Sparse Measurement-Based Reduced-Order Models of Fluid-Structure-Interactions

Introduction

Fluid-structure interaction (FSI) is a ubiquitous phenomenon in industries such as ocean engineering, biomedicine, aerospace, and is a driven idea for hydrokinetic energy conversion. Strong nonlinear interactions between flow and structures, as well as turbulent flow, bring a huge challenge for understanding and predicting the dynamics of FSI systems. Facing this challenge and the rising motivation of harnessing clean energy from natural fluids, SMARTFLUIDS aims to build a framework for developing Reduced Order Models (ROMs) with low complexity while retaining dominant physics for FSI. This will lead to a deeper understanding and efficient prediction of FSI.

Methodology

A low-dimensional subspace of the FSI will be extracted using modal analysis through Variational Autoencoders (VAE) based on deep learning (DL) of FSI data. The FSI data will be obtained by performing high-fidelity computational fluid dynamics (CFD) simulations of classic FSI problems.

Development of ROMs

The ROMs are developed by:

1. Physics-informed mapping of governing equations onto the low-dimensional subspace.
2. Data-driven techniques to deal with nonlinear and unresolved parts of the FSI.

Physical constraints are incorporated, and sparse measurements of the FSI will be used in building the ROMs.

Predictions

Dynamics and future states of the FSI, hydro- or aerodynamic loads on the structures, and structural responses can be predicted based on the ROMs.

Impact

SMARTFLUIDS will provide an innovative and systematic view of FSI by focusing on a few dominant features and enriching the knowledge of FSI physics by adopting the latest DL techniques. The project will bring a novel solution to reduce the time and cost of CFD and experiments in predicting FSI dynamics and structural responses for engineering design of:

• Pipelines
• Cables
• Wind turbine blades
• Airplane wings

Additionally, it will promote renewable energy development.