[ScottDean1982]

G'day guys,

I'm new to this forum, and this field, and it's almost not relevant but I thought you might have some wisdom or be able to point me in the right direction, please hear me out.

I am a university student studying avionics, we do a lot in control systems type stuff which brings me to my point.

I am working on a flight simulator that gets motion from a platform supported by three pneumatic actuators supplied by festo, run by three proportional control valves. This in turn is controlled by some basic software, not exactly custom set to run this sortof stuff called labview..

The major problems we are having is we have a basic PID control on each actuator, but this does not take into account the coupling of the actuators to each other through the motion platform and mass of the system (through moments of inertia), nor does it try to counter the pneumatic bounce of the system in any real way. As such the system suffers, and it is up to me to improve this performance through software coding, and modifications to the motion base (if possible).

Could any of you suggest any methods of combatting these problems?
How is the bounce of a pneumatic system compensated for in CNC applications?
etc..
any info on pneumatics applications or where to find said info would be greatly helpful.

Thanks Heaps
Scott.

[fkaCarel]

I once saw a Festo pneumatic servo on a trade show. It was fast, precise and.... horizontal.

In the system you describe you have to deal with the low frequency response of the proportional valves and stick slip. The stick-slip will prevent your system from working. The error will grow and grow until the actuator starts to move. Then the error is to big and the result is overshoot. You will have to define these characteristics in your software, either with hard numbers or an averaging learning system.

Carel

[Al_The_Man]

G'day, Personally I do not associate pneumatics with proportional control valves, unless there are some advances out there I am not aware of.
I would think for applications as precise as flight simulators, you would have to go with hydraulics with servo/proportional valves to get any kind of precise control.
Or at least go air over oil.
Are you using LabView? and what are you using for feedback?
Al.

[fkaCarel]

It works and is used in pick and place units. The advantages are: speed, reasonable positioning and no hydraulics.

Carel

[Al_The_Man]

It works and is used in pick and place units. The advantages are: speed, reasonable positioning and no hydraulics.

Carel

I could see for crude positioning, for Pick-and-place you could probably allow a large in position band, but applications like flight simulator, I just cannot see air working accurately over a non-compressible medium.
Air would not be my first choice, faced with an application like that.
Note: hydraulics can be equally as fast as pneumatics.
Al.

[fkaCarel]

The poster is given the task to make a pneumatic platform work. It's there. And of course it's elastic. But you and I can't change that. And for what it's worth: the Festo system positions better than 1 mm. Pneumatic positioning is more cost effective than hydraulics, because you don't need the maximum speed pump capacity.

Carel

[Geof]

Commercially available flight simulators use hydraulics for precisely the reason Al-the-man describes; pneumatics 'bounce'. Proportional control valves and pneumatics, that is pure pneumatics, are a contradiction in terms. Pneumatics have been used extensively for obtaining motion on automatic lathes but the control came from hydraulic dashpot cylinders, or alternatively air over oil systems where the hydraulic aspect provides the motion control and the pneumatic side provides the driving force. To me this project sounds like the typical impractical things university profs come up with.

[unterhaus]

It seems like you need to have some feedforward in the control loop because overshoot will kill you. I'm curious if anyone ever modelled this system. It seems like the bouncy behavior is entirely predictable. I would like to know what feedback you are using. The thing about a flight simulator is that the visual element is by far the most important. Motion does not have to be incredibly accurate as long as it doesn't make the user wonder what the heck is going on. Unfortunately, bouncing would have that effect.

I'm assuming this is a Stewart platform.
I searched for the terms stewart platform matlab model and got 75000 hits. http://www.google.com/search?hl=en&q...=Google+Search

You also could search for simulink model and see what happens. I added pneumatic to the search terms, and got some hits.

To me this project sounds like the typical impractical things university profs come up with.

I resemble that remark. Actually at my university job I would have used hydraulics. Granted, there is oil involved, but it is much better for this application. And I have a big hydraulic power unit in my lab. The other thing that can be used is the servomotor driven cylinders like Moog makes. They seem like a really good solution to this problem because the maximum force required is not that high. They would use much less energy. The other thing that should be considered is that someone with money may have wanted the university to use pneumatics, or the funding wasn't enough to use a better solution.

[ScottDean1982]

yeah, I know what u mean about hydraulics, however due to the classification of the 'lab' that the flight sim is in, we can't use hydraulics... sucks like you wouldn't believe.

For feedback we are using a linear sliding potentiometer attached to each actuator, that then goes into a Atmel micro-controller that converts to digital, does some moving average smoothing, then feeds back via RS232 (serial 57600) to the computer control running labview.

Anyone know if there are any free-source code for labview that helps to control these sort of things, eg adaptive, sliding mode, any controllers?

THinking of maybe going nuts with a fully detailed state-space analysis..

imagines a huge 12*12 state transition matrix.. *sigh*

Included: Photo of the Sim in all it's glory, can see the three pneumatics in a tripod type setup underneath.


I wish it was a stewart model, instead we got only 3 dof, pitch roll and vertical translation

[lwill]

Could you add air bags or springs to help support the structure and bring it to a more "counter balanced" state? Sort of like making it neutraly boyant? Then the cylenders would only be additive and not have to lift dead weight helping to cut down on the amount of force needed to start movement and not over shoot as bad.

[ScottDean1982]

hmm... *imagines large counterweights pivoted, and supplying lift to each corner of sim*
problem is, you'd have large amounts of travel for the coutnerweights to be effective (amd small enough), and also that would nearly double the inertia of the system, making it even more bouncy.

it would help to make the system more linear in some regards though.. interesting idea.. in some form it might help.

... but it's a bit out of scope of what I am meant to be doing (software control mainly)

[lwill]

Just a thought if all else fails and it has to go back to the drawing board.
Another question though, are the prop valves controling the air going into the cylinders? Generaly when controling how fast a cylinder moves the exahausting air is regulated allowing for higher pressures to be used and still be under control. With out knowing how the whole system is set up, I am just trying to help you eliminate any poosible design errors. I know how it can be some times. The mechanical guys blame it on the electronics, the EE's blame it on the software, and the programers blame it on the hardware, when there may be an under lying flaw that just can't be solved.
Good luck.

[ScottDean1982]

currently the valves are controlling air going into the piston, and out...

I am not sure I understand what you are saying about the exhaust being regulated (by the valve I assume)..

the valve itself has air supply in, two lines out to the cylinder, and two exhaust ports on the valve itself, as such it switches between supplying one end of the cylinder with pressure and opening the other to exhaust, and vice-versa

the valves are MPYE type from Festo (incase you are keen)
www.festo.com

hmm.. just a thought, do you mean that if the valve was hooked up so that the pressure supply went to both of the exhaust ports, and the (now) inlet port was turned into an exhaust instead?
could that be done?

[lwill]

Not exactly, most electro valves will only work one way. What I mean is in a normal "open loop" if you will, air system where positioning/speed is not as critical, air is supplied to one side of a double acting cylinder and the air coming out of the other end (on the other side of the piston) is gerneraly regulated by a flow valve to control speed. When you control the preasure/flow of the air coming into the cylinder, assuming the preasure in the cylinder is 0 to begin with, you will always have a delay filling up the line/cylinder and geting it up to preasure and overcomming friction. If the proportional valve is used and you want a small movement, you still have to overcome this amount to start the movement, so appling a small siginal will seem to have no effect at first, so if you increase the siginal to get it moving sooner the pressure/flow will then be much greater than you theoreticaly need to make the move causing sudden acceleration, causing overshoot. If you were to apply a high preassure to the one side of the cylinder with a standard valve, and then use the prop valve to control how fast the air comes out the other end this delay can be over come because there would be sufficent preasure availible imidiately to start movement. I will try to find a diagram to explain it better if you want.
Just so you know, I have worked on 4 and 6 DOF ride simulators, but they were hydrolic. They supply pressure to both sides of the cylender equaly and control both sides. Increase pressure on one side and decrease the other at the same time creating a natural brakeing effect, but oil does not compress as much as air so the effect is much more instantainious.
I am going to see what I can find on the MPYE valve you gave to see if I am off base or not.

[ScottDean1982]

I am starting to see what you mean now, so you;d have to have three valves, two standard open/shut types (call them (1) and (2)) , and one proportional (call it (3))

to move in one direction, you shut say (2) and open (1), then open (3) slightly to allow some air out of the other end of the piston

sounds like a lot more hardware, but I like the sounds of no spool time.

got any references that I can look at for a similar system?

[lwill]

Ok I now see the pic, and yes if it is hooked up as you described it is in fact controling both in and out and will control the speed correctly. But as I said, opening the valve a small amount to move the cylinder a small amount will still have a large delay since it is controlling both in and out at the same time? The way I described does get a lot more complicated needing more valves, but could give better performance.

[lwill]

We seem to be crossing posts let me see what I can find.

[lwill]

Clippard has alot of good info ike this:
http://www.clippard.com/downloads/ge...Pneumatics.pdf

This is sort of what I mean. To extend the cylender open valve 2 and control the speed with valve 3. To go down, use 1 and 4. You could poosibly tie the both exhusts together and use one proportional valve, but it is 3AM here and I am not thinking as straight (or drawing very good either)

I am sure there is some one else on this site that deals with this stuff more than I do that can verify this or correct me if I am wrong.

[ScottDean1982]

I was curious as to how that would affect the bounce of the system, having a fully pressurised cylinder, assuming a load 75kg's and by seeing how much it would compress under additonal load of 60kg's, to give me an indication.

I used excel to determine the effect, and interestingly enough as the pressure increases inside the piston the effect of adding load drops off from additional movement of 5mm when having one side pressurised and the other at atm (to count for weight) down to a little under 1mm when one side pressurised up to 600KPa and the other (initially) at around 570.

so this might be a win/win situation, faster response, less susceptibility to bounce and coupled movements, etc..

You've given me much to think about.

too bad that it requires more valves, that's a bit of a pain (additonal latencies are not desired) but it will be interesting to see if the pre-inflated state reduces spool times enough to make up for the increased latecy of command.

but the real crux of my work is to design a software controller, anyone got any advice?

[unterhaus]

I would build a model and try to control it first. The models of the Stewart platform can probably be modified fairly easily for this system. You probably would have to add in some of the pneumatic dynamics. I don't know if the (lack of) speed of the valves are hurting you, but some of their dynamics might need to be put in.

On the control side, if cross-coupling is a problem it seems like the proportional gain isn't high enough. If you can't raise the proportional gain, you need to change your controller structure. What I have found with hydraulics is that the systems are non-minimum phase. This means there is a zero in the right half plane, so if you increase your gain too high it goes unstable. It also means that root locus diagrams are drawn backwards. You might want to do a system id to find out if this is the case.

If it's this is a project for a degree requirement, they probably want you to learn something as well as fix the system. But Google works with the keywords: Stewart Platform Labview
Doubt anyone posted their control complete, but it might give you some ideas.