CFD benchmark with attack

基础网络以及对应的训练数据采用清华大学发表的preprint文章《CFDBench: A Large-Scale Benchmark for Machine Learning Methods in Fluid Dynamics》所用的数据集，详见github.com/luo-yining/CFDBench，该数据集包含四种物理模型，分别是dam, tube, cylinder以及cavity，并分别包含在改变几何模型，边界条件以及流体性质集中工况下的流场数据。本项目组在该数据的基础上，增加了对应的fgsm攻击和测试代码。The basic network and the corresponding training data adopt the dataset used in the preprint article “CFDBench: A Large-Scale Benchmark for Machine Learning Methods in Fluid Dynamics” by Tsinghua University, available at github.com/luo-yining/CFDBench.

This dataset contains four physical models—namely dam, tube, cylinder, and cavity—and includes flow field data under various working conditions involving changes in geometry, boundary conditions, and fluid properties. On the basis of this dataset, our project team has added corresponding FGSM attack and testing code.