What's new
What's new

Fanuc 21i - overshooting when changing axis direction

amh82

Plastic
Joined
Sep 2, 2020
Hi guys, hoping someone can offer some pointers for a problem we have with a Takumi mill.

The machine uses a Fanuc 21i-MB control.

The operator had mentioned that when milling a circle, he would have tight spots (when checking for a H7 fit) at the X and Y changeover points.

I initially started to check for backlash, there was a little, but not a lot - but whilst checking against a clock I noticed that when I changed direction the axis would jump. Say I moved 0.001mm on the control, the clock would move 0.01mm. If I continued to move the axis in the same direction, the initial overshoot would be made up (i.e if I carried on moving to say 0.1mm, the clock would also show 0.1mm).

If I reduced the backlash figures in 1851, then the jump would go away, but of course I would l just be left with backlash. I read of another CNC owner who had the exact same problem and fixed it by turning the backlash acceleration function off in 2003 bit#5 - I thought this could be it as some of the backlash acceleration parameters looked to be a long way from default, but it had no effect.

Could anyone suggest what could be causing this initial overshoot? I could be wrong, but I suspect this is what is causing the problem when trying to interpolate a circle, it seems to tie up with where the tight spots are on the circle at least. I did wonder about servo tuning, but I have little experience of this so have not wanted to dive straight in!

I contacted a calibration company to have the machine checked and adjusted against a ball bar and interferometer, but they are asking me to double check for any other problems first.

Could any mechanical problems cause this overshoot?

The machine runs very well other than this, and has not had a hard life at all. Really appreciate any suggestions - thanks!
 
To start, lets get a few more blanks filled in.

Both X and Y behave the same?

What values do you have in backlash parameters 1851?

Boxway machine or linear bearing guides and are they being lubed properly?

Have any mechanical aspects been checked (thrust bearings, ballscrew/nut, no loose mounts, etc.)?
 
Hi Vancbiker, thanks for your reply.

The same is seen in X and Y.

The initial values in 1851 were 44 (X) and 52 (Y) - from memory I had to increase them both by around 15 when checking with a clock.

The machine uses box ways - would stick/slip give this sort of problem? My experience is mostly with linear rail machines, and even then, I'm electrically biased, so having to learn the mechanics the hard way!

Again, I haven't got back to the machine since, but was going to put a clock on the end of the screw to see if there is any movement in the thrust bearings when changing direction, check for anything loose, etc - but any pointers will be really helpful!
 
With values that high in your comp parameters you have something wrong. Since both axes are affected, first place I'd check is oiling.

With high stiction (stick/slip) one tends to set too high of comp values when setting with a dial indicator (Americanese for a clock gage). When the machine is dynamic the effect of stiction is greatly reduced compared to manual movements with the handwheel.
 
Hi guys, hoping someone can offer some pointers for a problem we have with a Takumi mill.

The machine uses a Fanuc 21i-MB control.

The operator had mentioned that when milling a circle, he would have tight spots (when checking for a H7 fit) at the X and Y changeover points.

I initially started to check for backlash, there was a little, but not a lot - but whilst checking against a clock I noticed that when I changed direction the axis would jump. Say I moved 0.001mm on the control, the clock would move 0.01mm. If I continued to move the axis in the same direction, the initial overshoot would be made up (i.e if I carried on moving to say 0.1mm, the clock would also show 0.1mm).

If I reduced the backlash figures in 1851, then the jump would go away, but of course I would l just be left with backlash. I read of another CNC owner who had the exact same problem and fixed it by turning the backlash acceleration function off in 2003 bit#5 - I thought this could be it as some of the backlash acceleration parameters looked to be a long way from default, but it had no effect.

Could anyone suggest what could be causing this initial overshoot? I could be wrong, but I suspect this is what is causing the problem when trying to interpolate a circle, it seems to tie up with where the tight spots are on the circle at least. I did wonder about servo tuning, but I have little experience of this so have not wanted to dive straight in!

I contacted a calibration company to have the machine checked and adjusted against a ball bar and interferometer, but they are asking me to double check for any other problems first.

Could any mechanical problems cause this overshoot?

The machine runs very well other than this, and has not had a hard life at all. Really appreciate any suggestions - thanks!
Looks like classic REVERSAL SPIKES issue. Here is already several times advertised here procedure:
Fanuc 16 and above series controllers parameters for reversal spikes and backlash and servo loop gain


Important note the original settings in case of error

Backlash removal

The parameter is 1851. Remove positive backlash by increasing the value.


Reversal spike removal (for this to work there must be a value in the backlash parameter ie 1851 must be greater than 0)


Positive reversal spikes:

2003 bit 5 make a 1

2048 (increase/decrease by 50)

2071 time constant (2 ms units)

Negative reversal spikes:

2009 bit 6 to a 1

2048 (increase/decrease by 50)

2082 time constant (2 ms units)

Servo loop gain

Parameter 1825 where 0.01 seconds ie 10 milli seconds is one unit change in the parameter value
 
I would be looking into mechanical issues. Those are some massive comp numbers. I've played with backlash comp quite a bit in the past on some of my machines. I only have one mill that will interpolate a circle that's round to better than 0.0001" and it has 2 1 and 0 in the x y z comp settings. Maybe someone has had different experiences but I've yet to find where backlash comp will truly fix and hide more than a tenth or two.
 
Thanks for all of the replies guys!

The machine was demonstrating the problem interpolating a circle with figures of 44 and 52 microns in X & Y - I gather that around 50 microns is considered the limitation of compensation before a mechanical fix is required?

I added a little bit to these values when checking against a clock with a handwheel, but as you guys have suggested, I was a little concerned that this may not be a true representation having read elsewhere that BL comp set using a handwheel is usually too high. I will check it again running from a program.

I found the original parameter file, and from the factory this machine had 16 (X) and 19 (Y) microns of backlash comp.

Vancbiker - I will check the oiling/lube as you suggest. As you say, this problem is identical in X and Y.

Probe - I had read about reversal spikes causing this problem and had played with the figures but did not see any change, which is why I tried it with 2003 bit 5 turned off to see if I noticed any change - but still nothing. Did I have it wrong looking at positive reversal spikes, should I have been looking at negative reversal spikes?
 
.......I found the original parameter file, and from the factory this machine had 16 (X) and 19 (Y) microns of backlash comp.........

Those are high numbers for a new machine. I have heard of Takumi as a builder, but never worked on one to know what kind of quality they are. In the plant I retired from, we'd consider those values near to, or beyond, the point where the machine needed repair (depending on the size/type of machine).
 
This machine was built in 2006 which is before Hurco acquired the brand (have read that the quality suffered post Hurco takeover) - it's not a machine I am familiar with (my background is mostly German CNC Grinders) but the operator rates it highly. The machine is reasonably large, 2200 x 1025mm bed with 2000mm (X), 1066mm (Y) and 950mm (Z) capacity.
 
For a 40 x 80 travel boxway machine the factory numbers, while high, would be usable. 8-12 would be a good number. I suspect that this machine does not use tensioned ballscrews and probably undersized on diameter. Undersized diameter is a tough problem to manage due to torsional windup when reversing.
 
Probe - I had read about reversal spikes causing this problem and had played with the figures but did not see any change, which is why I tried it with 2003 bit 5 turned off to see if I noticed any change - but still nothing. Did I have it wrong looking at positive reversal spikes, should I have been looking at negative reversal spikes?
Let me know your email on PM. I will send you a document which covers comprehensive step by step procedure how to deal with reversal spikes problem. It simply is too heavy to attach it here.
 








 
Back
Top