Aerodynamic Shape Optimization Method of Non-Smooth Surfaces for Aerodynamic Drag Reduction on A Minivan

Round 1

Reviewer 1 Report

Review of "Aerodynamic Shape Optimization Method of Non-Smooth 2 Surfaces for Aerodynamic Drag Reduction on a Minivan"

by Zhendong Yang, Yifeng Jin and Zhengqi Gu

MDPI, Fluids

 

This manuscript explores the possibility of aerodynamic drag reduction of a minivan by applying a non-smooth surface structure consisting of semi-spherical indentations on the minivan’s roof. First, CFD simulations are conducted to determine the most promising surface structure arrangement. Subsequently, a parameter study based on a Design of Experiments method is carried out. The resulting drag data are then used to perform an optimization study, in which a Kriging surrogate model is employed. A CFD simulation is finally conducted for the optimized parameter set, in which the drag reduction results are confirmed.

Even though the results are certainly interesting and the methodology seems sound, this manuscript needs significant rework. The description of the CFD simulation setup lacks information necessary to assess the quality of the simulations. The results need to be presented in a more comprehensive manner, and the analysis of the results (i.e. how does mechanism of drag reduction work?) could be improved.

Major issues:

1. Literature:
   - reference 1: The book cited is edited(!) by Hucho; which chapter by which author(s) are you referring to? Did Hucho (or another author) really achieve drag reduction by improving local body shape or is he describing different measures in this book (chapter).
   - reference 2: Is this a book, a PhD-Thesis? Please provide details.
   - reference 12: the book title should be “Aerodynamic drag reduction of Passenger cars”
   - Citing Wikipedia in a scientific article should be avoided. There are many textbooks (e.g. Schlichting) that provide formulas for the development of the thickness of laminar and turbulent flat plate boundary layers.
2. Lines 41-46: The description of the effect of dimples on the surface of a golf ball is missing the most crucial aspect. The dimples trigger laminar-turbulent transition of the boundary layer. While the laminar boundary layer of a smooth ball is prone to separation in the presence of positive pressure gradients, the turbulent boundary layer remains attached much further downstream.
   While the golf ball is certainly a very prominent example of boundary layer manipulation due to surface structures, it should be pointed out that the mechanism is different from that observed in the minivan simulations.
3. Your literature review is not very well structured, since the order the articles are cited seems a bit random and inconsistent with respect to the different drag reduction mechanisms. Different methods mentioned in the cited literature are: microscale structures in order to delay laminar-turbulent transition; microscale structures which alter the near wall turbulence of turbulent boundary layers in order to reduce skin friction drag; structures which decrease the skin friction drag of laminar boundary layers due to stable vortices developing in the surface cavities; hydrophobic surface structures in order to reduce skin friction drag in liquids; measures to increase the base pressure of a bluff body by influencing the boundary layer development, e.g. by increasing boundary layer thickness by applying a rough forebody surface. Your approach falls somewhat in the latter category, since you try to influence the boundary layer of the roof in order to influence base pressure and hence reduce base drag by using macroscopic (centimeter-sized) structures. This is why you can conduct classical RANS-model simulations (which would not be possible e.g. for microscale riblets). Your approach can also be classified as a passive drag reduction method for bluff bodies (for which an ample body of literature exists).
4. Equations 1 and 2: You talk of ‘incompressible governing equations’, then why do you write the Reynolds-Averaged Navier-Stokes equations in a compressible formulation, where e.g. the volume dilatation term is still included in the stress tensor? It would be beneficial to point out that equations 1 and 2 are the RANS-equations and that the flow variables are averaged quantities.
5. It is essential that you mention the size of the measurement domain, e.g. in order to quantify the blockage effect. Concerning the comparison of wind-tunnel tests and simulations: Did the computational domain match the wind tunnel dimensions or did you take measures to correct for the blockage effect of the wind tunnel?
6. The quality of figure 1 is questionable. Subfigure (a) could benefit from showing the feature lines (as in figure 4). It is impossible to recognize the mesh in subfigure (b). The reader cannot learn anything from this picture. It would be beneficial to zoom in further and show the detailed surface mesh of e.g. 2-3 important parts of the geometry.
7. The surface mesh shown in Figure 4(b) is also not of sufficient quality. It is not even possible to figure out if you used triangular or quad cells on the surface.
8. Information about the mesh resolution of the wall boundary layer is missing, but is of crucial importance in this study. One way to provide this information would be a y+-contour plot, especially of the roof surface.
9. Equation 6 certainly does not describe the boundary layer thickness of laminar flow of a flat plate (which has a x^0.5 dependency). The formula seems to be derived for a turbulent boundary layer using a power-law velocity profile. Does the equation describe the displacement or the momentum thickness?
10. Axis labels are missing in figure 9.
11. What does ‘Position’ refer to in figure 11? Why does it start at 8m and end at approx. 1m? How does this relate to the position of the surface structure on the roof?
12. Figure 12 is needs to be improved. First of all, is the scale of those two plot really the same? If yes, why is the outer velocity so much higher in case (b)? Why do you only show such a small portion of the flow field? – the entire vortex structure in the wake of the minivan would be interesting. The resolution of the figure is very poor and the vector density is too high (if Ansys Fluent is used for post-processing, the sample-points option in the plane tool can be used to produce better vector plots). Can you also plot the wall contour? In this plot it seems like the rear edge is sharp in case (a) and rounded in case (b), which could at least explain part of the drag reduction.
13. Can you quantify how much of the drag reduction is related to base drag reduction and if skin-friction reduction on the roof plays a role?
14. Can you elaborate on how the dimples on the roof influence the boundary layer on the top of the minivan and how this is related to the base pressure?
15. In addition to figure 13, a comparison of the surface pressure distribution on the rear end of the minivan would be interesting.
16. Can you analyze why the drag reduction is more prominent when the design parameter L is reduced?
17. Your manuscript could greatly benefit from a careful read-through. There are a number of small errors (like ‘rib lets’ instead of riblets in line 51), but there are also a number of passages which are very hard to read and interpret, like to following:
    - line 67f: ‘The purpose of this study is to ascertain the aerodynamic drag improvement effect by using the Surrogate Model to the circular concave shape optimization of a minivan with non-smooth surface.’
    - Using the term ‘unit-body’ for the semi-spherical surface indentations is quite confusing.
    - ‘arrangement way’ is also confusing, it should simply mean ‘arrangement’.
    - line 214f: ‘so the rectangle arrangement of circular concave non-smooth surface is chosen as the optimized object,’ … is chosen for optimization?
    - line 293f: ‘The optimal design variables are put into the CFD simulation, the minivan model is updated, meshed and solved for aerodynamic drag coefficient by the simulation software, the simulation result for aerodynamic drag coefficient is 0.3342. Compared to the surrogate-model result, the tolerance is only 0.57%, shown in Table 8.’
    - line 302: ‘When the air flows through the circular surface…’

 

Further remarks:

• Is there a reason why you use the SST-k-omega model for your simulations?
  There is e.g. literature demonstrating that the SST-model is not able to reproduce the base pressure distribution of bluff road vehicles (e.g. DOIs: 10.1007/978-3-540-44419-0_21, 10.1007/978-3-319-20122-1_25)
• Which turbulence boundary conditions did you use at the inlet?
• In addition to your wind tunnel comparison, a grid dependence study with Richardson extrapolation of the drag coefficient results would be beneficial.
• Optimal Latin Hypercube is a well-established DOE method, still you might consider including a citation.
• Dimensionless variables (e.g. velocity divided by the freestream velocity, pressure coefficient instead of static pressure) could be used in the contour plots.
• The flow direction should be consistent in all plots (fig 10)

Author Response

Dear reviewer:

Thank you very much for your advice.We have revised the manuscript, and would like to re-submit it for your consideration. please see attached.

Author Response File: Author Response.pdf

Reviewer 2 Report

1. Introduction - Wrong syllable split

Rong Jianglei put forward a local retrofit method of aero-dy-namic characteristics based on iterative method, and verified its validity by the aer-ody-namic drag improvement practice of a car[2].

However, these methods have some limita-tions to further aerodynamic drag reductions.

Modifications of the model are based on tri-al and error experience, and the CFD calculations also need further valida-tion.

Line 223 - Specification of DOE abbreviation

Figure 10 - to be more visible

Author Response

Thank you for your useful comments and suggestions.

We have revised the manuscript accordingly, and would like to re-submit it for your consideration,please see attached.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

2nd Review of "Aerodynamic Shape Optimization Method of Non-Smooth Surfaces for Aerodynamic Drag Reduction on a Minivan"

by Zhendong Yang, Yifeng Jin and Zhengqi Gu

MDPI, Fluids

The manuscript has improved considerably. The figures have improved, grid dependency results have been included and the analysis of the results provides more information. However, there are still some errors in the manuscript, which must be fixed prior to publication (like obvious errors in the basic equations).

Major issues:

• Line 43: “To reduce skin friction drag of the blunt-based flight vehicle….”. This is not correct: The results of Whitmore et al (reference 7) show that, compared to a smooth surface’ an increase(!) of skin friction drag of the forebody due to transverse surface structures leads to a decrease of overall drag due to base drag reduction.
• Line 60f: You have added the laminar-turbulent transition to the explanations concerning golf balls. The previous passage, which is difficult to understand, i.e. “Because of the existence of circular concave, the boundary layer formed by airflow is close to the surface of golf balls during its flight, then the smooth airflow goes forward a bit more along the sphere, which delays the separation of the boundary layer…” is, however, still present, as well. It would be beneficial to provide a more concise explanation.
• Equation 2 is not correct. The momentum equation has been divided by density, which should be reflected in the pressure term and the diffusion term (kinematic viscosity).
  The bulk viscosity term is zero and should be omitted
• Equation 3 is not correct. The viscosity is not correct, and the bulk viscosity term should be omitted.
• The same error concerning the viscosity of the diffusion term is repeated in equations 4 and 5. The production term of k could be formulated without tau_ij (or tau_ij could be introduced in equation 3).
• Remark 19 of the first review has not been answered. When you apply a velocity inlet boundary condition, you need to specify values for k and omega (or related turbulence quantities from which k and omega can be calculated). This information could/should be provided in context of table 2.
• Concerning remark 22 of the first review:
  It is of course possible to produce dimensionless velocity contour or vector plots in Ansys Fluent. A dimensionless velocity field can be plotted by defining a Custom Field Function as “velocity-magnitude / 30” (assuming u_infinity=30m/s). If the inlet velocity and the fluid density are prescribed under ‘Reference Values’, the field ‘Pressure Coefficient’ will produce c_p-value plots.
• I do not think that way the concaves influence boundary layer thickness and reason why the concaves lead to reduced drag are adequately explained in the manuscript. In line 357f you write: “As the  roof  roughness  increases,  the  boundary  layer  on  the  roof thickens and reduces the shearing effect of external flow on the separated flow behind the base  region,  resulting  in  reduced  base  ” The wording is nearly copied from Whitmore et al, but with regard to the data presented, I doubt that this accurately describes the effects you encounter in these simulations.
  First of all, the friction drag on the roof seems to be greatly reduced by the concaves, at least this is what you demonstrate in figure 12 (even though plotting the skin friction coefficient for “some points behind the circular concaves are selected” seems a bit arbitrary and you do not provide the overall friction drag coefficient of the entire roof or the van and what proportion of the drag reduction is due to skin friction drag reduction).
  Second, in the second figure 12 (which should be figure 13) the boundary layer thickness at the end of the roof seems to be higher for the original van without dimples. In the response to Remark 14 you write, however, that dimples lead to an increase in the boundary layer thickness which in turn leads to a more advanced position of the separation line. From the vectors in figure 12 (13) it appears that the streamlines on the roof of the van with surface structures better follow the roof contour, so that the vectors have a downward component at the edge of the roof.
  I do not think that the data and the explanations you provide form a coherent argumentation, however the value of the manuscript would greatly increase if you could convincingly explain the influence of the roof boundary layer on the size of the recirculation area.  

Some small errors encountered:

• Line 137: total number of cells/elements
• Table 2: gauge pressure
• Line 155: ‘the cure of drag coefficien’
• There are two figures called “Figure 12”
• A careful read-through would still be beneficial in order to correct faulty expressions and grammatical errors: Just as an example, the first sentence of the abstract should read: “To reduce aerodynamic drag of a minivan, non-smooth surfaces are applied to the minivan’s roof panel design.”

Author Response

Thank you for your useful comments and suggestions on our manuscript.

Author Response File: Author Response.pdf