Surface normals / extrusion curves

Dear forum,

I have a question. I have a curve which serves as a ground plane for an extruded box (see attachment). When I extrude the curve, a box is created. When I explode the box and evaluate the surface normals, the normals are pointing outwards. But if I then offset these faces, some of them will move inwards, some of them outwards. I can't just flip the normals, because that won't solve the problem.

Solutions, anyone?

Attachments:

normals.3dm, 54 KB
surface normals.gh, 4 KB

Replies to This Discussion

Permalink Reply by Faberyayo on June 7, 2013 at 6:50am

I didn't check your files yet, but a possible solution is to:

- offset the faces both inwards and outwards (+distance, -distance)

- get area centroid of the faces

- get volume centroid of the box

- get the distance of the area centroids from the volume centroid

- sort the faces based on this distance

- pick the desired faces (closest/inwards offset or farthest/outwards offset)

Permalink Reply by Faberyayo on June 7, 2013 at 7:19am

Permalink Reply by jkanters on June 10, 2013 at 1:11am

Thanks for your input! It was easier to make a box of it first.. problem solved!

Permalink Reply by Pieter Segeren on June 7, 2013 at 9:34am

Hi jkanters, I can't say why this is happening, but I found an easy solution for it:

Permalink Reply by jkanters on June 10, 2013 at 1:10am

Thanks Pieter.. sometimes things are really simple :)

Permalink Reply by Faberyayo on June 10, 2013 at 1:30am

Touché!

Permalink Reply by David Stasiuk on June 10, 2013 at 2:39am

I think the offset direction for a surface is tied to the direction of the face loop...so the edge curve direction is the driver. It looks like making a box out of it resets these directions.

Permalink Reply by jkanters on June 10, 2013 at 3:01am

True.. or it least it seems so.

The problem is now that if your curve is NOT rectangular, the solution proposed by Pieter does not work anymore.. so I most probably need to do it more complicated in the way Faberyayo suggested.

Permalink Reply by David Stasiuk on June 10, 2013 at 3:19am

yes...I imagine you also could test for the direction of the edge loop, possibly using a plane constructed from the volume centroid to the face centroid...might make for a lighter calculation if you're doing it for a ton of objects at once...but just doing offsets and testing for proximity is probably easier.

Permalink Reply by Faberyayo on June 10, 2013 at 3:40am

True, I am not a big fan of too many area/volume centroid computations either.

I just noticed that if the normals of the faces are in the desired directions, then you can also do this: