Intermittent unwinder

jzhvymetal

Member
Join Date
Jul 2011
Location
chicago
Posts
3
I need some suggestion for calculations for an intermittent unwinder. The system consists of a servo controlled nip roller that moves film to a predetermined length and time. Then the nip roller delays for the remaining process on the machine. During the machine cycle the unwinder should continue to unwind if the machine gets consecutive machine cycles. Since the nip roller does not continuously index through the machine cycle I need a way to allow the unwind roll to slow down but not stop.

The film goes through a web with a dancer that has a senor to measure the position. The unwind roll diameter is being calculated via an ultrasonic sensor. Unlike a continuous unwind system I want the dancer to move up and down based on predetermined distance.


I would like to come up with equations that can calculate the speed required for VFD controlling the unwind roll. I know a PID loop is also required to compensate for error caused by slip, diameter error, and registration correction in the index.
I need some suggestions how to implement it.


I can calculate and have the following:

Nip Motion Profile: Triangular
Nip Index Time
Nip Index Length
Nip Expected Dwell
Machine Cycle Time= Nip Index Time + Nip Expected Dwell
Nip to Unwinder Ratio
Nip to Dancer movement Ratio

Dancer Movement Range Up/Down

Unwind Max Speed
Unwind Min Speed

Does anyone have experience in similar systems? [FONT=&quot]Any suggestions are welcome.[/FONT]


Thanks
 
I have done an unwinder very close to what you describe. I did it with a dancer arm & a VFD. I did a PID with the idle point of the dancer as a setpoint. The further the dancer moves from the idle position the faster the VFD would go (P). The PID loop had high gains on the P & the D and almost no I (actually I think I had no I). The high D was needed to kick the VFD in the Behind and speed up quickly when the film started to index & the same slowing down when the index stopped. The other reason I needed the high D was that the dancer arm/loop was too short for the amount of film I had to unwind. Putting in the high D allowed me to over come this mechanical design shortcoming.

Playing with the P & D gains should allow you to tune the loop so that the unwind will continue to run between indexes and still not pull the dancer tight at the midpoint of the index.

It was a passive system (it did not get feedback from the index servo), it only relied on the dancer.

The system had different characteristics with a full roll of material and a nearly empth roll. I was able to tune the PID so it worked at both ends of the spectrum.
 
bimini3,

Thanks for the replay. The problem with only PID control is the cycle time, index time and index length are variable in the machine. So when the operator changes them the PID will need to be tuned because the system will behave differently. I was hoping to use the additional index information to make it an active system that allows a smaller error correction. Then the PID is only correcting deviations in calculations, Nip roller slip , or registration corrections.
 
its just maths

(Unwind is a center unwind not surface unwind ???)

Nip Index length = unwind distance (sorry if i state the obvious)

Average Unwind Speed [mm/s] = index length / Cycle time

Use the unwind diameter and gearing to determine the unwind speed


Corrections / things that can go wrong

for Long index's and short dancer wraps you will need to change the speed
easiest way to measure is have the dancer at its loosest position with the film still tight (Unwind halted)
Jog the Nip roll (measuring distance) until the film is tight
This distance is absolute worst differential position permitted between the unwind and the nip roll (being an engineer you will design to use only some smaller % of this because you know murphy will appear)

You now have two index profiles that you have to connect
the Fast Nip move and the unwind move
Restriction that spring to mind include
"Differential position"
Unwind accel /decel rate

Advanced - Model where dancer should be within the cycle and use the (predicted - actual) error as a controller trim to the VFD speed
 
In my experiance, in 2012, dancers are a thing of the past. A high resolution encoder is much more accurate, and it doesent rely on a tension or anything but actual shaft speed. Just my thoughts
 
arkyny, that works well for sections of known and fixed diameter (print sections, draw sections, dryer drums, etc). But if you are dealing with a varying diameter over time, either long term from loss of material or short term from non-concentricity, accurately knowing the shaft speed is not enough.

jzhvymetal also has a case where he can optomize his drive system size if he doesn't have to accelerate his product roll at the same rate as his pull roll. His pull roll can use a relatively high index accel and speed to get product into the downstream process but the unwind doesn't need to achieve either the accel rate or the top speed. In that case his dancer is doubling as an accumulator.

Keith
 
I wouldn't involve PID with this, that would overcompliate it.

You know the three most important things

1) the rate you are drawing the material in with your nips.
2) the dancer position.
3) the diameter of your roll.

The average rate you are drawing the material in is simply the index distance divided by the cycle time (time between the starts of each index).

That means the average rate you need to unwind (assuming center unwind since you are measuring diameter) is the average surface speed you calculated above times 2Pi divided by your measured radius which will give you the RPMs you need to rotate the unwind at. (obviously you need to keep track of your units).

Now, just set up a limit control with a band.
On your dancer choose 4 spots.
Lowest = Declerate to a stop (material not being drawn)
Low = If you go below this setpoint send the core the LOW speed signal
High = If you go above this setpoint send the core the HIGH speed signal
Highest = Fault, unwind not keeping up.

Your LOW and HIGH speed signal should correspond to +/- some percentage of your average speed. I would target something comfortable like 10 to 15%. Actually, I often bias more toward the fast side. Therefore the slow speed is minus just a few % and the high side is maybe plus 15%. This favors a faster "catch up" time and a nice slow drop back down.

Obviously since it is indexing quickly you don't want to be making instantaneous decisions for the LOW and HIGH on the dancer. Maybe take a snapshot immediately after every index or some other consistent point. Obviously the TOO HIGH and TOO LOW levels may require an immediate response or just a slight debounce.

For some extra tweeking, you can adjust the accels and decels of the unwind drive to affect the response. You can also add a couple more levels on the dancer that maybe kick in a larger % influence to help get things back in line quicker. These things will all be system dependent.


I don't think there is anything wrong or archaic about using dancers. The are simple and easy to understand and maintain. They are also very forgiving. I've seen places spend weeks of their resources trying to get something to work without a Dancer in a process that really could have used one.


Most importantly, keep things simple!
 

Similar Topics

Hi everyone, I hope I don't butcher this up, please feel free to critique me wherever if I do, I have an issue I would equate to "chasing...
Replies
4
Views
242
I had a comms fault between my VFD and Controller (5069-L320ERS2) that started about a month ago and happened maybe once a day to now where it...
Replies
1
Views
244
Well, I'm working with this ABB plc project, and It's been a learning experience coming from Allen Bradley. The project can't be changed to an AB...
Replies
1
Views
1,140
We have an application where we're communicating between a 1769-SDN a 3rd party device. For the most part things work fine. We're reading maybe...
Replies
9
Views
2,712
HI all, i dont suppose anybody knows of some inverter range that might be specialized for running intermittently? we have a project where the...
Replies
7
Views
1,396
Back
Top Bottom