AI predicts properties of complex metamaterials – ScienceDaily


Can you flatten a 3D origami piece without damaging it? Just looking at the design makes it difficult to predict the answer because every single fold in the design must be compatible with flattening. This is an example of a combinatorial problem. New research led by the UvA Institute of Physics and the research institute AMOLF has shown that machine learning algorithms can accurately and efficiently answer these types of questions. This should give a boost to the design of complex and functional (meta)materials supported by artificial intelligence.

In her latest work, published in Physical Verification Letters This week, the research team tested how well artificial intelligence (AI) can predict the properties of so-called combinatorial mechanical metamaterials.

artificial materials

These are engineering materials whose properties are determined by their geometric structure rather than their chemical composition. A piece of origami is also a type of metamaterial whose ability to flatten out (a well-defined physical property) is determined by how it is folded (its structure) and not by the type of paper it is made of. In general, intelligent design allows us to control exactly where or how a metamaterial that can be used for everything from shock absorbers to deploying solar panels on a satellite in space bends, buckles, or buckles.

A typical combinatorial metamaterial studied in the laboratory consists of two or more types or orientations of building blocks that deform in different ways when a mechanical force is applied. When these building blocks are randomly combined, the material as a whole will not usually buckle under pressure, since not all building blocks can deform the way they want; they will jam. Where a building block wants to bulge outwards, its neighbor should be able to squeeze inwards. In order for the metamaterial to bend easily, all the deformed building blocks must fit together like a jigsaw puzzle. Just as changing a single fold can make a piece of origami unflat, changing a single block can make a “flabby” metamaterial rigid.

Difficult to predict

While metamaterials have many potential applications, developing a new one is challenging. Starting with a given set of building blocks, deriving the full set of metamaterial properties for different structures often amounts to trial and error. Nowadays we don’t want to do all this by hand anymore. However, because the properties of combinatorial metamaterials are so sensitive to changes in individual building blocks, traditional statistical and numerical methods are slow and error-prone.

Instead, the researchers found that machine learning could be the answer: even given a relatively small set of examples to learn from, so-called convolutional neural networks are able to accurately predict the metamaterial properties of any configuration of building blocks, down to the finest details.

“This far exceeded our expectations,” says 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 if we don’t know all the rules ourselves.”

This result suggests that we can use AI to design new complex metamaterials with useful properties. More broadly, applying neural networks to combinatorial problems allows us to ask many exciting questions. Maybe they can help us to solve (combinatorial) problems in other contexts. And conversely, the results can improve our understanding of neural networks themselves, for example by showing how the complexity of a neural network relates to the complexity of the problems it can solve.

story source:

Materials provided by University of Amsterdam. Note: Content can be edited for style and length.


Comments are closed.