2 Servos mechanically coupled; bad idea?

[James Fillmore]

WOW! Great info. That sounds kinda wild but I'll try it. Wish me luck that I don't break anything. I'm a big proponent of the pneumatic counterbalance cylinders for both safety reasons and to take the load off the servo(s). This is the first time after many jobs that or mechanical dept said, "NAH, we don't need those".

 

[Peter Nachtwey]

Gearing isn't the answer. Gearing should never be used if PLC is controlling both axis. One should follow the internally generated target position.

Gearing doesn't work when tuning mechanically linked actuators because it ASSUME the slave is already tuned. Then again, what do you gear too?

You need to write some sort of skew detect to make sure the axes don't get too skewed and bind. Do not go past go without doing this. If the two axes are similar they should both follow the internally generated target position with the same amount of error. Therefor the actual skew between the axes should not be large. HOWEVER, I have seen faulty mechanical systems where two identical axes with identical loads did not work the same and that is when the skew protect saved the day.

After the adding the skew detected the second thing to do is have some patience. Make point to point moves SLOWY and notice the following error. Increase the gains and the following error should drop and now you can increase the speed a little more and still stay within following or skew limits. Adjust the gains some more so both axes are tracking the same way and repeat until you get to the operational speed.

 

[James Fillmore]

Thanks Peter.

 

[OkiePC]

Gearing isn't the answer. Gearing should never be used if PLC is controlling both axis. One should follow the internally generated target position.

This is what the virtual axis is. The internally generated target position to which both axes are slaved, enabled, and then they follow the MAM applied to the virtual.

Gearing doesn't work when tuning mechanically linked actuators because it ASSUME the slave is already tuned. Then again, what do you gear too?

The gains better be close up front huh?

They'd be geared 1:1 at all times unless failed, or for alignment. You should be able to mechanically align them so that the load sharing is similar or use a raw torque command during the one direction auto tune operation, it could end up being a never-ending jerk circle. I foresee some good gains math coming...

Not that I am a fan of A/B autotune, Creonics was much better, they should have kept that through the s-class GML days. Hopefully it's better now in Controllogix than what we were stuck with in Commander.

You need to write some sort of skew detect to make sure the axes don't get too skewed and bind. Do not go past go without doing this. If the two axes are similar they should both follow the internally generated target position with the same amount of error. Therefor the actual skew between the axes should not be large. HOWEVER, I have seen faulty mechanical systems where two identical axes with identical loads did not work the same and that is when the skew protect saved the day.

This should be taken care of by the sum of the position error of the axes, plus the position error bits. You should be able to set the tolerances pretty tight once it's all tuned up. Using the position error flags, the fault trap is somewhat done for you this way. Just do the right thing with the virtual master axis when something goes wrong

After the adding the skew detected the second thing to do is have some patience. Make point to point moves SLOWY and notice the following error. Increase the gains and the following error should drop and now you can increase the speed a little more and still stay within following or skew limits. Adjust the gains some more so both axes are tracking the same way and repeat until you get to the operational speed.

 

[DamianInRochester]

I have seen this attempted many times. It is one of the applications I have seen that causes almost universal grief. There are a great number of pitfalls with this. The most complicated part used to be homing them in sync. Now I would not even attempt it without absolute feedback. The other problem is that even if you get it to work now, it will eventually fail. But they are not going to both fail the same way. They will wear differently. Then when you replace one the new one will behave different than the old. It is a nightmare setup for the guy who has to fix it 5 to 10 years later.

The biggest question ends up being, how much compliance do you have? How far out of sync can they get before you start binding.

In my opinion this configuration should be avoided if at all possible. And if you must, use absolute feedback, build in some compliance, and sync them both to the same virtual axis. I would tune it by uncoupling it from the mechanism and placing a weight of half the mechanism on it. It should get you close.

 

[Peter Nachtwey]

Yes, for the most part I am dealing with hydraulics and they have absolute position feed back. However, if encoders are used they can still be be moved in open loop and skew can be checked using relative counts.

Setting up linked actuators takes time and patience but it can be done. Start slow and work up to full speed.

 

[_Woody_]

Just my 10 cents.

1. Absolute encoder position feedback !!!!! Without it this will be difficult if not impossible to home.
2. I am in whole hearted agreement with DamianInRochester - Even if you get it going now it is going to be a nightmare for someone in 5-10 years.
3. If it must be done this way make both servos slaves of a master virtual axis as OckiePC has suggested. This way you program the motion for one axis (virtual) and the real world follows.

Did a platen lifter for a vacuum molding machine a few years back. This design used four ball screws in each corner. The top of each screw had a sprocket. This was linked by a chain wound around all four and they were driven by a single servo. Mecanically more solid, easier to maintain and homing the device was easy.