Hadn’t had any time to dig into this until this afternoon (work and all). First up, there are a lot of overlapping model lines as if you attempted to make them 3D at one point) which you’re projecting from that is causing part of the issue with families not placing correctly the first time out. Also some elements were complete circles, which will cause issues with placing 3 part adaptive components. I split these into two arcs to resolve that issue. Remember that as you go along, if you feed the computer garbage you’re only going to get garbage in return. Keep things neater and you’ll have better results. The length of some arcs was likely still too significant to work well, but I figured I’d let it rip anyway.
After cleaning these up that stuff I looked at your graph. There was a LOT of unnecessary list gymnastics (break apart lists, make new ones, flatten lists, etc). After reducing that and simplifying where I could, placement of the adaptive voids worked for all but 7 of the 1900 families. Other than being slow, which should be expected when dealing with this much geometry, and those seven issues (which may have been complete circles), and some of the larger circles had other issues with how the adaptive component formed, but overall it worked.
I then changed the lacing from shorted to longest on the Element.CutGeometry node and sure enough, every one of the place void families attempted to cut the floor. Not all worked due to some limitations of Revit’s geometry engine (which really isn’t meant for this degree of complexity), but it attempted to work and all but 253 were successful. Note that it also took a REALLY, REALLY, REALLY long time to run and basically caused my GPU to spend 20 minutes or so regenerating views and calculating new cuts.
I’d recommend that you either:
- Look into using another program for this level of visualization, such as 3DS Max
- Shift to using multiple floors and/or parts,
- Consider using a bumpmap properly formed and setup to get the visibility you’re after.
- Ditch the idea of adaptive components all together and use a generic model in place from a computationally driven set of lines extrusions.
A combination of #1 and #3 will likely look best.