Host Engineering Forum
General Category => General Discussion => Topic started by: jrayb on February 05, 2021, 11:42:34 PM
-
I have a plastic film winder that starts out with a 3" OD core. I am using BX-DM1E-M-D. I am using a potentiometer as my SP & a dancer as the PV. I am using a PID loop to control the VFD speed. My issue is that the dancer is "hunting". It gets worse as the roll diameter increases. It's not bad enough that we can't effectively use the winder, but I can see it & it's not "right". My accell & decel times are set at 0.1. I have attached a screen shot of the pid loop after auto tuning. Any help or a push in the right direction would be great!!
-
1) If you're oscillating, then less P, right? (In fact this may be confirmed by the problem worsening with increasing roll buildup, which increases process gain)
2) Make sure your dancer accumulates enough material to translate into adequate response time. Dancer accumulation makes ALL the difference!
3) Start without any I at all, till you get the system to respond smoothly.
4) Unless your name is Otto, do your own work.
-
5) Check and make sure nothing is making the dancer jump mechanically other than web speed.
-
I played around with the simulator I "built" found here:
https://forum.hosteng.com/index.php?topic=1925.msg16183#msg16183 (https://forum.hosteng.com/index.php?topic=1925.msg16183#msg16183)
Granted, I have improved on that version a little over time, but it's still crude as far as figuring out where to stuff values for the simulator. I have a C-more Micro that I would usually use to access the simulator, but don't have it on hand.
Anyway,
Those slow swings look like either integral time is too short, or you have a diameter calculator that's hunting (tail wagging the dog.)
I would set Rate(D) to zero until it was proven it actually helps. Usually it hurts on a winder because the dynamics change too much as the roll builds.
If you have a diameter calculator, try to see if freezing it for short periods allows the PID to stabilize. That would be you doing this manually and watching, not some code trickery. If it does, you need to find a way to make it less responsive, yet not trail behind too much, especially at core.
If it's not a diameter calculation issue, increase the Reset(I) time. If that drops the amplitude of the oscillations, but increases the period, that's probably indicating it's contributing too much too quickly.
Ideally, you want the Gain(P) to be the main contributor to the trim. With I and D at zero, increase it until it gets unstable (these will be very quick oscillations) and then decrease it to stable and then decrease a little more. Then add I back to bring the dancer to setpoint. But start with larger values, say 20 sec and come down.
Then there is feedforward, which is simply
(Line Speed %) * (Core Diameter Inches) / (Actual Diameter Inches) = Spindle Speed Reference %
where 90% ref to the Spindle gives a core surface speed that matches Line Speed.
Now you add the PID output to the line speed ref and multiply by Core Dia / Act Dia. This keeps the PID from getting to much authority as the diameter builds.
-
Thanks guys!!! This is my first winder "rebuild" & the only thing that's kicking my tail is getting the dancer to stabalize. I will give these things a shot.
-
Hello, jrayb
I work at a thermfoming facilty winding plastic film out of an extruder as well as winding corrugated paper
When I make a program for a winder I dont use a PID loop
I program it to follow a torque sensor or a encoder on the previous roller to match length of product
Does that help