03-10-2023, 07:52 AM
(02-01-2023, 09:09 PM)Albert Wrote:It is better if you contact us by email at info@www.sinclairbody.com to help you test TwinTool.
This is possible using a Keyence sensor with a flat contact point (25 mm recommended) and connected to a computer using USB, Socket or Ethernet/IP.
Hi Albert,
I tested TwinTool in the last two days. Unfortunately I don't get good results. My setup is:
- KUKA KR 10 R1100-2
- RoboDK 5.5.4.23146
- TwinTool 1.0.0
——日本基恩士与平面接触poi GT2-P12 / USBnt 25 mm, assembled on a magnet holder. I guess this could be a problem, but it seems stiff enough.
I used the same tool for testing as in the previous post:
Quote:1. Plane 1300 mm x 800 mm: [ -0.780005, -0.587233, 298.278491, 0.000000, 0.000000, 0.000000 ], Errors: mean 0.119, std 0.08, max 0.33
2. Plane 200 mm x 150 mm: [ -0.911481, -0.407068, 298.067150, 0.000000, 0.000000, 0.000000 ], Errors: mean 0.148, std 0.148, max 0.546
3. Tip-to-tip: [ -1.351774, -0.330700, 299.312497, 0.000000, 0.000000, 0.000000 ], Errors: mean 0.353, std 0.176, max 0.628
So Z should be around 299mm. While measuring I can see the Z Value "jump" by around 6-7mm and the end result is more or less random between 293mm and 301mm in the Z Axis after 217 measuring points. It seems that it depends highly on the max. error. 0.8mm max. error seems to be the threshold. When the value is smaller, the TCP Z Axis is around 299-300mm, when it is above it changes to 293-294mm.
I already increased the tolerance stabilization time to 1000 ms and the maximum joint ranges, but it does not seem to help.
This are the results:
Is it possible to somehow remove the outliers or are there any other tweaks I could try?
Otherwise I really like the idea of the "Calib XYZ by plane" method to get repeatable results independent of the person who does the calibration. But I am concerned with the accuracy since it seems to be off by 1 mm in the Z axis. Is this method supposed to be less accurate but more repeatable in comparison to tip-to-tip? Or maybe I did something wrong? I know there are many parameters, but it seems odd to get such different results with a new robot but different algorithms.
