A Novel Cementing Quality Evaluation Method Based on Convolutional Neural Network

Round 1

Reviewer 1 Report

Article "A Novel Cementing Quality Evaluation Method Based on Convolutional Neural Network" by Chunfei Fang, Zheng Wang, Xianzhi Song, Zhaopeng Zhu, Donghan Yang, Muchen Liu is relevant and written on an important issue. Done at a good level.

However, after reading, there were comments to the authors:

1) The main remark is to expand the introduction, since 4 references to works are not enough to disclose the problem studied in this paper.

2)Why did the authors decide to test the stability of the model described in another work (https://doi.org/10.1007/978-3-319-75193-413 1_50) by adding 3%, 6% and 9% white to the variable density image noise? What is the reason for such a step of 3%?

3) And how much white noise must be added in the end for an optimal result?

Author Response

Thank you for your valuable suggestions. I think it is very useful. I have revised the original article and expanded the introduction. For the research part of adding noise to variable density, I modified it again. I replied in detail in the document to the three modifications you proposed.

Thank you again for your valuable comments！

Author Response File: Author Response.docx

Reviewer 2 Report

1.       What are the full names of CBL and VDL?

2.       The sentence in line3 37 to 39 is not a logical expression.

3.       The expression in line 44 conflicts with the expression in line 47.

4.       The sentence in line 53 is confusing.

5.       Please revise the expression in Lines 66 and 67 and make it reasonable.

6.       What do T, R1, and R2 represent in Figure 1?

7.       The reason why the VDL brightness needs to be adjusted should be explained.

8.       In Line 164, “On the contrary” is not correct here.

9.       Table 1 lists four categories of cementing quality, but according to the response characteristics mentioned in Section 2.1.1, it should be five categories. Why do just take four categories?

10.   What is the filter size in the last Convolution layer in Figure 9? Figure 9 is a part of the proposed model. Please provide a complete architecture graph to show all components used in your model, including feature extraction, feature fusion, and classification.

11.   The procedure for feature fusion and quality classification should be added to the paper.

12.   Please add graphs to show accuracy convergence for VGG16, VGG19, ResNet-18, ResNet-34, and AlexNet methods.

13.   Accuracy from the multi-scale perception model is 0.90, but the accuracy in Figure 10 is greater than 0.90. Please explain the reason.

14.   It is better to use M (means 100000) as the unit of time complexity and space complexity in Table 2.

Author Response

Thank you for your valuable comments. I have revised my paper. Please see the attachment.

Author Response File: Author Response.docx