Order reduction in dynamic systems using machine learning
Academic Coordinator | Andrés Gómez Tato, Applications & Projects Department Manager at CESGA.
Business Coordinator | Ángel Rivero Jimenez, Senior scientist at Repsol Technology Center.
Specialist | Pablo Solano López, PhD student, Applied Physics Department (UPM)
Description | The goal is to reduce the degrees of freedom of a given problem (a non-linear PDE or a system of non-linear PDEs known) through neural networks (autoencoder or Boltzmann machines type).
Essentially this method of order reduction uses machine learning in 1D canonical problems (with intention to extend the method to 2D and 3D) of continuous media and known dynamics governed by non-linear EDPs (Burgers, Kuramoto-Shivasinskii, Cahn Hilliard or KdV).
Scope | Machine learning
