Diagnosis of Intermittently Faulty Units at System Level
Round 1
Reviewer 1 Report
The manuscript presents an interesting method for the diagnosis of intermittently faulty units.
One major point that could be further improved is the comparison with state of the art methods.
Author Response
One major point that could be further improved is the comparison with state of the art methods.
We thank you for this important remark.
To overcome this limitation, we have added a separate section (2. Related work), where we will briefly describe the modern methods of solving this problem. We believe that this additional section will allow the readers to better understand the methods we have proposed and to facilitate the understanding of the results obtained.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Congratulations. After the corrections and addition of the appendices this interesting paper has been improved significantly.
I would like to recommend this manuscript to be published.
Author Response
Dear Reviewer!
Thank you for your appreciation of our work. We once again read the comments of all reviewers and decided to add an additional section, which briefly describes modern methods for solving this problem (2. Related work).
Taking this opportunity, and once again, I want to thank you for the support.
Author Response File: Author Response.docx
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
Thank you for inviting me as a reviewer for manuscript titled Diagnosis of intermittently faulty units at system level. This paper presents interesting research of system level diagnosis. In my opinion the paper is almost ready for the publication. The paper would be more exiting if you implement below improvements:
- Line 15-23 - Abstract needs to be rewritten. Abstract is well written and includes introduction, problem statement, methodology, contributions and results, but abstract should be presented in no more then 250-300 words.
- In introduction section authors should clearly present gap in considered scientific area. Try to specific your research questions and build a case for your research. Please, focus on novelty. This should be presented in at least one paragraph.
- Please, clarify research questions in your research.
- Figures 2 and 6 – Provide figures in high resolution.
- Remove red highlights from your paper.
- Provide detail discussion what your model brings more then existing models in literature. What did you bring more? Add more detailed discussion in section 3.
- Please summarise the advantages and limitations of the proposed method in practical applications.
- The conclusion section also seems to rush to the end. Clearly state your unique research contributions in the conclusion section. Provide some future directions.
Author Response
Dear colleagues,
thank you very much for your review, the questions considered and the work you have done. We reviewed the comments you made, tried to eliminate them and to maximize answer your questions.
1. - Line 15-23 - Abstract needs to be rewritten. Abstract is well written and includes introduction, problem statement, methodology, contributions and results, but abstract should be presented in no more then 250-300 words.
Our reaction: first sentence of abstract was removed.
2. In introduction section authors should clearly present gap in considered scientific area. Try to specific your research questions and build a case for your research. Please, focus on novelty. This should be presented in at least one paragraph.
Our reaction:
After line 67 and before line 68 was inserted the following text:
Basis for diagnosis intermittent faults at system level was developed by S. Mallela and G. Masson [11]. They determined the condition (in this paper, denoted as RS ÎR0) which should be used in the diagnosis algorithms taking into account intermittent faults.
It worth noting that, authors did not consider the problem of implementation of the suggested method. In this paper, we present efficient algorithm allowing to perform verification of above mentioned condition. Moreover, we also have determined what should be done when the condition is not satisfied.
3. Please, clarify research questions in your research.
Our reaction:
The following text was inserted in conclusion after line 429 :
As compared to the existing approaches used for diagnosis intermittent faults at system level, we endeavored to make the implementation of algorithm which checks condition RS ÎR0 more efficient and simpler. We also suggest the way of how to resolve the situation when this condition is not met.
4. Figures 2 and 6 – Provide figures in high resolution.
Our reaction: done
5. Remove red highlights from your paper.
Our reaction: partly was done. Remaining part was corrected by adding appropriate references.
6. Provide detail discussion what your model brings more then existing models in literature. What did you bring more? Add more detailed discussion in section 3.
Our reaction:
Fig.7 was added.
Fig. 7 summarizes the presented approach to diagnosis at system level when units can be intermittently faulty. We only solve the task of how to improve the implementation of condition verification and suggest the right part (see Fig.7) of algorithm.
7. Please summarise the advantages and limitations of the proposed method in practical applications.
Our answer:
The suggested method can be used in the complex systems for which self-diagnosis at system level can be applied. It means that system unit should be able to perform tests on other units and execute diagnosis algorithm. Such systems as multi-agent systems, sensor networks, many-core processors etc. are met such requirement.
8. The conclusion section also seems to rush to the end. Clearly state your unique research contributions in the conclusion section. Provide some future directions.
Our reaction:
The following text was inserted in conclusion after line 429:
As compared to the existing approaches used for diagnosis intermittent faults at system level, we endeavored to make the implementation of algorithm which checks condition RS ÎR0 more efficient and simpler. We also suggest the way of how to resolve the situation when this condition is not met.
Thank you for your help, which allowed us to improve our article.
With deep respect,
Reviewer 2 Report
It is an interesting paper on fault diagnosis considering that both permanently and
intermittently faults can occur in the system. However the authors need to address the
following comments:
1. Several recent related papers of the area of model-free fault diagnosis are not cited nor compared to, such as a) Fault Identification in Distributed Sensor Networks Based on Universal Probabilistic Modeling, b) Fault diagnosis for smart grids in pragmatic conditions, and references therein. The authors need to highlight the differences between this line of fault diagnosis,
motivate their approach and point out the novelties of the present solution.
2. Figures are of low quality
3. The application scenario is not clear. Is it regarding only synthetic data? The authors should used real data too, such as http://roveri.faculty.polimi.it/software-and-datasets/
4. The figures of merits are not in line with the related literature. The authors should use false positive, negative, delay, etc. as done in the above mentioned papers
5. The most important point that needs to be addressed is comparison with related literature.
Right now, there is none so it is impossible to assess the relevance of this work.
Author Response
Dear colleagues, thank you very much for your review, the questions considered and the work you have done.
We reviewed the comments you made, tried to eliminate them and to maximize answer your questions.
1) Several recent related papers of the area of model-free fault diagnosis are not cited nor compared to, such as a) Fault Identification in Distributed Sensor Networks Based on Universal Probabilistic Modeling, b) Fault diagnosis for smart grids in pragmatic conditions, and references therein. The authors need to highlight the differences between this line of fault diagnosis, motivate their approach and point out the novelties of the present solution.
Our reaction:
The following text was inserted after line 70 and before line 71:
It is worth noting that, diagnosis of system units at system level considers a fault in the system as a failure of a system unit. We do not consider the details what was wrong with the unit which has failed. In view of this, we do not classify the faults as it was done, for example in [x1], [x2]. In essence, we consider each system unit as either failed or correct (which is denoted as “1” or “0”). Failure of system becomes when total number of failed units exceed the maximum value determined in advance.
x1. S. Ntalampiras. Fault Identification in Distributed Sensor Networks Based on Universal Probabilistic Modeling. IEEE Transactions on Neural Networks and Learning Systems, 2014, pp.1939-1949, DOI: 10.1109/TNNLS.2014.2362015
x2. S. Ntalampiras. Fault Diagnosis for Smart Grids in Pragmatic Conditions. IEEE Transactions on Smart Grids, 2016, DOI: 10.1109/TSG.2016.2604120
We also
After line 67 and before line 68 have inserted the following text:
Basis for diagnosis intermittent faults at system level was developed by S. Mallela and G. Masson [11]. They determined the condition (in this paper, denoted as RS ÎR0) which should be used in the diagnosis algorithms taking into account intermittent faults.
It worth noting that, authors did not consider the problem of implementation of the suggested method. In this paper, we present efficient algorithm allowing to perform verification of above mentioned condition. Moreover, we also have determined what should be done when the condition is not satisfied.
We also
have inserted after line 404 the following text:
The task of determining the probability of diagnosis result (fault omission and incorrect identification of fault-free units) is a separate task and it was not subject of this paper.
Fig. 7 summarizes the presented approach to diagnosis at system level when units can be intermittently faulty.
Figure 7. Diagnosis of system units when intermittent faults are allowed.
We also
have inserted in the conclusion after line 429 the following text:
As compared to the existing approaches used for diagnosis intermittent faults at system level, we endeavored to make the implementation of algorithm which checks condition RS ÎR0 more efficient and simpler. We also suggest the way of how to resolve the situation when this condition is not met.
2) Figures are of low quality
Our reaction: done.
3) The application scenario is not clear. Is it regarding only synthetic data? The authors should used real data too, such as http://roveri.faculty.polimi.it/software-and-datasets/
Our answer:
we deal with diagnosis at system level (i.e. the highest level of abstraction). At this level, we are abstracted from all details. We do not consider both the unit structure and details of interactions among system units. They are considered as atomic. We work with the model which includes only “1” and “0”. The problem consists in decoding the set of 1 and 0 so that to determine the units which should be considered as correct or failed. We do not research why a unit has failed.
4) The figures of merits are not in line with the related literature. The authors should use false positive, negative, delay, etc. as done in the above mentioned papers
Our answer:
System level self-diagnosis focus on credibility of obtained results. This credibility is expressed via parameter t. If the total number faulty units do not exceed this parameter, then diagnosis result should be considered as trustworthy.
5. The most important point that needs to be addressed is comparison with related literature.
Our reaction:
The most important items are shown in Fig.7 (was added). Left part of Figure was suggested by Mallela and Masson. We only try to improve its implementation. The right part of Figure was suggested by us. We aimed at achieving the diagnosis result even in the situations when existing diagnosis algorithms do not able to cope with it. We did not mention in this paper these diagnosis algorithms. There are many such algorithms.
Thank you for your help, which allowed us to improve our article.
With deep respect,
Volodymyr Lytvynenko
Reviewer 3 Report
The proposed research is not clear to the reader.
The state of the art presented on the paper is old.
The proposed method is poorly described, since no formalism is presented and unclear code is presented.
The flow of the paper is unclear and the sections names do not reflects its content.
The research presented was not validated.
No valid conclusions are presented.
Author Response
Dear Reviewer, thank you very much for your review, the questions considered and the work you have done.
We reviewed the comments you made, tried to eliminate them and to maximize answer your questions.
We also took into account the comments of previous reviewers, so the revised article is significantly different from the one that was originally sent to you. In this version of the article, we sought to eliminate the comments of all three reviewers. Thank you again for the comments you made to improve this article.
1)The state of the art presented on the paper is old.
Our reaction:
In the paper, we researched how to improve the implementation of the suggested by Mallela and Masson method. Thus, we used the restricted list of references.
2)The proposed method is poorly described, since no formalism is presented and unclear code is presented.
Our reaction:
Code in Julia is working and system parameters for its implementation are determined in the paper.
3)The flow of the paper is unclear and the sections names do not reflects its content.
Our reaction:
First, we presented the model of imtermittent fault and then we suggest how these faults can be diagnosed. We consider that this way is correct. We also added some content to the paper and made some corrections.
4)The research presented was not validated.
Our reaction:
Table 2 presents the characteristics needed for implementation of the suggested code. We have tested our code.
5)No valid conclusions are presented.
Our reaction:
Conclusions were extended.
Thank you for your help, which allowed us to improve our article.
With deep respect,
prof. Volodymyr Lytvynenko
Round 2
Reviewer 1 Report
I am very happy that the authors have addressed my concerns point by
point precisely. No further suggestions come from my side. Therefore, I
would like to recommend this manuscript to be published.
Reviewer 2 Report
I would like to thank the authors for implementing my comments. I think that the article has improved substantially. After a careful proofreading, it could be accepted for publication.
