FDM, What is it (actually) good for?

Ray Ma -

FDM ogres,
Have layers much like onions.
More donkey than Shrek

(note: I think I started writing this in January or something. Thought it was a good idea, then totally lost steam, which is how my posts typically go, but so much more so for this one)

I think it's safe to say now that 3D-printing has arrived. I think it's also safe to say that for the vast majority of average users (not researchers or pro-sumers), 3D-printing's been more hype than game-changer. Sure, the use of 3d-printing in industry has been really exciting, what with its use now in aerospace applications, customized prosthetics, and prototyping (in practically every hardware application), but I'd like to argue that despite a substantial decrease in price these past few years (thanks Monoprice!) the basic desktop FDM printer has little utility beyond spitting out cute paperweights.

Maybe the technology's not there yet####

There's certainly a lot of (deserved) attention paid to other processes like SLS and SLA becoming more accessible (thanks Formlabs!), but an argument could be made that the FDM process itself is still being refined, through new filament material and print hardware. Even when considering just PLA and ABS, there are so many variants nowadays, from glow-in-the-dark to carbon-fiber-reinforced to simply more optimized mixtures that optimize bed adherence and print consistency. Structural materials like nylon and compliant filaments are becoming more accessible and easier to print (thanks Taulman and Ninjaflex!). On the hardware end, print repeatability's been greatly improved thanks simply to better enclosure design (goodbye printed brackets and cantilevered 80/20s), more durable nozzle materials, and more refined motor controllers. Auto bed-tramming (or leveling, depending on who you ask) is becoming more prevalent, and machines now can go much longer before needing any manual servicing. Even multi-material printing has become more mainstream (with ongoing support too, thanks Lulzbot!), although it's arguably been put to use for more aesthetic goals than functional.

Although the humble tech behind FDM should continue to get better, more dependable, and cheaper, I don't think there's going to be any fundamentally ground-breaking developments that will revolutionize how we systematically squirt melted plastic out of a nozzle. The basic constraints of FDM (or most layered deposition processes) aren't going away:

  • Inter-layer delamination means parts are stronger in the plane determined by the build/print direction
  • Without dealing with support (removable or dissolvable), overhanging features won't print particularly well (if at all)
  • Even with variable nozzle diameters or or other forms of infill (like epoxy), FDM is relatively slow for larger models
  • Due to the print speed, FDM can't be expected to scale very well for mass-production (>1k units)
  • Surface finish is limited by the print speed, nozzle size, layer height; usually on the order of 50 microns at best.

I don't think the constraints above should be thought of as deal-breakers, even if they never get adequately addressed in future FDM implementations. Additive manufacturing, even with these constraints, still provides at least one key advantage over traditional manufacturing techniques:

  • Ease in making custom geometries at low cost

It's just a matter of figuring out what applications prioritize that need over part robustness and resolution.

Stuff That's Worked (or not)####

I feel like most of the so-called lauded examples of FDM (aside from nifty paperweights) just use it for short-run mass production when it's not economical to make a mold or use subtractive machining. Don't have the links on-hand anymore, but I remember that the early "success stories" included some support bracket for Square payment add-ons for smartphones. 3D-printing allowed the entrepreneur to skip the set-up costs usually required for casting/molding and profit much earlier. It filled a niche application that no other vendors were addressing just yet, and (apparently) customers were willing to accept a relatively poor quality plastic doohickey.

At a more industrial level, several airlines have proposed printing support brackets to minimize weight, under the assumption that geometries not easily achievable by conventional subtractive manufacturing could be more easily made via additive. These are typically simple supports for seats, but SpaceX has gone as far as to print flight hardware. These applications however, rely on metal SLS printing, so it's not exactly a readily application for the typical user.

Wearables and prosthetics have become a more interesting space for consideration with the introduction of flexible filaments. 3D-printing can produce a more customized fit for each individual. Normal earbuds scanned customers' ears to produce earphones with a better fit. Shoe companies have been trying out 3D-printed shoes for a while now. It seems like low-cost prosthetics initiatives and companies are popping up left and right to try and address the high degree of variability in impairments for the amputee community. The cost-to-benefit ratio for the consumer seem to be the primary pitfall here. 3D-printing doesn't do much for the assembly/integration process here, so at these low volumes, I'd wager that most (if not all) of the final production is done in the States, by hand. As a result, these might as well be luxury items unless there's a charitable organization driving some initiative.

Stuff That Should Work####

I'd argue that the best use cases for FDM should fulfill the following characteristics:

  • Non-structural (or only need to withstand small loads)
  • Complex geometry that would be tedious to recreate through traditional methods (ie. internal voids)
  • Acceptable to have short lifetimes (as consumables or easily swapped components)
  • Minimal assembly steps (or at least assembly steps so easy, the consumer can do it themselves)
  • Minimal reliance on extra components

For all its hype, 3D-printing produce standalone utility by itself. It produces a "glue" structure for other components, either as an enclosure or a connector. Maybe that's boring, but I think that's what 3D-printing is best tailored for producing. This Disney Research paper probably encapsulates what I consider to be the best use-case for 3D-printing. Arguably, it's also its primary use-case.

The old adage "Form follows Function" certainly applies here, but for me, it means that function is specified first, by traditional components, and then 3D-printing resolves the final form through the support structure. Instead of trying to replace conventional manufacturing in its entirely, I think the focus of additive manufacturing should be to enhance existing sub-components in interesting ways. To be fair, I'm still not sure if the dearth of interesting applications reflects a lack of creativity or the true limitations of FDM.

What's Needed for More Stuff to Work####

I think the key challenge is not only finding an application where custom geometries are necessary or highly desirable, but also figuring out the easiest way to specify the necessary geometry. Even with the ability to command arbitrary geometric parameters, I think most users are restricted to the limited intent of the original part designer. There's work to be done on streamlining the pipeline between functional requirements and printed part without having to rely on someone with an engineering background. Sure, there are parameterized designs, but I'm looking for another step (or a couple steps) beyond. For a bracket between a ledge and a shelf, for example, the user should just take pictures of the two interfacing surfaces. If I need a support bracket for my phone and some add-on, I should be able to do that with just the phone and add-on model, without having to measure and tolerance everything myself. Beyond that, there should be controls to fine-tune the geometric properties where allowed/necessary.

Better scanning? Perhaps the issue is in scanning/sensing. Noisy point clouds just don't cut it. Manual measuring and shape matching seems to trump even state of the art sensors. Even in the DARPA Robotics Challenge, some teams had human operators manually drag-and-drop shape primitives onto the scanned point clouds instead of trying to automatically match target shapes. I suppose trial and error is always a valid approach, but better scanning resolution or shape matching can make it way easier to make more interesting assemblies.

Better part libraries? On the other hand, what if companies made CAD for their products more widely available? I've talked up Teenage Engineering's approach before, and I'd point to custom mechanical keyboard keycaps as another instance of what's possible when the mating interfaces are either well-documented or well-known. Both of those cases are primary aesthetic, but imagine a set of printable connectors not just restricted to simple plates/panels but instead could be specified to mate with other off-the-shelf components.

Just Uber it? Hey, maybe we should just let engineers engineer. Nothing wrong with making it easier for customers to find a more experienced, CAD-capable individual to do the necessary work. It'd be interesting if an extensive survey of CAD requests would reveal certain classes or categories of parts that could be parameterized for future work.