Navier-Stokes Fluid Flow Dataset

该数据集包含了从二维纳维-斯托克斯流体绕圆柱流动问题的域中采样的点，用于训练物理信息神经网络（PINNs）以预测速度和压力场。数据集采用汉密尔顿采样方法，同时针对偏微分方程（PDE）和边界条件进行生成，旨在探究配置点对PINN预测的影响。在规模上，训练数据集包含10,000个点（其中7,500个为PDE点，2,500个为边界点）；测试集则包括来自域内的36,934个点以及来自边界的1,288个点。任务是通过使用PINNs来预测流体动力学问题中的速度场（U1、U2）和压力（P）。

This dataset contains points sampled from the computational domain of the 2D Navier-Stokes flow around a cylinder problem, and is used to train Physics-Informed Neural Networks (PINNs) for predicting velocity and pressure fields. The dataset is generated using the Hamiltonian sampling method, and is constructed for both the partial differential equation (PDE) and boundary conditions, aiming to investigate the effect of collocation points on PINN predictions. In terms of scale, the training dataset consists of 10,000 points, including 7,500 PDE collocation points and 2,500 boundary points; the test set includes 36,934 interior domain points and 1,288 boundary points. The task is to predict the velocity fields (U1, U2) and pressure (P) in this fluid dynamics problem using PINNs.