Can't thicken complex surface model

[mGVD3C]

Trigger warning: Self-taught amateur

I'm trying to model a (broken) latch for a rice cooker as an introduction to complex surfaces, and have tried to create it both as solid body and surface models, failing spectacularly both times. Lots of tutorials, lots of reading posts on this forum and I'm pretty sure I know (roughly) where the problem lies, but have no idea how to fix it (and my design approach and techniques). I have managed to get a 'working' model but it's got problems, a lot of problems.

The object to match:

rice cooker latch

More views:

https://imgur.com/a/Zqqm7vr

My (surface) model:

(edit: attached to post)

When I look at this, there's only one 'flat' surface (the bottom), and some geometrically simple posts and flanges for the actual support mechanism, and interior structural fins. No issues at all with these bits, they take only minutes to design and function exactly as expected. Everything else on this thing is a curve, and it's only me and my calipers trying to guess the rest of it.

The model above is complete, works (sort-of), but has a limitation of being unable to thicken more than 1mm, the part itself is 1.75mm, and I'd love to get it a tad thicker (to compensate for being a 3d print). It's also unwieldy and horrible to work with. Surface examinations show it to be utterly terrible, disjointed, and that's likely (I'm guessing) why I can't get it to function properly.

Let me describe my (surface) approach, because this is what I am trying to fix. The model will need to be re-done, that's fine. I don't even want to really fix this version because I feel it's too full of tiny errors and missteps.

1. Decided on modelling half, as it's a mirror down the middle 
2. Started with inserting reference photos, calibrated to 1:1 and logically placed
3. used those reference photos to create sketches on x/y/z planes, 2 per 'section', using projection and tangency constraints to get points lined up perfectly. 2-4 points per curve, mostly as fit-point splines. The design ended up having 3 'sections', the top angled bit, the front (mostly) cylindrical 'face', and the corner with the hinge post
4. project the intersection of these profiles, then loft the complex shapes using minimal rails where it was not matching the design closely enough
5. stitch 3 main panels together
6. thicken to desired thickness <-- this I couldn't achieve, as some of the strange curvature, probably the bits around the rear corners, seems really mis-matched and also not continuous across surfaces, so I think that's why it's failing. I did manage to get 1mm, so kinda functional, but not desirable as an outcome (or for when I am taking paid clients!). But I don't know how to approach a complex design like this in other ways, so I end up with these failures.
7. mirror, tidy surfaces, install interior structural fins, and 3d print

My approach for this seems to be flawed, but coming from a 2d vfx background, I may know my way around a 3d projection setup like the back of my hand but I can't figure out where I need to change my thinking to get it to work for CAD/solid/surface modelling.

(I've attempted to name all my sketches, bodies and planes, but there's a few in there which escaped. I did not name bodies I did not need, such as throwaway remnants from slicing another bit. Hope you don't mind.)

(I've also been thinking that I might need to design this as a simpler shape, slice off the top bit, then use the shell tool instead of thicken. Is this a solid idea? I'm still at the point where the sketches and projection setup took me the better part of 8h to do, both because I'm still becoming acquainted with F360 as well as the learning process of how the part was actually constructed as I spent more time with it; but I still need to know how to decide on the right tool for the job in more general terms, not only as it applies to this particular design.)

[Edit] to add the actual file, hopefully. It's attached before I hit the 'post' button.

[jhackney1972]

Your model cannot be downloaded either because you are using a Fusion 360 Personal license or you forgot to allow downloading when you shared it.  It is better to simply attach it to a reply post.  If you do not know how to attach your Fusion 360 model follow these easy steps. Open the model in Fusion 360, select the File menu, then Export and save as a F3D or F3Z file to your hard drive. Then use the Attachments section, of a forum post, to attach it.

[mGVD3C]

Ah, thanks! First time. I've edited the post to include the file.

[davebYYPCU]

Plenty of pictures in the archive, but no Fusion file as of yet.

Your description sounds about right, just need to see the data.

Might help...

[TheCADWhisperer]

---

@mGVD3C wrote:
Ah, thanks! First time. I've edited the post to include the file.

---

Do you see the *.f3d file Attached to your edited post?

You must give it time to fully upload before clicking Post.

[mGVD3C]

Tried again and this time I see it showing up. But I'm attaching here as well for posterity, just in case. Thanks for the patience with my initial n00bism, I promise I know more about curves than my posting suggests!

[TrippyLighting]

So lets roll the timeline back at the very first surface you lofted and look at the isocurve analysis:

Without actually looking at the loft dialogue we can observe that the loft was created from a curve to a point. Lofting into a singularity (the point) creates curvature problems and should generally be avoided. 

This single surface can be thickened to 1.5 mm, but should terminate in a single edge at the tip. The odd shape created by the thickening operation is yet another indication that something isn't OK with the surface. 

 Now lets take a look at isocurve analysis of the second surface:

Do you see the discontinuity in the curvature comb of the surface ?

That is the result of the 2ns loft profile:

Do you really need that ?

Less is often more!

The third surface is even worse. Here you also are not constraining all boundaries of the surface:

Also, there two surfaces are connected without tangency or continuity. The reason is that you are not even given the option in the loft dialogue is that you selected a curve for Rail 1 instead of an existing edge of the bordering surface and because the 2nd profile is not tangent or continuous to the "other" profile.

[mGVD3C]

First, an honest thank you for the detailed report on my surfacing, as I honestly did not realise how 'bad' these surfaces were in a parametric environment. I did see some of these issues but since Fusion allowed them, I was happy to accept 'poor practice' in service of 'moving on', but didn't realise that would bite so hard, so soon. The 'loft to one point' is admittedly not best practice anywhere, but the rest wouldn't faze me in a mesh world for visuals only. I'm not a modeller by profession but I do work directly with them, and this is common in the media world. Even if it leads to poor surfaces, as it obviously does, there's just so much smoothing going on that… well, gotta get it out the door.

So, in more general terms, and I'm guessing you're very much more advanced than I am, how would you approach this from the start?

- I don't have any drawings, only the object

- I want to learn the best method for a *very* precise model, even if it's not strictly needed for this exercise

- and I want the result to be as editable/modifiable as possible, reliably (or as much as is possible)

My bread-and-butter software, the one I know inside and out, is Foundry's 'Nuke' compositing software. It's node based, so creates a tree of editable nodes, not too dissimilar to Fusion's timeline. I get the tools, but I don't know the theory or what actually works in parametric land. Any advice on how you'd fix the approach would be worth thousands over how to fix this particular model. I extrapolate incredibly well 😉

And again, thank you for your in-depth and valuable response. There's a lot of great stuff there I will certainly keep in mind going forward, especially about QC'ing my early shapes instead of believing the software is working because it isn't throwing an error (and it does throw a lot of those!)

[TrippyLighting]

Some of the things I wrote are more meant to highlight limitations of Fusion 360 surfacing abilities and maybe provide guidelines to  better modeling workflows. The "cleaner" a surface is, the more likely is a successful surface model. 

Not all of it necessarily creates the thickening problem.

For example, rolling back the timeline, without the side face, the object can easily be thickened to 1.75mm:

I am still trying to decipher how the model was built 😉