View Full Version : Smooth blend of a cylinder and a wedge

03-15-2015, 05:44 AM
I'm trying to work out a way to smoothly blend a sort of wedge shape with a cylinder. (Image attached)

I've not gone too far to start again is that makes things better.

It may look scifi, but this is a serious proposal for a reusable sounding rocket. And the tubular forward hull needs a smooth join to the wedge shaped main body.

I tried some combinations of boolean and rounder, but as usual, I got a crashy mess.
Has anyone got an elegant way to smooth the area where the two shapes meet?

All clues very gratefully received.


03-15-2015, 07:29 AM
It could be done with sub-D, if you used the Edit Edges tool to add edges along the triangle's sides that line up with the tube. You could boolean the two shapes or add an edge around the tube and manually slide each point to be close to the edges on the sides of the triangle. Use tools such as slice to manually cut... actually, it's hard to explain. I might try to do a quick example.

03-15-2015, 08:51 AM
Sorry, I did it but there are 25 pages of screen captures and it's too much work to turn that into a step by step right at the moment. For the time being I'll give you the model as it might help to look at it.

Basically I used add edges and also the slice tool, and a bit of fiddly stuff re-directing edge loops and tris at the connection points.


And Multi-Shift as well.

03-15-2015, 10:04 AM
This looks like a task for a non-polygonal modeling tool.

03-15-2015, 11:59 AM
Is it totally Sub-D? or polygonal? I think you could take the panels and break them up, it would only add about 24 poly's to the model but if you have to have nothing but four point poly's or it has to be sub-D'd then it might not be the best solution. it might add way too many poly's to the model.

03-15-2015, 06:23 PM
Not elegant, but straight forward work.


But, yeah, probably easier to model with splines and patch.

03-15-2015, 07:19 PM
A few more views would help. I think I'll take a swing at this...

What would REALLY be interesting is a RACE (at least, elapsed times) for various solutions in different apps.

03-15-2015, 10:00 PM
I thought I had kind of solved it.


Of course that 3rd powers plugin would make short work of it.

03-15-2015, 10:40 PM
I thought I had kind of solved it.


Of course that 3rd powers plugin would make short work of it.

Pretty much the same conclusion I came to. I saw this at work and couldn't open the zip cause my LW seat is at home.

03-16-2015, 01:16 AM
Wow! Thanks for all the responses!

I have to admit I'm not clear on how all this was done. I'll post files with the base shapes later, so people can play with the source material.


03-16-2015, 01:22 AM

OK, here are the base shapes, thanks all!

03-16-2015, 06:57 AM
In case anyone is interested, this craft is an update on a project I did a while back, for a real re-usable sounding rocket. The new version is a lot bigger, (now 15m long), and as you can see has added a nose section. The idea is that you can use different numbers of engines, or even strap 3 together if you need a lot of thrust.


03-16-2015, 09:02 AM
My guess is that DJ's solution (I like the second one better) is a manipulation of the points of a single tube object knifed to get the diagonal (with a multi-shift or two in the top flat and a rounding on the diagonal). Not a combination/union of separate shapes or elements.
It's all quads so it could be subD'ed but it's dense enough that it's also fine as polys.

03-16-2015, 09:47 AM
If you look at the model file, the four layers sort of show the stages, but the last steps were pretty fiddly, I didn't use boolean, I sort of did a manual boolean, slicing and cutting away the sections that intersect, and manually knitting the points together. You could do it with boolean too I think, I can't remember why I chose not to.

03-16-2015, 10:57 AM
OK, thanks for the explanation.

I might even be able to just scale your one to match my bits!

03-16-2015, 11:34 AM
My apologies: I grabbed the zip but hadn't unpacked it. (Thumbnails apparently ROOL!) Nice work, dj.

One note for newbiesİ: AFAICT this object can be modeled in quarters, IOW 1/4th the total work, just mirrored twice. That would minimize tedious fiddling.

03-19-2015, 03:38 PM
After a lot of mucking about, I think I got it!

Thanks for the pointers. The wing edge is a bit sharp, bit IO think I can fix that easily.

03-19-2015, 04:49 PM
Looking good :thumbsup:

fwiw... totally forgot to throw my attempt up.

127553 127554 127555

03-20-2015, 01:16 AM
One thing I could use a hand with - I'm getting a bit of a crumple near where the wedge meets the cylinder and I do not know why.

Mesh shown before and after pressing tab.
Any ideas what I did wrong?

I don't normally model this way.

03-20-2015, 04:24 AM
Starbase if you have a retopo application my suggestion would be to block the shapes and the make the topology on top, so you dont have to worry about keeping the shape while you decide the topology. If not let the intrrsections dictate whats whats the mesh density required on either side.

But yeah... nurbs :)


03-20-2015, 07:31 AM
One thing I could use a hand with - I'm getting a bit of a crumple near where the wedge meets the cylinder and I do not know why.

Any ideas what I did wrong?

That crimp is suspiciously on the center line. You mirrored, yes?
I believe you have double points on one or the other side of the crimp itself.
(warning you might have double points along several spots. The crimp might move if you fix one spot. In that case use option 1 below.)
Option 1: Re-mirror on Y making sure to Zero out (v) all those center line points on the Y first.
Option 2: Grab the two offending points and merge. It might be hard to grab both in shaded solid mode. If you only get one, try selecting in wireframe.

03-20-2015, 07:39 AM
Good thinking, I'll check into that.

03-20-2015, 10:22 AM
Here's a nice feature request (could be CPU intensive): interactive Point Merge-- would allow the user to adjust the MERGE DISTANCE parameter interactively and see what the results are, instead of the iterative process we have now.

While this would probably break down at extreme (or even not-so-extreme) point counts, if users were disciplined enough to restrict the use to just Selected points it might be plenty fast, especially when you're playing "find the wrinkle" in Subd.

(IMO a lot of modelers don't restrict their actions enough with Selection, blindly applying tools to the entire mesh and then disappointed in the results/performance.)

Iterative processes seem inherently slow.

03-20-2015, 11:23 AM
Yep, I plan on trying a merge at a distance of a few mm when I get back to see if that changes anything - if it does I can undo, and check in more depth.

03-20-2015, 12:29 PM
I recreated your crimp on my model.
Because of the crimp "direction", the two points (if that is indeed what it is :hey:) are on the right.

The problem then becomes, how far around do you have doubles. I don't see the corresponding "bookend" crimp going in the other direction on your screen grab. So it's at least a couple polys away. Take a quick spin around the model.... see if it's there.
(If you haven't completed the back of the craft and the doubles go all the way to the open end, there obviously won't be one.)

And mine.... two points indicated in numerics, only one apparent visually (btw, it was hard to select both) 127564

03-20-2015, 01:38 PM
(btw, it was hard to select both)
In these situations WIREFRAME rules-- afaict, WIREFRAME will always select all points within a given lasso volume.

03-20-2015, 01:52 PM
In these situations WIREFRAME rules.

Yup :thumbsup:. I was proving my own point :hey:.
That's why I made a point of including that step in Option 2.

03-20-2015, 03:17 PM
Now fixed! I had the quadrants overlapping for some polygons, all now looks good.
I'll post some renders soon.

Thanks to all who gave advice, I've learned a new technique!

03-20-2015, 03:26 PM
wireframe image? : )

03-21-2015, 05:31 AM
OK, wires and first test render attached.

Short version of how I went about this:

Check the area of the wedge where I wanted to blend, and worked out how many edges I would get. Add some extras in areas they were widely spaced, and took the new total.
Made a cylinder with that many sides.
Did a boolean on extra copies of the parts, as a guide on where I wanted to cut.
Made cuts on both parts, just away from the point they would meet, then did it again a bit further away for smoothing
Took the two parts into one layer, still with the gap.
Took it down to 1/4 to take advantage of symmettry, and checked for 2 pint and 1 point polys etc.
Manually made the bridging quads
Replicated four times for the full set, merged points
Hit tab!

I cleaned out some loops after freezing again.


03-21-2015, 08:46 AM
Oh it all looks excellent now, your description of the process sounds close to what I did except I didn't do the boolean part that you used for reference, it was that part about putting a cut just back from the intersection point that was crucial. It's great to see someone make that leap forward in technique. It hardly matters but you might want to re-think your side fins as in achieve the same shape with less geometry, only because you are learning new ways to create meshes now.

03-21-2015, 09:14 AM
About the only thing that could be better in that process is finding a tool to create the bridging quads automagically.

Possibly.... QuadSew?

BETTER: I just tried BRIDGE by making a flat box w/divisions, deleting the center polys, selecting the facing Edges , and hitting BRIDGE, wallagh.

03-22-2015, 08:10 AM
There were not that many polys to make when working on a quarter of it.

Thanks again to all for the help, here's how the actual model is coming along.