Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The presented manuscript deals with a diesel engine real time monitoring based on deep learning technique LSTM. The aim is to evaluate in real time the in-cylinder pressure based on instantaneous torque measurement. A 16-cylinders diesel engine was studied both experimentally and numerically. In the experimental study the in-cylinder pressure of cylinder 1 and 9 were recorded at nine studied operating points. The engine was numerically studied in GT-Power software. Moreover, an LSTM algorithm was presented with his mathematical background and learning approach. Finally, the accuracy of the LSTM technic was evaluated by means of in-cylinder pressure comparison. The results shows good compliance between real data and that achieved based on instantaneous engine torque measurement for both cylinders. 

Despite of the importance of the problem especially in stationary engines, I have several questions and remarks to the authors.

• The main remark to the study is related to the chosen approach where the monitored in-cylinder pressure were evaluated only when the cylinders operates with their optimal performance. Does it means that the LSTM technic would give the correct response in case of cylinder failure. Please add such kind of analysis to prove that the approach is useful to detect engine damages.
• The in-cylinder pressure in Fig 2 to 10 looks quite smooth for experimentally measured pressure. Please clarify if the values corresponds to raw data or they are mathematically modified.
• At many points authors pointed out cylinder displacement measurement. It is clear that the displacement was not measured but can be easily calculated based on the crankshaft position and kinematics. 
• Please clarify why the engine was studied in GT-Power. It is only stated that the engine torque were evaluated with experimental value but there is no any comparison. I also suggest to evaluate the in-cylinder pressure.
• It is not clear how the engine torque was measured in Fig. 1. Also the other photos in Fig.1 are not clear. 
• At page 11 last paragraph you discussed about fuel injection pressure monitoring?  

Comments on the Quality of English Language

The English must be improved in the direction to be more fluent to the readers. 

Author Response

Point 1: The main remark to the study is related to the chosen approach where the monitored in-cylinder pressure were evaluated only when the cylinders operates with their optimal performance. Does it means that the LSTM technic would give the correct response in case of cylinder failure. Please add such kind of analysis to prove that the approach is useful to detect engine damages.

Response 1: As you mentioned, our team has only verified the correctness and accuracy of cylinder pressure testing methods on properly functioning engines. Evaluating the performance of this method in the face of abnormal states can help improve the robustness of the cylinder pressure detection method and promote its application in engineering practice. However, it should be pointed out that reproducing faults on multi-cylinder heavy-duty diesel engines, especially under multiple operating conditions, is still a challenging task. We will be researching the troubleshooting in our team（Li Liangyu,Tiexiong Su.Research on a small sample fault diagnosis method for a high-pressure common rail system[J].Advances in Mechanica Engineering, 2021, 13(9).See references[32]）On this basis, the research scope is further expanded, and the pressure detection method in the engine cylinder under abnormal working conditions is taken as the next research focus.

 

Point 2: The in-cylinder pressure in Fig 2 to 10 looks quite smooth for experimentally measured pressure. Please clarify if the values corresponds to raw data or they are mathematically modified.

Response 2: The cylinder pressure and torque experimental data used in this paper are all interpolated, filtered and averaged by optical smoothing, and this pretreatment process has been clarified in this paper.

 

Point 3:At many points authors pointed out cylinder displacement measurement. It is clear that the displacement was not measured but can be easily calculated based on the crankshaft position and kinematics. 

Response 3: As you mentioned, cylinder displacement cannot and does not need to be measured experimentally on the engine, only based on the crankshaft angle, crank link length, and initial angle.

 

Point 4: Please clarify why the engine was studied in GT-Power. It is only stated that the engine torque were evaluated with experimental value but there is no any comparison. I also suggest to evaluate the in-cylinder pressure.

Response 4: In this paper, the experimental calculated values, the effective power, the maximum burst pressure, the inlet air temperature, the pre-vortex exhaust temperature (row A/B) and the exhaust temperature after the vortex (row A/B) are listed in the simulation model establishment section. Through comparison, it can be seen that the simulation model used in this paper is consistent with the actual situation, and the performance calculation and analysis based on it are reliable. Due to the lack of measured data on the flow coefficient and mechanical loss of the intake and exhaust valves during the data input process, there are errors in the conversion of the effective flow area of the intake and exhaust pipelines, as well as the selection of the supercharger calculation model (there are simplified and full simulation models, and the measurement of data is not accurate when the full simulation model is selected), the selection of the combustion model (Vibe model) and the selection of heat transfer model (Woschni 1978 model is selected), resulting in some errors between the calculation results and the measured results.

 

Point 5: It is not clear how the engine torque was measured in Fig. 1. Also the other photos in Fig.1 are not clear. 

Response 5: Torque and cylinder pressure testing methods are not the focus of this article, so allow me to describe this process in my reply rather than in the main text. As shown in Figure 1, torque is measured using an electromagnetic dynamometer connected to the output (Figure 1c), while a cylinder pressure sensor is passed through the cylinder head (Figure 1b) to prevent at the fire surface. In the experiment, the dynamometer is first used to drag the engine backwards until the degree of the cylinder pressure sensor reaches the maximum value, and the crankshaft angle at this time is the dead center of the cylinder in the workpiece stage (recorded as 0°CA), and then the conversion relationship between cylinder pressure, torque and corresponding crankshaft angle can be obtained according to the arrangement position of each cylinder. In the formal experiment, the collector is first grounded and noise reduction, the engine is started, the speed is increased in steps to the target working condition, and the cylinder pressure and torque curves are repeatedly collected as the research data of this paper after stable operation (five minutes).

 

Point 6: At page 11 last paragraph you discussed about fuel injection pressure monitoring?  

Response 6: This is a clerical error that should be "training and verifying the pressure monitoring model in the cylinder of a diesel engine", which has been corrected in the text.

Reviewer 2 Report

Comments and Suggestions for Authors

This article discusses the development of an innovative method for assessing the variability of operation of an internal combustion engine, which is important for optimizing for high efficiency, durability, and reduced emissions (exhaust gases, noise, vibrations). The article is well-structured. The authors referenced a large number of studies by other researchers in this area, thus providing adequate background. The reviewer believes the descriptive section requires some additions (see comments below), but overall, the overall impression is positive and credible. The conclusions are accurate and relevant to the assumptions.
Detailed remarks:
- More engine data (e.g., main specifications, cranktrain geometry) would be appreciated; this could be important for other researchers, for example, regarding scale.
- The authors mention measuring crank angle and displacement. It is unclear from the article how displacement is measured. This is typically done by measuring the crankshaft rotation angle and utilizing geometric relationships in the crank mechanism. The authors are requested to explain how this was measured in this case.
- How is pressure measured relative to crankshaft rotation angle? Is the 5kHz frequency also applicable here, or is cylinder pressure taken for successive rotation angle values ​​(e.g., every 0.5 or 1.0 degrees).
- Table 1 - specifying torque and power to two decimal places is inappropriate for such large values. Please provide an accuracy of 1 decimal place.
- More model data in GT-Power would be appreciated.
- Comments on the scalability of the results would be appreciated (see first remark).

Author Response

Point 1: More engine data (e.g., main specifications, cranktrain geometry) would be appreciated; this could be important for other researchers, for example, regarding scale.

Response 1: The relevant structure and performance parameters of the engine are listed in the article, see "Table 1 Basic data of diesel engines

”

Point 2: The authors mention measuring crank angle and displacement. It is unclear from the article how displacement is measured. This is typically done by measuring the crankshaft rotation angle and utilizing geometric relationships in the crank mechanism. The authors are requested to explain how this was measured in this case.

Response 2: In the experiment, the dynamometer is first used to drag the engine backwards until the degree of the cylinder pressure sensor reaches the maximum value, and the crankshaft angle at this time is the top dead center of the cylinder in the workpiece stage (recorded as 0°CA), and then according to the arrangement position of each cylinder and the structural parameters of the crank connecting rod mechanism can be converted to obtain the conversion relationship of cylinder pressure, torque, piston position and corresponding crankshaft angle. Its calculation formula can be expressed as:

Where, is the piston displacement, is the engine stroke, is the crankshaft angle, andis the connecting rod ratio.
Point 3: How is pressure measured relative to crankshaft rotation angle? Is the 5kHz frequency also applicable here, or is cylinder pressure taken for successive rotation angle values (e.g., every 0.5 or 1.0 degrees).

Response 3: The sampling frequency given in this paper is the sampling frequency of torque sensor and cylinder pressure sensor. The cylinder pressure/torque curve with the top dead center (denoted as 0°CA) in the workpiece stage can be obtained by using the angle conversion method described in Response 2.

Point 4: Table 1 - specifying torque and power to two decimal places is inappropriate for such large values. Please provide an accuracy of 1 decimal place.

Response 4: The relevant data in this article has been modified to keep the torque to one decimal place.

 

Point 5: More model data in GT-Power would be appreciated.

Response 5: The parameters and model types set for the simulation model creation are listed in detail in Appendix 1.

Point 6: Comments on the scalability of the results would be appreciated (see first remark).

Response 6: At the end of the conclusion chapter, the next step of research is prospected.

Reviewer 3 Report

Comments and Suggestions for Authors

The paper corresponds to the scope of the journal. It is necessary to revise the manuscript and reduce the similarity index to approximately 10%, depending on the editorial requirements and the journal’s policy.

Within the introduction section, the text should be revised and standard terminology for this research field should be used. It is necessary to explain that pressure pulsations and variations in the in-cylinder pressure of diesel engines can primarily cause serious engine damage. From this aspect, the necessity of applying variable engine systems, such as the variable compression ratio (VCR diesel engine) and similar solutions, should be explained and emphasized. From this perspective, it is recommended to include the following reference in the literature:
Milojević, S., Savić, S., Marić, D., Stopka, O., Krstić, B., & Stojanović, B. (2022). Correlation between Emission and Combustion Characteristics with the Compression Ratio and Fuel Injection Timing in Tribologically Optimized Diesel Engine. Tehnički vjesnik, 29(4), 1210–1219. https://doi.org/10.17559/TV-20211220232130

The resolution of certain figures and diagrams should be verified and improved where necessary.

Data regarding the tested engine and measuring equipment should be presented in tabular form. The table containing test conditions should be checked, and standard designations and SI measurement units should be used.

The derivation of the presented equations should be clarified, and appropriate literature references should be provided. Information about the computer hardware, model, and software used (including versions and licenses) should also be included.

The obtained results are of great importance and represent a valuable database that can be used for comparison by other researchers. It is also recommended to indicate the potential for applying machine learning methods for processing the experimental results.

Guidelines for further research should be proposed, and the conclusions should be expanded to include key elements from the discussion of results. The number of references should also be increased.

After implementing all revisions required by the review process, a final technical proofreading of the manuscript should be performed.

Author Response

Point 1: From this perspective, it is recommended to include the following reference in the literature:Milojević, S., Savić, S., Marić, D., Stopka, O., Krstić, B., & Stojanović, B. (2022). Correlation between Emission and Combustion Characteristics with the Compression Ratio and Fuel Injection Timing in Tribologically Optimized Diesel Engine. Tehnički vjesnik, 29(4), 1210–1219. https://doi.org/10.17559/TV-20211220232130

Response 1: Suggested articles have been added to the references, and a description of intracylinder pressure fluctuations has been added to the introduction to reinforce the importance of this study, as detailed in the references[1]。

 

Point 2:The resolution of certain figures and diagrams should be verified and improved where necessary.

Response 2:Some icon formats have been refactored so that they now more clearly support the conclusions of the text.

 

Point 3:Data regarding the tested engine and measuring equipment should be presented in tabular form. The table containing test conditions should be checked, and standard designations and SI measurement units should be used.

Response 3: The relevant structure and performance parameters of the engine have been listed in the article, see "Table 1 Basic Data of Diesel Engines" in the article.

 

 

Point 4:The derivation of the presented equations should be clarified, and appropriate literature references should be provided. Information about the computer hardware, model, and software used (including versions and licenses) should also be included.

Response 4: Relevant references are added to the formula in the text, which can be seen in Reference [21] [22] [23] [24] [25] [29] [30] [31]. And at the beginning of Chapter 2, computer and software information is listed, which is expressed as "This paper takes a V-type 16-cylinder turbocharged diesel engine as the research object, and its basic structure and technical parameters are shown in Table 1. The sampling frequency of 5000Hz is used to collect relevant data such as cylinder pressure curve, crankshaft torque, crankshaft angle, and displacement of each cylinder. The diesel engine test bench is shown in Figure 1. In this paper, MATLAB R2020b is used to build various mathematical models, and GT-power V7.5 is used to establish a one-dimensional simulation model of diesel engine. The operating environment of the above software is Win10, and the workstation configuration is: Intel C621 series chipset; Intel Xeon Gold 6145 processor, 40 cores and 80 threads, clocked at 2.0GHz, maximum turbo frequency 3.7GHz, 128GB memory, NVIDIA GT1030, 2GB graphics card. ”

 

Point 5:The obtained results are of great importance and represent a valuable database that can be used for comparison by other researchers. It is also recommended to indicate the potential for applying machine learning methods for processing the experimental results.

Response 5: Thank you for your recognition of our team's research, and a directional summary of this research has been made in Conclusion 3.

 

Point 6:Guidelines for further research should be proposed, and the conclusions should be expanded to include key elements from the discussion of results. The number of references should also be increased.

Response 6: The discussion has been expanded at the end of the concluding chapter and the number of references has been increased, now from 22 to 32.

 

Point 7:After implementing all revisions required by the review process, a final technical proofreading of the manuscript should be performed.

Response 7: Thanks for the reminder, our team will do the final technical proofreading of the article after it is finalized.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for your prompt response to my remarques and questions. 

However, it doesn't clarify the main weaknesses of the study. Regarding Response 5, you stated that torque and pressure measurements are not the main focus of the article but all the results, analysis and contributions are based on measured torque and in-cylinder pressures. It is not understandable for the readers even for me how you measured the instantaneous torque using an electric machine rather than torque sensor. Please explain the torque testing method and equipment. The Fig. 1 (a, b and c) doesn't not bring clear information for the measuring equipment. Did your pressure sensor installed directly in the combustion chamber or not?

Regarding Response 4, it is still not clear what is the contribution of the 1D-0D modelling in GT-Power, moreover you cant trust the obtained numerical results. The normal approach should be to show some comparison graph for in-cylinder pressure and torque curves instead the output power and maximum pressure. Please add this comparison in order to justify the need of this modelisation. As well you should simulate abnormal operation of the engine by means of the simulation model and reveal if the proposed approach based on LSTM could be used for engine diagnostics.      

Comments on the Quality of English Language

The English must be improved in the direction to be more fluent to the readers. 

Author Response

Please see the attachment

Author Response File: Author Response.pdf