zoomy-core 0.1.11__py3-none-any.whl

zoomy_core/preprocessing/openfoam_moments.py

@@ -0,0 +1,452 @@
+try:
+    _HAVE_MATPLOTLIB = True
+    import matplotlib.pyplot as plt
+except ImportError:
+    _HAVE_MATPLOTLIB = False
+
+import numpy as np
+from copy import deepcopy
+import os
+from scipy.spatial import KDTree
+
+try:
+    import pyvista as pv
+
+    _HAVE_PYVISTA = True
+except ImportError:
+    _HAVE_PYVISTA = False
+
+from zoomy_core.mesh.fvm_mesh import Mesh
+from zoomy_core.misc.io import _save_fields_to_hdf5 as save_fields_to_hdf5
+import zoomy_core.misc.io as io
+
+
+main_dir = os.getenv("ZOOMY_DIR")
+
+
+def load_file(filename):
+    if not _HAVE_PYVISTA:
+        raise ImportError("pyvista is required for load_file function.")
+    reader = pv.get_reader(filename)
+    # get(0) gets us the internal (not boundary) data. The boundary data is non-existant anyways in our case
+    vtkfile = reader.read().get(0)
+    return vtkfile
+
+
+def get_fields(vtkfile, fieldnames):
+    fields = []
+    for fieldname in fieldnames:
+        # point field)s
+        # field = vtkfile.point_data[fieldname]
+        field = vtkfile.cell_data[fieldname]
+        fields.append(np.array(field))
+    return fields
+    # return np.array(fields)
+
+
+def get_coordinates(vtkfile):
+    # point coordinates
+    # return np.array(vtkfile.points)
+    # cell centers
+    return vtkfile.cell_centers().points
+
+
+def get_time(vtkfile):
+    return vtkfile.field_data["TimeValue"][0]
+
+
+def sort_data(coordinates, fields):
+    # Sort by z, y, x
+    order = np.lexsort((coordinates[:, 0], coordinates[:, 1], coordinates[:, 2]))
+    return order
+
+
+def apply_order(coordinates, fields, order):
+    coordinates_copy = np.array(coordinates)
+    fields_copy = deepcopy(fields)
+    for d in range(coordinates.shape[1]):
+        coordinates[:, d] = coordinates_copy[order, d]
+    for field, field_copy in zip(fields, fields_copy):
+        field[:] = field_copy[order]
+    return coordinates, fields
+
+
+def get_number_of_layers_and_elements_in_plane(coordinates):
+    layers = len(np.unique(coordinates[:, 2]))
+    n_elements_plane = int(coordinates[:, 0].size / layers)
+    return layers, n_elements_plane
+
+
+def get_layer(data, layer, n_elements_plane):
+    return data[layer * n_elements_plane : (layer + 1) * n_elements_plane]
+
+
+def compute_height(
+    alpha, n_layers, n_elements_per_layer, total_height=1.0, threshold=0.5
+):
+    height = np.zeros(n_elements_per_layer)
+    active = np.ones(n_elements_per_layer, dtype=bool)
+    dh = total_height / n_layers
+    for i in range(n_layers):
+        alpha_layer = get_layer(alpha, i, n_elements_per_layer)
+        height += np.where(np.logical_and((alpha_layer >= threshold), active), dh, 0)
+    return height
+
+
+def extract_faces(mesh):
+    faces = []
+    i, offset = 0, 0
+    cc = mesh.cells  # fetch up front
+    while i < mesh.n_cells:
+        nn = cc[offset]
+        faces.append(cc[offset + 1 : offset + 1 + nn])
+        offset += nn + 1
+        i += 1
+    return np.array(faces)
+
+
+def sort_faces(faces, order):
+    for face in faces:
+        for i, point in enumerate(face):
+            face[i] = order[point]
+    faces = faces[order]
+    return faces
+
+
+def add_noslip_layer(U, n_elements_per_layer):
+    Unew = np.zeros((U.shape[0] + n_elements_per_layer, U.shape[1]))
+    for d in range(U.shape[1]):
+        Unew[n_elements_per_layer:, d] = U[:, d]
+    return Unew
+
+
+def extract_velocity_column_at_coordinate(U, n_elements_per_layer, n_layer, coordinate):
+    u = np.array(
+        [
+            get_layer(U[:, 0], i, n_elements_per_layer)[coordinate]
+            for i in range(n_layer + 1)
+        ]
+    )
+    v = np.array(
+        [
+            get_layer(U[:, 1], i, n_elements_per_layer)[coordinate]
+            for i in range(n_layer + 1)
+        ]
+    )
+    w = np.array(
+        [
+            get_layer(U[:, 2], i, n_elements_per_layer)[coordinate]
+            for i in range(n_layer + 1)
+        ]
+    )
+    return u, v, w
+
+
+def shift_integration_interval(xi):
+    return (xi + 1) / 2
+
+
+def plot_basis(basis_generator):
+    fig, ax = plt.subplots()
+    basis = [basis_generator(n) for n in range(0, 8)]
+    X = np.linspace(0, 1, 100)
+    for i in range(8):
+        ax.plot(basis[i](X), X)
+    return fig, ax
+
+
+def moment_projection(field, n_layers, basis, integration_order=None):
+    if integration_order is None:
+        integration_order = max(len(basis), n_layers)
+    xi, wi = np.polynomial.legendre.leggauss(integration_order)
+    xi = shift_integration_interval(xi)
+    dz = 1 / (n_layers)
+    xp = np.arange(dz / 2, 1 - dz / 2, dz)
+    xp = np.insert(xp, 0, 0)
+    fp = field
+    field_xi = np.interp(xi, xp, fp)
+    basis_xi = [basis[i](xi) for i in range(len(basis))]
+    projections = np.zeros(len(basis))
+    for i in range(len(basis)):
+        projections[i] = np.sum(field_xi * basis_xi[i] * wi)
+    return projections
+
+
+def convert_openfoam_to_moments_single(filename, n_levels):
+    vtkfile = load_file(filename)
+    coordinates = get_coordinates(vtkfile)
+    n_layer, n_elements_per_layer = get_number_of_layers_and_elements_in_plane(
+        coordinates
+    )
+    fields = get_fields(vtkfile, ["alpha.water", "U"])
+    sort_order = sort_data(coordinates, fields)
+    coordinates, fields = apply_order(coordinates, fields, sort_order)
+    fields[1] = add_noslip_layer(fields[1], n_elements_per_layer)
+    time = get_time(vtkfile)
+    Q, basis = compute_shallow_moment_projection(fields, coordinates, n_levels)
+    return coordinates, Q, time, basis
+
+
+def convert_openfoam_to_moments(
+    filepath, n_levels, filepath_mesh_for_order, meshtype_order="triganle"
+):
+    filepath_vtk = os.path.join(filepath, "VTK")
+    filename = "fields_openfoam.hdf5"
+    sort_order = None
+
+    # file order
+    file_number_list = []
+    file_start = ""
+    for i_file, file in enumerate(os.listdir(filepath_vtk)):
+        if file.endswith(".vtm"):
+            file_start = file.split("_")[0]
+            file_start = file.rsplit("_", 1)[0]
+            index = int(file.rsplit("_", 1)[1].split(".")[0])
+            # index = int(re.findall(r'\d+', file)[-1])
+            file_number_list.append(index)
+
+    file_number_list.sort()
+    print(file_number_list)
+
+    # first iteration
+    file0 = (
+        os.path.join(filepath_vtk, file_start) + "_" + str(file_number_list[0]) + ".vtm"
+    )
+    coordinates, Q, time, basis = convert_openfoam_to_moments_single(file0, n_levels)
+    mesh_order = Mesh.load_gmsh(filepath_mesh_for_order, meshtype_order)
+    Q, sort_order = sort_fields_by_mesh(mesh_order, coordinates, Q)
+    if os.path.exists(os.path.join(filepath, filename)):
+        os.remove(os.path.join(filepath, filename))
+    save_fields_to_hdf5(filepath, 0, time, Q.T, Qaux=None, filename=filename)
+
+    # loop iterations
+    iter = 1
+    total = len(file_number_list)
+    print(f"Conversion {iter}/{total} completed.")
+    iter += 1
+    # loop iterations
+    for i, i_file in enumerate(file_number_list[1:]):
+        file = os.path.join(filepath_vtk, file_start) + "_" + str(i_file) + ".vtm"
+        coordinates, Q, time, basis = convert_openfoam_to_moments_single(file, n_levels)
+        Q = apply_order_to_fields(sort_order, Q)
+        save_fields_to_hdf5(filepath, i + 1, time, Q.T, Qaux=None, filename=filename)
+        print(f"Conversion {iter}/{total} completed.")
+        iter += 1
+
+
+def sort_fields_by_mesh(mesh, coordinates, Q):
+    # coordinates are still 3d, while Q is 2d and the mesh is 2d
+    n_layers, n_elements_per_layer = get_number_of_layers_and_elements_in_plane(
+        coordinates
+    )
+    coords_2d = coordinates[:n_elements_per_layer, :2]
+    coords_ordered = mesh.element_center
+    # Construct a KDTree from coords_2d
+    tree = KDTree(coords_2d)
+
+    # Find the indices of the nearest points in coords_2d to the points in coords_ordered
+    _, indices = tree.query(coords_ordered)
+
+    # Use these indices to sort Q and coords_2d
+    Q_sorted = Q[:, indices]
+    return Q_sorted, indices
+
+
+def apply_order_to_fields(order, Q):
+    return Q[:, order]
+
+
+def plot_contour(coordinates, field):
+    if not _HAVE_MATPLOTLIB:
+        raise ImportError("matplotlib is required for plot_contour function.")
+    fig, ax = plt.subplots()
+    n_layer, n_elements_per_layer = get_number_of_layers_and_elements_in_plane(
+        coordinates
+    )
+    X = get_layer(coordinates[:, 0], 0, n_elements_per_layer)
+    Y = get_layer(coordinates[:, 1], 0, n_elements_per_layer)
+    # Z = get_layer(fields[1][:, 0], 1, n_elements_per_layer)
+    # Z = get_layer(fields[0], 12, n_elements_per_layer)
+    # Z = compute_height(fields[0], n_layer, n_elements_per_layer)
+    Z = field
+    colorbar = ax.tricontourf(X, Y, Z)
+    ax.set_aspect("equal")
+
+    circle = plt.Circle((0.5, 1), radius=0.2, fc="silver", zorder=10, edgecolor="k")
+    plt.gca().add_patch(circle)
+    fig.colorbar(colorbar)
+    return fig, ax
+
+
+def basis_legendre(i):
+    def f(x):
+        basis = np.polynomial.legendre.Legendre.basis(i, domain=[0, 1], window=[-1, 1])
+        return basis(x) * basis(0)
+
+    return f
+
+
+def compute_shallow_moment_projection(
+    fields, coordinates, n_levels, basis_generator=basis_legendre
+):
+    n_layers, n_elements_per_layer = get_number_of_layers_and_elements_in_plane(
+        coordinates
+    )
+    height = compute_height(fields[0], n_layers, n_elements_per_layer)
+    basis = [basis_generator(n) for n in range(0, n_levels + 1)]
+    alphas = np.zeros((n_elements_per_layer, n_levels + 1))
+    betas = np.zeros((n_elements_per_layer, n_levels + 1))
+    for i in range(n_elements_per_layer):
+        u, v, w = extract_velocity_column_at_coordinate(
+            fields[1], n_elements_per_layer, n_layers, i
+        )
+        alphas[i] = moment_projection(u, n_layers + 1, basis)
+        betas[i] = moment_projection(v, n_layers + 1, basis)
+    return np.concatenate(
+        (height.reshape((n_elements_per_layer, 1)), alphas, betas), axis=1
+    ).T, basis
+
+
+def plot_data_vs_moments(fields, coordinates, Q, basis, coordinate_index):
+    if _HAVE_MATPLOTLIB is False:
+        raise ImportError("matplotlib is required for plot_data_vs_moments function.")
+    n_layers, n_elements_per_layer = get_number_of_layers_and_elements_in_plane(
+        coordinates
+    )
+    n_levels = int((Q.shape[0] - 1) / 2) - 1
+    X = np.linspace(0, 1, 100)
+    U = np.array(
+        [basis[i](X) * Q[1 + i, coordinate_index] for i in range(len(basis))]
+    ).sum(axis=0)
+    V = np.array(
+        [
+            basis[i](X) * Q[1 + n_levels + 1 + i, coordinate_index]
+            for i in range(len(basis))
+        ]
+    ).sum(axis=0)
+    dz = 1 / (n_layers)
+    x = np.linspace(dz / 2, 1 - dz / 2, n_layers)
+    x = np.insert(x, 0, 0)
+    u, v, w = extract_velocity_column_at_coordinate(
+        fields[1], n_elements_per_layer, n_layers, coordinate_index
+    )
+    fig, ax = plt.subplots(2)
+    ax[0].plot(U, X)
+    ax[0].plot(u, x)
+    ax[1].plot(V, X)
+    ax[1].plot(v, x)
+    return fig, ax
+
+
+def load_openfoam_file(filepath):
+    vtkfile = load_file(filepath)
+    coordinates = get_coordinates(vtkfile)
+    n_layer, n_elements_per_layer = get_number_of_layers_and_elements_in_plane(
+        coordinates
+    )
+    fields = get_fields(vtkfile, ["alpha.water", "U"])
+    sort_order = sort_data(coordinates, fields)
+    coordinates, fields = apply_order(coordinates, fields, sort_order)
+    fields[1] = add_noslip_layer(fields[1], n_elements_per_layer)
+    time = get_time(vtkfile)
+    return coordinates, fields, time
+
+
+def test_load():
+    filepath = os.path.join(
+        os.path.join(main_dir, "openfoam_data/channelflow_coarse"),
+        "channelflow_coarse_0.vtm",
+    )
+    coordinates, fields, time = load_openfoam_file(filepath)
+
+
+def test_moment_projection():
+    filepath = os.path.join(
+        os.path.join(main_dir, "openfoam_data/channelflow_coarse"),
+        "channelflow_coarse_0.vtm",
+    )
+    coordinates, fields, time = load_openfoam_file(filepath)
+    Q, basis = compute_shallow_moment_projection(fields, coordinates, 3)
+
+
+def test_convert_openfoam_single():
+    filepath = os.path.join(
+        os.path.join(main_dir, "openfoam_data/channelflow_coarse"),
+        "channelflow_coarse_0.vtm",
+    )
+    X, Q, t, basis = convert_openfoam_to_moments_single(filepath, 3)
+
+
+def test_plots():
+    if _HAVE_MATPLOTLIB is False:
+        raise ImportError("matplotlib is required for test_plots function.")    
+    filepath = os.path.join(
+        os.path.join(main_dir, "openfoam_data/channelflow_coarse"),
+        "channelflow_coarse_0.vtm",
+    )
+    coordinates, fields, time = load_openfoam_file(filepath)
+    X, Q, t, basis = convert_openfoam_to_moments_single(filepath, 3)
+    fig, ax = plot_data_vs_moments(fields, coordinates, Q, basis, 0)
+    plt.show()
+    fig, ax = plot_basis(basis_legendre)
+    plt.show()
+    fig, ax = plot_contour(coordinates, Q[0])
+    plt.show()
+
+
+def test_sort():
+    filepath = os.path.join(
+        os.path.join(main_dir, "openfoam_data/channelflow_coarse"),
+        "channelflow_coarse_0.vtm",
+    )
+    filepath_gmsh = os.path.join(
+        os.path.join(main_dir, "meshes/channel_openfoam/mesh_coarse_2d.msh")
+    )
+
+    mesh_gmsh = Mesh.load_gmsh(filepath_gmsh, "triangle")
+    mesh_gmsh.write_to_hdf5(
+        os.path.join(os.path.join(main_dir, "openfoam_data/channelflow_coarse"))
+    )
+
+    filepath_hdf_mesh = os.path.join(
+        os.path.join(main_dir, "openfoam_data/channelflow_coarse/mesh.hdf5")
+    )
+    mesh = Mesh.from_hdf5(filepath_hdf_mesh)
+
+    coordinates, fields, time = load_openfoam_file(filepath)
+    Q, basis = compute_shallow_moment_projection(fields, coordinates, 3)
+
+    # mesh_comparison = Mesh.load_gmsh(os.path.join(main_dir, 'meshes/channel_openfoam/mesh_coarse_3d.msh'), 'tetra')
+
+    sort_fields_by_mesh(mesh, coordinates, Q)
+
+
+def test_convert_openfoam_to_moments(level=0):
+    # filepath = os.path.join(main_dir, 'openfoam_data/channelflow_coarse')
+    foam_sim = os.getenv("FOAM_SIM")
+    filepath = os.path.join(foam_sim, "multiphase/interFoam/RAS/channelflow_mid")
+    filepath_target_mesh = os.path.join(
+        os.path.join(main_dir, "meshes/simple_openfoam/mesh_2d_mid.msh")
+    )
+    convert_openfoam_to_moments(
+        filepath, level, filepath_target_mesh, meshtype_order="triangle"
+    )
+
+
+if __name__ == "__main__":
+    # test_load()
+    # test_moment_projection()
+    # test_convert_openfoam_single()
+    # test_plots()
+    # test_sort()
+    test_convert_openfoam_to_moments(level=1)
+    filepath = os.path.join(main_dir, "openfoam_data/channelflow_mid")
+    filepath_mesh = os.path.join(main_dir, "meshes/simple_openfoam/mesh_2d_mid.msh")
+    io.generate_vtk(
+        filepath,
+        filepath_gmsh=filepath_mesh,
+        gmsh_mesh_type="triangle",
+        filename_fields="fields_openfoam.hdf5",
+        filename_out="fields_openfoam_vtk",
+        skip_aux=True,
+    )

zoomy_core/transformation/helpers.py

@@ -0,0 +1,25 @@
+from sympy import MatrixSymbol, fraction, cancel, Matrix
+
+from zoomy_core.misc.misc import Zstruct
+
+def regularize_denominator(expr, regularization_constant = 10**(-4), regularize = False):
+    if not regularize:
+        return expr
+    def regularize(expr):
+        (nom, den) = fraction(cancel(expr))
+        return nom * den / (den*2 + regularization_constant)
+    for i in range(expr.shape[0]):
+        for j in range(expr.shape[1]):
+            expr[i,j] = regularize(expr[i,j])
+    return expr
+
+def substitute_sympy_attributes_with_symbol_matrix(expr: Matrix, attr: Zstruct, attr_matrix: MatrixSymbol):
+    if expr is None:
+        return None
+    if type(attr) is Zstruct:
+        assert attr.length() <= attr_matrix.shape[0]
+        for i, k in enumerate(attr.get_list()):
+            expr = Matrix(expr).subs(k, attr_matrix[i])
+    else:
+        expr = Matrix(expr).subs(attr, attr_matrix)
+    return expr