Researchers from the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS) have developed a machine learning platform designed to "program" two-dimensional stretchable materials — so they can morph into specific three-dimensional shapes.
"While machine learning methods have been classically employed for image recognition and language processing, they have also recently emerged as powerful tools to solve mechanics problems," claims Katia Bertoldi, the William and Ami Kuan Danoff Professor of Applied Mechanics at SEAS and the study's senior author. "In this work we demonstrate that these tools can be extended to study the mechanics of transformable, inflatable systems."
The team's work centers around an inflatable membrane, divided into a 100 "pixel" grid of 10×10 squares. Each individual square in the grid can be either "soft" or "stiff," and the membrane as a whole can be inflated to a variety of pressures — and while 100 pixels doesn't sound like many, the possible combinations are numerous. Too numerous, in fact, to program manually — which is where the new machine learning platform comes in.
"Once the machine learning model was trained, we came up with an arbitrary 3D shape and passed it to the model," explains Antonio Elia Forte, first author of the paper. "The neural network then outputs the membrane design and the pressure at which we should inflate such membrane to obtain the desired 3D shape."
The approach has already proven its worth: The team successfully employed the platform to build a mechanotherapeutic device designed to stimulate the tissue around a scar in order to improve healing and reduce the patient's recovery time — a process that requires the device to map closely to the shape of both the scar and the area of the body being treated.
“This platform has potential to quickly and effectively design patient-specific devices for mechanotherapy and beyond," says Forte. "Before this research, we didn't know how to use machine learning to unravel non-linear mappings in inflatable systems but it turns out that they are very powerful for these purposes. Machine learning could push the boundaries of currently known design strategies and allow us to design and build fully reconfigurable shape-morphing material."
The team's work has been published in the journal Advanced Functional Materials under open-access terms.