Flowfield calculations in javafoil

Discussion in 'Hydrodynamics and Aerodynamics' started by johan gronvall, Dec 30, 2014.

  1. markdrela
    Joined: Jun 2004
    Posts: 307
    Likes: 30, Points: 28, Legacy Rep: 324
    Location: MIT Aero & Astro

    markdrela Senior Member

    You're obviously seeing some bad numerical effect from this panel method. All those airfoils should have nearly identical lift and profile drag curves. I'd keep increasing the thickness until the lift curve doesn't change significantly. Then don't use a section thinner than that.
     
  2. tspeer
    Joined: Feb 2002
    Posts: 2,319
    Likes: 303, Points: 83, Legacy Rep: 1673
    Location: Port Gamble, Washington, USA

    tspeer Senior Member

    I would think that if you're getting that much variation from such a minor change in thickness, that something is seriously amiss and I wouldn't trust those results.

    Edit: Hi, Mark! I didn't see your post. It's good to see we're saying the same thing. These Javafoil results make me appreciate XFOIL even more.
     
  3. johan gronvall
    Joined: Dec 2014
    Posts: 7
    Likes: 0, Points: 0, Legacy Rep: 10
    Location: Sweden

    johan gronvall Junior Member

    Alright, when I have time I'll try XFOIL, maybe from XFLR5 - do you have experience regarding ease of use?
     
  4. tspeer
    Joined: Feb 2002
    Posts: 2,319
    Likes: 303, Points: 83, Legacy Rep: 1673
    Location: Port Gamble, Washington, USA

    tspeer Senior Member

    XFOIL has a command-line user interface that has a definite learning curve. It's easy to use once you know what steps you need to take. There's a GUI for it called Profili, which I haven't used, but many say it makes XFOIL much easier.

    I think the hardest part about learning XFOIL is picturing the work flow of a typical session.
    - You start at the top level by LOADing the coordinates or generating a NACA section as a starting point. You may want to use PANE to interpolate well-distributed points around the section.
    - Next, you drop into GDES to look at the section and make any geometric modifications (thickness, camber, flap deflection). Use EXEC to save the changes in the airfoil buffer, and return to the top level. You can SAVE the coordinates from the top level.
    - Then go to OPER and use ALFA to compute the pressure distribution at an angle of attack. From here, you can either run a polar sweep or you can go back to the top and use one of the inverse design modes to reshape the section. You will probably want to use VPAR to change some of the boundary layer parameters from their default values.
    - If you want to run a polar sweep, you use PACC to open a polar data file to accumulate the data, and then use ASEQ to compute a range of angles of attack for the polar.
    - If you want to redesign the entire section, then from the top level you use MDES. You can edit the pressure distribution and EXEC to create a new shape that will have approximately the specified pressures. There are some mathematical constraints on the pressures to ensure the section is a closed shape, etc., so you may not get quite what you specified. Use PANE at the top level to make the redesigned section your new working section, and SAVE the coordinates if you want to keep it.

    That's a rough outline of a typical XFOIL session.
     
  5. 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

    XFLR5 is imo definitely an easier way to go for a beginner. It has a very nice GUI and a VLM solver for multiple lifting surfaces.

    Cheers
     
  6. Humberto
    Joined: Mar 2008
    Posts: 47
    Likes: 2, Points: 0, Legacy Rep: 55
    Location: Spain

    Humberto Junior Member

    I just want to share some tests I have made with JavaFoil. First of all, I would say JavaFoil is quite "picky" regarding sampling points, that is, the results are quite influenced on the distribution of these.

    I routinely use JavaFoil in order to evaluate keel lift and drag, so that I can study how small variations affect its characteristics. The math behind is I do 2D calculations at different keel sections, and then estimate the final 3D flow with empirical formulations. I am not that interested right now in the 2D to 3D business, but in a good estimation of the 2D characteristics.

    I have developed a software that does all the geometry processing (slices the 3D NURBS representing the keel and produces the foil sections) and the calls JavaFoil for each of the foil sections (my software is written in Java, the main reason to use JavaFoil). Because the keel shape has continuity, the lift and drag of each slice should not differ too much related to its neighbors. If we plot the 2D lift coefficients of each section, the curve must be continuous as well.

    When I first tried this procedure I did not get any reasonable plot. There was way too much variation among close by sections, as shown in the left topmost plot of the screenshots. But sampling quality was good enough (at least in my opinion), as shown on the centre topmost drawing of the screenshots. Wha I finally did was to modify the sampling process so that all samples maintained symmetry along the camber line. In theses case, things went much better. I attach both screenshots. I believe that now the JavaFoil results are more or less reasonable.

    Best regards
    Humberto
     

    Attached Files:

  7. 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

    Very nice work, Humberto.
    It is interesting to note that characteristic curves of the finite-wing did not change at all after the smoothing of the spanwise lift distribution. How plausible is that?
     
  8. Humberto
    Joined: Mar 2008
    Posts: 47
    Likes: 2, Points: 0, Legacy Rep: 55
    Location: Spain

    Humberto Junior Member

    Well, there are somebdifferences, but they are barely noticeable in the plots. In the order of 1%
     
  9. 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

    Ok, but 1% is an extremely small difference, considering how different are the shapes of the spanwise lift distributions before and after smoothing.
    So my question is - how come such a big difference in the spanwise lift distribution gives an almost negligible difference in the characteristic curves of the wing?
    If the Cl-Alpha and Cd-Alpha curves of the wing have been obtained through the integration (summation) of the spanwise lift distribution, then different spanwise distributions should be reflected in the integrated values. Unless you have used a different mathematical model for the calculation of the finite wing, which doesn't take into account the results of the spanwise airfoil calculations?
    Just thinking out loud...
     
  10. Humberto
    Joined: Mar 2008
    Posts: 47
    Likes: 2, Points: 0, Legacy Rep: 55
    Location: Spain

    Humberto Junior Member

    The procedure I follow is more or less what you say. But in this ver example, if you integrate the lift distribution, weighted by the actual chord length of the sections, the differences are not that large. But in an optimization process, they are definitively important, because they introduce some "random noise"
     
  11. 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

    Frankly, I find it a bit odd, and imo you should investigate this thing.
    How about running different test cases, with different airfoils and wing shapes and see what happens before and after smoothing the spanwise lift distribution? If you see that the finite-wing curves keep being unaffected by the smoothing procedure, then you have two options:
    1) you cannot find a rational explanation for this behavior. In that case, if I were you I would be suspicious of the results obtained through these calculations.
    2) you can explain it physically or mathematically, in which case you will prove that my doubts are unfounded and hence will be able to trust the results. :)

    Cheers

    P.S.:
    Out of curiosity, I have overlayed the two lift distributions (before and after smoothing), scaled them properly in order to make the axis scales coincide and, finally, have colored the positive and negative differences. This is the result:

    Lift distribution comparison.jpg

    In order to have identical values of the integrals of the two curves, and hence the same finite-wing lift, the overall red area should be nearly identical in size to the total blue area. The positive differences would then compensate the negative differences between two curves, and the net difference between their integrals would be nearly zero.

    But it is evident that the red area is much bigger (visually, much more than 1%), hence the integral of the non-smoothed lift distribution should have a lower value than the integral of the smoothed curve.

    Another reason to double-check the results of the software's output, IMO.
     
    Last edited: Feb 19, 2015
  12. brian eiland
    Joined: Jun 2002
    Posts: 5,067
    Likes: 216, Points: 73, Legacy Rep: 1903
    Location: St Augustine Fl, Thailand

    brian eiland Senior Member

    Just found this subject thread, and read thru it while 'attempting' to understand it....ha...ha. I think this computer analyzing technology is very interesting, but beyond my capabilities to use or understand it. I'll leave it to the younger guys. :D

    That said Johan I have a question to ask of you, as I was also intrigued by Arvey Gentry's explanations concerned mainsail and jib interactions. I have utilized his observations to justify a mast-aft sailing rig where I favor the genoa over the mainsail to the extend that I replace the traditional mainsail with a parallel staysail (thus 2 headsails), and claim a higher potential performance....
    http://www.boatdesign.net/forums/sailboats/aftmast-rigs-623-2.html#post51730

    http://www.boatdesign.net/forums/sailboats/aftmast-rigs-623.html#post23390

    http://www.boatdesign.net/forums/sailboats/aftmast-rigs-623-2.html#post110983



    Tom Speer has contributed a number of posting to that subject thread in the past, and has suggested that I will likely end up with considerable drag factors that will degrade the performance gains I was seeking (and I do very much respect his views). But I have also had some posters who do not feel that my double headsail configuration will be any better than a normal sloop configuration, and this combined with the extra drag would be 2-strikes against it?

    I'm just wondering if your new 'research' might shed any new light on the double headsail arrangement on my rig. If you have the time, I thought this subject would be aligned with the subject you are studying

    thanks and regards,
     
  13. Humberto
    Joined: Mar 2008
    Posts: 47
    Likes: 2, Points: 0, Legacy Rep: 55
    Location: Spain

    Humberto Junior Member

    Well, I can have something wrong. As I said previously, I was more interested in the 2D lift than the finite span wing. Said that, you overlayed the wrong curves. Those curves plot the Cl of the differnte sections. But it is not a constant chord wing. In fact, it is an L shaped keel-bulb, so that lower chords are larger, and as such, they have a different influence in the finite wing lift. Additionally, I follow a modified wessinger lifting line wing model for the finite wing calculations. For each section, the term that you put in the integral is the section lift weighted by the local chord and a term related to an asumption on the lift distribution and the IACs, aerodynamic coeffcients.

    While I am still not very sure of the accuracy of this estimation, I have run a series of VPPs and the results do not look like they are that far. For optimization purposes gives reasonable results. IMHO
     
  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

    Ok, I couldn't know that. :)
    I thought that you have shown the screenshots of the same wing and same sections in your post #21. Actually, everything does look identical between the two pictures, except for the shape of the lift distribution.
    Cheers!
     

  15. Humberto
    Joined: Mar 2008
    Posts: 47
    Likes: 2, Points: 0, Legacy Rep: 55
    Location: Spain

    Humberto Junior Member

    The keel is the same for both screenshots. The left topmost curve shows the quality of the 2D section computations, but for the finite wing, what matters is the left bottommost curve, which includes weighting of the local Cl with the actual section chords. In this example the difrences are not that large in this case, but this is not true with other appendage shapes.

    In any case what I wanted to show is how sampling affects quality in JavaFoil. And, although the differences might not be too large for a rough approximation, in an optimization process (which is what I use this for) the sampling introduces random noise, that is differnces in lift and drag that are not due to sections characteristics but on sampling, which depends on how the NURBS surface is evaluated
     
Loading...
Similar Threads
  1. Rabah
    Replies:
    12
    Views:
    7,028
  2. vejas
    Replies:
    34
    Views:
    11,945
  3. Surfer Naval Architect
    Replies:
    0
    Views:
    1,775
  4. TrulsT
    Replies:
    2
    Views:
    11,024
  5. Jgerstemeier
    Replies:
    8
    Views:
    2,175
  6. samar87
    Replies:
    12
    Views:
    3,050
  7. marlin974
    Replies:
    2
    Views:
    2,175
  8. athvas
    Replies:
    15
    Views:
    9,965
  9. sottorf
    Replies:
    39
    Views:
    9,200
  10. ADAM87
    Replies:
    22
    Views:
    13,185
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.