Methods and Applications of Analysis

Volume 26 (2019)

Number 3

Special Issue in Honor of Roland Glowinski (Part 2 of 2)

Guest Editors: Xiaoping Wang (Hong Kong University of Science and Technology) and Xiaoming Yuan (The University of Hong Kong)

Parameter identification of a fluid-structure system by deep-learning with an Eulerian formulation

Pages: 281 – 290

DOI: https://dx.doi.org/10.4310/MAA.2019.v26.n3.a5

Author

Olivier Pironneau (Laboratoire J.-L. Lions , University of Paris VI, Paris, France)

Abstract

A simple fluid-structure problem is considered as a test to assess the feasibility of deep-learning algorithms for parameter identification. Tensorflow by Google is used and as it is a stochastic algorithm, provision must be made for the robustness of the large displacement fluidstructure simulator with respect to a wide range of values for the Lamé coefficients and the density of the solid. Hence an Eulerian monolithic solver is introduced. The numerical tests validate the deep-learning approach.

Keywords

parameter identification, fluid-structure interaction, genetic algorithm, neural network, deep learning

2010 Mathematics Subject Classification

35Q30, 65N30, 68T01, 90C15

Received 25 December 2017

Received revised 30 July 2019

Published 2 April 2020