Welcome to the Onshape forum! Ask questions and join in the discussions about everything Onshape.
First time visiting? Here are some places to start:- Looking for a certain topic? Check out the categories filter or use Search (upper right).
- Need support? Ask a question to our Community Support category.
- Please submit support tickets for bugs but you can request improvements in the Product Feedback category.
- Be respectful, on topic and if you see a problem, Flag it.
If you would like to contact our Community Manager personally, feel free to send a private message or an email.
Struggling to turn a curvy 3D-scan mesh into a worthy surface
kenn_sebesta167
Member Posts: 85 ✭✭
I have an cowling on an old airplane, and said cowling has a massive amount of cooling compared to what it needs. (Basically, they spec'ed the same cooling for engines from 100hp to 200hp, and this engine is on the 100hp side).
The 1950s designed cowling looks something like this:
whereas modern aerodynamic intakes are a little less cavernous:
That massive intake on the first plane is responsible for a significant amount of drag on the airplane, and more importantly the air-cooled 1940s engine can't reach operating temperature in the winter.
For winterization, some people stick duct tape or metal plates at the entry to reduce the area. But I'd like to do better, and actually have the aerodynamics get better at the same time!
I did a 3D scan of the cowling:
and imported it into OnShape: https://cad.onshape.com/documents/a5b5595cfd2591c26e078e2e/w/354f988ef8500216ed719495/e/3d5c0a9352239fe7aff47a53?renderMode=0&uiState=67ad0c6b3a0d8067e8ef41e2
Now I'm trying to build a smooth surface to represent the cowling entry. Once I have that, I can build on top of it to make any kind of intake shape I can dream of. But my approach to surfacing obviously lacks. My surfaces are ugly and don't conform well to the mesh. My approach is to make lots of guide curves by placing planes (yay, best fit plane featurescript!) where it makes sense to me, and the projecting nearby mesh points onto the plane to make a 3D spline which is planar.
To my eye, that looks pretty decent, but the resulting surfaces are horrid. Just look at the wrinkles!
Ideally all the surfaces would blend into each other with matching curvature, but I can't even get a basic fill to work when working with curvature matching.
If I use lofts, instead of fills, it gets much, much harder to do even minor modifications, since lofts can't use random points to help converge.
Can anyone suggest a different approach? I'm really at a loss here for where to go next. The more individual mesh points I put in, the more waviness I will have. But when I try guidelines with precise positioning, I get fill failures very quickly.
(Same link as above: https://cad.onshape.com/documents/a5b5595cfd2591c26e078e2e/w/354f988ef8500216ed719495/e/3d5c0a9352239fe7aff47a53)
Comments
I think Loft is going to be a much easier workflow, ultimately. Your Fills are getting way too many constraints, that's why they are getting lumpy.
I like how you're using Edit curve, but wonder about some of the reapproximations you've applied - there is one seen below which is at degree 8 and has a very crisscrossed set of control points. The resulting curvature combs are bad.
3D Spline 18 and 3D Spline 6 meet at nearly 10 degrees, so any tangent connection between surfaces that follow these guides will be impossible…
I'll look deeper, but these are my first takes.
{edit #2]
How much deviation from the scan to surface is acceptable?
^^^ is with Lofts. The end two are multi-profile, and have guides along their loft direction. The middle Loft is tangent at each end and skips the profile created in the Curve 11 folder.
I created a couple of new G1-G1 Bridging curves (seen in blue here) so that they could act as guides. By overlaying the original scan you can get a sense of the deviations.
Without tweaking anything, some rough checks look like < 1mm deviations
Thanks for the thoughts! I feel like you're seeing the same thing I am.
I have gotten decent results with the loft, but it is such a rigid approach. It's a heck of a lot of work, and it's not clear until things break where the errors are coming from. There's a lot which is missing in OnShape to make it easier for this, for instance the ability to select a point on a a 3D curve for creating splines. This means that if I add an extra profile for a loft, I can no longer use the prior guidelines until I go back and modify a series of sketches and points.
Edit curve is also very useful, but it doesn't seem to have a mode to inject new points while preserving the old ones. That, too, breaks loft guidelines.
I'm getting some traction, though:
It was little work to get here, but I feel like it's going to be a lot of work to do refinement steps. Everything, everything, everything I do breaks lofts.
(For context, I'm more interested in generalizable approaches of how to do useful things with the 3D scanner and meshes than I am in "gotta get this done and then never again".)
Unfortunately, if you don't have clean curves, your surfaces aren't going to get any better. In fact, any wobbles or bad curvature in your curves will be amplified by the surfaces. To get smooth curves and surfaces you want to use as few control points and guides as possible.
The mesh itself has some surprising wobbles, especially looking from left to right along the top and bottom front of the opening. I don't know if those are artifacts of the scanning process or the real world parts are wobbly.
You have kind of the right idea with where you've created curves, but there are probably 2x the number you need/want. The areas that you recreated with Fill 1 and Fill 3 should be simple 4-sided boundary surfaces with the straighter curves being degree 2 or 3 if possible. The intermediate curves and complexity are making everything too complicated.
Edit curve is a useful tool, but using degree 3 with a lot of spans to approximate the curves isn't helping you. For the front view, I would try to construct the four main "sides" of the shape with degree 2 or 3 single span curves. Then add the corners with the bridging curve tool. If at all possible I would recommend manually sketching the cross sections with single span Bézier curves with limited numbers of control points (maybe degree 5-7 for the cross section curves). If you do that carefully you can keep the degree and spacing of the CVs even and consistent around the opening. Then surface the four sides with boundary surfaces. You might be able to do the corners with boundary surfaces, but there's a change you'll need lofts and or fills.
I would highly recommend going through some of the tutorials and info on the Alias Help. The underlying math behind the surfacing is essentially the same, and it's important to understand how to keep your curves/surfaces as clean as possible.
https://help.autodesk.com/view/ALIAS/2025/ENU/?guid=GUID-21501AEB-9E7A-4F9F-A0B3-0A4B3431B9BD
I used the edit curves to reduce the number of control points and play with tolerances. The subsequent lofts came out much soother. Can still use some tweaking for improvement. The more imported measurements you have the more tweaking of tolerances needed for smoothing. I've not been able to Boolean these 2 surfaces into 1 as yet and end up with 2 parts when thickening.
https://cad.onshape.com/documents/0eeaea78be73d0251e28a531/w/9b37b9dac8fd537d2c234292/e/46f915bd92ba4828ee1b9a88
And here I had thought that I was doing well with the edit curves by backing off the default 15 control points. Thanks to both @glen_dewsbury and @S1mon for this advice.
@S1mon if I understand you correctly, you're saying that instead of using the mesh points I should try to make some very simple curves which follow the gist of the fiberglass cowling, and then build from there? Neat to realize that to a trained eye there's so much ability to work with curves.
Re your question about the wobbles, it's likely the scanner if what you're seeing is how at the inside lip it gets wavy. I did it in infrared mode on a Creality Raptor, and while the results are stunningly good for the effort, they're nowhere near as good as the blue laser approach. But, also, the physical molds for this airplane's fiberglass cowling likely date back to 1964 so they were probably done with straightedges and compasses. Alas, I have to mate to that, so I inherit all the imprecision of 60 years ago.
Thanks, everyone, this was exactly the kind of education I was hoping for. Now to figure out how to pass that on to my students…
It seems like lofts really was the way to go. This bezel should work on the first print with some gentle prodding.
@GregBrown Not a whole lot of deviation is allowable, since at the end of the day there's a lot of airflow and nearby propeller blades. But 1mm seems not unreasonable.
When possible I try to lead with extrudes, revolves, drafts, fillets, and then come in with surface and curve work where I can't get those tools to provide a solution. Nothing wrong with doing curves and surfaces straight away and I can see myself steering toward that for cases like these but I can also see a case for an attempt with regular tools modeling tools on this object.
@glen_dewsbury check your end conditions on your guide curves 3D fit spline 6&7. One curve is running past and they both need an end condition defined that flows with the curve they are adjoining to. This should allow you to define the loft 2 start and end to tangency and join them. Also why did you choose to define points and a 3D fit spline for the loft profiles? Why not just trace out and adjust a spline/bezier curve in the sketch and use that and skip the 3d fit splines and points? Just curious.
Good morning MDesign. I played with the profile splines but not the guides. Hopefully would have found and improved those as well but needed some sleep. The distortions on an old cowl are not a surprise, From the time frame these Citabria Cowls were made the original plug was probably hand shaped with the inaccuracies that involves. Over time they will also distort some due the wear and tare and stressing as well as patches under the paint. The idea is to get your development faired as best you can. Some hand blended fillers around the edges and across surfaces during install.
For a few years I did this kind of work as a side line. One customer had molds that had been stored so badly that I couldn't work with them and had to rebuild or simply build new. I was surprised at how much distortion could creep into fiber glass parts and molds over time. Had some time this morning. Replaced guide curves after editing profile curves and opened up tolerances. Doesn't look too bad. Only 1 or 2 spots I'd like to smooth a little more but need to get on with other stuff.
https://cad.onshape.com/documents/0eeaea78be73d0251e28a531/w/9b37b9dac8fd537d2c234292/e/46f915bd92ba4828ee1b9a88
@glen_dewsbury Thanks for that last bit of advice! I'm pretty satisfied with how it's turning out. The print needed more support material so as not to string, but it's going to be good enough for a first-fit test.