Scan/execution rate affect on 'I' parameter 2

[controlnovice]

I am new to this company and investigating a few control issues.

We have a three layered cascade system. Temperature 1 sends Output to setpoint of Temperature 2 which sends its output to setpoint of Pressure controller (steam valve).

Each controller, or control module, was running at a scan rate of 1 sec.

I want to change the scan, or execution, rates so the pressure controller is 1 sec, the Temp 2 is around 5 sec, and Temp 1 controller at 20 or 30 sec.

On Temp 2 or Temp 1 controller, if the reset, or Integral parameter is set less than 5 sec, how does this affect the control if the scan rate is equal to or greater than 5 sec?

______________________________________________________________________________
This is normally the space where people post something insightful.

 

[PNachtwey]

I want to change the scan, or execution, rates so the pressure controller is 1 sec, the Temp 2 is around 5 sec, and Temp 1 controller at 20 or 30 sec.

Why?

I want to change the scan, or execution, rates so the pressure controller is 1 sec, the Temp 2 is around 5 sec, and Temp 1 controller at 20 or 30 sec.

A good PID will take the sample or update time into account.

On Temp 2 or Temp 1 controller, if the reset, or Integral parameter is set less than 5 sec

What? Normally the integrator gain is set to about the time constant of the system it is controlling. Are you sure your plant has time constants that are that fast? Second, the update or sample time should be 10 times faster than the plant time constant or the closed loop time constant which ever is shorter.

I see trouble ahead because if what you are saying is true then your plant is barely working now with the sample time only 5 times faster than the integrator time constant which should be about the same as the time constant(s) of the system that it is controlling. Everything gets worse if you have dead time that is a significant part of the plant time constant(s).

I don't see why you want to slow down your sample rate. Usually the faster the update the better the control. There are practical limits to how fast the update rate should be. If the feed back has poor resolution and there is lots of sample jitter the derivative terms will cause more harm than good but you didn't say you were using the derivative term.

 

[roydm]

As I understand it the scan rate just sets the frequency of the calculation
e.g.
higher scan rate = more smaller steps,
lower scan rate = fewer large steps
At the end of a minute the Integral change should be the same, percent wise.
If it's a PLC based system, all bets are off. Some of those don't seem to understand PID at all.

I can sort of follow your reasoning Temp 2 is a slow loop and Temp 1 is very slow so why do the calculation so often?
I think you would find every 5 seconds and again at 25 seconds the pressure loop would make a sudden step changes and become unstable.
Your 5 second integral time seems very fast to me, that must be for the pressure loop.
Regards
Roy

 

[FrancisL]

Ask youself how quickly the process can change, the sample time should be in proportion, you gain nothing by sampling too fast, and indeed some PID calcs work worse when the samples are too fast due to the loss of the changes into the noise.
So yes, make the outer loops slower.
By the way properly programmed PLC's can do it, they should run the PID loops on a time based interrupt, not every scan.
Francis

 

[controlnovice]

Thank you for your replies. They do help, but I would like to clarify my question:

Ignore the example I gave above. Don't worry about pressure, temperature or speed of the control.

A PID module or block is calculating only once every 5 seconds.

The seconds/reset = 3 seconds.

Since the block only runs every 5 seconds, but the PID reset is every 3 seconds, how does the PID cacluate the Integral term in the block?

______________________________________________________________________________
This is normally the space where people post something insightful.

 

[FrancisL]

Your Reset is not something that is 'every 3 seconds', it is 3 secs.
Description of the PID Algorithm

http://www.jashaw.com/pid/description.html

Have you used any methods or software to tune the loops?

 

[PNachtwey]

The PID in most PLCs is
Error=SP-PV
CO=Kc*(Error+(T/Ti)*Error+(Td/T)*(Error-Error(n-1))

In theory you should be able to do what you want to do. You can see that the sample time (T) changes the effective gain of the integrator and derivative time constants. In your case T ( scan time ) would be 5 seconds and the integrator time constant ( Ti ) would be 3 seconds. If the PID can properly handle the math you will be OK. A lot depends on the quality of the PLC implementation. Some PLCs have 16 bit CPUs without floating point processors so all math is done using only 16 bits of precision where truncation errors can be significant.