ML4Fluids Challenges

We are pleased to announce two exciting challenges associated with the 3rd ERCOFTAC Workshop on ML4Fluids, each designed to advance machine learning applications in computational fluid dynamics.

AI Surrogate Challenge

The AI Surrogate Challenge focuses on neural operators, Physics-Informed Neural Networks (PINNs), and direct ML prediction approaches for fluid dynamics applications. This challenge is directly affiliated with the ML4Fluids conference and aims to benchmark state-of-the-art surrogate modeling techniques for complex flow simulations. A dedicated webpage with full challenge details, datasets, evaluation metrics, and submission guidelines is currently under development and will be announced separately.

Organizers: Neil Ashton, Johannes Branstetter, and Siddhartha Mishra

Data-Driven Turbulence Modelling Challenge

Building on NASA's successful 2022 turbulence modelling challenge, this initiative focuses on advancing data-driven turbulence modelling approaches for 3D cases of industrial relevance. Participants will develop and test novel RANS models, enhanced turbulence closures, hybrid RANS-LES approaches, or other data-driven turbulence modeling techniques on a set of challenging test cases spanning various flow phenomena such as periodic hills, smooth-body separation, automotive wake flows, and wing-body junction flows. Evaluation is based on accuracy, computational cost, and generalizability. The community-driven challenge welcomes test case suggestions from participants. Visit the dedicated challenge webpage at github.com/rmcconke/ml-turbulence-benchmark for complete technical specifications, final test case details, training datasets, boundary conditions, and evaluation protocols.

Organizers: Tyler Buchanan, Paola Cinnella, Richard Dwight, and Ryley McConkey
To participate: Send an expression of interest to t.s.b.buchanan@tudelft.nl