OptiMatter provides data on a 3D printed materials, with the aim of including all the materials and all the printing technologies available on the market. We are constantly testing and adding new materials to our database using the methodology we are presenting here.
OptiMatter is a multivariate regression model that forecasts the performance, visual quality and processability of 3D printed parts depending on the technology, material and printing parameters. The regression is based on thousands of mechanical and visual quality tests we conduct for each material present in the software.
Technology: the highest level of material classification is by technology, i.e. the 3D printing forming process. The technology defines what format the material will come in: filament, powder, liquid… etc. Examples of technologies are FDM, SLA or SLS
Material type: For each technology, materials are sorted by material type. The material type is the main chemical component of the material. It determines how the material behave and how it will react to the printing process. Examples of material types are PLA, ABS, Nylon
For each material type, we investigate and model the influence of the printing parameters on the material properties, as they affect material types in different ways.
Product / brand: The specific product/supplier used within a material type. For example, Makerbot’s PLA will not have exactly the same properties as Esun’s PLA. For products with additives or that are blended with other polymers, we use the majority base polymer to categorize them. For example, Colorfabb’s PLA/PHA will be found in the PLA material type.
The other inputs to the model are the printing parameters. So far, we have included:
Infill%: Percentage of the object that is filled with material. Example: 70% infill
Layer Height: thickness of each layer constituting the object. Example: 0.2mm
Infill pattern: pattern the printhead is drawing to fill the object. Example: honeycomb, diagonal (45°/-45°), linear (0°/90°)
Section area: The section area represents how big the typical section of the object is. If the object is big, it will behave more like its infill, whereas if it is small, it will behave more like the outer surface (= shells, top solid layers and bottom solid layers). This input is modeled using the “% of outer surface” for a given cross section.
Orientation: By nature of the 3D printing process (layer by layer), 3D-printed objects are anisotropic, meaning that their properties depend on the direction considered. OptiMatter provides data for both X/Y axis and Z axis.
Note: as of August 2016, only infill %, layer height and orientation are modeled by material type, the other two are still modeled as having the same influence for all material types.
We aim to include other influencing factors as inputs to the model, such as the slicer used.
The following table shows the mechanical tests we conduct, the ASTM or ISO standard we base our procedure on, the number of specimens we print, and the properties that we measure via these tests:
- To conduct tensile testing on materials at low infill, we have designed a specimen with a larger cross section. Indeed, the cross section of ASTM D638 / ISO 527 specimens is too small: there are not enough cells at low infill to be statistically significant.
- All specimens are measured and weighed to make sure they have been printed correctly.
- We print all our FDM specimens without outer surface (i.e. 0 shells, 0 top solid layers and 0 bottom solid layers). The outer surface is modeled separately.
- FDM specimens are printed on rafts to avoid the smoothening effect of printing directly on the bed.
- SLA specimens are post-cured for two hours.
Testing a new product
We conduct Tensile testing on every product in our database. We conduct Charpy testing on rigid materials, and we conduct Tensile Set testing and Hardness testing on flexible materials.
We test the product at 100% infill, at a default layer height of 0.2mm and usually using a linear pattern (0°/90°).
If this new product is part of a material type we have already modeled, then we can include it in OptiMatter directly. If not, we need to model this new material type.
Introducing a new material type
There are too many combinations of printing parameters to test all of them, so we are creating a regression model for each material type.
We model each printing parameter by making it vary while keeping all other printing parameters constant. This way, we get a first curve for the influence of that parameter under the baseline conditions. We then extrapolate this influence along the other parameters. We use a design of experiment methodology to refine the model, in particular to quantify the cross-terms of the regression, as the parameters are not independent.
For quality testing, we are using Testman, a file 3D Matter specifically designed to test various aspects of what defines Visual Quality. To learn more about the file, visit this page
To grade a given product, we print several Testmans with that product, tweaking the printing parameters (such as temperature, cooling… etc.) to find the best quality. Then we elect the best Testman of the selection and compare it to other products. We usually use one standard product we already know the grades for in the comparison process in order to adjust the grades to the right level.
The comparison is done regarding two main categories: Details & texture and Geometrical accuracy.
We also model the influence of layer height on visual quality: a small layer height usually increases resolution, and therefore improves the quality grades.
We assess the ease of printing a material to give the user a sense of the difficulty they may face in the printing process. This assessment is mostly based on an overall feeling after having used a given product for a while, but here are some example criteria we consider:
- Ease to feed into the printer
- Convenience of spool supplied (FDM)
- Frequency of filament getting tangled (FDM)
- Adhesion issues / degree of warping
- Support removal
- Post-processing requirements
Besides the parameters we are modeling in the Input section, we need to give consideration to the other printing parameters. In general, we strive to keep them constant to ensure comparability of results. But we also have them vary to check that our data is still relevant under other parameters. For example, we conduct the tests on several printers, and then average out the results. Here is a summary of how we handle other key printing parameters:
Based on our data set, we estimate our model to have an accuracy within 10% to 40% depending on the material and the output considered. For example, we believe that our data is replicable within 15% for PLA’s max stress at a high infill, but that it is only replicable within 30% for PET’s max stress at low infill. Please send an email to email@example.com if you would like more detailed info on the replicability of our data.
The main source of error at this point is linked to the variability in printers, and more specifically the extruder type and the slicer used. 3D Matter is planning to model these two variables as well in the future to improve data accuracy.