Bending and scale

Hi Daniel,

you mentioned in a recent discussion that Kangaroo 2.5 addresses an old bug where bending/angle simulations are affected by scale.

I have just updated and can see a number of exciting new K2 components, but I still have the bending and scale issue. Any thoughts?

Attached is a file, where I am reducing the scale to a 20th of the original. It solves in a few seconds using 16k iterations, while the original (red colour) takes many, many minutes to achieve the same (?)

Also, I was wondering how this bug may affect K2Eng definitions.


190304 Bending and (26.7 KB)

Hi Mathias,

That’s not a bug…
I should have clarified - the updated angle goal is still giving numerically correct forces when using real Young’s modulus.
Since the strength depends on the cross section, this does mean if you scale your model by X, you’d need to scale the angle strength by X^2 to get the same convergence behaviour.
As far as I know, K2Eng still uses the same angle formulation as was in the previous version of Kangaroo, so will converge to the same thing, but slow for large scales, and potentially unstable for very small scales.