Control of air jet and servo motor

[Uridan]

Hi,

I am trying to control the height of a ball in a jet stream of air. The height is being calculated by a camera and the air jet coming out controlled by a servo motor (connected with a butterfly valve).

I found the open loop response by giving a step input, i.e I opened the value and then closed it so the ball with rise. And the response was a normal under dumped wave form.

Now I connected the system in a closed loop and the system response has gone totally nuts, the closed loop was done buy a simple unity feedback where it gives the height from the camera sensor subtracted with the Users reference height and the error is fed to the servo motor equation.

What am I doing wrong ? Or is this normal where the closed loop response becomes worst then the open loop response?

Note that No controller has yet been applied.

Thanks
Regards
Uridan

 

[Mech_Engineer]

What do you mean by "totally nuts"? It sounds like the response of the control system is underdamped, are you planning on utilizing a PID loop?

http://en.wikipedia.org/wiki/PID_loop

 

[Uridan]

sorry for that miss understanding saying lol.

When I introduced a feedback to the system by inserting a camera to give the actual position of the ball the servo motor will keep moving and as a result it keeps the ball oscillating and thus, the response is just plane oscillations without any decaying or settling time at any step response.

This is the first System that I am building and I am not sure if this is normal when a feedback is introduced without control, the response of the system becomes worst then open loop.

Yes my whole point of this system is to study different controllers such as PID and Lead lag compensator for a highly none linear system (in this case I am testing a jet stream of air coming out of a nozzle and a ball flouting in it to see its stability as it goes higher).

I haven't introduced any controller nor on the feedback nor at the feed forward part so I can assume that the response would be inappropriate but not at this magnitude.

Thanks
Uridan

 

[Mike_In_Plano]

Hello Uridan,

Mike here. I'm a EE, so I've gotten stuck with a good many control system problems.

Here's my short tutorial on controls:
- Once you have a second order system (ie mass + spring), your complicated enough. Beyond that, your control systems engineer is looking for something else to work on.
- If you don't need your system to have 0 error, than avoid the I term. Integration will make stability more difficult and generally slow things down.
- Add another term. Call it B - for bias. Get the system so it's near the operating point without P, I, or D.
- Add P. A lot of times, it will pull in with enough P and later begin to oscillate with too much P.
- Add some D to make it more stable, some more P to make it accurate...
- Make sure your sensor's not sticky, noisy, or abrupt.
- Do the same for your actuator - you want the action to be smooth.
- If you have a resonance, you can tune the system more to slow down (more d) and add a notch filter at the resonance. This knocks the gain off the resonance, but your system still has to slow down.
- If your actuator is not directly controlling the needed input (ie The air delivery doesn't change perfectly with the valve) Then feel free to make a second "embedded" control loop that reads your parameter (air pressure / flow?) and controls the valve to make it follow your command. This is a powerful technique, because it takes down the order of the larger control loop. Make your embedded loops fast - a lot of P - nothing else.

I'd start by shimming the ball up and down to see that the position feedback is operating fairly linear and doesn't have any sudden transitions (ie dv/dx at "magic" position).

Next, I'd make a little force gauge by taping some spring wire to your ball. Then vary the valve and make sure it's controlling smoothly. Flapper valves are notoriously non linear - especially near cracking.

Finally, don't trust PCs for control systems. The lag associated with a PC can be much shorter than the time constants and still make it unstable. Me, I like to make little proto boards with pots. The only difficulty with it is that sometimes you need a really slow integrator and it's hard to do with caps. Then again, most of life is pretty sweet with 'em.

Best Regards,

Mike