Static friction and PID Control

[racrowd]

I'm running an electronic regulator that is being used to pilot a self exhausting pneumatic regulator.  I've tuned it fairly well, but I think I'm getting some integral wind up from the static friction in the regulator.  Does any body have any suggestions on how to compensate for this?  I believe it is due to a guide on the piston that helps with alignment because it seems to happen on both the opening and closing.

[Controls Guy]

A bit slower I and/or more P or D might help if it gets the actuator unstuck before the I can wind up.  If you're not using differential, it might help this situation.   One thing I've done in DL PLC's to limit integral windup is to limit the bias until near setpoint.  IOW, let P&D (if any) do their thing to get you near setpoint, and only use I for reset.

In a case where an actuator can stick, people also use dither, or superimpose a slight wave action on the CV.  If the loop comes up with 68%, vary the output between 66% and 70% a couple times a second or something.  That helps keep the actuator from getting stuck, and is also used (sometimes in larger amounts or higher speed) with actuators that have a deadband such that they ignore very small command signal changes.  I've seen that used in electric motorized valves feeding heat exchangers and so on, where a little noise in the CV won't make it through the system to the PV.

I assume the PV is the pressure downstream of the piloted regulator.  If that's the case, and if the variation from the dither or the differential makes it through, AND if that's an issue, you might try  some accumulation downstream of the piloted regulator.

[racrowd]

My system is a 2" NPT regulator (proportion air regulator R000C) with about 35 gallons downstream.  I'm using a sentronic plus 1/4" NPT 0-450 psi to pilot the regulator.  Adding more volume would help, but that wouldn't be possible due to the machine limits.  I think you spelled out the problem a little better with having a large deadband.  I'm really trying to improve the accuracy of this device using a PID controller (as to make small adjustments to get within range).

More P shows to be a problem making the system unstable at higher pressure (300 psi) since I have to clamp the pilot to 320psi.  I'll look at the effects of more D and clamping the Bias.

I like the idea of adding some value to my output momentarily to overcome the static friction.  I'll try this if adjusting the PI&D values don't help.

Thanks for the help. 

[Controls Guy]

What more often causes your PV-SP error, load changes or setpoint changes or both?

I wouldn't anticipate an [intentional] deadband in the piloted regulator, but there might be one in your I/P, and my suggestions would be the same.  Are you sure whether the time lag is in the main regulator and not in the I/P?  Is there a gauge on the pilot control line or a DP between the pilot line and the PV to see how well the 2" regulator is following?

If you already have 35 gallons on the outlet, I'd think that would be plenty of volume to overcome the dither or differential noise.

Also, try if you can to verify that there's no or minimal mechanical backlash in the I/P and main actuator.  That can look like deadband as well and cause windup.

[racrowd]

Let me start out with my equipment, I'm making an automated valve tester for 1" up to 2-1/2" NPT diaphragm valves.  One of the tests is to verify that it actually operates at the extreme pressures that we state that it will.  Some diaphragm valves are sensitive to flow, so too much restriction in the plumbing will make the pressure drop across the valve too low and it will not close properly.  Therefore, my regulator driving all of this is intended to go to a pressure so the valve downstream can dump a small amount of air (just enough for the machine to verify that it actually fully opened).  So, the amount of time it takes for the regulator to recover is critical to my operation and cycle time.

I was finding that the PID loop controlling the regulator wanted to go really high on the pilot pressure even when it wasn't neccessary, so I thought I would try to protect my pilot equipment and clamp the outlet of the regulator to +/- 25 psi of the goal.  After repeated use on really large test valves, the inlet pressure drops so low that it makes the pressure drop across my regulator only 50 psi where it was 250, so my cfm drops and makes my control loop continually wanting to climb.  Since it is clamped, the pilot isn't increasing crazily and the regulator is fully open, but the integral windup is still there.  So, I'm trying to come up with some solutions, and maybe I had some ideas.

1. Is there a way to limit integral term so that it doesn't make it over shoot so much when the recovery slows down due to a decrease in the pressure drop across my regulator.  One thought I had was to change my Ti so that when it gets really big integral then to increase my Ti so that it has less effect on the control, but I think this is complicated and probably wouldn't work. Maybe change this to increase the Ti term when the pressure drop decreases since I know it will take longer.  This would keep it from being unstable, but it wouldn't really help with the recovery time.

2. Another thought would be to use the goal pressure directly to the regulator, but it is only accurate to +/- 10 psi.  So, then I would run a PID loop in parallel to correct this.  The PID loop will go through a LERP where 0% output will be -10 psi and 100% is +10 psi.  This will only help if the integral windup doesn't continue beyond the 100% output.  I'm hoping that there is some cutoff internally when the loop maxes out, so I figured I would ask you guys.  Attached is a picture of what I'm trying to do.

This gets pretty confusing fast, so I figured I would share my thoughts and see what you guys think.  I posted this here because it had some other details about the equipment I'm using, but I don't think it has too much to do with the friction issues I was having before.

[racrowd]

Actually, I saw that there was an error in the image I attached.  The feedback should be coming from the pressure transducer.

[Controls Guy]

I was finding that the PID loop controlling the regulator wanted to go really high on the pilot pressure even when it wasn't neccessary, so I thought I would try to protect my pilot equipment and clamp the outlet of the regulator to +/- 25 psi of the goal.  After repeated use on really large test valves, the inlet pressure drops so low that it makes the pressure drop across my regulator only 50 psi where it was 250, so my cfm drops and makes my control loop continually wanting to climb.  Since it is clamped, the pilot isn't increasing crazily and the regulator is fully open, but the integral windup is still there.  So, I'm trying to come up with some solutions, and maybe I had some ideas.

Could you do a plumbing schematic so I can make sure I'm understanding this paragraph?  If I'm following you, it sounds as if when you test large valves, you're able to complete tests fast enough to overwhelm the primary air supply to the test equipment.  Is that right?

If so, I don't think you're going to be able to have a solid fix for that with a better PID.  If you do improve it and get a few more tests with the system working satisfactorily, the usage rate of air is still exceeding the capability of the system to make more so you're going to crash eventually.  You'd have to have more compressed air capacity or test slower.

You might be able to regulate the incoming air to 50psi over what you want to regulate to at the test valve inlet, and then the flow variation due to recent consumption goes away at least partially, and the PID would become better behaved.  I have a hunch that might even fix the problem entirely or alleviate it because the lower flow rate through the piloted regulator (which the PID could deal with because now it's not variable), might slow the test rate to where the supply compressor could keep up.  Implementing my last paragraph de facto.

[racrowd]

yes, you are correct about comprehending my paragraph.

You might be able to regulate the incoming air to 50psi over what you want to regulate to at the test valve inlet, and then the flow variation due to recent consumption goes away at least partially, and the PID would become better behaved.  I have a hunch that might even fix the problem entirely or alleviate it because the lower flow rate through the piloted regulator (which the PID could deal with because now it's not variable), might slow the test rate to where the supply compressor could keep up.  Implementing my last paragraph de facto.

What you are saying makes sense, but the flow feeding it will have to be higher than what is going out of the regulator which might be a little more complicated.  It really has me thinking though.

Regarding my other idea about using the PID as a correcting term, is this a bad idea?  I'm not familiar enough with the Do-More PID loop calculations to understand what will happen.  I think it might alleviate the integral windup and offer output clamping.  I was trying to understand the "Freeze Bias", and based on my understanding, if the output of the control loop is outside of the 0%-100% range, then it no longer accumulates (DoMore Help topic DMD0352 section "Reset Windup Protection"). But it probably won't start out at 50% which will be ok, because it will correct itself accordingly I think.

[Controls Guy]

You can't fix rust with a paint job.  You have to either test more slowly, have (way) more local accumulation, or have a higher capacity air supply.  (Even then, the local accumulation is only a bandaid of sorts that might get you through a batch of tests if you do them in batches rather than totally continuously)

Improving your PID is at best a stopgap that would get you a few more tests before the supply pressure is so low you can't continue to test.  If one aspect of the system is inadequate it doesn't work well to try to make something else so fantastic that it compensates.  Directly attack the thing that doesn't work, and you'll probably find out everything else is good enough already.

[Controls Guy]

Imagine if you had some electric heaters that you were controlling with PID via some SCR's, but under high load, you could actually drag down the service voltage to the plant.  You wouldn't try to fix the (admittedly) screwed up control performance by heroic efforts on the control system design and tuning, you'd reduce the load or secure more capacity.  You wouldn't even attempt to fix that with the heater control PID.

[racrowd]

You are right.  My solution when I found that the plant air couldn't keep up was to slow the cycle rate down on the larger valves and this solved my problem.  I know that more accumulation before and even after the regulator will definitely help.  I'm just looking into ways to improve the throughput, and letting you know how my system is currently responding.  Based on what you are saying there is nothing else I can do from a control perspective.  All things point to the volume of air and rate at which I operate my valves.  I'm following your logic.  If I don't have consistent pressure on the inlet of my regulator, then how can I expect a consistent output pressure through my control loop.  Really though, thank you for your input.  I'm still trying to learn control loops and I think I'm starting to understand it's limitations.

[Controls Guy]

Good.  And if anyone tries to paint it as if it's your control system limiting throughput, don't let them get away with it.  No matter how good your PID is, it can't make more air come out the compressor in an hour!