Junk rig CFD

  • 30 Oct 2024 14:04
    Reply # 13425206 on 13423701

    hi paul

    ii'm still not sure if your comparison of different sails based on the best l/d ratio is that useful, as long as most sails will give those 'best numbers' in a range which is not usable in the real world.

    the actual america's cup yachts with (compared to us narmal sailors) generate almost no drag from the hull may use angles of attack of something around 10°. but even with those high-tech-foilers 5° are not a reasonable value.

    by comparing your 9% cambered sail on both tacks, the best l/d ratio (around 5°) is just a little bit better on stb tack than on port tack. but by comparing at 9° the difference between the two tacks is something around two to three with a much higher total force on the tack of the better l/d ratio. we should interpret your data like that as we need some minimal power to propel our boats…

    ueli

    Last modified: 30 Oct 2024 14:05 | Anonymous member
  • 30 Oct 2024 11:13
    Reply # 13425126 on 13423701

    Hi,

    Eric, we know that every geometry in moving air has aerodynamic forces acting on it. They can be summed up in one total force vector, having a length and an angle. It is now just a question of definition, how we decompose this total force vector. The Lift (Cl) and Drag (Cd) concept decomposes the total force vector into one part parallel to the flow direction and one part normal to the flow direction. Useful for different cases, i.e. aviation. Now, there is no objection in decomposing the total force vector into different directions: for example into the boat’s course direction and lateral direction. Useful, as well. However, just because we decide to use another convention of representing the forces on the geometrie does not change the actual total force vector. The conclusion we can draw from both conventions are the same!

    Cl is the Lift, Cd is the drag, just written down in dimensionless form to get comparable coefficients instead of specific forces.

    Arne, I repeat: the aim of my 2D profile simulations is not to find the perfect AoA of a real sail to the wind. This information could simply not be drawn from my 2D simulations. I chose the LDmax point to have a variable on which to compare the different profiles. I take it on my head, that this preliminary choice of mine (to use LDmax as the comparable variable) was not the best. That’s why I suggested to open up from one variable to an interval of AoA for comparison between different profiles. It would also be possible to just compare max Cl between for different profiles, but I think it won’t be good as at max Cl the profile is already stalling.

    What is sure from Marchaj’s diagrams here, is that the  5° AoA, with the lowest drag angle, is not the setting which will get a boat fastest to windward. One needs some brute force as well. Don’t forget the boat. – Arne K.

    Yes, exactly! I do not argue on this one.

    In the end, I want to compare profiles to each other. A hull does not have an effect on this comparison. To give a real-world example: think about Ilvy, being rigged as she is now, cambered, and next year rigged as SJR. Given the same weight of the hull (plus 2-3 kg gained body weight from Christmas), the hull can be ruled out when comparing those two rigs.

    To compare the profiles, my suggestion is to compare the ClCd curve, at the interval between deflating luff and fully stalling flow (which is about 5° to 15° AoA). Anyway, I’m not emotional on this: Any suggestions or better ideas on how to compare the performance of profiles in this case are very welcome!

    Telltales at the leech  -  useful or not. – Arne K.

    Arne, I think leech telltales are very useful, indeed! They have been in use with junk rigs as well as Bermuda rigs, airfoil flow visualizations as well as kites. I’m not debating on their usefulness, when I try to investigate what exactly they are showing, and question if they show the most efficient sheeting angle. Regardless if they do or don’t point to the most efficient angle, they depict a very effective, practical and proven way of feedback from the sail – which works!

    (Btw. when watching videos on Youtube, showing stall-tests of aeroplanes, the tufted wings clearly indicate that separation starts at the trailing edge...) – Arne K.

    I assume you watched stall-tests of thick profiles – at least thicker than a single-ply sailcloth can ever be. That nose radius on an airfoil has a significant effect: i.e. the luff is detaching way later when the nose is not sharp but has a radius. Slieve nicely described it in his “Some thoughts” writeup, and also David must have thought about it a lot when designing his wingsail – and for a good reason. The thickness of the nose, the nose radius, does the trick! Regarding performance, a single-ply sheet, even if cambered, can not be as good as a profile with a nose radius.

     

    Graeme, I’m sorry, I should have used the common sail-related terms: suction side is the leeward side of the sail, pressure side is the windward side of the sail.


    Cheers,

    Paul

  • 30 Oct 2024 09:12
    Reply # 13425103 on 13423701
    Anonymous member (Administrator)

    Keeping one boot in the boat...

    Marchaj’s test foil, which was used to produce Fig. 83, was rigid, and no doubt just made of a sheet of metal, so would not collapse even with the AoA set to 0°. Still, the lowest AoA he tested here, was with it at 5°, and that on all three foils.
    This was probably because he knew that real sails, made of sailcloth, would collapse well before that.

    I have now drawn in the AoA = 5° case of sail 2; the sail with 10% camber. I found that  the drag angle (Epsilon) drops from 8.9 to 5.3°. However, this comes at a cost  -  the lift drops with 32%.
    In next diagram (Fig 84, not shown) Marchaj tilts Fig. 83 to show three sailing angles, 30, 60 and 90° to the apparent wind. To my surprise, sail no 3 (12.3% camber) comes out with the highest drive force when close-hauled, with AoA at 15°. Sail 2 comes in just behind, also with AoA at 15°.
    The price to pay for sail 3 is higher side force, so not every boat could make use it.

    What is sure from Marchaj’s diagrams here, is that the  5° AoA, with the lowest drag angle, is not the setting which will propel a boat fastest to windward. One needs some brute force as well. Don’t forget the boat.

    Telltales at the leech  -  useful or not.

    My leech telltales are cheap and not that accurate way of warning against a stall.

    • ·         First of all, they are helpful in shouting to me when a wind shift or bad helming have stalled the sail completely.
    • ·         Sailing upwind in light winds, I tend to sail close to stalling, with these telltales flying –  mostly. This clearly gives the best speed, at the cost of pointing angle.
    • ·         As the wind picks up and the toerail approaches water, I usually head up a little, to point higher. Keeping an eye on the wind indicator helps. As the wind increases more, I end up flirting with luffing angle, until I find it wiser to drop a panel.

    Whether the leech telltales warn me too late or not about stalling, I dunno. Maybe that is true, but it cannot be that bad, and anyway, that’s what I have.

    (Btw. when watching videos on Youtube, showing stall-tests of aeroplanes, the tufted wings clearly indicate that separation starts at the trailing edge...)

    Arne


    Last modified: 30 Oct 2024 10:23 | Anonymous member (Administrator)
  • 30 Oct 2024 02:21
    Reply # 13425064 on 13423701
    Anonymous member (Administrator)

    Paul, I am having a bit of trouble understanding your interpretation of these diagrams.

    Take these two, for example:

    The model of AoA = 2 dgerees shows a bubble of detached flow on the pressure side of the luff - then the model of AoA=10 degrees shows a bubble on the "suction" side?

    I must be misinterpreting the images.

    I presume the detached bubble is the part in dark blue/violet?

    Then, could you please mark on the diagram, or describe the boundary between the pressure and "suction" sides? (Or put it another way, which part of the diagram depicts the sail section?)


    Last modified: 30 Oct 2024 02:35 | Anonymous member (Administrator)
  • 29 Oct 2024 21:00
    Reply # 13424976 on 13424703
    Anonymous member (Administrator)
    Anonymous wrote:

    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).  – Eric A.

    Eric, I do not quite understand what you mean. Are you referring to True Wind Angle (TWA) and Apparent Wind Angle (AWA)? The simulation I ran have an Angle of Attack (AoA) towards an incoming flow. Transferred to a boats coordinate system, that AoA from the simulation translates to the angle of the sail towards the AWA – not the TWA.

    The Cl/Cd approach has been invented in the aeronautical world, but it is not at all relevant in boating.

    In aeronautic, the speed vector of the aircraft is almost in line with the zero lift cord of the wing (forget stalling aerobatics).

    So the thrust of the engine (or the horizontal gravity force in a glider) must be almost equal and opposed to the drug of the wing (the aircraft in fact). So the Cl/Cd value is relevant to evaluate the efficiency of a wing. 

    On a sailing vessel it is not at all relevant. At first, the drug of the sail... is rediculous in regard to the drug of the hull in the water. segondly, the drug vector of the wing is ofset from the boat speed vector from about 40° headwing. Downwind it is just a non-sence; the drug of the sail is the powering force.

    For a sailing vessel, the important decomposition of the thrust vector of the sail is the powering thrust and the heeling force. The heeling force should be sufficiently low to allow the boat to sail in good condition in order to maximize the powering thrust.

    As such, it might be interresting, for example, to ease the sheet to increase the powering thrust while reducing the heeling force and, incidentaly, the thrust vector. It is what is done on racing boats when you ease the mainsail sheet to reduce the heeling moment.

    Eric

    Last modified: 29 Oct 2024 21:07 | Anonymous member (Administrator)
  • 29 Oct 2024 16:33
    Reply # 13424821 on 13423701

    Hi again,

    Now to the critics part. Thanks, Arne, for your good rethinking and critics. I hope to be able to convince you, that the simulation results are not faulty – we just do need to know how to interpret them. Let’s see:

    L/D maximum

    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. – Arne K.

    Now, you compared the L/D ratio of Marchaj at 15°, which is 6.4, with my L/Dmax, which is about +-20 at 4-5°. That is comparing apples with bananas. Either you compare L/Dmax of Marchaj with L/Dmax of me, or you compare L/D at 15° from Marchaj and L/D at 15° from me.

    L/Dmax is achieved where the tangent to the Cl/Cd curve has the highest slope (which is at 5° in Marchaj’s graph). I added that line to the drawing, have a look:

    it gives L/Dmax = 12, which is double of what you wrote. It is of course still lower than my value, but then, Marchaj’s data comes from measurements with a real model sail, including 3D effects like tip vortex, while mine is a 2D-section which cannot include 3D effects like the tip vortex (resistance from the tip vortex is substantially high). Thus, the L/D ratios of Marchaj can only be lower than mine.

    In Addition: the Cl/Cd line has quite a high slope at 5°. Marchaj did only put one measurement point at 5°, the next one at 10°. I do not know of the accuracy of his setup, but chances are that this one measurement point can have a little measurement deviation, which would have a huge effect on L/Dmax due to the high slope of the curve.

    Better AoA for L/D reference

    However, I definitely see your point about me using L/Dmax at 4-5° is faulty (more to that later). If we compare Marchaj and me at, let’ say, 10°, it looks like this: L/D_Marchaj = 9 and L/D_Paul = 13. Pretty close together, given that my 2D simulation does not include induced drag/tip vortex and thus naturally has to be higher than Marchaj’s measurements!

    Choice of AoA range

    You are absolutely right in claiming that my fixed geometry is not real at low AoAs. In reality, at AoAs too small, the sail would simply collapse at the luff – which my fixed geometry can’t do.

    A simple solution to this is to have a look at the airflow around the luff. If the flow on the pressure side of the luff is detaching, then this simulation run is not modelling reality and has to be discarded. Only the suction side of the luff is allowed to show detached flow. Have a look at the following series:

    (this image is also attached, hopefully in better quality.)

    It can be observed, that for very low AoA’s < 4° the flow is detaching on the pressure side. This would not happen in a real sail, instead the sail would collapse. Conclusion: Those runs have to be discarded.

    Next observation: the flow starts detaching at the suction side of the luff at AoA’s > 7°. That would happen in a real sail. You could see it by the lee telltale (at the luff part of the sail) starting to wobble. à if racing, you try to keep both telltales, windward and leeward, streaming in parallel. The simulation shows, this is the case for 4° < AoA < 7°. What a narrow interval! However, it matches my practical experience: I have to be quite focused at the helm if I want to keep windward as well as leeward telltale streaming in parallel. Now we are talking about alpha-tolerance.

    4° - 7° AoA, keep that in mind when looking again at Marchaj’s or my Cl/Cd curves! It is right on that steep slope. That’s where we want to operate our sail.

    The max peak in the Cl/Cd graph of Marchaj (my graph shows the same) is not the most efficient one! That peak, at 15° for the sail “2” in Marchaj’s fig. 83, is the point where the detached air bubble on the suction side reaches the leech. From about 7° to 15°, the detached bubble gets longer and longer, starting from the luff, before it finally reaches the leech. At this very point, when the detached bubble reaches the leech, the leech telltales start to wobble – a situation that Arne does not get tired of to sensitise people, and he is right doing so, in my opinion. However, it can now be seen very clearly, that just at the point where the leech telltales start to wobble, it is already “too late”. The sail is more efficient when the AoA is lower than at that point.

    -->  Again: my simulation is of a section/profile! The Cl/Cd-curve gets distorted when 3D effects get into play. Still, comparing a 2D profile with another 2D profile gives absolutely valid, comparational results.

    Summed up: From all my simulation runs between 0 to 20° (for each tack), the ones below 4° and the ones above the first max peak (about 10-15°) can be discarded. A good learning for the next series of simulations!

    No patience with a stalling profile

    I should mention, that with a fully stalled flow around the profile (any profile), the simulation isn't static anymore but gets into a periodic state: A Karman Vortex Trail builts up behind the profile, and does not let the simulation converge. Convergence is essential for numerical simulation (there are many books written about this topic). However, as my computational power is quite limited, I stopped the simulation runs at 15 sec simulated time. That is enough time to converge for partly stalled flow, but as soon as the profile is fully stalling, this strategy of me aborting the run “before it is finished” gives faulty results. But, good thing is, it doesn’t matter. We are not interested in fully stalled situations. I just need to find the max peak in Cl/Cd, everything beyond is of no interest (for now). Having explained that, you can observe that effect when looking at the diagram below: after the max peak, moving towards higher Cd, the lines go down and then up again. This second rise is faulty. It is due to me stopping the simulation before it was finished (if it ever would finish).

     

    Cl/Cd plot with equally scaled axes

    What I would like to see is a (computer-generated) Lift/Drag diagram with the coefficient of Lift at the Y-axis and the Coefficient of Drag along the X-axes, both to the same scale. – Arne K.

    I already posted such a diagram, but without the equal scale on both axes. No problem to do, here you are! Same simulations, same data, different diagram layout:

     How good that it’s nasty outside, otherwise the temptation to go sailing today, instead of sitting at the laptop, would have triumphed :-)

     

    Hope to have you convinced and on board, Arne!

    Cheers,

    Paul
    3 files
    Last modified: 30 Oct 2024 06:11 | Anonymous member
  • 29 Oct 2024 13:22
    Reply # 13424703 on 13423701

    Hi,

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

    Yes, Eric, that is exactly what I intended to show. Before looking at improvements/innovations, the simulation needs to be capable to reproduce trivial and obvious results first, to prove it is reliable. Which it does, read on!

    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).  – Eric A.

    Eric, I do not quite understand what you mean. Are you referring to True Wind Angle (TWA) and Apparent Wind Angle (AWA)? The simulation I ran have an Angle of Attack (AoA) towards an incoming flow. Transferred to a boats coordinate system, that AoA from the simulation translates to the angle of the sail towards the AWA – not the TWA.

    Also, Eric, I had that photogrammetry idea, too. As far as I know, a modern Iphone Pro has LIDAR-sensors built in, which are capable to do a 3D scan. As you proposed, doing this in a berth, with the sails set and sheeted in properly (leech telltales flying, luff not deflating), should give some pretty accurate results. A 3D-model of at least one panel would be sufficient, it could be sliced into 2D and here we go. However, I do not own such a device, but am having a sharp lookout within my friends in Kiel.

    Doing multi-physic coupling, that is combining flow simulation with textile strength simulation, as you thought of, Eric, is in another league. Possible, but really a tremendous amount of work and quite costly regarding computational power. If time was unlimited, I would go for it, but unfortunately it is budgeted as anything else. And one would need to ask oneself: what is the advantage? If we get the actual 3D shape of an inflated panel from i.e. photogrammetry, what better can it get?

    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? – Kurt R.

    Interesting approach, Kurt. However, CFD uses multi-core distribution in a way, that the “virtual wind tunnel” is divided into smaller segments, one segment for each core. On my 6-core laptop, the virtual wind tunnel is divided into 6 segments. Now at the end of each timestep, those segments need to communicate with each other so that the flow is able to advance downstream, so to say in a simple way. There are about 1.5 million timesteps necessary for each run. If you imagine different computers all over the world to communicate via internet such oftenly, that would be quite slow. I think in that case, unfortunately, my one-core smartphone would even be faster in the end. There are applications that are prone to do what you suggested, like i.e. SETI, but CFD calculations are working differently. Still, interesting approach. CFD is highly interesting for the industry, so there is a lot of improvement going on. Who knows what's next... I know of some very clever people working right now on using GPU-cores instead of CPU-cores. Looks promising, but still no miracle should be expected…

    … the purpose of these basic models is to find comparative data, on just an arbitrary section of a junk sail panel. – Graeme K.

    The numbers themselves may not be absolute, but the comparison is still well made, and we have learned something – or at least verified something. – Graeme K.

    Graeme, you are spot on with your explanation! I am not aiming at actually calculating absolute values of windward performance of an actual yacht. The approach I take here is not intended to do so, and definitely not able to do so. However, instead of absolute results, it gives relative results: How is the port tack compared to the starboard tack. What effect has 10% mast balance to 25% mast balance? How are split junk profiles working compared to flat cut profiles compared to cambered profiles compared to flat cut profiles (watch out, I intendedly wrote “profiles” - not rigs or sails!). Those results can be transferred to a real sail/rig, but in reality are probably blurred by other things like 3D-effects (yard angle, batten angle, sail plan form), environmental influence (seastate, gusts) and yacht related issues (hull shape, weight, mast weight, mast stiffness, …). However, the effects from the 2D view are still there. They are simply blurred. (sorry for my unaccurate English!)

     

    I hope Paul will put things right, and not be afraid of telling me. – Arne K.

    Right now there is too big difference between Machaj’s and Paul’s results – Arne K.

    Actually, Arne, the difference is marginal. I’m not afraid of proving it to you – but in a separate post ;-)

    Thanks for all your interest and input so far! It is fun and valuable to not do this alone in my lonely corner, but discuss it with you and thus improve it.

    Cheers,

    Paul

    Last modified: 29 Oct 2024 13:22 | Anonymous member
  • 28 Oct 2024 03:12
    Reply # 13424123 on 13423701
    Anonymous member (Administrator)

    In the last line of that posting I wrote:

    The next diagram shows a more realistic sail with AR =3.

    Thanks Arne

    I misunderstood that sentence.


    Last modified: 28 Oct 2024 03:20 | Anonymous member (Administrator)
  • 27 Oct 2024 23:25
    Reply # 13424072 on 13424056
    Anonymous member (Administrator)
    Graeme wrote:

    Hi Arne, I am not sure if we are maybe talking past each other here.

    What I meant was - I was referring strictly to “comparative” data. For example, for a given AoA the model can make a comparison between the numbers arising from port tack and starboard tack (the effect of the mast). The numbers themselves may not be absolute, but the comparison is still well made, and we have learned something – or at least verified something.

    That was all I meant, but it is only a small point, I suppose.

    Having made that one small point, and assuming the AoA figures are not so far out that the conclusion would be reversed, I don't have enough knowledge in theory or practice to question what you are suggesting.



    (By the way, just to correct the written record, should you not edit your previous post referring to Marchaj's AR=3 ? Or, am I missing something? I know it's not an important point).

    In the last line of that posting I wrote:

    The next diagram shows a more realistic sail with AR =3.

    What I meant with that was that the sails on Fig 99 show worse results (L/D) than the reference sail with AR = 5 (fig. 83)-

    Maybe I have missed something of importance. I bet I will then be told so rather soon.
    Maybe my writing here looks a bit blunt, but I can assure you, I am just technically focused, and then finer diplomacy must give way.

    I hope Paul will put things right, and not be afraid of telling me.

    Arne



    Last modified: 28 Oct 2024 07:18 | Anonymous member (Administrator)
  • 27 Oct 2024 22:37
    Reply # 13424056 on 13423701
    Anonymous member (Administrator)

    Hi Arne, I am not sure if we are maybe talking past each other here.

    What I meant was - I was referring strictly to “comparative” data. For example, for a given AoA the model can make a comparison between the numbers arising from port tack and starboard tack (the effect of the mast). The numbers themselves may not be absolute, but the comparison is still well made, and we have learned something – or at least verified something.

    That was all I meant, but it is only a small point, I suppose.

    Having made that one small point, and assuming the AoA figures are not so far out that the conclusion would be reversed, I don't have enough knowledge in theory or practice to question what you are suggesting.



    (By the way, just to correct the written record, should you not edit your previous post referring to Marchaj's AR=3 ? Or, am I missing something? I know it's not an important point).

    Last modified: 27 Oct 2024 22:53 | Anonymous member (Administrator)
       " ...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