CFDBench: Computational Fluid Dynamics Benchmark Dataset

CFDBench is a large-scale benchmark dataset specifically designed for machine learning methods in computational fluid dynamics. It contains four classic CFD problems: lid-driven cavity flow, laminar pipe flow, step dam-break flow, and flow around a cylinder. The dataset is intended to evaluate the generalization capabilities of neural operators when facing unseen boundary conditions, fluid properties, and geometries. This benchmark supports research in applying deep learning to physical simulations and PDE solving.