Jump to content

Size of the 3D print vs Actual size


Recommended Posts

  • 4 weeks later...

Every plastic shrinks a bit when it's cooled down. This percentage is different for the different materials, but it's a good idea to scale up your model in the slicer with 0,05%. Also, some composites have minimal shrinkage and they should be on choice for prints, which requires high dimensional accuracy. I'm using Silk PLA (85% PLA, 15% Polyester, some other additives) with great success, because the shrinkage is lower than the natural PLA. On the graphic I measured a model of Aberrant arteria subclavia dextra - the CT scan with a dicom viewer, the model with Autodesk Fusion 360 and the printed model with vernier caliper. I calculated average deviation on 95% confidence interval and this is what I got. Note that the caliper was too tick and large for subclavia dextra, which resulted in the difference of the result (my tool is too large). I think those results are quite promising and I'm planing to close the circle - to measure cadaver, CT of the cadaver, generated 3D model and 3D printed model, when the ethical commission allow me to.

 

Picture1.png

Link to comment
Share on other sites

The shrinking is a major and complicated issue for all the 3D printing methods in general. I didn't cared about it, until I had to made really accurate, industrial-grade model, in which every micron matters. I'm not a material science engineer, but I have to address this issue even further, because my research group is planning to go in the direction of the temporary and permanent implants, in which even the slightest deviation can cause slow recovery, pain, inflammatory reaction or rejection. This is especially true for the 3d printed dentures and maxillofacial implants, in which even 50 micron deviation can cause significant problems for the patient. It's so good for me, that I'm not working for the dental medicine faculty, because for the bone implants deviation of 200 microns is perfectly acceptable. For the diagnostic models (presurgical, demonstration, teaching etc.) the acceptable deviation is even bigger - up to 500 microns.
Every 3d printing method have some base shrinkage and warping, which is combined with the base deviation of the STL model itself and with the base shrinkage of the material. To make the issue even harder, the manufacturers didn't post the basic deviation of their materials, because this can ruin their trade secret. Every material in the market is not just a single polymer, but a composite of several polymers, with a lot of additives, which makes the issue even more complicated.

There are several general rules:
1. Material Jetting > SLS > STL/DLP/FDM. From the termoplastic modalities, the material jetting is most accurate method, while STL/DLP/FDM are relatively equal in their accuracy. STL and DLP have better porstprocessing option than FDM, though. Overall, for the lower industrial segment, FDM is better in the printing of large objects (more than 10 cm), while STL and DLP are better in the smaller part printing.
2. Stiffer materials (high Young's modulus) are more accurate than the elastic materials. PLA (high stiffness) is more accurate than PETG or Nylon. This also depends on the 3d printing method.
3. Material with lower extrusion temperature are more accurate than materials with high extrusion temperature. PLA shrinks with 2-5% (depending on the manufacturer), while ABS ~8%. You can control this process even further by tuning the cooling fan. 
4. Larger objects shrinks more than the smaller ones. It's better to cut your model into several attachable parts, printed individually, than to print it as a whole.
5. Enclosed system is more accurate than unenclosed one. The constant temperature in the enclosure will make the shrinkage more uniform in all the segments of the object, resulting on overall more accurate final product.

The best practice is to start measuring your models.
1. When you're done with the model, reimport the stl again in the medical data program, used for segmentation, convert it into a label map and compare the geometry with the voxels of the source dataset. If you're using control points on the antropometric points (separate segmentation on the classical points of orientation in the human body, several voxels in size), every deviation will be visible and editable. If you're not, you'll have to compare it visually.
2. Use a professional CAD program (Fusion360, Solidworks) to measure the parts of the STL model. The digital simulations of those programs makes the quality control even better.
3. Use dicom viewer to measure the same distances in the DICOM dataset.
4. Use a caliper to measure the same distances on the 3d printed object.
5. 6-8 measurements per distance will be more than enough. 4-6 distances will be enough for a model.
6. Calculate average deviation at 90% confidence interval.
7. The resulting score will give you the accuracy of your model.
8. If you're not specifically targeting high accuracy or you're not a detail freak as myself, point 1 will be enough.

You can check this guide, in which the issue is well explained. You can also check Form Lab's guide for dimensional accuracy, since you're regularly using Form2. For now, I can achieve without a problem 0,5 mm accuracy from CT scan to the 3D printed object (in some cases 0,2 mm), but I'm constantly fighting for every micron and I won't quit, until I achieve 0,2 mm on all of my models. I guess on some point I'll have to add a material science engineer to my research group. I'm also considering machine learning algorithm for quality control, but this is way over my head right now.

To check how accurate are your models, you can use 3d printed tests. You can try the tolerance test and the all-in-one test. They work only for FDM, though.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...