Research on Prediction Method of Hydraulic Pump Remaining Useful Life Based on KPCA and JITL

Round 1

Reviewer 1 Report

• lots of incomprehendable equations seem to disguise their irrelevance by artificially complicating simple relations: The volumetric efficiency decreases and the pump fails (well known). This decreasing efficiency  is then recalculated by the said numerous and cude equations without any visible improvement of the finding that the pump fails (what it did not as the test simply was stopped).
• fig. 6 is a lough
• it is not outlined which stress level (fig. 8) was measured and how
• the relevance to RUL, KPCA, JITL is doubted
• many own citations 
• the scientific-engineering-practical relevance is not visible 
• That such a robust gear pump fails after only 1000 hrs of operation is most surprising, respectless of whether the tests were performed with or without step-stess acceleration method.
• it is normal that the volumetric efficiency of any rotary positive displacement pump decreases with increasing counter pressure. This is not necessarily an indication of decreasing RUL.
• If the volumetric efficiency decreases below 70% (due to high counter pressure), the test is stopped (without a pump failure) but the pump have failed.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper proposes a new method to detect the failure of hydraulic geared pumps and to predict their remaining useful life. Authors employed the wavelet packet method to denoise the sensor signals. The kernel principal component analysis and the just-in-time learning were used for the detection of the failure and estimation of its remaining useful life, respectively. The paper provides theoretical backgrounds, which are applied to experimental data. The paper presents analysis results and valid conclusions of the gear pumps. The paper can be further improved for the publication, not limited to the following comments.

• Authors defined the failure of the pump as their volumetric efficiency which becomes lower than 70%. The volumetric efficiency will change with the pressure and speed at which a pump is operated. When a single number is given for volumetric efficiency, it will typically be at the rated pressure and speed. Authors planned the stepwise loading schedule with 23, 25, and 27 MPa. Higher the pressure is loaded, lower the efficiency will be, probably not because of the pump failure itself due to component degradations as shown in Figure 9. Authors need to explain the valid justification of the pump failure definition and experimental setting.
• Authors need to explain why 70% level of volumetric efficiency is set to the threshold of the failure. Is 70% come from any physical understanding?
• The structural failures such as surface wear or functional factors like loosening can be correlated to the way of signal representations. With the considerations of the characteristics from the signal data, a certain algorithm for denoising and extracting a degradation index can be better suitable, and KPCA or JITL can be chosen in that reason. In other words, the theories presented in the paper are pretty general, which means they are less appealing.
• The conclusion section looks like an introduction. In conclusions, they need to summarize their work and to provide the implications of their data, leading to the conclusions.
• What does the performance parameter mean in Figure 1? It does not make sense that the failure is at its highest level of the performance. Typically, the performance of mechanical equipment goes down like a slower motor as it degrades. Also, the Figure 1 does not seem to provide much information, if not the trend of typical mechanical failure accelerated to the end of its life is shown. If it is intended so, the manuscript needs a quote referring the physical meaning of a specific example.
• The paper says 14 time-domain indices in Figure 10 are from the extraction of the gear pump pressure. Then, what is the unit “g” of the mean value? Isn’t it supposed to be “MPa”? Also, what sensor data are processed for the statistical measures as time goes by, even though there are only one pressure sensor per pump?
• Is RUL meaning right in Figure 2? Also, the Figure 2 does not seem to provide much information.
• It has to be physically explained how P can be determined in Figure 2.
• Authors need to explain how h and g for filters in Equation (2) are selected?
• What are multiple characteristic parameters at section 3?
• The hypothesis that the k-VNN method which considers both the Euclidean distance and angle relationship between two vectors is superior than k-NN needs to be provided at section 4, and then subsequently it can be proved in the end of the manuscript.
• “A certain load” does not sound scientific at section 5.
• What is no-substitute time tac-tail life? Please put a reference if you use it from somewhere.
• What is the time metric t/h?
• Table 8 indicates that both MRE and ARE of the gear pump RUL prediction method proposed in this paper are better than those of the traditional k-NN-based RUL prediction algorithm. Can authors provide more discussion in detail on the reason why the results were obtained like that?
• Some sentences are unnecessarily written as they are too general. For example, the paragraph from lines 134-143, it does not deliver useful information. So, trimming down is recommended in the context.
• More references are recommended in the field of fault detection as of followings (of course, not limited to):

Song, Yongxing, et al. "A novel demodulation method for rotating machinery based on time-frequency analysis and principal component analysis." Journal of Sound and Vibration 442 (2019): 645-656.

Siano, D., and M. A. Panza. "Diagnostic method by using vibration analysis for pump fault detection." Energy Procedia 148 (2018): 10-17.

Gajjar, Shriram, Murat Kulahci, and Ahmet Palazoglu. "Real-time fault detection and diagnosis using sparse principal component analysis." Journal of Process Control 67 (2018): 112-128.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

 Research on Prediction Method of Hydraulic Pump Remaining  Useful Life Based on KPCA and JITL

 

Abstract should be rewritten to show better the methodology and novel of the manuscript

Line 46-65 more references should be used and they should be discussed not only enumerated (line 59). Review this all document.

Line 65 add point at the end of the sentence

Line 70 discuss references

Line 83-92 more reference are necessary

Line 97-105 add references to justify the sentences

A flowchart is necessary to clarify the used methodology

Figure 3 should be described better and improved

Figure 4 and Figure 5 should be discussed

The experimental campaign should be described

The manuscript should be reorganized on introduction, methodology and results, in my opinion seven sections are too much

The conclusions should be rewritten and they cannot show the summarize of the research. The conclusions section should show the main novel of the research and its applicability in the real life.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

The authors clarified the different suggestions