AI Researchers from the Netherlands Propose a Machine Learning-based Method to Design New Complex Metamaterials with Useful Properties – MarkTechPost

Combinatorial problems often arise in puzzles, origami, and metamaterial design. Such problems have rare collections of solutions that generate intricate and distinct boundaries in configuration space. Using standard statistical and numerical techniques, capturing these boundaries is often quite challenging. Is it possible to flatten a 3D origami piece without causing damage? This question is one such combinatorial issue. As each fold needs to be consistent with flattening, such results are difficult to predict simply by glancing at the design. To answer such questions, the UvA Institute of Physics and the research center AMOLF have shown that researchers may more effectively and precisely respond to such queries by using machine learning techniques.

Despite employing severely undersampled training sets, Convolutional Neural Networks (CNNs) can learn to distinguish these boundaries for metamaterials in minute detail. This raises the possibility of complex material design by indicating that the network infers the underlying combinatorial rules from the sparse training set. The research team thinks this will facilitate the development of sophisticated, functional metamaterials with artificial intelligence. The team’s recent study examined the accuracy of forecasting the characteristics of these combinatorial mechanical metamaterials using artificial intelligence. Their work has also been published in the Physical Review Letters publication.

The attributes of artificial materials, which are engineered materials, are governed by their geometrical structure rather than their chemical makeup. Origami is one such metamaterial. The capacity of an origami piece to flatten is governed by how it is folded, i.e., its structure, and not by the sort of paper it is made of. More generally, the clever design enables us to accurately regulate a metamaterial’s bending, buckling, or bulging. This can be used for many different things, from satellite solar panels that unfurl to shock absorbers.

A combinatorial metamaterial typically consists of two or more different orientations of building pieces. These building blocks deform differently in response to an external mechanical force. The material will not typically give way under pressure if these building pieces are mixed randomly because not all of them will be able to deform how they wish to. A neighboring building block should be able to squeeze inward where one wants to protrude outward. All distorted construction components must fit together like a jigsaw puzzle for the metamaterial to buckle quickly. A ‘floppy’ metamaterial can become rigid by modifying a single block, just as altering a single fold can render an origami piece not flatten.

Although metamaterials have a wide range of potential uses, creating a new one is difficult due to their unpredictable behavior. It typically comes down to trial and error to determine the general metamaterial properties of different structures starting from a specific set of building blocks. Recent technological developments make it unnecessary for researchers to perform all this work by hand. However, typical statistical and numerical methods are slow and prone to errors because the properties of combinatorial metamaterials are so sensitive to changes to individual building blocks. This is where machine learning comes into play. CNNs can precisely predict the metamaterial properties of any configuration of building blocks down to the smallest detail, even with only a minimal set of examples to learn from.

The CNN results were astounding and far beyond expectations. The accuracy of the predictions showed that the neural networks had truly mastered the fundamental mathematical principles governing the behavior of metamaterials, which still need to be better understood by researchers themselves. These results indicate that complicated metamaterials with relevant properties can be created using AI. More generally, neural network applications can assist researchers in solving combinatorial problems in various contexts and raise various intriguing concerns. The results can also help understand neural networks by showing how a neural network’s complexity correlates with the complexity of the issues it can handle.

This Article is written as a research summary article by Marktechpost Staff based on the research paper 'Machine Learning of Implicit Combinatorial Rules in Mechanical Metamaterials'. All Credit For This Research Goes To Researchers on This Project. Check out the paper and reference article.

Please Don't Forget To Join Our ML Subreddit

Khushboo Gupta is a consulting intern at MarktechPost. She is currently pursuing her B.Tech from the Indian Institute of Technology(IIT), Goa. She is passionate about the fields of Machine Learning, Natural Language Processing and Web Development. She enjoys learning more about the technical field by participating in several challenges.