Session: S-07 Computing Metamaterials

Paper Number: 171992

171992 - Elastoacoustic Mechanical Intelligence: The Pursuit of Smart, Multi-Tasking Waves 

Mechanical intelligence has gained prominence in recent years due to advances in metamaterials and fabrication techniques, providing analog computing mechanisms with self-embedded sensory and actuation capabilities. Owing to their physical nature, such systems achieve a level of autonomy, through thermal, acoustic, and mechanical forms of material response, that enable them to seamlessly integrate with surrounding environments. In the presence of elevated temperatures, ionizing radiations, low electric power, or any extreme conditions that render electronic and digital components inoperable, such systems provide supplementary computations and enable information processing at a level that maybe otherwise infeasible. Most of the early efforts in this space have relied on static and quasi-static deformations, infusing elements of bi-stability, buckling, and employing origami concepts to create logic gates and mechanical bits. On the other hand, wave-based mechanical computing is the process of executing complex mathematical operations by manipulating incident waves through guided elastic wave scattering.

In this talk, we provide an overview of distinct classes of wave-based mechanical computers that leverage the rich dynamics of periodic and phononic materials, while integrating mechanical memory and thermal adaptivity within the computing system. We demonstrate that these systems can undertake high-order computations such as differentiation and convolution or embody mechanical intelligence by realizing physical neural architectures. While promising, the overarching constraints on elastic wave propagation including reciprocity, transmission symmetry, and preset dispersion patterns, inherently limit these systems to single-task configurations. As such, their inability to re-adapt to new information or concurrently perform multiple tasks (i.e., compute in parallel) has remained elusive. As such, we will show how to exploit dynamically modulated metasurfaces to unlock parallel processing of incident information. The metasurfaces simultaneously conduct independent computational tasks within the same elastic medium via frequency multiplexing. By breaking time invariance in an array of tunable locally resonant cells, multiple frequency-shifted beams are self-generated at higher and lower harmonics, which absorb notable energy amounts from the fundamental signal. The onset of these tunable harmonics enables distinct computational tasks to be assigned to different independent “channels,” effectively allowing a wave-based mechanical computer to multitask.

Finally, we will shed light on a class of multifunctional mechanical computing circuits that incorporate memory-integrated components, configured via shape-memory alloy (SMA)-based metamaterial cells, which trigger specific actions upon thermal activation. These systems lay the foundation for substantial advancements in both acoustic and optical domains, and open up several avenues which thus far have been elusive in physical and reservoir computing.

Presenting Author: Mohamed Mousa University at Buffalo (SUNY)

Presenting Author Biography: Mohamed is a senior PhD candidate in the Mechanical and Aerospace Engineering department at the University at Buffalo (SUNY), currently conducting research in the Sound and Vibrations Laboratory. His research centers on acoustic and elastic metamaterials, exploring their unique properties. His work involves modeling analysis, utilizing machine learning techniques, and performing experimental testing on these systems, with the goal of realizing innovative mechanical computing. He holds a BS and MS from Cairo University in Mechanical Power Engineering and has 5+ years of experience as a Teaching assistant at CU and a Turbomachinery Control engineer at Schneider Electric.