Understanding Wing Technology

It depends on the transition mechanism. At low to medium chord Reynolds number, on the order of 10^5 to 10^6, transition typically occurs via laminar separation and turbulent reattachment, forming a laminar separation bubble. An adverse pressure gradient is required to cause the laminar separation. Whether or not the flow reattaches depends on how adverse the pressure gradient is.

When the fow separates, the pressure distribution becomes nearly flat. Once the flow transitions to tubulent, the pressure increases at nearly the maximum that can be sustained by a turbulent boundary layer, assuming a Stratford-like recovery from the transition point. If the inviscid pressure distribution has a slope that is less than the recovery profile, the two pressure distributions intersect and turbulent reattachment occurs at about that point. The flat pressure at separation combined with the pressure recovery at transition forms the characteristic shelf in the overall pressure distribution that marks the presence of a laminar separation bubble.

If the inviscid pressure distribution has a steeper slope, the two pressure distributions do not intersect and you get massive stall due to laminar separation. This is what gives the NACA 6-series their nasty leading edge stall at low Reynolds numbers. The NACA 6-series carries the rooftop right to the leading edge, which causes it to form a short, sharp pressure peak when operated above the design lift, and the steep backside of that peak can lead to laminar separation without reattachment.

If the inviscid and turbulent recovery pressure distributions are nearly parallel, they intersect well back on the section and you get the long separation bubble that causes an increase in profile drag at moderate angles of attack. With a modest increase in angle of attack, the curves may not intersect and you get bursting of the long separation bubble. The cure for this is to use a convex shape to the pressure distribution, at least in the transition area. The convex inviscid pressure distribution intersects the concave turbulent pressure recovery profile very quickly, resulting in a short laminar separation bubble. This has a much lower drag penalty compared to the long separtion bubble, and is quite robust. The convex pressure distribution becomes gradually more adverse as angle of attack increases, resulting in smooth movement of the transtion point with angle of attack.

At higher Reynolds numbers, transition can occur via Tollmein-Schlicting waves being amplified in the boundary layer until they become nonlinear and kink up into the structures that are embedded in the turbulent boundary layer. A favorable pressure gradient has a dampening effect on the waves, and an adverse pressure gradient is destabilizing. The amplification occurs faster at high Reynolds numbers and slower a low Reynolds numbers. Below a Reynolds number of 250,000 it is not possible get transition this way, even when using the maximum adverse pressure gradient that the laminar boundary layer can stand without separation, all the way to the trailing edge - the amplification occurs too slowly for the disturbance to reach critical size before the boundary layer runs out of chord. A boundary layer velocity profie that has an inflection point in it is particuarly destabilizing. This is one reason why transition occurs fairly quickly once laminar separtion is experienced - the separated velocity profile has an inflection point between the outer flow and the recirculating flow next to the wall.

The NACA 6-series sections used a flat rooftop to delay this kind of transition, which is how they got the moniker of laminar flow sections. As the Reynolds number increases, it takes a stronger favorable pressure gradient to suppress the instabilities, and transition occurs earlier for the same pressure distribution.

The e^n method of predicting transition is based on Tollmein-Schlicting instabilities. The idea is that transition occurs when the disturbances get amplified to a critical size. If the ambient air is very smooth, then the disturbances have to be amplified by a large factor before they reach the critical size. If the ambient air has large disturbances in it already, then only a little amplification is needed to reach the same critical size. The critical amplification factor, e^Ncrit, is how XFOIL models the ambient environment. The default value, Ncrit = 9, is for very smooth air because the disturbances have to be amplified by over 8,000 before they become critical. Ncrit = 1 means the disturbances need only be amplified by less than a factor of 3 to cause transition.

It is common to insert a transition segment between the rooftop and the pressure recovery segments of a design pressure distribution. The transition segment can take the form of a ramp or a convex rounding of the corner between rooftop and pressure recovery. The purpose of the transition segment is to have a modest adverse pressure gradient that causes laminar separation or amplification of the Tollmein-Schlichting waves, so as to result in transition before beginning the pressure recovery. Turbulent flow is deisrable for the start of the pressure recovery to avoid forming a long laminar separation bubble there. Especially if the recovery is fairly aggressive so as to allow a longer laminar rooftop.

A short convex segment is also used at the leading edge ahead of the rooftop to produce a lower, more rounded leading ege pressure peak at high angles of attack.

What you will see in the section data for an airfoil that is designed this way is transition occurs fairly far back, but moves slowly forward along the transtion segment as angle of attack increases. Then it will move forward more rapidly to near the leading edge as the rooftop becomes adverse and a leading edge pressure peak forms. Then transition remains close to the leading edge until stall.

Smith was concerned with high lift. At high lift, transition is typically at the end of the rooftop, or has already moved to the leading edge.

The rooftop is designed to avoid that. It can be flat or have a favorable gradient, depending on the Reynolds number. The rapid acceleration at the leading edge has a strong dampening effect, and then the constant pressure rooftop results in a boundary layer velocity profile that looks like the profile on a flat plate.

No, there is only one transition point - once the flow is turbulent, Humpty-Dumpty is not going to be put back together again. Smith was using the Michel criteria to determine when transition occurred.

The pressure recovery is due to the shape of the section. It isn't triggered, but rather a constraint on the design pressure distribution. The pressure distribution can't have any arbitrary shape. It has to result in a closed body that does not intersect itself. At the trailing edge, the pressure has to get back to near ambient pressure. So a pressure recovery region is necessary to get from the low pressure of the rooftop to the higher pressure at the trailing edge.

In addition to the leading edge, rooftop, transition, and pressure recovery segments in the design pressure distribution, Wortmann used a short steep adverse segment called the closure contribution. It completes the pressure distribution for aft loaded sections, because they have a pressure recovery that doesn't come all the way down to ambient pressure.

The shape of the surface mirrors the shape of the pressure distribution to some degree. When you use a concave pressure recovery it will typically result in a concave shape to the surface, resulting in an inflection point.

Such designs can result in a long separation bubble, for the reason I describe above.

No. Lift is the difference between the pressures on each side. There's not a direct relationship between the dumping velocity and lift. When you raise the dumping velocity of the element ahead of a slot, it tends to be done by increasing the aft loading on the forward element because the velocity will still be low on the pressure side ahead of the slot. But if both sides come to the dumping velocity with the same slope, there's no change in lift with the dumping velocity.

 

Thank you very much Mr Speer.

Just to say I have to chew for a few weeks or months I don't no yet, that is why I prefer to post now a little message instead of being silent for weeks.

My former attitude was not very "forum friendly" Sorry Dough, but Fluid Dynamics is a very serious matter, and I am more a benchwarmer than a star.

On the other side I have to understand the structural engineering issues, Euler's formula and other "modulus" which are very new for me.

But without appropriate spars no wingsail.

In fact my project is to achieve a "generic wing structure" for an A-Cat, I mean spars like a C-Cat with twisting first element.

Then it will be possible to "wrap" around it any kind of wing concept either a classic one with 2 symetric elements, either a single one which can morph.

But in the meantime I think I have to make serious "homework" to improve my understanding of fluid dynamics.

Thank you again Mr Speer for taking time to write this long post, I greatly appreciate.

Hope many other wing mates will take advantage of your explanations.

Best regards to all

Erwan

 

Hello Erwan,
Do you know of any wings that twist using internal control vs external restraint and a "fixed" spar? I have been contemplating this problem for some time. I'm a structures person so am grappling with the structures and control side of the wing problem. A C-Class twists about 15degs so an A-Class may need less. Cheers Peter S

 

Hi Peter,

Not sure I understand your question in depth, but about twist, find below what I know so far.

First remark: First element twist (Cogito-style) does not required external control, but the one at the mast foot, 20 cm above the trampoline, with a very very marginal drag, I guess.

The slotted flap requires external controls.

1- Don't know for the twist angle for C-Cat. Longer wing, more twist than an A-Cat, it makes sense.
2- for A-Cat, my first assumption was 17°, not considering the potential consequence when sailing wild thing on the wire, it is new.
That is why 22° max twist could be the required. (Just look at the Australian National Video)

These assumptions were for a single sail.

With a 2 elements wing, you can achieve twist by an appropriate evolution of the 1 element chord/ flap chord ratio combined with the flap twist.

But this second point is Mr Steve Clark intelectual property, just read his comments on his open wing.

It is probably the smarter solution to achieve twist, a positive side effect is the relative simplicity for the 1 st element structural issue.

I think that this twist concept, provide a kind of twist "efficiency benchmark".

The arbitrage between the Open Wing "Synthetic Twist Concept" as mentioned above and the Cogito-style twist system is like an economic choice with regards to the "Law of Decreasing Returns". or "Law of Marginal returns" I don't know exactly the correct wording in English, but I am sure you know it.

In the wing-technology universe this law could be translated the following way

"The marginal efficiency of complexity".

In both cases the flap contributes to the twist and in both cases you need "control arms" for the flap, and I am afraid that putting the control lines inside, will iincrreased signiificantly the loads on the ribs, requiring heavier ribs and so on..and big loads are probably not tack or gybe-friendly.

The drag on the command struts is a bit the "last stage refinment".

Your twist system must consider the gybe in the breeze, with some "g" accelerations on the flap, so a tight (no elasticity) limit at 15° might create some unexpected leakage or failures in the rig during hard gybes, especially if combined wiith internal controls wihich have to stand high loads.

Thank you for your kind consideration in asking me advises. But do not forget, I am in the benchwarmer category !!

Cheers Mate

EK

 

Hello Erwin,
Thanks for the reply. I shall explain a bit better. The "law" you refer to is the "Law of dimenishing returns". Leading edge Panel 1 twist is currently acheived by using an internal spar which is "fixed" at the hounds by the rigging. By controlling the angle of Panel 1 at its bottom the LE twist is controlled. Trailing edge twist is controlled by multiple panels. By internal twist control I meant are you aware of any other form of leading edge control other than the "fixed" spar method? I say this because I'm interested in a free standing rig design and in this case the internal fixed spar can't work. So an "internal" twist control would need to be used vs an external stay/shroud system as is now used. I am also not clear as to how the Panel 1 twist control works when tacked/gybed is it automatic or needs to be corrected each time? I have not found a clear discription of this mechanism yet. Plus does anyone out there know when the next C-class worlds are on and where? Regards Peter S

 

==========
Peter, as far as I know the next Cup is 2013 in England. Here is the Little America's Cup 2010 thread: http://www.boatdesign.net/forums/multihulls/little-americas-cup-2010-c-class-real-one-32301-9.html

Steve Clark is a member here and you could PM or e-mail him with your twist questions and about the next LAC.

 

With regard to twist, there are two different reasons for twisting. The first is to control the spanwise lift distribution to minimize the induced drag for the available heeling moment. This can be done by twisting the flap or by twisting the whole wing. In either case, you can think of this as controlling the angle of attack of the flap. The main element may carry the majority of the load, but the load is largely determined by the flap. Twisting just a 50% chord flap vs twisting the whole wing gives 80% of the effect on lift.

Given a desired level lift at a given station, there is a combination of angle of attack and flap angle that produces the minimum profile drag for that lift. Minimizing the profile drag requires the ability to twist the main element.

Using the spar as a torque tube to twist the shell of the wing will result in shear stress in the shell. If you had a different means of applying the same shear stress, then it would result in the same twist. The torque tube approach has the advantage that the spar can take the bending and compression loads largely independent of the shell. If you go for the integrated approach, your control system may have to react the shear loads due to bending as well as the shear associated with twist.

 

Thanks Tom,
I understand the aero reasons for twist. I'm addressing the mechanical approachs to controlling it. If the wing is twistable and its full of telltales then you can change the twist until all the telltales are streaming rearward (same as on softsail) In this condition the induced drag (profile drag) is at minimum. (I'm correct?)

The torque tube approach works, but I have a thing for free standing masts so why not a free standing wing? Planes gave up rigging a long time ago. Would allow wing full rotation for mooring and aero unloading in heavy wind conditons. Answer trouble to get twist. But take our forearm it has a radius and an ulner bone that allows twist nicely. So I'm working along these lines. Yes its true that the shear stress in the skin has to be addressed but if the skin was like a hang glider construction (or the Omer Wing sail)where the skins just dealt with the local aero loads and the internal skeleton dealt with the global structural loads then alls good! Peter S


Note
If boats continue to use code zeros and gennakers and a wing even if the wing is free standing they will have to use running backstays. I've modelled the new AC72 rig along these lines and with the forsails it needs extra support to get reasonable stay tension for the sails. Classes that do not use foresails could go free standing. There is a freestanding Moth mast in Perth, A-Class and C-Class could go free standing . Rigging is quite a drag at speed!

 

No. Induced drag and profile drag are two different sources of drag and behave very differently. Induced drag decreases with speed squared when the lift is held constant, while profile drag increases with speed squared. Induced drag is inversely proportional to the span squared, while profile drag is proportional to the planform area. Induced drag comes from producing lift over, and leaving a wake with, a finite span. Profile drag comes from viscous losses in the boundary layer and pressure changes due to the effective change in shape due to the displacement of the outer flow by the boundary layer. Telltales will tell you if there is separation of the boundary layer, which can avoid the worst of the profile drag, but they don't tell you anything about the induced drag.

With regard to effecting twist control, there's no reason the torque tube approach can't be used for a freestanding mast as well as a stayed mast. The spar needs to be mounted with a bearing, just like a conventional cantilevered mast. The spar and shell are connected at the hounds or at the head, as with the C-class.

Aerodynamic load on the head will make the head twist off and rotate the spar. The sheeting load is applied to the foot of the shell. A control arm attached to the spar would have a limiter line going to the foot of the shell. When this line is tight, you get no twist - the load path for the sheet load goes from the foot of the shell through the limiter to the spar, and is transmitted via torsion in the spar to the head. This approach is simple and self-tacking. The main disadvantage is it depends on aerodynamic loads to effect the twist, and in light winds there may not be enough air load to overcome friction between the shell and spar. However, with a cantilevered mast, you don't have torques from the stays to overcome.

If the spar can be rotated under positive control, and the wing sheeted in either direction, then you have complete control over the twist. A cantilevered mast actually makes twist control easier, although there is a lot more friction in the bearings than for a stayed mast sitting on a ball.

 

Wing Sail Twist Control

Hi Tom,
Thanks for clarifing the drag bit. I see where you are going with the limiting line, I have gone down that thought path but as you say a direct control system is the best way. Using proper bearings rather than a ball & cup probably has less friction. Friction coefficients are about 0.1-0.05 teflon to steel and roller bearings are <0.001. Being mechanical engineer spending half my time designing machines and half my time designing yacht structures a wing mast seems to be right for me! Peter S

 

Twist Control on Wing Mast

Hi Tom,
Heres a dwg of what you have described & I have thought about. Is this whats on most C's now? Really its functionally the same as what we do with soft sails. The Wing Panel has to be torsionally soft enough to cope with the twist and maintain its aero shape (say +/-20degs? suppose it depends on mast height) and provide structure to attach rear panels to... the next step! But can't we do the same thing with the rear "panel/s" so there is no segmentation on the rear panel? I think this has been done on one C-Class? So we need a better system than the limiting line so we have positive control. I'll work on that.

Peter S
The future can't be predicted, but futures can be invented. D Gabor.

 

Hi Peter,

Just like Mr Speer's remarks, I think a free standing mast does not rule out the torque tube system.

But with regards to twist and leading edge related stress, in my former "research", I identify the Element1 trailing edge (structural) as the driver of twist. (I dissociate the aerodynamic TE and structural TE).

The wing main control line is on the flap, creating a leverage with the flap hinge which is on the Elt1 structural TE.

The aerodynamic forces will push the whole wing until lines limit it.

If you take a rear wiew of the boat, and you want twist.
Your torque tube is straight, just behind it the Elemnt 1 TE is sagging from the bottom, creating a curve to leeward.

If this curve could meet the wing gradient it will be great.

This curve is pushing on the twist control arm at the hound.

So an appropriate stiffness of the structural TE will drive your twist, but you must add the structural leading edge of the flap.

Not sure, but I guess that the best approach could be to consider both the Element 1 structural TE and flap structural LE as an homogenous system of 2parallel spars tied to each other by command struts, and flap control arms.

For an A-Cat the max load on the rig is around 700N, the first 20% of the wing section (leading edge) will transfer the aerodynamic load directly to the torque tube. Even if you do not like integral calculation, looking at the pressure distribution, you can see that the first 20 % of a wing section carries more than 20% of the aerodynamic loads. So the remaining aerodynamic load on the aft part of the wing sail is not hudge.

Using wing section Cmo and highest/lowest Cl assumptions, it could be easy to appraise the required flexion stiffness of the combo: (Elt1- TE + Flap LE)

When providing info about his new Aethon wing, Mr Steve Clark said that the leading edge has a zero torsionnal stiffness.

I guess that when molding the leading edge, you need some special care to achieve the twist while maintaining the geometry of the section.

It could be very sharp V fence on the leading edge at each rib level and parallel to the ribs of course.

These sharp V could be 1 mm large and 5mm depth at the end of the leading edge nose.

The depth of these V could be increasing from 0 increasing to 5 mm along the leading edge nose until the maximum wing section thickness.

A good brush with epoxy and laminate, can provide very useful experience in your garage. Starting with basic parabola could be insightfull enough I guess.

I did not try it yet, so it is far from sure it could be a solution.

Not sure it can help, but so far it is the only contribution I can provide.


Cheers

Erwan

 

Hi Erwan,
I'll be building an FEA model of the wing so will figure out the stiffness/strength required of the various elements/parts. If I use a suitable section such as has been mentioned on various discussions (I'll look this up) if I can get a suitable pressure plot of the wing surface then the model will be very accurate. If I can't get an accurate pressure plot then I shall estimate the distribution (but I'm not an aero guy). For structural purposes I don't need a completely accurate pressure plot. I have been playing with a wingmast+stiff rear sail combination. But it does not have a slot and there seems to be emphasis placed on the slot being the big aero advantage more than the camber. The avi shows a wing section with stiff rear "sail" that cambers nicely. I have been grappling with TE of the wing and the stiffness of the "sail" so it gives a smooth shape when deflected. Is this as good as a two element solution? or should I forget this idea and move to a two element solution. Cheers Peter S

 

Hi Peter,

Nice job, I am a sucker with this kind of software, I need time to answer in good (or not too bad)English.

At first glance our philosophy is quite similar.

For the load, according to your project, I suggest you to use XFOIL, and find the pressure plot for the S 1223 wing section at max AoA. 5lift coef is around 2,1 for a single sail. Pressure distribution for a single sail at hight lift are not very different from a 2 elet wing. With the slotted wing you have a thin "pie slice" in the pressure distribution for the slot and a peak pressure at the flap LE as mentionnned above by Mr Speer. I guess that on average it si not so different for global load, whatever is the aerodynamic solution.

in any case F= 0,5*p*V^2*Area*Cl. and max F is limited by the righting moment. So 700N for 150 sqr feet with 60% of the load on the first 40% chord is a bit rude approximation, but I am not sure a 2 decimales approach will improve the solution.
I guess it will provide you with good idea of the aero loads along the chord.

For which kind of boat are you working on ?

Cheers Mate

Erwan

 

Hi Erwan,
Your English is fine Erwan. I've played around with softsail design but my usual trouble is getting an accurate pressure plot. I can do an aero-elastic analysis if the pressure plot is known. I use the approximation as you suggest with 60% of load at front and rest at rear. The wing sail will be easier as I don't have to wrestle with getting the flying shape of the sail going using membrane analysis. Although I've become good at sorting this. I even have figured out how to put in battens without having to use full blown non linear contact approaches which are a pain to resolve. But I'm working towards an A-Class wing or a new class so I don't have restrictions. I would design a catamaran that is the smallest that a person could sail fast using a wing (perhaps 14ft), a bit smaller than an A then build two so my wife and I can race each other!. In this way it will be the cheapest to develop. I have also been doing a study on the structures of the new AC72 (see attached movies for hull structures). Havn't got to the Wing yet. Attached is a hogging load case for the AC72 and a soft sail that is fluttering and has not yet settled down, the foot is moving the wrong way. Would need to run the solver for more time to allow it to settle. Cheers Peter S