PID or something else?

[BLW]

Hello all,

 I have a gas control system running through a set of chambers.  The repsonse of the system is very slow at first but then very fast and so the PID is not able to prevent an overshoot.  I've been trying to adapt my system to use the PID or PID Autotune.  I can't get either one of these controlers to perform as my system requires.

An ideal control system would put the petal to the metal intitially but not overshoot the goal/setpoint.  So, my goal is to achieve the setpoint very quickly with no or very little overshoot.

What are my options, do I have other options?

Thanks,

BLW

[LV_Pro]

Is there enough time between the transition from slow response to fast response that you could use a strictly linear control up to near the transition point, then switch? Does the transition occur with some predictibility, say at an approximate flow or pressure?

[johnsold]

It sounds like you need an adaptive system.  As Putnam asked, can you detect or predict reliably when the change in response speed occurs?  If you can detect that, then switch the controller from a set of parameters which works well for the slow phase to a different set chosen for the fast phase.  Make sure the controller does not generate some transient step at the output when you change parameters.

Lynn 

[gsussman]

It sounds like some form of gain scheduling might be the ticket, however the PID advanced VIs or the PID function block have a setting called Linearity that will allow a sort of continuous gain scheduling. By adjusting the linearity setting between .1 and 1 you can generate an essentially parabolic response of the controller.

Linearity = 1 will give a true linear response (gain will be applied evenly across all error magnitudes)

Linearity = .1 will give a parabolic response (gain will be applied at the full setting when error is large, and will be reduced parabolicly as the error signal diminishes).

The idea is that when errors are large, the higher gain settings will allow quicker system response but as we get closer to the setpoint, the gains are continually reduced to provide finer control to get you in range.

The limitation that I see with this is that you said you don't want any overshoot (critically damped). This seems to be at odds with getting to setpoint quickly.

http://zone.ni.com/reference/en-XX/help/370622F-01/lvfbhelp/pid_fb/

[BLW]

I think all are leading to very viable directions.  Here's more info.

Time from setpoint change to first realized change in gas mixture is about 8 seconds then the gas mixture accelerates to the setpoint and overshoots at about 30 seconds.  Here's another caviat, we're occilating gas which to operate properly needs some type of adaptive settings because the gases are much more adaptive or "responsive" when moving our setpoint down rather than up.

Any more ideas?  I'll start trying suggestions...  what did you mean by an adaptive system... what type of LabVIEW controller?

BLW

[johnsold]

I do not have any specific LV examples for adaptive control.  Generally adaptive control modifies the controller according to some function of the state of the system.

Lynn 

[Barp · Accepted solution]

Which version of LabVIEW you have? After version 8.2, we added several examples to the shipping PID (now called PID and Fuzzy Logic Toolkit) that shows how to do "Gain Scheduling" or how to use the "Advanced PID" VIs that has the nonlinear parameter. Notice also that when you change parameters on the PID.vi, it should try to change with bumpless transition. I recommend start with:

\LabVIEW 2009\examples\control\pid\prctrlex.llb\Gain Scheduling Simulator.vi

Maybe the main challenge you will have is to identify how your plant behavior and how you can tune the PID for the different areas. If you do not have an initial PI parameters that makes the plant stable, you should not use the Autotuning PID VI. This algorithm require you to have at least some initial PI controller and the algorithm will try to improve it.

Let me know if you need more information.

[BLW]

Thanks Barp!  I briefly looked into the Advanced PID but didn't even know what a plant was.  Now that I know this is a viable option I can focus on the few solutions you have mentioned.

Thanks again everyone and I'll let you know how it goes.

BLW