Damage Diagnosis of Frame Structure Based on Convolutional Neural Network with SE-Res2Net Module

Round 1

Reviewer 1 Report

 

In this paper, damage diagnosis of frame structure based on convolutional neural network with SE-Res2Net module is proposed. The paper is well organized but several parts of paper are exactly copied from [R1], the authors should carefully revised the paper and modified these parts. The paper can be reconsidered after major revision. Some comments are listed as follows:

 

 

 

-There are some typos should be modified for example

a. In line 18,   “30and 99.78 % when” a space must be added between numbers and letters.

 

b. There is no Table 1. (Table 1. Selection basis of the step size of the first convolution layer. line 206)

 

 

-Quality of Fig.8 and Fig.9, should be improved.

 

-Plagiarism check index is so high and several parts of papers are exactly copied from other works, like [R1] which should be modified.

 

- Figures which are not belong to authors should be cited. For example Fig7. is copied from [R1] and [R2].

 

 

-Training curves versus epoch should be added and sufficient explanations about training, and learning rate of the applied algorithm should be added.

 

-Table 3 are exactly copied from [R1], which should be modified.

 

 

-Provide schematic diagram of damage location diagnosis.

 

 

[R1] Ren J, Cai C, Chi Y, Xue Y. Integrated Damage Location Diagnosis of Frame Structure Based on Convolutional Neural Network with Inception Module. Sensors. 2022 Dec 30;23(1):418.

 

[R2] Xu H, Cai C, Chi Y. Damage Location Diagnosis of Frame Structure Based on a Novel Convolutional Neural Network. Applied Sciences. 2022 Jan;12(23):12411.

 

Author Response

Thank you for your comment. Our detailed answers can be found in the attachment.

Author Response File: Author Response.doc

Reviewer 2 Report

The paper presented by the authors is interresting and well structures.

Although it needs a further English reading and verification.

Authors must add table1 that is missing.

Author Response

Thank you for your comment. Our detailed answers can be found in the attachment

Author Response File: Author Response.doc

Reviewer 3 Report

In the manuscript entitled “Damage Diagnosis of Frame Structure Based on Convolutional Neural Network with SE-Res2Net module” the authors proposed a new model for damage diagnosis of frame structures by implanting squeeze-and-excitation neural network (SENet) and Res2Net modules. The study is relevant to the journal's audience but has several shortcomings related to writing, text organization, deepening the results, including referenced comparisons. Therefore, the manuscript should be reconsidered after a major review.

 

1) Improve the introduction. Some examples:

a.       Introduzir melhor o assunto de “frame struture”;

b.      In the sentence: “Moreover, the working environment of the frame structure is harsh, and it is susceptible to vibration, shock, high temperature, corrosion and other factors [3]”, explain why.

c.       Improve this sentence: “The use of frame structure involves the safety of personnel…”

d.      In the sentence: “There are some problems when using the traditional method to diagnose the frame structure because of the complex composition.”, what are the traditional methods? It wasn't clear.

2) In section 2, the authors review the main modules without going into too much depth on the topics. I suggest checking the need for this section and if it is really necessary, the authors should rewrite it, leaving the information clearer, more precise and in-depth.

3) Ideally, a manuscript should have the following sections: Introduction, experimental and/or theoretical methods, results and discussion, conclusion, references, etc.

4) Figures need to be better described in the text, such as Figure 5.

5) Figures 8 and 9 need to be divided into a) and b). What is the difference between the two graphics? Authors should better describe graphics in the text.

6) Figures in general need to have more explanatory captions, as is the case of Figure 11, which presents various data.

Author Response

Thank you for your comment.Our detailed answers can be found in the attachment.

Author Response File: Author Response.doc

Reviewer 4 Report

In this paper the authors discuss the damage diagnosis of a frame structure using squeeze and excitation neural network.
They have demonstrated the gains in the prediction using the method for the frame structure they have used.

The methodology of the paper seems to be correct and the advantages of the prediction method are clearly stated. In that
regard the work could be publishable. However, the authors should address the following points:

1. How far the method is general? They have used a particular structure from the University of British Coloumbia. How far
do they think the methods are applicable for other similar systems?

2. The paper uses a lot of acronyms. Although those are all defined, it restricts the flow of reading. At least for the
conclusion section, it would be better to write in more general term.

Author Response

Thank you for your comment.Our detailed answers can be found in the attachment.

Author Response File: Author Response.doc

Round 2

Reviewer 1 Report

Authors have addressed most of my concerns.

Reviewer 3 Report

The authors have answered all the questions/queries and revised/improved the manuscript accordingly. In my opinion, the present manuscript can be published in the Applied Sciences.