embarrassing question regarding PID tuning

[ganutenator]

I haven't had to tune a PID loop in a very long time.
It's actually a PI loop for a pulse width modulation s.v.
What was the name of that tuning method?
Zieglar or something w/ the quarter response curve blah blah blah or something?
Am I supposed to graph the error.
Sorry, I forgot. I'm old.

[ganutenator]

I think that I may be able to do it right on the plc output card as it is solid state, the amps are correct, and the process engineer is stating that he needs something like a 2 to 4 Hz.
And the card states a maximum of 1kHz switching frequency.

[ganutenator]

Obviously, there are still a lot of questions, and this project is in infant stage.
I've programmed a lot of PID loops in the oil fields back in the 2000's but I suck at tuning loops. I had starting numbers to plug in at that time.
This is different, as it will be my first PWM control.

[JLand]

I believe you are thinking of Ziegler-Nichols:

https://en.wikipedia.org/wiki/Ziegle...Nichols_method

And I think you are talking about quarter amplitude dampening. QAD is what the Z-N algorithm attempts to achieve. It means that a process variable oscillates around the setpoint where a given peak is a quarter the amplitude of the peak preceding it (peak to peak or trough to trough, not peak to trough or vice-versa).

https://blog.opticontrols.com/archives/1066

[JLand]

You can graph whatever you find easiest to graph, but Z-N method has some graphical elements to the calculation (which you can do analytically) that relies on you graphing your setpoint and process variable with respect to time.

If you are trying to tune a highly aggressive loop, you could consider Cohen-Coon:
https://blog.opticontrols.com/archives/383

If you cannot overshoot, you can use IMC. I've never used it though, so I can't comment on it. In my relatively limited experience, Z-N and C-C get you a good starting point.

https://blog.opticontrols.com/archives/260

[ganutenator]

thank you very much JLand.
So, If I graph the pv and the sp and the error, should I look at the quarter amplitude dampening on the error?

[ganutenator]

p.s. I love Wisconsin.
Everyone is so friendly there.
I once was on a job and went to the gas station.
I slapped my credit card on the counter after I filled up my rental car w/ gas.
The lady said: "We don't take credit cards"
I said: "Oh no, where is the nearest ATM?"
she said: "A T what?"
then: "Just come back when you have the money"
Even the TSA were polite.

[JLand]

You can look at the quarter amplitude dampening on the error, yes.

Because error is just PV-SP, you really don't need to graph your error, you just look at how the PV oscillates around the SP. It gets you the same thing as graphing the error, just with the x-axis shifted from zero to the SP.

It's really up to you on how you find it easiest to visualize a PID loop. Looking for QAD on the error about y=0 or QAD on the PV about y=SP, it doesn't matter.

That said, you need to use your engineering judgement to decide if QAD is even desirable. If the setpoint is close to the limits of the machine, or if you desire rock steady values at the expense of response time, QAD is inappropriate to target. QAD is just a solid heuristic to target with Z-N and C-C tuning methods for when you want to hit your SP quick.

[JLand]

Hah - thanks for the nice words about WI. As a transplant, I like it here and have noticed similar.

[ganutenator]

One of these days I am going to go back and pay her. jk

[KuulKuum]

Originally Posted by ganutenator

One of these days I am going to go back and pay her. jk

With the current high price of gas and if she applies interest thrown in for the time past... you probably will not have the funds to repay... LGB

[ganutenator]

I actually paid the same day. But I appreciate analytical brained people w/ a good sense of humor. Yes, gas prices are insane.

[Peter Nachtwey]

All info on Z-N should be scrubbed from all sources. I would look at the www.controlguru.com site for how to tune systems.
The control guru pushes IMC tuning with is essentially pole cancelation.
I think pole placement is better but more complicated. Ron Beaufort once called that lambda tuning. Same thing.
In the end, it is worth putting in the extra effort to tune a system/plant right because the payoff is forever as opposed a few extra hours of work
BTW, you don't tune a PID. You tune a system/plant.
You haven't told us what kind of system/plant you are tuning. Shame, you should know better.
I am retired but not retarded. Age is no excuse.

[Puddle]

Originally Posted by ganutenator

gas prices are insane.

Petrol station at the end of my street is currently showing £1.90 per litre, which is $8.82 for a US gallon. This time last year it was around £1.30.



On topic question, which PLC is it? I've got very lazy with using auto-tune, which isn't always perfect but it always gets the settings into the right ballpark and I can adjust from there.

[ganutenator]

It is a wago 750-862.
The process:
We are currently monitoring flow rate of a chemical that is being introduced into the water for a car wash application.
Customer wants to be able to set the flow rate from his office chair instead of driving to the car wash.
Customer knows that he can buy a device to do this, but wants to do this on the cheap.
So we think that we can buy a solenoid valve that accepts pulse width modulation to control the flow and do this with the output card of the wago (750-1500).
We are still awaiting info from the valve manufacturer.