Easter project 2009: Hydrofoil!

[lmfoils]

Originally Posted by Doug Lord

================
Two independent flaps for each foil,right? Very interesting video, thanks! Whats the next step?

Thanks! Yes two independent flaps per foil.

The next step is to recalculate the model, in order to improve it's accuracy, add a 9 degrees of freedom IMU and optimize the controller. We hope to be able to stabilize the system even when standing up on the surfboard.

[Doug Lord]

Originally Posted by lmfoils

Thanks! Yes two independent flaps per foil.

The next step is to recalculate the model, in order to improve it's accuracy, add a 9 degrees of freedom IMU and optimize the controller. We hope to be able to stabilize the system even when standing up on the surfboard.

-------------------
Will you offer the electronics for sale at some point?

[lmfoils]

Originally Posted by Doug Lord

-------------------
Will you offer the electronics for sale at some point?

The problem is that the dynamics of every system have to be modelled, so unless you plan to fly a surfboard exactly identical to ours the electronics with our code will be useless.

What do you have in mind?

[Doug Lord]

Nothing right now but down the line I think sailing foilers-particularly large ones will benefit from electronic augmentation of their stability. I've done a lot of experiments with rc foilers and one in particular I want to redo with electronic control of stability. I did a 36" model of a Moth with movable ballast and two foils. I want to redo do it as a scale model of a 60' Moth this time 5 or 6' LOA. But I want to be able to fly upwind at a constant 20 degree angle and tack with the ballast auomatically controlled. Not going to be ready for this for some time to come. I've talked to some electronics guys in the states that think they might have a solution so I was just wondering what you might be looking at down the line.

36" Moth-(manual altitude control and manual movable ballast-very, very difficult to sail)- and my RC F3(similar to a Rave with dual independent wands on the main foils-very easy to sail) :

[tspeer]

Originally Posted by lmfoils

Here is the video!

Notice the good roll stability given the noisy input of the wands.

Feedback can be a wonderful thing! When used appropriately.

I suspect the reason your surface piercing foils did not perform as well as the fully submerged foils is because the fully submerged foils were connected across their whole span, while the surface piercing foils had a large gap in the middle of their span. This causes a large increase in induced drag. It might be worth resurrecting the surface piercing foils and putting a piece across the middle to see the difference.

With regard to pitch control, I suggest you combine the height control and pitch control together again, but this time apply them in different frequency ranges. As you saw, pitch control was fighting the height control at low frequencies. Part of the problem is the steady-state pitch angle varies as a function of speed, and you don't know what value to command,and before takeoff the floating attitude is different from the flying attitude. But height control has a problem because waves look like a change in flying height, and even without waves, by the time the height starts to change the system is already late in compensating for a change in pitch. So you'd like to use pitch control to reject the high frequency disturbances and use height control for the overall flying height at low frequencies. You can achieve this frequency separation by low-pass filtering the height control and using a complementary high-pass filter for the pitch control.

Say you are using a pure proportional,fixed gain control law right now. This would be of the form:

delta = Kheight *(Hdesired - Hmeasured) + Gcontrol*Stick

delta = commanded flap angle
Hcommand = desired flying height
Hmeasured = measured flying height
Stick = manual stick input
Kheight = height feedback gain
Gcontrol = manual control sensitivity

Now apply filters and combine with pitch feedback:

delta(s) = Kheight*a/(s+a) *(Hdesired - Hmeasured) + Kpitch*(s*a)/(s+a)*(Pitch_desired - Pitch_measured) + Gcontrol*Stick

In the steady state (s=0), this reduces to the original control law. That means the craft is free to trim at any pitch angle, and the controller will not fight the height control.

delta(s=0) = Kheight*(Hdesired - Hmeasured) + Gcontrol*Stick

In the limit of high frequencies (s=infinity), this becomes

delta(s=infinity) = Kpitch*(Pitch_desired - Pitch_measured) + Gcontrol*Stick

and you have a pure pitch command control law that doesn't pay any attention at all to the measured height. That means waves will not affect the pitch control.

Although I've shown a desired pitch angle, it really doesn't make any difference and can be deleted

delta(s) = Kheight*a/(s+a) *(Hdesired - Hmeasured) + Kpitch*(s*a)/(s+a)*Pitch_desired - Kpitch*(s*a)/(s+a)*Pitch_measured + Gcontrol*Stick

Kpitch*(s*a)/(s+a)*Pitch_desired = 0 at low frequencies, and if Kpitch is constant, Kpitch(s)=0 at high frequencies, so there's no contribution from this term at all due to the washout function. The control law then becomes

delta(s) = Kheight*a/(s+a) *(Hdesired - Hmeasured) - Kpitch*(s*a)/(s+a)*Pitch_measured + Gcontrol*Stick

The way you implement this filtering is to turn it into an integral equation:

delta*(s+a) = Kheight*a*(Hdesired-Hmeasured) - Kpitch*s*a*Pitch_measured+(s+a)*Gcontrol*Stick

Solving for, and dividing through by the highest power of s:
delta*s = Kheight*a*(Hdesired-Hmeasured) - Kpitch*s*a*Pitch_measured +(s+a)*Gcontrol*Stick - a*delta

delta = Gcontrol*Stick - Kpitch*a*Pitch_measured + 1/s*a*(Kheight*(Hdesired-Hmeasured) + Gcontrol*Stick)

And you end up with a proportional-plus integral control law. The pitch control is pure proportional control, and the height control is pure integral control.

"a" is the cross-over frequency where the response is an equal blend of the two. You pick this frequency based on the wave frequencies you want to ignore. The craft will platform waves that are substantially higher in frequency, and will follow waves that are substantially lower in frequency, than a.

Integrating a signal is easy in a digital computer. The simplest way is to multiply the integrand times the time step and add it to the previous value of the integrator. (There are more accurate ways, too.) The two tricky bits are initializing it and applying limiters. You want to initialize the integrator so that the commanded flap position equals the measured flap position - don't just set the integrator to zero. That way, there will not be any sudden transient when the system turns on, no matter where the stick may be.

There is a limit to how large the flap angle, delta, can be, and you want to stop integrating when you get to that limit - don't just reset the value downstream of the integrator. Otherwise, the integrator will continue past the saturation point and then have to integrate back down until it is less than the saturation level before you see any change in the flap angle. This is called integrator wind-up, and is a bad thing.

[lmfoils]

Hello Tom,
thanks for your interesting input!

Originally Posted by tspeer

I suspect the reason your surface piercing foils did not perform as well as the fully submerged foils is because the fully submerged foils were connected across their whole span, while the surface piercing foils had a large gap in the middle of their span. This causes a large increase in induced drag. It might be worth resurrecting the surface piercing foils and putting a piece across the middle to see the difference.

I assume you are referring to the Summer Project 2010. Many parts of the design were suboptimal. The idea behind a pair of surface piercing foils with a large gap of the middle was to give the craft sufficient roll stability. Connecting the two foils would have given us too much area and putting them close together would have resulted in almost no roll stability.
If we were to do it again we would certainly choose higher aspect ratio foils, we did not dare do this at the time because we did not yet have complete confidence in our structural calculations with composites. During the first tests we had heavy ventilation issues because the immersed part of the foils had an aspect ratio of 2 to 3. A partial solution to this was to transfer more load forward and to apply fences. The losses due to induced drag were not really relevant: the 15hp two stroke had more than enough power. We never managed a full throttle run.
We were still experimenting with various angles of attack as the local police sent us out of the water (our craft is not exactly legal). However we learnt a lot and hope to test the boat to it's full capabilities somewhere sometime.


Your idea of height-pitch controller is very interesting. However I am not able to follow you in two points:

Originally Posted by tspeer

Kpitch*(s*a)/(s+a)*Pitch_desired = 0 at low frequencies, and if Kpitch is constant, Kpitch(s)=0 at high frequencies, so there's no contribution from this term at all due to the washout function. The control law then becomes

delta(s) = Kheight*a/(s+a) *(Hdesired - Hmeasured) - Kpitch*(s*a)/(s+a)*Pitch_measured + Gcontrol*Stick

The whole pitch term indeed is 0 at low frequencies and constant for high frequencies (see attached bode plot for a=100rad/s) but what do you mean with Kpitch(s)=0 at high frequencies?

Originally Posted by tspeer

Solving for, and dividing through by the highest power of s:
delta*s = Kheight*a*(Hdesired-Hmeasured) - Kpitch*s*a*Pitch_measured +(s+a)*Gcontrol*Stick - a*delta

delta = Gcontrol*Stick - Kpitch*a*Pitch_measured + 1/s*a*(Kheight*(Hdesired-Hmeasured) + Gcontrol*Stick)

Why can the a*delta term be eliminated?

Marco

[tspeer]

Originally Posted by lmfoils

...what do you mean with Kpitch(s)=0 at high frequencies?

My mistake. I meant to say Pitch_desired has no content at high frequencies.

Why can the a*delta term be eliminated?

I knew I should have done a proper job of deriving the equations instead of simply banging them out in a message! I inadvertently dropped it off when going to the last equation. The proper form is:

delta = Gcontrol*Stick - Kpitch*a*Pitch_measured + 1/s*a*(Kheight*(Hdesired-Hmeasured) + Gcontrol*Stick-delta)

My main point was that height control and pitch control can be combined together. It's easy to see that the sum of a/(s+a) and s/(s+a) is unity over the whole frequency range, so this makes it possible to blend things together on a frequency-weighted basis, just like A*(1-k) + B*k algebraically blends from A to B as k goes from 0 to 1.

There's another really good reason for feeding back pitch to the height control loop. At constant speed, the rate of change of height is proportional to the flight-path angle, pitch = flight-path + angle of attack (ignoring heel), and angle of attack is approximately constant when not doing sudden maneuvers. So the rate of change of height is approximately proportional to pitch attitude, and feeding back pitch attitude has same effect as feeding back the rate of change of height. It adds damping to the height loop. When you use pure position feedback, like height-only feedback, you create an oscillator as you increase the loop gain. The pitch feedback makes the craft settle down faster after a disturbance, and it allows you to use more gain (sensitivity) on the height feedback for more precise control without causing instability.

[lmfoils]

Originally Posted by tspeer

There's another really good reason for feeding back pitch to the height control loop. At constant speed, the rate of change of height is proportional to the flight-path angle, pitch = flight-path + angle of attack (ignoring heel), and angle of attack is approximately constant when not doing sudden maneuvers. So the rate of change of height is approximately proportional to pitch attitude, and feeding back pitch attitude has same effect as feeding back the rate of change of height. It adds damping to the height loop. When you use pure position feedback, like height-only feedback, you create an oscillator as you increase the loop gain. The pitch feedback makes the craft settle down faster after a disturbance, and it allows you to use more gain (sensitivity) on the height feedback for more precise control without causing instability.

Thanks for the clarifications. We will of course consider your input for the design of the second version of the controller.

Do you happen to know of papers treating active stabilization of hydrofoil craft?

There seems to be some valid information in
Saito et al, 1991, Rough water capabilities of fully submerged hydrofoil craft “Jetfoil”
but I still have to get hold of this.

[lmfoils]

As usual, one would be tempted to say, the team met at Christmas to test the latest development. This time it was an upgraded version of the craft tested during Summer Project 2010.
The front beam was lengthened, the front foils were also extended with a tapered section. The modifications proved to be very effective. More details can be found in the report on our website Link to report

Here is the video!

[BPL]

Now that is a Christmas card photograph!
This is what it's all about. Good on you lmfoils.

[intrepid71]

Originally Posted by lmfoils

As usual, one would be tempted to say, the team met at Christmas to test the latest development. This time it was an upgraded version of the craft tested during Summer Project 2010.
The front beam was lengthened, the front foils were also extended with a tapered section. The modifications proved to be very effective. More details can be found in the report on our website Link to report

Here is the video!

Nice video. The hydrofoil seemed to perform well in smooth water. The glassy conditions were beautiful.

Your impromptu dive at 1:10 when you attempted to pass the wake of the camera boat is a very clear demonstration of the shortcomings of fixed hydrofoils. The downward velocity component of the water on the backside of the wake reduces the effective angle of attack of the water flow into the foil. This reduces the the lift rather suddenly, resulting in the instantaneous morphing of your boat into a submarine. I was surprised how dramatic the failure was. Good thing you were wearing wetsuits!

Do you have plans to turn the fully submerged and controlled foils you had previously towed into a self-propelled craft with an outboard?

[lmfoils]

BPL
your understanding of our philosophy is appreciated!

Intrepid71
Thanks! Indeed this is always a problem with surface piercing configurations, our foil configuration however is not optimal. If we were to start again we would choose a higher aspect ratio planform with taper along a larger portion of the span, furthermore extending the foils in the outboard direction with respect to the struts would provide some extra area (also helpful in turns).
Crossing the turbulent and air rich wake of the propeller did not help.
You can see some stills of the crash on our flickr page under http://www.flickr.com/photos/foils/6...ream/lightbox/ (click on the button "newer" at the top of the page to view the stills in order)

Yes, a new set of fully submerged foils for hydrofoil 3 is planned. The foils used for the surfboard are not capable of supporting higher weights and speeds, mainly because of structural limitations. A possibility might be to experiment with an electric motor...

Next on the roadmap is the test of a new and more sophisticated control system for the surfboard, if everything goes as planned it will be used for hydrofoil 3.

[Doug Lord]

Originally Posted by lmfoils

BPL
your understanding of our philosophy is appreciated!

Intrepid71
Thanks! Indeed this is always a problem with surface piercing configurations, our foil configuration however is not optimal. If we were to start again we would choose a higher aspect ratio planform with taper along a larger portion of the span, furthermore extending the foils in the outboard direction with respect to the struts would provide some extra area (also helpful in turns).
Crossing the turbulent and air rich wake of the propeller did not help.
You can see some stills of the crash on our flickr page under http://www.flickr.com/photos/foils/6...ream/lightbox/ (click on the button "newer" at the top of the page to view the stills in order)

Yes, a new set of fully submerged foils for hydrofoil 3 is planned. The foils used for the surfboard are not capable of supporting higher weights and speeds, many because of structural limitations. A possibility might be to experiment with an electric motor...

Next on the roadmap is the test of a new and more sophisticated control system for the surfboard, if everything goes as planned it will be used for hydrofoil 3.

======================
Great to hear about your development! Thanks for keeping us informed....

[intrepid71]

Originally Posted by lmfoils

BPL
your understanding of our philosophy is appreciated!

Intrepid71
Thanks! Indeed this is always a problem with surface piercing configurations, our foil configuration however is not optimal. If we were to start again we would choose a higher aspect ratio planform with taper along a larger portion of the span, furthermore extending the foils in the outboard direction with respect to the struts would provide some extra area (also helpful in turns).
Crossing the turbulent and air rich wake of the propeller did not help.
You can see some stills of the crash on our flickr page under http://www.flickr.com/photos/foils/6...ream/lightbox/ (click on the button "newer" at the top of the page to view the stills in order)

Yes, a new set of fully submerged foils for hydrofoil 3 is planned. The foils used for the surfboard are not capable of supporting higher weights and speeds, mainly because of structural limitations. A possibility might be to experiment with an electric motor...

Next on the roadmap is the test of a new and more sophisticated control system for the surfboard, if everything goes as planned it will be used for hydrofoil 3.

You might find this youtube video of a similar smalll surface-piercing hydrofoil trying to pass a wake interesting. They have similar success although your crash was more dramatic.

http://www.youtube.com/watch?v=5zGiH1vgpnk