Fanuc 18i-MA servo tuning

[Milling man]

Hello everyone, I hope someone can help a little with my old friend, Fanuk 18i-MA (Serfo Software 90A0-5). When any linear axis is reversed, something strange happens. For example, if I run the code:
G91
X0.0004 F20.
X-0.0004
X0.0004
and so on.
In this case, the X axis at the beginning of each line will first “jump” in the desired direction by about 0.002 and then return back by 0.0016 As a result, the axis ends up in the desired position, but it goes through a strange path))))) This situation is on all axes - X, Y and Z. In this case, the lower the feed speed, the smaller the “jump”. Circle interpolation with a radius of 0.02 seems to work well.
I control the movement of the axis using a dial indicator - the electronic indicator does not have time to display the devilry that is happening, the process is too fast.
Obviously it's a matter of servo settings. This is an old rack, it cannot be directly connected to a PC and use the servo guide. I need to adjust the settings manually. The problem is that I still have no idea what parameter can be used to fix this.
I tried changing the parameters on the servo tuning screen - the “jump” value did not change. I tried disabling the backlash compensation and changing the amount of backlash on all axes to 0 (it was about a few microns for each axis, the actual backlash was about 10 microns or 0.0004 inches) - it didn't help. I tried reducing the acceleration of the axes - it didn't help.
System composition:
Mazak FF-660
Fanuk 18i-MA
Servo drive - A06B-6096-H107 (SVM1-240)
Servomotors - alpha M40/3000 with fan
Software versions:
System: BDF2-0017
Servo: 90A0-0005

 

[SeeFair]

I have one comment and one question:
1. A dial indicator is probably not the best tool to measure fast transient motion. I don't think the tip of the indicator stays in contact with your axis during the move: the indicator and tip have their own dynamics. You describe a 200% overshoot. Something very dramatic would have to happen to all 3 axes for that to occur. So, I don't think it's real.
2. What problem or machine behavior led you to investigate 0.0008" moves? Does it overshoot on large moves also? (I hope not 200%). The small moves are the hardest for a servo to do. If you optimize the servo tuning for very small moves, you are likely to see some really bad, violent behavior when you use those parameters on large moves.

Caveat: I know nothing about this machine, and my experience with servos is quite dated. So maybe this machine does some fancy gain scheduling where the gains are different for big and small moves, I don't know. But I DO know that small and large signal behavior in DC servos is very different, and you can't just focus on small moves and use those settings for large moves.

 

[Milling man]

1. A dial indicator is probably not the best tool to measure fast transient motion. I don't think the tip of the indicator stays in contact with your axis during the move: the indicator and tip have their own dynamics. You describe a 200% overshoot. Something very dramatic would have to happen to all 3 axes for that to occur. So, I don't think it's real.

This doesn't seem to be a problem - I make contact with the indicator on some part of the machine before starting the test. And to be on the safe side, I tried pressing the tip of the indicator with my hand I didn’t see any difference between “with a hand” and “without a hand”.

2. What problem or machine behavior led you to investigate 0.0008" moves? Does it overshoot on large moves also? (I hope not 200%). The small moves are the hardest for a servo to do. If you optimize the servo tuning for very small moves, you are likely to see some really bad, violent behavior when you use those parameters on large moves.

The initial problem was that the machine broke thread mills like matches. I already knew about the “movement features” described above, which I discovered during normal everyday work - and immediately assumed a connection between these two factors.
In general, the machine works well at fairly high feeds, for example F200, and works well with small tools - but these jerks when reversing the axis spoil everything.
The amount of movement chosen is very simple - this is a metric machine, and I'm a metric guy, so it's 10 microns I tried movements at larger distances, for example 0.05 inches - but the amount of jerk depends on the feed - that's what I found out for sure.

 

[SeeFair]

I tried pressing the tip of the indicator with my hand I didn’t see any difference between “with a hand” and “without a hand”.

I think this means you were sure the tip was in contact with the axis? I get that, but I still suggest that tip gets launched like a catapult when the move happens.