The aim of the Winter Project was to test the control system Sandro had developed for his bachelor thesis during the spring semester of 2010.
The starting point were the foils built in summer 2009, back then control was manual with an old RC transmitter, this worked quite well but flying the craft proved to be quite challenging and frequent crashes were unavoidable.
We hoped that the observer based LQR controller would make life easier. Several types of sensors were considered and we concluded that one of the most simple and affordable ways to measure flight height, pitch and roll angles was by using four wands attached to potentiometers. The height is the average of the measured values, the pitch and roll angles are obtained by computing the normal vector of the plane spanned by the two segments joining the diagonally opposite corners and computing the angle to the vertical vector.
In summer 2009 waterproofing was very basic, some large balloons were used to protect servos and micro-controller, this time a more thorough approach was selected: micro-controller and battery were housed in an Otter Box. The servo cases were opened, the top packed with grease, the lower part with silicone, built back together and all seams were closed with silicone.
On the top of each strut a small plate was added in which a potentiometer with 5k Ohm was mounted. The wands are simple aluminium T bars.
From November onwards several dry tests were performed and some bugs in the software corrected. We also found out that our beloved Klepper surfboard had disappeared form the yacht club under mysterious circumstances, fortunately a friend came to the rescue and lent us an old board of his. The Klepper weighted 17kg, the “new” one 21kg.
On the 23rd December the core part of the project started. First thing was to adapt the strut supports to the new board. This was accomplished with a sandwich consisting of glass fiber – foam – carbon fiber. The carbon was laid on the board (covered with cellophane), the foam saturated with resin and the glass fiber was laid on top. The sandwich was then wrapped in cellophane and struts and foils were attached. After aligning everything we waited for the resin to cure.
The result was not great but usable. Although we had used a whopping 700g of resin the foam was not completely saturated in some parts and thus very frail.
At this point we just had to attach the wands, plug everything together and test the software. This task involved some intensive troubleshooting and took a full two days. On of the most challenging parts was calibrating the potentiometers, in the process we also changed all four to more rugged Multiplex servo potentiometers with the same resistance.
On the morning of the 26th all equipment was transferred to the yacht club for final assembly and the foils were once more aligned to have an angle of attack of 6 degrees.
Trying not to have too high hopes we hit the lake at eleven o'clock. It was immediately clear that the controller did have some kind of effect, the stability on the roll axis was exceptional even during gusts and as sets of larger waves rolled by. The craft however would not fly well: an extreme pitch up attitude had to be forced (with a manual override) to convince it to take off and stay airborne for a few meters. After about 40 minutes of tests a detached wand forced us to return to the shore.
Some more dry tests were performed and we noticed a misbehavior of the height-pitch controller. During the debugging session one of the push rods broke at the beginning of the tread for the clip: a classic fatigue failure. The threaded part has a diameter of only 2mm and was subjected to repeated alternated bending and tension/compression loads. From the look of the fracture, corrosion must have also played a role. An emergency repair was quickly made.
After an hour or so of debugging it became clear that the cause of the misbehaving height-pitch controller was the very fact that the latter was responsible for both states. Specifically, the controller anticipated a loss of height as a result of non-zero pitch angles, and saturated the entire controller output with a flap correction to compensate for the imminent height loss – the pitch correction itself was lost. This made it basically impossible for the surfboard to become airborne since the large flap angles caused by the height correction significantly increased the drag of the foils.
As the sun was going down, as a quick workaround, we forced the measured pitch angle to zero. With this hack the height-pitch controller became a height controller.
To our amazement the surfboard now flew almost perfectly, a short pull of the pitch stick was sufficient to get it airborne! On a straight course without large cross-waves the craft stayed in the air without any user input. After about twenty minutes the freezing temperatures made us return to the yacht club. On the way back we tested the craft without pilot and it flew flawlessly!
Interestingly the foils seemed much more efficient than in summer 2009, takeoff speeds were lower and even at slower speeds stability was great. A controller is clearly superior in maintaining a dynamically unstable system in equilibrium than a human pilot. The servos however work a lot: we had an average total current drain of about 1.15A during both test runs.
The objectives for the future are to separate the height and pitch controllers and to optimize the model.