Detection and Isolation of Incipiently Developing Fault Using Wasserstein Distance

Round 1

Reviewer 1 Report

The proposed study shows the significance of Wasserstein Distance in faults isolation. But the study should also incorporate some points and address the following comments for possible publications:

• The effectiveness of Wasserstein Distance is evaluated from the simulated dataset of CSTR. Validate the effectiveness and efficiency of the Wasserstein Distance using a real-time dataset. 
• Also, nowadays many machine learning and explainable AI techniques are used for fault identification and isolation. Then how could the Wasserstein Distance proves to be significant in these recent scenarios? 
• Compare the performance matrices of Wasserstein Distance with the other AI/ML techniques then conclude the final outcome. 
• Also, in between the articles, the information flow is missed. Kindly restructure those sentences or paragraphs.

  Update your literature review to include recent and important related review and research articles.  Make critical analysis and comparison of the literature in order to justify your research.

Describe clearly the innovation in this paper as compared to the literature and your previously published papers.

Author Response

We would like to thank Reviewer 1 for the helpful suggestions for our article. We appreciate the insightful and constructive comments that improve the presentation of the material described in our article. We believe that our modifications adequately address these concerns and appended a point-by-point response to the comments. Please see the attachment for point-by-point response.

Author Response File: Author Response.doc

Reviewer 2 Report

This paper develops the idea of Wasserstein distance, starting with Gaussian variables and ending with two practical examples: linearly dependent variables and the CSTR process.

There are several significant caveats in the text:

1. I disagree that Figure 3 proves the W approach. In addition to True-positives, I see True-negatives at 1010 and True-negatives at the 1100-1270 range.

 

2. It is unclear how the authors incorporated 44a and 44b knowledge into the W approach. It brings two major concerns: a) there is a chance that authors were testing the stationary phase of the process, which might mislead them; b) no initial and process conditions were presented in the text, so most probably, the dynamics were not fully accounted.

3. In many industrial processes, usually multiple nonlinearities and boundary conditions are to be followed, which are not mentioned in the paper. Therefore, two equations do not mean that it is an industrial process.

4. Moreover, systems know some of the controls and infer some virtual variables. The authors did not discuss how parameters were validated on an actual object, and I recommend dropping the "industrial" term altogether.

5. No variances were provided and validated for inputs and outputs of  44a and 44b formulas. It is unfair to claim that Feed fault in observations is the correct disturbance to analyze. Abrupt pressure changes are an even more complex issue that disturbs several variables at once, and I am not sure that such a small challenge is enough for scientific paper content.

However, I admit that theoretical considerations are engaging in the paper, so depending on how the authors rewrite the text and reconstruct the claims, I will recommend either rejecting this study or accepting it.

 

 

Author Response

We would like to thank Reviewer 2 for the thorough assessment of our manuscript and for offering insightful and constructive comments that improve the presentation of the paper. Please find a point-by-point response to your comments appended in the attachment.

Author Response File: Author Response.doc

Round 2

Reviewer 1 Report

Minor English spell check and grammar check required. 

Some texts are not cited appropriately. Update the same with proper citations. 

Author Response

We would like to thank Reviewer 1 for offering helpful and constructive comments that improve the presentation of the paper. We believe that our modifications adequately address these concerns and appended a point-by-point response to the comments. Please see the attachment for point-by-point response.

Author Response File: Author Response.doc

Reviewer 2 Report

1. I recommend keeping the CSTR part (including the authors' answer on my comment #3) in the text in the addition of the water boiler case. Include your derivations in the paper, too, not just in response to my comments.

2. The furnace wall withstands typically high temperatures, but the bigger problem is its gradient, i.e., the temperature difference, not the absolute temperature.

3. Please provide the principle or technical schematics of the Table 2 variables because workflow and layout are not clear from the text.

4. Are you sure you are using the correct name for "wind"? Isn't it an airflow?

5. The authors did not address my 4th comment on the CSTR case. Please elaborate on it in the text when you reinsert the CSTR.

6. Re-iterate Tables 2 variables through the equations 1-42, presenting the numeric values for the outputs of the equations so that the reader understands the study qualitatively and the reviewer can perform some double-check quantitatively.

 

Author Response

We would like to thank Reviewer 2 for the thorough assessment of our manuscript and for offering constructive suggestions that improve the presentation of the paper. Please find a point-by-point response to your comments appended in the attachment.

Author Response File: Author Response.doc