Non-intrusive reduced-order model for unsteady fluid flow and fluid-structure interaction problems

Sugar-Gabor, O; Fairchild, D

Abstract

A non-intrusive reduced-order model applicable to time-dependent parametric systems is
developed. The model is based on extracting a reduced-order basis from high-order snapshots via proper
orthogonal decomposition. Multi-layered feedforward artificial neural networks are used to capture the
unsteady behaviour of the reduced-order coefficients. Nonlinear regression networks using Radial Basis
Functions are used to approximate the parameter space behaviour of the reduced-order coefficients,
ensuring the method is applicable to problems having an arbitrary number of parameter samples. An
adaptive sampling approach based on determining the relative influence of each parameter sample on the
overall approximation is implemented. This increases the quality of the neural network prediction while
minimising the required number of parameter samples and the number of high-order snapshots. The
resulting model is fully non-intrusive, it is independent of the high-order computational method and can be
used together with black-box solvers. First, numerical studies for two canonical fluid flow test cases,
unsteady incompressible laminar flow around a circular cylinder and transonic inviscid flow around a pitching
NACA 0012 aerofoil. Next, two fluid-structure interaction test cases are considered, sinusoidal pressure
waves interacting with a metallic rod, and large-displacement oscillation of an elastic beam in incompressible
laminar flow. Results confirm the accuracy of the non-intrusive approach for the various tests considered.

Citation

Sugar-Gabor, O., & Fairchild, D. (2022, October). Non-intrusive reduced-order model for unsteady fluid flow and fluid-structure interaction problems. Presented at International Conference of Aerospace Sciences AEROSPATIAL 2022, INCAS, Bucharest, Romania

• Presentation Conference Type: Keynote
• Conference Name: International Conference of Aerospace Sciences AEROSPATIAL 2022
• Conference Location: INCAS, Bucharest, Romania
• Start Date: Oct 13, 2022
• End Date: Oct 14, 2022
• Publication Date: Oct 13, 2022
• Deposit Date: Nov 25, 2022
• Publisher: National Institute for Aerospace Research “Elie Carafoli”
• Publisher URL: https://aerospatial-2022.incas.ro/
• Additional Information: Event Type : Conference