Slow PI reaction time???

[RPax]

Hi, not to familiar with tuning loops but was wondering if I could get some advise, recommended liturature or a link to a another similiar question posted on the site. I have a PI temperature control loop which we're commisioning. At the moment the valve closes when the Process variable goes behond the setpoint and opens when below the setpoint.


However,the loop reacts very slow. For instance, when the valve proceeds or ramps to a close when the temperature is to high, the temperature still rises until the effect of the valve closing comes into affect. The same happens when temperature is low. It has come to the stage where the control valve is practically just opening and closing in the loop without much control. Is there anyway to controlling such a senario. Would changing the scan rate of the PI help. Appreciate any help on this. Thanks

 

[curlyandshemp]

Normally what I do is first set the I to 0.
Gradually increase the P until the loop begins to oscillate, control valve will go to 100% to 0% to 100%. Divide the value of P by 2 and enter that.

Allow the loop to stablize, with P only the actual temperature and setpoint temperature should be close with an offset of a few degrees. Gradually introduce I, start with 0.001 and monitor how the loop behaves. Gradually increase the amount of I and monitor. Once the P and I appear to accomplish what you desired, change the setpoint to ensure the PI loop responds as desired.

Its a start

Ian

 

[RPax]

Nice one Ian, will get me started, thanks for this

 

[Tom Jenkins]

PI (or PID) control is poorly suited for systems with long response delays. It is very difficult to tune them to get reasonable response and simultaneously avoid hunting.

Look at feed forward control or else a floating control (like a thermostat) system. It will probably give tighter control and better stability.

 

[curlyandshemp]

Nice one Ian, will get me started, thanks for this

Also

http://www.controlguru.com/

this web site has some good stuff on PID

 

[Peter Nachtwey]

It isn't clear to me what is slow.

Is it the plant time constant(s), dead time or the PI gains are too small?

RPax needs to find the plant gain, time constants(s) and dead time. The rest will be easy then unless the dead time is large relative to the other plant time constants.

I also recommend studying the control guru suite.