homehome Home chatchat Notifications


Could machine learning help us develop next-generation materials? These researchers believe so

Algorithms could help usher in a new material revolution.

Mihai Andrei
October 18, 2021 @ 6:25 pm

share Share

In the past few years, 3D printing has seen a massive growth in popularity — and it’s not just for toys and trinkets anymore. Scientists and engineers are 3D printing everything from boats to bridges to nuclear plants components. But as 3D printing becomes a more and more integral part of modern engineering, it’s also important to develop new innovative materials.

To cut down on the time and resources required to develop these new materials, researchers at MIT have used machine learning to help them find new materials with the desired characteristics (like toughness and strength).

Image in public domain.

Materials development is still very much a manual process, says Mike Foshey, a mechanical engineer and project manager in the Computational Design and Fabrication Group (CDFG) of the Computer Science and Artificial Intelligence Laboratory (CSAIL), and co-lead author of the paper.

“A chemist goes into a lab, mixes ingredients by hand, makes samples, tests them, and comes to a final formulation. But rather than having a chemist who can only do a couple of iterations over a span of days, our system can do hundreds of iterations over the same time span,” says Foshey.

In the new process, a scientist selects some ingredients, inputs their chemical compositions into the algorithm, and defines what mechanical properties he wants the new material to have. Then, instead of having the researcher do trial-and-error themselves, the algorithm increases and decreases the amounts of these properties and checks how each version would affect the material’s properties and make it most similar to what is desired.

Then, the researcher would actually create the material in the way recommended by the algorithm and test it. Streamlining this development process not only saves a lot of time and effort but also has a positive environmental impact by reducing the amount of chemical waste. In addition, the algorithm could find some combinations that may escape human intuition.

“We think, for a number of applications, this would outperform the conventional method because you can rely more heavily on the optimization algorithm to find the optimal solution. You wouldn’t need an expert chemist on hand to preselect the material formulations,” Foshey says.

The team trialed the system by asking it to optimize formulations for a 3D-printing ink that only hardens when exposed to ultraviolet light. They found six chemicals that could be used in the mix and asked the algorithm to find the best material that can be made from those six chemicals, in terms of toughness, stiffness, and strength.

This was a particularly challenging task because the properties can be contradictory — the strongest material may not be the toughest or the stiffest. However, the team was impressed to see just how many different materials the algorithm suggested — and how good the properties of these materials were. Ultimately, the material zoomed in on 12 top-performing materials that had optimal tradeoffs between the desired properties.

To encourage other researchers to use it, the researchers have also created a free, open-source materials optimization platform called AutoOED that incorporates the algorithm. AutoOED is a full software package that encourages exploration and allows researchers to optimize the process.

Researchers expect algorithm-driven material development to become more and more important over the next few years. Overall, the approach promises great improvements over the old-fashioned way of doing things.

“This has broad applications across materials science in general. For instance, if you wanted to design new types of batteries that were higher efficiency and lower cost, you could use a system like this to do it. Or if you wanted to optimize paint for a car that performed well and was environmentally friendly, this system could do that, too,” Foshey concludes.

Journal Reference: Timothy Erps, Accelerated Discovery of 3D Printing Materials Using Data-Driven Multi-Objective Optimization, Science Advances (2021). DOI: 10.1126/sciadv.abf7435www.science.org/doi/10.1126/sciadv.abf7435

share Share

This 5,500-year-old Kish tablet is the oldest written document

Beer, goats, and grains: here's what the oldest document reveals.

A Huge, Lazy Black Hole Is Redefining the Early Universe

Astronomers using the James Webb Space Telescope have discovered a massive, dormant black hole from just 800 million years after the Big Bang.

Did Columbus Bring Syphilis to Europe? Ancient DNA Suggests So

A new study pinpoints the origin of the STD to South America.

The Magnetic North Pole Has Shifted Again. Here’s Why It Matters

The magnetic North pole is now closer to Siberia than it is to Canada, and scientists aren't sure why.

For better or worse, machine learning is shaping biology research

Machine learning tools can increase the pace of biology research and open the door to new research questions, but the benefits don’t come without risks.

This Babylonian Student's 4,000-Year-Old Math Blunder Is Still Relatable Today

More than memorializing a math mistake, stone tablets show just how advanced the Babylonians were in their time.

Sixty Years Ago, We Nearly Wiped Out Bed Bugs. Then, They Started Changing

Driven to the brink of extinction, bed bugs adapted—and now pesticides are almost useless against them.

LG’s $60,000 Transparent TV Is So Luxe It’s Practically Invisible

This TV screen vanishes at the push of a button.

Couple Finds Giant Teeth in Backyard Belonging to 13,000-year-old Mastodon

A New York couple stumble upon an ancient mastodon fossil beneath their lawn.

Worms and Dogs Thrive in Chernobyl’s Radioactive Zone — and Scientists are Intrigued

In the Chernobyl Exclusion Zone, worms show no genetic damage despite living in highly radioactive soil, and free-ranging dogs persist despite contamination.