Hydrofoil exercise to validate CFD analysis

Discussion in 'Hydrodynamics and Aerodynamics' started by quequen, Oct 1, 2014.

  1. Joakim
    Joined: Apr 2004
    Posts: 892
    Likes: 53, Points: 28, Legacy Rep: 422
    Location: Finland

    Joakim Senior Member

    How do you know? What is y+ on the surface? You also need a dense grid behind the foil, since a lot happens there. Also it is very important to extend the computational domain far away from the foil, especially downstream. If you don't get the wake correctly, you won't get the correct Cl and Cd.
     
  2. groper
    Joined: Jun 2011
    Posts: 2,483
    Likes: 144, Points: 73, Legacy Rep: 693
    Location: australia

    groper Senior Member

    For any of this to have real meaning, leeway and lift need to be accounted for, so the entire foil vertical and horizontal needs to be modeled accurately and also a flow angle which is appropriate for the leeway.

    It's an interesting problem, and certainly not an easy one. For comparisons to experimental data, I don't think there would be much data available in the public domain for this type of problem, these types of foils havnt been around very long and the people who have done the work so far are not going to give away any of their information, as they naturally seek to protect their competitive edge.

    When I looked into CFD simulations a few years back, the simulation needed to be run on a dedicated supercomputer server in order to get results with adequate resolution. The time could be bought from various service providers. Fine meshes of the domain, need to extend well downstream and a good distance all around. It takes a lot of computing power to handle it.

    I'm curious, What kind of boundary condition is suitable for the free surface anyway?
     
  3. Mikko Brummer
    Joined: May 2006
    Posts: 574
    Likes: 83, Points: 28, Legacy Rep: 258
    Location: Finland

    Mikko Brummer Senior Member

    Flowsim uses wall functions to resolve the boundary layer, so y+ is not relevant, basically you need enough cells to get a decent velocity distribution along the foil. You would need to do several runs at increasing resolution and see how it affects the forces. Yes, extending the wake is important, but as I recall wake refinement can be set automatically in the software based on velocity gradients. And yes, you want to use a large enough domain, but i believe 7-10 chord lengths behind the foil is sufficient. Here too you would run a few trials and compare how it affects the results. As i recall, not much with flowsim.
     
  4. Joakim
    Joined: Apr 2004
    Posts: 892
    Likes: 53, Points: 28, Legacy Rep: 422
    Location: Finland

    Joakim Senior Member

    Yes it is. You need to have y+ preferably 30-100 for wall functions to operate well. The upper limit can be higher in some cases or when the accuracy requirements aren't that high. 300 or even 1000 is sometimes suggested.
     
  5. Mikko Brummer
    Joined: May 2006
    Posts: 574
    Likes: 83, Points: 28, Legacy Rep: 258
    Location: Finland

    Mikko Brummer Senior Member

    I suppose this depends on the software, how the near wall treatment is done... there are near wall models that provide an analysis in the near wall region, as opposed to bridging the region from the turbulent main flow to the wall, with different requirements. My experience with FloEFD which we used extensively between 2005 to 2010, was that there was little difference in forces between 1 million or 10 million cells. There was more differences in the flow patterns. Even just 500,000 cells would give reasonable results.

    When it comes to lift, Flowsim should give good results for a foil like Quequen's as long as there is not much separation (with the exeption of the free surf effects). Maybe not quite as good for drag, but acceptable. Agree with Groper that you should include leeway, else the numbers are pretty academic for your foiler dinghy.

    Also agree with Daiquiri that the L/D appears too high - are you sure your drag is all the force in X-direction, not just skin friction?
     
  6. Joakim
    Joined: Apr 2004
    Posts: 892
    Likes: 53, Points: 28, Legacy Rep: 422
    Location: Finland

    Joakim Senior Member

    They seem to use some non-standard wall functions and use a different model for y+<300 than for a courser grid.

    They show results of Cf only for small y+ values in validation + for some higher values for bluff bodies, which are really far from a foil.

    http://www.hawkridgesys.com/file/solidworks-flow-simulation/enhanced-turbulence-modeling.pdf

    For a foil drag is mostly Cf at small AoA, thus you really need to get that right.

    What are the Cd, Cf and Cl in the results by quequen?
     
  7. quequen
    Joined: Jul 2009
    Posts: 370
    Likes: 15, Points: 28, Legacy Rep: 199
    Location: argentina

    quequen Senior Member

    Joakim, thanks for that file, I didn't know it.
    Mikko, perhaps you are right and I'm taking just friction. I'll check that.
    Regarding y+ even a high value like 1000 yields a cell thickness of 3mm or less, wich is unable to handle with cartesian grids and my PC capabilities.
    I'll try nesting 3 or more grid densities.
    In fact, trying to refine the grid around the vertical branch, I lost all my calculations in a hang-out. It seems that I can't handle more than 2000000 cells, may be less, with my configuration.
    I'll start all again in a few days.
     
  8. PatHanley
    Joined: Oct 2014
    Posts: 7
    Likes: 1, Points: 0, Legacy Rep: 23
    Location: United States

    PatHanley Junior Member

    The strut should also contribute to the drag (as modeled in the simulation) since it will produce a side force (due to the cambered section). This should decrease the overall L/D.
     
  9. Joakim
    Joined: Apr 2004
    Posts: 892
    Likes: 53, Points: 28, Legacy Rep: 422
    Location: Finland

    Joakim Senior Member

    Your profile may give even up to 100 L/D at 4 degree AoA and about half of that at 0 degree: http://www.tspeer.com/Hydrofoils/h105/h105.htm

    Your foil seems to have more area in the vertical part compared to the horizontal one. Also 3D effects will reduce L/D considerably. Thus you can't reach even half of the values above.

    Your simulation don't include free surface effects. Thus getting L/D of 30-40 may be OK for higher Re (>1M) at 4 degree AoA, if you don't have more area in the vertical part and you have a high aspect ratio.
     
  10. PatHanley
    Joined: Oct 2014
    Posts: 7
    Likes: 1, Points: 0, Legacy Rep: 23
    Location: United States

    PatHanley Junior Member

    This is a quick run on Stallion 3D with about 630,000 cells (Euler solver).

    @8 m/s and AOA = 4 deg
    Lift = 439 N
    Side = 431 N
    Pressure Drag = 24.8 N

    @8 m/s and AOA = 4 deg
    Lift = 1755 N
    Side = 1723 N
    Pressure Drag = 99 N

    The image shows the grid and the pressure coefficient along a cross section near the strut.

    [​IMG]

    Patrick Hanley, Ph.D.
    http://www.hanleyinnovations.com
     
    Last edited: Oct 8, 2014
    1 person likes this.
  11. Mikko Brummer
    Joined: May 2006
    Posts: 574
    Likes: 83, Points: 28, Legacy Rep: 258
    Location: Finland

    Mikko Brummer Senior Member

    Nice PatHanley. So it would seem Quequen's run was showing just the friction drag, although lacking some in the lift as well. Would the Stallion NS show the friction drag as well?

    Quequen - I understand that $1900 is a lot of money, but it is still a very good price for CFD software - an order of magnitude cheaper than most other commercial codes, including Solidworks. I guess you have access to some friends copy of Sworks.
     
  12. PatHanley
    Joined: Oct 2014
    Posts: 7
    Likes: 1, Points: 0, Legacy Rep: 23
    Location: United States

    PatHanley Junior Member

    Hi Mikko. Stallion 3D has a capability for analyzing the Navier-Stokes but currently only for laminar flows (low Re). With the inviscid analysis, it does not include friction drag in the Cd. The pressure drag accounts for form drag and vortex (induced) drag. In this case, a good approximation for the total drag is to add Cdo for the airfoil. Stallion 3D has a built-in airfoil analysis tool as well.
     
  13. daiquiri
    Joined: May 2004
    Posts: 5,371
    Likes: 258, Points: 93, Legacy Rep: 3380
    Location: Italy (Garda Lake) and Croatia (Istria)

    daiquiri Engineering and Design

    The doubts are even stronger after having manually checked the value of the drag.
    The minimum drag coefficient of Tom Speer's H105 airfoil is Cd,0=0.005 (according to the fig.20 of this paper: http://www.foils.org/hysecdes.pdf).

    The hydrofoil has the following reference areas (measured from the 3-D model attached in the OP):
    Wetted area: A = 0.579 sq.m
    Developed planform area: Sdev = 0.289 sq.m
    Horizontal projected area: Sh = 0.109 sq.m
    Lateral projected area: Slat = 0.187 sq.m​

    It means that at 8 m/s the bare profile drag of the complete hydrofoil is (in fresh water, Rho = 1000 kg/cu.m):
    D = 0.5 Rho V^2 Sdev Cd,0 = 46.3 N

    This value is the minimum theoretical drag for this hydrofoil. The actual total drag will be higher, because lift-induced, strut-interference and wave drag components have to be added too.

    So the value of 36.4 N cannot be correct, because it is less than 80% of the profile drag value which, as said, is the mathematical minimum.

    And what happens when AoA=4°? The drag of the H105 airfoil will be even higher. Again, from the Figs. 19 and 20 of Mr. Speer's paper, at AoA=4° we have the following values (approximate, because read from the graph):
    Cl = 0.55 (acting on the horizontal projected area, set at 4° AoA)
    Cd,h = 0.0065 (acting on the horizontal projected area, set at 4° AoA)
    Cd,lat = 0.0055 (acting on the lateral projected area, set at 0° AoA)​

    So, without taking into account the 3-D (finite AR), interference and free-surface effects, the max. Lift and min. Drag of this hydrofoil at 4° of AoA, 0° yaw angle and 8 m/s are:
    Lmax,vertical = 0.5 rho V^2 Sh Cl = 1918 N (vertical lift)
    Dmin = 0.5 rho V^2 ( Sh Cd,h + Slat Cd,lat ) = 56 N

    The 3-D, interference and surface effects will considerably decrease lift and increase drag.

    Assuming that Pat Hanley's inviscid analysis is correct, we could imo get closer to the real characteristics of this hydrofoil by assuming the following Lift and Drag values:
    1) Lift = from the inviscid analysis by P. Hanley = 1723 N
    2) Drag = Dmin + induced drag from the inviscid analysis = 56 N + 99 N = 155 N
    3) Hence, L/D = 11.1
    And it will still lack the free-surface and interference effect, which will deteriorate all the above numbers.
    The L/D ratio is relatively low in the analysed configuration because vertical lift is produced only by the horizontal portion of the foil, while the drag force is due to both horizontal and vertical parts of the blade. This situation would certainly improve by lifting the foil closer to the water surface, but only up to a certain point. The wave and aeration drag components become progressively bigger as the horizontal part gets closer to the surface - which means that, probably, there is an optimum foil depth.

    Sorry Quequen for not taking your calculated values into account, but your drag value is below the minimum value, and hence imo it has to be doubted until proven correct. ;)

    Cheers
     
    Last edited: Oct 7, 2014
  14. daiquiri
    Joined: May 2004
    Posts: 5,371
    Likes: 258, Points: 93, Legacy Rep: 3380
    Location: Italy (Garda Lake) and Croatia (Istria)

    daiquiri Engineering and Design

    Edit: expanded a bit and corrected some numbers.
     

  15. Joakim
    Joined: Apr 2004
    Posts: 892
    Likes: 53, Points: 28, Legacy Rep: 422
    Location: Finland

    Joakim Senior Member


    OK so the horizontal area is only 38% of the total area. You can't get high L/D with that.

    Isn't the Cl at least 0.8 for 4° AoA and some reasonable Re (2M?)? The rather low aspect ratio gives clearly less 3D lift in than 2D prediction.
     
Loading...
Similar Threads
  1. Maarten88
    Replies:
    5
    Views:
    999
  2. S V
    Replies:
    0
    Views:
    1,876
  3. Eytan Levi
    Replies:
    0
    Views:
    1,688
  4. container
    Replies:
    7
    Views:
    4,086
  5. vejas
    Replies:
    34
    Views:
    11,948
  6. revintage
    Replies:
    8
    Views:
    2,604
  7. revintage
    Replies:
    14
    Views:
    17,342
  8. rallyhybrid
    Replies:
    39
    Views:
    6,409
  9. optiwings
    Replies:
    1
    Views:
    2,482
  10. Tommifin
    Replies:
    5
    Views:
    3,102
Forum posts represent the experience, opinion, and view of individual users. Boat Design Net does not necessarily endorse nor share the view of each individual post.
When making potentially dangerous or financial decisions, always employ and consult appropriate professionals. Your circumstances or experience may be different.