At my last company we spent countless hours trying to decide how to manufacture the parts that we needed to build PlantLink. One of the resources that we relied on was the Rapid Prototyping Lab at the University of Illinois at Urbana-Champaign. The lab has four different 3D printing technologies available: Stereolithography (SLA), Polyjet, Fused Deposition Modeling (FDM), and Selective Laser Sintering (SLS). Each of these 3D printing methods has a unique set of pros and cons that I won't get into just yet. In the end it didn't matter which process we choose - we would always run into some limitation that would prevent us from getting an exact copy of the parts we wanted to produce at quantity. Luckily for us our parts were relatively simple, so we were able to pull the trigger on injection molding without having to produce perfect prototypes first.
Fast forward a few years and now I'm working for a 3D printer company. We make an incredible desktop SLA printer called the Form 2. One of the perks of working here is unlimited access to our printers and materials for personal projects. It's a pretty awesome perk that I don't take advantage of as often as I should. Most of the time I've just printed things for friends or as party favors for guests at a couple of events my wife and I have hosted.
Since I'm constantly exposed to SLA parts these days I know more about their pros and cons than I did when I was making parts for PlantLink. In broad terms the detail and surface finish of SLA parts is better than most other processes and you can use a wider variety of materials. Sadly most of those materials are a poor match for dynamic applications due to their brittle nature. There are a lot of exception and caveats to all of those statements, but that's how I think of it when I'm trying to decide if I want to prototype something on a Form 2.
Ok, so I know more about SLA printing than I used to. So what?
Well, my knowledge of other additive manufacturing processes has been pretty stagnant since my PlantLink days. A few weeks ago I decided I wanted to learn more about some aspects of 3D printed parts across a variety of processes - specifically dimensional accuracy and material wear resistance. I had a few different ideas of how to go about testing these things, but I decided on one relatively simple design for a machine that would satisfy my curiosity. I designed a system using small gears produced from six different manufacturing methods: injection molding, SLA printing, FDM printing, SLS printing, metal casting, and laser cutting.
The concept here is pretty simple. A central gear drives each of the five arms of fixed spur gears, which in turn work together to spin an outer ring gear. That's it. There's no greater purpose or value that this mechanism creates. It's not a clock, it's not a clever power transmission. It's just a few gears spinning around in circles.
Quick side note: I used Onshape to design and document this entire project. You can access all of the files - from the math for designing the gears to the part files I printed on my Form 2 - on the public project page.
Since those gears are all ostensibly produced from identical sets of plans they should be exactly the same size. Obviously this wouldn't be a very interesting exercise if that was the case, so I set out on a journey to produce gears six different ways and compare the results. I wish that I could say there was an elaborate matrix that I constructed to select the production methods I chose, but in reality I just picked a few different processes that I was curious about and moved ahead with things from there.
Finding injection molded gears that also come with detailed CAD drawings might seem like a difficult task at first. Thankfully, McMaster-Carr is there to save the day. All I had to do was browse the McMaster product page for spur gears until I found a few geometries that I liked. Once I picked out my gears I just downloaded a solidworks file for each one and uploaded it to Onshape. I repeated this process for all of the mounting hardware I needed for the project so that I could create a precise model of my assembly before building everything.
Once my assembly was complete I purchased several sets of the gears from McMaster Carr and exported STL files for the gears from Onshape so that I could 3D print the other four arms of the design. I chose binary format STLs in inches at high resolution when I exported my files.
Getting my hands on FDM gears was simple enough. My friend Brian has an UP mini printer at home that he used for a lot of his projects before he started working at Formlabs. He's a cool guy, so getting the gears involved sending him an email and asking nicely. Having engineers for friends is pretty awesome.
Producing the SLA gears was a little bit more complicated. I wanted the gears to mesh as well as possible with one another after I printed them, but I also wanted them to have some resilience. This seems like the perfect application for our Tough resin. I wasn't sure which support structures or orientations would make the best parts so I printed several copies of the gears in one print.
My print came out great, but I forgot to take a picture of the build platform before removing all of the parts. I need to get better at this whole project documentation thing...
Since I don't have access to a multi-hundred thousand dollar SLS printer or commercial metal casting facility I had to go through a service bureau for those versions of my gears. I chose Shapeways since I've visited their facility in New York once before and they seem to be the industry leaders in the space right now. Using Shapeways means that I had a bit of a wait on my hands, but until there's a low cost solution for SLS printing and metal casting I'll just have to learn to be patient.
It ended up taking Shapeways around three weeks to ship me my order. I've had much faster results in the past when ordering only SLS parts, so I assume that the casting process was the bottle neck here. Once I received my Shapeways order I was able to compare the results from each manufacturing process.
I broke out some calipers and captured three measurements for each gear: height, gear outside diameter, and base outside diameter. Get ready for some bar charts.
Here's that same data presented as percent difference from the specifications on the drawings. Positive values mean that the measured dimension was larger than the specification, while negative values mean that the measured dimension was smaller than the specification.
Some quick observations:
- I don't know what orientations the SLS and cast gears were placed in the build volume when they were printed, so drawing conclusions about how growth or shrinkage happened in those processes is challenging.
- The injection molded gears were typically closer to the specification than the other processes, but not always, and not by very much.
- Brian's FDM printer is really well calibrated. His prints were much closer to the specifications than I expected them to be.
- SLA printing is unsurprisingly quite a bit more dimensionally accurate than SLS printing.
- The cast gears were smaller in the diameter dimensions but larger in the height dimension. I was expecting them to be smaller in all dimensions based on some common casting issues.
Overall their was a fair amount of discrepancy between each production method and the specifications. If you are interested in doing your own statistical analysis the raw measurements can be found in the Onshape project or directly in this google sheet.
Assembling the device should have been a fairly straight forward task, but I ran into a few small snags. The inside diameter of each of the additive manufactured gears was too small to fit on the shoulder bolts that I selected for the shafts. I used a small lathe at work to remove a few hundredths of an inch from each gear. The laser cutter that I used to make the enclosure and the ring gear (my 6th gear preparation method) had some strange kerf issue that made me have to re-cut all of the acrylic a couple of times. All in all pretty basic stuff, but good things to keep in mind for the future.
Once I had everything together I snuck into the photo studio at work to use the lighting setup that was leftover from some recent product shots. Don't get excited - the image quality won't be too high. I still used my iPhone to take the photos and videos.
The finally assembly looks pretty close to what I had in mind when I made the CAD model. Nothing all that impressive, but I'm still happy with the results.
Turning on the DC motor causes everything to rotate exactly as planned.
From here I'm planning on leaving the system running for a week or so and then measuring the outside diameter of each gear over again. Hopefully I'll update this post when that happens. I'd also like to build some kind of automatic data logging system for the setup at some point. I've got a spare Photon sitting around that I got at a conference last year. If I run a motor off of that and attach a rotary encoder to the drive shaft in the box I should be able to count total revolutions for each gear pretty easily. This might just be the perfect useless object to connect to the internet.
Overall I'd say that it was pretty simple to design, fabricate, and assemble this device. McMaster-Carr makes it easy to get CAD files for standard parts, Onshape makes it easy to manipulate CAD files, accessible 3D printers like the Form 2 and UpMINI make it easy to produce my own gears, and Shapeways makes it easy to order gears from materials I don't have access to otherwise. It's a pretty amazing time to be a mechanical engineer. Next time I'll have to find a more ambitious project to tackle with all of these amazing tools.