Hi all,

I'm exploring a method to make round ended channels using boolean Intersect and
curve blend to build the cutter. Here's what I've got so far. http://www.divshare.com/download/1265457-dea
I'd like to do some brainstorming to come up with the most efficient method...any ideas?


-jdk-

Hi Jesse, cool project, it's already looking great!

I have an idea for a different way to build the cutter, I think it's several fewer steps.

Looking at the final cutter result, it primarily has a cylindrical shape but ending with a rounded cut on the ends. So the idea behind this different approach is to follow the "big picture" of that shape and literally build a cylinder and cut rounded parts off.

Your method is a very incremental approach. I think that a lot of times this is a great method to use because it is just easier in general to make forward progress by small steps that build a piece at a time. It's kind of a more guaranteed way to finish any particular model. It doesn't always make things happen in the fewest number of steps, but a lot of times it will take more time to worry about the fewest number of steps than to just make incremental progress and finish it up! (anyway, I'm sure you're aware of this, just mentioning for others as well). So there is absolutely nothing wrong with your approach, it's kind of a real life approach. But of course it is nice to optimize things for a tutorial though too.

Anyway here is a more "big picture", "create it like the final thing looks" approach. I took the curves from your second step, used Edit/Separate on the semi-circle to throw away the straight line part, and mirrored that arc over, to get this curve setup:



Create the main body part by selecting the 2 large circles, and using Construct / Extrude, turning on "Both sides", and using a distance of 0.85 (the radius of your arcs there), then I deleted the 2 large circles which gets the main body piece:



Now select the cylinder, and run Construct / Boolean / Difference, and select the 2 arcs as the subtracing/cutting objects. Each arc will project all the way through the ring and you will end up with 4 solid pieces. Delete the 3 pieces you don't want, and you are left with this:



This sequence leverages MoI's auto-extrusion of planar shapes inside of the booleans to avoid extra steps.

You method kind of follows a "defensive" approach where your booleans are clearly pushing all the way through each other. That tends to guarantee successful booleans which is nice. The way I show above has booleans where the cutters and the objects to be cut skim right along each other, sometimes that can cause problems although MoI's geometry engine seems to handle this type of stuff pretty well especially with simple shapes, so I'm getting less hesitant as time goes on to try those types of things. But things definitely have to be snapped together nicely for this type of boolean to work well.

Another possible approach would be to focus on just the outside single surface of the cylinder, and trim that to the 2 arcs and then thicken it using Shell.

Hope this helps!

- Michael

Hi Michael.,

Wow, I had a hunch there was an easier way to make the channel cutter, but I didn't think it would be so simple.
I see what you mean about my approach being defensive...coming from a Rhino 3 perspective, booleans aren't
always my friend, so I'm in the habit of making sure that objects extend past the edge of where they have to cut..
that's why it didn't even occur to me to use the semi circles as boolean cutters

This is a great example of how MoI redefines NURBS modeling.

Thanks,

Jesse

>>coming from a Rhino 3 perspective, booleans aren't
>>always my friend, so I'm in the habit of making sure that objects
>>extend past the edge of where they have to cut..
>>that's why it didn't even occur to me to use the semi circles as boolean cutters

Hi Jesse, ...in some cases, Rhino V4 has trouble to make this kind of booleans with overlapping/sharing surfaces as well.. MoI seems to have more powerful geometry kernel related to booleans.... -Petr

Hi Petr,

I wasn't sure, but I think you're right...that's why I specified Rhino 3 rather than Rhino 4, but yeah, it seems like MoI is better at booleans than Rhino 4.

Regards,

Jesse

Yep, I had attempted to redo something I had started in MOI using the Rhino v4 demo - this was when I was working on that spacestation model - in addition, a simple curve projection onto the sphere didn't work for some reason where in MOI, it was working without a second thought!

I see a lot can be taken for granted when using MOI, which is a good thing because you end up in a modeling bliss instead of running into snags and dealing with annoying little errors and problems!

-Will

i haven't had any problems with booleans in Rhino 4. Even with shared edges, faces, etc I have no problems. there were some known issues while in beta, but it seems they fixed a lot of it before release...

jonah

Hi Jonah - it took me about 1 minute to come up with the attached simple example. This boolean union with shared surfaces and edges is calculated fine here in the current MoI beta, but fails in Rhino 4.0 (Eval, version 11-Jun-2007).

Strangely, it works fine in Rhino 3.0, so it seems to be an area where there has been some regressions. I don't doubt that they will eventually be able to tune it up, but this simple example should prove that others are not exaggerating above.


Boolean operations are proving to be a particularly strong area in MoI, both in reliability of calculations as well as interface. The ability to have solids and curves interact directly can save extra steps in many simple construction scenarios.

- Michael

Hi Michael. I hope you don't think I'm knocking Moi in any way. I still plan on buying it for several reasons, and one of them is to get past Rhino's intersection hiccups... So I pasted your test object into a Rhino file in mm units. And yes it did fail. Then i raised my tolerance to .01 and it worked. So i am curious what is the default tolerance of Moi ?

jonah

> Hi Michael. I hope you don't think I'm knocking Moi in any way.

Hi Jonah, not at all. I just sometimes see the attitude out there in general that Rhino is all-powerful in modeling and MoI has only a tiny fraction of the power. It's easy for things to be perceived in that way since it is certainly true that there are a lot of functions in Rhino that MoI does not do.

This was just a chance for me to illustrate that things are more complex than a simple "Rhino = powerful, MoI = not powerful" type of generalization.


The tolerance in MoI is 0.001.

That's very interesting that adjusting the tolerance makes it work in Rhino4. But I can't see any reason why it should be necessary - the parts that are touching are very precisely aligned, the overlapping surfaces are exact duplicates of one another (just one is trimmed down) and the edges that are overlapping all come from the same initial curve...

I did an experiment with reconstructing the surfaces and solids completely within Rhino4 from the original curves, to see if that made any difference. However, I was unable to complete all the steps because the boolean intersection between the top slab and the sphere failed in Rhino4 (attached here also), it's another one that works in MoI and also in Rhino3 as well. This intersection failure does not seem to get fixed by adjusting the tolerance like the previous one...

- Michael

I took a look at this new example file, and the intersection is incomplete where the sphere's seam intersects the side face of the slab. Again back to the intersection weirdness of Rhino. I guess it's true what they say- you fix one thing, and break a few more in the process... :)

jonah