Flow wrapper with a two axis gantry on the infeed, Kinetix 5700 with a 2198-D020 dual axis module on a 1756-L83E. The Y axis faults Excessive Position Error every few cycles, always on the fast return move, right where it comes off about 1.2m/s into the decel. X on the same module never faults, and the machine has run two years.
Mechanically it looks clean, I checked the belt tension and the carriage moves freely with the drive disabled. Widening the position error tolerance from 5mm to 10mm made it fault less but not stop, and dropping the return speed 20 percent did nothing at all.
Is the motor going, or the feedback cable? It only goes on the fast return though, so would a motor or a cable really pick that one move out of the cycle?
Post the trend. And Axis Properties, Position Loop tab, what are the gains and what's Velocity Feedforward set to? Also gearbox ratio and whether tuning was ever run on this axis or copied from X. Widening the tolerance is hiding it, put it back to 5mm.
Trend attached. Error climbs with the speed, sits around 8 or 9mm through the fast part, then collapses to 0.2mm once it stops. Position Loop tab: Position Loop Bandwidth 24Hz, Velocity Feedforward 0 percent, Acceleration Feedforward 0. X shows 100 percent and 0. Gearbox 5 to 1 on both. X was tuned, Y I think was copied and edited, that's before my time.
Copied and edited is where I'd stop looking at the mechanical side, pretty sure.
what does feedforward do that the gain doesn't? Is it just more gain?
Not more gain. With feedforward at 0 the position loop only pushes when there's already error. At speed it trails the command by whatever error it takes to make the proportional term push that hard, and at 24Hz and 1.2m/s that works out around 8mm. Feedforward hands the loop the commanded velocity directly, so it doesn't have to build error to move. Your trend shape, error tracking the speed and collapsing at rest, is exactly that.
Axis Properties, Position Loop tab, Velocity Feedforward to 100 percent, same as X. Re-run the move and watch the trend, it should sit under 1mm. Tolerance back to 5mm after. Writeup on the loop side of this here: https://plctr.com/utilizing-motion-control-in-plc-programming/
Velocity Feedforward to 100 percent on Y, tolerance back to 5mm. Ran 3000 cycles overnight, no fault. Trend now peaks at 0.6mm on the same move, same profile.
So feedforward at 0 on Y, the loop chasing the move on gain alone and trailing past the tolerance. 100 percent velocity feedforward, matching X, and it's gone.
What I don't know is who set it to 0, or whether it was 0 from day one and something else slipped that made it show now. Two years of it working says something changed. Belt stretched a little maybe. No way to tell now.
Thanks mattr, carlh, and finn.
Keep that trend screenshot with the project. Next person will want it.