Seems to me like you could use your dancer position as the PV into your PID loop, then control the output to your brakes, based on how far off you are from the desired position of your dancer. The point is to maintain a uniform tension on the web as the machine stops?
I suppose you could leave the brake pressure constant, and then control air pressure to the dancer to try to maintain dancer position, it just seems like that would be less effective.
We have a similar machine (rewinding) that we have been having problems maintaining even tension on the unwind end. On it the unwind (mill roll) is driven off the same shaft as the rest of the machine, only with a v-belt and variable speed pulley. The pulley is controlled mechanically with a dancer roll that, as it moves up and down, opens and closes a linear regulator that feeds air pressure into something called a "pneumatic tension regulator" The incoming air from the dancer works on a diaphragm that operates ANOTHER regulator that will vary the position of an air cylinder attached to this monstrosity that in turn will move the tension roller for the unwind drive belt, thereby changing where the belt rides on the variable pulley. This allows the unwind roll to keep a (somewhat) constant surface speed matching the drive speed of the machine. I guess you could call it "air over air", pretty amazing, really that it works. I guess that is what guys used back in the 40's before VFD's existed. (I'm not kidding, we have mechanical drawings for these things from PCMC dating from 1947!)
Anyway, hats off to the genius that invented that thing, but it's beyond obsolete. I'm wanting to put an ABB drive with an independent motor and timing belt drive to it, using one AI from the main drive with another AI for trim from a dancer pot, and a diameter calc in the drive for the speed reference.
Not sure if this kind of set-up would do you any good in your application, just thought I'd put it out there.
Cheers,
Dustin