Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
The Effect of Data Selection on the Sparse Identification of Dynamical Systems
Submission Author:
Davi Saadi de Almeida Lettieri , RJ , Brazil
Co-Authors:
Davi Saadi de Almeida Lettieri, Leonardo Santos de Brito Alves
Presenter: Davi Saadi de Almeida Lettieri
doi://10.26678/ABCM.COBEM2021.COB2021-0316
Abstract
Machine learning has continuously provided new and more efficient techniques for the discovery of governing equations. A well known branch of such techniques is known as symbolic regression. Although it had been left aside for a long time due to its high costs, it has been recently brought back through the development of SINDy, i.e. Sparse Identification of Nonlinear Dynamics. This technique is analyzed here when applied to the identification of partial differential equations, more specifically the unsteady, one-dimensional and viscous Burgers’ equation. The analysis is performed by choosing different temporal data sets generated at different spatial locations during different time spans in order to investigate SINDy’s ability to rediscover the Burgers’ equation while employing different libraries of candidate functions. The results indicate a strong sensitivity to the spatial location and time period chosen, even when the library contains only the original terms of this equation.
Keywords
machine learning, symbolic regression, SINDy, Burgers' Equation, Data Selection
