sparseSpatialSampling.export module

Interpolate CFD data for the specified fields and time steps onto the coarse grid sampled using the \(S^3\) algorithm. Export the interpolated data to HDF5 and XDMF to enable visualization in ParaView.

class sparseSpatialSampling.export.ExportData(s_cube, write_new_file_for_each_field=False, n_jobs=None, n_neighbors=None, interpolate_at_vertices=False, write_times=None, append_existing=False)[source]

Bases: object

Initialize an Export object for interpolating original snapshots onto the grid generated by \(S^3\) and exporting them to HDF5.

Parameters:

  • s_cube (SparseSpatialSampling) – SparseSpatialSampling object containing the sampled grid

  • write_new_file_for_each_field (bool) – If True, each field is written to a separate HDF5 file; if False, all fields are written to a single file. Gets disables if append_existing = True

  • n_jobs (int | None) – Number of CPUs used for interpolation. If None, the same number of CPUs used for executing \(S^3\) will be used

  • n_neighbors (int | None) – Number of neighbors for the KNN. If None, defaults are 8 for 2D and 26 for 3D

  • interpolate_at_vertices (bool) – If True, interpolate a solution at cell vertices in addition to cell centers

  • write_times (str | list[int | float | str] | None) – Numerical time steps of the simulation to be exported

  • append_existing (bool) –

    flag to append a field to an existing HFD5 file.

    Note

    The s_cube objects and therefore the meshes must be identical in order to append the field. The mesh consistency is not checked.

export(coordinates, data, field_name, n_snapshots_total=None, chunk_size=100000)[source]

Interpolate the provided CFD data onto the grid generated by \(S^3\) and export it to HDF5 and XDMF files for all specified time steps.

Note

The field data from CFD must have dimensions [N_cells, N_dimensions, N_snapshots]. For scalar fields, N_dimensions = 1.

Parameters:

  • coordinates (pt.Tensor) – Coordinates of the original CFD grid

  • data (pt.Tensor) – Original field data with dimensions [N_cells, N_dimensions, N_snapshots]. - N_snapshots can represent all snapshots, a batch of snapshots, or a single snapshot

  • field_name (str) – Name of the field to export (e.g., 'p' for the pressure field)

  • n_snapshots_total (int | None) – Total number of snapshots to export - If None, it is assumed that all snapshots are included in data

  • chunk_size (int) – Number of cells to interpolate at once. Helpful if the memory is the restricting factor and

Return type:

None

the batch size is already 1. The memory requirements scale linearly with the chunks size, so decreasing it saves memory. :type chunk_size: int :return: None :rtype: None

property new_file: bool

Flag indicating whether a new HDF5 file will be created for each field.

Returns:

True if a new file is created for each field, False otherwise

Return type:

bool

property save_dir: str

Get the directory where output files will be saved.

Returns:

Path of the save directory

Return type:

str

property save_name: str

Get the base name of the output files.

Returns:

Name of the file used for saving the grid and data

Return type:

str

property write_times: list

Get the list of available write times

Returns:

List of time steps

Return type:

list

class sparseSpatialSampling.export.Fields(centers=None, vertices=None)[source]

Bases: object

Initialize a container for storing interpolated field values at cell centers and vertices.

Parameters:

  • centers (pt.Tensor | None) – Interpolated field values at cell centers

  • vertices (pt.Tensor | None) – Interpolated field values at cell vertices

sparseSpatialSampling.export.interpolate_data(weights, idx_weights, data, chunk_size=100000)[source]

Interpolate the data matrix from CFD onto the \(S^3\) grid. To avoid memory issues, we loop over chunks of cells

Parameters:

  • weights (pt.Tensor) – computed KNN weights for each point in the CFD grid

  • idx_weights (pt.Tensor) – indices of the k nearest neighbors for each point in the CFD grid

  • data (pt.Tensor) – data matrix from CFD, which should be interpolated

  • chunk_size (int) – number of cells to be interpolated at once

Returns:

interpolated datamatrix

Return type:

pt.Tensor