I have an object and some edges seem to fillet fine, but others just refuse to do anything when I apply a fillet to them.

These edges will fillet



But these won't.



Anyone have an idea how to make it work, or am I just out of luck?

Thanks.

Hi Sun, there are some little slivery surfaces in the side walls of those holes, like here:



Having little slivery fragments like that is not good for filleting - it makes for little fragmented fillet pieces trying to be constructed, and it's particularly difficult for it if the little fragmented pieces are not smooth with one another because corner patch pieces are created by the fillter where 2 fillet pieces meet at a crease. That's particularly bad if the crease is only very shallow like 5 or 10 degrees tangent deviation because it makes for a tiny corner piece as well.

So probably those holes will need to be remodeled to be a cleaner more simple surface structure that is just made up of one surface instead of being fragmented into numerous slivers like that. If you can post your model I can demonstrate what I mean by that.

- Michael

The file is pretty big, but I'll try and clean it out a bit and attach it... Couldn't do it without compressing it so I left extra things in the file that will give some idea of how it was constructed.

Hi Sun, so it looks like the holes that won't fillet are formed from lofts between sets of these 2 curves like this?



The problem is that those curves are made up of several segments, if you select one of them and run Edit > Separate, you can then select the various segment pieces like this:



How were these curves created, did they come from 2 surfaces being intersected with each other? If so then the surface intersector did not do a great job with that and it needs some cleanup on it before using it for construction. The easiest way to clean it up is to run the Rebuild command on it:
http://moi3d.com/2.0/docs/moi_command_reference10.htm#rebuild

Running rebuild on those curves using a tolerance of 0.001 (and check "Delete input objects") will make a new version of them that is all one single segment instead of made up of multiple sub segments. Those will then correspondingly create just a single surface from the loft instead of multiple surface fragments, and that should help a lot with filleting.

You probably want to do that with the other ones that were actually filleting as well since having those hole side walls just as single surfaces will make for a cleaner and lighter model overall.

- Michael

Hi Sun, here are some cutting surfaces that have been simplified as described above.

I have not had a chance yet to test cutting your main piece with these and doing the filleting, but it should have a much better chance of working with these as the side wall cuts instead of the other previous ones.

You should hopefully be able to select these cutting pieces, then do a Transform > Array > Circular to replicate them, then select your base ring solid (when it was plain without any cuts in it yet), and also these pieces and do a Boolean Merge operation to do all the cutting.

- Michael

Yes, the basis for those curves was originally a surface intersection. The surface intersection curve was then scaled to create two other curves (a bigger one outside the ring and one smaller one inside the ring) and then those two curves where used to create the loft which was then used to boolean out the holes. I'll experiment with the 'rebuild' as you've suggested and see what I can do.

Thanks!

*edit* Oh wow, I just saw your 2nd post with the file attached. Thanks again!

Hi Sun, so Transform > Array > Circular with 24 items should work to replicate those surfaces into the pattern you want - it's easier to replicate the surfaces that way than to have a whole lot of separate curves created that you then need to loft in pairs.

So then take your main uncut ring base object and select it and also select all the cutting surface objects too and run Construct > Boolean > Merge - that will maybe crunch away for a minute or two and slice up everything. Then you can select the main body piece you want to keep, and do a Select > Invert to flip the selection and delete all the little pieces. That's how I got the attached 3DM model version.

By the way, it's handy to put invert selection on a keyboard shortcut for I so you can just hit I to flip the selection, to do that go to Options > Shortcut keys and add in a new entry, put in I for the key and for the command part paste in the following:
script:moi.geometryDatabase.invertSelection();

So with this simplified surface structure it does seem to fillet much better, like this:


So the tricky part is that the surface intersection command produced some curves that were kind of a bit messy to be used directly for creating new surfaces off of since they were made up of several segments. I think that in the future I'm going to automatically run the results of the intersection through that Rebuild process but for now you'll need to run them through Rebuild as an extra manual step in order to get a more simple curve structure that will then create some more simple surface structure when you build with them.

Hope this helps!

- Michael

I'll try as you recommend, except I want to keep all the little pieces too. Thanks for all the additional tips too. I didn't know about boolean 'merge' so I was previously using multiple operations to cut the holes and generate the intersection pieces. Maybe one of these days I'll learn to RTFM instead of just diving right in. lol.

Hi Sun, yeah boolean merge is convenient if you want to slice a solid up into multiple pieces and keep all the pieces, you just select everything all at once before running it.

So just skip the part in the above steps about doing the invert selection and the delete to keep all the pieces around, let me know if you have any problems getting that.

- Michael

I tried it and it certainly cleaned up the curves and resultant surfaces, but some of the edges still won't fillet.

Hi Sun, what fillet radius are you trying to use?

Since this piece has rotational symmetry, it should be possible to replicate one successfully filleted piece around and trim it into place, if you can let me know what fillet radius you want I can get it on the model that way.

One other thing that you may be running into is also related to short fillet segments - when you have 2 curved surfaces meeting each other instead of having one curved and one straight, then it is actually not so great to have edges between them that come to a single common point like this:



When each side of an edge like that is a curved surface, the fillets will actually get constructed in some separate segments in a case like that, see this previous post for an illustration:
http://moi3d.com/forum/index.php?webtag=MOI&msg=4023.10

That's because fillet surfaces are created at an offset of the current existing surfaces, and when you've got curved surfaces on either side it means that the fillet pieces on either side of that point where the edges touch each other in the original surfaces will be offset slightly, leaving a kind of small gap to be filled. Small gap situations do tend to be a complex area to handle where the filleter can get confused more easily than when it deals with larger fillet sections.


If those side pieces had a "seam edge" that was not in exactly the same spot as the edge on the outer ring surface, that would actually be easier to fillet. That can be done by dragging the seam point markers that show up in the Loft command when creating the loft initially - when you're in the loft command inside the viewport there will be 2 points that show up that you can drag around to control the starting connection point between 2 closed loft profiles.

But I think it should be possible to replicate any one successful fillet piece around to finish off your piece instead of doing any changes to it, if you can let me know what fillet radius you want I can make that for you.

- Michael

Hi Sun, so another note - if you can eliminate these edges completely:



That would be another simplification and cleaning step for the model, and would also simplify the fillet calculations quite a bit as well since there would be fewer edges colliding with one another throughout the model.

It looks like that edge is a juncture between 2 revolved pieces:



But this shape could be more simply and cleanly formed as just one larger revolved surface there instead of 2 identical ones joined together.

You can get that simpler shape there by deleting the existing surfaces there, drawing in a vertical line segment, doing a revolve (turn end caps off), and then joining the revolved surface into the other pieces, that's what I did to get the base shape in the attached 3DM file.

Having that as the base model will reduce quite a lot of edge junctures and small fillet pieces that the fillter would otherwise have to deal with.

As you've seen, filleting tends to be a rather sensitive area - filleting has a sequence of several complex operations that it has to do in a row, like offsetting surfaces, intersecting those offsets, building fillet pieces and potentially extending them and intersecting the extended pieces with each other, building corner patches between areas with kinks, ... it can run into problems if there is difficulty in any one of those steps. So the more simple structure (generally meaning the fewest edges and faces that are necessary) that you give it to work with, the better.

- Michael

Wow, I can see now that I need to consider my construction methods a bit more carefully. I'll do some rebuilding and see how it works out.

Oh, I was trying to apply a 0.05 fillet. But I'm not sure if that will be the final or what. My intent is to try and export into modo and sub-d it so I'll probably have to do a few experiments to find a combination of settings that works.

Hi Sun, I've attached a new "fully cleaned" 3DM file version that seems to be able to fillet everything ok at radius = 0.05, giving this:



I've attached the base unfilleted version so you will be able to apply other fillet values if you want.

Here's a tip on building the selection for the fillet more easily - start by zooming in a bit and selecting just one of the side-wall faces of the model like this one:



With that one face selected it will make window selection target faces also, and with just a couple of strategic window selects you can get all the other side wall faces.

Do one window select like this:



And another window select like this:



And then that will end up with all the side wall pieces selected like this:



You can now run fillet directly after that last selection - when you give fillet a face selection all the edges belonging to those faces will be targeted for selection.

In some cases like this it can be more convenient to select faces for the input to fillet rather than edges, since there are kind of fewer face pieces than edges to wrangle with.

Note that on the window selection steps you want to start the window on the let and drag it towards the right. By dragging in the left-to-right direction, only things completely contained inside the window will be targeted for selection. If you start on the right-hand side and drag towards the left you will get what is called a "crossing" window (the outline box will show as dashed instead of solid in crossing mode) which grabs anything that intersects the window even just partially. That crossing mode can be useful in other situations but for this one you want to do 2 non-crossing window selections because that then avoids the wider faces at the top and bottom from being selected.


When you then do the fillet it will crunch away for possibly a few minutes but then seems to get everything done ok this time.


Hope this helps!

- Michael

Don't try and apply sub-d in Modo. There is no need and it won't work properly.

Thank you for all the help and great tips Michael. That looks great now.

SteveMacc,

Yeah, getting a meshed model that can be sub-d is tricky. The problem with leaving it as plain polygons is that while the vertex normals will smooth the look of the surfaces, you can see the polygonal shape on the edges sometimes if you view them from the right angle. Right now I'm experimenting with getting meshes from MOI into MODO. Hopefully if I build the model right in MOI I can avoid having to sub-d, but I'm not sure that will always be possible.

Right now this model looks good when I render it whether I sub-d it or not, so I'm inclined not to.

Hi Sun - re: polygonal silhouettes - the main method to eliminate those is to increase the density of the export from MoI, you can just move the slider in the meshing options dialog towards the "More polygons" side to produce a finer mesh with more polygons in it and no other work will be involved.

Trying to do a sub-d refinement of the MoI generated mesh on the other hand is going to involve quite a lot of additional work because sub-d smoothing requires a particular kind of polygon topology to work well. MoI generates a kind of sparse efficient polygon topology that resembles the original NURBS model and it's a totally different kind of topology than what sub-d actually needs.

Here's a quick example object to help illustrate it more clearly:



So note there that MoI's mesher generates a single planar n-gon on the top there, which matches the original object which had just a single planar surface for the top face. That's a great result for getting a clean and efficient polygon structure for uv mapping or rendering but it's very very different than the kind of topology that you need for sub-d smoothing.

For use as a sub-d cage you would want that top piece to be tiled into a bunch of quads that radiate out from the outer edge and try to have some good organization where they then collide into each other. That would mean generating quite a lot more little polygons for that case than what MoI's current mesher does, and it would mean producing the mesh in a completely different way.

So it's generally not feasible to apply sub-d smoothing to polygon output from MoI expect in some special cases. If you want to apply sub-d smoothing you're probably better off creating the polygons by hand because then you'll be able to control the specific kind of topology that is required for the sub-d smoothing - either that or you can use a retopologizing tool like TopoGun or some of the tools in 3D Coat but you should expect to have quite a bit of work and effort involved in doing that.

Meanwhile for making a denser mesh when exporting from MoI only involves moving a slider over a bit so that's a couple of seconds of work as compared to a huge amount of work for re-topo and then sub-d smoothing.

- Michael

Yes, I've been experimenting with the various meshing parameters to come up with something that doesn't need to be sub-d, or if that doesn't work well, will sub-d in an acceptable way. I know that making a mesh suitable for sub-d is going to be tricky, and may require some editing in modo.

My own experience with Moi to Modo is that if you set the mesh angle to 6, you will not get any faceting in Modo. You can then adjust the other parameters to reduce the poly count. "Smaller Than" is a useful one where you have small fillets. The rendered poly count is similar to a mesh constructed in Modo with a subdivision level high enough to avoid faceting.