Polybevel on two merged objects . Topology Help

[lajosch]

Hi,

I am new to Houdini and cannot solve this problem:

I created two open circle objects (circle > sweep > merge). The dimensions are identical.

I want to combine them and then bevel the edges. It looks fine before I apply the polybevel (Image: Polybevel_01.1 and 01.2), but with the polybevel applied, the result looks broken (Polybevel_2.1 and 02.2).

It makes no difference if I merge the objects, or if I use a boolean node.

Can anyone help me with this?

Thanks

[Sepu]

use the bridge sop to connect the 2 polys properly then bevel so is one piece or you can also convert them to Vdbs and back to polys 

[lajosch]

Hi Sepu,

thanks for your reply.

Both circles are animated, so I cannot bridge the polys.

I tried vdbs, but the scene gets very heavy.

I will upload a scene file, so it makes more sense.

Thanks

[lacer8]

Merge only combines two (or more) objects. It isn't doing any welding of points or boolean functions. Not knowing exactly what you're trying to do makes it tough to answer, but if VDB's work for what you're trying to achieve, you might get more success increasing the voxel size before converting back to polys. The larger the voxels, the less detailed the volume, and the less heavy it'll be as a mesh after converting back. 

[lajosch]

Hi Peter,

thanks. I am just wondering, why there is no difference between a merge and a boolean, when the boolean is supposed to "create" a new object.

So attached is the hip file.

Later on I want to make a lot of copies of both circles in the x and y axis (like 200) so that they create a pattern. And they have to keep their form. That is why I think VDB's would be overkill, because the voxel size would have to be pretty small.

Thanks again for your help.

 

 

Polybevel.hip

[AntoineSfx]

17 hours ago, lajosch said:

Hi,

I am new to Houdini and cannot solve this problem:

I created two open circle objects (circle > sweep > merge). The dimensions are identical.

I want to combine them and then bevel the edges. It looks fine before I apply the polybevel (Image: Polybevel_01.1 and 01.2), but with the polybevel applied, the result looks broken (Polybevel_2.1 and 02.2).

It makes no difference if I merge the objects, or if I use a boolean node.

Can anyone help me with this?

Thanks

Boolean SOP is too unstable to do this properly. On top of that, Polybevel is also very numerically unstable.

It will try its best to slide the bevel on edges, but with an intersection of two polygonal (meaning a lot of points and straight edges), it will lead to a different result depending on the relative locations of the two intersection objects. It will break  / jump as soon as there is a topology change in the result, and it will if you animate it.

Also, you're using two beams of the exact same width, which leads to a bad edge case in boolean where you end up triangulating a set of coplanar points, which is guaranteed to give random results ( look at the tooltip help on Random seed in triangulate2d SOP parameter: I think it's related, not necessarily in term of internal implementation, but the general idea)

Can you modify the scene so that one of the beam is larger than the other by the amount of bevel you intent to put on the seam (see abseams in Boolean SOP)

If you animate it, can you retrain it so that the primitives that intersect in each input of the boolean SOP remain the same ? I know it's restrictive..

Also, see the other output groups in Boolean SOP.

 

 

 

[lajosch]

Hi Antoine,

thanks a lot for taking the time. That is a lot to digest

I cannot change the width of one of the beams unfortunately, because this is supposed to look like it is one piece. And with a different width the bevel would not be even across both of the beams. I will give it a try though. And I will check out your other suggestions as well. Thanks a lot - much appreciated!

[Aizatulin]

Hi,

boolean seems to work for me, you can also group->dissolve edges with small angle and fuse close points (optional).

Polybevel_bool.hipnc

[lajosch]

Wow! Thanks Aizatulin.

While trying to fix this problem I had some OK result with the Redshift RS Round Corners node.

But this would have been just a compromise.

Your solution is so much better and pretty much what I was looking for.

Thanks for the effort! You made my day

Have a nice weekend.