Printers in ConTension

Over the past few weeks, I’ve been running tensile tests on 3D printed components coming off of the variety of printers I have access to at Olin, which include a MakerBot Replicator 2, A MakerBot Replicator 2X, a Stratasys Dimension 1200es, and a Mark One from Markforged. While not all of my tests are complete, I’d like to share some of my initial findings and analysis of those findings because they turned out to be pretty interesting! One thing to note is that I am not directly comparing printers: rather, I am looking at the different behaviors in tension that specimens on each printer exhibit.

Some of the interesting things I found involve tensile behaviors across different axes on the build plate based on how the build plate is fixed and how the slicing software on each printer slices part files. One of the biggest disparities I have seen between the four printers I have been working with is the way that each printer’s slicing software determines the infill (the filament layout on the inside of the part) pattern for a given part. When setting up a print on the Stratasys or Markforged printer, no matter how many parts you are printing at once, the filament path and infill for each part is individually generated. This means that if you were to print the same part on the same printer at the same time, their profiles would end up nearly identical. On the two MakerBot printers, the infill structure seems to be determined by the part’s location on the build plate. Thus, if printing the same part twice, changes in the infill density can lead to stronger or weaker parts depending upon a part’s placement. As my preliminary results show, changes in part layout along the build plate may affect part strength depending on what printer you use.

Notice the differences in infill structure, making the parts not quite equivalent.

My testing procedures have been performed in accordance to ASTM Standard D638-14 (Standard Test Method for Tensile Properties of Plastics) and ASTM Standard F2971-13 (Standard Practice for Reporting Data for Test Specimens Prepared by Additive Manufacturing). On each machine, I printed 3 sets of 3 ASTM Type IV tensile test specimens  on each printer: one set  running along the X axis of the printer (with the length of the part parallel to the front of the build plate), one set along the Y axis of the printer (parallel to the side of the build plate), and one set along the Z axis of the printer (with the length of the specimen extruding up from the build plate). All of the specimens were printed on each printer’s standard settings. While this means that parts from the different printers may have different infill densities or wall thickness settings, these parts reflect how the printers produce parts under usual conditions.

Using a universal testing machine, I ran tensile tests on each sample. Each sample was extended at a rate of 10mm/min. The load was collected over the course of the extension and the dimensions of each sample were used to convert the load deformation curve to a stress-strain curve. From this curve, I then calculated the modulus of elasticity (the stiffness normalized to the specimen dimensions), the yield strength (the stress at which the material will deform permanently), and the toughness of the material (how much energy it absorbs before failure). By comparing these values across different printers and across axes on each printer, we can begin to understand when certain kinds of printers are appropriate for certain applications, and how to best print a part if we know how it will be loaded when in use. Although I have not finished all of my testing and I want to run some more analysis on my data, here are some of my initial findings.

Stratasys Dimension 1200es

Let’s first start with the Stratasys Dimension 1200es. These parts were printed in ABS on Sparse – High Density filling, according to the slicing software. As you can see from the graph below, the printer seems to produce relatively consistent results. Parts printed on the X (red) and Y (green) axes seem relatively consistent: they each have standard deviations of only about 2% in their modulus and yield strength, and about 8% in their strain at fracture and their toughness. There is no clear advantage between the two axes, which I believe is because the build platform is firmly fixed inside the machine, the gantry is just as precise in the X as in the Y direction, and the extruder head automatically levels to the build plate. While the Z axis, in blue, has a fairly similar modulus to the other two axes, the part is noticeably more brittle along that axis. There is a very sharp drop off in stress at a much lower strain value. Consequently, the strain at fracture and toughness are much lower. This means that the part does not deform much before breaking; instead it quickly snaps, as shown by the GIF below on the right. The GIF to the left, showing fracture along the X axis, does not break as cleanly because the testing machine is pulling along the strands of the part instead of the seams.

Stress-strain curves for the Stratasys Dimension 1200es printer. The parts seem to be quite consistent.

Because the Stratasys Dimension is an FDM machine, the Z axis behaves differently than the other two axes. While there are “strands” of filament running back and forth along the X and Y axes, the connections up the Z axis are only composed of stacked layers of filament. When you pull on the samples in the X and Y directions, you are pulling on the strands, whereas in the Z direction you are pulling the strands apart. The strength of the Z axis specimens are thus all dependent upon how well the ABS adheres to itself. As this is a high quality machine, the Stratasys produces fairly consistent results, overall no standard deviation exceeded 10%. This result could be because of the very controlled environment that parts are printed in: parts print in a heated enclosure so their temperatures are regulated, and the printer auto-levels its build plate before each print. Additionally, the printer’s slicing software slices parts individually, and then populates them on the build plate, so each part on the X and Y axes actually have the exact same path profile, just rotated 90 degrees on the build plate. Parts printed on the Stratasys are very consistent and repeatable.

While the moduli across all parts are fairly close, the samples along the Z axis absorb less energy at failure because parts printed along that axis are more brittle.

MakerBot Replicator 2

Next, let’s move on to the Makerbot Replicator 2. You can see from the graphs below that the curves are much less consistent. These parts were printed on Standard Quality, which has 10% infill and 2 layer walls. The Z axis tests are still fairly brittle and weak, and overall parts printed on this printer seem to inconsistently. From the table below, you can see the Z axis specimen data consistently has standard deviations above 20%, despite its mean values being so small all around. On the other two axes, the data is much more consistent but still fairly variable. The parts extend more before fracture, are stronger and absorb more energy before failure. The GIFs below again show the differences between breakages in the X and Z axes: like before, the Z axis specimen has a clean break, whereas the X axis specimen doesn’t have as clean of a break.

The stress strain curve for parts off of a Makerbot Replicator 2. The data has more deviation than the Stratasys.
The stress strain curve for parts off of a Makerbot Replicator 2. The data has more deviation than the Stratasys.

The X and Y axis curves are much less consistent than those of the Stratasys printers, and this is where the differences between each printer show. As I explained above, on the Makerbots, the entire build plate is sliced at once, while on the Stratasys, each part is sliced individually. Each specimen on the Makerbot has a slightly different infill pattern because each is placed on a different part of the build plate. This changes the internal structure of the part and the cross sectional area of the filament intersecting the test region of the specimen. The larger deviation in the data could also be affected by a less stable build plate, as it is leveled by hand and more unstable to begin with. I hypothesize that parts on the MakerBot have a wider range of potential failure because the build plate is not as consistent and the infill density can change the structure about different regions on a part depending upon how they are printed.

Data obtained from the test. Note the wide range of values across the test samples.

Makerbot Replicator 2X

The Replicator 2 and 2X have the same frame and the same slicing software. As you can see the X and Y curves have a lot of variability because of the infill issue brought up earlier. The Z axis seems to be a lot stronger than that of the Replicator 2, which may mean that it is due to differences in material adhesion as the Replicator 2X prints in ABS and the Replicator 2 prints in PLA, but the curves are fairly spread apart. This potentially means that ABS is better at adhering to itself than PLA. The modulus for all of the specimens seems to be fairly consistent, while the strain at failure (and also the toughness, because they are related) seems to be pretty variable. The same qualitative properties of fracture on the X and Z axes are exhibited below, with a clean break on the Z axis and a more ragged break on the X axis.

The stress strain curve for parts on the Replicator 2X seems more consistent than the Replicator 2s, but the Y axis still seems stronger than the X.

One of the interesting things about the two MakerBot printers is that there is a lot more variability between the X and Y axes. On the data from the Replicator 2, it even seems like the Y axis is a bit more brittle than the X axis seems a bit stronger than the X axis, while on the Replicator 2X, the X axis seems more brittle. While I cannot demonstrate this with my current data, it seems like this may be because of how the build plate is constrained. On the Replicator 2, the plate is fixed mostly along the Y axis. You get much more deflection along the X axis than the Y if you were to try and rock the build plate which could mean that parts end up printing better along the Y axis than the X on the printer. For the 2X, the build plate is better constrained and cannot be removed, so the data is a bit less variable but still pretty bad. The Stratasys printer, on the other hand, has a much more rigid build plate, so the parts it produces may be more consistent because of that.

Stress strain data from the test. Parts off the 2X seems more similar to the Stratasys in terms of consistency and behavior.

Markforged (Nylon)

So far, I’ve only been able to test parts on the Markforged printer with pure nylon and with a few layers of Kevlar. For pure nylon with 30% triangular infill, the Z axis had a huge range, and I think this is due to inconsistencies in the printer, because specimens are pretty tricky to print vertically and on a couple of occasions (on all printers apart from the Stratasys), I’ve come to pick up my print only to find the specimens fallen over when only halfway complete. You’ll see that one of the Z axis prints far outperformed the others, which I think is because the samples with lower yield strengths had some issues printing. With more data, I can verify how parts tend to behave on the Z axis. The strain at fracture for specimens printed on three axes is highly variable, probably due to the nylon tearing in slightly different ways due to material inconsistencies.

The stress strain curve for Nylon on the X and Y axes shows that the parts deform a lot more and absorb more energy before failure. The Z axis had a lot of variability due to inconsistencies in nylon adhesion.

On the X and Y axes, notice the nylon behavior. It is much more flexible and absorbs a lot more energy (check out the GIF on the X axis specimen below). The nylon slowly tears apart, and I believe the stepwise behavior on the graph of the failure is due to internal strands of nylon from the infill pattern failing section by section. This does not show as much on previous printers because none of them were nearly as elastic. The Markforged  printer slices by individual part, so each print was an exact copy in terms of its infill pattern, unlike the MakerBots. There is still some variability in the data, so I’m not quite sure what the cause of that is yet. Notice that like the MakerBots, the Y axis does seem to absorb more energy overall than the X.

Data from parts on the Markforged with pure nylon. It’s stepwise behavior is due to sections of nylon tearing apart during the test. The strain at fracture and the toughness are calculated based on the initial, large failure in each plot, not the total curve.

Markforged (Kevlar)

The Markforged prints with Kevlar acted like more traditional composites. I printed these with 30% nylon infill, and 3 layers of Kevlar each on the top and bottom faces. On each layer of Kevlar, two strands pass through the testing region (the thin region) of the part, making 12 strands of fiber passing through the region total. As you can see, the parts have a much higher yield strength than those without Kevlar, behave pretty interestingly after failure, as shown in the first graph below. The Kevlar fails first because it can’t extend as much as the nylon (although much stronger). You can see this behavior in the GIF as well; the Kevlar snaps after pretty low extension, and then the nylon stretches the rest of the way. After the Kevlar fails, the nylon has yet to fail, and thus the part begins to act like semi-extended nylon, hence the messy decrease in stress for all of the tests after the sharp drop off corresponding to the Kevlar snapping. There is not enough space for Kevlar to fit in the part when printed vertically, so this was not tested in the Z axis. Even if I had, the Kevlar strands are separated every layer, thus a part with Kevlar would behave as the nylon would along the Z axis.


The nylon absorbs a lot more energy during failure after the Kevlar has broken, and becomes much more inconsistent afterwards as well.
The same data zoomed in on the Kevlar behavior up until its failure during the test. The Kevlar is much more brittle than the Nylon composite, but can handle a lot more force before failure.

One thing you may notice is that the Kevlar adds consistency to the nylon, especially with regard to the strain at fracture and the toughness. For both tests involving the Markforged printer, I evaluated the strain at fracture and the toughness at the first “fracture” point, those being at the first large drop in the pure nylon specimen curve and after the Kevlar failure in the Kevlar stress-strain curve.

With the addition of Kevlar, the Markforged printer seems to produce parts that are more consistent and more brittle, but overall stronger.


Four key things that I’ve observed from this data include:

  • The way that a slicer slices its parts for the printer can have a significant effect on the consistency of the parts that are printed. Slicers that generate a toolpath for an entire build plate may produce different internal structures for parts of the same design, whereas “individual part slicers” make copies of each design with exactly the same toolpath.
  • On desktop printers, the Y axis (which I define as parallel to the front of the printer) is more constrained and thus may produce stronger parts along that axis, while parts are overall more consistent on industrial printers.
  • Parts in tension along their Z axis for many FDM 3D printers are much more brittle and weaker than those of the X and Y axes, as expected.
  • Embedded fibers behave like composite materials in that the more brittle material dominates first, and after it fails the part behaves more like the ductile material.

I’ll be running more tests both to reproduce and understand these findings in the future. Look for more posts soon!

Happy printing!



3 thoughts on “Printers in ConTension

  1. Fascinating, sounds to me like Makerbot cheaped out on their slicing software!
    Also, beware making specimens using the default printer settings of infill/pattern/wall thickness (rather than changing the settings on all printers so that they all match), this sounds like a huge potential source of unintended variations.


    1. You’re right, it will mean that each printer cannot be compared eye to eye. I made the choice because the default printer settings are how the printers are usually used in fabrication. The data is more relevant to parts printed in normal scenarios on each printer, so people can predict failure modes of their parts more accurately.

      Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s