A fast and inexpensive imaging method has been found that analyzes the structure of 3D-printed metal parts and provides information about material quality.
3D-printed metal alloys consist of numerous microscopic crystals that differ in shape, size, and atomic lattice orientation. By mapping this information, scientists and engineers can deduce properties of the alloy, such as strength and toughness.
According to scientists at the Nanyang University of Technology, this technology could benefit a number of industries related to the maintenance, repair and overhaul industry, including aerospace, where a rapid and low-cost evaluation of critical parts could be beneficial.
Until now, analysis of this microstructure in 3D-printed metal alloys has been achieved through laborious and time-consuming measurements using expensive scanning electron microscopes.
The method, created by assistant professor Matteo Seita and his team in Nanyang, is expected to provide the same quality information in minutes using a system consisting of an optical camera, a flashlight, and a laptop computer running proprietary machine learning software developed by the team.
Fast Imaging Method
The team’s new method requires first treating the metal surface with chemicals to reveal the microstructure, then placing the sample facing the camera and taking multiple optical images while the flashlight illuminates the metal in different directions.
The software then analyzes the patterns produced by the light reflecting off the surface of the different metallic crystals and extracts their orientation. The whole process takes about 15 minutes. The team’s findings were published in the npj Computer materials.
“Thanks to our inexpensive and fast imaging method, we can easily distinguish good 3D-printed metal parts from defective ones. At present, it’s impossible to tell the difference unless the microstructure of the material has been thoroughly evaluated,” he said.
No 3D-printed metal parts are equal, even if they are manufactured using the same technique and have the same geometry. Conceptually, this is similar to two identical wooden artifacts, each with a different grain structure.
Assistant Professor Seita thinks the imaging method can simplify the certification and quality assessment of metal alloy parts produced by 3D printing.
Instead of using a complex computer program to measure the orientation of the crystal from the resulting optical signals, the software developed by Professor Ast Seita and his team uses a neural network. The team then uses machine learning to program the software by feeding it hundreds of optical images.
Eventually, their software learned to predict the orientation of crystals in metal from images based on differences in the scattering of light on the metal surface. It was then tested to create a complete “crystal orientation map” that provides complete information about the shape, size, and orientation of the crystal’s atomic lattice.
