airfrans.dataset

download(root, file_name='Dataset', unzip=True, OpenFOAM=False)[source]

Download AirfRANS dataset.

Parameters

  • root (str) – Root directory where the dataset will be downloaded and unzipped.

  • file_name (str, optional) – Name of the dataset file. Default: 'Dataset'

  • unzip (bool, optional) – If True, unzip the dataset file. Default: True

  • OpenFOAM (bool, optional) – If True, it will download the raw OpenFOAM simulation with no post-processing to manipulate it through PyVista. If False, it will download the .vtu and .vtp of cropped simulations with a reduced quantity of features. Those cropped simulations have been used to train models proposed in the associated paper. Default: False

load(root, task, train=True)[source]

The different tasks ('full', 'scarce', 'reynolds', 'aoa') define the utilized training and test splits. Please note that the test set for the 'full' and 'scarce' tasks are the same. Each simulation is given as a point cloud defined via the nodes of the simulation mesh. Each point of a point cloud is described via 7 features: its position (in meters), the inlet velocity (two components in meter per second), the distance to the airfoil (one component in meter), and the normals (two components in meter, set to 0 if the point is not on the airfoil).

Each point is given a target of 4 components for the underlying regression task: the velocity (two components in meter per second), the pressure divided by the specific mass (one component in meter squared per second squared), the turbulent kinematic viscosity (one component in meter squared per second).

Finally, a boolean is attached to each point to inform if this point lies on the airfoil or not.

The output is a tuple of a list of np.ndarray of shape (N, 7 + 4 + 1), where N is the number of points in each simulation and where the features are ordered as presented in this documentation, and a list of name for the each corresponding simulation.

We highly recommend to handle those data with the help of a Geometric Deep Learning library such as PyTorch Geometric or Deep Graph Library.

Parameters

  • root (string) – Root directory where the simulation directories have been saved.

  • task (string) – The task to study ('full', 'scarce', 'reynolds', 'aoa') that defines the utilized training and test splits.

  • train (bool, optional) – If True, loads the training dataset, otherwise the test dataset. Default: True