Skip to main content
3D Printing

Build It Up: 3D Printing To Get Boost From World’s Fastest Computer

Vince Beiser
January 22, 2020
Tennessee’s Oak Ridge National Laboratory made history as part of the Manhattan Project, demonstrating the crucial ability to produce plutonium from uranium in a nuclear reactor. Today ORNL is at the forefront of another potentially world-changing technology involving tiny bits of matter: additive manufacturing.
Better known as 3D printing, additive manufacturing is the computer-controlled process of building objects layer by layer. Most systems use plastic or other polymers, but ORNL is pioneering the use of metal powders, which can be used to create objects that need to withstand tremendous heat and pressure, like jet engine parts. Printing parts, rather than assembling them from pieces of conventionally machined metal, can save time, money and weight — all prized commodities in the world of manufacturing.

Through partnerships with private companies, ORNL — which is sponsored by the U.S. Department of Energy — is exploring which alloys of steel, aluminum and other metals are most effective for particular applications. The potential for additive manufacturing with metal powders is tremendous, says Brian Thompson, GE Additive’s manager of design and development.

In October, GE Additive signed a five-year cooperative research and development agreement with ORNL to combine the lab’s research capabilities with GE’s experience in developing real-world products.

The agreement is an extension of ORNL’s work with Arcam, a company GE acquired in 2016. Arcam manufactures refrigerator-sized 3D printers that use electron beams to weld together millions of grains of fine-powdered metal, one hair-thin layer at a time, to build things like intricately detailed jet engine parts.

GE’s new Catalyst turboprop engine, for instance, includes a dozen 3D-printed parts that previously would have to have been built out of some 800 pieces, but that can now be printed as a single unit. A fuel nozzle for GE’s LEAP jet engine is now printed as one component instead of 20 pieces welded together. Distilling all those pieces down to one saves on weight, fuel consumption and cost.

 width= Above: Tucked into the hills of eastern Tennessee, the storied Oak Ridge National Laboratory is exploring the tremendous potential of metal powders in 3D printing, aka additive manufacturing. Image credit: Oak Ridge National Laboratory. Top: Scientists at ORNL can crunch huge amounts of manufacturing data with the help of Summit, the most powerful computer in the world. Image credit: Carlos Jones/ORNL.

But this kind of manufacturing is tremendously complex, with a huge risk of error. Slight changes in heat, the chemistry of the powders, even the age of the equipment can make a significant difference in how the final product turns out. And when you’re dealing with jet engine parts or something as small as a new hip joint, even a fraction-of-a-millimeter mistake can ruin an entire build. “We’re still trying to understand how all the variations in materials and the machines can affect part quality,” says Christine Furstoss, vice president of advanced manufacturing at GE Additive.

For additive manufacturing to flourish, engineers need to be sure their parts will match exacting specifications every time. This challenge will be at the heart of ORNL’s ongoing work with GE Additive. Metal powder 3D printers already in use at GE facilities such as Avio Aero in Italy are equipped with sensors that can collect data from each build. That data can add up to a full terabyte.

Analyzing that much information on traditional computers takes a very long time — but ORNL is also home to Summit, a scientific supercomputer that is ranked as the most powerful on Earth, capable of performing 200 quadrillion calculations per second. (It’s slated to be joined in 2021 by the even mightier Frontier, capable of one quintillion calculations per second.)

“That kind of computing power unlocks some exciting potential. We can analyze the data much, much faster,” says Thompson. That data will be plugged into sophisticated computer models to help researchers understand how the many variables interact, and fine-tune the process. Those findings in turn can be integrated into AI-driven software that can make adjustments to the equipment during a build, in real time.

The goal is to make the 3D-printing process more efficient and dependable — which should in turn motivate more companies to start adopting the process. “The holy grail is being able to just push a button and have the product come out,” says Thompson. That’s still some ways off, though. “This industry is still in its infancy,” says Furstoss. “We don’t have 50 or 60 years of processing knowledge and people studying it, like traditional manufacturing. That’s why relationships like this are so important.”