Strange PID adjustment technique (Experts Only Please)

[Alaskan]

Hi

I've worked as a system integrator for a few years and have run across an interesting setup at a local plant. One of the electricians has been doing most of the programming over the years and he has an interesting technique of tuning his loops.

Essentially what he does is change the kp value depending on the current process value eg
kp=(pv-A)/B+C
where A and B are gain values and C is an offset

His value of ki is dependant on the value of kp e.g.
ki=kp/D+E

And for the derivative term he scales this depending on the value of ki e.g
if ki>F then kd = G
if ki<F then kd = H
where G and H are also constants

I have never seen this before and was wondering if any of you have and if you could comment on whether this idea has any merit at all. He also has other code where the PID values are different depending on whether the PV is higher or lower than the SP.
Personally I don't like it because it makes the loops harder to tune (since you have to think about how the PID parameters are going to change depending on where the pv is).
I could see this being useful if the process was inherently nonlinear but is it really necessary for lime addition into a tank to control pH? (simple, first order system with a time constant of about 3 min)
Is this some advanced control technique or just a lot of hot air?

Thanks in advance

 

[xnuke]

If you think pH control using lime is a simple first-order linear system, you are mistaken, since pH is the negative logarithm of the hydrogen ion concentration. Neutralization is a highly nonlinear process. The pH controller gains are often adapted based on the process variable and where it falls on the titration curve for the system. This adaptation works well as long as the titration curve is constant and the system flow is constant. If either of these are changing, the system becomes more difficult to control.

xnuke
"Live and act within the limit of your knowledge and keep expanding it to the limit of your life." Ayn Rand, Atlas Shrugged.
Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Alaskan]

The question was about PID tuning. Please keep on topic.

This form of parameter adjustment is being used for flow controllers, level controllers and pressure control loops. All of which are have parameters and gains that can be linearized around the operating point. Any further thoughts?

 

[xnuke]

I did keep on topic - exactly on the topic you described. My reply was about PID tuning for the system you indicated, "lime addition into a tank to control pH."

xnuke
"Live and act within the limit of your knowledge and keep expanding it to the limit of your life." Ayn Rand, Atlas Shrugged.
Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Skogsgurra]

Alaskan,

I'm afraid you blew it. That attitude is not working well for you.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[roydm]

I did a quick spreadsheet substituting values of pv from 0-100%. It doesnt make sense as the gain just keeps going up with pv.
If you substitute deviation for pv, kp increases on either side of setpoint which makes more sense.
Sometimes just PID is not the best control e.g. pH or in some levels where the gain in one direction is completely different to the gain in the other (I am thinking of float cell level control).
Its not a method I would use
Roy

 

[Alaskan]

Thanks Roydm, I had done the same simulations in excel and had come up with the same results. I'm trying to tie these loops together in a feed-back-trim arrangement for disturbance compensation and the changing PID values are giving me a headache.

 

[PNachtwey]

The method described has no basis is control theory. The formulas in the first post did not mention the plant parameters. Finding the plant gains and time constants is the most difficult part of auto tuning. There are many types of plants that are type 0, type 1 and perhaps type 2. These plants can have one or more poles and if there are two poles or more then some of the poles may be imaginary. Therefore, there is a different set of equations for calculating the gains for a plant depending on the plant transfer function. To make things worse, the system may be non linear. Then the gains and perhaps the time constants will change as a function of the PV. Heat exchangers, PH systems are non-linear but they can be tuned around a operating point or fixed SP assuming the system is linear at that point.

To start one should identify the plant transfer function Gp(s). Then one can compute the controller transfer function Gc(s) by using the equation
Gp(s)*Gc(s)/(1+Gp(s)*Gc(s))=CLTF(s) where CLTF is the desired closed loop transfer function. Now one can solve for Gc(s), the controller gains.

There has been no mention of what the plant is. Know that would help a lot, but like I said, the system identification is the hardest and the most important part of tuning. How can you control what you don't understand?