Need help troubleshooting LOFT_FAILED

elif

Hi Dave,
I finally got a chance to take a look at the case you sent in. It looks like we can actually create the loft, but then the result fails a geometry check, and so we fail the feature. It is not clear to me that the failure is necessary in this case, so I filed a ticket with our kernel to double check that.

In the meantime I'm attaching some images of what the resulting loft would look like. There's an area where two faces get very close to each other (near the cusps as expected). And that is where the fault occurs. Until I hear back from our third party, it might be possible to alleviate the issue there by adding a guide and manipulating it away from the opposite face. 
It would be difficult to do interactively since we don't yet have 3d curves, but it might be doable for you since you're creating your curves programmatically.





Hope this helps.
We'll keep you posted via your support ticket as we find out more information.

dave_cowden

Hi, thanks for the analysis.

I can create programmatically guide curves.. But what would be a wise choice of vertices to use for the guide curves?

I don't know how to go about addressing this incident, though I have the flexibility to do so. The problem is that I really don't 'know' which points are the ones near the cusp. These come from a big list of points.

Do you think this would be resolved by re-fitting the splines so that each profile has an equivalent number of points? Of will these two surfaces close to each other still be a problem.

Honest I'm having problems imagining how to get started on creating a guide curve. I will try it manually, but in code it's hard to whi which surfaces are the problem ones

elif

In this case (the one you sent in via support), the profiles were already broken up. You can create guides that pass through those points that split the profiles. (the guide location i pointed out was where there is currently a split).  IIRC the top profiles already have the same number of splits. You might be able to get a better behavior by splitting the last one (the one with no cusps) to have one more segment.

I can appreciate that this is pretty difficult to do with the current set of tools we provide. We will continue working on making it better and easier to use. 

dave_cowden

Oh, right! Ok good point, let me try creating guides at those splits.  I can also try splitting the profiles inro an equal number of segments so that the guides work out. I will report back with what I find 

billy2

1st thing is to build one side and then build the other using sheets. Generate 2 surface patches (sheets) and see if they are intersecting. You might not be able to see it because it's beyond pixel resolution.


I doubt if your sections have problems, it's in between the sections that could have the problems. My approach is to get it working first. Loft individual sheets and confirm that you'll get a clean manifold.



I was playing with lofts this weekend and was able to get through my design. 


The error messages from lofts isn't very useful, they never are in any system. The issue here is that I'm creating a 2 sided surface and surfaces want to be 4 sided. This means that 2 sides are collapsed resulting in poles (at least that's what we used to call a zero length side "a pole"). In the earlier days of SW, poles would make your model unstable. Maybe not with that feature creation, but in subsequent operations ie.. cutting with a 3 sided surface would eventually crash the system. It's good to construct well formed surface patches (4 sides). These days it won't crash, but it'll slow things down. Just because it'll build it doesn't mean it's constructed well.

So with OS, you can detach the curve from the end nodes. Why? I don't know, but playing around you can create those sides 3 & 4 by simply adding a gap between the curves.

Detached curve from end nodes:




Don't let this freak you out because all curves require a trim capability that shows/hides the curve.

Here's an example:


See the 2 orange curves created by the trim operation with the circle? In reality, they're the same curve, they just don't show the original curve inside the circle. This is an expensive dance and causes havoc for first time surface dudes learning how to construct surfaces. OS doesn't re-evaluate the 2 orange segments because they won't be accurate enough mathematically and therefore they play games with display on/off strategies with curves. If you believe this, then you'll kinda understand what a trimmed surface is, it's the same construct.

I'm not sure why I dove into the deep side of the pool here. Surfacing is pretty straight forward once you know the basics. Trying to surface programmatically is impressive. Keep it working, possibly lofting sheets and seeing the geometric behaviors. I know there's no knit yet, but breaking the problem down into manageable chunks might work better for you. There's a reason I built your geometry using the UI first. If you can't do it that way, I don't think you'll get the program to build it. Eventually you might be able to loft sheets that'll form a manifold (solid) after understanding the loft issues.

Good luck, please share what you learn 
 

dave_cowden

Wow, thanks for the insights! I am going to play around with these ideas and see what u can come up with