Surface Roughness Drag Analysis

Hello, is anyone familiar with drag analysis in turbulent flow due to surface roughness using Ansys Fluent? I could really use your help.

I'm analyzing drag on a flat plate with homogeneous surface roughness and inhomogeneous surface roughness. I use 3 variations of surface roughness that are equivalent to sand grain roughness: P150 (111 microns), P100 (163 microns), and P60 (264 microns). I use tetra mesh, with y+ slightly over 30 to capture the effects of the surface roughness. I use the k-omega SST turbulence model. I've tried analyzing the drag with all three variations of surface roughness, but the differences between the results were not significant. Somehow, the P60 roughness resulted in a slightly smaller drag than P150 roughness. The validation I'm using is a wind tunnel experiment with the same roughness variations. In that simulation, the results between the roughness variations are quite reasonable, with P60 giving the highest drag.

Up to this point, I am so confused about what had gone wrong that I have been oblivious to. If anyone can shed a light on this issue then it would be very much appreciated. Thanks!

Comments

  • KremellaKremella Admin
    edited June 1

    Hello,

     The results you are seeing a somewhat surprising. I'm assuming that your convergence (in each case) is deep, right? I'm assuming that you are obtaining a steady value of drag. What is your flow Reynolds number?

    Also, since this is a flat plate, could you please try a y+ =1 mesh and test the same thing?

    Thank you.

    Karthik

  • regitasyaliregitasyali Member
    edited June 1

    I set the residuals to 1e-4 for the convergence criteria for all cases. Yes, somehow I have been obtaining a similar drag value for different roughnesses. I have to run each variation of surface roughness on 5 Reynolds number (110k, 220k, 320k, 430k, and 540k). 

    I have also tried to run the simulation using a y+ = 1 and still got the same results. Although, I have a question regarding the first layer thickness that I calculated using this site https://www.cfd-online.com/Tools/yplus.php. Is the estimated wall distance measured from the wall to the centroid of the first cell adjacent to the wall or is it measured from the wall to the top of the cell adjacent to the wall? Do you think it would account for a significant change in the results? I've tried both possibilities but still, I got the same results.

    Thank you for your concern!

  • KremellaKremella Admin
    edited June 2

    It is up to the centroid of the first cell. Moreover, as long as you have sufficient resolution, this does not matter.

    How similar are your values? Also, just to ask you again, the drag values are all constant with the number of iterations, and is not fluctuating - right?

    Does the value change if you try to converge the solution to 1e-5 or 1e-6?

    You are also opening Fluent in double-precision, right?

    Thanks.

    Karthik

     

  • regitasyaliregitasyali Member
    edited June 2

    For the 0.111mm and 0.163 mm roughness, the drag increase reached 8.5% and 12.7% from the smooth condition respectively. For the 0.264 however, it only reached 2.4%. Whereas in the wind tunnel experiment, the drag increase for each roughness should be 35% for 0.111 mm, 90% for the 0.163 mm, and 130% for 0.264 mm.

    "Also, just to ask you again, the drag values are all constant with the number of iterations, and is not fluctuating - right?"
    - From what I can tell from the Drag plot during the simulation, yes the values are constant, always below 0.002.  

    I've always tried to converge to 1e-4, but now I'm trying to converge to 1e-5 and see what happens. Yes, I have been using Fluent in double-precision.

    Where do you think the problem is?

    Thank you!

  • KremellaKremella Admin
    edited June 4

    The reference values of area, density, and velocity are correct? Between your experiments and simulation?

    Also, what is the Y+ value of your mesh? And which turbulence model are you using?

    Thank you.

  • regitasyaliregitasyali Member
    edited June 4

    I used the default reference value of area, because once I tried using the actual value, the results went below expected. I also used a different density but recalculated the results so it would match the density I'm using. I only maintained the values for the velocities.

    I am using y+ = 1 and k-omega SST turbulence model. 

    Thank you!

  • KremellaKremella Admin
    edited June 8

    Especially when you are comparing values using experiments, it is very important to use the same set of scaling parameters. I'd strongly recommend finding out the reference values used in experiments and compare your simulations using the same set of values.If they were published in a paper, writing to the authors and asking them about it might help.

    Thank you.

    Karthik

Sign In or Register to comment.