Hello,

i am a student and currently doing an experiment on p-only controller for the flow control system. so i came across a few result. The experiment is done by using multiple PB(P)% value, which is 999%(max), 200%, and 100% . This will gives Kc(gain) value of nearly zero gain, 0.5 gain and 1 gain. the controller used is yokogawa controller using DCS (yokogawa CENTUM CS 3000 R3). and the plant is solteq model: se 113 experimental stand for ratio control. here I'll also upload the result.

so the question is, is my result correct? based on theory my results seem so wrong. the response never stabilize. it always under constant cycling for the PB value of 200% and 100%. theoretically the response will stable but with an offset for proportional control so i hoping some of you can help me...tq..

pb999 ->http://oi68.tinypic.com/2mhu8wg.jpg
pb200 ->http://oi65.tinypic.com/2r6cjs9.jpg
pb100 ->http://oi68.tinypic.com/2lv1oa9.jpg

 

With only proportional control, response will be oscillating type.
You please go thru the theory, response of 2nd order system with P control.

results are correct.

 

Assuming this was an assignment, your instructor is trying to show you that proportional control alone does not usually work. To stabilize the control you need some integral action. Think about what you are doing with proportional only control - as the error approaches zero, the controller output will decrease, which will, in turn, increase the error!

 

"...the response never stabilize...

I had a difficult time understanding the specifics of your question. However, I can tell you that a proportional-only control might well respond that way. "Never stabilizes" is why we have the I & D parts of PID. You might try making your Kp much smaller, so that it at least gets closer to your desired output, but it's pretty rare that a proportional-only gain will be particularly accurate.

-James Ingraham
Sage Automation, Inc.

 

Hello w4n0b1366,

I acknowledge you are expecting an offset with Proportional-only control.

I am interpreting your question that you are expecting that oscillatory behaviour will occur for this type of plant, though you are theoretically expecting the oscillations to typically decay in the closed-loop step response?

Try repeating your experiments including various other gains. As part of your investigation on decaying oscillations.

I have not sought to directly answer your question; nor reconcile your implementation.

Kind regards,
Flash

 

thank you for your reply..

i do continue to decrease the PB%(increase Kc gain) and the result are:
at pb 95% ->http://oi67.tinypic.com/1e14ow.jpg
at pb 90% ->http://oi67.tinypic.com/2nu681t.jpg
at pb 85% ->http://oi67.tinypic.com/xc57qt.jpg
at pb 80% ->http://oi67.tinypic.com/2qx2bv4.jpg

the whole experiment is to determine the PI parameter for the PI control system using closed loop(continuous cycling) technique.
so the proportional part is to determine the PB ultimate and T ultimate from the PB% with min 5 perfect cycle..so at 80% pb the result is pretty much got 5 min same amplitude:

result->http://oi64.tinypic.com/j63q00.jpg

so the PBultimate value is 80% and T ultimate is 62s and the pid value as new PB for PI control is PBnew=PBultimate/0.45=177.777%
and Tnew=Tultimate/1.2=51.6667...(ref ziegler & nichols method).for->PID value:
http://oi63.tinypic.com/xcofaw.jpg

and at the end the result is ->http://oi67.tinypic.com/r950rt.jpg. the result is not as expected...so..i just want to investigate and find the cause of the problem.

 

Hello w4n0b1366,

Have you resolved your original concerns about reconciling Proportional-only control (theory and results)? What did you conclude?

Kind regards,
Flash

 

> With only proportional control, response will be oscillating type.
> You please go thru the theory, response of 2nd order system with P control.

Not necessarily.
See this.
http://ctms.engin.umich.edu/CTMS/index.php?example=Introduction§ion=ControlP
What value of Kp results in a critically damped response?

 

Dear Flash

> Have you resolved your original concerns about reconciling
> Proportional-only control (theory and results)? What did you
> conclude?

in my early conclusion, i can say that the end PI tuning control response may be caused by the P-only control that have very low gain for stability. Or seems that the output at low gain does not being effected by the error change, so rather than oscillate the output, it does not produce any result...e.g.PB% at 999...Kc=nearly zero. now i'am still trying to resolve the P-only response to make it stable at a decent value of gain (0.5~1). theoretically the response should be oscillating when the the too much Kc and moves toward perfect/stable oscillation. i'm still trying to figure the other tuning parameter (not just P,I,and D) for the yokogawa controller itself, since i'm not familiar with these practical PID controller (only learn theoretical part before).

so, maybe other people that are professional in these area of work can guide/point the mistake. it would truly appreciate it and my i ask if anything you want to point out or your opinion Flash..

Thanks

 

>What value of Kp results in a critically damped response?

thank you for your reply.

well that is exactly that is the problem i assume.. i never actually found the critically damped value..it always at over-damped at 999% PB

-> http://oi68.tinypic.com/2mhu8wg.jpg

if you assume that is kind of over-damped..and never found the critically damped gain ..and next is the under-damped that constantly cycling e.g at

PB%300-> http://oi66.tinypic.com/10zzkoj.jpg
PB%400-> http://oi67.tinypic.com/42w5c.jpg
for other value you can refer previous post...

as the p value is inserted and put in auto mode. it start to oscillate no mater what the gain/PB value inserted. the difference is only that the smaller the PB value(larger gain) the larger the time taken for one complete oscillation..

this experiment is done for the flow control (water flow control). and is it possible that these type of response always oscillate? that's mean no critically dumped value at all...or there is something wrong in this experiment?

thank you for your time..

 

Dear w4n0b1366,

I am unsure what you are theoretically expecting with Proportional-only closed loop feedback under various situations; and whether you have some misconceptions on the theoretical outcomes.

I suggest you go back through your notes, as well as seek out additional information. It may also help to seek out typical Proportional-only feedback responses (under a variety of situations, not just ideal cases); that will assist in recognising and interpreting your results.

Furthermore, I suggest you experiment a lot more with the Proportional-only setup, to better appreciate how it responds under a variety of situations; and what it can offer (plus its limitations).

Whilst the insights can be many, I hope you will observe this process with damped partial-stability; critical stability [approximately], and instability. You can then better interpret what your test results are telling you; as well as identify the causes that lead to those responses. Have you considered broader experiments investigating gain? It is not the only factor you will need to experiment with.

In respect to your related attempts to do Zeigler-Nicholes Closed-Loop tuning, P-only experimentation:
Carefully go back through your notes. You may want to seek out additional sources, to deepen your understanding and complement your technique. Whilst you may have a general sense of this trial-and-error method, carefully consider some of its subtleties.

<i>Not all sustained periodic oscillations are created equal</i>

As this is part of your education, I have only provided general nudges; rather than directly answer the questions for you.

Kind regards,
Flash

 

i've never got the critically damped response, only under-damped and over-damped. the response oscillate through all the PB/gain value. it only changes the period of oscillation based on the value of PB/gain.

 

thank you for your reply..

my expectation is actually based on ideal Proportional control response, and now i'll try to go through all the note and theory as well as experimenting all other variable. i do notice the lack of my knowledge here. i'll update on progress and ask for opinion.

thank you for your help

 

First, in that example do you know what value of Kp results in a critically damped response?

Do you have any idea what the open loop transfer function is like? The example on the CTM website is a two pole system over damped system. You should be able to find a value for Kp that is stable for a single pole system or a double pole over damped system. Unfortunately too many systems are tuned this way

Something is wrong. If you adjust the DCS flow signal manually will the flow change instantly? Is there a dead time? Does the flow oscillate around the final value? You can't control what you don't understand unless you are lucky or are willing to do a lot of trial and error.