Data-Driven Sliding Bearing Temperature Model for Condition Monitoring in Internal Combustion Engines

Round 1

Reviewer 1 Report

The main objective of the paper is to develop a surrogate model to describe the temperature of heavy-duty ICE main bearings. For this, the authors have used three different machine learning algorithms on a set containing only experimental data. The paper is understandable and well written, the methodology being presented with a high level of details. However, there are a few points which this reviewer thinks could be improved or more discussed:

a) The idea of creating a surrogate model from experimental data is not new, and the application to internal combustion engines makes the novelty just marginal. The authors should think carefully what the novelty of their work is and what are the advances made to the state of the art.

b) As far, the authors have a way to compute the main bearings’ temperatures. How the plan to use this information for condition monitoring of the system? They should discuss the applicability of their model on the conditioning monitoring of the system (Entire engine? Only the bearing? Crankshaft?).

c) All equations should be numbered, and when talking about an equation its number should be cited.

d) The authors could give more details about the test rig used.

e) Would the results be similar if the authors have defined one independent equation for each bearing (with independent coefficients for each parameter) instead of the Boolean approached used?

f) The paper lacks conclusions.

Author Response

Dear Reviewer 1,

Thank you for your valuable comments. Please find our answers to your comments below:

1. “The idea of creating a surrogate model from experimental data is not new, and the application to internal combustion engines makes the novelty just marginal. The authors should think carefully what the novelty of their work is and what are the advances made to the state of the art.”
   
   We fully agree that the methods used are not new when considered separately ­— neither machine learning methods for condition monitoring nor thermocouples for temperature measurements. However, as our extensive literature review has shown (cf. lines 86–102), the presented approach to use readily accessible engine data to predict sliding bearing temperatures is indeed novel. Furthermore, the approach presented lays the foundation for a new form of condition monitoring for sliding bearings in internal combustion engines. Therefore, we are not only convinced about the scientific novelty of our approach, but also that our manuscript will be of particular interest to the readership of the special issue "Tribology in Mobility".
   
   
2. “As far, the authors have a way to compute the main bearings’ temperatures. How the [sic] plan to use this information for condition monitoring of the system? They should discuss the applicability of their model on the conditioning monitoring of the system (Entire engine? Only the bearing? Crankshaft?).”
   
   With the high accuracy achieved, we are confident that the derived modeling procedure could serve as an advantageous system for condition comparison. We have already addressed this potential in lines 103–110. However, there is no specific application in a production engine planned yet. We believe that the presented approach is particularly suitable for the field of large engines, since the cost for bearing instrumentation is relatively low compared to the cost of an engine failure caused by the bearing system. As discussed in lines 187–206, the modeled bearing temperature solely depends on engine parameters that are available on a production engine. Therefore, our approach could be easily adopted by engine manufacturers or engine part suppliers.
   
   
3. “All equations should be numbered, and when talking about an equation its number should be cited.”
   
   While writing our manuscript, we used “Occam’s Rule” [1]. According to this rule, equations are only numbered if they are referred to in other text passages. Since no equation is directly referred to, there are no numbered equations in the submitted manuscript. However, based on your input (also referred to as “Fisher’s Rule” [1]), we have decided to number all equations and have added additional references to appropriate passages.
   
   
4. “The authors could give more details about the test rig used.”
   
   As outlined in lines 151–152, the test rig was explained in detail in [2]. Therefore, we only provided a summary of the experimental setup in our manuscript. We believe that further details about the test rig are not of particular interest considering the focus of our manuscript (i.e., data-driven modeling and condition monitoring of bearing temperatures). However, if specific information is requested by the reviewer, we will gladly provide it.
   
   
5. “Would the results be similar if the authors have defined one independent equation for each bearing (with independent coefficients for each parameter) instead of the Boolean approached used?”
   
   Yes, in terms of error measures, the results are indeed very similar.
   
   
6. “The paper lacks conclusions.”
   
   Thank you for this input. As mentioned in the MDPI template, the section Conclusion “is not mandatory, but may be added if there are patents resulting from the work reported in this manuscript”. Based on the reviewers’ comments, we have added this section to our revised manuscript and have discussed the conclusions in more detail.
   
   

Sincerely,

Christian Laubichler on behalf of all authors

[1] Hwang, A. D. (1995). Writing in the Age of LATEX. Notices of the AMS, 42(8).
[2] Marques da Silva et al. (2021). Experimental Investigation of the Influence of Engine Operating and Lubricant Oil Parameters on Sliding Bearing and Friction Behavior in a Heavy-Duty Diesel Engine. In Internal Combustion Engine Division Fall Technical Conference (Vol. 85512). American Society of Mechanical Engineers.

Reviewer 2 Report

In this research article, preventing condition monitoring was performed using modeling. Different machine learning methods are thoroughly tested in terms of their prediction error with the help of a repeated nested cross-validation. Different linear regression approaches were used to predict the bearing temperatures on the basis of engine operation parameters.

The technique is attractive, and the work is of significance. However, the results are poorly discussed and highly qualitative. There are still some obvious shortcomings needed to be addressed by the authors before publication. Therefore, I suggest that this manuscript requires a minor revision.
Some other comments are given below:

1. Add the most significant results in abstract.
2. "All relevant parameters such as engine torque and speed as well as media temperatures, pressures and flow rates were measured and recorded with measuring instruments and a data acquisition system." Add the measuring instrument and measuring procedure.
3. Add the engine specification.
4. Figure 2: correct unit of speed and recheck the operating torque of engine.
5. Viscosity measuring procedure and instrument used.
6. "fuel consumption is calculated from fuel mass flow and engine power" fuel consumption is measured by fuel mass flowmeter which depend on the engine torque and speed/ power.
7. table 1: correct the unit of speed-which can be in terms of rpm.
8. Air inlet pressure upstream of EGR admixing"    admixing?
9. Author can add the governing equations of the regression and predictive models models.
10. the formulation and detailed discussion of bearing temperature model.
11.  line 538:  comparatively slowly to swift" may be comparatively slow.
12. conclusion can be added.

Author Response

Dear Reviewer 2,

Thank you for your comments and suggestions. In addition to the replies to your specific comments provided later, we would like to address the following two points:

1. You’ve selected that “Moderate English changes are required”.
   
   Before submission, our manuscript was proofread by a professional proofreader (native American English speaker). Since we would like to discuss this topic with our proofreader, could you please point out specific parts that require revision?
   
   
2. You’ve stated that “the results are poorly discussed and highly qualitative”.
   
   We feel compelled to disagree with the assertion that our results are "poorly discussed". However, we are gladly willing to provide additional evaluations on the results if the reviewer would be so kind to specify them in more detail. Please note that the following analyses on the results are already discussed in our manuscript:
   
   
   • Cross-validation (CV) error comparison using density plots (cf. Figure 8) and summary measures (cf. Table 4). Both (R)MSE and MAE error measures are provided for this purpose. The presented discussion includes an interpretation how large the (smallest) errors are compared to the overall temperature range (cf. lines 460–462). The ranking of the modeling procedures does also assess the stability of the CV results (cf. lines 462–468).
   • The results on the test data first include the optimal hyperparameters (cf. Table 5). The SVR model is then discussed in detail using a scatter and a residual plot (cf. Figure 9, lines 473–479). In addition, the results per bearing (cf. Table 6) are also addressed there (cf. lines 475–479).
   • For a better interpretation, the LM approximation is first motivated (cf. lines 480–483) and then presented as formula (cf. line 487–488).
   • How the LM coefficients can be interpretated is discussed in detail using Pearson correlations and Hoeffding’s D statistics (cf. Figure 10, lines 489–502).
   • Additional discussions on the results are then also included in the “Discussion” in Section 4.

Based on your other comments, we are further not sure if comments 9. and 10. address the theoretical model formulations or the results. If the latter, you could you please elaborate on what specifically is missing at the discussion and analysis of the results (especially concerning the quantification of the results)?

Please find below our answers to your other comments:

1. “Add the most significant results in abstract.”
   
   Thank you for this input. We have added the most significant results to the abstract.
   
   
2. “‘All relevant parameters such as engine torque and speed as well as media temperatures, pressures and flow rates were measured and recorded with measuring instruments and a data acquisition system.’ Add the measuring instrument and measuring procedure.”
   
   As outlined in lines 151–152, the measuring instruments and the measurement procedure were explained in detail in [1]. Therefore, we only provided a summary of the experimental setup in our manuscript. We believe that further details about the test rig are not of particular interest considering the focus of our manuscript (i.e., data-driven modeling and condition monitoring of bearing temperatures). However, if specific information is requested by the reviewer, we will gladly provide it.
   
   
3. “Add the engine specification.”
   
   In lines 148–150 and Figure 2 the engine specification is already addressed including engine type, number of cylinders, displacement, and engine operating points including speed and torque performance data. More details were presented in [1]. If specific additional information is required by the reviewer, we will gladly provide it.
   
   
4. “Figure 2: correct unit of speed and recheck the operating torque of engine.”
   
   Could you please elaborate what the correct unit of speed is? If as intended in comment 7. rpm should be used, we do not see any difference since 1 rpm = 1 min^-1. Of course, we can also provide an SI-conform unit (e.g., s^-1).
   
   The engine operating map shown in Figure 2 is based on technical data from the engine manufacturer (MAN). Could you please specify in what way we should check it again?
   
   
5. “Viscosity measuring procedure and instrument used.”
   
   Thank you for this input. In the course of our research, we did not measure the oil viscosity ourselves. Instead, the oil supplier provided us with the viscosity curves shown in Figure 2. We have added this information to the revised manuscript.
   
   
6. “‘fuel consumption is calculated from fuel mass flow and engine power’ fuel consumption is measured by fuel mass flowmeter which depend [sic] on the engine torque and speed/ power.”
   
   Thank you for pointing this out. Actually, the “brake-specific fuel consumption” is meant in this sentence. We have changed this in the revised manuscript.
   
   
7. “table 1: correct the unit of speed-which can be in terms of rpm.”
   
   Please see reply to comment 4.
   
   
8. “Air inlet pressure upstream of EGR admixing"   admixing?”
   
   Could you please elaborate why you are concerned about “admixing”? In our experience, it is a proper term (cf. [2]) that is regularly used in connection with internal combustion engines.
   
   
9. “Author can add the governing equations of the regression and predictive models models [sic].”
   
   Could you please elaborate if this comment refers to the theory in Section 2.3 or the results in Section 3? Please also see reply to point b.
   
   
10. “the formulation and detailed discussion of bearing temperature model.”
    
    Could you please elaborate if this comment refers to the theory in Section 2.3 or the results in Section 3? Please also see reply to point b.
    
    
11. “line 538:  comparatively slowly to swift" may be comparatively slow.”
    
    Because “slowly” is an adverb of “react” in the addressed sentence (“Since bearing temperature reacts comparatively slowly to swift changes in engine operating conditions such as engine speed and engine torque, …”), the “-ly” is required.
    
    
12. “conclusion can be added.”
    
    Thank you for this input. As mentioned in the MDPI template, the section Conclusion “is not mandatory, but may be added if there are patents resulting from the work reported in this manuscript”. Based on the reviewers’ comments, we have added this section to our revised manuscript and have discussed the conclusions in more detail.
    
    

Sincerely,

Christian Laubichler on behalf of all authors

[1] Marques da Silva et al. (2021). Experimental Investigation of the Influence of Engine Operating and Lubricant Oil Parameters on Sliding Bearing and Friction Behavior in a Heavy-Duty Diesel Engine. In Internal Combustion Engine Division Fall Technical Conference (Vol. 85512). American Society of Mechanical Engineers.
[2] https://www.merriam-webster.com/dictionary/admix

Round 2

Reviewer 1 Report

Dear authors,

The paper would improve if the you could add to it the comments and dicussions about novelty of 1). In addition, the authors should consider changing  their objective to create a deatiled procedure for developing a data-driven temperature model for ICE bearings. I think the novelty lies more in this aspect than in creating a data-driven model, which is not original at all. The authors should consider change the paper's title as well.

Regarding the test rig description, not all readers will have access to paper [7]. Therefore, a brief description of the test rig would benefit the current work. It does not need to be as detailed as [7], but it should contain how the test rig works, which sensors are used and what their measurement range and errors are, what data acquisition system was used, and so on.

Author Response

Dear Reviewer 1,

Thank you for your valuable comment. We agree that the novelty should be emphasized more in our manuscript. For this reason, we have revised the passages concerning the goal of the paper as well as the final discussion. In addition, we have also changed the title of the paper to better reflect the generalizability of our approach. In our revised manuscript you will also find additional information on the test rig. We believe that this should provide a sufficiently detailed overview. Of course, if additional information is requested, we will gladly provide it.

Sincerely,

Christian Laubichler on behalf of all authors