Eventos Anais de eventos
DINAME 2023
XIX International Symposium on Dynamic Problems of Mechanics
Transformers Surrogates for Vortex-Induced Vibrations Computational Simulations
Submission Author:
Rodolfo S. M. Freitas , RJ
Co-Authors:
Rodolfo S. M. Freitas, Fernando Rochinha
Presenter: Fernando Rochinha
doi://10.26678/ABCM.DINAME2023.DIN2023-0074
Abstract
The accurate prediction of structural instability caused by vortex shedding behind bodies or by nonlinear unsteady aerodynamics is fundamental to avoiding the degradation of structural performance or even failure of the system. Numerous approaches can represent analytical models to model both the structure and fluid. The CFD (Computational Fluid Dynamics) approaches consists of solving the Navier-Stokes equations directly, mostly limited by heavily computational costs that, many times, are tough to satisfy in practical engineering. To increase the expectations of solving practical problems, surrogate models are an alternative approach to the underlying physics. Such models have become an essential tool to simplify the analysis and can be a very useful tool in broad industrial applications. In this work, we propose the self-attention transformers model to act as a surrogate for vortex-induced vibrations (VIV) dynamics. We show by numerical experimentation that the surrogate model can accurately predict the VIV dynamics, and more importantly, it can be a suitable tool for many-query applications like sensitivity analysis, design, optimization, or uncertainty quantification.
Keywords
vortex induced vibrations, Transformers, Deep learning, Surrogate modeling
