Prediction of Blade Tip Timing Sensor Waveforms Based on Radial Basis Function Neural Network
Round 1
Reviewer 1 Report
The main three problems of the current study are as follows. First, to what extent you think using a neural network with 10 (or less) input points is beneficial (a NN is known to be data-demanding technique). Second, there is no benchmarking (comparison) against other simpler methods and there is no performance metrics for deciding about the “goodness” of your approach. Performance metrics may include RMSE, MAPE, etc. Third, the authors did not specify how they determine the RBF NN hyperparameters such as the number of neurons in each layer, etc.
1. The literature review should be more critical including the used technique, the dataset, the obtained results, the pros and cons of each study. Otherwise, your current literature review is too narrative and can’t contribute to well situate the contributions of your paper. You can include a summary of the previous works on the topic in a table for more readability.
2. The contributions’ part should be rewritten to include clearly the study contributions. Particularly, the authors should specify whether the RBF technique was not used. If used, please, specify the novelties of your work.
3. Line 159: What you mean by living things?
4. Please, include the source of Figure3 if it is not yours.
5. The results of all compared models should be included in a unique figure to allow the reader sees clearly the difference.
6. Please, add performance metrics in a table for better comparing the results of different models.
Decision: Reject (Encourage resubmission).
English should be polished.
Author Response
Dear Editors and Reviewers,
Thank you very much for your letter and for the reviewers' comments concerning our manuscript entitled " Prediction of blade tip timing sensor waveforms based on Radial Basis Function Neural Network" (ID: applsci-2551883). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. The main corrections in the manuscript and the responds to the reviewer’s comments are listed as follows. The corrections parts in the manuscript are marked in blue.
Reviewer 1
The main three problems of the current study are as follows. First, to what extent you think using a neural network with 10 (or less) input points is beneficial (a NN is known to be data-demanding technique). Second, there is no benchmarking (comparison) against other simpler methods and there is no performance metrics for deciding about the “goodness” of your approach. Performance metrics may include RMSE, MAPE, etc. Third, the authors did not specify how they determine the RBF NN hyperparameters such as the number of neurons in each layer, etc.
Response: Thanks for your suggestions. As the current blade tip timing(BTT) method has the problems of under-sampling and lower recognition accuracy, it is chosen to establish RBF NN to analyze the static calibration experimental data, while, because of the simple structure of RBF NN, which is simple to train and has fast learning convergence, it can approximate the arbitrary nonlinear function, therefore, combining both of them to predict the waveforms of BTT sensors, to reduce prediction errors that caused by the lack of sampling points, then comparing the prediction curves of multiple models. Currently, the output voltage of the BTT sensor is predicted with the data of the blade tip clearance and the angle at which the blade cuts the magnetic susceptor in 10 input points, which still shows good performance with fewer input points, and "Reddy, R.R.K.; Ganguli, R. Structural damage detection in a helicopter rotor blade using radial basis function neural networks. Smart materials Structures 2003, 12, 232. " also uses only three modal rotational frequency sets to detect the structural damage of a helicopter rotor blade, and still obtains good damage detection accuracy. This model is of great benefit in making up for the missing data caused by the under-sampling of the BTT sensor, enriching the waveform data of the BTT sensor, and expanding the database of the BTT; In the paper, RBF NN is compared with Kriging model and polynomial function fitting prediction curve, and RBF NN is the better model in comprehensive comparison, meanwhile, the table about the comparison of performance indexes between RBF NN and Kriging model has been added in the paper; And the choice of hyperparameters for RBF NN is now embodied in the table of the paper.
Comment 1: The literature review should be more critical including the used technique, the dataset, the obtained results, the pros and cons of each study. Otherwise, your current literature review is too narrative and can’t contribute to well situate the contributions of your paper. You can include a summary of the previous works on the topic in a table for more readability.
Response: A tabular list of previous work on vibration detection methods for rotating blades is now included in the manuscript, and the advantages, disadvantages, as well as constraints of each technique studied have been recounted.
Comment 2: The contributions’ part should be rewritten to include clearly the study contributions. Particularly, the authors should specify whether the RBF technique was not used. If used, please, specify the novelties of your work.
Response: The contributions’ part recounts the contributions of others who have conducted various studies on using neural network methods, and compares the advantages of the RBF NN used in this paper.
Comment 3: Line 159: What you mean by living things?
Response: By living things in 159 is meant animals, that is, the central nervous system of animals. The explanation of this term has now been added to the manuscript.
Comment 4: Please, include the source of Figure3 if it is not yours.
Response: Thanks for your suggestion, and a reference to Figure 3 has been added to the manuscript.
Comment 5: The results of all compared models should be included in a unique figure to allow the reader sees clearly the difference.
Response: The comparative results regarding all models are presented in a more visual way in Fig. 6, Fig. 10, Fig. 14, Fig. 18.
Comment 6: Please, add performance metrics in a table for better comparing the results of different models.
Response: The performance metrics table about RBF model, Kriging model has now been added to the manuscript.
Finally, we appreciate for Editors/Reviewers' warm work earnestly, and hope that the correction will meet with approval. Once again, thank you very much for your comments and suggestions.
Best regards,
Liang Zhang, Cong Chen, Yiming Xia, Qingxi Song and Junjun Cao
Address: 169 Shiying Street, Guta District, Jinzhou City, Liaoning Province, China. Faculty of Mechanical Engineering and Automation. Liaoning University of Technology.
Phone: +8615941611803
E-mail address: [email protected]
August 18, 2023
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Damage to axial compressor rotor blades by fatigue fractures is a important problem of current aircraft jet engines, but also stationary power equipment. Analysis and accurate determination of the applied loads that provoke the fatigue load on the rotor blades of axial compressors makes it possible to solve this problem.
The authors of the thesis solve this problem using a mathematical model and experimental measurements. The results obtained by mathematical modelling and compared with the results of experimental measurements show a very good agreement.
The work is processed at a very good content, formal and graphic level. Minor formal shortcomings appear in the work (lack of a more detailed description of the formulas used, formal shortcomings in writing commas, periods and spaces, etc.).
Due to the topicality of the work, I recommend publishing it after eliminating the above formal shortcomings.
Author Response
Dear Editors and Reviewers,
Thank you very much for your letter and for the reviewers' comments concerning our manuscript entitled " Prediction of blade tip timing sensor waveforms based on Radial Basis Function Neural Network" (ID: applsci-2551883). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. The main corrections in the manuscript and the responds to the reviewer’s comments are listed as follows. The corrections parts in the manuscript are marked in blue.
Reviewer 2
Damage to axial compressor rotor blades by fatigue fractures is a important problem of current aircraft jet engines, but also stationary power equipment. Analysis and accurate determination of the applied loads that provoke the fatigue load on the rotor blades of axial compressors makes it possible to solve this problem.
The authors of the thesis solve this problem using a mathematical model and experimental measurements. The results obtained by mathematical modelling and compared with the results of experimental measurements show a very good agreement.
The work is processed at a very good content, formal and graphic level. Minor formal shortcomings appear in the work (lack of a more detailed description of the formulas used, formal shortcomings in writing commas, periods and spaces, etc.).
Due to the topicality of the work, I recommend publishing it after eliminating the above formal shortcomings.
Response: Thanks to your comments, the problems with the form of writing formulas in the article have now been changed, as well as the descriptions of some of the formulas.
Finally, we appreciate for Editors/Reviewers' warm work earnestly, and hope that the correction will meet with approval. Once again, thank you very much for your comments and suggestions.
Best regards,
Liang Zhang, Cong Chen, Yiming Xia, Qingxi Song and Junjun Cao
Address: 169 Shiying Street, Guta District, Jinzhou City, Liaoning Province, China. Faculty of Mechanical Engineering and Automation. Liaoning University of Technology.
Phone: +8615941611803
E-mail address: [email protected]
August 18, 2023
Author Response File: Author Response.pdf
Reviewer 3 Report
A good and actual research work.
Introduction is appropriate. The references are mostly form Asian authors. More non-Asian authored references will improve the paper value.
The research design is appropriate.
The results are clearly presented.
The conclusions are supported by the results.
The aim of the research have to be defined more clear and this has to be the main point of the conclusion - it is made ... and the result shows .... Some applications of the results as well as further investigations will improve the paper.
There are also other methods for evaluating the vibration phenomena of the rotor blades. On the other hand further research on different type of blades is one of the proper way to continue this research. Presented work does not cover all type of blades.
Author Response
Dear Editors and Reviewers,
Thank you very much for your letter and for the reviewers' comments concerning our manuscript entitled " Prediction of blade tip timing sensor waveforms based on Radial Basis Function Neural Network" (ID: applsci-2551883). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. The main corrections in the manuscript and the responds to the reviewer’s comments are listed as follows. The corrections parts in the manuscript are marked in blue.
Reviewer 3
A good and actual research work.
Introduction is appropriate. The references are mostly form Asian authors. More non-Asian authored references will improve the paper value.
The research design is appropriate.
The results are clearly presented.
The conclusions are supported by the results.
The aim of the research have to be defined more clear and this has to be the main point of the conclusion - it is made ... and the result shows .... Some applications of the results as well as further investigations will improve the paper.
There are also other methods for evaluating the vibration phenomena of the rotor blades. On the other hand further research on different type of blades is one of the proper way to continue this research. Presented work does not cover all type of blades.
Response: Thanks for your advice, this paper has now added references to non-Asian authors, and we will continue to deepen our research and continue to conduct multiple types of blade analyses.
Finally, we appreciate for Editors/Reviewers' warm work earnestly, and hope that the correction will meet with approval. Once again, thank you very much for your comments and suggestions.
Best regards,
Liang Zhang, Cong Chen, Yiming Xia, Qingxi Song and Junjun Cao
Address: 169 Shiying Street, Guta District, Jinzhou City, Liaoning Province, China. Faculty of Mechanical Engineering and Automation. Liaoning University of Technology.
Phone: +8615941611803
E-mail address: [email protected]
August 18, 2023
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
1. The results of Table 2 don't show that your model is better than kriging model. A lot of mistakes exist in the table. Please, check it carefully.
2. In the same Table2, what you mean by number of neurons = 0 (zero)??? seems abnormal.
3. The authors should comapre to analytic models since the number of used samples is small.
English is almost good.
Author Response
Dear Editors and Reviewers,
Thank you very much for your letter and for the reviewers' comments concerning our manuscript entitled " Prediction of blade tip timing sensor waveforms based on Radial Basis Function Neural Network" (ID: applsci-2551883). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. The main corrections in the manuscript and the responds to the reviewer’s comments are listed as follows. The corrections parts in the manuscript are marked in blue.
Reviewer 1
Comment 1: The results of Table 2 don't show that your model is better than kriging model. A lot of mistakes exist in the table. Please, check it carefully.
Response: Thanks for your suggestions, the errors in the table have now been corrected, previously there were problems with the output due to MATLAB setup issues, and now the lines have been corrected.
Comment 2: In the same Table2, what you mean by number of neurons = 0 (zero)??? seems abnormal.
Response: Thanks for the suggestions, the neuron = 0 was intended to be a contrast for this group, and the data in the table has now been modified.
Comment 3: The authors should compare to analytic models since the number of used samples is small.
Response: The four groups of data in the original table (10, 8, 6, and 3) were each containing 109 data points, and because of an error in the presentation, the representations in the table have now been corrected, and the results of the comparison between the RBF NN and the Kriging model are presented in the table.
Finally, we appreciate for Editors/Reviewers' warm work earnestly, and hope that the correction will meet with approval. Once again, thank you very much for your comments and suggestions.
Best regards,
Liang Zhang, Cong Chen, Yiming Xia, Qingxi Song and Junjun Cao
Address: 169 Shiying Street, Guta District, Jinzhou City, Liaoning Province, China. Faculty of Mechanical Engineering and Automation. Liaoning University of Technology.
Phone: +8615941611803
E-mail address: [email protected]
August 25, 2023
Author Response File: Author Response.pdf
