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 fluid-structure 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.