Hi there, did anyone here succeeded in reducing or getting rid of backlash. Especially when you mill a flat surface, backlash is a very disappointing trouble as it leaves marks over the surface. Any advise about that to get closer to a gantry surface quality?
Very Best,
Marco
Backlash reduction during milling
-
Marco -
December 19, 2016 at 6:27 PM -
Thread is Resolved
-
Usually concern is A1 and problem is solved by:
a) mounting robot tilted slightly (10 deg)
b) squaring world coordinate system by entering tilt angle into $robroot -
Usually concern is A1 and problem is solved by:
a) mounting robot tilted slightly (10 deg)
b) squaring world coordinatte system by entering tilt angle into $robrootWell, in may case I would not say that the issue is with the A1. My spindle axis is mounted parallel to the x flange axis and sometimes I notice strange movements, like pitching, given be the A5 (I would say). My TCP is 450 mm away from the center of the flange (due to a tool changer+spindle support+spindle assembly which make this strange movements quite evident. I calibrated the load with kukaload determination software and my load is admitted (52kg)...Is there any other thing I could do to correct this strange pitching?
-
Well, first you need to compare your payload to the robot's load capacity. If you are using (for example) a KR150, you only have the robot loaded to about 30% capacity and this can make the Load-Ident results suspect.
Also, it's best to weight your payload carefully and then apply that value to the M variable in LOAD_DATA, because the robot's ability to accurately measure the payload mass is not 100% trustworthy, especially when the robot is underloaded.
So, best practice: obtain the Mass value for the tool separately, determine the CoG values empirically or via FEA, and only use Load-Ident for the Inertial values. -
For flat surfaces, the best results i achieved were by milling in only one way (climbing or traditional)
When using extended spindle holder any error you may have is multiplying... -
Yes, robots are far more vulnerable to backlash than typical gantry-style milling machines. Any axis that reverses direction during a cut will leave marks. For example, imagine a robot making a linear move along the World Y axis, essentially drawing a straight line from left to right (Y position value going from + to -)in front of itself. Over the course of this motion, A1, A4, A5, and A6 will probably maintain a constant direction of motion, though at varying rates. But A2 and A3 will both reverse direction as A1 passes through 0deg. And this will leave behind definite evidence in the cut being made.
That's an easy example, but A2 and A3 are some of the most backlash-immune axes on the robot, thanks to gravity and the counterbalance (although A2 has definite issues near -90deg). The wrist axes are vulnerable, and any reversal of A1 is likely to have large effects simply because A1's backlash usually has the longest "lever arm" to act over. -
Yes, robots are far more vulnerable to backlash than typical gantry-style milling machines. Any axis that reverses direction during a cut will leave marks. For example, imagine a robot making a linear move along the World Y axis, essentially drawing a straight line from left to right (Y position value going from + to -)in front of itself. Over the course of this motion, A1, A4, A5, and A6 will probably maintain a constant direction of motion, though at varying rates. But A2 and A3 will both reverse direction as A1 passes through 0deg. And this will leave behind definite evidence in the cut being made.
That's an easy example, but A2 and A3 are some of the most backlash-immune axes on the robot, thanks to gravity and the counterbalance (although A2 has definite issues near -90deg). The wrist axes are vulnerable, and any reversal of A1 is likely to have large effects simply because A1's backlash usually has the longest "lever arm" to act over.Thanks for this explanations. So it is not about some trick to tweak this issue. So I should check again load, if I can set my Sprutacam as to avoid the robot passing by the zero of the all axis. I will also check the effect of climbing and traditional setup as I'm actually using both to save some milling time reducing idle time.
-
Well, it's not as simple as "avoiding 0" -- directional reversals can occur at many different angles, depending on the robot's current pose, direction of motion, and the TCP mounting. The O-Scope function on the teach pendant can be a lifesaver, here -- you can set it up to record the axis velocity of all 6 axes over a time span (which can be triggerd from inside your program) and then look at the traces to determine which axes reversed direction and when.
-
Usually concern is A1 and problem is solved by:
a) mounting robot tilted slightly (10 deg)
b) squaring world coordinate system by entering tilt angle into $robrootI'm interested in this concept, and having the same problems.. the idea must be to have A1 always be fighting uphill while milling?
Just to confirm the idea, if my cell is set up that the robot spindle approaches from world +Y, and the table is directly in front of the robot (2000, 0, 0 or something), I could just tilt the robot base in a slight (-X) rotation?
Such a simple solution if it's that easy! There's a fitting saying from my old mechanisms professor: "it's really hard to make a simple design."
Advertising from our partners
