Hi, all! I ran into an issue that has me stumped.

Here's the part:



What I'd like to do is Shell the selected area at 1mm. At present the largest wall thickness I can get is .5mm. And that causes strange geometry around the fillets.

The model is a solid, so that's not the issue. I'm wondering if the geometry is flawed around the fillet detail. Or maybe the angles are too steep for it to make a wall of the desired thickness?

Not sure, since I'm not clear on what the command requires in order to work.

I suppose I could do this the hard way, but I'd rather understand how to better utilize the supplied tool.

.3dm file is attached for your perusal.

Thanks,
Patrick

Hi Patrick:try to shell just half cylinder without the four cuts,then revolve four closed profiles filleted and then bool subtract them
you'll see that 1mm is too thin

Hi Patrick, the thickness of 0.5 giving weird results looks like a bug in the geometry library - something is messed up in the resulting trim curves of the generated object and that makes for strange "leaky" type of shapes. I will try to investigate it, but it's a pretty tricky area to mess around with without a huge amount of time involved.

I do have an updated version of the geometry library that I'm going to be rolling out with the next v3 beta release, hopefully it will fix some various issues.

A distance of 1.0 is probably not going to work because it involves some faces being entirely consumed away to nothing, like this one here is almost consumed at 0.5mm already:




The geometry library that MoI uses is not so good at handling cases where entire faces disappear like that, it's just not a strong area of the library currently. Some other CAD programs like ones based on the ACIS kernel should handle that case better probably.

- Michael

In cases like these Michael, is it worth scaling the model up, then doing the routine(s) with questionable abilities to shrink the model back to scale, or would the geometry still goof ? I only ask as i have had this issue a few times but only just thought of enlarging and shrinking as a work around.

Martin

Hi Martin, usually scaling up is something that can help when you've got some pretty small features in the object, like pieces with a radius or length that starts to get close to around 0.001 units.

I wouldn't have thought that it would help for this particular case, but actually it does! Scaling this object up by 10 times in size (select it, run Transform > Scale and then type in 0 and push Enter to set the scale origin to 0,0,0 and type in 10 and push enter to scale up by 10 times) then doing a shell of 5mm instead of 0.5 does work ok.

But really the object does not have small enough features to where I would normally think that would help.

I think it's something of a matter of luck in this case that it happens to avoid the particular bug - it does appear that the update to the geometry library fixes a UV trim curve generation problem that should fix this case at its original scale and also fix a variety of other boundary mangling cases as well probably, but there is also a separate side effect that it's only generating the inner core of the shell and not booleaning it out from the main solid automatically so I need to investigate that.

- Michael

Fascinating stuff.

I ended up doing it the hard way, which didn't turn out to be so hard after all. In the process I realized that 1mm is actually too thin, so I increased it to 2mm for stability. This is a part that gets screwed on rather tightly.



The underside isn't meant to be seen, so it doesn't matter what it looks like. I was just trying to remove as much material as possible for 3D printing purposes.

Michael, I know it can be frustrating at times, but just know that you're the hero for tons of us out here. I can't wait to see the new update. Keep up the great work.

Thanks to everyone for the ideas and assistance!
Patrick

Hi Patrick, your result looks great! Yup sometimes when the "automatic" type processes fail or get confused your best bet is to just manually construct a cutting piece and "do it yourself".

I think the fix to the UV trim curve generation with the latest geometry library should probably be a big help in general for robustness, there were a couple of different bugs in that area causing mangled boundaries that could happen in a variety of different operations.

It looks like I've tracked down the area of the library that changed that made the final boolean step of the shell not work quite correctly with the updated library, so I can undo that particular change but keep the other fixes.

That's one thing that's really pretty frustrating about the geometry library being updated, it's not unusual for some things that get changed to help out in one particular case but then cause problems in other cases...

I was going to wait until cramming in a couple of more features before releasing a new beta, but I think that maybe I'll go ahead and roll one out now, the geometry library update and an update to Network are enough to merit a new release now I guess, I will try to roll out one later tonight.

In that new release the 0.5mm thick shell of your original posted file should then work ok.

Thanks for posting the example file so I could test with it!

- Michael

Hello,

You could also use the curve offset and Boolean out the inside to a specific width, remembering to carve out the offset along the flat plane where the hole. One and two millimeter wall thicknesses are in the pic.

Hi Patrick, almost ready with a new beta with those geometry library bug fixes, but it look like it will take me until sometime tomorrow to finish it instead of tonight.

- Michael

Cool. :-)

Peace,
Rich_Art. ;-)

Michael - Broadband routers wait with baited bytes ....

Martin

If you design the 2 pieces in meters then at the end make a scale / 100
Does you obtain the same result than in mm ?
Does Moi's functions will be helped?

Hi Pilou,

> If you design the 2 pieces in meters then at the end make a scale / 100
> Does you obtain the same result than in mm ?
> Does Moi's functions will be helped?

You could do something like that - if you switch the unit system in Moi it will apply a scale for you.

But anyway in this particular case any scaling should not be necessary, I think it's just a matter of luck that the different scaled version did not happen to run into the uv trim curve bug.

With this next upcoming release that particular bug should be fixed. Hopefully there are not too many new ones introduced, we'll see!

- Michael

Michael <<Hopefully there are not too many new ones introduced>>

It's not in my nature to complain ...and i think you can't expect everything by one software..but i got an old silly question for you Michael:

a simple square 10x10,if i want to fillet r=5 ,it should be done: Not fillet...if i choose 4,99 it works
this happenned with all Moi's versions.I remember Autocad 2000-2005 did it without problems
i changed unit precision settings from 1 to 1,00.same thing..(maybe Moi don't want to share same points of the fillet..)
let you say something about

Well, if you fillet each point individually and not all 4 at once it works.



Regards
Stefan

You're right Stefan...i'd like just two at once,not all 4..no problem !

my compliments for your 3d doodles (linking stars II in blue is really nice)

Hi Mauro - the way curve filleting works is that each curve segment is trimmed by the arc pieces that are generated.

In your particular case there the original curve segment would be trimmed away to nothing at all, so the code that tries to do the trimming gets an error in that case and that's why it does not work.

It should probably not be too difficult to fix up, it just needs a special case added for this particular circumstance. I can change it to just throw out the original curve piece in this case rather than trying to trim it. I've added it to my list of stuff to fix.

In the meantime the easiest workaround is to just draw a circle in a case like this and just delete the square rather than using fillet - the result you are looking for is a full circle, right?

Thanks for reporting it!

- Michael

Thank you m-dynamics, much appreciated!

Regards
Stefan

Ok, the new beta is out there: http://moi3d.com/beta.htm?rev=k

Main new stuff is - updated Solids++ geometry library to the latest version, mostly various bug fixes but please also be on the lookout for any regressions (things that used to work previously but don't anymore after the update).

Updates to the Construct > Network command - fixed bug that tended to cause little lumps in the network and are also now some options that you can adjust to control the network behavior to choose to get a lighter but less accurate result if you want.

The Mac OSX version should now be working ok on OSX 10.8 Mountain Lion, please let me know if you see any problems launching this version on Mountain Lion, it should be ok now though.

I will start a new thread with a proper announcement in a bit.

- Michael

Cool :)