3d calibration for milling tasks??

[instituto_idf]

Last week, a man from KUKA-Spain visited me. At this time i'm fighting with the calibration of the externat axis (linear track and rotatory table) in order to prepare it for milling tasks. The main problem seems to be a small mismatch between the $robrooth coord system (and hence $world) and the real mecanical movement of the linear track.

Talking to him, he explained that there is a service of "3D calibration", quite expensive, consisting of carrying here a special equipment form Germany (only by boat to avoid pressure effects on this equipment), and then the one week long work of calibration. Thats the commercial way to minimize the error when moving in a cartesian coordinate system in such a task (milling), different of other tasks where the memorized positions include the possible error of the cell (and so the repeteability is really great).

please, ANY INFORMATION ABOUT THIS 3D CALIBRATION??

Many thanks

[TygerDawg]

This **sounds like** the KUKA method they use on the KR30-HA "High Accuracy" arms.  The put a laser target on the arm and run the arm through pre-defined motions in the 3D work envelope.  A laser tracker is used to track the target to get "error values versus commanded position" data in the workspace.  Then an error matrix is calculated and applied to the kinematic model to adjust for the errors.

Or **something** like that.

[bert]

**Something** like that indeed :)

I just thought I'd clear a few terms up here.

There are two different terms used by Kuka - firstly "Absolutevermessen" (or absolute accuracy) and H-A (or high accuracy).

The first of therse is essentially a calibration applied to the robot to make it's static positioning more accurate. This calibration file is generated in Augsburg using a measurement system from Metris and software developed by Kuka (AROB or XROB). This generates a completely new kinematic model for the robot and also measures the flexibility of the robot so that it can account for different payloads. When this procedure has been followed in the factory, you will see a sticker on the robot near to the serial number plate stating so. Robots so measured typically have an static spacial accuracy better than 1mm.

High accuracy robots are a little bit more mysterious... I have not yet got a consistant definition of what the difference between a normal and H-A robot actually is. What it should mean, though, is that it is better able to follow a path than a normal robot. Whether this involves differences in the motors and gearboxes is not quite clear to me at the moment.

Anyway, I don't know if this is interesting, but there it is...

[SkyeFire]

Bert's description is pretty close, except that even "Absolute Accuracy" robots have varying levels of accuracy depending on their model.  The 16-kg robots, for example, have tighter accuracies than the 500-kg robot, even if both have been given AA calibration.
However, AA is not a panacea.  It doesn't make the robot *more* accurate as much as it makes the robot's accuracy more consistent across the movement envelope.

"High Accuracy" is only available on a limited number of robot models, topping out at (I think) the KR240 at the moment.  My (possibly mistaken) understanding is that the HA robots combine AA software calibration with slightly better mechanical components to improve accuracy beyond what AA can accomplish alone.

However, going back to the original poster's question, as far as I know this process does not apply to external axes at all, only to a robot arm as delivered from Germany.  I suppose it's *possible* that KUKA might offer this service for KUKA-brand external axes (like the KL1500 linear slide axis), but I've never heard of it. 
One problem here is that even with an AA robot, the accuracy of the external axis will be the limiting factor.  If the linear axis is only accurate to 2.0mm, say, then no amount of tweaking the *robot's* accuracy is going to help. 

[pliedas]

There is one other company called Dynalog that performs calibration of robot as well as external axis. You can check them out at Dynalog-us.com.

[mookie]

as bert as skyfyre said...

due to machining tolerances and inaccuracies between the blueprint of the model and the actual machined robot product, the metris is used to measure the actual link lengths,etc of the robot and regenerate a new kinematic model based upon the results.

with that being said, it depends as well upon your required accuracy. the KUKA is a mechanically accurate robot but only to a degree, the metris/krypton solution measures the robot but i dont believe that it can account for backlash/slop for the gearing.

this would potentially help your situation between your robot and the external axis but remember that your also going to have slop on your slide as well. that slop will vary greatly depending on the manufacturer of your slide.

if your trying to take cad/catia/offline programming and apply it to your robot, I would also suggest maybe mounting some sort of reference points out on your actual base fixture to measure where the offline program says the base/part is in relationship to where it really is.

it just takes chipping it down to a component level one step at a time but as some of around here are proving, it can be done, but sometimes not without a couple of FU's in a meeting or needing to leave and vacation at a science fiction convention.