2 Servos mechanically coupled; bad idea?

[James Fillmore]

I've done 100 single axis AB units now and always used the auto tune and then tweaked the parameters. I’ve been dealt a new application; lifting a 2000lb plate to multiple positions with 2 AB MPL430 servo mounted to 24" Exlar actuators. I can't autotune this since the 2 servos are MECHANICALLY coupled, one on each side. It looks to me like a bad mechanical design. I suggested like a scissor lift in the middle but this is what I have now and customer is on his way.

IS this a bad servo application? Mechanically coupling 2 servos? Do you ever see applications like this?

How the heck would I tune this? What parameters would I use for each motor if I just did the calculations by hand?

 

[cottagewood]

I've seen this application several times, the amplifiers are set up as Master and Slave

 

[OkiePC]

I have seen few designs that look mechanically bad that end up being good.

You may end up with best results if you can switch one of them between torque and position follower mode. If you run master/follower, there is inherently going to be a lag. If you run them from a master virtual axis with both in positioning mode as slaves (once they're balanced) and tight position tolerances it may do even better.

Interlocking each amp position error and other fault responses and brake circuit timing may require special attention.

 

[James Fillmore]

You'd think they'd fight each other even a small move on a 3' x 8' x 6" steel plate supported on each end by linear bearing does have much give. A tiny bit outta sync would be a HUGE imbalance and one servo basically carrying all the load. Inertia matching at it's worst I would think.

 

[TConnolly]

The CLX and its motion instruction set supports electronic gearing. Is there a reason why that won't work?

 

[James Fillmore]

I have used the MAG instruction in the past. It works great. I used it on a machine were I was advancing a long plastic part and then feeding insulation from a roll from a 2nd axis. But they weren't mechanically connected. With this how do I tune, home, estop?

 

[OkiePC]

I would try the MAG (one to the other) and closely monitor current and position error on both drives to see if they "fight".

Then try a virtual with both axes MAG'd to it, move the virtual through the range of motion to see if it helps with any irregularities in another observation of drive position error and current/balance of load.

I think you will have to tune uncoupled if allowed, or do a static only type of tune and skip dynamic tuning if your drive allows.

All action will require the gearing status to be interlocked, so you're going to be rolling your own homing routine.

If (when) one of them gets in a bind relative to the other, this is where running in torque mode briefly during homing, may be a cheap and dirty way of getting them aligned after, say a drive belt replacement or other change that may have shifted the relationship slightly.

I would need to see the mechanism to qualify any of the above comments.

 

[James Fillmore]

Pic test

 

[Randy Turner]

Gearing mechanically coupled servos

I had to do this on an X, Y, Z pick and place system. The X axis was the only one with a single motor. Both the Y and Z axes had a motor on each end, with a common beam between them. The MAG will work, and you can also do an incremental move on top of the MAG instruction if one of the axes lag the other by your predetermined amount. For tuning, our Rockwell sales engineer recommended starting with the gains displayed in the servo sizing software when you simulate a move. Beyond that, you can also autotune one of the axes with the other axis geared to it. The slave axis will move while the master tunes.

Randy

 

[OkiePC]

I am not sure I can clearly make out what are the actuators versus guides, but I would expect you will (as passively as possible) home this thing from a maintenance screen with a ruler or fixture to move one to align it with the other, then store that as the current position for both axes, and enable and monitor gearing from that point forward, keep the logic supply powered up or use absolute feedback to minimize the need to home.

It appears there is going to be enough give between the frames sides to allow for a small tolerance for misalignment, beyond which catastrophic mechanical failure may occur. So yes, carefully scrutinize how the drives are braked and interlocked on the schematics as well as the software side of things.

Be sure to set the current level low enough so that either drive will trip well before misalignment gets extreme.

How fast and how far does it move?

You should getpaint.net a layer of fat brightly colored arrows on a copy of frtview.jpg and add notes to help us see what is what.

 

[James Fillmore]

Pictures of vertically mounted actuator not yet attached to 2000lb lower platen,
24" Exlar actuator wit AB servo motor (there are 2 of these, one on each side) and a scissor lift idea that I suggested with 2 large, pneu counterbalance cylinders on each side.

 

[Peter Nachtwey]

You don't want to gear. If you issue the same MAM command to the two axes with they will move together, at least their target positions will. The problem is getting them tuned. We and our customers have done lots of these applications where the axes are tied together. One trick I use is to issue the same open loop command to the two axes at the same time. They should track reasonably close. It is best to have some sort of skew detection incase the axes don't track well in which case the mechanical people should be asked why. From the data, trends, graphs, etc, I collect I can get a pretty good idea of what the feed forwards should be and the gains.

 

[OkiePC]

Beyond that, you can also autotune one of the axes with the other axis geared to it. The slave axis will move while the master tunes.

Randy

That's interesting. I suppose they could take turns to get them both tuned reasonably well.

As to Peter's point, using the virtual axis and gearing both real axes to it does the same thing as simultaneous MAM at least it did in the GML days.

I think this should make it simpler to keep them aligned if either should fault, just move the virtual to the position of the faulted one and stop it (brake if applicable).

 

[James Fillmore]

How do you gear an axis to another to tune? Do the MAG and then initiate the tune in software?

 

[Randy Turner]

Gearing slave before tuning

Yes, you can use the direct motion commands and gear the second axis to the one that you want to auto tune. When the
first axis does it's movements to calculate inertia, the second axis will follow.