An extraordinary research at TU Delft. University researchers have managed to create new material using only artificial intelligence.
The new material was developed without doing any experimental research. This super-compressible but strong material was designed using only artificial intelligence. "Artificial intelligence gives you a treasure map, and as a scientist you have to find the treasure," says Miguel Bessa, first author of the publication on the subject in Advanced Materials.
Folding bike
Miguel Bessa, associate professor in Materials Science and Engineering at TU Delft, got inspiration for this research project during his time at the California Institute of Technology. In the Space Structures Lab, he saw a satellite structure that could open long solar sails from a very small package. He wondered if it would be possible to design a highly compressible but strong material that could be compressed to a small fraction of its volume. "If that were possible, you could fold everyday objects like bicycles, dining tables and umbrellas and carry them in your pocket," says Bessa.
Reversing the design process
The next generation of materials must be adaptive, versatile and customisable. This is possible with so-called metamaterials, in which structure is central, and in which totally new properties and functionalities can be realised through new geometries. "However, design of metamaterials has so far been based on extensive experimentation and trial-and-error," continues Bessa. "We advocate reversing the process, using machine learning to explore new design possibilities, and keeping experiments to an absolute minimum."
Machine learning
"We use data-driven computer calculations to research a new concept for a metamaterial, and to adapt the concept to the desired properties, choice of base materials, length scales and production processes." Using machine learning, Bessa fabricated two designs at different length scales, converting brittle polymers into lightweight, recyclable and super-compressible metamaterials. In the macro-scale design, he aimed for maximum compressibility, while in the micro-scale one, he aimed for high strength and stiffness.
Undiscovered areas
Yet Bessa argues that the most important aspect of the work is not that they managed to make this particular material, but that they managed to enter unexplored areas of the design space via machine learning. "Most importantly, machine learning gives us the ability to reverse the design process: by moving from experimental research to data-driven computer calculations, even when information is missing from the computer models. The essential requirements are that there is 'sufficient' data on the problem at hand and that the data is accurate enough." Bessa is a strong advocate of data-driven research in mechanics and materials science. "Data-driven science will revolutionise the way we arrive at new discoveries, and I can't wait to see what the future will bring us."
Source: TU Delft
