Nonlinear Dynamic Process Monitoring Using Canonical Variate Kernel Analysis

Round 1

Reviewer 1 Report

There is a very small number of recent references (5 out of 21), considering the last five years (2018-2022). This makes it difficult to position the contribution of the work in the state of the art. I strongly suggest the authors include some comments about recent references regarding the theme of the proposed work.

 

In the list of references, the fonts of some years are in boldface style and others are not. I suggest uniform formatting.

 

I suggest the authors use the term "dynamical systems" instead of "dynamic systems". The latter one is widely used in the manuscript.

 

The same suggestion applies for the "dynamic kernel".

 

The text needs comprehensive English language revision. For instance, on Page 2, 

 

- line 59: The sentence "... In a previous work, we has combined..." ;

 

I suggest authors improve the quality of the graphs shown in Figures 3-7.

 

 

Page 12, line 238: Authors need to be extremely careful with statements like "... Hence, CVKA is a more efficient method than others." The proposed method is applied in only one type of system. What needs to be discussed and highlighted is that this performance was achieved in the type of dynamical system considered in the case of study of the present work. Besides, many aspects may be considered before overestimating the performance of systems and algorithms.

 

Author Response

Reviewer: 1

1.1 Comments to the Author

There is a very small number of recent references (5 out of 21), considering the last five years (2018-2022). This makes it difficult to position the contribution of the work in the state of the art. I strongly suggest the authors include some comments about recent references regarding the theme of the proposed work.

1.1 Response:

Thank you for your advice and several new relevant references published in the last five years have been added and compared in the revised manuscript.

1.2 Comments to the Author

In the list of references, the fonts of some years are in boldface style and others are not. I suggest uniform formatting.

1.2 Response:

It has been revised and all the years are in boldface style.

1.3 Comments to the Author

I suggest the authors use the term "dynamical systems" instead of "dynamic systems". The latter one is widely used in the manuscript. The same suggestion applies for the "dynamic kernel".

1.3 Response:

The relevant expression has been modified

1.4 Comments to the Author

The text needs comprehensive English language revision. For instance, on Page 2, - line 59: The sentence "... In a previous work, we has combined..." ;

1.4 Response:

We are sorry for some language errors. The revised manuscript has been carefully checked through to ensure correctness and accuracy.

1.5 Comments to the Author

I suggest authors improve the quality of the graphs shown in Figures 3-7.

1.5 Response:

We have used clearer graphs in place of the original graphs.

1.6 Comments to the Author

Page 12, line 238: Authors need to be extremely careful with statements like "... Hence, CVKA is a more efficient method than others." The proposed method is applied in only one type of system. What needs to be discussed and highlighted is that this performance was achieved in the type of dynamical system considered in the case of study of the present work. Besides, many aspects may be considered before overestimating the performance of systems and algorithms.

1.6 Response:

Thank you for your advice and we have emphasized that the superiority of the proposed method is reflected in the dynamical nonlinear system for TE-like process.

Author Response File: Author Response.docx

Reviewer 2 Report

In this paper, a novel nonlinear dynamic process monitoring technique is proposed in which KPCA is applied to the CVA residual space, called CVKA. The new CVKA is then applied to a TE process case study, proving the excellent performance of CVKA to other common approaches. However, the following problems still exist

(1) Through reading this paper, many spelling typos and grammatical errors can be found, it is suggested to proofread your paper carefully. For example, only "as follow" is followed by ":". Punctuation marks exist after mathematical equations. Articles should pay attention to grammatical details.

(2) Formulas (7), (8), and (9) need to be reformulated, and the same pattern is not repeatedly written.

(3) Some sentences do not read smoothly, which should be checked by authors to make it more understandable to the readers. Please correct these.

(4) In the first part of this paper, when the current status of research on dimensionality reduction methods in multivariate statistical analysis, is there any other research on dimensionality reduction methods?

(5) Section 1 cites some references, but it does not provide a sufficiently exhaustive overview and critical discussion of the state of the art of related literature. Moreover, as we know, PCA is a kind of dimensionality reduction in Multivariate statistical process monitoring. Therefore, it is recommended to refer to some work in the "Deep PCA based real-time incipient fault detection and diagnosis methodology for PCA electrical drive in high-speed trains" literature.

(6) The literature review is poor in this paper. I hope that the authors can add some new references in order to improve the reviews and the connection with the literature. It is suggested to add some literature on fault detection in the first part of this paper, for example, 10.1109/TCYB.2021.3060766 and 10.1109/TCYB.2022.3163301.

(7) Figure 2 shows the flowchart of the proposed algorithm, but the authors should summarize the offline training and online monitoring as Algorithm 1 and Algorithm 2 through pseudo-code, respectively, to facilitate the reader's understanding.

(8) Do not use faint lines and/or lettering and check that all lines and lettering within the figures are legible at the final size.

(9) To ensure the accuracy and intuitiveness of the experiment, I would like the authors to simulate the fault detection results of the proposed method for different faults.

(10) The conclusions and motivations for the work should be added in a clearer way.

Author Response

Reviewer: 2

In this paper, a novel nonlinear dynamic process monitoring technique is proposed in which KPCA is applied to the CVA residual space, called CVKA. The new CVKA is then applied to a TE process case study, proving the excellent performance of CVKA to other common approaches. However, the following problems still exist.

Response: We thank the reviewer for these excellent advices. All comments have been considered as explained below one by one.

2.1 Comments to the Author

Through reading this paper, many spelling typos and grammatical errors can be found, it is suggested to proofread your paper carefully. For example, only "as follow" is followed by ":". Punctuation marks exist after mathematical equations. Articles should pay attention to grammatical details.

2.1 Response:

We are sorry for some spelling typos and grammatical errors. The revised manuscript has been carefully checked through to ensure correctness and accuracy.

2.2 Comments to the Author

Formulas (7), (8), and (9) need to be reformulated, and the same pattern is not repeatedly written.

2.2 Response:

We thank the reviewer for this advice. We revised the expression and preserve one formula to avoid repetition.

2.3 Comments to the Author

Some sentences do not read smoothly, which should be checked by authors to make it more understandable to the readers. Please correct these

2.3 Response:

We are sorry for the carelessness. We have checked this revised manuscript several times again to ensure accuracy and polished it for good readability.

2.4 Comments to the Author

In the first part of this paper, when the current status of research on dimensionality reduction methods in multivariate statistical analysis, is there any other research on dimensionality reduction methods?

2.4 Response:

Of course, there are some other dimensionality reduction methods such as PLS, ICA et al. They are all static methods. Although some studies have extended them to monitor dynamic processes, the performances of them are not as good as CVA.

2.5 Comments to the Author

Section 1 cites some references, but it does not provide a sufficiently exhaustive overview and critical discussion of the state of the art of related literature. Moreover, as we know, PCA is a kind of dimensionality reduction in Multivariate statistical process monitoring. Therefore, it is recommended to refer to some work in the "Deep PCA based real-time incipient fault detection and diagnosis methodology for PCA electrical drive in high-speed trains" literature.

2.5 Response:

Thank you for the recommendation and the work in literature "Deep PCA based real-time incipient fault detection and diagnosis methodology for PCA electrical drive in high-speed trains" is worthy of reference and has been referred in page 2.

2.6 Comments to the Author

The literature review is poor in this paper. I hope that the authors can add some new references in order to improve the reviews and the connection with the literature. It is suggested to add some literature on fault detection in the first part of this paper, for example, 10.1109/TCYB.2021.3060766 and 10.1109/TCYB.2022.3163301.

2.6 Response:

More relevant research has been added in the literature review including 10.1109/TCYB.2021.3060766 and 10.1109/TCYB.2022.3163301.

2.7 Comments to the Author

Figure 2 shows the flowchart of the proposed algorithm, but the authors should summarize the offline training and online monitoring as Algorithm 1 and Algorithm 2 through pseudo-code, respectively, to facilitate the reader's understanding.

2.7 Response:

This is a worthwhile suggestion. Before submitting the manuscript, the pseudo-code was removed given its similarity to algorithm flowcharts. Now the pseudo-code is added on page 8.

2.8Comments to the Author

Do not use faint lines and/or lettering and check that all lines and lettering within the figures are legible at the final size.

2.8 Response:

Relevant figures has been revised to be as clear as possible.

2.9 Comments to the Author

To ensure the accuracy and intuitiveness of the experiment, I would like the authors to simulate the fault detection results of the proposed method for different faults.

2.9 Response:

In TE process, we monitored all of 20 different faults available in the benchmark process. We may not have made it clear that there are 4 types, e.g. step, random, slow drift and sticking faults plus 5 unknown faults. The detection results of different faults can ensure the accuracy and intuitiveness of the experiment.

 

2.10 Comments to the Author

The conclusions and motivations for the work should be added in a clearer way.

2.10 Response:

We have fully taken this suggestion into consideration. Relevant content has been modified on page 12.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

I would like to thank the authors for their effort to improve the manuscript and enrich the presentation of the proposal. The current version of the manuscript fulfills all the fragile issues raised by this reviewer relative to the first version. I consider that the authors' work on this version has reached a sufficient level to be considered a journal article.

Reviewer 2 Report

The reviewer appreciates the revision from all the authors. Now, my suggestion is "accept."