Issue: Loft Rail Inverts surface flow

[cekuhnen]

at the mid point the surface bends incorrectly with rail option

[Anonymous]

It's not incorrect, it just doesn't have enough information to do what you want! 😉 (it's not a sweep).  Looks like choosing centerline option gives desired result, I'm guessing due to centerline meaning that it must be at the "crest" of the surface curve.  For rails, would need at least another probably to give it a better idea what want.

Jesse

[cekuhnen]

I disagree

Loft in Fusion has a rail/rails option

point of a rail is to move it along a rail otherwise center line would do the same

shouldn't it?


sweep on the other hand in Fusion can only move one profile a long a rial

[Anonymous]

The loft in Fusion (I'm not familiar with other modeling programs) uses some sohpisticated wizardly to determine a good curve to connect between two profiles.  The cross section of the curve as it moves from the first to second profile is not forced to keep the same geometry as does a sweep.  As the loft is moving from one profile to the other, it's trying to figure out how the paths of the points along the first profile are going to ultimately match up with points along the second profile (hence that line that appears with lofts that can be adjusted to redirect profile to profile mapping without using a rail...or multiple lines that appear if profiles have multiple sharp corners).  Rails dictate explicitely how a specific point on one profile is to map to a point on the second profile (as well as the path followed in space from one end point to other), and then the loft region adjacent to the rail is extrapolated from that information in some way.  Due to this, it's sometimes important to choose well what is going to be profiles, and what is going to be rails.  So with the method you are trying here, I tried adding another rail, but can see the difficulties 😉

It's not extrapolating well, so best to choose another method. 

My guess is the centerline rail option for this method works well, because it gives the extrapolation algorithm good information, specifically that if you had an imaginary duplicate surface mirrored about the centerline, that the two would have either G1 or G2 (not sure which) at the centerline.

Well, I got to get some work done, interesting seeing what you brought up and talking to you!  G'night. 🙂

Jesse

[cekuhnen]

@Anonymous

yeah who knows - I dont know their code.

In Fusion they kinda mix what normaly is a loft with rail functions.

And then they offer a sweep one rail command.

But a one rail 2 profiles command should work like moving the profiles along the rail

here is it in Rhino

[Anonymous]

Coulda shoulda 😉 Fusion is Fusion.  Probably confusion may arise for you because they use the term "rail" in the loft tool, but it's not really a rail in the strict sense at all as applied to a sweep.  Strictly, it's only a single point/line rail (or with the centerline rail, having a G1 or G2 with imaginary mirrored loft surface), and what ever happens laterally to that is fair game. 

[cekuhnen]

@Anonymous

Yeah but still think about if loft along a rail then it is a sweep.

I m curious what some AD devs would say.

cause here it works just fine

of course with a single rail centerline might be better for an even move

[Simon.Hern]

Hi Claas.

I've been doing some development on the Loft algorithms used by Fusion recently.  I think differences in terminology between Fusion and other applications may be causing confusion here.

In Fusion, Sweep makes a surface by sweeping a Profile along a Path (no surprise there!).  The surface's cross-section keeps the same shape as the original Profile.

The approach used by Loft is fundamentally different from Sweep.  Loft makes a surface by covering the framework of 'Sections' (both Profiles and Rails) with what are known as Coons Patches.  Mathematically a Coons Patch doesn't see any difference between a Profile and a Rail - they're both just Sections - although of course Fusion does handle them a bit differently (for example, a Loft must have at least two Profiles, but it can have one Rail or no Rails, and the Profiles may be non-G1 but the Rails must be G1).  This means that you can't expect cross-sections of a Loft to stick to the same shape as its Profiles (or indeed its Rails).  (If it helps I'm told that Loft in Fusion is quite similar to the Rhino NetworkSrf command, although I'm not a Rhino user myself.)

And then there's Loft with Centerline, which is different again.  That can be thought of as a sort of hybrid between Loft and Sweep.  The Profiles are swept along the Centerline, but with the surface's cross-section morphing between the shape of one Profile and the next.

For your example above, the ideal operation would be, as you say, a Two-Rail Sweep - but that's not currently something Fusion supports.  Lacking that, normal Sweep and Centerline Loft both appear to give useful results in this case.  However Loft (with Rails) simply isn't guaranteed to give a surface that respects the shape of the Profile in the way you want.

Regards,

Simon

[cekuhnen]

@Simon.Hern

Hi Simon,

So if one wants to see it that way then Loft is maybe similar to CurveNetwork in Rhino.

There you select your profiles (boundary edges and if present profiles inside the future patch) and then create a surface

with those profiles and define how aling the boundary the surface with flow into the neighbour surfaces.

This is the original form - I split the center surface to create the red window

CurveNet with G2 settings note center area blends well

all edges are positional and center area is linear not inverted like it is in the Fusion file

And here with internal network profiles - the main reason to use curve network

Ok not but what if we do not make a surf G1 or G2 to other surfaces:

I hope this explains a little where the similarities are and where things are different.

Rhino Sweep 2 Rail can only offer SC aling the rails but not the start and end profile even if that is part of actual sufaces.

I keep the sweep open so you see it better.

So this brings up some interesting questions or suggestions: are the current namings good and or could the functionality be improved.

Fusion sweep can only do one rail one profile and has no SC ability making it more or less an extrude along path command.

Fusion loft that can work like a loft or sweep 2 rails where SC is only possible along the profiles not rails making it not work like surfnetwork.

Rhino can do a sweep along one rail - it can blend between two profiles alone one rail - it can blend two profiles along two rails

Rhino CurveNetwork can SC to all boundary edges

I am not sure if Sweep should be or could be renamed into Extrude along Path

because I find it always confusing that what in most surfacing apps is sweep in Fusion honestly is more loft.

To my understanding this month the loft command finally will get SC along rails a much needed improvement.

I am actually quite ok with making the loft command in Fusion a loft/curvenetwork/sweepRails etc combo tool because while Rhino has many specialized commands it just inflastes the UI a lot and packing that amount into one icon is genious in Fusion.

But then the sweep command itself falls out naming wise.

[Simon.Hern]

Thanks Claas!  That's interesting stuff.  It does seem that Rhino's "Surface from Curve Network" is very close in principle to Fusion's Loft (using Rails rather than Centerline) even if they don't give quite the same results.

Taking a closer look at the loft in your original example, the 'inverted' surface shape does seem to come about because the algorithm views that as a good way of connecting the profiles and rails.  In particular, it seems to be trying to connect the two profiles as directly as possible, and that's pulling the surface inwards.  Adding a middle profile (as in the picture below) gives the algorithm a much better idea of the overall shape you're trying to achieve in this case.

Simon

[cekuhnen]

Thats pretty true but this requires the code requires you as a user to add an extra profile by hand.

It would be nice or better if that would not be needed.

However I am curious about how this would also be if the rail can have SC abilities so it works like a patch surface in Alias or with internal profiles like a surface network in Rhino.

[Anonymous]

It's not a matter of being "better," it's just what that tool is designed to do, and in many cases its algorithm of finding the optimal surface like it does is just what is needed 😉

Jesse

[cekuhnen]

I actually disagree - if a code requires more steps then it is not ideal.

however center line does the job when you only have one rial - with two rails it is not really an issue.

with a center line as demonstrated this works but obviously because of the lack of rail SC the edges break - looking forward to this month update.

[Anonymous]

I didn't mean optimal as in "ideal workflow", but rather optimal as in optimal surface, specifically minimal surface, see https://en.wikipedia.org/wiki/Minimal_surface In many cases, a minimal surface is what is desired for lofts, just not in this particular case you illustrate. Jesse

[Anonymous]

I would add that's just what I'm guessing, not in any way being a loft expert 😉

[cekuhnen]

in general you want clean surfaces. that is why the patch command often works good but along the edges has some issue (noise in the zebra)

Alias Patch command makes size

also there along the center line one needs to adjust the weight so that you get curvature 

[Anonymous]

Yes but there are an infinite number of possible surfaces that can exist between two profiles.  How do you/algorithm choose which one?

[cekuhnen]

@Anonymous

my biggest concern is surface quality

the combs show G2 and G1 and you can see that when it comes to the result it is not optimal - the center spline actually making things even worse

because the thing is the loft command messes up the suface - then you say well lets add a center profile to stabalize it 

now the result is less noticable but still there and instead of one now at two areas each between a set of two profiles

everything other than this is well not good:

Alias Square surface result (needed center weight help)

Alias Rail command 2 profiles 2 rails

I am looking forward to the July update and if the loft then can do 2 profile 2 rail surfaces + surface continuity aling edges and get equally good results.

if not that would be a bummer because the Fusion results in this post I dont really like