Forum: 3D Modeling


Subject: Modeling aircraft wing roots (and other nasty joints)

MarcoCraine opened this issue on Sep 21, 2013 · 20 posts


MarcoCraine posted Sat, 21 September 2013 at 8:19 AM

Greetings,

I'd consider myself a more or less routined hobbyist, but I've run into an interesting problem I'd like to share (and hear) some thoughts about:

I've been working on modeling fictional WWII-style aircraft, and I can't seem to wrap my mind around a particularly interesting problem: Modeling a realistic, smooth wing root. I'm working with Blender, although the question concerning the "right" technique is probably universal.

When it comes to wing roots on aircraft, you'll generally find two approaches in people's models:

a) The wing and the fuselage blend together fluently, in a realistic manner. The author obviously bothered to go through great lengths to achieve that.

b) The wing simply sticks out the fuselage, with little to no effort to smooth the wing root, leave alone to create a closed mesh. Seems you'll often get away with that if the viewer/user doesn't look too closely, but technically... I've noticed this is a weak spot an many aircraft meshes that would otherwise look great.

The problem about coming up with a general-purpose solution seems to be that both the fuselage and the wings of an aircraft are rather specific shapes, and no matter what technique you use to create each part, they never lend themselves particularly well to merging them to a single, tidy mesh.

Any thoughts or ideas on a problem like this?

Personally, I seem to achieve fair results by starting with two low-res cylinders, roughly laying out the general wing profile and hull shape, and then merging the two while they're still low-poly. It comes with much manual fiddling, but I feel it's a fair compromise. Still, this method seems to have a price in terms of poly-efficiency: It still requires wild cutting through polys and merging of vertices to create a realistic joint. Doing it before applying subdivisions requires the addition of "buffer" polygons around the joint to brute the curvature in shape, and that leads to lots and lots of useless little polys once the model is subdivided. And even then, it seems to be more trial-&-error than actual tidy work. But once the joint is there, detail work on wings and fuselage parts is rather hassle-free.

I suppose I'll go with this technique for my current project as it seems to look rather pretty - but I feel there must be some smarter techniques for joining two shapes that just don't seem to fit too well.

While I hope my rather primitive approach might help provide insights to other artists with similiar problems, I'm also curious about any ideas on how to handle such nasty headscratchers.

Best Regards,

Abendwind


LuxXeon posted Sat, 21 September 2013 at 3:45 PM

I've never attempted to model an aircraft, so I really have no idea what the wing roots are, or how they should look on an aircraft.  I'm trying to envision where the problem of extruding wings from the main fuselage would present an issue, however, and I can' t seem to come up with anything just from a standard polygon modeling standpoint.

I took a very brief look at some planes on Google for reference, and quickly extruded a wing from a rudimentary fuselage (basically just a subdivided rectangle), just to try and understand what you mean.  I think if I were going to model a plane, I'd start off somewhere in this neighborhood, although it may not be how an expert on planes would approach it.

Could you point out the problem here to help me better understand what you mean?  I'll try to help with the basic modeling dilemma, but when it comes to actual knowledge of planes, I have none.

______________________________________

My Store
My Free Models
My Video Tutorials
My CG Animations
Instagram: @luxxeon3d
Facebook: https://www.facebook.com/luxxeon


MarcoCraine posted Sun, 22 September 2013 at 5:30 AM

Thank you for the quick reply, and forgive my superficialty. Here's he current state of affairs. It looks acceptable, but my approach seems to have its problems (extrusion has its own issues, but more on that later):

Wing Root Problem - Overview

Note that a wing profile is supposed to be quite a delicate shape with a rounded, higher-poly leading edge and a sharp trailing edge that can be much easier on the poly density. At the same time, the fuselage is a rather tidy tubular structure. Naturally, the two go together like chocolate and sauerkraut.

So what did I do to achieve the above result, and what is it that I don't like about it? Well, the following is a glimpse at the mesh resolution I've been working in (it started out with two low-res cylinders, one of which ended up as the fuselage, the other one as the wing stub). Note how I mindlessly cut through quads to make it all fit:

Wing Root Problem - Ugly Spots

This solution just seems to feel untidy and may or may not get me into trouble later. It gets worse once we take a look at the subdivided mesh:

Wing Root Problem - Ugly Areas

As I mentioned earlier, I think I'll go with this solution for my current purpose, but I feel there must be better ways to achieve the overall shape.

I perfectly see how plain old extrusion is attractive: It gets you tidy edge loops around the wing roots (no messy triangles), but given how nasty a shape a wing actually is, I've never achieved satisfactory results with that. The vertical stabilizer you can see at the tail is extruded, as this one is less demanding in terms of actual airfoil geometry. But a wing will probably get you in trouble if you aim for a certain degree of realism. The manual fiddling to get the profile right (if at all) is probably much worse than with my approach of just bruting two separate cylinder-based meshes together. However, if anybody comes up with a nifty technique for extruding a satisfactory wing from a fuselage, I'll be more than happy to learn!

Now I actually regret I didn't document the process I used to achieve the above results. I have yet to do the horizontal stabilizers (the little wing-ish thingies at the tail that are still missing), which will be pretty much the same procedure. If anybody is interested, I can document the progress with a couple of screenshots to...:

a) show how it's (messily) done in case you're still learning.

b) Point out the problematic issues during the procedure in case you can provide some advice.

Pardon the lengthy post, I hope I could clarify the problem.


LuxXeon posted Sun, 22 September 2013 at 11:37 AM

Well, the problem doesn't seem as bad as you seem to think.  You have a nasty 5-point pole there, but it's in a spot that isn't going to cause that much trouble if you decided to leave it.  If you think about it, the wing isn't going to deform in animation, and it's not a spot where texture pinching/stretching is going to be a major issue anyway.  The subdivided area doesn't appear to be grossly pinched.

You dont actually need to end up with any tris there.  You could just run those loops a little differently (perhaps add an extra loop), and that would take care of it.  You may even be able to take care of that pole in the process.  The way that looks after subd isn't as messy as you think.  Did you try relaxing the polys in that area after subd?

If you just change the way loops are running on there a little, I think you can fix that up in no time.  It doesn't look unsolvable, if you simply add/remove some edges.  The shape of the wing, to me, isn't much different than a katana blade, which I have modeled before.

I would extrude an edge like this, all quads:

Which would subdivide nicely like this:

Now, there's no tris and no poles here, and I think the shape is nearly exactly what you were showing above (a little vertex tweaking to angle the back pointed part a little isn't a big deal, and could easily be done just by pulling verts around a bit).

I'd reconsider how you are modeling the fuselage to start.  All it depends, in my opinion, on starting out with the correct amount of edges to cut a hole in the side, and extrude the shape you want from quad topology.

______________________________________

My Store
My Free Models
My Video Tutorials
My CG Animations
Instagram: @luxxeon3d
Facebook: https://www.facebook.com/luxxeon


LuxXeon posted Sun, 22 September 2013 at 12:20 PM

After two seconds of just tweaking the verts in the back end of the wing, and adding one extra loop to the root...

I think this is looking similar to your shape now, but I am still all quads with no poles.  Actually, I see I need to straighten the bottom of the wing, and curve the top a bit more to be correct...

______________________________________

My Store
My Free Models
My Video Tutorials
My CG Animations
Instagram: @luxxeon3d
Facebook: https://www.facebook.com/luxxeon


LuxXeon posted Sun, 22 September 2013 at 12:32 PM

Edit:  To correct the shape, and have it being planar (flat) on the bottom and curved onthe top, I simply added two more edge loops to control the shape a bit more, and here's the result.  Still all quads, no poles.

This is just a very quick job, but I'm sure it's in the right direction, because the shape of the wing could be moved into any variety of curvature at this point.

Is there anything inherantly wrong with this method?  Seems to take care of the issues you were having.

______________________________________

My Store
My Free Models
My Video Tutorials
My CG Animations
Instagram: @luxxeon3d
Facebook: https://www.facebook.com/luxxeon


SinnerSaint posted Sun, 22 September 2013 at 2:36 PM

Quote - Edit:  To correct the shape, and have it being planar (flat) on the bottom and curved onthe top, I simply added two more edge loops to control the shape a bit more, and here's the result.  Still all quads, no poles.

This is just a very quick job, but I'm sure it's in the right direction, because the shape of the wing could be moved into any variety of curvature at this point.

Is there anything inherantly wrong with this method?  Seems to take care of the issues you were having.

NO poles, Lux??  I see FOUR areas there which contain a 5-way junction.  Take a closer look.  This is unavoidable, as 5-way junctions are derived from extrusions.

But that's not a bad thing!  An E-Pole (5-edge junction) is an important part of surface modelling and getting proper edge loop topology, so why is that a big deal here?

There's nothing wrong in having tris on this kind of model either.  It's a plane.  The wings don't move in animation, so what's the big deal about keeping quad topology?  People think you need all quads all the time, but most people don't even know WHY.  Why do you need or want all quads?  Usually the answer is deformation or UV mapping.  However, this kind of model will not have a problem with either of those two issues, so the concern over it being 100% quads, and no tris is moot.

I don't see much of a problem with the OP's original topology at all.  Does it texture well?  Test it out and see.  Just relax the UV's in that area like luxxeon suggested, and you shouldn't find any horrible issues with your texturing there.  IF you do, it shouldn't be hard to cover up, because how much detail is really going in that corner of the wing to begin with?

Lux's solution will keep it all quads, but I don't kwow what he is talking about with the poles.  There's two kinds of poles in proper edge loop modelling.  E-Poles (5-edge junctions) and N-Poles (3-edge junctions).  Both of these are important for redirecting edge loop flow.


airflamesred posted Sun, 22 September 2013 at 3:32 PM

Yes the e poles are inevitable from the extrusion and, I would guess, be followed by some n-poles at the end of the wings.

I would agree with Lux on the quads though. In this case there is no need for tris.


LuxXeon posted Sun, 22 September 2013 at 4:44 PM

Quote -
NO poles, Lux??  I see FOUR areas there which contain a 5-way junction.  Take a closer look.  This is unavoidable, as 5-way junctions are derived from extrusions.

Well, yes, there are those.  However, I was specifically addressing the "problem" area that Abendwind was concerned with; what he calls the "wing root".  In his example, there's several 5 edge junctions in a cluster at one end, and a triangle at the other end.  When he subdivides this, he doesn't seem to care for the way the edges are flowing in those areas specifically.  This approach I am providing is an easy way to handle that.

I see no reason to ever keep a triangle on any mesh if you don't have to.  In this case, it may be a matter of taste, but it's also good practice too.  Theoretically, if you can't figure out how to quadrify a hard-surface object efficiently, how would you ever learn correct edge loop procedure when it comes to organic modeling, where it's more than a matter of aesthetics?

______________________________________

My Store
My Free Models
My Video Tutorials
My CG Animations
Instagram: @luxxeon3d
Facebook: https://www.facebook.com/luxxeon


MarcoCraine posted Mon, 23 September 2013 at 6:19 AM

I suppose I'm with Lux there in that my primary concern is an aesthetic one: I feel I just want to get it 'right', though I perfectly see there's different opinions on whether that's actually necessary in this project.

Lux, your approach seems to illustrate the problem about my current mesh: While I did plan for the trailing edge of the wing to join the fuselage along an existing division, I failed to properly account for the position of the leading edge when I created the fuselage mesh. Having another division along the fuselage there would have eased up the whole affair.

I tried to see what I can do about that leading edge mess earlier today, and it seems the first level of subdivision lends itself better to manual corrections than I thought. It'll still be some fiddling from the way it started out, but I think I can close another smooth edge loop and repair some of the worse distortions on that level.

Minor distortion and poles (so yes, I learned a new term ;) ) around the wing root area won't hurt my current project too much. I'm just slightly concerned about panels and other fuselage details I may have to cut out as I progress, so limiting the untidy spots is pretty much mandatory for me.

Anyhow, your perspectives have already helped me think outside the box in a way. Thank you for all your feedback so far. I'll see what I can do about this.


airflamesred posted Mon, 23 September 2013 at 6:23 AM

I'd leave the panel detail to bump rather than mesh.


SinnerSaint posted Mon, 23 September 2013 at 2:10 PM

I understand there is merit in knowing how to keep a nice clean quad mesh.  It looks pretty when you show off your wires, and of course it's essential in character modelling.  The other side of the coin is that it also serves a purpose to know when and where it is ok to allow tris, if it's going to speed up workflow, and have little or no impact on the result.  I'm not refering to anyone here, but there's a lot of people out there who have no idea why they model in quads, and think that they can never have even one triangle in their meshes. This is especiallyl true for anyone who might want to get a job with any kind of  studio, where you could be expected to finish a model in a ridiculously short period of time.  Or if you ever intend to model for RT/game engines, you'll learn quickly where to collapse verts, and that tris are your friend.


EricofSD posted Tue, 24 September 2013 at 11:27 PM

Good advice above.  I built two aircraft so far.  The jet fuselage/wings was made from a single cube.

(In the chant voice of the worms in the locker on MIB... "All Hail the Cube".)  ... If you make a cube correctly and adjust the points, you get a sphere like the earth.  Hence the flat earth society.  But I digress.

There is some extrusion and point welding, but the fuselage is organic looking and is a cube.  I'm sure you can do this in blender.


EricofSD posted Tue, 24 September 2013 at 11:33 PM

For fun...

MarcoCraine posted Wed, 25 September 2013 at 3:53 AM

Saint/Lux:

Both sides of the debate sound like they have some merit for my purpose. Saint, I'm actually glad to see this debate continue in the thread you opened, as it's interesting to see people's opinions on a matter that I've frankly been rather ignorant of.

Eric:

Neat indeed. It think I can spot a rather harsh division at the biplane's wing root, but the smooth low-poly topology on the jet is interesting. It's hard to judge from the total side view, but I think the wing profile is somewhat unrealistic. People probably won't spot it in a render, but judging from the wireframe view, the wing profile looks a bit too symetrical to generate lift in level flight (but I'm getting picky and speculative here ;) ).


pauljs75 posted Sun, 29 September 2013 at 10:13 PM

I think it gets tricky when the wing attaches to the fuselage at the very top or bottom. There's different sets of edge flows that you have to make work together. If the wings are attached midway, it's a heck of a lot easier since you can let Catmull-Clarke take care of most things for you. (And if not wing/fuselage edge flows, then there may be wing/nacelle-bulge edge flows, or similar topology occuring with tail surfaces.)

Not sure if it'd be any help, but I have a few free aircraft at ShareCG that you could download and study. I think all of them have the wing attach at the bottom, and for the most part I think I've done what I can to take care of edge flows in a reasonable manner. Sometimes you also have to tweak those verts manually to get the tension or transition right.

Also if the mesh doesn't deform, and you don't plan on smoothing it anymore, it's ok to have a few tris here and there. And sometimes a tri may be needed to split what would be a non-planar quad in order to take out any ambiguity a render engine might have.

BTW, that screen grab looks like Blender there... I'm not new to 3D, but I'm starting out at Blender after they got to the 2.6x versions and sufficiently noob-proofed it. So I've been using it a lot now. Great rendering and animation (Cycles is awesome), but for working with polygon meshes - I still prefer an old version of Wings3D. You might want to give that a shot since it seems to deal with edge flow better than Blender (imo), and has faster workflow when it comes to models and how their own geometry relates to them. (Blender might be able to do some of the same things, but it's not as obvious or involves a lot more steps.)


Barbequed Pixels?

Your friendly neighborhood Wings3D nut.
Also feel free to browse my freebies at ShareCG.
There might be something worth downloading.


pauljs75 posted Sun, 29 September 2013 at 10:54 PM

I noticed some of this stuff with topology while looking at the one pic. So I drew a little over it. Not sure how much that helps.

Barbequed Pixels?

Your friendly neighborhood Wings3D nut.
Also feel free to browse my freebies at ShareCG.
There might be something worth downloading.


MarcoCraine posted Mon, 30 September 2013 at 4:31 AM

Ah, thanks a lot for that graphic effort. I feel I can get somewhat blind to more obvious solutions once I have stared at my own mesh long enough. This might actually help a bit.

I perfectly agree the low-wing configuration gets into the way of more convenient solutions. It makes the geometry around the wing root particularly tricky and calls for nasty transitions at the wing edges. Those would arguably be more straightforward to handle on a mid-wing aircraft.

To be perfectly honest, I've never tried Wings 3D. I've been playing around with Blender for quite a while now, and I love how it always gets you a nifty solution to this or that if only you keep banging your head against the keyboard hard enough ;-).


EricofSD posted Wed, 02 October 2013 at 11:11 PM

Abendwind, thanks for the observation.  The wing is not quite symmetrical.  The lower surface is flatter and the upper has more of an arc to it.  So there is an airfoil profile, however, probably not enough.  I can cage it and make it a bit more prominent.

It is a bit like an NACA 2415 airfoil which I thought would go with a high aspect ratio wing.  We really don't know what the L133 looked like in detail, only guesses.  I'm pretty sure the landing gear configuration is incorrect. 


EricofSD posted Sat, 12 October 2013 at 7:32 PM

Oh, I was totally missing it.  You were referring to the biplane, not the jet.

The biplane has a good airfoil.  The lower wing is simply a separate object from the fuselage.  Hence the sharp edge.

The jet has the wing root connected to the fuselage so that is where the smothing is.