Reproduce “A Model Hierarchy for Predicting the Flow in Stirred Tanks with Physics-Informed Neural Networks” 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
Veronika Trávníková, Daniel Wolff, Nico Dirkes, Stefanie Elgeti, Eric von Lieres, Marek Behr
A hierarchy of Physics-Informed Neural Network (PINN) models was developed to predict fluid flow in stirred tank reactors. The models incorporate physical insights, such as strong boundary conditions and domain decomposition, to achieve ℓ1 velocity errors less than 1% and enable accurate predictions across a range of Reynolds numbers.
Key insights
- Integrating physical insights, such as strong Dirichlet boundary conditions and tailored domain decomposition, into Physics-Informed Neural Networks (PINNs) substantially enhances their accuracy and robustness for complex fluid flow problems.
- A novel adaptation of the XPINN/DPINN approach, solving different sets of governing equations (e.g., simplified ODEs vs. full PDEs) in distinct subdomains, effectively handles geometrically intricate domains and varying flow characteristics.
- Parameterized PINNs can create fast-to-evaluate, robust surrogate models capable of accurately interpolating flow fields across a range of operational conditions, reducing the need for extensive high-fidelity simulations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.