The ultimate validation?

Yes, I agree. Looks amazing from the computational point of view, too.
Mikko, I would like to know what hardware do you use for your CFD job, and how much time did it take to compute the airflow for that video?

 

Congratulations Mikko! and thanks for sharing what seems to me a very "top secret knowledge" (your competitors will be happy too, I'm afraid!).
Two medals in a row is not just coincidence: sail design in a one-design class has a huge percent on success.
Question: main over jib suffers big perturbation while pitching, is it just the bigger acceleration of that part, or the jib presence has a benefical influence in the lower part of the main (or both)?.

 

The solver is not RANSE at all, but LES with wall functions, and inherently transient.

We have a 8 core workstation with 32 GB memory - 8 cores is double as fast as 4 cores - something like here, a 6 sec simulation runs in about 72 hours. This is not HiRes, rather medium resolution. A HiRes run will take about a week, it will give mostly more detail in the wake but the forces won't change much.

 

I don't think we are revealing huge secrets here, although I doubt that our competitors have ever seen something like this. We are not giving the sail shapes, or wind gradient & shear etc., so they can at best compare with their own results. But I think everything we learn should be shared, it will come back to us some day. "If I were to play violin alone in the forest, would I make a sound?". Yes, the angular velocity is higher in the top of the main and the local flow goes more forward there - and I think the jib does appear to "protect" the main a little, since in the Finn (no jib) the disturbance goes all the way down to the tack. I can post a Finn simulation later if I find the time.

 

I'm sorry Tom, I did not mean to disturb your sleep ;-). But yes, I was at awe when I saw a LIC-simulation at the first time. Like a sailing friend said you can almost feel the wind on the sails.

Here's a couple more...

 

I found some papers online that might be of interest to those using CFD to
model sails. There's quite a bit of experimental data and some comparisons with
a variety of computational techniques.

The papers are at:
http://www.ignazioviola.com/ignazio_maria_viola/download.html

The author, Ignazio Viola, is a lecturer at Newcastle University. See:
http://www.ignazioviola.com/ignazio_maria_viola/home.html

 

I have seen these, of course, and I've met & spoken with Ignazio on many occasions. Ignazio shows rather an incredible accuracy between full scale/wind tunnel/CFD comparisons, but I'm sure he is showing only half the truth, chosing to show the results that fit... For instance, in his latest paper, in Ocean Engineering 38 (2011), he gets a 0,5% difference only between CFD & windtunnel tests (on one point/awa). Yet in the CFD a uniform inflow is used (no wind gradient or twist), while in the wind tunnel they would both be present, not to mention the real world, full scale measurements.

In general, I think pressure measurements on sails are problematic. Even in an open windtunnel, like the Auckland twisted flow tunnel, the reference pressure (and ref. wind speed or ref. AWS, if you like) is difficult to choose, as it varies form upstream to the location of the model, and also varies with height around the model. So how do you define Cp, to draw your Cp-curves? Do you have a different ref. flow speed at every height? The problem is similar to the one referred by Tom Speer in the asymmetric polars-thread. This makes a big difference, the difference in predicted forces between uniform flow and twisted flow can be 25%. The Cp-surface plot so descriptive and easy to read for airplanes gets much less meaningful when used for sails.

In another paper, Ignazio provides pressure data for an AC33 upwind sails measured in the wind tunnel on rigid fiberglass sandwich sails. He was kind enough to send me the IGS-files for the 3D-models of the sails, but I was not able to match our RANS-code results in a good manner to the measured pressure values.

For the fullscale pressure measurements, a further complication is it is impossible to define the 'reference' dynamic pressure. This should be somewhere in the "free stream" far ahead of the boat, but even if we could measure the dynamic pressure there it would be quite different that the sails are experiencing at the time. From the paper "The dynamic pressure qN was measured with Pitot- static tubes fixed onto a pole attached to the stern of the yacht. Subsequent analysis showed that the dynamic pressure mea- sured at this location was about 20% higher than the dynamic pressure at the reference height." So they tuned the dyn pressure to better fit the wind tunnle results. And the static ref. pressure is no simpler, measured here inside the cabin, which is different than the far-field.

All this has troubled me for quite some time, so I have given up the idea of measuring absolute pressures. Pressure difference between the two sides of the sail is independent of all these problems and does reflect the force on the sail. But you would need a whole lot of pressure taps on the sails to integrate well, and you also need to know the shape of the sail to get a direction for the pressure vector... mind you, this is being done, too, by David le Pelley in Auckland: David Le Pelley, Dale Morris & Peter Richards
Aerodynamic force deduction on yacht sails using pressure and shape measurements in real time.

 

There are a number of people working on pressure instrumentation for sails. The people I've spoken to seem to have the idea that this will be useful in some way for trimming the sails in real time, but do not have a solid notion of how to use the data. It's sort of a Field of Dreams philosophy ("If we build it, they will come").

To me, pressure measurements are problematic for two reasons. First, outside of a narrow range of ideal angles of attack, there will be a narrow pressure peak at the leading edge that is difficult to resolve with practical instrumentation. But my main objection is this: Assume that the instrumentation is perfect - it covers the whole surface with high resolution at a high sample rate with high precision. You can integrate the pressures to get the spanwise and chordwise load distributions on the sail. What does the ideal pressure distribution or the ideal load distribution look like?

There are a limited number of aerodynamic objectives in trimming a sail:
- achieve maximum lift without stalling
- minimize induced drag through control of the spanwise load distribution
- avoid local areas of flow separation so as to
- shape the camber to minimize profile drag

Flow separation can be visualized using telltales. Differences in profile drag due to camber shape are minor. So the main contributions that pressure measurements can make to sail trim is to minimize the induced drag through the spanwise load distribution.

The problem here is you can't look at a spanwise load distribution and know if it is optimal or not - even when you add the load distributions from multiple sails to get the combined load distribution of the whole rig.

I think pressure instrumentation would be most valuable for validating CFD predictions. You have to know what you are looking for, and be interested in knowing if you are achieving it or not.

 

Quite agree with you Tom. Yet, if I had a display in front of me showing TWS, total drive and heeling moment in real time, I would be quite happy ;-).

For CFD validation, how would you measure the real world "inflow", even if you had the perfect pressure tapping on your sails? The apparent wind varying in both strenght & direction with height, needed as input for the CFD? And turbulence intensity?

 

Cool simulations! As an atmospheric science grad student with a fondness for boats and aero who does LES modeling of boundary layer cloud as a day job, I can't help but think that at some point we will see a nested grid for the boat inside an atmospheric LES that will give a nice boundary layer wind profile, along with the inherent variability of the boundary layer winds. One could take this yet farther by including the wave action, as in the simulations done by Sullivan et al. in the paper attached. Not sure how much it would improve the sail optimization, but I'm sure it would make for more pretty videos...