02 Jul 2019
Technique, using diodes borrowed from LLNL's National Ignition Facility cuts residual stress by as much as 90 percent.
Researchers at Lawrence Livermore National Laboratory (LLNL) and the University of California, Davis are addressing this problem by using laser diodes, high-powered lasers borrowed from technology created for LLNL's National Ignition Facility, to rapidly heat the printed layers during a build.
The new technique, described in a paper published in Additive Manufacturing, results in the reduction of effective residual stress in metal 3D-printed test parts by 90 percent, by enabling researchers to reduce temperature gradient and control cooling rates.
“In metals it’s really hard to overcome these stresses,” commented lead author John Roehling. “There has been a lot of work on changing the scanning strategy to redistribute the residual stresses, but our approach is to get rid of them while we are making the part, to remove those problems. By this approach, we can effectively get rid of residual stresses to the point that there are no part failures during the build anymore.”
Laser powder bed fusion
LLNL engineer and co-lead author Will Smith built small, bridge-like structures from 316L stainless steel using the laser powder bed fusion process. He let each layer solidify before illuminating the surfaces with the diodes, initially at full power and immediately ramping down the intensity over a period of 20 seconds. The result was akin to putting the part in a furnace after each layer, as surface temperatures reached about 1000°C (1832°F).
The finished parts, with their thick legs and thin overhang section, allowed researchers to measure how much residual stress was relieved by cutting one of the legs off and analyzing how much the weaker overhang section moved. When the diodes were used, the bridge did not deflect anymore, the researchers said.
Smith said, “Building the parts was similar to how a normal metal 3D printer works, but the novel part of our machine is we use a secondary laser that projects over a larger area and that post-heats the part afterwards — it raises the temperature up rapidly and slowly cools it down in a controlled fashion.
“When we used the diodes, we saw that there was a trend in the reduction of residual stress, and it compared to what is done traditionally by annealing a part in an oven afterwards. This was a good result, and it was promising as to how effective our technique was.”
The approach is an offshoot of a previous project in which laser diodes, developed to smooth out lasers in NIF, were used to 3D-print entire metal layers in one shot. It improves on other common methods for reducing residual stress in metal parts, such as altering the scanning strategy or using a heated build plate, Roehling said. Because the approach heats from the top, there’s no limit on how tall the parts can be.
Researchers plan to next perform a more in-depth study, turning their attention to increasing the number of layers per heating cycle to see if they can reduce residual stress to the same degree, attempt more complex parts and use more quantitative techniques to gain a more in-depth understanding of the process.
'Room for improvement'
“This technology is something that could be scaled up, because right now we’re projecting over a relatively small area and there’s still a lot of room for improvement,” Smith said. “By adding more of the diode lasers, we could add more heating area if someone wanted to integrate this into a system with a larger printing area.”
More importantly, Roehling said, researchers will explore controlling phase transformations in titanium alloy (Ti64). Typically, when building with Ti64, phase transformation causes the metal to become extremely brittle, causing parts to crack. If researchers could avoid the transformation by cooling the part slowly, it could make the material ductile enough to meet aerospace standards, Roehling said, adding that preliminary results are promising.
The project was funded through the Laboratory Directed Research and Development program. Co-authors included LLNL researchers Gabe Guss, Tien Roehling, Bey Vrancken, Joe McKeown and Ibo Matthews, as well as UC Davis professor Michael Hill.