The Fast Way to Tune Slow Loops

The Fast Way to Tune Slow Loops
The Fast Way to Tune Slow Loops

Tuning slow loops does not have to take a long time.

Using the strategy outlined in this presentation, you can tune slow loops very quickly.
To tune a control loop, analysis software must process some loop data. You must test or bump the loop somehow. For example a setpoint change in automatic or a controller output change in manual mode. For more information see the presentation: "Plant Data You Should Use for PID Tuning and Modeling" available from the articles page on ExperTune.com.

The trick we use for tuning slow loops, is a fast test. A fast test:

  • Is up to five times faster than a step test

  • Upsets the process little

  • Allows less time for outside upsets to interfere (with the test)

  • You don't need initial PID tuning

  • You don't need any process model estimates

  • You don't need to know the process dead time


FastTestStep1.gif

AbFastTestPulseCut2.gifove is an example step test performed on a slow temperature loop. The white line in the upper portion is the temperature. The green line in the lower portion is the controller output.
The controller output steps up at about 20 minutes. After about 1000 minutes the temperature is settled enough to provide good data for PID tuning and modeling. This is not a fast test.

To the left is the fast test performed on the same loop.

The temperature reached a temporarily settled area at about 140 minutes. This fast test data will result in the same PID tuning as using the step test data and it went 5 times faster.



Fast Test StepsFastTestSteps3.gif

  1. Check to be sure the controller output is between about 30% and 70%. If it is outside 30% and 70%, change the controller output to between this area if possible because valves are often non-linear at limits.

  2. Let the loop settle out and direct the analysis software to start archiving or recording data.

  3. Check to be sure the controller output is between about 30% and 70%. If it is outside 30% and 70%, change the controller output to between this area if possible because valves are often non-linear at limits.
  4. Let the loop settle out and direct the analysis software to start archiving or recording data.
  5. Put the controller in manual and be ready with a stopwatch or timer.

  6. Change the controller output quickly by 10% to 20% (This controller output change can be relatively large since we will be moving it back fairly soon.) As soon as you make the change, start timing.

  7. After a short amount of time or when you see the process variable move from the change you made, move the output in the opposite direction by twice as much as the change in the previous step. Also immediately record the time that has passed since the first change - call this time, T, and start timing again.

  8. As soon as time T has passed, return the output to its original value.

  9. Wait until the PV seems to settle from the changes made before using the data for tuning. The temperature here is not settled. You need to wait longer as shown in the next slide.

 
FastTestExa4.gifTo the left is the temperature loop test data after more time has gone by. You need to wait until the PV seems to settle or turn the corner from the pulses made - about at time 260 minutes in this case. The red region shows the time window of data to use for tuning. This window of data results in excellent PID tuning.

When you do a fast test you probably will not change the controller output at exactly a symmetrical interval - the first half will either be longer or shorter than the second half - and this is recommended. In the previous example, the second time period is slightly longer than the first. The shape of the response of the PV will depend on your process dynamics and the time durations of your pulse.

FastTestExb5.gif

In this fast test on the same temperature loop, the first portion of the controller output pulse is shorter. For tuning, you can stop taking data at about 180 minutes. The data shown in the red shaded area gives excellent results.
Again we have the same temperature loop.

This time the first half of the controller output pulse is longer than the second half. You only need to collect data in this case until about 130 minutes. This test, the previous one and a step test all give almost identical tuning parameters from ExperTune's PID tuning software.

FastTestExc6.gif

If your temperature loop responds differently up than down, it is best to take data in both directions and use the more conservative tuning. ExperTune's Advanced PID tuner includes a Loop Summary Table that lists the results of several tests, automatically picking the more conservative tuning for you.

Use the fast test combined with the analysis power of ExperTune software to quickly get PID tuning on slow loops. From this data ExperTune's Tuner/analyzer also gives you a high fidelity model of the loop, simulation, and robustness analysis.

See the response in simulation and know how robust the loop is before you download tuning.

© 2002 ExperTune Inc.
 


Did you enjoy this great article?

Check out our free e-newsletters to read more great articles..

Subscribe