Embedded Files
Co-chairs:Â
Xiaodong Li (University of Southampton)*; Liang Fang (Imperial College London)
Key words: Computational Fluid Mechanics, Reduced-Order Modeling, Shadowing-based Methods.
ABSTRACT
Scale-resolved simulations are essential for capturing the complex dynamics of turbulent flows, particularly in cases involving flow separation or high unsteadiness, where traditional Reynolds-averaged Navier-Stokes (RANS) methods often fail to provide accurate predictions. High-fidelity simulations, such as Large-Eddy Simulation (LES) and Direct Numerical Simulation (DNS), offer a more reliable representation of coherent structures in turbulent flows. However, performing sensitivity analysis (e.g., adjoint methods) within such simulations is challenging due to the chaotic nature of turbulence.
To make LES/DNS applicable in engineering tasks such as optimization and flow control, innovative methods must be developed to efficiently compute gradients in scale-resolved simulations. One promising research area is shadowing theory, which provides strategies to address instability in sensitivity analysis caused by turbulence. However, further improvements are necessary to reduce computational costs for high-Reynolds-number turbulent flows. Another avenue of exploration involves Unstable Periodic Orbits (UPOs), which can prevent the exponential growth of adjoint variables in response to small perturbations. Efficient methods for identifying UPOs in high-Re turbulent flows, however, are still under development.
This proposed mini-symposium will serve as a forum for researchers to exchange ideas and present recent advancements in computational methods aimed at improving gradient computations in scale-resolved simulations. We welcome contributions from a variety of methodologies, including (but not limited to):
New algorithms to accelerate shadowing-based methods, such as Least-Squares Shadowing (LSS) and Non-Intrusive LSS.
Efficient methods for identifying UPOs, including reduced-order modeling techniques.
Data-driven analysis approaches for turbulent flows, such as Proper Orthogonal Decomposition (POD) and Spectral POD (SPOD).
Machine learning applications (e.g., autoencoders and neural networks) for enhancing flow analysis and prediction.
Page updated
Google Sites
Report abuse