Bend a ruler, and it springs back when you let go. This change of shape can also happen when a metal part is freed from its support structures after a 3D printing process called laser powder bed fusion (LPBF).
In LPBF, rapid heating and cooling of metal powder leave internal forces, known as residual stress, locked inside the printed parts. The stress can cause warping in the finished product, requiring costly corrections. Predicting how such distortion occurs, and how to counter it, remains a challenge.
“We wanted to show direct evidence linking residual stress to deformation,” said Pan Wang, a Senior Scientist at the A*STAR Singapore Institute of Manufacturing Technology (A*STAR SIMTech). Working with collaborators at Osaka University in Japan; Xi’an Jiaotong University in China; and Nanyang Technological University, Singapore; Pan Wang and Senior Scientist Chen Wang investigated the factors influencing the warpage behaviour of stainless steel beams.
The team manufactured beams at different orientations relative to the recoater, a blade that spreads fresh powder across the printing bed, and cut these to different thicknesses. When aligned parallel to recoater movement, beams thicker than about 1.3 mm warped upwards, while their thinner counterparts bent downwards. In contrast, beams rotated 90 degrees showed much less sensitivity to thickness.
According to Chen Wang, the stronger deformation of recoater-aligned beams likely arose from the physical collision of the recoater with every new powder layer. “When the beam is in the same orientation as the recoater, the repeated coating introduces additional mechanical disturbance and shear effects layer after layer, leading to higher residual stress.”
After measuring surface stress with X-ray diffraction, the researchers used contour mapping to reconstruct the internal stress based on the warpage. This revealed how stress was distributed throughout the beams. They found an uneven balance of stretching and compressive forces, with the thicker beams bearing more tensile stress (stretching) in their upper areas. Cutting away material reversed that balance. Stress mainly accumulated in the lower region of the thinner parts, leading to the downward deformation.
The findings also exposed a blind spot in a common simulation method, which predicted that the two beam orientations would deflect almost identically at 1.69 and 1.67 mm. To better capture the observed differences, the researchers trained machine learning models on 64 experimental data points, determining the most suitable model types for linking beam rigidity and residual stress to deflection and warpage profile, respectively.
The team now plans to test other print settings such as laser power and layer thickness while combining experimental measurements and machine learning to develop an integrated framework. “Ultimately, our goal is to enable more reliable distortion prediction and compensation, helping move metal additive manufacturing closer to first-time-right production,” Pan Wang said.
The A*STAR-affiliated researchers contributing to this research are from the A*STAR Singapore Institute of Manufacturing Technology (A*STAR SIMTech).
