Fault Diagnosis in Partially Observable Petri Nets with Quantum Bayesian Learning
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsIn the current manuscript, the authors investigate the Quantum Bayesian probability estimation of fault diagnosis based on partially observed petri nets (POPN) for a liquid engine system. It is shown that for the studied system model, the algorithm's response time and the accuracy of the diagnosis are improved with the use of Quantum Bayesian Petri nets (QBPN). The manuscript has a lot of merits but there are some points that need to be considered:
1) Please follow a specific referencing style. For instance, in the Introduction, the authors use two styles of referencing.
2) The authors need to highlight more in the Introduction, the merits of this study and what the novelty is. Has this been studied before? What is the difference between this study and previous studies? This is a major part that needs to be included, as it gives a better understanding to the readers about the significance and the impact of the study.
3) Regarding Figure 1, the figure is small and cannot be read easily. Furthermore, the authors have named the places as p in Table 1 but on the figure these places are stated as s (?). The figure needs to align with the tables, to give a better understanding to the readers.
4) For Figures 7 and 8, the axis titles need to be in English.
5) The language needs to be improved. For instance, "Estimate the system fault probability by the QBPN fault diagnosis method. Firstly, we structure the QBPN models of the fault transition t10 and the fault transition t11 ,respectively, by using our algorithm. Then calculate the firing probability of fault transitions. Choose the maximum probability as the system fault probability.".
This language could be used in an algorithm explanation section, but the main body of the manuscript needs the text to address the readers. So, the above text could be revised as follows "To estimate the system fault probability by the QBPN fault diagnosis method, we firstly structure the QBPN models of the fault transition t10 and the fault transition t11 ,respectively, by using our algorithm. Then, we calculate the firing probability of fault transitions. Finally, we choose the maximum probability as the system fault probability."
Comments on the Quality of English LanguageImprovements are needed. Some sentences need to change based on my previous comment (please see comment 5). The main text should not be like a section that explains algorithmic steps; it should have a flow so the readers can comprehend better.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsI would like to thank the authors for writing up the results of the work and the editor for giving me the opportunity to review the manuscript.
The manuscript "Fault Diagnosis in Partially Observable Petri Nets with Quantum Bayesian Learning" discusses the use of Petri nets in fault detection. The example of a liquid-propellant rocket engine is presented. Its Petri net is only partially observed, so it cannot always be deterministically stated whether a fault transition has occurred given the observed place. Quantum Bayesian probability estimation is used to determine the probability of the fault.
I have not performed in-depth review of the content of the manuscript as the manuscript is not yet ready for that:
1. The language is inadequate. There is a lot of non-standard terminology and grammatically incorrect sentence structures. As a result, the manuscript is too hard to follow for me. For example, the first sentence of the Abstract introduces the term "liquid engine", which is not a standard name for a liquid-propellant rocket engine (it would be much better to use "liquid-propellant rocket engine" first, to let the reader know what the object of the study is, and to call it "engine" when the full name is unnecessary). In the next sentence, it is unclear what the meaning of the first "that" is: what does it refer to and which verb goes with it? Similarly malformed sentences are then common throughout the manuscript. Deciphering the text would thus be too hard for either me or a more typical reader, and one could never be sure if one understood the message correctly or not.
2. The structure isn't suitable. Neither the Abstract nor the Introduction makes it clear what the point of the article is, what is the new result that is being presented (it needs to be stated in both places). Without this piece of information, nobody will have a reason to read the paper. There are also practically no references cited outside of the Introduction. I thus can't tell from the manuscript whether Quantum Bayesian fault diagnosis is novel (and I should check it) or standard (and I should look up the literature in case I have trouble understanding). Same for partially observable Petri nets. Same for the application of Quantum Bayesian fault diagnosis to partially observable Petri nets. The section on methods should be strongly supported with references that your methods are built upon. The discussion section should discuss your results in the context of already published results as well and cite the relevant references (by the way, the discussion is missing). The Conclusion section summarises some findings about the QBPN model but it is unclear which findings are originally yours and which ones are well known but supported by your work.
I have my doubts about the Quantum Bayesian computations. It feels like they may be in disagreement with the probability axioms. Since I don't know how original the claims are, I don't know whether I should clarify my doubts by studying the relevant literature or through scrutinising the presented results.
Regarding the content details, I only have a few remarks until I study an easier-to-follow version of the manuscript:
- The notation log(), introduced just before Definition 2 in section 2.2, seems not to imply a logarithm. If so, please rename it.
- The citations in the text don't follow a single system, some are given as [number] and some as Author (year).
- There is no data availability statement, which is a significant omission for a simulation study. Please refer to https://www.mdpi.com/journal/applsci/instructions#suppmaterials. A particularly relevant paragraph states that "For work where novel computer code was developed, authors should release the code either by depositing in a recognized, public repository such as GitHub or uploading as supplementary information to the publication. The name, version, corporation and location information for all software used should be clearly indicated. Please include all the parameters used to run software/programs analyses." None of these requirements are satisfied. In the current state, the manuscript simply reports what you (or even someone before you, it is not clear) found out. However, one should attempt repeatability: the article should enable the reader to repeat your work, check your findings, and progress from them.
I encourage the authors to rewrite the manuscript in such a way that it will advance the science and be able to serve as the basis of possible further studies.
There is a lot of non-standard terminology and grammatically incorrect sentence structures. As a result, the manuscript is too hard to follow for me. For example, the first sentence of the Abstract introduces the term "liquid engine", which is not a standard name for a liquid-propellant rocket engine (it would be much better to use "liquid-propellant rocket engine" first, to let the reader know what the object of the study is, and to call it "engine" when the full name is unnecessary). In the next sentence, it is unclear what the meaning of the first "that" is: what does it refer to and which verb goes with it? Similarly malformed sentences are then common throughout the manuscript.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsThe manuscript is still not in great shape and hard to follow. However, from the current version I was able to deduce that the Quantum Bayesian probability computation in partially observable Petri nets is the main novel contribution of the manuscript. I believe such computation is wrong, as illustrated in the attached file. I invite you to double-check your work; if you believe that I am wrong, please provide more support for validity of your idea, including by building on the available literature. The claim that the classical way of computing probabilities in Petri nets is strong, therefore it requires strong arguments!
Comments for author File: Comments.pdf
If the manuscript was written better language-wise, it would be easier to follow.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 3
Reviewer 2 Report
Comments and Suggestions for AuthorsI am not currently able to review the manuscript in depth, but the latest authors' explanation clarifies how the proposed probability calculations actually do adhere to probability axioms, just as quantum mechanics does. However, a new free parameter, interference angle, is introduced. It is unclear how to choose it (why maximize the probability of fault?), whether the quantum approach results in any improvement in accuracy (or does it make the predictions worse), and how the simulations described in 4.3 are performed.
Since I find it possible (but unlikely) that Quantum Bayesian Learning has an advantage compared to the established methods and I cannot offer much helpful advice on improving the manuscript, I do not oppose its publication.
