Convolutional Neural Networks for the Construction of Surrogate Models of Fluid Flows

Author(s)

Lui, Hugo F. S. ; Wolf, William R.

• Document type: Journal article
• Source: AIAA SCITECH 2021 FORUM; v. N/A, p. 15-pg., 2021-01-01.

Abstract

In this work, we present a numerical methodology for construction of surrogate models of fluid flows which combine data-driven system identification and convolutional neural networks. The framework is implemented in a context similar to that of the sparse identification of nonlinear dynamics (SINDy) algorithm with some modifications regarding the regression step. The approach presented in this work allows us to obtain an ODE for each flow variable at each mesh point. This should be beneficial for flow control approaches since every flow state can be modified by a control law. The method is tested for two unsteady compressible flows: the flow past a cylinder at low Reynolds number and the turbulent flow past a plunging airfoil under deep dynamical stall. Results demonstrate that the current methodology provides accurate reconstructions of the high fidelity model. (AU)

• FAPESP's process: 13/08293-7 - CCES - Center for Computational Engineering and Sciences
• Grantee: Munir Salomao Skaf
• Support Opportunities: Research Grants - Research, Innovation and Dissemination Centers - RIDC
• FAPESP's process: 13/07375-0 - CeMEAI - Center for Mathematical Sciences Applied to Industry
• Grantee: Francisco Louzada Neto
• Support Opportunities: Research Grants - Research, Innovation and Dissemination Centers - RIDC
• FAPESP's process: 19/26196-5 - Large-eddy simulations of supersonic axial turbines
• Grantee: Hugo Felippe da Silva Lui
• Support Opportunities: Scholarships in Brazil - Doctorate