ABB sent an experienced wind/unwind engineer to work through the issues. Coincidentally, he was one of the guys who helped start up this machine back in 2009, before he worked for ABB, and was able to relate some of the headaches they had back then. He had a lot of experience with wind/unwind applications, and I got a lot of information from his methodical process.
We went through every detail of setting up each roll individually, with the anticipation that perfection in the individual drives would make the application fall together effortlessly. I recorded the laptop screen on the first day (without sound) so I could digest everything afterwards. It was good... I got a lot out of it. But it didn't fall together like we anticipated.
He got the torque mode to work, which was a huge stumbling block for me. It didn't come easy, but he got it to work on the second or third attempt. Another issue I had was getting the drive-to-drive communication to work for speed reference from the master drive. He couldn't get it to work as written in the manual, either. I contend there is a bug in the firmware for that function. He gave up on the standard path and looked for a work-around. We really needed the ramped speed reference from the master drive to be sent to the unwinders, and they needed to apply the diameter to that reference to get the RPM target.
It wasn't going well. We could get a reference value into the unwinders, but we couldn't get the dimension calculation to work right. There is a scaling parameter involved, but every time we moved it toward the value that made sense, it got worse. Finally, I figured out that the relationship was probably inverse... which didn't make sense, but it was the only thing we didn't try. ....and it worked!
So now everything was looking promising. The application was working for the first time, and we started slow and increased the line speed each time. At 1,050 (ish) FPM, all hell broke loose. Long story short, the ramped speed reference that we sent from the master had a scaling factor applied in the unwinders. When the value got to a certain point, the number turned negative. (I don't remember the details well enough to insert them, and I don't want to get it wrong.) After working on that for a while, he found a smaller number that worked as a ramped speed reference.
After that, we worked on the details of running the application, like inertia compensation , PID tuning, and trim, etc. When he left after two days, it wasn't perfect, but it was close enough to making finished product that I could take it from there. I had the following Monday to work with the machine operator to train her and get it good enough to make finished product while I was gone for the rest of the week. I had her limit the line speed to 800FPM, and that kept the machine productive while I was gone.
On Monday, I just watched the machine for a couple of hours. Initially, it looked like the dimension function in the unwinders was having issues. On ramp-up, the poly tension would shoot high, the PID would respond, then after ramp up, the poly tension would dip to almost nothing before coming back to set point. I watched the dimension calculation struggle a little during this, and started working on a theory that it was the cause of the PID not being accurate. I saw the PID output parameter respond high enough that it should have kept the error low... but it didn't. And then after ramp-up, the PID output was too high and it caused the value to overshoot in the other direction. It seemed like the drive just wasn't responding to speed changes fast enough. Then it hit me.... the ramp values! Early in the setup, the ramp values were an issue and we set them to a value close to the master drive. That worked before we changed the speed reference to the drive-to-drive value. But now, the speed reference was a "ramped" value. I reduced the ramp up and down values to 1 second on each of the unwinders and it all fell together. Since then I've been tweeking the PID. ....I was able to dial the gain back up, and now I'm just working on the overspeed issue at the end of the cycle, and the ramp to 0 speed.
Hopefully, someone out there might find value in my learning curve. I'll throw some more on this post when it's all done... probably put a few visuals in as well. I'll see about sharing the parameter files as well.