Kuka PRC/ Incremental Sheet Metal Forming

Hi, I’m developing a definition for incremental sheet metal forming. I don’t have a problem with the toolpath, but I’m having a problem with simulating a metal deformation in parallel. Does somebody have a good example how to simulate it during the tool path simulation?

Thank you in advance!

Mainly I’m interested in the top left image.

Hello Katya,

Cross posted from the Robots in Architecture Forum: For our experiments with incremental sheet forming (using PET, not metal) we went the other, easier way and designed a form, which then defined the toolpaths. See e.g. at a workshop in Shanghai: https://vimeo.com/158804677
Approximating material behaviour in parallel to the simulation might work with Kangaroo, but I wouldn’t expect to achieve an accurate simulation, as the material interactions are really complex. E.g. with PET, heat was a very significant factor, which you cannot really simulate that well in realtime.
The nicest approach towards material simulation of incremental sheet metal forming are the machine learning strategies that Mateusz developed for “A Bridge Too Far” at CITA. The disadvantages as with all ML being that you first need to collect plenty good data.

Let me know if you need any specific data from KUKA|prc to link it with your material simulation!


An idea of an approach, which might work. This looks like the material is machine hammered in sequence, with the sheet metal backed by either a hard foam, or a medium durometer polyurethane. A way of simulating this might be to treat the backing as a constant volume, so when an impression is made, the volume of the depression is moved into the adjacent non impressed areas, ‘inflating’ those areas, I think Dan Piker has done some excellent work, published on this forum, about mesh inflation. So, yes, could be done,

The geometric deformation/stress analysis method developed by @dave_stasiuk for the preceding CITA project Stressed Skins might serve as an alternative. Specifically the elliptical analysis seen here:

As I recall there were several papers presenting this approach, same with @MateuszZwierzycki’s ML approach (was in the latest Fabricate proceedings I believe).


It’s usually not hammered, but just deformed through the pressure of the tool. The process looks very similar to milling, you just do not subtract material, but push it to the side. Sometimes two machines/robots are used so that a panel can be processed from both sides at the same time.
I’m not sure if a backing of hard foam would be a good solution - XPS can take >30tons/m² (of course it would be a point load for the metal forming, but still).
For a rough approximation I would probably start with a grid of points, and when the tool touches a point it gets moved down in Z to the height of the tip of the tool. The distance between points should give you an idea of the stretching at each point, and it will be fast enough to run in real-time. The grid could then be coupled with Kangaroo for some material simulation, but you will still not be able to predict the exact bounce-back of the metal, folding, and other effects.
But I guess that you can get nice pictures with color gradients out of it :wink:


Can you share information about generation the toolpath ?
I am a new architect and i have to make some custom made lights from brass.