Reproduce “Deep neural operators can serve as accurate surrogates for shape optimization: A case study for airfoils” 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
Khemraj Shukla, Vivek Oommen, Ahmad Peyvan, Michael Penwarden, Luis Bravo, Anindya Ghoshal, Robert M. Kirby, George Em Karniadakis
Deep Neural Operators (DeepONets) accurately infer full high-dimensional flow fields around airfoils, serving as efficient surrogate models for aerodynamic shape optimization. This approach yields an over 32,000-fold speedup in online optimization compared to high-fidelity CFD, maintaining fidelity for lift-to-drag ratio predictions on unseen geometries.
Key insights
- DeepONets effectively learn the operator mapping from airfoil geometry to high-dimensional flow fields, exhibiting low generalization error (relative L2 errors around 4-7e-03) on unseen geometries.
- The ability to predict entire flow fields offers enhanced flexibility for defining complex and varied objective functions in optimization, beyond simple scalar performance metrics.
- Automatic differentiation through the DeepONet's computational graph provides an accurate and efficient method for calculating derivatives of physical quantities, such as wall shear stress, necessary for objective function evaluation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.