A scanning electron microscope image of a single laser scan cross-section of a tested nickel and zinc alloy structure. Image via Texas A&M University.
Texas A&M’s ML-led method
Having identified solidification speed as key to avoiding defects, and realising that increasing scanning velocity may not be feasible in some cases, the team opted to map how alloys behave to develop a means of ‘controlling’ the process instead.
To gather the required data, the researchers investigated the behavior of four nickel-based alloys during printing, with each including different concentrations of zinc, zirconium and aluminum. A total of 46 single-track laser tests were then carried out, in which the metals’ physical states were monitored at different temperatures, yielding data that could be converted into detailed phase diagrams.
Once the team had determined how chemical composition could be optimized to achieve minimal microsegregation, they carried out further experiments to find out how laser settings affect part porosity too. Interestingly, alloys with higher melting temperatures were found to be more susceptible to keyhole defects, as they exhibited shallower melt pool structures, which resulted in a lack of fusion.
Later, with the aim of finding additional trends in the data they’d collected, the researchers fed it into a machine learning algorithm, which was trained to calculate a given part’s potential error rate and level of accuracy. Results showed that as predicted, temperature, laser power, partition coefficient, and freezing range all had an impact on the resulting model, but scan speed was the most important input.
Backed by the United States Army Research Office and the National Science Foundation, the team say that their research has yielded a simplified method of 3D printing crack-free parts with any alloy, that could now be adopted within a wide variety of industries.
“Our methodology eases the successful use of alloys of different compositions for AM without the concern of introducing defects, even at the microscale,” added Ibrahim Karaman, Head of Texas A&M’s Materials Science and Engineering Department. “This work will be of great benefit to the aerospace, automotive and defense industries, that are constantly looking for better ways to build custom metal parts.”