PID and hard stop causing problem?

I had a problem on a hydraulic servo controlled cylinder, which I believe was caused by the PID loop and the cylinder reaching a hard stop at one end. I want to ask the PID experts out there (PETER)if my thinking on what I saw is correct.

Here is the setup: I have a cylinder that extends to a home position, when a board on a lug chain is indexed to the right position the cylinder retracts to pull the board onto another chain (there are 4 identical cylinders, only two are used depending on length of the board). The control (1771-QB) uses a .250" PID band, and outside of the band uses feed forward to drive the servo.

What would happen is if a few lugs were empty or the line had to stop for a few minutes, on the next one or two retract strokes it would not get into position, but once it was continuously running it acted right. I also noticed it would never get to the target home position, but since it was within the in position band of .200" it would still run. I came in over the weekend to look at it closely, and found that at the home position was physically hitting the hard stop of the pin slide (the slide shaft was replaced recently) while the other cylinders had about of inch clearance. After a few measurements I offset the zero move the home position away from the hard stop by about .200" and the problem went away.

I think what was happening is when it was at home position, the integral term would wind up causing the control to be biased away from the retract position, and it not getting into position. The operator would turn off then turn on the hydraulics (which sends a reset to the QB) which resets the integral term and it runs normally until the next pause. Does this sound plausible?

I am working to determine why the slide is not correct, either not install correctly or it was the wrong length. We were trying everything we have ever seen cause similar problems (cylinder, servo valve, hyd pump, accumulator, temposonic) without any results. This one was stumping me, but if my conclusion is right, I have another nugget to store away in my brain.

Thanks guys.

The 1771 module implies this is a PLC-5. That implementation has anti-reset windup, which stops integrating when the CV hits the limit.

I would be more concerned with using the feedforward feature. In my experience it will eventually bias the loop so much, the PID impact is overrun. You might consider using transitions to software manual to manipulate the output between cycle, and then switch to auto when the output is creating good feedback to the loop.

The other three identical cylinders have been operating fine with the exact same parameters, and this one did too until recently. The slide shaft had been changed out due to the old one breaking, that seems to be when this problem started.

We have plans on the table to upgrade the PLC5 to a Controllogix next year, likely with Delta RMC controllers. But this has been working like this, as well as other machines with similar setups, for years (>10years). So I don't think the setup is wrong or we would be seeing this problem other places.

My earlier comment was in the context of the PID instruction in RS-Logix 5, not necessarily the implementation in the 1771-QB. I was not familiar with the QB until reading about it in more detail. With the QB, you do have the option to limit the integral term in the parameter block if wind-up is a concern.

With the recent maintenance work, examining setup and wiring (if touched) would be advised. For example, did the feedback signal change with the new shaft? Home switch? Limit switches?

The QB module works good, but of course the technology behind it is dated. It is made specifically for hydraulic servo control, saving a lot of programming other than initial configuration and set points.

There are no limit switches of any kind on this system, the Temposonic is a magnetostrictive linear sensor which mounts in the hydraulic cylinder and uses a magnet attached to the piston for absolute position feedback. All the wiring is fine, the shaft that was replaced is what the pin rides on. The pin is moved by the cylinder.

The reason I suspect integral windup is that with the rod being too short (by either manufacture or installation) the pin cannot reach the home position since it hits the shaft mount. When it sits there for a period of time, the integral will windup (but yes,it is limited in the QB)and when it goes to do its next stroke, the integral term biases it towards the home position until the term is reset or it sits long enough for the integral unwinds to get into position. That would be why when I offset the zero away from home, it could then get there and the integral does not windup. By the way, our home position is actually at 37", and it retracts towards zero which is the center of the chain it is pulling the board to.

Mark Snodgrass said:

The reason I suspect integral windup is that with the rod being too short (by either manufacture or installation) the pin cannot reach the home position since it hits the shaft mount.

Click to expand...

It seems as though you are facing a choice to make the mechanical parts like they were before the maintenance work, or modify the motion profile and/or logic to match the new physical arrangement.

Since it may not be practical to fix the rod/pin/mount situation, a new profile with reachable positions may be the most appropriate option. In scanning through the QB manual, I noticed a command bit named "Integral Disable" that could prevent the suspected wind-up. The challenge with this is whether you reliably know when to deactivate the integral action. I imagine there are other logic-based approaches, though more like band-aids instead of root cause corrections.

Yeah, we are going to fix the rod. Probably just need a small spacer to get it where it should be. A quarter of an inch is all we really need to get it right.

Sounds like a sawmill edger pulling boards on to a feed chain. It is hard to tell exactly what is happening without more info but the QB modules didn't provide much debug info.

Yes, if the integrator winds up the there would be a bias but it would tend to be away from the retract position. It would be trying to force itself to extend. This would keep the position from reaching the retract set point if the integrator didn't have time to unwind. Since edgers move quickly there isn't much time.

It could also be that if the actuator didn't reach home or the retract position then the in position bit never came on and that messed up the sequencing.

Hey Peter, it is an edger. An Omega feed table to be exact. It definitely was not getting into the home position (extend stroke), about .150" from reaching the target. It was up against the mount for the guide rod. The other fetchers have about an inch of space when at home. I only really noticed that when I came in this weekend to test it while the mill was not in production. A zero offset away from the extend position allowed it to reach home, and the problem went away.

I am glad you agree that an integrator windup would cause the symptoms I was seeing. It was the best explanation I was able to come up with.

We just need to figure out why the guide rod is shorter than the rest, probably a spacer missing where there should be one.

The good news is USNR is quoting us a new optimizer package which includes a Controllogix upgrade, including your RMC controls for the 10 hydraulic axes. I don't have all the details yet, but it will be installed next year.


Thanks for responding

If the problem is with the integrator winding up it should unwind if it is moved a location where it can move freely.

I/we have seen all sorts of weird stuff. If the position is always off by 0.150 inches is that by external measurement or is that what the QB module says. This is important. We have seen cases where the whole cylinder moves. In this case the motion controller thinks it is in position but all the boards are too wide or thick when measured by the QC person.

You will like the USNR edgers. They are fast. You will like RMCTools too. RMCTools provides lots of debugging information.

I measured the position manually, and the QB position matches. And visually I can see it was up against the guide rod end support. It is the way the millwrights have it installed currently, I will just have to get a time to be able to remove the extra spacers they have put in at the fixed end of the rod.

I think if the operator was patient enough, the integrator would unwind. I think with some better tuning I could make the fetchers run better, but when mechanically sound they do the job. When we put the new controls in next year, we will definitely dial in the tuning then.

Unfortunately we are only getting the optimizer and not the entire edger. We are just upgrading the controls on the old Baxley/Omega hybrid we have now. Of our 11 mills here in the south ours is the most productive with the oldest equipment, so we have a real hard time finding much ROI for most new equipment. We have become good a taking a machine designed for a certain production level and getting twice that. This mill was built for 25,000BF/hr and we exceed 50,000BF/hr as standard now making small incremental improvements.

I have looked at the RMCTools, definitely easier to navigate than a data table in PLC5.