PIDE warnings rslogix5000

Hello,
I am trying to modify an existing program because the temperature is not stable (according to the operator it was never precise). So, I decided to use PIDE instruction along with a SRTP block to do the job. When I tried the autotune feature I received the following warning “Warning: Observed PV change was too small., Warning: Identified process gain is too small., Warning: Long deadtime process identified.”
I changed the CV Step size to different values (10, 20, 40& 80%) but I receive the same warnings and none of the slow, medium and fast response gave me a good result. This is my first time using the PIDE, I believe that I am doing something wrong. I attached the file. Thanks in advance for any inputs.

As a rule of thumb always put PID/PIDE tasks in a periodic task.

When you make a change in open loop measure the time it takes for the PV to react. This doesn't have to be super precise just get close. I take this time value and multiply by 20 and use that as my time interval for the periodic task.

It sounds like you are executing a slow process at too fast of a scan rate.

Keep in mind the autotune will get you close, you will most likely have to do some trending and tweaking to get it where it needs to be.

Good luck!!

Yes, you are right. The scan rate is too fast for this process, I was just trying to keep it as it was originally. Today I decided to create two new tasks. I created a 2000ms task for the PIDE and a 10ms task with higher priority for the SRTP to make it more accurate. In addition, I changed the SRTP Cycle time from 5 to 10 secs. When I tried the autotune again, I only received one warning “Warning: Observed PV change was too small.” should I set a higher setpoint and repeat the autotuning?
However, I will proceed to calculate the time interval for the PIDE periodic task as you suggest.

ELake20 said:

As a rule of thumb always put PID/PIDE tasks in a periodic task.

When you make a change in open loop measure the time it takes for the PV to react. This doesn't have to be super precise just get close. I take this time value and multiply by 20 and use that as my time interval for the periodic task.

It sounds like you are executing a slow process at too fast of a scan rate.

Keep in mind the autotune will get you close, you will most likely have to do some trending and tweaking to get it where it needs to be.

Good luck!!

Click to expand...

Did you mean divide by 20? If it takes 2 seconds to see a temp rise I can't imagine a 40 second loop time.

Maybe look at the IMC block??

robertmee said:

Did you mean divide by 20? If it takes 2 seconds to see a temp rise I can't imagine a 40 second loop time.

Click to expand...

Correct sorry..

You need to up the output so the autotune sees a 'noticeable' change in the PV

TimeToMove / 20 = Periodic task time

The time that took to react was 45sec approximately, so the periodic task should be around of 2.25sec. However, I repeated the autotuning using the 2sec task that I previously set and a higher setpoint. This time I didn’t receive any warnings from the PIDE. What I can see now is that the autotune gains are lower than my first test. I should probably repeat the test using a 2.25secs period task to see how this affect the gains.

You aren't really issuing a setpoint with autotune..you are setting the CV to a % and the instruction waits to see how PV reacts and then calculates the gains.

Use the autotune as a baseline to get you started.

You will most likely need to play with PI&D to get the desired functionality.

Building a trend screen will help with this and allow you to see the deviation.

Good luck!!

Would it be possible to attach an image of a trend chart showing CV and PV during the autotune? Just because the autotune status is "OK" does not necessarily mean the resulting model is good enough for tuning. With such a long time constant, this process could be vulnerable to disturbances during the test, thus weakening the model.

Well I didn’t set a higher setpoint, I kept the same SP but the ratio between the process value and the setpoint was bigger. For the picture with the “Warning: Observed PV change was too small” the SP was 176°F and the CV step size was 10% but the process value was 158°F. Less than 18°F between the SP and PV.
Then I repeated the autotuning with the same parameters SP 176°F and CV 10% but the PV was 66°F when I started the autotuning and I didn’t receive any warnings.

Sorry Mispeld I didn’t save it but if I repeat the autotuning I will post it.

Have your looked through the auto tune getting results guide?

Yes I have. I will continuing running some tests and I will post it

Per the document posted in reply #11, this auto-tuner is described as: "It provides a simple, open-loop Autotuner built right into the PIDE instruction." As such, it will execute one or more CV changes with the loop in manual (see also p 17 flow chart), and use the PV response to compute model parameters. Your best results will usually come if the CV step results in as much PV response as your process can tolerate. In this case you mention a starting PV of 66 and SP of 176. Ideally, you want the autotune CV step to move the process by 100 F over the full response time. One approach is to look at history for CV values when the PV was around 176 F. Let's say that is 75% (to randomly pick a number). If the CV that gives 66 F is 10%, then a good autotune step will be 65% (starting from 10%)
The PIDE autotune should not care what the SP is, because all of this is happening in manual. I suspect that the warning you were getting was because the starting and ending PV difference was too small, not the SP - PV difference. If you can get a 100 F change in PV during the autotune procedure, and the load on the system is typical, you will likely get a good model and resulting tuning.

I will add that if you are going to be "bold" with auto-tuning, or even manual step tuning, you will want to watch the PV directly or with logic to keep it from going dangerously high or low. In the example above, if 75% CV typically yields a PV around 176 F under normal system load, and you auto-tune under significantly different loading, you might get a much higher response temperature.
Best case is to manual- or auto-tune under normal, unvarying load conditions. Unfortunately this is not always possible or unacceptably costly, so compromises must be made.
One last thought: be patient. If your first order time constant really is 2800 sec, you will be waiting at least a few hours to reach 99% of the ultimate PV response. Cutting off early will underestimate process gain. If you have to do this, choose conservative tuning.

Mispeld said:

The PIDE autotune should not care what the SP is, because all of this is happening in manual.

Click to expand...

Correct the autotuner does not care what the SP is.
You must also raise the PV limit.