Sounds like a singularity issue. Try to avoid A5 close to zero.
Fubini
There is nothing to fix. This is normal system behaviour when doing a bco motion on a spline. You probably did a touch up, which triggers a bco move. Reset the program and run from the beginning.
Fubini
You could reduce your velocities.
Also you should read the read first post and try to find documentation. After careful studying try to be much more precise in your problem description.
And most important try to get at least basic KUKA training. This forum can not be a substitute for that.
Fubini
Usually it is covered in the programming manual or in more advanced detail in the system integrators manuals.
Fubini
where do you draw your conclusions from?
I would try the suggestions above by reducing your velocity during certain moves. My experience, every time I program something in Kuka that has more of a majestic curve moves, it takes more time trying to figure out jerk, velocity, etc and always prone to errors. That’s where I move to a graphical interface where I can see the move, manipulate curves and so on. Have you ever heard of Mimic? Check it out...mimicformaya.com
Spot welds, or seam welds? Two very different things. "More accurate" is entirely a matter of definition, as well. In absolute terms, Spline motion and "classic" motion are equally accurate and repeatable, but the devil is in the details.
Put very (overly?) simply, Spline motion approximates the motion between points, while "classic" motion approximates motion at/near points.
Spline motion was created to address a shortcoming in "classic" robot motion for continuous-path operations -- basically, that "classic" motion achieved path smoothing by passing near points, but not explicitly through them. So if the points A,B,C represented a sharp 90-deg corner, and B was approximated, the TCP would "cut the corner" past B while executing the sequence A->B->C.
So, a path for laying down a line of adhesive on a specific, non-straight path, would be hand-taught using trial and error, teaching points "off path" and adjusting their approximation radii until the adhesive was laid down where the programmer wanted it. This worked well enough for decades, but was expensive in terms of man-hours. It was also very hostile to CAD-to-path programming styles.
Spline motion uses a completely different model for path smoothing, one that is much more computationally intensive. In general terms, it pre-calculates a smooth path that passes through every point in the Spline block, and this path ideally maintains a constant TCP velocity throughout. One potential drawback is that multiple points too close together can create very odd motions between them -- the calculated motions are less constrained, and therefor less predictable. Spline motions can also encounter additional mathematical singularities that "classic" motions generally aren't bothered with.
all of motion instructions are ideal for something and poor for something else.
spline is suited for complex smooth curves.
PTP motion is the robot's most "natural" motion. But that's b/c it only cares about the start and end positions of the motion, and does little to regulate the "in between" motion. That is, if you make a PTP move from P1 to P2, the robot calculates how much each axis must rotate to get to P2, adjusts each axis's speed so that all 6 axes will start and stop at the same time, and then executes that motion. What space the arm passes through en route is not measured or controlled. As such, the path the TCP traces through space is not controlled or predictable. So, depending on S&T, singularity conditions or wind-up, it's possible to have two points 1 inch apart that will cause the wrist to completely flip over moving between them with a PTP motion.
For any type of "seam" motion, you need interpolated motions (LIN, CIRC, Spline), or a great number of closely-spaced PTP motions. Technically, any interpolated motion is made up of a large number of PTP motions spaced only a few microns apart, handled at the motion-planner level, but writing an application-level program that way is self-defeating.
