What's new

Issues with the loft tool, a nice clean example of where AD seems to always fail me

cdub

Member
Now that AD is actually being maintained again, I am going to take a few minutes and document some problems I have with AD whenever I attempt to do geometry more complicated than rectangles and holes.

I have been working on a 1/43 slot car design, with the intent of 3D printing the majority of the car. Modelling the chassis is very straight forward, as it is just simple extrudes and cuts.

However, I'd like to model a body that has a few curves to it.

The game plan is to do 3 lofts, one for the front, one for the rear, and one for the cabin. Then shell them out. Front and rear "fascias" will be done as separate pieces.

Three sketches on 3 planes, all of which are done by drawing one half, then using the sketch "mirror" tool to make the symmetrical other half.

Here are the sketches:

Here are the results:

So, the result isn't symmetrical. One side has an odd bulge, and the other an odd twist. No combination of check boxes in the loft tool fixes this. I know better than to sketch half of the profiles and do the loft on just half, because the central plane will be warped so badly that mirroring the feature wont yield usable results.

Okay though, one half of this looks usable, so cut off the odd half and soldier onward.

Again, 3 sketches and a loft. This loft performs worse than the first loft. I already picked a side when I cut the first loft, I am stuck with a weird twist in the rear of the car that I don't want, but fine. Cut off half and move on.
Sketches:

results:

results:

By this point, I'm tired and am giving up. Instead of a 3rd loft, I just do an extrude with some draft to form the greenhouse. I doesn't look very good, but I've already spent twice the amount of time on this as I wanted to.

Lop off half of the greenhouse and move on.

Now it's time for the shell.

Right out of the gate, there is a problem. I'd like to select the bottom of the body as a face to remove, but for some reason when it solved the Loft, it didn't split this face into two pieces.

So, I'm stuck just picking the "interior" face and adding additional cuts later to knock out the bottom.

This is where AD always gets weird. For the first attempt, I go for the gusto and attempt a 2mm thick inward shell. AD grinds for a long time, then pops an error message that the operation has failed. No surprise. Then I turn the thickness down to 0.2 mm. AD Grinds again for a long time, and to my amazement succeeds! I see it rendered on the screen! Of course, an 0.2mm thick shell couldn't be more useless to me. I try to bump the thickness back up to 1 mm, and it fails. I turn it back down to 0.2mm and it fails. Wait a minute, why did it work the first time? I am never able to repeat that first success. Now, this isn't my first rodeo. I know that shelling outward is usually easier to solve, so I check the box and shell outward.

A 2mm shell fails. A 1.68mm shell fails, but a 1mm shell solves!
Not what I wanted, but I think I can do something with it.

Now for the easy stuff. I knock a couple of holes through the shell for the wheels. But what happened here???



Finally, I cut out the "bottom".

Okay, we are halfway there. Ideally, now I could just "mirror" all of these features and boom, I have a body shell. Of course, it can't solve that. I won't even bother with a screenshot. But, hope is not lost. I can use the Boolean Tools and insert two of these, mirror one, and "stick" them together!


Fire up a new part, do a few Boolean operations, and hey, I have an okay looking model, with one small issue.

Now, for the moment of truth, export an STL file and see if we can print something?

Here is how the STL looks when I import it into Simplify 3D:

More than half of the part is missing!

USELESS!

This little experience has played out probably a dozen times for me in the last few years. Basically, I just don't do projects that require anything other than rectangles and holes because I know Alibre simply can't handle it. It is so frustrating to be so limited! I have a hard time recommending people drop $1K or more on this software when I know it just chokes on anything slightly more complicated than plates with holes.

I don't expect to be able to model every swoop and curve of the latest Ferrari, but I do expect to be able to throw 3 lofts together and pop in a shell in a reasonable manner. I have experience with Pro/E and SolidWorks which leads me to believe that it absolutely can be done. I know, I know that AD costs less, and I know I know that it is a "great value" and that it does lots of things well. I'm also aware that there are lots of modelling techniques that "could be" used. At the end of the day though, I paid the price of the ticket for parametric 3D CAD software (in the economy class) and I expect the whole plane to fly all the way to its destination.

This turned into a bit of a rant and I apologize. The Loft feature has performed very poorly for me the entire 10 years or so I've owned this software. Is it going to get any better in the next 10?
 

DavidJ

Administrator
Staff member
Lofts can be tricky. Others here have had much more practice (and success) than me, but I know there are some tricks that can help. If the car body is symmetric about the centre line (or at least mainly so). Just produce one side only with lofts, mirror when you've finished modelling one side. That's much easier than trying to get both sides of a loft to behave exactly the same.

Another thing - you can often avoid needing to use lofts at all, yet still achieve some pretty good 'organic' shapes - look at http://www.cadalyst.com/cad/top-down-design-curvy-parts-alibre-design-tips-11195

To get lofts to behave - I often find it necessary to use guide curves. I tend to avoid the use of lofts where possible, because I agree it can be hard work and frustrating.
 

cdub

Member
Modelling half of the loft sounds great, and almost works, except on the "neutral plane" of the part, the Loft invariably introduces some "twist" and you lose the ability to really work with that face. I went back and reworked the sketches and lofts this way, and this screenshot shows the effect the best:
 

DavidJ

Administrator
Staff member
Can you mirror what you have there (leave the shell until after the mirror) ?
 

bigseb

Alibre Super User
If I may offer a tip:

Start with a block and whittle away at it. Subtractive rather than additive is the way to go with most solid modellers. Ultimately work of this nature is best done using surface modelling software like Rhino or Moi (or Catia if your budget can stretch that far).
 

JST

Alibre Super User
This problem is INHERENT IN ALIBRE as far as I can see. I posted and reported it long ago to 3DSYS, so it must be somewhere in a dusty recess of the bug list.

What I found was that the loft tool seems to act as if the surface is a rubber balloon. You put guide curves in and that is all very well, but at some point in the process, the TOOL seems to blow up the balloon inside of the "cage" you establish with the curves.

I COMMONLY FIND that some parts of the form have BULGED OUT in between guide curves. That is usually rather subtle, but sometimes is really blatant, as in your case.

I try to avoid the LOFT tool for that reason.
 

bigseb

Alibre Super User
This problem is INHERENT IN ALIBRE as far as I can see. I posted and reported it long ago to 3DSYS, so it must be somewhere in a dusty recess of the bug list.

What I found was that the loft tool seems to act as if the surface is a rubber balloon. You put guide curves in and that is all very well, but at some point in the process, the TOOL seems to blow up the balloon inside of the "cage" you establish with the curves.

I COMMONLY FIND that some parts of the form have BULGED OUT in between guide curves. That is usually rather subtle, but sometimes is really blatant, as in your case.

I try to avoid the LOFT tool for that reason.
I recall your thread HERE where you had a misbehaving loft. And the solution I suggested then was to do it subtractive. Not saying Alibre shouldn't look at fixing this issue, only that there is a simpler way of doing it. :)
 

cdub

Member
Alright guys, it's time to take this idea that "lofts are hard to do", or "you can just whittle it away with extrudes and drafts" out back and shoot it in the head.

I just logged into my old OnShape account, which I haven't touched in over 6 months and modeled this thing up exactly how I wanted to in about 30 minutes.


Not only did the lofts behave the way you would think they should, the Shell function works and gives me a 2.5 mm thick inward shell. Also, "feature mirror" works. I don't have to jack around with creating another part file and importing two halves, and mirroring, etc..

Finally, I export an STL file and it is usable in Simplify 3D.


I'm so sick of hearing about how you need Rhino or freaking Catia to do this kind of stuff. Maybe 20 years ago, yes, but good grief, not today.

No more excuses on this stuff. I demand better.
 
Last edited:

JST

Alibre Super User
I recall your thread HERE where you had a misbehaving loft. And the solution I suggested then was to do it subtractive. Not saying Alibre shouldn't look at fixing this issue, only that there is a simpler way of doing it. :)

You did, and I did, and it was perfectly OK done that way.

Point being that the LOFT approach SHOULD have given the same results, only it did not.

All it would have taken to make the loft the preferred method in the case of my particular instance, is a slight change of shape that would not allow the method you suggested and I used. Then the use of the fillet command with start and end radii would not have worked well.

Now, I assume that Onshape and Alibre use different "engines", and it appears that the ancient version of ACIS that Alibre uses has problems. There are always problems with fillets, also, and fillets are really just special cases of the LOFT command.

Just as cdub says.... "no more excuses", I demand that lofts and fillets get better.

If one program can do it, there is no excuse for another program that has the same feature failing to get it done. But lofts are buggy, and fillets commonly fail.
 

cdub

Member
Everyone is so hung up on the Lofts, and working around the fact that the Lofts tool doesn't work. That is only 1/3 of the problem. The shell command doesn't work worth a darn and neither does the feature mirror tool. Oh yeah, and fillets are flaky too, so that is really 4 problems. Fillets are always dicey in CAD tools and of the items involved here, that is the one I'm least worried about.
 

cdub

Member
I went ahead and put a support ticket in on these issues. I pointed them to the forum discussion and attached the Alibre files. If anyone else has a similar story to share, maybe put it on here. It is free to sign up for OnShape, but the free version is limited to only "public" files. More examples of people failing in Alibre and succeeding in a web browser based free online tool might help drive the point home.....
 

JST

Alibre Super User
Everyone is so hung up on the Lofts, and working around the fact that the Lofts tool doesn't work. That is only 1/3 of the problem. The shell command doesn't work worth a darn and neither does the feature mirror tool. Oh yeah, and fillets are flaky too, so that is really 4 problems. Fillets are always dicey in CAD tools and of the items involved here, that is the one I'm least worried about.

Hunh....?

Shell works pretty well, although I have had problems shelling a solid that was LOFTED. That can give weird partly transparent portions and so forth. Seems to be something to do with the whole loft deal, the lofted surfaces are in some way not as "real" as other surfaces.

Loft also does not seem to be "deterministic". If you loft the same thing several times, it may not end up with the same result every time. The same seems to be the case with fillet, which I think is a special case of loft. Fillet may fail 5 times in a row, and work the 6th, no clue why.

Feature mirror has been FLAWLESS for me.

I think, however, that features created in certain ways do not mirror or shell in a useful way. That seems to be more a function of the way the feature was made, and not a problem with shell or mirror. Meaning that some stuff does not mirror, etc.
 

oldfox

Alibre Super User
Feature mirror has been FLAWLESS for me.

Same for me.

Fillet may fail 5 times in a row, and work the 6th, no clue why.

I've had problems with fillets in the past. Now I can pretty much make them work every time if the fillet radius is NOT greater than the 2
edges you wish to fillet. (Always a fail) I have noticed a distinct habit of fillets needing to be set in a particular order and then the ones
that failed will fall right into place. I have done this on some very intricate edges. (read: going in many directions but all connected)

Just my $0.02
 

cdub

Member
I should have stated that more clearly. The Loft working or not is just the first step. The other tools that you then apply to the loft to finish the geometry not working is just as big of a problem. My experience is similar to yours. The shell command is flaky with extrude/revolve based geometry and I typically have stability problems when editing models with shell commands, but I'm usually able to get a usable result. The feature mirror command works okay with extrude/revolve geometry as long as you don't try to mirror the fillets as well. However, neither feature works very well (or hardly at all) when applied to a loft.
 

cdub

Member
Here is an example of shell/mirror issues, no loft involved.

It's a simple part. In AD, Feature mirror won't let you select the shell. You can select it, and it highlights, but doesn't get mirrored.


In OS, it behaves as you would expect.


AD will do a complete geometry transform and create a mirror of the part, but it deletes the existing geometry in the process. Seems like other tools I've used had a checkbox for this option. Interestingly, the first time I did the Geo Transform, I was left with a surface from the part on the original side. I deleted the mirror. Saved, closed, reopened, and recreated the mirror and it didn't do it again.

Honestly, when I created this part, I expected the fillets to fail in the feature mirror too and they didn't. Perhaps some improvements have been made recently?
 

Attachments

  • SimpleEx.AD_PRT
    337.5 KB · Views: 6

DavidJ

Administrator
Staff member
Put the shell after the mirror - then it works perfectly.

Remember that it is generally the operation being mirrored - having 2 shells in one part can lead to all kinds of problems, so I can see why you aren't allowed to mirror it.

Onshape seems to take a less literal view of what it means to mirror the shell - instead mirroring the result perhaps.

I'm not suggesting one is right, the other wrong, just that there are different ways the commands can be interpreted.
 
Last edited:
  • Like
Reactions: JST

cdub

Member
We both know that. The point of this file is to be a simplified example. There are plenty of geometries where you would want to do the shell first. Why can't I mirror the shell in AD? I can mirror it in every other CAD program on the market. If it doesn't work in this most basic case, just think of all of the others where it falls apart.
 

oldfox

Alibre Super User
OK. Here is the *simpler* version of your example and it didn't fall apart.
 

Attachments

  • I did it my way....AD_PRT
    487 KB · Views: 12
Top