Junk rig CFD

  • 27 Oct 2024 21:35
    Reply # 13424040 on 13423701

    Very interesting and neat, Paul!

    As the price of the runs on AWS or similar is so (too! wow...) high, what about if instead of chipping in money, we chipped in computer cycles - can the calcs be run in a distributed manner, a la Seti@Home, or Bitcoin blockchain, or something along those lines?

    Apache HADOOP?
    BOINC maybe?  https://boinc.berkeley.edu/

    Just a thought. Thank you for the work you have put in to answering these questions of the ages! ;)

  • 27 Oct 2024 20:13
    Reply # 13424022 on 13423701
    Anonymous member (Administrator)

    To both Arne and Ueli I would suggest: don’t forget, the purpose of these basic models is to find comparative data, on just an arbitrary section of a junk sail panel.

    It’s a first step towards establishing some basic principles, including that huge mystery, the effect of the mast on a lug sail.

    A "partial view". It is not (and it is not meant to be) the same thing as designing a sail – much less, designing a complete boat and hull package. 

    By the way Arne, are you sure your diagram from Marchaj of the “more realistic sail” depicts an A.R of 3? Not that it matters much in regard to the point you are making, but for the record, it looks a lot more than that to me.


    Eric wrote "... it could be possible and not too costly, to use some photogrammetry technology to model a given sail ... the surface could then be "sliced" into 2D profiles to feed the CFD". 

    David can provide real life profiles for his articulated-batten wing sail, but for a cambered sail on rigid battens we can't do that, of course.

    Eric's idea is good, especially if you have the skill and equipment to do it, but if a boat is tied up to a pontoon with an inflated sail, I would suggest a better way to start with, just take one "slice" - simply take a series of measurements along the centre chord of a panel, and get the profile by direct measurement. I think that would be easy enough to do. We only need one "slice" for a 2-D model to start getting good comparative data from the computer model, as Paul has demonstrated.


    [Some time ago I made a series of different model sail panels from cloth, inflated them and took measurements along a vertical line between the battens, in order to look at the profile (shape of section) along that vertical line. 

    It was not difficult to do, (and it also showed that the vertical section curve has a lot less to do with the way the soft cloth sail is constructed than most people would think - but that's another story  The same may well be true regarding the horizontal section curve, which Paul requires).

    I am kicking myself now that I did not at the same time take measurements along the horizontal line, the central chord, but that was not what I was thinking about at the time. I threw out all the model panels a few months ago. Anyway, a real life sail could be measured up just about as easily, to obtain the profile of a single "slice", (and it would be interesting also to compare the horizontal sections of differently constructed sail panels).

    (So much for rigid batten cambered sails. How you would deal with the camber that is supposed to be induced by twisting flat cut fanned sails is in the "too hard basket" for me, but I dare say it could be done somehow - maybe by using Eric's photography idea). ]


    Last modified: 29 Oct 2024 23:08 | Anonymous member (Administrator)
  • 27 Oct 2024 15:23
    Reply # 13423969 on 13423701

    hi arne

    Arne Kverneland wrote:

    Paul,
    hm, now I have slept on it and then opened C.A. Marchaj’s Sailing Theory and Practice.
    His numbers regarding resulting Lift/drag ratios differs quite a lot from the numbers you have found.

    Marchaj operates with alpha angles between 5 and 20 mostly, and for upwind work these are 10-15°. The resulting lift/drag will then only be 6/1 on a High-AR sail, that is, very far from the 20:1 and thereabouts as you find.

    The very small alpha angle you are working with is unrealistic on a soft working sail…

    you got a good point, as the l/d ratio of a sail is far from being the number that shows the real windward performance, as the better part of a boat's drag is in the water… 

    checking 'sail 2' from marchaj's book shows similar l/d ratios as the simulation from paul schnabel at an alpha of ~5° – not much drag, but not enough lift to move the boat on a speed to let the keel really work.

    sailing is more complicated. it works with two wings (at least), one in the air and the other in the water.

    the pure sail simulations still show interesting results, but we need to interpret them according to the real world 'below the sail'.

    ueli

  • 27 Oct 2024 14:21
    Reply # 13423951 on 13423701
    Anonymous member (Administrator)

    Paul,
    hm, now I have slept on it and then opened C.A. Marchaj’s Sailing Theory and Practice.
    His numbers regarding resulting Lift/drag ratios differs quite a lot from the numbers you have found.

    Marchaj operates with alpha angles between 5 and 20 mostly, and for upwind work these are 10-15°. The resulting lift/drag will then only be 6/1 on a High-AR sail, that is, very far from the 20:1 and thereabouts as you find.

    The very small alpha angle you are working with is unrealistic on a soft working sail.
    Fig 83, below shows a super-good ‘sail’ and even this has a lift/drag of just 6.4:1 and with Alpha = 15°.
    The next diagram shows a more realistic sail with AR =3.

    Arne

    (Diagrams under Arne's sketches, section 8, photos 13 and 14)

    Last modified: 27 Oct 2024 14:23 | Anonymous member (Administrator)
  • 27 Oct 2024 11:26
    Reply # 13423919 on 13423701
    Anonymous member (Administrator)

    Bonjour

    It's a very interesting approach but the start of a very long process.

    As always the firsts cases are rather trivial and the results are obvious.

    Obviously a flat panel is not rigid and it develops some sort of curves under the airflow pressure, but next to the batten. So the flat profile that has been computed represents the upper or under limit of a pannel.

    In the case of a junkrig, it seems quite reasonable to consider a pannel as an independant 3D aerodynamic surface.

    The Cl/Cd approach is interesting to work on the flow but is not representative of the thrust and drug on a boat as the angle of attack is ofset of the boat speed vector from about 45° (at best). The theoretical benefits of a wingsail over a sail for cruising boats

    As the sail is a 2D surface, it could be possible and not too costly, to use some photogrammetry technology to model a given sail ( ideally at different incidence angle and wing speed) with a boat tided to a ponton. The surface could then be "sliced" into 2D profiles to feed the CFD.

    A 2D strength of material software tool could be used to compute the forces (pressures) involved to provide the pannel 2D shape identified by the photogrammetry. 

    Eric

    Last modified: 27 Oct 2024 15:00 | Anonymous member (Administrator)
  • 27 Oct 2024 09:23
    Reply # 13423907 on 13423701

    Some flow visualizations of the flat cut profile:

    Port tack: 2, 4, 6, 8, 10 deg AoA

    Starboard tack: -2, -4, -6, -8, -10 deg AoA


    Basically, the dark blue/ violet areas are detached zones. I would say, at about 7-8° on both tacks the detachment bubble reaches the luff and the sail is fully stalling - but still drawing.

    A flat cut sail profile is called a "flat plate" in aerodynamics. A lot has been investigated with this most basic profile over the past decades. The flat plate detaches almost immediately at the leech with just a tiny little bit of AoA.

    (interesting enough, that the detachment bubble resembles the shape of a cambered profile at the L/Dmax point. It is then acting pretty similar to the cambered profile, with the detachment bubble "being the camber". However, the difference is the highly increased friction and the flat pressure side of the profile).


    Cheers,

    Paul

  • 27 Oct 2024 08:59
    Reply # 13423905 on 13423701

    Fantastic, David, that's exactly what I was looking for!!

    It's fine like this, I'll need to remodel anyway.

    Thanks,

    Paul

    Last modified: 27 Oct 2024 08:59 | Anonymous member
  • 27 Oct 2024 08:51
    Reply # 13423904 on 13423701

    Paul, here are sections through the final form of Weaverbird's wingsail, which, as a seat-of-the-pants sailor, I found to be quite alpha-tolerant. Let me know if it would be helpful to have the DXF file.

    2 files
  • 27 Oct 2024 08:10
    Reply # 13423899 on 13423701

    Thank you, for your kind replies and interesting thoughts!

    Though I am relatively new to the JRA, it appeared to me pretty soon that the one thing missing (or at least very rare) among us is the possibility to evaluate and put numbers to those ideas that a lot of clever people develop in the JRA. If those wonderful ideas can’t be evaluated and tested, they stay theories and speculation. And that would be an absolute pity! Wind tunnel tests are naturally very expensive, and for a reason: huge machinery involved, very sensitive sensors and data acquisition, lots of people involved. Though CFD is still a huge amount of work, frustration and debugging, it is so much less compared to wind tunnel tests.

    One thing: I would like to give advice from my professional experience: we all tend to like the idea of looking at the whole thing. The whole airplane, the whole car, the whole junk rig in 3D. From what I experienced, that is a very human wish. However, there is a lot to be learned from only partial views, like a 2D case. You do not need to simulate the entire car if you want to install a new side mirror :-) Also, the budget is always limited. Always. It would also be limited in case crowdfunding would produce reasonable results. In my opinion, it is best to simulate as simple cases as possible – sticking with the KISS principle. For example: mast balance can be easily investigated in 2D, without relevant compromises. On the other hand: the different sail plan forms, yard angle, batten angle  or the tip vortex would have to be simulated in 3D. Looking at the complete rig in 3D in a CFD simulation is what I, personally, would really love to do. But let’s stick to what we have available right now, 2D, which is already huge and can be used for a vast amount of information.

    I’m not so sure if crowdfunding from the JRA members would be sufficient to pay the machine rent necessary. I can only estimate right now, but I think a 3D case simulation would be about 500-2000€ for each run. A run is one data point in my graphs. To produce the graphs of my first post below, I already had to do 70 runs. If done so in 3D, that would be 35000 – 140000 €. Of course, a lot of those data points can be omitted (more to that later), but still. Now imagine doing different velocities on top, different mast balances, etc…

    However, I could imagine doing a phd about this topic (in which case I would also publish in peer reviewed journals, Mauro ;-) ). I am actively looking for universities/insitutes to host such a topic – and also pay a little salary to feed me. Most of the technical universities are equipped with quite some computational power, so the costs of machine rent wouldn’t be such an issue. Let’s see if the topic is relevant enough to be of interest. If you have contacts to anyone related, please let me know!

    Graeme, Paul: about a JRA CFD/aerodynamic seminar: why not? That would be an interesting group!

    Back to the technical part:


    Good/bad tack

    Graeme, I agree with you. Good questions from your side! Other than with the flat cut sail, I also think that the difference is hardly sensible out on the water with a cambered sail, but maybe measurable. The L/D ratio is not directly translatable to windward performance, other factors like the hull shape might “reduce” (but not turn) the difference between good and bad tack.


    Alpha tolerance

    Again, Graeme, your explanation is spot on! The sharpness/bluntness of those lines, with respect to the tangent, is the alpha-tolerance. To have it clearer, I made a graph of L/D ratio over angle of attack. Same data, just a different way of plotting it:

    The blunter the curves, the more alpha-tolerant the profile. The sharper the hill (the area of max L/D), the more you need to focus while at the helm. The flat cut profile on stb tack definitely has the highest alpha-tolerance compared to the three other curves – however, at a price: it is least efficient. Actually, it’s alpha-tolerance is so good because it stalls almost immediately when sheeting in…


    Further calculations

    Yes, the mast balance is one of the things I will have a look at. Not sure which results to expect in advance, quite adventurous :-) Also, the SJR, wingsail and aerojunk are on top of my list.


    Process Speedup

    Those first simulations I did from 0 to +-15°/20° AoA, to get an impression how everything works out. However, to evaluate the performance of the junk rig, we do not need that much data points. Of main interest will be the L/Dmax point, which is around 4-5°, as well as the alpha-tolerance, which is described be the curvature of the L/D over AoA graph. The points of interest of the already finished simulations I did mark in the following graph:

    Also, the empty data in between the data points can be easily interpolated by a spline curve. So, instead of calculating every degree from -15° to +15° (30 simulation runs), I would only do every second degree from -8° to -2° and +2° to +8° (8 simulation runs). Quite a speedup! If in the data evaluation phase, after all runs are finished, it appears that it would be good to go i.e. up to 10° or have one further data point necessary between 2° and 4°, it is no problem to execute another run afterwards.

     

     

    I hope to start a lot of discussions with this topic, some ruffling of feathers but also some calming of feathers. I am very open to all of your input! For example, I would be very interested if one of the SJR/wingsail/aerojunk experts might draw a typical profile curve, which I could then use for simulation. Also, I would be very interested in feedback to the cambered profile I used (and simply made up).

     

    Cheers,

    Paul


    Edit:

    David, you have been faster typing than me :) The flat cut sail develops a detached area almost immediately when sheeting in. I'll post some more material to illustrate this.

    I fully agree about your 3D panel shape comment. That's why this CFD data still has to be interpreted wisely, and not taken granted as it is. Do you have any actual shape measurements of the cambered panel at hand?

    Yes, I intend to do a magazine article about this. But I would like to discuss things first here, to soften the edges, rule out errors from my side and put together a more round story.

    Last modified: 27 Oct 2024 08:22 | Anonymous member
  • 27 Oct 2024 07:59
    Reply # 13423898 on 13423701

    Fascinating, Paul! Well done!

    As well as the points that Graeme has made, I think I see that the flat sail stalls at about 8˚ - 10˚ on both tacks, and I can't imagine why. 

    Also, the  3D cambered panel is more like a flat sail at and near the batten, only developing full Cl/Cd at and near its mid height; so the actual "real life" figures are going to be somewhere in between those two cases.

    No doubt we'll be seeing all this laid out in a magazine article in due course. I look forward to it!

       " ...there is nothing - absolutely nothing - half so much worth doing as simply messing about in junk-rigged boats" 
                                                               - the Chinese Water Rat

                                                              Site contents © the Junk Rig Association and/or individual authors

Powered by Wild Apricot Membership Software