Artificial intelligence is making a lot of headlines these days. In fact, from smart classrooms to transatlantic voyages, this technology is fundamentally transforming a wide range of engineering and scientific disciplines. Today, MIT developers are exploring a new application area: materials engineering.More specifically, they are using machine learningis a technique for training AI systems to solve problems autonomously by analyzing millions of combinations.
Exploring new materials has so far required computationally intensive simulations. Each variation had to prove its toughness and behavior at the atomic scale, including the trajectory of each atom, so computing the optimal combination took hours or days. A new machine learning-driven approach can perform the same process in milliseconds.
The research team’s goal was to assess how cracks propagate throughout the material’s molecular structure. Unlike previous methods where breaking points were established by analyzing each combination, machine learning allows artificial intelligence to detect relationships between combinations that are typical patterns for the most robust and most fragile materials. .
this innovative technology project We performed atomic simulations of layered coatings made of crystalline materials and found the hardest structures almost instantly. The goal of this project was to create new coatings for the aerospace industry, such as the ceramic plates that protect space shuttles. However, applications may cover numerous areas, from body prostheses to buildings.
hyper-compressible materials
MIT has used artificial intelligence to develop an ultra-tough material, while other research teams are exploring different properties. An example of this is a study done by Miguel Bessa, assistant professor at his TU Delft in the Netherlands. The way a satellite can open a long solar sail from a very small package inspired him to develop a highly compressible yet durable material. Allows the manufacture of bicycles and umbrellas. Like her MIT colleagues, Bessa knew that traditional approaches were computationally expensive and trial and error didn’t work.
So his team decided to go down the machine learning path, reducing the need for physics experiments. Using this software, they focused on a series of brittle polymers that are highly compressible on the macroscopic scale, yet tough and resistant on the microscopic scale.
Technically, these algorithms are useful for developing new materials despite using incomplete data sets. Once enough accurate data is available, the platform can autonomously perform the calculations and find the best combination.
sauce: When, Physics
