Session: 01-02-04: Phononic Devices

Paper Number: 174015

Memory-Integrated Metamaterial Computing Systems for Signal Gating and Delay 

The ability of mechanical systems to perform basic computations has gained traction in recent years, providing a robust supplement to digital computing in off-grid, low-power, and extreme environments where most electronic systems fail. However, much of the prior work has focused on mechanical logic applied via static deformations of origami-based structures or snap-through mechanisms of bistable lattices. In this work, we present a dynamic, wave-based mechanical computing platform squarely focused on the temporal control of mechanical signals. The metamaterial-based system exploits temperature-dependent phase change properties of shape memory alloys (SMAs) to enable functionalities such as gating, retention, and delay of impinging mechanical loads which represent a computational signal. In doing so, we aim to address a new set of challenges by manipulating transient mechanical waves through dispersive and lossy media. To achieve this objective, we implement the foundational concepts of “mechanical signal memory” within the mechanical computing framework. Specifically, we introduce a metamaterial-based mechanical computing system which selectively admits, retains, and releases incident waves with full, and precise, control over the relevant time stamps.

The proposed system utilizes thermal stimulation (i.e., heating or cooling) from the external environment to trigger phase changes in the SMA components of the metamaterial unit cells in a manner that enables precise trapping of an input signal and full control over its release, effectively forming a temporal gate. By dynamically changing the mechanical properties of SMAs within the computing substrate, the system can exhibit programmable delays in wave propagation, a critical requisite for memory-capable systems. Phase transitions between martensite and austenite are used to alter the metamaterial’s dispersion behavior and shift the underlying band gap frequencies. The system’s unit cells are engineered to exhibit high transmission in the martensite phase and create strong attenuation in the austenite phase at different target frequencies. To optimize the gating behavior and programmable delay characteristics, a systematic study of the geometric parameters of the SMA unit cells was conducted to quantify the effects imposed by variations in the unit cell thickness, neck width, and resonator shape, all of which play a pivotal role in tuning the resonant frequencies and band gap location.

To showcase the capabilities of the system, we demonstrate the notion of mechanical signal memory in the context of two examples, i.e., computational tasks. First, we achieve wave filtering by tuning the band gaps of individual SMA unit cells to either block or transmit selected wave modes. An incident signal in the form of a broadband wave packet containing a range of frequencies is introduced into the system. While the majority of the signal propagates through, a targeted frequency window is trapped and temporarily held within an intermediate layer, akin to saved or memorized data. The signal, which is fully preserved, is then released after a specified interval. Following the release, a probe of the output signal confirms the reconstruction of the packet, confirming that the system can isolate, store, and recover targeted frequency content with precision. In the second example, we achieve a mechanical wave-based realization of the “input shaping” paradigm which is widely utilized in controls theory. We demonstrate input shaping by trapping specific frequency components of the incident wave using SMA unit cells and releasing them after a controlled delay. The delayed wave then combines with the ongoing signal, producing constructive or destructive interference based on their relative phase which can be used to instigate cancelation or amplification of the outputted waveform. Frequency domain analysis of the downstream signal in the computing system further confirms this selective superposition effect, validating the suppression of a designated frequency band. This work paves the way for higher-order operations such as classification, signal separation, and analog summation through sequencing and time alignment of mechanical signals.

Presenting Author: Saeed Behboodi University at Buffalo (SUNY)

Presenting Author Biography: Saeed Behboodi is a PhD student in the Department of Mechanical and Aerospace Engineering at the University at Buffalo (SUNY). He conducts his research in the Sound and Vibrations Laboratory.

Authors:

Saeed Behboodi University at Buffalo (SUNY)
Mohamed Mousa University at Buffalo (SUNY)
Mostafa Nouh University at Buffalo (SUNY)