PID Control: Is Droop Always Against Setpoint?

[russ_watters]

Possibly easy/concise question:

Is droop in PID control alway against the setpoint? (as the name implies?)

Background:
I have a humidifier that is consistently controlling high. The control point is room dewpoint, the room has outside air supply (dry in winter) and the humidifier uses plant steam to generate clean steam. The valve position fluctuates with the weather (OA Dewpoint), indicating it is under control and responsive, but *always* ends up with about the same offset above setpoint.

I see no physical reason why this should happen. I'm not controlling to a proxy like duct dewpoint, I'm controlling to the point I want to control to, so it should eventually get there, right?

I thought droop was a function of proportional gain - essentially a certain amount below setpoint causes a certain amount the valve is opened. Since the OA is always trying to pull the humidity in the room down, I would expect the error to be low.

Or, when the weather changes, the humidity in the room should creep up, reversing the push/pull direction? Or only if it changes fast enough to get you on the other side of the setpoint?

Is this a simple tuning issue, needing more integral control to bring the room conditions in-line with the setpoint or am I missing something in the control theory?

As primarily a design engineer, I design systems and assume the controls can make it happen. But on this project (study), I'm working with the commissioning department because the client isn't sure if it is a control problem or a design problem. So this is a little more down in the weeds than I usually get.

 

[ChemAir]

Various control manufacturers (Foxboro, Honeywell, Yokogawa, Allen Bradley, others) are pretty casual with their use of control textbook terminology e.g.--proportional band, damping, gain, reset, windup, droop, direct acting, reverse acting, increase/decrease, etc.. Sometimes it is difficult to be sure terminology is matching up during communication.

In most cases, proportional control alone will never make set point, it just gets to an error level proportional to the load on the controller. Adjusting the Proportional band (gain) will change the offset (difference between process measurement and control point) at a given control load. Getting too much gain will result in an unstable controller output.

I expect some integral action would be helpful in the description above. Reset (integral) gets the process measurement the rest of the way to the setpoint based on size and time a control variable has been in error. The longer the error, the larger a given reset response. It can also be adjusted too much and can overshoot/become unstable.

It would be good to watch the controller output percentage, on a chart with the process variable in real time when it is operating to see what it is doing. This can tell you if it is beyond it's ability to control--Controller output at its maximum or minimum and things aren't changing like they should. This plot also can show how changes in tuning parameters affect the controller performance. You can figure this out without a plot, but it is easier if you have the visual.

 

[jrmichler]

Are you measuring the room humidity with the same sensor that is feeding the humidifier controller? Humidity sensors are known to drift.

Is it simple feedback control, where the only sensor is the room humidity sensor? Or is there an OA feedforward sensor in the system?

What is the system response to a step input? Blow some dry air (from a desiccant air dryer), or dry nitrogen on the sensor, then wrap the sensor in a wet rag. Is the integral gain even turned on? Is the derivative gain turned on? Did somebody try to tune the controller by random parameter adjusting? The Ziegler-Nichols tuning method is not the best, but is well documented and should get you in the ball park.

 

[russ_watters]

Thanks guys.

Various control manufacturers (Foxboro, Honeywell, Yokogawa, Allen Bradley, others) are pretty casual with their use of control textbook terminology e.g.--proportional band, damping, gain, reset, windup, droop, direct acting, reverse acting, increase/decrease, etc.. Sometimes it is difficult to be sure terminology is matching up during communication.

Not surprising. I've just never gotten that far into the weeds on some of those terms.

It would be good to watch the controller output percentage, on a chart with the process variable in real time when it is operating to see what it is doing. This can tell you if it is beyond it's ability to control--Controller output at its maximum or minimum and things aren't changing like they should. This plot also can show how changes in tuning parameters affect the controller performance. You can figure this out without a plot, but it is easier if you have the visual.

Yes, I have a commissioning engineer going to watch it in operation on Monday. Hopefully he'll be able to get that level of access to what the program is doing.

Are you measuring the room humidity with the same sensor that is feeding the humidifier controller? Humidity sensors are known to drift.

Is it simple feedback control, where the only sensor is the room humidity sensor?

Ding, ding.

I was *told* that the control system was controlling to supply duct dewpoint (for responsiveness) with room dewpoint reset (to control to the sensor that has the acceptance criteria). After noticing the offset I started asking more pointed questions and it turns out, that is not the case. There is code for that, but it apparently "wasn't working well", so it was disabled and now the duct sensor is the only input. This alone would explain the offset: there is indeed a surprising amount of what is probably calibration offset between the sensors.

That doesn't preclude other problems, but one known problem at a time...