Review Reports

Round 1

Reviewer 1 Report

The authors introduced an integrated approach for recognizing AS state that combines data fusion and Convolutional Neural Networks (CNNs). The method takes denoised vibration signal's wavelet transform time-frequency images as input and employs Convolutional Neural Networks for learning and supervising the data. The proposed approach extracts deep data structures layer by layer and enhances the classification results through data fusion technology. The use of Convolutional Neural Networks and data fusion techniques in the proposed method is highly effective, allowing for accurate and reliable recognition of assembled structure states. The paper is easy to read and quite well- organized, with clear explanations of the methodology and results. However, the paper's proposed method was tested on a small dataset, limiting its generalizability to other datasets. Also, It is suggested that the paper should include and discuss literature reviews on previous works related to classical neural networks. This would enhance the paper's content and provide a more comprehensive overview of the field. For instance, the following papers related to classical neural networks should be included and discussed in detail: Conditional Random Fields Meet Deep Neural Networks for Semantic Segmentation: Combining Probabilistic Graphical Models with Deep Learning for Structured Prediction. IEEE Signal Process. Mag. 35(1): 37-52 (2018), Vision-Based Detection of Guitar Players' Fingertips Without Markers. CGIV 2007: 419-428, LSTM recurrent networks learn simple context-free and context-sensitive languages. IEEE Trans. Neural Networks 12(6): 1333-1340 (2001), and Unsupervised Learning in LSTM Recurrent Neural Networks. ICANN 2001: 684-691. Furthermore, the paper did not mention the process used for parameter tuning, and it is not clear whether the proposed method's performance is optimal. Moreover, the proposed method relies on pre-processing techniques such as denoising and wavelet transform, which may not be suitable for all datasets or scenarios. Overall, I recommend that this manuscript can be accepted after major changes. 

Author Response

Point 1: the paper's proposed method was tested on a small dataset, limiting its generalizability to other datasets.

Response 1: We acknowledge the limitation of our study due to the small dataset used for testing, and we agree that it may affect the generalizability of our findings to other datasets. As you suggested, we plan to conduct further experiments with larger datasets in the future to verify the robustness and effectiveness of our proposed method.

Please provide your response for Point 1. (in red)

 

Point 2: Also, It is suggested that the paper should include and discuss literature reviews on previous works related to classical neural networks. This would enhance the paper's content and provide a more comprehensive overview of the field. For instance, the following papers related to classical neural networks should be included and discussed in detail: Conditional Random Fields Meet Deep Neural Networks for Semantic Segmentation: Combining Probabilistic Graphical Models with Deep Learning for Structured Prediction. IEEE Signal Process. Mag. 35(1): 37-52 (2018), Vision-Based Detection of Guitar Players' Fingertips Without Markers. CGIV 2007: 419-428, LSTM recurrent networks learn simple context-free and context-sensitive languages. IEEE Trans. Neural Networks 12(6): 1333-1340 (2001), and Unsupervised Learning in LSTM Recurrent Neural Networks. ICANN 2001: 684-691.

Response 2: We also appreciate your suggestion to include a literature review of classical neural networks relevant to our work. We will review the papers you mentioned and incorporate them into our discussion to provide a more comprehensive overview of the field.

 

Point 3: Furthermore, the paper did not mention the process used for parameter tuning, and it is not clear whether the proposed method's performance is optimal.

Response 3: Regarding the process of parameter adjustment, we have adjusted the parameters several times to obtain better CNN parameters, and we will include this information in Section 2.2.1 to make it more transparent.

 

Point 4: Furthermore, the paper did not mention the process used for parameter tuning, and it is not clear whether the proposed method's performance is optimal.

Response 4: Finally, we agree that our proposed method relies on preprocessing techniques such as denoising and wavelet transform, which may not be applicable to all datasets or scenarios. We will address this issue by discussing the limitations and potential challenges of our method in the revised manuscript. In Section 4 Discussion, The method proposed in this paper has been tested with a small sample size due to the limited shaking table test data of AS and the lack of actual engineering data. Therefore, the generalizability of our proposed method to other datasets may be limited. To validate the robustness and effectiveness of our approach, we plan to conduct further experiments on larger datasets in the future. Additionally, it should be noted that our method relies on preprocessing techniques such as denoising and wavelet transform, which may not be ap-plicable to all datasets or scenarios.

Author Response File: Author Response.docx

Reviewer 2 Report

This paper introduces the assembly structure state recognition method based on convolution neural network and data fusion. This method takes the wavelet transform time-frequency image of the de-noised vibration signal as the input, uses cnn to monitor and learn the data, extracts the deep data structure layer by layer, and improves the classification results through data fusion technology. To sum up, the study is interesting, but there are still serious aspects to be corrected in the paper:

1. In terms of content, the following problems need to be corrected:

(1) There are repeated subheadings;

(2) Some sentences are not concise and accurate enough, and even contain grammatical errors and semantic confusion, which makes me wonder what you want to express.

2. The introduction of this paper lacks a meaningful analysis of the current research situation. It can be carried out according to the following steps:<1>After a meaningful analysis of the current research situation, it provides a very broad and comprehensive perspective;<2>to cut into the topic, analyze and summarize the algorithm cases of other scholars, and highlight your contributions;<3>to focus on the problems and difficulties faced by the current research, and put forward your own views and solutions.

3. Vision technology applications in various engineering fields, should also be introduced for a full glance of the scope of related areas. For object detection, please refer to Novel visual crack width measurement based on backbone double-scale features for improved detection automation.

4. The second part of the paper has too much theoretical knowledge, only qualitative description, and lacks quantitative and theoretical analysis. Readers may lack interest in the paper.

5. The title of the three-line table in the third part of the paper should be aligned in the middle, and the subtitles in 3.2.3 and 3.2.4 should be repeated.

6. In the final conclusion, we should not simply restate the experimental results of the algorithm model. It should be based on experiments and analyze more in-depth and instructive conclusions.

 

 

Author Response

Point 1: In terms of content, the following problems need to be corrected:

(1) There are repeated subheadings;

(2) Some sentences are not concise and accurate enough, and even contain grammatical errors and semantic confusion, which makes me wonder what you want to express.

Response 1: We will carefully review the manuscript and correct the repeated subheadings, clarify sentences that lack precision or contain grammatical errors, and ensure that the paper's meaning is clear and unambiguous.

 

Point 2: The introduction of this paper lacks a meaningful analysis of the current research situation. It can be carried out according to the following steps:<1>After a meaningful analysis of the current research situation, it provides a very broad and comprehensive perspective;<2>to cut into the topic, analyze and summarize the algorithm cases of other scholars, and highlight your contributions;<3>to focus on the problems and difficulties faced by the current research, and put forward your own views and solutions.

Response 2: We agree that the introduction needs to provide a more meaningful analysis of the current research situation. We will follow your suggested steps to provide a comprehensive overview of the research field, analyze and summarize the algorithm cases of other scholars, and highlight our contributions to the field. We will also focus on the problems and difficulties faced by the current research and provide our own solutions.

 

Point 3: Vision technology applications in various engineering fields, should also be introduced for a full glance of the scope of related areas. For object detection, please refer to Novel visual crack width measurement based on backbone double-scale features for improved detection automation.

Response 3: We agree that the paper should include an introduction to the applications of vision technology in various engineering fields, and we will reference the paper you suggested, "Novel visual crack width measurement based on backbone double-scale features for improved detection automation. Tang et al. [26] proposed a new fracture trunk thinning algorithm and width measurement scheme, which can improve the detection automation and has potential engineering application value.

 

Point 4: The second part of the paper has too much theoretical knowledge, only qualitative description, and lacks quantitative and theoretical analysis. Readers may lack interest in the paper.

Response 4: We will add quantitative and theoretical analysis to the second part of the paper to make it more engaging and informative to readers.

 

Point 5: The title of the three-line table in the third part of the paper should be aligned in the middle, and the subtitles in 3.2.3 and 3.2.4 should be repeated.

Response 5:  We will align the title of the three-line table in the third part of the paper in the middle and correct the subtitles in sections 3.2.3 and 3.2.4.

 

Point 6: In the final conclusion, we should not simply restate the experimental results of the algorithm model. It should be based on experiments and analyze more in-depth and instructive conclusions.

Response 6:  We agree that the final conclusion should be based on more in-depth and instructive conclusions drawn from the experiments. We will ensure that the conclusion section provides meaningful insights and valuable guidance for future research.

5. Conclusions

This paper proposed a three-stage state recognition method for Assembled Structures (AS) using Convolutional Neural Networks (CNNs) and data fusion. The method was evaluated on shaking table vibration data from a three-story AS subjected to different wave patterns and earthquake excitations. The experimental results demonstrate the effectiveness and generalization capability of the proposed method, with CNN-3 achieving the highest overall identification accuracy (IA) of 96.3% for training samples and 94.0% for test samples.

The proposed method has the potential to significantly improve the safety and maintenance of assembly structures by accurately identifying their state and facilitating informed decision-making. However, further research is required to fully validate its effectiveness under different conditions and in real-world scenarios. Future studies should focus on testing the method on various types of AS under different environmental conditions, while also comparing its performance to other existing methods. By doing so, we can better understand the overall accuracy, reliability, and efficiency of the approach and its potential for practical implementation.

Author Response File: Author Response.docx

Reviewer 3 Report

Dear Author,  

I have carefully reviewed your work and found it to be a valuable contribution to the field. Your study is well-designed and executed, and the results are clearly presented and analyzed. 

I recommend your manuscript for publication in its present form.

Author Response

 

Dear Reviewer,

Thank you very much for taking the time to review my work and for your positive feedback. I am glad that you found my study to be a valuable contribution to the field and that you believe that it is suitable for publication in its present form.

I appreciate the time and effort that you have put into reviewing my manuscript and providing your thoughtful comments. Your feedback has been invaluable in improving the quality of my work. I will carefully consider any suggestions you may have for further improvement, and will do my best to address them.

 

Once again, thank you for your valuable feedback and support.

 

Best regards

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Some comments have been achieved, such as "the paper did not mention the process used for parameter tuning, and it is not clear whether the proposed method's performance is optimal." and "the paper did not mention the process used for parameter tuning, and it is not clear whether the proposed method's performance is optima". I am quite OK and satisfied for the answers and revisions.  

However, the revised manuscript is not complete with every issue from the comments of the reviewer, such as  "it is suggested that the paper should include and discuss literature reviews on previous works related to classical neural networks. This would enhance the paper's content and provide a more comprehensive overview of the field. For instance, the following papers related to classical neural networks should be included and discussed in detail: Conditional Random Fields Meet Deep Neural Networks for Semantic Segmentation: Combining Probabilistic Graphical Models with Deep Learning for Structured Prediction. IEEE Signal Process. Mag. 35(1): 37-52 (2018), Vision-Based Detection of Guitar Players' Fingertips Without Markers. CGIV 2007: 419-428, LSTM recurrent networks learn simple context-free and context-sensitive languages. IEEE Trans. Neural Networks 12(6): 1333-1340 (2001), and Unsupervised Learning in LSTM Recurrent Neural Networks. ICANN 2001: 684-691." Some previous recommended works still do not mention and discuss clearly. These should not be missed. In conclusion, I recommend the authors revise again for reconsideration.

Author Response

Dear Reviewer,

Thank you for taking the time to review our manuscript and providing us with valuable feedback. We appreciate the constructive comments and suggestions you have provided, and we have taken them into consideration while revising our manuscript.

We are sorry to hear that our revised manuscript did not address every issue that you had raised, particularly with regards to the literature review on previous works related to classical neural networks.

Upon reviewing our revised manuscript, we realized that we had missed out on some of the recommended papers that you had mentioned. We will ensure that these papers are included and discussed in detail in our manuscript.

Thank you once again for your time and feedback.

Best regards

Author Response File: Author Response.docx

Round 3

Reviewer 1 Report

The revised manuscript is now better and ready to be published. I recommend that it should be accepted.