Sandia National Labs announced it has successfully deployed machine learning to aid in developing novel alloys and materials, speeding up a process that can take weeks or months into moments.
Running simulations, Sandia researchers have sped up the computer simulations that predict how the modification of materials will affect a given product’s overall design and structure. In other words, making slight modifications to materials in their development process requires significant computer testing time. Now, those modifications can be made via a machine learning simulation, and the algorithms can return the data almost immediately.
BACKGROUND: What is Sandia?
Sandia National Labs “has delivered essential science and technology to resolve the nation’s most challenging security issues,” according to its website.
A seven-decade old laboratory, Sandia works on numerous projects for various private and government agencies. The U.S. Department of Energy funded this project. Testing took place at the Center for Integrated Nanotechnologies, a jointly operated facility headed up by Sandia and Los Alamos national labs.
The team at Sandia published their results on January 4th in Nature’s partner journal, Computational Materials. The overall project was to accelerate material science calculations, which are cumbersome, into a faster and effective model using machine learning.
ANALYSIS: WHy is Sandia On To Something?
“Materials are essential elements in advanced technologies such as aerospace, optics, or microelectronics for example,” Rémi Dingreville, one of the authors of the Sandia study, told The Debrief in a written statement. “Here we provide a machine-learning materials acceleration platform to not only accelerate the discovery of new materials, but also to speed up our understanding and predictions of the relationship between manufacturing and materials performance.”
Depending on the application requirements, different products, from smartphones to hypersonic aircraft, require various materials, alloys, and metamaterials to achieve top performance and efficiency.
As those products are tested, modifications may need to be made to the materials themselves to boost any number of factors, such as flexibility, performance, or strength.
“Conventionally, materials scientists use computer simulations to evaluate hundreds or thousands of iterations of potential candidate materials. Although simulations are clearly quicker and cheaper than physically creating a couple hundreds of prototypes, it still can take months—sometimes more than a year—to complete all these calculations even on a high-performance computer. Our research shows that machine learning can cut this research phase down to a matter of minutes to efficiently and rapidly explore the manufacturing-materials-performance relationships,” Sandia’s David Montes de Oca Zapiain told The Debrief.
OUTLOOK: More Technology More Quickly
“We see this machine-learning materials discovery platform as a key tool that can not only drive down the design cycle of new technologies—something that’s often measured in years—but also a new tool that enables us to navigate the near-infinite range of possible designs for targeted applications,” Dingreville explained. “There are exciting opportunities ahead of us to start integrating this platform with manufacturing processes to co-design materials, processes, and products for the next generation of manufacturing.”
Moving into the future, Sandia’s achievement is less about new products and materials but more about the pace at which it can design and produce them. Often, the biggest hurdle to progress is time itself, and Sandia seems to have reduced the lag time between an idea and making that idea a reality.