PID Loop: CV and PV are independent

[mondi_19]

Hello and thanks for reading my question/problem.

I am working on an a application where I have a servo controlled gate that allows product (dry) into a channel. The gate has to stay open to maintain a certain amount of product height. Downstream of the gate a level sensor monitors the product height. As the machine speeds up/down the measured height changes. The servo driven gate needs to move and maintain the desired height.

Controller: Control Logix 5069-L320ERM. Kinetix 5500 servo drive for the gate.

The problem that I am facing by trying to use a traditional PID function is that when setpoint (SP) equals the process variable (height sensor, PV) the output goes to zero and this closes the gate (CV), starving the system of product. After being starved the PID increases the output flooding the system, which causes the PID loop to go to zero and then starving the system of product. The cycle repeats.

 

[rdrast]

Sounds like you have just a proportional controller, with no I (integral) term.

 

[Ken Moore]

Post screen shots of your PID parameters, so everyone can see what you have.

 

[drbitboy]

Post screen shots of your PID parameters, so everyone can see what you have.

+1


Your query to this point is basically asking "How long is a piece of string?"

The first thing to understand is the PID equation, where CV comprises three terms, P, I, D, which terms are summed:

• P = Proportional = Error * Kp
• I = Integral = (Sum of Errors at updates over time) * Ki
• D = Derivative = (Rate of error change since last update) * Kd
• CV = P + I + D

Kp, Ki, and Kd are the primary tuning parameters. It sounds like at least Ki is 0, so the I = (SumErrors)*0 = 0, possibly also Kd, so D = 0. So when the level (PV) equals the setpoint (SP), the error* is zero, so P = (0) * Kp - 0. So CV 0+0+0 = 0.

* error = PV-SP, or possibly SP-PV, depending how the CV drives the servo.


You could add a bias term (PID setup dialog; also PID structure .BIAS element) so the gate remains open when the error is zero, but the better solution, as implied by @rdrast, would be to have a non-zero Ki, which would cause the I term to effectively add bias to the CV.

Tuning this process is going to be far, far more complicated than that, but that is a start.

And until we know more, any help will not get much better than that.

 

[mort81]

Speed?

How fast is the process changing? On the surface sounds like a step control might be more favorable, but do need to understand more of what is going on.

 

[drbitboy]

This is essentially a level control, no different from a tank with a fluid.

The gate controls the inlet flow rate; there might be a second-, or even first-, order effect on the flow rate from the machine speed.

The level at the exit (or at the PV level measurement) of the channel* times the machine speed is the outlet flow rate.

The distance, between the inlet gate and the PV level measurement, divided by the speed of the material will be dead time, which will affect the tuning of any control algorithm.



* or filled cross-sectional area, same thing monotonically but perhaps not linear with level.

@mort81 suggested step contro; there was a recent thread with a non-PID control method, look for posts in the past couple of months by Tom Jenkins.

 

[mondi_19]

@ drbitboy you are correct. This is very similar to level control.
Today I discovered that I need to look at the Smith Predictor. I have to determine the system model to predict the system delay as some are suggesting below.

Thanks for your feedback.

 

[Peter Nachtwey]

As the machine speeds up/down the measured height changes.

Why?
Is it because the material is not being dumped in the channel fast enough?
Are you really trying to control level or volume of material?
The volume of material will change as the speed changes if the level is fixed.

 

[mondi_19]

The task is to maintain level/height. It's the path and time the material takes that's long and that is what i need to determine. The delay of the system. The gate height needs to change to meet the demand for product but also maintain a level that does not starve or flood the system with product. I will continue to test and document more tomorrow by observing the system response as I incrementally adjust the gate position.

The channel that i mention before is actually a fill plate with fill tubes. The machine is rotary volumetric filler. As the machine rotates product is being drawn by the fill plate with fill tubes. I need to maintain a fixed height to cover the fill tubes. The gate controls how much product goes into the filling machine. The size of the fill tubes and a second gate (volume adjustment) effect the position the product gate (device that I need to control). If the density of the product remains the same then by trial and error the position of the product gate can be determined. However as the density changes the volume adjustment gate needs to go up or down to maintain weight/volume. It's this movement of the volume adjustment gate that contributes to the product height hence the product gate needing to move. The system delay seems to be different based on the height of the product gate. This what I need to predict/determine.
I apologize for not including all this information in the beginning. I started with a PID but noticed through trial and error that the PID instruction alone was not enough even. I tried changing the cyle time for PID instructions. Turning it on/off but these test did not provide any good results.

 

[drbitboy]

like this: https://www.youtube.com/watch?v=QOwri12F6S

?
https://www.youtube.com/watch?v=QOwri12F6So

 

[mondi_19]

Yes, like that.

 

[drbitboy]

Yes, like that.

So you want to get a fixed height (i.e. volume) in each fill tube.

And somehow the combination of the upstream "product gate" and downstream "volume adjustment gate" controls that fixed height? Is the fixed height in a fill tube used for the PV? Or is there another level measurement used for the PV e.g. of the bulk material moving toward, or after, the vol. adj. gate?

Where are the fill tubes? Are they between the gates? Are they after the vol. adj. gate? What controls the vol. adj. gate?

What makes the material move (can we call it flow)? Vibration? Fluidization?

 

[mondi_19]

The PV is the height sensor, it is located before the fill tubes, in the center of the machine ( fill plate). As the machine rotates it moves the product from the center to the sides where the fill tubes are located. The product then covers the fill tubes. The fill tubes are on the outside of the machine. There is an area of the machine where we have to fill the tubes before they reach the volume adjustment gate. After this gate the product goes into a container and the process repeats.
The volume adjustment gate is controlled by a checkweigher that sends signals to make incremantal moves. The rotation of the machine makes the material move. As the machine rotates it moves the produt.

 

[drbitboy]

Let me write back what I am hearing:

• each fill tube needs to have at least enough material,
  • to fill one container
    • when the fill tube reaches the volume adjustment gate,
  • BUT each fill tube must not be overfull?
• and the volume adjustment gate position determines how much material will be let out of each fill tube and eventually go into the container from each fill tube
  • and the volume adjustment gate position is controlled by the checkweigher feedback control
• So the volume adjustment gate position will depend on how full each tube is
  • which is in turn dependent on
    • the position of the product gate,
    • plus the speed of the machine

As the machine rotates it moves the product from the center to the sides where the fill tubes are located

"Si muove?" Why? Vibration? Angle of repose? Centrifugal force (cf. here)? A combination of some or all of those? Is the rate of movement affected by the machine speed? Is the rate of movement affected by the material height at the center of the fill plate i.e. by the quantity measured for the PV?

 

[drbitboy]

I still don't understand the process fully, but I would think you would want to adjust the product gate to keep the volume adjustment gate near the center of its range*, so the volume adjustment gate has room to adjust for changing process conditions e.g. process density, humidity, etc.

* whatever that means, e.g. it may be non-linear