brooklin boatyard

ullman banner ad

seadeck banner2

pyi vendee ad

pyi-kiwi-12-16

hh 66 banner

fish side 5 17

melges m 14 banner ad

trogear ad

west system fp banner ad 1 15

diab big

gunboat-banner-ad-png

learning to fly

Part two of Oxnard Shores by Kevin Hall. Brought to you by Mauri Pro Sailing. Part one is here.

We’ll also get to the wing later. Let’s get our heads around the last axis of our six which might dominate a neural-net approach to improving our flying technique : heave. The vertical movement of the boat, the whole package. What happens when we lift it out of the water with a crane, or push it up with foils – that’s heave. If we can understand and marshal the setups and control inputs to smooth out the heave, we will really have a weapon. There are enough variables in the equation, as I hope is becoming apparent, that this is where our big data approach might start coming in. It may not always be clear which things manipulate heave behavior the most. Candidates are, in no particular order and a non-exhaustive list: rudder AOA, effective cant, daggerfoil rake, wing twist, daggerfoil rake response rate, center of gravity, wind puffiness &/or shear, helmsman, sea-state, to name a few. We could use our computers to look for the variables which have the strongest relationships with heave. We would be wise to check for plausibility with the sailors before announcing our discoveries. I imagine the sailors at Oracle Team USA were pretty open to input born of silicon and bits during this America’s Cup.

One of the main reasons computer models of foiling craft tend to fall (very) slightly short of full-scale real world behavior may be that the foils those computer models are flying through their virtual fluids are not the same shape as the foils which are flying through the actual air and water. They are certainly unlikely to be changing shape in the model in response to changes of boatspeed and applied/produced force like they do in the real world. At least today. This will come : I recently saw a 10-billion element model running two F1 cars in a virtual wind tunnel, wheels and all. The virtual tunnel has to be many car lengths long and run the turbulence in front of and way behind the cars to be accurate. It was impressive, but it will likely prove to have its quirks, as all models do.

As we apply force in the real world to a foil which is not infinitely strong (or in fact is likely designed specifically to deform a certain way with an expected force), the deformed foil has different characteristics – a different section and effective camber from the pressure on the trailing edge – to the original, stationary foil. These deformed foils’ forces at a given speed, will deform the foil differently to the model of forces of the non-deformed one. When we feed those new forces back to our structural model, it will deform more accurately in response to them, which will give us a new section shape to run, which we will feed back to our structural model…  If we have fibre optics in our foils we can measure the strain or compression where our sensors are located, but this still doesn’t give us the deformation of the foil. For that, we need a model of the foil’s behavior which would produce identical strains to those we measure while sailing. This model depends on where we assume the center of lift of the foil is, even after our thorough jig testing in the shed to significant load. Unfortunately this applied load is not applied as evenly in the jig as the water applies load, but it will have to do.

Let’s review. Our computer model of our foils on our simulated AC72 about to take off relies on a model of foil performance, which relies on a model of real-world foil shapes, which relies on a model of foil deformation derived from matching jig-induced (modeled) load, and measuring Fibre Optic stretch or compression of the foil itself during jig testing. There are a hundred places our models can go wrong enough to be useless, or, worse, misleading.

How the boat behaves in heave as it starts flying is determined by its existing inertia (are we coming out of the water quickly or gradually? is the bow pitching up and staying, or pitching up more and more?), and the vertical component of lift our foils are producing. Those components are sensitive to pitch, roll, and leeway, turn rate, and rate of change of those things. Working hard with the data to find things – anything! – that seems to work well and be repeatable, we could make some technique suggestions. We might learn that being slightly more patient to get flying, letting the boat accelerate a little more on its own time, can be better than slamming in daggerfoil rake and popping the bow, then stern up, only to come back down. We might find we seem to fly sooner when sailing upwind when the boat is relatively flat, and aim to keep the windward daggerfoil just out but the rudder horizontal just in. We may find that each time we tend toward one edge of the targets – be they foil position, wing setup, or BSP – something good (or bad) happens. We have probably already systematically tried effective cants, foil pitches, and rudder lifts to understand which combinations make the boat porpoise more and which make the ride smooth during our one and two boat testing leading up to the races. We may stumble on different hydraulic hoses hoping for self-regulation, some release under increasing pressure. We may put a delay in a certain input to try to take some guess work out, or work to eliminate delays everywhere in all the systems. We might try to make the daggerfoil move quickly at the ends of the stroke, and more gradually in the middle. We may realize there is a magic daggerfoil rake setting, and strive to make achieving that setting as easy and “hands off” for our sailors as possible. Or, we may just be hanging on for dear life the whole time. I’ve only watched an actually foiling AC72 on TV, and that’s what it looks like the guys are doing to me.

The wing setup might also affect how the boat behaves in leeway and hence subtly in heave. One example is what we call the span-wise lift distribution of the wing’s lift. Don’t be fooled, this is just a fancy way to basically say twist, which designers use for job security. If we make the wing really cambered in the bottom part, and reduce, eliminate, or invert the camber in the top part, we will produce a certain roll moment. Imagine pushing with your finger to tip it over on the middle of the wing of your 3D printed model of your virtual AC72, while one hull is kept from slipping sideways. Your finger is producing roll moment. A wing which is very deep in the bottom and flat or inverted in the top is pushing sideways hard at the bottom, and may even be pulling a little at the top. This produces a lot of side force that the hull and vertical parts of the foils have to resist. When the hull comes clear of the water it is only the foils opposing leeway. The good thing about a wing like that is it produces a lot of driving force for a given roll moment. However, if we are trying to keep the boat flat, to have the rudders produce more lift during takeoff, we have more righting moment than if we are heeled much. Maybe its worth trading some wing span-wise balance against this righting moment to reduce leeway. If we add a little camber to the middle of the wing and flatten the bottom of it just a little, we will have about the same driving force, a little more roll moment, a little more drag, but a lot less sideforce trying to cause leeway. This might be a worthwhile tradeoff for flying upwind.

Our computer models flying through the air have similar challenges to those flying through the water. Now lets add some left shear at the top of the wing which is variable in time and location on the course, and the fact that as you camber part of the wing it deforms more, which may or may not reduce the lift produced there. It would be great to know what the air is really doing over the wing in the real world. That would help us tune our computer models a lot. Fortunately, with some little holes and some plastic tubing and then some cool electronics, we can measure the local pressure at each hole. We might put them along the chord of the wing at a number of heights, say four or six heights along the wing from top to bottom, and at 1/16, 1/8, 1/4, and 7/16 along the chord of E1. I suppose there is a case for having them in E2 as well if someone will actually look at and eventually use the data. We could have a big argument about exactly where to put them, but for sure we will learn something from them as long as they are there or thereabouts in E1 to start with, and checked for blockage and false readings. For example, they tell us immediately when the wing is stalled at one height, or, worse, along the whole thing.

I once learned the hard way on a windy reach at the moth worlds that playing the mainsheet with an outhaul which is too loose has significant impact on ride height behavior if there is any windward heel. I imagine there is an ideal setup for an upwind AC72 in left shear that is hoping to foil most efficiently. We might be able to troll through our data to find times when the wing setup seemed the best compromise between what would be best for cat at 3.2 deg of heel with a hull producing sideforce, and what is ideal for a flatter cat with no hull producing sideforce. Or maybe we never got time to try any experimentation with that, because we were still at the broad stroke phase.

OK, so we have all this stuff and all this data, and we’ve finally learned to get down the run on the foils. As they go faster and faster, our rudder foils produce more lift and the stern comes up and stays up. As we slow down in a gybe, the lift from the daggerfoils reduces from speed but the foils double in area (though not necessarily lift since we should be able to put the new foil in at a rake we have chosen). Pretty soon we figure out all the subtleties of the gybes and our percentages go up. Especially when we have two boats out, we could experiment a little with foiling upwind but we’ve probably optimized our setup for downwind heave stability already, and we don’t yet handle more lift on the rudder very well. Adding rudder lift requires us to take too much lift off the daggerfoils to keep the boat near level pitch. Every time we try this setup, and then pitch a little nose down the effective pitch of the daggerfoil horizontal goes neutral and we fall out of the sky. No good.

Any aero drag will always be an enemy to foiling upwind, so we’re glad the aero guys tipped the balance of our total package so far to that corner of the box. Some of what we learned in 2010 probably helped here. We’ve ended up with a tantalizing endplate for the wing, small frontal area bows, and lots of care to aero drag around the boat. At 27, 28, 29, 30 knots BSP that whole drag increasing with the square of the change in apparent windspeed starts to sound like a really big deal! Maybe our front beam fairings even vary in AOA to account for the widely different flows between the center and the hull. Maybe we tried a tiny bit of lift on the rear fairing and rejected it for the induced drag penalty. We saved some weight in the self-tacking jib, so that could go back into the structure to make up for its aero-oriented topology. Turns out that wasn’t the primary difference in tacking performance after all! Put it on the long list of AC red herrings.

What else can we do if we are desperate to foil upwind? We could try to move the center of gravity (CG) aft. Wing rake does this, as does taking off the front of the pole we don’t really need. This will require more relative lift on the rudder for the downwind settings we are otherwise used to, and we may be able to get away with slightly less effective pitch on the daggerfoils before we fall out of the sky as well. Since we have invested so much time and effort in our models, we might be able to get a feel from them for how the boat would behave at this new CG as we nudged the rudder lift up. Note that it is very subtle. If we normally run 2 degrees of effective angle of attack (AOA = section shape + orientation in boat), a change of just half a degree is a 25% change). Maybe we toyed with foiling upwind in our two boat testing, and it showed some promise but we could never prove it was better. Or maybe we did lots of race testing with one boat foiling and the other not, but since we didn’t know how to tack, the foiling boat always lost.

None of this theory helps a lick until someone does the work to make the changes. Here’s to all the boat builders and wing builders, hydraulic and electronics guys without whom there would be no AC72s flying in the first place. Finally, a special shout out to the entire shore team of Oracle Team USA for your success in September 2013.

 

January 13th, 2014

ewol banner ad

front-banner

cst banner ad

hancock ad 6 16

hh 55 new

http://www.camet.com/

rapido banner

schwab tower ad 1 15

fareast-28-ad