An Explainable DL-Based Condition Monitoring Framework for Water-Emulsified Diesel CR Systems

Round 1

Reviewer 1 Report

1. Page1 Line 14 with an average validation accuracy of 95.9%
2. Page1 Line 22 “which contribute nearly 30% of greenhouse effects and a host many health and environmental problems” I don’t understand the meaning of “a host” here.
3. Page 1 Line 31 causing emissions while maintaining engine efficiency.
4. Page 2 Line 51 CNN is a subset of DNN. They shouldn’t be listed together.
5. page 5 Line 125, mu is the mean of the distribution.
6. In equation 4, the definition of Softmax is wrong. Please double-check it.
7. In Figure 7, looks like the accuracy and loss don’t change much during the training. Could the authors comment on this?

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

For water-emulsified diesel CR systems, this paper proposes an interesting condition monitoring framework. The structure of this paper is well arranged, but there are still some errors. In general, reviewer suggest minor revisions.

1. It is recommended that figures A6, A7, A8, A9, and A10 have the same size.
2. Please give comparative experiments to reflect the superiority of the algorithm proposed in this paper.
3. Please list the training time of the model proposed in this paper.
4. The reviewer suggests that the author make a profound summary in the conclusion and add further research tasks.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper present a new, interesting approach for condition monitoring of common-rail diesel engines based on convolutional neural networks.

The use of LIME is a great addition to improve the model trustworthiness.

I have only minor remarks regarding the work.

1 - What is the practical application of this algorithm? One application that comes to mind is fraud prevention, or to evaluate the separation of the emulsion during operations. In any case, I would like to know the authors' take on this.

2- I'd like the authors to better explain CNNs, describing at a high level how this particular type of neural networks operate in a general sense before describing this particular application case, expanding on Subsection 3.2.

3 - Referring to Fig.8 (the correlation matrix), the comment is incorrect. In fact, if conventional approach is used, uncorrelated features should have a correlation factor close to 0, rather than a negative value close to -1 (high negative correlation). Please correct the comments on lines 252-254 to address this.

4 - Finally, some minor english typos are present throughout the text, e.g. lines 252 "uncorrleated", "dicriminnance", etc.

 

 

 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 4 Report

This is a good paper, that addresses an interesting problem and gives a solution based on convolutional networks. The paper well organized and written, clear and to the point. The experimental results indicate the suitability of the proposed method for the condition monitoring of the emulsified Diesel CR system.

Author Response

We appreciate the reviewer for carefully reviewing our manuscript. We are glad that our manuscript interests the reviewer.