A Novel Principal Component Analysis-Informer Model for Fault Prediction of Nuclear Valves
, Mifeng Ren 2, Wei Feng 3, Bo Sun 4, Jun Ling 5, Huanlin Chen 1, Weihua Chen 1, Yalin Luo 1 and Zhile Yang 3Round 1
Reviewer 1 Report
This paper deals with fault detection and failure prediction based on the combination of PCA and informer model. PCA is used for dimensionality reduction and feature extraction, and the former model is used for failure prediction by modeling the feature trend. The fault detection thresholds are calculated usingT-square and Q-statistic metrics.
The paper is well written and easy to read and understand. The use of the former model in this context can be considered as a contribution. The results obtained show the effectiveness of the proposed approach. The paper can be improved on the following points:
1) In section 2 of the paper, the authors gave a reminder of filtering methods, such as the Fourier Transform. Was this method applied to raw signals before applying PCAs, if so, why?
2) In line (297 to 300) the following sentence, “In Figure 7, a, b, and c show the wavelet transform time-frequency diagram of different types of fault signals respectively, and d shows the wavelet transform time-frequency diagram of normal signals. In Figure 8, the features of the data are visualized, and the results are divided into three categories: A, B, and C. “
Why did the authors use WT, and what are categories A, B, and C?
3)The introduction of the paper can be improved by review papers of existing tools or already used for the prediction of time series, such as for example: A systematic review on model selection in high-dimensional regression. Journal of the Korean Statistical Society 48 (2019) 1–12. A survey of modeling for prognosis and health management of industrial equipment Advanced Engineering Informatics 50 (2021) 101404 .And papiers dealing with hybrid methods using physical models for feature identification and data-driven approaches for failure prognosis, as for example: Failure Prognosis Based on Relevant Measurements Identification and Data-Driven Trend-Modeling: Application to a Fuel Cell System".Processes2021, V.9, 328.
Author Response
Dear reviewer,
Many thanks for your kindly support and careful review. We have now addressed all the comments and prepared the revised version of our paper. Our responses are attached, hope this revision meets your requirements.
All authors.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The manuscript discusses an important subject - fault prediction of nuclear valves. The authors of this paper propose a deep learning fault detection and prediction framework combining Principle Component Analysis and Informer to solve the problem of online monitoring of nuclear power valves. The acoustic signals generated by the valve are taken into consideration as an object of monitoring and analysis. The acoustic signals derive of mechanical phenomenon which inform about bad state of the valve, e.g. internal leakage or increasing level of friction between parts of the valve.
The manuscript develops knowledge on the area of hydraulic valve online monitoring and fault prediction of them.
The proposals for improvements, observations and minor shortcomings, which are difficult to avoid, are shown below:
- The figures 1. and 2. are illegible. I suggest to enlarge the describing of graph axis. Moreover the scale of the “normal state” and “leakage state” graphs are different and therefore the comparing of them is difficult. It would be better to show the graphs in the same scale.
- It seems that numbering of the figure in the phrase “In a word, for fault detection, regions in Fig.1 with data points in the red area indicate abnormal events” is mismatched (lines 186/187). There should be “Fig.3” instead of “Fig.1”.
- It would be good to explain better experiments, especially how the sound signals were achieved and describing the research object (valve).
- The results are very poor discussed – please discuss them more widely.
Author Response
Dear reviewer,
Many thanks for your kindly support and careful review. We have now addressed all the comments and prepared the revised version of our paper. Our responses are attached, hope this revision meets your requirements.
All authors.
Author Response File: Author Response.pdf
