I have the following fairly simple system and want to install a PID loop to control it. Problem is I have had zero luck tuning the thing.

Layout -- A VFD controls a pump moving glycol through a system. A pressure transducer reads the pressure in the piping. We want the PID to control VFD speed to maintian 10 units of pressure.

Current Configuration of the PID shows below. PLC logic takes the PID CV % and multiplies by 327.67 and sends it to VFD after applying limits of 8000 low and 27000 high.

Any Ideas?

[attachmentid=4627]

[attachmentid=4628]

A couple hours playing with RSTune and it it works now.

The oscillation on the left of this first picture is what we where getting with the original tuning settings.
[attachmentid=4630]

This print is from 0 CV in manual switched into auto.
[attachmentid=4631]

This print is from 100 CV in manual switched into auto.
[attachmentid=4632]

This print is changing SP from 10 to 12 and back to 10 with PID in auto.
[attachmentid=4633]

Comments guru's. Did I get this one right?

Looks good to me. Can you post the before tuning parameters, just curious.

ken




It is displayed in the left bottom of rs tune in the image itself

That is still a long time.


If it meets your specificaitions then it is alright. It is certainly MUCH better. This is a good example of what a tuning package can do for you. There are still people on sci.engr.control that say that gains are tuned, not calculated. This is another example of how they are calculated.

I have never used a tuning software but am not opposed to it if it works. What do you think about this software? Is four hours typical to tune a loop with it? That seems extremely long. I couldnt imagine tuning a whole system at that rate.

Part of that time was learning curve for the dummy on the keyboard. Also this was a relatively complex process with lots of load upsets occurring during the attempted tuning. I like the software just for how it exposes the PID parameters and gives you a nice graph of perforance. Something not seen easily in RSLGX5,