PID control - why is the overshoot so long?

[KF_]

Concerning the PID-values, I have a program which minimizes the settling time and gives very nice response. My program only works with first and second order transfer function with or without time delay though. I would be happy to run the program for you if you give me a first or second order function.

Regards

KF_

[adamkse]

My PID values are something like:

Kc= 36

Ti = 0.078

Td= 0.012 

I tried the Ziegler Nichols method but could not get it to work.

Another problem is that I cannot derive a plant transfer function because I do not have enough information about the system so I can not give you one. I tried using the system identification toolkit but I could not get an answer that was correct, I was most likely using it wrong, I tried both the reference and control signal as the stimulus but the results were not even close.

I cannot post the VI until Monday but I will let you all have a look then.

Thanks for all the help so far.

Adam 

[KF_]

Hi Adam,

I don't know exactly what you mean by reference signal, but an open loop response is what is needed to calculate good PID-settings. If you can provide open loop response output data with a step input on the control signal (20 or 30% difference), I can help.

Best regards

KF_

[Andy_C.]

Hey adamkse,

I'm not sure what you meant by you cannot get a good response from system identification toolkit. You just need to provide reference signal and the response from the system to sys id. The algorithm will then provide an estimate model.

[adamkse]

KF. My reference signal is my set point or desired response. My plant is open loop unstable so it goes to infinity (saturation) in around 1-2 seconds.

Andy. I did wire the reference and outputs to the block but it gives me terrible answers. I put them in matlab and it does not give me a result anything like I would expect. I have tried many different orders but no luck. 

[adamkse]

KF. My reference signal is my set point or desired output. My plant is open loop unstable so it goes to infinity (saturation) in around 1-2 seconds.

Andy. I did wire the reference and outputs to the block but it gives me terrible answers. I put them in matlab and it does not give me a result anything like I would expect. I have tried many different orders but no luck. 

[KF_]

Hi Adam,

I can also determine good PID-settings if your open loop plant is a simple integrator. Does your plant have time delay?

Best regards

KF_

[adamkse]

I do not think it is a simple integrator. I have a servo motor in the plant. Am I right to say that servos are technically closed loop systems?

[KF_]

Hi Adam,

If you cant open the loop and you dont know the open loop transfer function, you have little options to tune the PID. You have to use the Ziegler Nichols rules posted earlier, you can determine the ultimate gain by making the plant oscillate or use autotuning, where you induce a small oscillation by a relay signal, but in the end you also have to use the ZN tuning rules. I am not sure if they apply on an unstable plant, I think they do apply on an integrator.

Best regards

KF_

[Soroush]

Unfortunately, I do not check this forum very frequently. In my last reply, I did not mean to comment on PID tuning. My main point was about your PWM converter. If you have not solved the problem yet, I strongly suggest that you work on your PWM converter.

Even your initial graphs are showing a reasonable tuning parameters, but it seems to me that you are quantizing the output of PID into resolution limited pulse lengths. So, I think increasing this resolution could help.

Soroush