PB
P. Belardinelli
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
1 records found
1
Micro- and nanomechanical resonators are compliant systems that can exhibit nonlinear behaviour at relatively low actuation levels. Accurately identifying the governing modal parameters of such nonlinear systems remains a significant challenge. Here, we present a deep learning-based methodology for nonlinear system identification that predicts modal parameters from a single frequency sweep. By training neural networks on non-dimensionalised data, the approach is scale-invariant and broadly applicable to resonators ranging from the nano- to the macroscale. To generate training data efficiently, we introduce a parallel time-integration scheme for rapid frequency-sweep simulation, implemented in the open-source Python package Poscidyn. The methodology is validated on a two-degree-of-freedom benchmark system exhibiting 1:2 internal resonance with nonlinear modal coupling. The results demonstrate accurate parameter prediction and highlight the potential of the approach for data-driven identification of nonlinear systems. This work demonstrates a new route for nonlinear system identification that can easily be extended to systems with higher degrees of freedom and more complex nonlinear interactions.
...
Micro- and nanomechanical resonators are compliant systems that can exhibit nonlinear behaviour at relatively low actuation levels. Accurately identifying the governing modal parameters of such nonlinear systems remains a significant challenge. Here, we present a deep learning-based methodology for nonlinear system identification that predicts modal parameters from a single frequency sweep. By training neural networks on non-dimensionalised data, the approach is scale-invariant and broadly applicable to resonators ranging from the nano- to the macroscale. To generate training data efficiently, we introduce a parallel time-integration scheme for rapid frequency-sweep simulation, implemented in the open-source Python package Poscidyn. The methodology is validated on a two-degree-of-freedom benchmark system exhibiting 1:2 internal resonance with nonlinear modal coupling. The results demonstrate accurate parameter prediction and highlight the potential of the approach for data-driven identification of nonlinear systems. This work demonstrates a new route for nonlinear system identification that can easily be extended to systems with higher degrees of freedom and more complex nonlinear interactions.