PID tuning methods

[OysterMan]

Hello,

I have somewhat minimal experience tuning PID systems. Ive done a small servo controlled camera gantry. In the process of tuning sometimes the camera would flail around wildly but it was okay due to how small and safe the system was. Im now faced with another system thats much larger and cant really be simulated easily so Im curious how PID is tuned for potentially dangerous systems. If I wanted to make a simulation it would take quite a while to set it up. Do you have any reading material that would help me learn about how the pros do it? Currently I know a little about how the values are related and how they affect each other but Im not quite experienced enough to apply it without the 'change this and see what happens method' which doesnt fly with something like a rocket or nuclear reactor lol.

[GaryS]

Here are a few links that may help you there are others you may want to check out


https://www.youtube.com/watch?v=qKy98Cbcltw

https://www.youtube.com/watch?v=NVLXCwc8HzM

https://www.youtube.com/watch?v=JFTJ2SS4xyA

[drbitboy]

Wow @ GaryS, those are great videos, and that last video by ElectroNoob is especially brilliant.


OP is going to have to either be very careful in the tuning as they note, or invest time in modeling. The latter is, to answer the OP's query, "how the pros do it," but will require significant effort to get good data and rigorous analysis to adequately characterize the system. The former is risky, but we almost always end up there. Actually, a simple conservative model might be useful to get a safe starting point for the former. Small moves, Ellie, small moves.


I can't think of other options, but that doesn't mean there aren't any.


TL;DR


As far as the mechanics of modeling, I found that AdvancedHMI is a simple and useful way to tie a simulation to a PLC control system. VB is a simple language; assuming one has the math and programming chops it is straightforward to make a model. I have also used simulation via Python as a another approach for the theory. As far as judging whether the model adequately matches the real process, I doubt that, short of a PO, anyone here can do more than encourage.




I am NOT* suggesting it here, but as an example, Zeigler-Nichols, is a form of characterizing (modeling) the process, and using that model to tune the PID. However, even if Z-N's model is close to reality, it's tuning goals (disturbance rejection) may be inconsistent with OP's (e.g. command tracking?). My point is that OP will need some kind of model to start. Computer CPU and programming is cheap; OP's time is not, but I suspect it is a lot cheaper than replacing a damaged process.



* Z-N is applicable to some processes (first-order and sluggish, IIRC) and not to others; we don't know enough about OP's process to say whether it would work here.

[rupej]

This is the most conclusive explanation of how to tune that I have found. This is a whole series that starts with the basics.

https://youtu.be/gZSEnzmaht8

[geniusintraining]

Originally Posted by OysterMan

so Im curious how PID is tuned for potentially dangerous systems

Very carefully.... sorry had too

Originally Posted by drbitboy

and that last video by ElectroNoob is especially brilliant.

I agree, I know what PIDs are and have been setting them up for years but I still watched that complete video... thinking about building something like his as a trainer

EDIT: Not sure what PLC your using or if a PLC but Tim has made some good videos for Allen Bradley PID's https://www.youtube.com/playlist?lis...8jYqlNbwMr5AwP