PID tuning

initially how to tune a PID for temp control. is there any relation among P, I & D while tuning.

Regards
Manasi

manasi said:

initially how to tune a PID for temp control. is there any relation among P, I & D while tuning.

Regards
Manasi

Click to expand...

Short answer yes. Each can affect the other.
I suggest you do a search on P.I.D. tutorials. There are several. Look for those with the easy math. You'll see what I mean when you look into it.

I would do a search at this forum first, if that doesn't pan out, try Google.

You do not always need P.I.D. depends on the system, and amount of control and response needed. Most tines you need the P and I, on a dynamic system you will most likely need all 3.

manasi said:

initially how to tune a PID for temp control. is there any relation among P, I & D while tuning.

Regards
Manasi

Click to expand...

Hello manasi;
Since you seem to be in your first steps in PID tuning, I would recommend that you search on the posts presented in this forum (and on his own site) by a gentleman named Ron Beaufort.(Look for the SEARCH link at the top of the page). He especially has 3 linked articles that you should look at: What is "P" in PID, and same for "I" and "D". They offer a very good, clear and low-math initiation into the principles and application of the PID control principles; they also explain the interactions of the different parameters on the control output, and on each other.
You would be much better prepared to implement PID in your project and future ones.
Hope this helps,
Daniel Chartier

Edit: here is a link to "P in PID", it containk the required links to "I in PID" and "D in PID". Good reading!
http://www.plctalk.net/qanda/showthread.php?t=13459

Manasi,

While I am by no means an expert with PID control I have found that temperature is one of the easiest processes to control. That is, compared to others I have had to do such as chlorine, pH, and ozone. I pretty much never use Derivative and normally use a small amount of Integral with the Proportional being my main adjustment.

manasi said:

initially how to tune a PID for temp control. is there any relation among P, I & D while tuning.

Click to expand...

Yes, auto tuning programs have worked out the math that calculates the gains after analyzing there the response of the system to excitation.

Every system is different so what works for one system may not work for another system. I would go to www.controlguru.com and read the part about tuning a heat exchanger.

Actually flow loops are the easiest to tune because the response to a change is almost instantaneous whereas temperature can have lots of lag. As rta53 says D is the most difficult term to utilize successfully.
I suggest you forget the math and do a search for Ziegler and Nichols, their tuning method is quite easy to understand.
Roy

Roy Matson said:

As rta53 says D is the most difficult term to utilize successfully.

Click to expand...

That is because:
1. some systems don't require derivative gain mathematically. Do you know which ones?
2. The feed back resolution plays a big part. Most of the noise people complain about is really due to quantizing.
3. Sample jitter. PLCs do not have a constant scan time and even the interrupts have a lot of jitter. Sample jitter results in poor rate estimations.

I suggest you forget the math and do a search for Ziegler and Nichols, their tuning method is quite easy to understand.
Roy

Click to expand...

ZN results in poor response that over shoots and the oscillates around the set point a few times. One can do much better with just a little effort. The methods on the www.controlguru.com are much better than ZN.
Once you understand the math you will never see PID tuning the same.

I would make one guess: In flow controll you should not use derivative term?

PID is the best technique bUt however P,PI.PD can be used in some processes with great results.these parameters rely on each other especially in critical loops

Hi, Manshi,

I have some experience with temp. loops, try the following, hope you will get good result. but before that I need to understand what is your temp. source, ( thermic fluid, heat exchanger, electric heat generator Star/delta) etc.

keep, i & d =0.

put p (kp) to a little high, try to check three cycle, if required put some I, to adjust. you might not need D, use dead band to stabilize ( if your process permit.

Thanks Bhaskar

TurpoUrpo said:

I would make one guess: In flow controll you should not use derivative term?

Click to expand...

You are right. The general rule of thumb is that one gain is required for every pole in a system. Do you know what a pole is?

PID is the best technique bUt however P,PI.PD can be used in some processes with great results.these parameters rely on each other especially in critical loops

Click to expand...

PID isn't always the best technique. It really does depend on the application. Ki is usually required on type 1 systems where reaching the set point exactly isn't required. Kd isn't required in most speed control applications. There is a mathematical justification for these statements.

I have some experience with temp. loops, try the following, hope you will get good result. but before that I need to understand what is your temp. source, ( thermic fluid, heat exchanger, electric heat generator Star/delta) etc.

keep, i & d =0.

put p (kp) to a little high, try to check three cycle, if required put some I, to adjust. you might not need D, use dead band to stabilize ( if your process permit.

Click to expand...

Superstition without mathematical justification. A temperature control system has dead time because it takes time for heat to move. It has one time constant which is heating the thermal mass or the load. There may be another time constant and that is the time it takes the temperature sensor to respond to the temperature of the mass being heated. A type 0, temperature or velocity control system, requires a integrator. If there is a time constant due to the thermal mass. This requires a P gain. If there is also a time constant due to the temperature then a another gain is required, the derivative gain.

No i do not know what a pole is (google does not give anything i recognice to explain what pole is). But i believe you'll explain it to me.

That flow controll thing i knew only becouse my teacher some six years ago once told us that we should not use D-term in flow controll. Altough i cant remember why that was. At only 23 years i have lot to learn


Thanks in advance.

Hopefully this will explain

http://en.wikibooks.org/wiki/Control_Systems/Poles_and_Zeros
http://virtual.cvut.cz/dynlabmodules/ihtml/dynlabmodules/syscontrol/node18.html

If not then I will try tomorrow.

I agree on ziegler nichols tuning methods, the initial parameters that are calculated via the Z-N tables are really just a starting point but will most probably require some fine tuning.
Just a comment on Derivative and Flow control, i believe the reason for not using derivative is that with flow you may get some spikes and fluctuation in the signal and due to derivative term operating on the "rate of change" of the error this would generate derivative "jerk".

Derivative jerk is not a problem if the derivative gain only acts on changes of the process variable.

ZN has no mathematical proof. The formulas for calculating the gains are derived empirically.