Both smash and UnrollSrfUV dont come near (90% h and 83% w) to matching simple shape

compound curve , tried smash and then UnrollSrfUV, neither are even near the right shapes dimensions.

correct shape at right. Both width and height an inch and more out, 83% on width, 90% on height.
Is this proportional I wonder, can I scale up to get my two shapes (two days on establishing the size of a simple rectangle, ( a week on fabric, text and now an apparent simple rectangle or two, even further behind deadline)

What should I use to Unfold this shape ? somewhat accurately more than this ?

Also why did it destroy the most planar area at hole ?

This is an eye opener for the smash command though. UnrollSrfUv a tad better.

analyze smash.3dm (958.3 KB)

Something that creates isocurves at intervals over length of the surface, then lays them out keeping the intervals and places a surface to their tips !


Hi Steve- the sort answer is you cannot cleanly unroll non-developable surfaces.
Squish may help.
Note your actual canvas will not be the same shape as this surface - it will have folded in places to make up the difference between that non-devlopable compound curved shape and the flat canvas.


cleanly being a too great an error to be of any use., 17 and 10 % eeek !
Can I scale those would they be proportional or best avoided ?
I will explore also Squish.
um…it wont allow me to select the curves rectangles, which I need to see true flat shape of.
so cant use squish.

I am after establishing these rectangles (canvas tunnels) when flat, thus what do I do ?

My idea of a command that does the isocurves, runs intersect with the surface edge, analyses lengths of each intersect and spacing and lays them out etc, can that be turned into a useful command ?


1 Like

You’re off in left field here, what you’re trying to do is fundamentally a mistake. Fabric doesn’t behave anything like what those tools do. This isn’t a 3D CAD problem, this is something you need to work out more practically.

1 Like

I am there but for unroling the surface with rectangles.
I dont think having to construct a massive dome from cardboard, trying to then plot these rectangles on it, (how ?) and then removing the paper and laying out flat is going to be seen as being up to modern methods, if smash worked then not a problem, squish doesnt allow selection of the rectangles.
1 week on ripple recreation, forget that, then text on a compound curve, forget that, (no persevered 2 days and eventually got surfaces trimmed to text, (decals no use) then a simple unroll surface, I dont mind a few % BUT NOT 17%. Would anyone accept such on a simple shape ?

What other way ?


Smash works - your surface is not developable.


As its a planar surface flowed then pull to a dome made from an elipse rotated, what is it about such a surface that is not developable, if I was to trim the dome with the edge of my surface would the result be developable or is it that any portion of a surface made from revolving a quarter elipse curve is not a developable surface ?

I scaled the shape UnrollSrfUV gave to correct length, the rectangles are not proportionally correct.
Width of tunnel 1.2 versus 1.4 on dome.
length 4 verses 3.8 on dome.

What do I need to do to make the surface developable ?

whole day on a simple rectangle. and at the moment its stuck on the dome unable to be removed and analysed for true dimensions.


Have you ever noticed how the panels of a hot-air ballon are cut and arranged? I suggest you first cut your object into strips parallel to the long axis. Make the strips narrow enough (trial and error) that the curve in the width direction is very shallow. Then the resulting surfaces will be very nearly developable to the point that the error in the width when flat will be small and the length error will be practically non-existent. Flatten them.

1 Like

It has compound curvature, it’s curved in 2 directions, it can’t be “flattened” without stretching. Of course that’s not how fabric things are made, so trying to make it developable isn’t going to accomplish much.

What is this all even for? If you’re trying to make a render of something fabric, or trying to reverse-engineer something fabric, those are totally different, unrelated tasks.

though it shrank. !

I thought SMASH was for compond curves, the little video shows it, and youtube video analysing all three methods shows a sphere with sliced top and bottom and hole in it.
rhino help says:-

‘Makes an approximate 2-D development of surfaces that have compound curvature.’

What has Rhino to use for a compound curve then , ?

I am trying to establish the 8 petals and canvas tunnel that existed on an 80 yr old device, so projecting the tunnel in a photo to a dome of diameter and size as it was, should give me the size and location.
The entire device has no plans, is extinct and my work will bring it back from extinction and give public pleasure in seeing it on display again.
if I get length 3.85 inches I will be made as 4.

as such I am measuring instead the curve lengths of it on the dome, do it that way,
enough time messing about with smash. should have been drawing plans up 3 days ago as I finally promised, and 3 weeks ago to meet with original target. until I met barriers on text on compound surfaces, and projecting pics to objects.


Yeah that’s what it does. And it’s APPROXIMATE. Once material starts stretching/compressing, all bets are off, trying to “simulate” it is the purview of very specialized simulation software and a lot of real-world experience. It’s absolutely not intended to simulate fabric, trying to use it for that is a mistake, if anything it’s most like sheet metal in a mold…that’s about as far from fabric as you can get.

I did once develop a method for making 3D of hot-air balloon envelopes based on given pattern specs in order to place designs on them and turn those into properly-distorted 2D patterns. That’s a far more straightforward situation than your lumpy bundle, I’ve forgotten more about Rhino than you’ll ever know, and it took weeks. So forget about finding some magic zero-dollar Rhino 5 push-button solution for this and focus on what you actually need to do.

Getting some paper and making patterns and seeing what happens–which even itself it a very rough approximation–makes a damn sight more sense than what you’ve been doing. Or–well in your case I think it’s “and”–use brute-force sculpting to make a lump of surfaces that approximates the look of the final thing for now, with zero concern for how it might “actually” unfold into patterns.


I dont have it as a folded fabric item though, I have it as a Rhino smooth surface, a dome formed from a quarter elipse. The fabric doesnt even come into it at all, the final object will be made from fabric BUT I am dealing with rhino surfaces, using rhino tools designed for those non rippled normal surfaces.

How do I make a 1/8 slice of a dome developable ?

Smash is for compound curves it said, my one eight slice from a dome is such.

As for getting cardboard cross sections of a dome, gluing together, covering with paper, smoothing it all out, then placing the photo in a slide projector, aim at dome, altering the angle of the paper construction until a fit is obtained, and tracing the canvas tunnel rectangle on the paper assembly, then cutting round it and flattening it out on a table, just to find out its width and length and distance from dome centre, that has to be questioning the failure of a CAD program if the user of that program is reduced to tactics 50 yrs old. Imagine expressing interest in buying a CAD prog and being told one must do that. I am just trying to see if its 4inch long or 4.5, its width and distance from centre, is it 2 inch or 2.5 inch.

to trace it on photo, angle the object, match to photo, and then project to object, that is how we do it in 2021. To then flatten out that petal is to see the true shape. If it has to be turned into a series of thin poited surfaces as the world is seen in some maps flattened out so be it.

As such I measured the tunnels length and width whilst on the dime using curve length, fabric though plays no part in this at all. I had a smooth surface and wanted to flatten it out, its only got a slight dome to it, start and finish areas are relatively flat. why the most flat part at dome center got split and warped outwards I dont know.

My manual method somewhat time consuming was extract a set of isocurves e.g. 0.5 inch apart on petal surface, intersect with petal edge, redraw these as planar linear curves adopting the lengths found, join up the ends, hey presto the shape !
If that could be coded thats the solution.

and asking again, How do I make a 1/8 slice of a dome developable ?

another idea, the petal was planar, then given flow, then pull. What if the flow surface was shown again, the shape on the surface pull commanded to the flow surface, then flow surface was given UnrollSrf, that would do it.


I am assuming the dome shape is axisymmetric. Create two curves on the dome which are coplanar with the dome surface; ie curves of the edges of the the slice. Loft between the curves. The developable surface will be straight between the curves.

Try ExactFlat. It works.

I’ve been using Rhino to make patterns for at least 15 years. Having tried with several different applications, including those specifically for making patterns for sewn goods, I can say that none of them are going to nail that with 100% accuracy. As said above, Rhino would have to stretch some areas, shrink some areas and in the end would still get it wrong because it can’t tell what your final intention is. When I use Rhino to make patterns, which I’m doing pretty much weekly and often for products that contain over a hundred separate fabric panels, I often have to make two models. The first is fully rounded out and looking like I hope the final product will look after sewing, etc. That model is purely for presentation and getting design approval. Afterward, I remodel, often using many of the same curves I used to make the original model but trying to make the surfaces flatter or even developable wherever possible. I always try to have at least one developable surface in the mix because it will unroll with accuracy and I can use it as a reference. When this model is finished, I’ll often scale it up by some amount. This amount varies depending on the material, thread and stitch count. The reason is that a tightly woven fabric sewn with a heavier thread will “shrink” where sewn. That’s because the fibers in the fabric are being displaced by the thread, making some of those fibers take a more circuitous route from start to finish of the sewn seam. That causes the fabric to shorten along the seams. It also has the added advantage of adding roundness back to the sewn panels. Then I use squish to flatten that second model. Sometimes I’ll even make a rounded panel, what I really want and a flatter, developable panel of the same pattern piece, then squish or unroll as appropriate and compare the two, using the unrolled developable surface as a reference to see how I can modify the squished pattern to get what I want. Even then, your work isn’t done. Now you have to check each panel against its mates and make sure things add up. You may also have to add ease or any number of pattern maker’s tricks to make the final product look the way you want. You’re not going to see any software that can do that perfectly. I’ve spent a bunch of time with Exactflat and found it no better. It can be if you get the material properties exactly right, but even when I did, I found it to be way slower and still less accurate than using Rhino on its own along with some basic pattern making skills.
With Rhino, I’m way faster and more accurate than any pattern maker I’ve ever worked with, even when they’re using Gerber, the de facto standard and the first round prototypes are typically require significantly fewer samples to final approval. CAD is awesome, bit it’s still no substitute for specific knowledge.


Hi Steve,

I am not sure if you have found the solution to your problem, here are my two cents.

I have found this post by Tom_ where he goes in some details about what makes surface not-developable At the end there is a video that explains the ideas of flattening in detail. How to Unrolling doubly curved surfaces - #3 by Tom_

One way is to do the analysis of gaussian curvature. The tool in Rhino 7 is called CurvatureAnalysis, I do not know if it (and the commands mentioned further down) exists in Rhino 5. The rule of thumb I use to imagine what surfaces are developable is to think of a surface being created by straight ‘stick’ that travels on top of two curves. The surface created can bend only in the direction of the stick’s travel, but not along the stick itself.

My understanding that Loft tool does not necessarily have to create developable surface and it does not work on the curves that cross. DevLoft should create developable surface. For it to work though it’ is best to make sure that the curves are quite smooth with not too many control points as I had experienced it giving unexpected results if the curves were not very ‘clean’. Help menu of each of these commands includes a fair share of information as well on developable surfaces.

I fully agree with bjorf’s point as well, that every model is only an approximation of reality. The question of tool selection in the model will depend on the accuracy needed and tolerances required to produce a ‘good enough’ model.

dear @krzys thanks for the crosslink to my other post.

what you describe is a ruled surface
what you describe as “stick” … is the ruler ;-D

all developable surfaces are ruled surface. (“all volkswagen are cars”)
but not all ruled surfaces are developable. (“not all cars are volkswagen”)

a hyperboloid is a simple example for a ruled surface that is not developable - because it is doubly curved

hope that helps - kind regards -tom

Thanks Tom!