Tuning PID controllers?

Francisco,

There are soooo many things that need to be known about the process before somene can give you a reasonable starting point.

One of the most important things is - how long before a change in the heat source output makes a change in the medium who's temperature is being controlled.

You can play around with the Kp, Ki, Kd and gain all day long, but without some awareness of how the process is behaving, you could spend a very long time getting the process tuned.

Mike

Auto-tune system applied in a high loop gain circuit causes more problem than coming to the optimum controlled set point.

First set controller to manual, and the Proportional Band to 100% or Gain 1, Reset (Integral) to 1 minute and cut off the Derivative (Pre-act). Try Foxboro's quarter amplitude method. However on manual control, the system has difficulty to stabalize, check the control valve. Good chance this is oversized. If you are using steam to heat the process, then lowering the inlet steam pressure helps to reduce the heat transfer reaction rate.

Many process control system designers oversize the control valve. If your heating system is electrical heaters, and if by reducing the supply voltage to the firing tyristors helps to stablize the overshoot and oscillation, then it is oversizing of the heating elements. There are two waveforms: oscillating and hunting. You can reduce the gain if oscillation starts, and increase it for hunting. To reduce offset, reduce the Reset time.

There is one common oversight by control operators, and that is reset windup. If a controller is laid to rest with a large offset, the system is not running, the reset will build up a high signal due to the constant offset. When the system starts, this has to be discharged with the measurement above the set point, therefore it is also good practice to start the system on manual.

Tijit

There are software that do that automatically nowadays. Just push a button and the thing manipulates the controls and calculates the PID values based on the reaction of the process.

If you want to read, there used to be a book titled "PID without the Math". I haven't read it myself but hear that it's a good book on the subject. Just Google for it.

If you want to learn intuitive tuning, it's going to take a lot of practice. You may be limited in the types of loops that you may practice on, however. I wouldn't practice on the controls of a nuclear power plant, that's for sure !

My early attempts was to get a PID controller and hook it up to a temperature sensor, a soldering iron and, if I had old paper, a recorder. I'd put the sensor on the iron try to control its temperature. It was crude and took a long time but it taught me the basics.

When I learned how to program PLCs and HMI, I developed a program that simulated different processes. Having the HMI meant that I didn't need a recorder since the HMI can be configured to produce a trend.

It's taken years but now, I'm the go-to guy when it comes to tuning (the company's a little tight when it comes to money, won't buy any tuning software).

Hello Francisco,

If you have a very fast system, and there is no danger in playing with it you could tune it by the ziegler nichols method. You can find it in the literature and even at the Wikipedia, under PID control.

With that orientation you could even start tuning it on the feel.

But if your system is not that fast and the system can be easily broken, or it is too expensive to be playing with it then you must model your aplication on the basis of input output.

Most temperature applications I have been working on are mainly first order plants.

Then you should identify the parameters of your model. Once you have done this you should be able to control it.

If it is your first PID controller, forget about Derivative with proportional and Integrative it should be ok.

With the Proportional gain you control the speed of your actuation, and with the Integral you get a zero permanet error.

Good luck!

Put in 20% prop band (gain 5) and 30 to 40 secs integral then suck it and see. Turn D term off until you have a feeling for responses. If your controller has auto tune give it a go as long as you can keep process under control.

I 've been using these figures as a starting point for many years on vastly different applications.

Another way is the recommended method where you start with no I or D and wind down the P (wind up the gain) until a slight change in setpoint causes oscilation.You can then adjust the P,I and D accordiing to the transient function - I can give you more details if you want.

I agree...leave I and D out of it until you get P working somewhat. Then fool around with D to correct overshoot...and as the overshoot is brought under control you can increase the P a bit more.

In fast response (i.e. small DC motor response) I've experienced that I needs to stay very small or it can cause instability...I can't really say how "playing with it" would effect a slower response system like heating.

Guest

Go to omega.com and order a free temperature handbook. It has some very good info in the back for tuning p.i.d. controller. They also have free telephone tech support.

congradulations,this is the first commentary that I have read where the subject matter stayed on course. with some fairly good suggestions to try. remember the KISS method keep it simple stupid! slow the process down and change your numbers slowly and have patience let the controller work .I've found that some time we try to keep the parameters to tight and cause the controller to hunt . sometimes while working on a system you have to lie to the process from the sensors, this usually means calibration ,we could go on and on .have patience and don't let it get the best of you. good luck

Have to agree with you sloebear, there are a lot of people asking questions on this site who do not seem to have a clue about engineering or any common sense - a major shortfall nowadays!

In the early 90's, an electronics engineer and I developed an electro-pheumatic controller at Leslie Controls, Tampa, Fla. that took most of the wierd PID stuff out of the tuners hands. By the second generation in the mid 90's, we refined it to the point where if you can tune a radio, you can tune this controller. IT IS DOWNRIGHT SIMPLE. You can take someone off the street and have them trained in about an hour.

I have since moved on and am not connected with the company and the Design Engineer is no longer with us but his patented concept is. (Hi Mahesh where ever you are) At that time, late 90's, Sealift Command had approved and was installing up to five systems each on many of its vessels as they came due for refitment. They use some of them to successfully control engine coolant temperature. You can imagine the varying conditions. Ask about Leslie PMC I controller. (Try one of their steam valves too. They supply catapult valves to the Navy, at least for now until they switch to magnetic motors) Tell um Bob sent ya.

Like always... it depends on your budget... I know a software which calculates the parameters for a P/PI/PID/PID-T1 automatically. The name is "ecICP" form a German software developer called "ExpertControl". We use it for hydraulic systems with cascade controllers.

http://openpr.com/news/78392/ecICP-from-ExpertControl-for-PID-controllers-and-beyond-completely-automatic-parameterization-based-on-measured-data-in-less-than-4-seconds.html

cheers, honno