One of my projects this summer has been analyzing the mechanical properties of 3D printed parts using a homemade testing machine. I’ve been trying to figure out how to optimize the printer settings to maximize the physical strength of parts while minimizing the plastic needed to make them.
As simple as smashing a few pieces of plastic sounds, I’ve had to face a fair amount of challenges on this little project. There’s a reason for the general lack of test data on this topic. It’s not just a matter of sitting down and pumping it out, but rather spending a lot of time redoing tests and staring at graphs that don’t make sense.
In this post I want to discuss some of the challenges I’ve encountered and share the results I’ve been able to produce so far, including what I now consider to be the best infill pattern.
Here’s the quick recap of the experimental test setup:
- All specimens were 3D printed on my FDM type Reprap PrintrBot, using Slic3r and Repetier Host software.
- All specimens were made of ABS filament from the same roll.
- All tests were run in a static 4 point bend configuration using TestrBot.
- Three of each specimen were used for each variable tested.
For science to work properly I had to double check that my results were actually caused by the variable I was trying to test rather than by something else that interfered. And there were many potential sources of unintended interference:
Artifacts of Software Processing Magic: The programs I use to prepare my 3D models for printing are good, but they aren’t perfect. There’s a lot going on behind the scenes in this software and there are many seemingly innocuous settings that can have huge effects on the printed part.
One example I encountered was that playing with the extrusion width can greatly affect the mass and relative strength of the finished part, even while keeping all other settings constant. The crazy thing is that this effect is completely unpredictable! Try to find a pattern in this fancy pants graph.
Ya know, maybe I’ll just leave the extrusion width settings at their default values.
Artifacts of Physical Manufacturing: My 3D printer is home built and has its own unique set of quirks (by which I mean I have to hit it sometimes). One of the biggest concerns here was the need to be aware of minute printing imperfections and even subtle amounts of warping. Any warping could cause the specimen to fall over prematurely, rather than break, resulting in a lower ultimate load. Testing three specimens per variable helps weed out outliers, unless of course your printer messes up three times in a row!
Artifacts of Testing: Human interference, changing ambient environment, and test fixture instability can all become the cause for a bad test run. In the example below the specimen yielded so much that it eventually came into contact with a screw on the test fixture, causing the load to spike. All data beyond the purple arrow is useless.
Results and Recommendations:
Despite the above challenges I was able to run enough good tests to confidently make some reasonable evidence based conclusions. These are my recommendations for maximizing the strength of 3D printed parts:
- The rectilinear infill pattern is the most efficient use of plastic and it also prints at the fastest rate! (The first part was unexpected!) The other two most useful infill patterns, hexagonal and 3D honeycomb, were slightly weaker in terms of yield stress per unit weight of plastic used. They also took 30% longer to print than rectilinear.
The primary problem with the hex and honeycomb patterns is their inefficient use of plastic infill near inside walls and their asymmetrical lattice structures. The dimensions of the combs are determined by the percent infill you choose rather than as a function of the space they are trying to fill. The result is that they don’t fit evenly into the shape and one side ends up being weaker.
- To maximize the strength of prints for specific applications, orient your infill pattern to be normal to applied bending loads, and at 45 degrees from applied tensile loads.
- Printing the specimens from the ‘top’ orientation was a loss all around. The resulting failure mode was a delamination of the outer layers. (Note that if I had printed these specimens oriented vertically their bending strength would undoubtedly have been much worse.)
The lesson here is that both the load orientation and print orientation have a significant effect on the strength of FDM 3D printed parts.
- Always print parts with 3 perimeters on all sides. The overall quality and print strength is greatly increased by having more than one perimeter, but the returns diminish soon after for multiple perimeters. (The exact diminishing rate will vary based on the part geometry & dimensions.)
- The strength of materials as a function of infill percentage can be adequately approximated as a linear relationship. I saw no diminished return in strength no matter how much plastic you use.
I decided to re-evaluate the above conclusion when I noticed that even though the graph was linear its slope was less than 1. That means I DID observe a reduced effectiveness for adding additional material… Doing some more research on testing done by others I determined that linear 1:1 scaling estimates work well with tensile stress, but not with bending.
What I mean to say is that under tensile stress a part with 50% infill will be 50% as strong as that part with 100% infill. BUT under bending stress that same part will be more than 50% as strong.
The reason is that bending stiffness is mostly a function of extreme fibers. (Remember: Stress = Mc/I?). Adding more material near the center of the part provides very little additional strength in bending applications.
- And finally, my earlier experiments showed that acetone vapor treatment reduces the strength of ABS plastic parts by a measurable amount; for me it was almost 20 % under bending loads.
I’m happy with the progress I’ve made so far. Not only has my cheap little testing machine proven adequate for my needs, but I’ve recently become aware of two separate individuals who are attempting to build their own TestrBot to continue this line of work!
There’s a lot I’m still curious about and I’m not done with the work I started here. But for now I need to take a break from testing so I can focus on an upcoming electric vehicle project. 🙂 (You all know how I feel about electric vehicles.)