valuable feedback and suggestions, as they have greatly contributed to enhancing the quality and impact of our research. We have thoroughly examined each of your comments and are pleased to inform you that we have successfully addressed the required revisions. In light of your feedback, we have implemented the following revisions:

Point 1: The paper is well written & presents an interesting incremental advance on the field. I have only a few minor comments:1. PHM is usually an abbreviation for "Performance and Health Management", on L21 you say "Prognostics.." and reference [1]. I do not find a reference to "Prognostics...". Do you want to change it?

Response 1: Based on your advice, I have reviewed the content of reference [1] and other literature. The literature indicates that in the field of remaining useful life prediction for equipment, PHM is commonly referred to as "Prognostic and Health Management." Since this paper also falls within this research domain, I have retained "Prognostic and Health Management" in the text and did not change it to "Performance and Health Management."

Point 2: 2. your references are good in the introduction. Below I list 3 other references that you did not refer to. Do you want to include them?Potential additional references:

Qiao, Xianpeng, et al. "Advances and limitations in machine learning approaches applied to remaining useful life predictions: a critical review." The International Journal of Advanced Manufacturing Technology (2024): 1-18.

Song, Ya, et al. "Remaining useful life prediction of turbofan engine using hybrid model based on autoencoder and bidirectional long short-term memory." Journal of Shanghai Jiaotong University (Science) 23 (2018): 85-94.

Li, Jie, et al. "Remaining useful life prediction of turbofan engines using CNN-LSTM-SAM approach." IEEE Sensors Journal 23.9 (2023): 10241-10251.

Response 2:Based on your advice, I have carefully read the three references you provided and found them to be highly valuable. I have incorporated citations to these three papers in Section 3 of the manuscript. The citations for these three references in the paper are [26], [27], and [28], respectively. The main contributions are as follows: (Line 182)

“In existing studies, Qiao et al. [26] discussed data and results from RUL prediction 167

experiments on turbofan engines, finding that indirect mapping approaches yield better 168

results, despite being more challenging and time-consuming to implement. Song et al. [27],

considering the high dimensionality and volume of engine monitoring data, designed a bidirectional long short-term memory network for RUL prediction. Li et al. [28] respectively utilized CNN and LSTM networks to extract spatial features, perform data fusion, and employed the SAM method to obtain feature weights.”

26. Xianpeng, Q.; Veronica, L.J.; Lim, C.S.; Tiyamike, B. Advances and limitations in machine learning approaches applied toremaining useful life predictions: a critical review.The International Journal of Advanced Manufacturing Technology. 2024,133(8), 4059–4076.
27. Ya, S.; Guo, S.; Leyi, C.; Xinpei, H.; Tangbin, X. Remaining Useful Life Prediction of Turbofan Engine Using Hybrid Model Basedon Autoencoder and Bidirectional Long Short-Term Memory.Journal of Shanghai Jiaotong University (Science). 2018, 23(12), 85–94.
28. Jie, L.; Yuanjie, J.; Mingbo, N.; Wei, Z.; Fanxi, M. Remaining Useful Life Prediction of Turbofan Engines Using CNN-LSTM-SAM Approach.IEEE Sensors Journal. 2023, 23(9), 10241–10251.

 

Point 3: 3.  Figure 10 is very hard to understand & adds little value to the presentation. I suggest either deleting it, or showing the error between the RUL and RUL-estimate.

Response 3: Based on your suggestion, we have supplemented the explanation of the contribution of Figure 10 in the paper. We have added a description of Figure 10, clarifying the true and predicted values of Remaining Useful Life (RUL), aiming to enhance readers' understanding of its significance. The lines representing RUL true and predicted values in Figure 10 are closely fitted, indicating minimal error. The specific details of prediction errors in Figure 10 are further elaborated in Figure 12 below. The following is the supplementary explanation for Figure 10: (Line 309)

“In the figure, the horizontal axis represents the engine number, and the vertical axis represents the RUL of the engines. The solid blue line indicates the actual RUL values of the engines, while the dashed red line represents the predicted RUL results using the method proposed in this paper. From Figure 10, it can be observed that the blue solid line fits closely with the red dashed line, indicating a small prediction error and hence a good predictive performance.”

Comments on the Quality of English Language

Point 4: Some editorial mistakes in the language - Missing letters or spaces, extra hyphens, Capitalisation.  See, for example L60, L275.

Response 4:Based on your advice, the entire manuscript has been rechecked and proofread. In line 60 of the paper, there was an extra word "and" at the end of the sentence, which has been removed. (Line 64)

The sentence in line 275 has been reviewed again.

We express our sincere gratitude for your positive evaluation of the manuscript's organization and publication potential, as well as your recognition of the practical significance of our research. Taking into account your valuable suggestion, we have diligently revised the manuscript to enhance its robustness and suitability for publication.

We firmly believe that these revisions have significantly elevated the overall quality and comprehensiveness of our manuscript. Once again, we extend our appreciation for your valuable feedback and the considerable time and effort you invested in reviewing our work. Your guidance has provided us with the opportunity to strengthen our research further. We are hopeful that our revised manuscript meets your expectations and eagerly await your further guidance.

Sincerely,

The Authors

 

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

1. TITLE - it is not necessary to have "Research on" in the title. It is better to have the word Method instead. So, better title could be "Method for Remaining Useful Life Prediction of ..."

 

2. ABSTRACT - should contain full words with abbreviations at first appearance, starting with RUL in line 1.

 

3. MANUSCRIPT STRUCTURE should be rearranged - Common structure of an article type of manuscript is: 1. Introduction (motivation, short current status of related work, research gap, contribution, announcement of the rest of the paper), 2.Background (having basic terms from the title explained), 3.Related work or Literature Review (presenting previous studies in the area), 4. The Proposed Method (or Method), 5. Experiment (with experiment setup having methods, tools, sample -size, features; results), 6. Discussion, 7. Conclusion, References. Sometimes introduction is extended with background paragraph and detailed related work, but it is better to have them separated from the introduction.

 

3.1.) Currently, in this manuscript there is Introduction and Method section immediately after the Introduction. In generally suggested structure for MDPI articles, it is properly organized. It is expected to have novel approaches and methods under the section entitled "Method", which emphasizes the contribution. However, subsection "2.1. Self-attention" as a well-known mechanism used in deep learning has been put under the section "2. Method". It is expected to have this text placed under Background section, that would be placed after the Introduction. If authors wish to present improved SAM mechanism, then improvements should be clearly stated and compared to basic SAM within the Method section. Since text under 2.1. completely explains basic SAM, it should be placed in the Background section.The same situation is with 2.2. Temporal Convolutional Network. Text presented here explains this term with formulas and figure. This section should be shifted into Background section. The same is with 2.3. Adam optimization algorithm - to be shifted into Background section.

 

3.2.) Contribution 

 

3.2.1.) Section "Method" should be entitled "The proposed method". 

 

3.2.2.) Within the text of the proposed Method, authors should compare their proposal of integration SAM with TCN with contribution from paper: Yan Zhen, Junyi Fang, Xiaoming Zhao, Jiawang Ge, Yifei Xiao (2022): Temporal convolution network based on attention mechanism for well production prediction, Journal of Petroleum Science and Engineering,Volume 218, November 2022, Elsevier, 2022.

(https://www.sciencedirect.com/science/article/abs/pii/S0920410522008956). 

 

3.2.3.) The section "4. The Proposed Method" should contain subsections from 3.3. SAM-TCN, but entitled "4.1. SAM-TCN method", then 4.2. should be "The Adam-SAM-TCN Remaining Useful Life Prediction Model"

 

3.4.) After the proposed Method, there should be section 5. Experiment, which will include 5.1. Experimental Data (currently it is 3.1.), 5.2. Evaluation Metrics for Predicted Results (currently 3.2.), 5.3. Experimental results (currently it is section 4. entitled Analysis of Experimental results).

 

3.5) Section 6. should be placed after Experiment, and entitled 6. Discussion. Here, text should include discussion about experiment results (could be placed under 6.1. Experimental results dicussion - currently it is text in the conclusion under (3)), and the proposed method compared with other similar methods (currently 4.2.2. Comparative study could be 6.2. Comparative study).

 

4. Referencing the source for statements and equations - it is necessary to have every statement, figure or equation taken from any source referred properly according to references list. If authors do not provide sources, then it is supposed that text, figures and equations represent their contribution, but within the self-attention mechanism explaining text, it is obvious that it is not their contribution.Of course, figures could be created for the purpose of explaining of basic terms, but they should also be sourced (in figure caption there could be "according to [2])

First example - text explaining Self attention in 2.1. is based on sources, but there are only two references [17] and [18] put in the beginning and end of this text (so equations remain uncovered with the sources).  

Second example - Temporal convolutional network.

 

5. REFERENCES LIST

5.1.) Item no 5. starts with Author 1 - this should be corrected.

5.2.) Item No 11. has two names of an author without space between, i.e. "GiduthuriSateesh" should be "Giduthuri Sateesh"

5.3.) Check formatting of references - does title or name of the conference require italic letters? Do appropriate formatting for item No 9, 11, 24, 25 

5.4.) Item no 15 is not complete with data - is it Master thesis, PhD thesis?

5.5.) Some items have name of conference or journal set immediately after the publication title, without any space between - see and correct item no: 12, 13, 14, 16-20, 23, 26, 28, 29.  

5.6.) Why is name of the conference bold in the item 21? Please check formatting rules for references list. 

5.7.) Item no 22 starts with small letter, it should be Gouhua instead of gouhua.

Comments on the Quality of English Language

1. Title should be written in Title Case. Some words should be capitalized, such as Useful, Based.

2. Articles should be used properly. For example, in line 17, instead of "for aircraft" there should be "for an aircraft".

3. The use of abbreviations should be avoided in the title, but the use of whole words, if possible.

4. Abbreviations should be explained with full words at the first appearance. For example, in line 1 (abstract) there is RUL abbreviation, not fully explained. Of course, it stands for Remaining Useful Life, but it is expected to have these words before or after RUL at the first appearance in text. This abbreviation was explained in line 25, but not at the first appearance, in abstract. ADAM and SAM are explained in lines 78-80 but it should be set earlier (with the first appearance of these abbreviations), TCN is explained in lines 66-68.

5. It is necessary to have full names for terms, such as network - it is better to use "neural network" (if possible, in title, but starting with abstract - line 3), then "self-attention" in line 92 should be "self-attention mechanism in deep learning".

6. Sentences formation - sometimes some sentence parts are missing. For example, in line 93: "The SAM primarily used for handling complex data" should be "The SAM is primarily used for handling complex data".

7. Comma symbol "," should be used properly. For example, in line 125, instead of "up to time t ,effectively" should be "up to time t, effectively".

 

 

Author Response

Response to Reviewer 3 Comments

Dear Reviewer,

Thank you for reviewing our manuscript entitled "Method for Remaining Useful Life Prediction of Turbofan Engines Combining Adam Optimization Based Self-attention Mechanism with Temporal Convolutional Networks"[applsci-3093032]. We sincerely appreciate your valuable feedback and suggestions, as they have greatly contributed to enhancing the quality and impact of our research. We have thoroughly examined each of your comments and are pleased to inform you that we have successfully addressed the required revisions. In light of your feedback, we have implemented the following revisions:

Point 1: 1. TITLE - it is not necessary to have "Research on" in the title. It is better to have the word Method instead. So, better title could be "Method for Remaining Useful Life Prediction of ...".1.

Response 1: Based on your advice, the original title "Research on Remaining useful Life Prediction of Turbofan Engines based on Adam-optimized SAM-TCN Network" has been revised to "Method for Remaining Useful Life Prediction of Turbofan Engines based on Adam-optimized SAM-TCN Network" in the manuscript.

Point 2: 2. ABSTRACT - should contain full words with abbreviations at first appearance, starting with RUL in line 1. 2.

Response 2: Based on your advice, I have verified the English abbreviations throughout the manuscript and noted the full terms where abbreviations first appear. Below are the lines where the abbreviations first appear after the modifications:

remaining useful life (RUL)  (Line 1)

adaptive moment estimation (Adam)  (Line 5)

self attention mechanism-temporal convolutional network (SAM-TCN)  (Line 6)

temporal convolutional network (TCN)  (Line 9)

Self attention mechanism (SAM)  (Line 11)

root mean square error (RMSE)   (Line 13)

recurrent neural network (RNN)   (Line 61)

convolutional neural network (CNN)  (Line 63)

Point 3: 3. MANUSCRIPT STRUCTURE should be rearranged - Common structure of an article type of manuscript is: 1. Introduction (motivation, short current status of related work, research gap, contribution, announcement of the rest of the paper), 2.Background (having basic terms from the title explained), 3.Related work or Literature Review (presenting previous studies in the area), 4. The Proposed Method (or Method), 5. Experiment (with experiment setup having methods, tools, sample -size, features; results), 6. Discussion, 7. Conclusion, References. Sometimes introduction is extended with background paragraph and detailed related work, but it is better to have them separated from the introduction.

Response 3: According to your advice, the structure of the paper has been reorganized. The revised table of contents structure is as follows:

1. Introduction
2. Background

2.1 Self-Attention

2.2 Temporal Convolutional Network

2.3 Adam Optimization Algorithm

3. Related Work
4. The Proposed Method

4.1 SAM-TCN

4.2 The Adam-SAM-TCN Remaining Useful Life Prediction Model

5.Experiment

5.1. Experimental Data

5.2.Evaluation Metrics for Predicted Results

5.3.Experimental Result

5.3.1 Data Processing

      Feature Selection

      Data Normalization and RUL Labeling

5.3.2 Predicted RUL Results

6.Discussion

6.1 Experimental Results Dicussion

6.2 Comparative Study

7. Conclusion

References

Point 4: 3.1.) Currently, in this manuscript there is Introduction and Method section immediately after the Introduction. In generally suggested structure for MDPI articles, it is properly organized. It is expected to have novel approaches and methods under the section entitled "Method", which emphasizes the contribution. However, subsection "2.1. Self-attention" as a well-known mechanism used in deep learning has been put under the section "2. Method". It is expected to have this text placed under Background section, that would be placed after the Introduction. If authors wish to present improved SAM mechanism, then improvements should be clearly stated and compared to basic SAM within the Method section. Since text under 2.1. completely explains basic SAM, it should be placed in the Background section.The same situation is with 2.2. Temporal Convolutional Network. Text presented here explains this term with formulas and figure. This section should be shifted into Background section. The same is with 2.3. Adam optimization algorithm - to be shifted into Background section.

Response 4: Based on your advice, the section arrangement of Section 2 in the paper has been adjusted. The revised content is as follows: (Line 107) 

2. Background

2.1 Self-Attention

2.2 Temporal Convolutional Network

2.3 Adam Optimization Algorithm

Point 5: 3.2.) Contribution 3.2.1.) Section "Method" should be entitled "The proposed method". 

Response 5:Based on your advice, I have modified "Method" in the section of the paper to "The proposed method". (Line 197)

Point 6: 3.2.2.) Within the text of the proposed Method, authors should compare their proposal of integration SAM with TCN with contribution from paper:

Yan Zhen, Junyi Fang, Xiaoming Zhao, Jiawang Ge, Yifei Xiao (2022): Temporal convolution network based on attention mechanism for well production prediction, Journal of Petroleum Science and Engineering,Volume 218, November 2022, Elsevier, 2022.

(https://www.sciencedirect.com/science/article/abs/pii/S0920410522008956).

Response 6: Based on your advice, a comparison with the SAM and TCN integrated method proposed in the referenced literature [29] has been added to the content of Section 3 of the manuscript. The specific content is as follows: (Line 188) 

“Zhen et al. [29] proposed a TCN-attention model for oil well production prediction to overcome issues with traditional neural networks such as poor data processing effects and gradient vanishing. Although this method mitigates the shortcomings of traditional neural networks, its prediction accuracy and overall performance still require improvement.

To address these challenges effectively, this study introduces a self-attention mechanism into the temporal convolutional network to obtain different feature weights and utilizes the Adam optimization algorithm to enhance the overall performance of the prediction model, significantly improving prediction accuracy.”

29. Yan, Z,; Junyi, F,; Xiaoming, Z,; Jiawang, G.; Yifei, X. Temporal convolution network based on attention mechanism for well production prediction.Journal of Petroleum Science and Engineering. 2022, 218(11), 111043.

 

Point 7: 3.2.3.) The section "4. The Proposed Method" should contain subsections from 3.3. SAM-TCN, but entitled "4.1. SAM-TCN method", then 4.2. should be "The Adam-SAM-TCN Remaining Useful Life Prediction Model".

Response 7: Based on your suggestion, the sections of the paper have been reorganized, and Section 4 has been revised to:  (Line 197)

4. The Proposed Method

4.1 SAM-TCN

4.2 The Adam-SAM-TCN Remaining Useful Life Prediction Model

Point 8:3.4.) After the proposed Method, there should be section 5. Experiment, which will include 5.1. Experimental Data (currently it is 3.1.), 5.2. Evaluation Metrics for Predicted Results (currently 3.2.), 5.3. Experimental results (currently it is section 4. entitled Analysis of Experimental results).

Response 8: According to your advice, the section order of the experimental part in Section 5 has been adjusted. The revised content is as follows: (Line 232)  

5. Experiment

5.1. Experimental Data

5.2.Evaluation Metrics for Predicted Results

5.3.Experimental Result

5.3.1 Data Processing

5.3.1.1  Feature Selection

5.3.1.2  Data Normalization and RUL Labeling

5.3.2 Predicted RUL Results

 

Point 9: 3.5) Section 6. should be placed after Experiment, and entitled 6. Discussion. Here, text should include discussion about experiment results (could be placed under 6.1. Experimental results dicussion - currently it is text in the conclusion under (3)), and the proposed method compared with other similar methods (currently 4.2.2. Comparative study could be 6.2. Comparative study).

Response 9: According to your suggestion, the content of Section 6 has been rearranged, and the revised section title is as follows: (Line 332)

6. Discussion

6.1 Experimental Results Dicussion

6.2 Comparative Study

The discussion section of the experimental results in 6.1 has been supplemented with the following content:

“This paper proposes a hybrid network structure using Adam optimization and SAM-TCN for predicting the remaining useful life of engines. It utilizes a self-attention mechanism to measure the contributions of different features to the engine's remaining life. The results are evaluated using comprehensive performance metrics, with RMSE and Score on four datasets reported as follows: 11.50, 16.45, 11.62, and 15.47 for RMSE, and 225.32, 1136.27, 259.79, and 1365.40 for Score. From these metrics, it is evident that the proposed method performs well on the FD001 and FD003 datasets with smaller errors. This is primarily because the other two datasets contain more noise and complex operating conditions, making it challenging for the prediction model to capture these variations, resulting in less satisfactory performance. However, overall, the proposed model achieves good predictive accuracy. To comprehensively assess the prediction effectiveness of the proposed method, comparisons and analyses with methods from other literature are conducted in the next section.”

Point 10: 4. Referencing the source for statements and equations - it is necessary to have every statement, figure or equation taken from any source referred properly according to references list. If authors do not provide sources, then it is supposed that text, figures and equations represent their contribution, but within the self-attention mechanism explaining text, it is obvious that it is not their contribution.Of course, figures could be created for the purpose of explaining of basic terms, but they should also be sourced (in figure caption there could be "according to [2])

First example - text explaining Self attention in 2.1. is based on sources, but there are only two references [17] and [18] put in the beginning and end of this text (so equations remain uncovered with the sources).  

Second example - Temporal convolutional network

Response 10: According to your feedback, references have been added to citations in figures and equations throughout the text.

In Section 2.1, the formula for the attention mechanism now includes a citation as follows:

"According to [18], its computational formula is as follows:" (Line 120)

18. Ye Qin. Research on Maintenance Decision-making of Turbofan Engines Based on Attention-based Temporal Convolutional Network and Evolutionary Game. Master’s thesis, Huazhong University of Science and Technology, Hubei, 2021: 1-102.

The source citation for Figure 1 in Section 2.1 is [19]. (Line 130)

19. Zhiqiang, X.; Yujie, Z.; Jianguo, M.; Qiang, M. Global attention mechanism based deep learning for remaining useful life prediction of aero-engine.Measurement. 2023, 217(8), 1-10.  

The source of the formulas in the time convolutional network in section 2.2 has been cited in the literature, with reference [22]: (Line 149)

22. Gengwei, Z. Rolling bearing life prediction based on Temporal Convolutional Network. Master’s thesis, Nanjing University of Aeronautics and Astronautics, Jiangsu, March 2022: 1-76.

The reference source for Figure 2 in Section 2.2 has been annotated, with reference [23]: (Line 160)

23. Yu, S.; Qiang, D.; Yudong X.; Cong L. Chiller fault diagnosis based on combination of multiblock and self-attention TCN.The Chinese Journal of Process Engineering. 2024, 24(2), 162–171.

The formula source in section 2.3 has been cited, with reference [25]: (Line 171)

25. Yang, L. Prediction of residual life of gearbox bearing of offshore wind turbine. Master’s thesis, Jiangsu University of Science and Technology, Jiangsu, May 2022: 1-75.

 

The citation source for formula (14) has been marked, with reference [33]: (Line 280)

33. Wang, H.; Li, D.; Li, D.; Liu, C.; Yang, X.; Zhu, G. Remaining Useful Life Prediction of Aircraft Turbofan Engine Based on Random Forest Feature Selection and Multi-Layer Perceptron.Applied Sciences. 2023, 13(12), 7186.

The citation source for formula (15) has been indicated, with reference [34]: (Line 291)

34. Junwei, W.; Baojia, C.; Zhengkun, C.; Gang, W.; Zhuxin, T.; Qiang, L. Application of SSA-TCN in Prediction of Remaining UsefulLife of Turbofan Engine.J of China Three Gorges Univ.(Natural Sciences). 2023, 45(6), 92-100.

Point 11: 5. REFERENCES LIST

5.1.) Item no 5. starts with Author 1 - this should be corrected.

Response11 : Based on your suggestion, the redundant first author in reference number 5 has been removed.

Point 12: 5.2.) Item No 11. has two names of an author without space between, i.e. "GiduthuriSateesh" should be "Giduthuri Sateesh".

Response 12: According to your suggestion, reference 11 has been modified. "GiduthuriSateesh" has been changed to "Giduthuri Sateesh".

Point 13: 5.3.) Check formatting of references - does title or name of the conference require italic letters? Do appropriate formatting for item No 9, 11, 24, 25. 

Response 13: According to your advice, the journal titles and conference names in references 9, 11, 24, and 25 have been italicized, and appropriate formatting adjustments have been made.

Point 14: 5.4.) Item no 15 is not complete with data - is it Master thesis, PhD thesis?

Response 14: According to your advice, the content of reference number 15 has been supplemented. This paper is a master's thesis.

Point 15: 5.5.) Some items have name of conference or journal set immediately after the publication title, without any space between - see and correct item no: 12, 13, 14, 16-20, 23, 26, 28, 29.  

Response 15: According to your suggestion, the references section of the paper has been rechecked and revised. Spaces have been added in references 12, 13, 14, and 16 to 20, as well as in 23, 26, 28, and 29, and errors have been corrected.

Point 16: 5.6.) Why is name of the conference bold in the item 21? Please check formatting rules for references list.

Response 16: According to your suggestion, reference number 21 has been checked and revised to ensure its format is correct.

Point 17: 5.7.) Item no 22 starts with small letter, it should be Gouhua instead of gouhua.

Response 17: According to your advice, the term "gouhua" in reference number 22 (now [31]) has been changed to "Gouhua" with a capital "G".

Comments on the Quality of English Language

Point 18: 1. Title should be written in Title Case. Some words should be capitalized, such as Useful, Based.1.

Response 18: According to your suggestion, the lowercase words "useful" and "based" in the title have been changed to uppercase "Useful" and "Based".

Point 19: 2. Articles should be used properly. For example, in line 17, instead of "for aircraft" there should be "for an aircraft".

Response19: According to your suggestion, "for aircraft" in the paper has been modified to "for an aircraft".

Point 20: 3. The use of abbreviations should be avoided in the title, but the use of whole words, if possible.3.

Response 20: According to your suggestion, abbreviations in the title have been replaced with their full words. The revised title is:

“Method for Remaining Useful Life Prediction of Turbofan Engines Combining Adam Optimization Based Self-attention Mechanism with Temporal Convolutional Networks”

Point 21: 4. Abbreviations should be explained with full words at the first appearance. For example, in line 1 (abstract) there is RUL abbreviation, not fully explained. Of course, it stands for Remaining Useful Life, but it is expected to have these words before or after RUL at the first appearance in text. This abbreviation was explained in line 25, but not at the first appearance, in abstract. ADAM and SAM are explained in lines 78-80 but it should be set earlier (with the first appearance of these abbreviations), TCN is explained in lines 66-68.

Response 21: According to your suggestion, abbreviations that first appear in the text have been reviewed, and full explanations have been provided. The following is the list of abbreviations and their explanations as they appear for the first time in the revised paper:

remaining useful life (RUL)  (Line 1)

adaptive moment estimation (Adam)  (Line 5)

self attention mechanism-temporal convolutional network (SAM-TCN)  (Line 6)

temporal convolutional network (TCN)  (Line 9)

Self attention mechanism (SAM)  (Line 11)

root mean square error (RMSE)   (Line 13)

recurrent neural network (RNN)   (Line 61)

convolutional neural network (CNN)  (Line 63)

Point 22: 5. It is necessary to have full names for terms, such as network - it is better to use "neural network" (if possible, in title, but starting with abstract - line 3), then "self-attention" in line 92 should be "self-attention mechanism in deep learning".

Response 22: According to your suggestion, in the text, "network" has been changed to "neural network", and "self-attention" has been changed to "self-attention mechanism in deep learning". (Line 76)

Point 23: 6. Sentences formation - sometimes some sentence parts are missing. For example, in line 93: "The SAM primarily used for handling complex data" should be "The SAM is primarily used for handling complex data".

Response 23: According to your suggestion, sentences in the document have been checked and revised. "The SAM primarily used for handling complex data" has been modified to "The SAM is primarily used for handling complex data".(Line 109)

Point 24: 7. Comma symbol "," should be used properly. For example, in line 125, instead of "up to time t ,effectively" should be "up to time t, effectively".

Response 24: According to your suggestion, punctuation in the text has been checked, and "up to time t ,effectively" has been modified to "up to time t, effectively". (Line 143)

We express our sincere gratitude for your positive evaluation of the manuscript's organization and publication potential, as well as your recognition of the practical significance of our research. Taking into account your valuable suggestion, we have diligently revised the manuscript to enhance its robustness and suitability for publication.

We firmly believe that these revisions have significantly elevated the overall quality and comprehensiveness of our manuscript. Once again, we extend our appreciation for your valuable feedback and the considerable time and effort you invested in reviewing our work. Your guidance has provided us with the opportunity to strengthen our research further. We are hopeful that our revised manuscript meets your expectations and eagerly await your further guidance.

Sincerely,

The Authors

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have addressed correctly my comments but the format used to present the responses is not the optimal to follow corrections accurately.