In tests of four small titanium-alloy bridges, a laser-printing pattern meant to reduce stress produced higher stresses than the alternative. Researchers from the National Institute of Standards and Technology (NIST), Lawrence Livermore and Los Alamos national laboratories and other institutions found higher residual stresses in bridges built with island scanning than in those made with continuous scans. NIST announced the findings on May 19, 2021. For manufacturers trying to prevent printed parts from warping or cracking, the result challenges the idea that choosing a familiar scan pattern is enough. [S1][S2]
Laser powder bed fusion builds a metal part by melting successive layers of powder. Repeated heating and cooling can leave stresses locked into the finished material. In this experiment, the team printed four Ti-6Al-4V bridges measuring 21 millimeters by 5 millimeters by 9 millimeters. The laser either swept continuously or worked through square islands 2.5 millimeters across, with two different starting orientations. Laser power stayed at 100 watts and scanning speed at 600 millimeters per second. The bridges were built without support structures or contour scans, so the result belongs to a specified build rather than to metal printing in general. [S1][S2]
The researchers used high-energy X-ray diffraction at the Cornell High Energy Synchrotron Source to examine what was happening inside the metal. Diffraction measurements revealed atomic spacing; comparison with reference samples allowed the team to calculate strain, then infer stress using the material’s elastic properties. They mapped three directional components across a central plane of each bridge. This was a laboratory investigation of internal stress, not a demonstration of an inspection sensor operating inside a production printer. [S2]
A separate check showed how the stored stress affected shape. The team cut one leg free from the build plate on two specimens: the continuous-scan bridge aligned with the sample axes, called C0, and its island-scan counterpart, I0. The island-scanned bridge distorted by 0.15 millimeters over its 21-millimeter length, against 0.06 millimeters for the continuous-scan bridge. That comparison concerns those two specimens after partial release from the plate; it is not a rejection rate or a tolerance limit for customer parts. [S2]
Individual crystal grains can retain stresses different from those acting across a whole metal part. The authors warn that treating selected grain-scale measurements as representative can even produce a bulk-stress estimate with the wrong sign. For this titanium alloy, they favor analysis across many diffraction peaks, accounting for the distribution of grain orientations, rather than relying on a single peak. [S2]
For teams checking a computer model against the experiment, the paper gives a less obvious instruction: compare the measured strain components directly. Calculating stress combines strain measurements that have different uncertainties and spatial resolutions. That calculation reduces the accuracy and resolution of some components relative to their underlying strain measurements. The authors therefore prefer strain data for comparison with simulations, even though stress is the quantity of interest for distortion and fracture. [S2]
The experiment does not establish a universally better print recipe. Apart from the scan pattern, the build settings were held fixed, not optimized separately for each bridge. NIST also cautioned that island scanning might work under other conditions. For a manufacturing team, the practical implication is to request evidence for the proposed material, geometry and settings, together with the reference measurements and uncertainty treatment behind any stress claim. [S1][S2]
Source documents
- New Research Could Help Manufacturers Avoid 3D-Printing Pitfallhttps://www.nist.gov/news-events/news/2021/05/new-research-could-help-manufacturers-avoid-3d-printing-pitfall
- Effect of the scanning strategy on the formation of residual stresses in additively manufactured Ti-6Al-4V — accepted manuscripthttps://www.osti.gov/servlets/purl/1785480