In partnership with:
Improving the mechanical performance of a printed part often comes at the expense of printing speed, affordability and quality. In this study we quantify the impact of different parameters on performance, and we try to help users choose the optimal settings by clearly laying out the trade-offs faced by the user. We provide the settings we would pick depending on the application requirements.
The key parameters we look into are infill %, layer height and infill pattern. In the main body of this study, we provide a detailed description of the influence these parameters have on max stress, elongation at break, rigidity (Young Modulus) and yield stress.
To best sum up the large amount of data we gathered, we present the user with the table of preferred settings we would choose, depending on the application requirements. Does your print need strength or quality? Are you trying to minimize cost or are you trying to save time? Or is it – as is often the case – a combination of these requirements?
These conclusions are based on our interpretation of the trade-offs presented in the following tables:
The Strength, Speed and Cost tables were extrapolated from the mechanical tests we made:
- Strength corresponds to the maximum stress the specimen can take before breaking
- Speed describes the printing time of a specimen
- Cost is derived from the actual weight of the specimen, and assumes 30€/kg
- Quality relates to the general aspect of printed parts based on their layer height. Quality is not the focus of this study but the relationship between quality and layer height is generally accepted.
A much more detailed analysis of the mechanical tests is provided in the rest of the study. In particular, we show that elongation at break is lowest around 90% infill, which is not necessarily intuitive. Although elongation at break is not part of the requirements presented in the tables above, it could influence the decision on the settings the user ends up choosing.
The other key parameter we looked into was infill pattern. We show that generally, the best patterns to use are Linear or Diagonal (=Linear tilted 45°).
Indeed, decorative patterns such as Moroccan stars and Catfill show poor performance and should only be used if they are exposed and are part of the design. The real debate was between Linear, Diagonal and Hexagonal (a.k.a. Honeycomb).
- At a low infill %, we show that all three are fairly equivalent. Because Hexagonal is more demanding on the printer (more directional changes), we suggest using Linear or Diagonal.
- At a high infill %, Hexagonal is essentially the same as Linear and therefore the discussion is really between Linear and Diagonal. We show that Diagonal is ~10% stronger than Linear.
Finally we made a test on the anisotropy of a 3D printed part: this means that 3D printed parts are weaker along the Z-axis than they are along the X and Y-axis. We show that the parts are 20% to 30% weaker along the Z-axis, and that elongation at break is about half.
About the testing procedure
For each specimen we tested, we measured the following mechanical properties:
- Max stress (Ultimate Tensile Strength)
- Elongation at break
- Young modulus (Rigidity)
- Yield stress
The material used is PLA, and the 3D printing process is Fused Deposition Modeling (FDM). The three parameters we studied are:
- Infill %: percentage of the object’s volume (inside) that is filled with material
- Layer height: thickness of each layer constituting the object
- Infill pattern: pattern the nozzle is drawing to fill the object
We performed the tensile tests with a universal testing machine, at the PIMM lab of Arts et Métiers Paristech. We aggregated the results and chose to display the tables and graphs we think are most relevant.
The characteristics of PLA are well-known when the material is formed by injection molding. The point of this study is to better understand its behavior once it has been 3D-printed with the FDM process. To give a point of comparison, here are the characteristics of injection molding PLA:
|Strength [MPa]||40 – 70|
|Elongation at break||4% – 6%|
|Young modulus (rigidity) [GPa]||2 – 4|
Detailed results: Infill % test
- Considerations on strength
Unsurprisingly, the specimen strength increases with infill %, from 10MPa at 10% to 46MPa at 100%. However, it is interesting to note that the evolution is not linear: the strength gained per percentage point of infill also increases. To put it another way, reducing infill from 100% decreases resistance less and less.
Increasing infill % means a higher amount of material is used (= higher cost) and printing time is longer. This has several interesting consequences on ratios such as [strength / speed] and [strength / weight]:
The graph on the right shows that the 30% to 50% range is least efficient from both cost (material usage) and printing time standpoints, as they have the lowest ratios.
- Other performance results
Elongation at break:
The most surprising result of this study probably lies in this test. Elongation at break is remarkably constant around 2.8%, except at 90% infill where it drops to 2.0%. We checked that there had not been an error by testing another batch of two specimens for infill % 80, 90 and 100, and the results are as follows:
This second test confirms the drop around 90% infill.
Our hypothesis is the following:
- For infill below 80%, the extruded PLA filaments constituting each layer (we will just call them “filaments”) do not touch each other along the specimen axis: there are clear gaps in the mesh. So the filaments can elongate in parallel the same amount before breaking, regardless of infill%.
- For infill around 90%, the filaments touch and form a continuous 3D material, but it is porous because there are lots of small air voids in it (~10% of the specimen). In this case, the stress concentrates around the voids so the strain is localized around the void areas. The voids behave like faults that expand to eventually join and break, but lead to a lower elongation at break. This seems to be confirmed by the fact that the break is slanted for 90%, while it is straight for 70% or 100% (see picture below).
- For 100% infill, the plastic filaments also touch but there are (nearly) no more air voids in the material. Therefore the plastic deformation is not localized anymore and the whole specimen behaves as a single plastic filament would. Therefore we find the same elongation at break as the case where filaments elongate in parallel (below 80% infill).
This point deserves a more in-depth analysis to confirm our hypothesis.
Yield stress increases from 8 MPa at 10% infill to 28 MPa at 90% infill, before decreasing back to 23MPa at 100% infill. The fact that yield stress is higher at 90% than at 100% infill is in line with our hypothesis on elongation at break: the stress is localized around the air voids at 90% so at a macro level, the material yields at a higher stress.
Young modulus (rigidity):
Because the specimen is porous (except at 100%), there are two ways to calculate rigidity:
- We can count the void inside as part of the material and calculate the Young modulus by dividing by the full cross-section.
- Or we can adjust the calculation by multiplying the cross-section area by the infill %.
The non-adjusted curve shows a positive and linear relationship between infill % and rigidity. Consequently, the adjusted curve is mostly constant around 3.0GPa, right in the range of PLA’s rigidity (see section About the procedure), except as we approach a 0% infill, because then the weight and role of the shells become non-negligible, so the infill % adjustment is not accurate.
Detailed results: Layer Height test
- Considerations on strength
Layer height influences the strength of a printed part when it becomes thin. A printed part at 0.1mm shows a max stress of only 29MPa, as opposed to 35MPa for 0.2mm (21% increase).
Past 0.2mm, the max stress remains fairly constant around 36 MPa (we confirmed this conclusion with an extra test at 0.4mm, not shown here because it was not part of the same batch).
Normalizing max stress by weight smoothens the curve a bit, from 4.7MPa/g at 0.1mm to 5.6MPa/g at 0.3mm. In theory it should show the same evolution as the absolute numbers, because a constant infill should lead to constant weight, regardless of the layer height. But in practice – 3D Matter weighs all specimens – the Replicator adds less material on lower layer heights such as 0.1mm and 0.15mm.
The other result is not surprising: it takes longer to print at lower layer heights, so the max stress divided by printing time shows a curve that is increasing linearly.
- Other performance results
As shown on the stress-strain curves, the specimens behave the same way on the first part of the curve: The Young modulus remains constant around 2.9GPa, again well within the range of PLA’s rigidity. And yield stress is also fairly stable around 19MPa.
The curves differ later, for the max stress (as previously seen) and for elongation at break: it increases linearly with layer height, from 2.1% to 3.0%. This is in line with the fact that the material is weaker at lower layer heights, possibly linked to lower accuracy of a thinner deposit.
Detailed results: Infill pattern test
Linear, Diagonal and Hexagonal are all fairly comparable in terms of strength. Linear is ~10% stronger than the other two, but with a fairly wide error bar. Catfill and Moroccan stars are clearly weaker, as we would expect from these suboptimal structures.
It is important to note that the infill % chosen (10%) is very low and does not necessarily extrapolate well for higher infill %. The box on Anisotropy (see later in this section) also compares (for a different purpose) Linear with Diagonal at 100% infill and gives a slightly different conclusion: Diagonal is 10% stronger than Linear. This low infill % was chosen because we realized that 1) Catfill and Moroccan stars are not printable at a high %, and 2) Hexagonal starts looking very similar to Linear past 30% infill.
Elongation at break is between 1.8% and 2.5% but with very wide error bars, so we can consider that it is in the same range of 2%.
In conclusion, while decorative patterns such as Moroccan stars and Catfill show clearly poorer performance, Linear, Diagonal and Hexagonal are comparable at 10% infill.
Testing the mechanical performance of 3D printed PLA depending on infill %, layer height and infill pattern allowed us to define the trade-offs the user faces when choosing his settings. While the focus of the study was on mechanical performance, we made sure to include quality, cost and speed as key requirements on top of strength for 3D printer users.
Purely on mechanical performance, 3D Matter also found interesting results, such as the fact that elongation at break is lowest around 90% infill, that a lower layer height weakens the object, and that Linear, Diagonal and Hexagonal patterns show fairly equivalent performance.
We also got an interesting data point regarding the difference between printing directions, and while the Z-axis is, as expected, weaker than other directions, the decrease in max stress is only 20-30%.
- We didn’t test printing quality to come up with our recommendations. We took from experience that the quality decreases when the layer height increases. Also, because of the bumpy top layer, 100% infill prints are of lower quality.
- We opened several interesting research leads in this paper but some of the results on mechanical performance deserve further investigation to prove hypotheses we formulated – for example regarding the behavior of printed PLA around 90% infill.
- We did not investigate the influence of other key printing parameters, in particular extrusion temperature and printing speed.
- The specimens were printed without roof or floor, but with two shells. We did not investigate the influence of roof, floor or shell thickness.
- This study is valid for PLA, the conclusions might change for other materials such as ABS.
- The printing parameters we used and results we got are specific to the Makerbot Replicator. They might be slightly different on other printers.
 We measured actual values for strength (max stress), weight and printing time for all infill % at 0.2mm and for all layer heights at 80% infill, and extrapolated the rest of the tables from these values
 Infill % does not matter with regard to quality (it is inside) except at 100% infill where we have observed that the prints were not as smooth due to an excessive amount of material extruded
 Slic3r manual, Makerware, http://airwolf3d.com/wiki/slicing-1/
 Loughborough University, Universiti Malaysia Pahang, Makeitfrom.com