AI predicts properties of complex metamaterials — ScienceDaily

Given a 3D piece of origami, can you flatten it without damage? By looking only at the design, it is difficult to answer, because every stroke of the design is compatible with the flat. This is an example of a combinatorial problem. Recent research led by the UvA Institute of Physics and the AMOLF research center has shown that machine learning algorithms can answer these questions accurately and efficiently. This is expected to give a boost to the design that is aided by the intelligence of complex objects and (meta) work.

In their new work, published on Physical examination letters this week, the research team tested how well artificial intelligence (AI) can predict the properties of so-called combinatorial mechanical metamaterials.

Work tool

These are engineering materials whose properties are determined by their geometrical properties rather than their chemical properties. A piece of origami is a kind of metamaterial, whose ability to flatten (a well-defined property) is determined by the way it is folded (its shape), rather than the shape of the paper it is made of. as far as. Basically, the smart design that allows us to control exactly where and how the metamaterial bends, bends or bends, can be used for all kinds of things, from explosives to removing solar panels on a satellite in space.

Also Read :  Conversational Artificial Intelligence (AI) Market (Discount up to 70%) to See Incredible Growth, Business Insights, Developments, Future Assessment and COVID-19 Analysis by 2029|, IBM, AffectLayer

A typical combinatorial metamaterial studied in the lab is composed of two or more shapes or arrangements of building blocks, which change into different shapes when a mechanical force is applied. If these building blocks are put together, the whole is not always under pressure because not all the blocks can change the way they want; they will hold. Where one building block wants to go out, its neighbor can go in. For the metamaterial to be easily assembled, all the deformed building blocks must fit together like a jigsaw puzzle. Just as changing a single print can create an origami piece that cannot be folded, changing a single block can create a stable ‘floppy’ metamaterial.

Also Read :  Home Internet Cheat Sheet: Your Guide to Broadband Basics and Beyond

It’s hard to predict

Although metamaterials have many potential applications, designing a new material is difficult. Starting with a set of building blocks, sorting out the overall metamaterial properties for different buildings often starts with trial and error. Nowadays, we don’t want to do these things manually. However, because the properties of combinatorial metamaterials are easy to modify in the building blocks, conventional computational and computational methods are slow and prone to errors.

But researchers have found that machine learning is the answer: even if given only a small set of examples to learn, so-called convolutional neural networks can accurately predict the metamaterial properties of each set of building blocks to the finest detail.

Also Read :  Essential Signaling Pathway for Neuronal Connectivity During Brain Development Identified

“This exceeded our expectations,” said PhD student and first author Ryan van Mastrigt. “The accuracy of the predictions tells us that the neural networks have indeed learned the mathematical rules underlying the metamaterial properties, even though we don’t know all the rules ourselves.”

This discovery shows that we can use AI to design new complex metamaterials with useful properties. More broadly, the use of neural networks in combinatorial problems allows us to pose interesting questions. They may help us solve combinatorial problems in other ways. On the other hand, the findings can improve our understanding of neural networks themselves, such as showing how the complexity of a neural network is related to the complexity of possible problems. it is up to him to fix it.


Items provided by University of Amsterdam. Note: Content can be edited for content and length.


Leave a Reply

Your email address will not be published.

Related Articles

Back to top button