Designing precise dynamical steady states in disordered networks
Elastic structures can be designed to exhibit precise, complex, and exotic functions. While recent work has focused on the quasistatic limit governed by force balance, the mechanics at a finite driving rate are governed by Newton’s equations. The goal of this work is to study the feasibility, constr...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IOP Publishing
2025-01-01
|
| Series: | Machine Learning: Science and Technology |
| Subjects: | |
| Online Access: | https://doi.org/10.1088/2632-2153/ade590 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849418279620182016 |
|---|---|
| author | Marc Berneman Daniel Hexner |
| author_facet | Marc Berneman Daniel Hexner |
| author_sort | Marc Berneman |
| collection | DOAJ |
| description | Elastic structures can be designed to exhibit precise, complex, and exotic functions. While recent work has focused on the quasistatic limit governed by force balance, the mechanics at a finite driving rate are governed by Newton’s equations. The goal of this work is to study the feasibility, constraints, and implications of creating disordered structures with exotic properties in the dynamic regime. The dynamical regime offers responses that cannot be realized in quasistatics, such as responses at an arbitrary phase, frequency-selective responses, and history-dependent responses. We employ backpropagation through time and gradient descent to design spatially specific steady states in disordered spring networks. We find that a broad range of steady states can be achieved with small alterations to the structure, operating both at small and large amplitudes. We study the effect of varying the damping, which interpolates between the underdamped and the overdamped regime, as well as the amplitude, frequency, and phase. We show that convergence depends on several competing effects, including chaos, large relaxation times, a gradient bias due to finite time simulations, and strong attenuation. By studying the eigenmodes of the linearized system, we show that the systems adapt very specifically to the task they were trained to perform. Our work demonstrates that within physical bounds, a broad array of exotic behaviors in the dynamic regime can be obtained, allowing for a richer range of possible applications. |
| format | Article |
| id | doaj-art-8c1b54df2bc14b239ba1d52dd4c8ea84 |
| institution | Kabale University |
| issn | 2632-2153 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IOP Publishing |
| record_format | Article |
| series | Machine Learning: Science and Technology |
| spelling | doaj-art-8c1b54df2bc14b239ba1d52dd4c8ea842025-08-20T03:32:28ZengIOP PublishingMachine Learning: Science and Technology2632-21532025-01-016202507310.1088/2632-2153/ade590Designing precise dynamical steady states in disordered networksMarc Berneman0https://orcid.org/0000-0002-3821-1933Daniel Hexner1https://orcid.org/0000-0002-7241-0741Faculty of Mechanical Engineering, Technion—Israel Institute of Technology , Haifa, IsraelFaculty of Mechanical Engineering, Technion—Israel Institute of Technology , Haifa, IsraelElastic structures can be designed to exhibit precise, complex, and exotic functions. While recent work has focused on the quasistatic limit governed by force balance, the mechanics at a finite driving rate are governed by Newton’s equations. The goal of this work is to study the feasibility, constraints, and implications of creating disordered structures with exotic properties in the dynamic regime. The dynamical regime offers responses that cannot be realized in quasistatics, such as responses at an arbitrary phase, frequency-selective responses, and history-dependent responses. We employ backpropagation through time and gradient descent to design spatially specific steady states in disordered spring networks. We find that a broad range of steady states can be achieved with small alterations to the structure, operating both at small and large amplitudes. We study the effect of varying the damping, which interpolates between the underdamped and the overdamped regime, as well as the amplitude, frequency, and phase. We show that convergence depends on several competing effects, including chaos, large relaxation times, a gradient bias due to finite time simulations, and strong attenuation. By studying the eigenmodes of the linearized system, we show that the systems adapt very specifically to the task they were trained to perform. Our work demonstrates that within physical bounds, a broad array of exotic behaviors in the dynamic regime can be obtained, allowing for a richer range of possible applications.https://doi.org/10.1088/2632-2153/ade590disordered structuresdynamical steady statesautomatic differentiation |
| spellingShingle | Marc Berneman Daniel Hexner Designing precise dynamical steady states in disordered networks Machine Learning: Science and Technology disordered structures dynamical steady states automatic differentiation |
| title | Designing precise dynamical steady states in disordered networks |
| title_full | Designing precise dynamical steady states in disordered networks |
| title_fullStr | Designing precise dynamical steady states in disordered networks |
| title_full_unstemmed | Designing precise dynamical steady states in disordered networks |
| title_short | Designing precise dynamical steady states in disordered networks |
| title_sort | designing precise dynamical steady states in disordered networks |
| topic | disordered structures dynamical steady states automatic differentiation |
| url | https://doi.org/10.1088/2632-2153/ade590 |
| work_keys_str_mv | AT marcberneman designingprecisedynamicalsteadystatesindisorderednetworks AT danielhexner designingprecisedynamicalsteadystatesindisorderednetworks |