Design for 3D-Printing

Great article. This is all above the skill level of the average part on thingiverse or printables, but the good parts on there are going to follow similar ideas. Love the mouse ears, press-fit holes and step-by-step alignment of layers to build impossible bridges.

Notably, in fusion 360 this would all be designed in "plastics" mode, and yet that mode is oblivious to whether the part is printed or moulded. I wonder if any CAD engine can do "production-aware design" that constrains design to the capabilities of standardized machines, e.g. keeping a metal part 3-d millable. I've seen strict design rule enforcement with PCBs, and I have seen sheet metal macros, but nothing for general mechanical CAD.

I've investigated this space, and I'm not entirely sure its even a desired goal from the perspective of a mechanical designer. The benefit tends to be for smaller aspects (ensuring hole sizes are appropriate for the desired thread, or that holes aren't too close to a bend line on a sheet metal part, etc) but the final design of a 3d part is so non-deterministic, and the variety of manufacturing methods are so varied and unique, it might just cause more issues than benefits.

> I wonder if any CAD engine can do "production-aware design" that constrains design to the capabilities of standardized machines

At this time, none that I'm aware of. I am considering some manner of FreeCAD workbench that would integrate slicing to allow specific printing techniques to be applied to specific features of the part. I'm still not sure exactly what it would look like or integrate into the workflow yet.

The people I've talked to found it similarly unnecessary. But mainly because they weren't able to imagine a good implementation.

"But what if I want to do x" is what I heard the most. Like, sure, if you want to make your part on a 3-axis router and then drill one sideways hole, then put that in the markup. CAD always seems to have a feature stack, so apply a 3-axis design rule and discard it before the last step. Similar for multiple setups on a mill, or for surface treatment.

The gold standard still seems to be a signed and printed drawing that is never complete and full of implications. Mapping a design to a factory, or even pricing it, is an art form that has resisted automation. I expected this to change with all the "industry 4.0" push from ten years ago, but somehow that just meant adding wi-fi.

There are DFM tools that you can use on the back end of your design process e.g. through fictiv/protolabs etc. However there is a lot of stuff that is "technically machineable" but way more expensive to do, and it really is an engineer's job to both understand how the part is made, talk to the machinists if they don't, and to trade off the design complexity vs. the engineering needs.

Totally just spitballing, I know nothing about this.

Thinking about the problem, it seems like it would be extremely difficult to come up with a set of design rules that cover everything somebody might want to print.

But would it be possible to literally simulate the printing process? Maybe using some kind of CFD code? I mean, for arbitrary designs this could get really complex. But, there’s a hard limit—the thing actually has to get printed, which is a slow mechanical layer-by-layer process, and the end result has to fit in the print chamber, haha.

I'm actually working on something along the lines of:

>CAD engine can do "production-aware design" that constrains design to the capabilities of standardized machines, e.g. keeping a metal part 3-d millable.

by modeling a part by only using subtraction based on tooling:

https://github.com/WillAdams/gcodepreview

you'll need: https://pythonscad.org/ but it's allowed me to do pretty much everything I've tried out in it thus far, and I'm putting the finishing touches on a joinery module which should let one make pretty much anything of wood, and metals should be much the same --- even turned out a thread cutting program as a proof of concept a while back.

CAD has a number of features to help design engineers with manufacturability. In Fusion, take a look at the Inspect panel - you'll find analysis tools that will highlight areas on your geometry that might have insufficient draft to release from a mold, or areas that might be difficult or impossible to access with a milling tool. With the right extension, you can set EDA-style design rules for injection molded parts.

There are a number of software packages dedicated to design analysis for injection molding, although the price is far out of reach of hobbyists.

Fusion has CAM integrated, which is already a small revolution.

Used to be that CAM was entirely separated software, operated by a separate person who usually spoke fluent g-code.(No seriously, old mills had pushbuttons where you could punch in your G72 with coordinates, and the grey beards would do that with no hesitation.)

CAM software simulates the machining process, not dynamically but with constraints and filters that use the results from dynamic measurements. Printer software is actually quite sophisticated there, Klipper knows about machine dynamics as well as how to compensate for acceleration in the molten plastic.

My issue is that all CAM integration only happens after the design stage. I draw up whatever, then check for manufacturability, and then realize that it's failed again. Electronics CAD software for a while now has had real-time design rule checks that don't let me draw an impossible line in the first place, and I kinda want that.

Oh, that looks nifty. I'm going to take a closer look at this!

Never heard of pythonscad before, but I've always thought it would be a good idea (if only I honestly had the skill to make this myself). Thanks for the links!

I use many of these techniques and have modules in openscad to create build orientation aware parts and have for years. You can be less not entirely sure, because many people already do this.

It was absolutely game-changing for me --- I'd been flailing at this for a long while, and was most of the way there w/ RapSCAD, but stuck on it not being very popular --- within a week of release I had a working version of my library --- it's been nothing but upside since then.