Reproduce “Identifying regions of importance in wall-bounded turbulence through explainable deep learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Andres Cremades, Sergio Hoyas, Rahul Deshpande, Pedro Quintero, Martin Lellep, Will Junghoon Lee, Jason Monty, Nicholas Hutchins and 3 more
An explainable deep learning framework combines U-net architecture with SHAP to objectively quantify the dynamic importance of coherent structures in wall-bounded turbulence. This approach demonstrates that wall-attached ejections and sweeps are crucial for predicting future flow states, and it identifies smaller, dynamically significant structures often overlooked by traditional Reynolds shear stress contributions, successfully applying to both DNS and experimental data.
Key insights
- Explainable AI, specifically SHAP, provides an objective, data-driven metric for quantifying the dynamic importance of turbulent structures, moving beyond traditional Reynolds shear stress contributions.
- Scaling SHAP values by structure volume/surface reveals that smaller, wall-detached structures can be highly dynamically important for future flow evolution, even if they contribute minimally to Reynolds shear stress.
- The framework successfully applies to both high-fidelity DNS data and experimental data at higher Reynolds numbers, demonstrating its robustness and potential for broader applicability in turbulence research and flow control.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.