Materials made from mechanical neural networks can learn to adapt to their physical properties

Materials made from mechanical neural networks can learn to adapt to their physical properties

A new type of material can learn and improve its ability to handle unexpected forces thanks to a unique lattice structure with joints of variable stiffness, such as It is described in a new paper of my colleagues and I.

A new material is a type of engineered material, which gets its properties primarily from the specific geometry and features of its design rather than what it is made of. Take hook-and-loop fabric closures like Velcro, for example. It does not matter if it is made of cotton, plastic or any other material. As long as one side is a fabric with solid hooks and the other side has fluffy loops, the material will have the adhesive properties of Velcro.

My colleagues and I based our new material architecture on an artificial neural network – layers of interconnected nodes that can Learn to do tasks By changing the amount of importance or weight they put on each connection. We hypothesized that a mechanical network with physical nodes can be trained to acquire certain mechanical properties by tuning the stiffness of each connection.

Engineered materials – like this 3D mesh – get their properties not from what they’re made of, but from their structure. Image credit: Ryan Lee, CC BY-ND

To see if a mechanical network could adopt and maintain new properties — such as taking a new shape or changing directional strength — we started by building a computer model. Then we chose the desired shape of the material as well as the input forces and had a computer algorithm to adjust the communication tensions so that the input forces produced the desired shape. We ran this training on 200 different mesh structures and found that the triangular mesh was the best at achieving all the shapes we tested.

Once you have set several connections to achieve a set of tasks, the material will continue to interact in the desired way. Training is remembered – in a sense – in the structure of the material itself.

Then we built a physical model grid with adjustable electromechanical springs arranged in a triangular grid. The prototype consisted of 6-inch joints and was about 2 feet long and 1½ feet wide. It has succeeded. When the network and algorithm worked together, the material was able to learn and change shape in certain ways when exposed to different forces. We call this new material a mechanical neural network.

Picture of hydraulic springs arranged in a triangular lattice
The prototype is 2D, but a 3D version of this material can have many uses. Image Credit: Jonathan Hopkins, CC BY-ND

why does it matter

along with some live tissueVery little material can learn how to be better at handling unexpected loads. Imagine the wing of an airplane suddenly picking up a gust of wind and being forced in an unexpected direction. The pavilion could not change its design to be stronger in this direction.

The modular mesh materials we designed can adapt to changing or unknown conditions. In the wing, for example, these changes could be a buildup of internal damage, changes in how the wing attaches to a vehicle, or fluctuations in external loads. Every time a suite made up of a mechanical neural network encountered one of these scenarios, it could strengthen and loosen its connections to maintain desirable attributes such as directional strength. Over time, through successive modifications made by the algorithm, the suite adopts and maintains new properties, adding each behavior to the rest as a kind of muscle memory.

This type of material can have far-reaching applications for the longevity and efficiency of built structures. A wing made of mechanical neural network material could not only be stronger, but could also be trained to transform into shapes that maximize fuel efficiency in response to the changing conditions around it.

What is still unknown

So far, our team has only worked with 2D synapses. But using computer modeling, we expect that 3D synapses will have a greater ability to learn and adapt. This increase is due to the fact that a three-dimensional structure can have dozens of connections, or springs, that do not intersect with each other. However, the mechanisms we used in our first model are too complex to support in a large 3D architecture.

What’s Next

The material created by my colleagues and I is a proof of concept and demonstrates the potential of mechanical neural networks. But to bring this idea into the real world, it would require knowing how to make individual pieces smaller with precise properties of bending and tension.

Hope new research in Micron-scale material manufacturingas well as working on it New material with adjustable hardnesswill lead to advances that will make powerful, micron-elementally intelligent mechanical neural networks and dense 3D connections a ubiquitous reality in the near future.

This article has been republished from Conversation Under Creative Commons License. Read the original article.

image rights: UCLA Resilient Research Group


Leave a Reply

Your email address will not be published. Required fields are marked *

Google Pixel 7a just appeared on Amazon
Google Pixel 7a just appeared on Amazon

Google Pixel 7a just appeared on Amazon

We expect to see a file Google Pixel 7a It arrives in the coming months,

AI disrupts long-held assumptions about general rules
AI disrupts long-held assumptions about general rules

AI disrupts long-held assumptions about general rules

Contrasting carefully Textual dialogue is found in most books and movies, the

You May Also Like