Science & Technology

Will AI help us discover new and innovative materials?

In an endless quest for better materials, researchers have discovered an unexpected ally. An ally who can easily explore large datasets and calculate how materials behave at different temperatures and pressures. This ally, commonly known as artificial intelligence (or AI), has the potential to usher in a new era of materials science.

Computing materials

This is a material issue. There are many things that can be obtained in bulk. new material It has exciting properties, but it takes time, money and effort. Instead, what researchers do before actually making a new material is to create the model on a computer.

Current forecasting methods work well and have become very standard, but they also require a lot of computational power. Often, these simulations require a supercomputer and can consume many resources that are inaccessible to many researchers and businesses.

“Usually, solving this problem requires a lot of physics-based simulations,” said Mark Messner, chief mechanical engineer at the Argonne National Laboratory at the US Department of Energy (DOE). I will.

Instead, Mesner and his colleagues looked for a shortcut.The shortcuts were provided in the form of AI that reveals patterns in large datasets (those that neural networks are particularly good at) and simulates what happens next. material In extreme conditions, it uses much less processing power. If it works, it’s much more efficient and faster than the existing methods … but does it work?

In a new study, Mesner and his team say so.

AI, organize this

Their new study calculated material properties 2,000 times faster than standard modeling approaches, and many of the required calculations could be performed on a typical laptop. The team used a convolutional neural network, a relatively simple class of deep neural networks most commonly applied to image analysis, to recognize the structural properties of a material.

“My idea was that the structure of the material is no different from a 3D image,” he said. “It makes sense for a 3D version of this neural network to recognize structural properties well, just as a neural network learns that an image is a cat or something,” says Mesner.

To put an approach to testing, Mesner first designed a square with bricks. This is somewhat similar to how an image is constructed from pixels. We then took a random sample of that design and used the simulation to create 2 million data points. This associated the design structure with physical properties such as density and stiffness. These 2 million data points were fed into the neural network and the network was trained to look for the desired property. Finally, he used another type of AI ( Genetic algorithm, Commonly used for optimization), find the overall structure that matches the desired property.

This simulation shows the steps that neural networks and genetic algorithms perform to find the overall structure that matches a particular material property. Credit: Image by Argonne National Laboratory.

With this approach, the AI ​​method found the correct structure in 0.00075 seconds, compared to 0.207 seconds, which was the standard physics-based model. If the same ratio can be maintained for more complex calculations, this approach may help resource-poor laboratories and companies enter the field of material manufacturing.

This is especially true in the field of renewable energy, where materials must withstand high temperatures, high pressures, corrosion and last for decades.Another promising way 3D printing material — Creating structures layer by layer gives you more flexibility than traditional methods and tells your machine exactly what to produce.

“Give the structure determined by the neural network to someone who has a 3D printer and they will print it with the properties you want,” he said. “We are not there yet, but we hope.”

Messner and the team Molten salt reactor, Uses molten salt as a coolant and can operate at much lower pressure than existing Reactor — But researchers first need to make sure that the stainless steel needed for a nuclear reactor works properly under extreme conditions for decades.

The future of mechanical engineering looks bright. Increasing computing power, 3D printing, and smarter algorithms allow engineers to fine-tune materials to create the innovative materials needed for the industry to thrive.

the study Published In Journal of Mechanical Design..

Will AI help us discover new and innovative materials? Will AI help us discover new and innovative materials?

Back to top button