I thought I would take some time to let you know what PSAS has been up to.
Just as a quick history refresher: our recovery system failed in 2005 and we have spent some time getting back into the swing of things. I'd say we are now finally moving forward. Last year (May 2009) we launched a rebuilt LV2 rocket that proved our new recovery system. It was devoid of any interesting electronics. We noticed is that it spun badly—the fault of poorly aligned fins. While this was not detrimental in any way, but it did give us an idea: active roll control. So all winter we worked on a scheme to actively control the spin rate of the rocket. We settled on a canard design—which seems obvious enough. We had to scrub due to weather in May, but we were able to launch June 27th.
Roll Control
This page on our wiki describes the mechanical design and some of the maths involved.
From the beginning the aerodynamics of the design were a sticking point for us. We don't really know any aerodynamicists. We decided that we don't understand how to predict the forces on the canards well enough for model predictive control and instead used a nested PID loop control program (nested to provide gain scheduling since the control authority of the system is non-linear with vehicle velocity).
We even wondered if the canards would interfere in the airflow around the rocket in some way and cause a destabilizing effect. But we had no way of knowing so we just flew it with fingers crossed.
Launch
Other than the roll control we again flew nothing fancy. We still haven't finished the battery enclosure or carrier board for our "real" flight computer. Maybe next time? We finally had good weather in June, a perfectly clear, warm day.
Another thing we did differently was try to nail down our launch procedures. I printed up preflight an launch procedure checklists which we used on the launch day. This proved very helpful both in not forgetting to do anything important, but also to get everyone coordinated and on task.
The launch itself was gorgeous.
From the ground everything went great. The motor lit very quickly and we had real time ground tracking data plus real time (though pre-rendered) simulation data. The recovery system behaved nominally, though further analysis showed the drogue chute deploying later than desired. The rocket flew to 15,700 feet with a max acceleration of around 10g's and a max speed of around Mach 1.1.
Post flight Analysis
We had a bunch of little boards on the rocket that were recording various data, so we have more to look at than anyone has time to do. The really interesting thing is that the roll control was a failure.
At first glance it looked like the canards weren't able to provide enough force to spin the rocket, but upon closer inspection we noticed that, worse, the rocket was spinning in the opposite direction than the canards were pointed! After a week of researching old missile design books we learned that canard based roll control is fundamental difficult because of vortex shedding off the canards interacts with the fins to produce an induced roll in the opposite direction of intended control. This is large enough in our case to cause a control reversal and overall loss of control.
Here is the video from on board showing the struggling canards and no roll control:
So we are still working on what to do next. Clearly control is going to be much more complicated than originally hoped.
This is a good lesson and reminder that rocket science is hard!
Get all of our data and see more pictures and video on our launch data page:
Thanks for posting this. For me, it was nice to read. Even though you failed to prevent the rocket from rolling, I at least learned how not to prevent rolling :)
I was also interested in how well the measured data correlates to the predicted data from OpenRocket. For the most part, it was quite accurate. There were places where the data didn't line up correctly, but it does give a good prediction. It was quite accurate in predicting things like apogee, max speed and max acceleration.
I don't have access to Solid Works, but I might have some tools that can view these files. I'lll take a look later, and see if there is anything I can learn from your methods. Thank you for posting these.
10:01 am July 19, 2010
sampo
Espoo, Finland
Member
posts 9
3
0
Hi,
That's very interesting, since we have also attempted roll control using rectangular canards. We had very similar experience, the control worked for a few seconds and after that the rocket started rolling even though the control fins were fully against the roll. We have assumed that the cause was the fins being too small compared to an unfortunately places asymmetric camera, but the effect you described would explain it much better.
Can you provide any more details on the sources where you found the info? Have you any ideas on how to overcome the problem?
You might also be interested in the simulation listener features in OpenRocket. Basically they allow controlling and even modifying the rocket during the simulation. We first simulated the control software in Java code running on the simulation computer, and later transmitted simulated flight data to the actual Arduino controller over a serial link and received control commands from it.
Note however that the OpenRocket simulations are limited in accuracy when canted fins are used. Especially single canted fins might not be completely correctly simulated, so I suggest using canting only for fin sets with at least two fins. Interference effects between fins are of course not simulated, that would require full CFD simulation to work.
Cheers,
Sampo
12:31 am July 20, 2010
natronics
Portland, OR
Member
posts 17
4
0
@rpulkrabek:
We are really happy with OpenRocket. There is a lot more data analysis to be done including a section devoted specifically to quantifying how well OpenRocket performed and where and how it was wrong. But it's obvious that it does really well as a general purpose rocket flight and stability simulation program.
Unfortunately open source CAD software just isn't very good yet. We are using SolidWorks (which is very good and very expensive) under an academic license. But this makes it hard to share our work. I think they at least make some free design viewers:
We found a little about your roll control while researching for our project, but it's mostly in Finnish. We also found one other person who has flown a roll control system (in Australia I think?) with similar, though ambiguous, results.
We have been researching tirelessly for the last month for some answers. It turns out if you know the right key words to look for there is a lot of information. Try searching for "canard induced roll"
Most of the results are papers (abstracts) from various aeronautical journals. If you have access to a good research library you should be able to get some of the full papers. I also might be able to get you a copy of a paper that someone gave us that explains the effect and attempts to quantify it. Send me an email if you're interested.
With caveats that I am not an aerodynamicist this is what I have come to understand about the phenomenon:
When a canard is at an angle of attack it forms vortices along the leading edge of the wing which separates and continues in the air stream behind the wing. When these get near another surface (like fins in the back of a rocket) they create lift (either by lowering the pressure on one side of the fin or by the vortex itself hitting the fin). In certain configurations the induced roll created by the fin-vortex interaction is more than the roll created by the canards – in our case producing a control reversal. At the very least it lowers the effectiveness of the canards.
Mitigation
This is a big problem for missile design folks so, luckily for us, a fair amount of work on the subject. Apparently the most popular way to fix the problem is to have free-spinning fins. That is, they actually put the fin can on bearings and allow it to spin freely. That way the induced roll simply spins the fin can and not the rocket.
I have seen one paper that suggests that if the canard span (how far off the rocket body the canard pokes out) is more than 75% the fin span then the effect is minimized. This is because the vortex sheds off the end of the canard, so if you make a wide canard the vortexes fall far enough away from the fins to not interfere too much. There is also an area ratio component that is probably important.
Another method is to have control surfaces on the fins to directly oppose the effect. In other words, the brute force method.
I read about the simulation listeners when I first looked at OpenRocket. It think it's a great feature! We tested our control system entirely in simulation with MATLAB. We tried some simulations in OpenRocket but we noticed that the roll predictions didn't seem to add up to what we were expecting. We decided just to fly it and see what happens. And it turns out that full CFD or wind tunnel tests are the only thing that would describe this problem anyway.
I did find a company that claims to have prediction software for the effect. They list some potentially helpful papers.
My hope is in the coming months we will come to understand what is happening in some detail. We have started a page about it on our wiki off the launch data page:
Thanks for taking the effort to keep us up to date on PSAS's activities, Nathan. Even though your roll control experiment was unfortunately a failure, I have to say that the fact that PSAS was able to design, manufacture and fly as mechanically-complicated a device as your canard system is a genuine inspiration to CSTART – and the quality of video you recovered from that flight is excellent! It's fantastic to hear that OpenRocket is serving you well as an accurate flight simulator.
Sorry if this is a naive question from someone still learning his way around rocketry, but what is PSAS's motivation for developing high quality roll control? My research seems to suggest that some degree of rolling is, in fact, a good thing from the perspective of stability. It helps to minimse weathercocking, and also helps maintain stability at higher altitudes where the atmospheric density is low and so fins don't do an awful lot. I can see that, on lower altitude flights, this advantage of roll is basically negated, but is it actually a net negative in any way? Does it contribute significantly to drag? Or is the motivation just to get nicer video from a non-rolling camera?
Main CLLARE workgroups: Mission Planning, Navigation and Guidance. I do maths, physics, C, Python and Java.
2:36 pm July 20, 2010
natronics
Portland, OR
Member
posts 17
6
0
That's an excellent question!
We first came up with the idea just to get better video footage, but there is another very good reason for doing it: practice.
Since we want to eventually build a proof of concept orbital vehicle, we are going to eventually need to do real 6DOF control of a rocket. We plan to do this in a series of reasonable steps, making mistakes and learning from them in the process.
The roll control is step one. It's a reasonably simple, 1DOF control problem. However, as you saw, nothing is easy in rocket science. There always seems to be hidden problems. So we start with roll and if we can make that work move on to roll/pitch or the like. Eventually we should have an active stabilized rocket (that goes exactly straight up). Then we know we can do guidance and we can start to think about how to get to orbit.
In general you are right about spin stabilized rockets, they should fly straighter and will be stable even after they leave the atmosphere. Eventually we will have to worry about RCS instead of fins, but for now it's easier to practice with fins. We have learned a great deal though this project about designing control systems and some of the complexities of rocket maneuvering. For that standpoint it's been a big success.
8:15 pm July 20, 2010
Luke Maurits
Adelaide, Australia
Admin
posts 1483
7
0
And that's an excellent answer! It's truly inspiring to see a group on your scale talking seriously and openly about amateur orbital flight and taking the first steps to solving some of the sub-problems on the way. I wish you every success with your future roll control flights!
Main CLLARE workgroups: Mission Planning, Navigation and Guidance. I do maths, physics, C, Python and Java.
Login 
Register 
Search 
Home
Topic RSS 
