コード例 #1
0
ファイル: interpolate.py プロジェクト: daveb-dev/flowtools
 def set_probe_points(self, pts):
     self.probes = StatisticsProbes(pts.flatten(), self.V, True)
     if self.h5fp_str is not None:
         self.probes_p = StatisticsProbes(pts.flatten(), self.V, False)
     if self.compute_stress:
         self.probes_grad_u = dict()
         for dim in xrange(3):
             self.probes_grad_u[dim] = StatisticsProbes(
                 pts.flatten(), self.Vv)
コード例 #2
0
def test_StatisticsProbes_vector_3D(VF3):
    u0 = interpolate(Expression(('x[0]', 'x[1]', 'x[2]'), degree=1), VF3)
    x = array([[0.5, 0.25, 0.25], [0.4, 0.4, 0.4], [0.3, 0.3, 0.3]])
    probes = StatisticsProbes(x.flatten(), VF3)

    for i in range(5):
        probes(u0)
        
    p = probes.array()
    if MPI.rank(mpi_comm_world()) == 0:
        assert round(p[0,0] - 2.5, 7) == 0
        assert round(p[0,4] - 0.3125, 7) == 0
コード例 #3
0
def test_StatisticsProbes_vector_3D(VF3):
    u0 = interpolate(Expression(('x[0]', 'x[1]', 'x[2]'), degree=1), VF3)
    x = array([[0.5, 0.25, 0.25], [0.4, 0.4, 0.4], [0.3, 0.3, 0.3]])
    probes = StatisticsProbes(x.flatten(), VF3)

    for i in range(5):
        probes(u0)

    p = probes.array()
    if MPI.rank(MPI.comm_world) == 0:
        assert round(p[0,0] - 2.5, 7) == 0
        assert round(p[0,4] - 0.3125, 7) == 0
コード例 #4
0
def test_StatisticsProbes_segregated_2D(V2):
    u0 = interpolate(Expression('x[0]', degree=1), V2)
    v0 = interpolate(Expression('x[1]', degree=1), V2)
    x = array([[0.5, 0.25], [0.4, 0.4], [0.3, 0.3]])
    probes = StatisticsProbes(x.flatten(), V2, True)

    for i in range(5):
        probes(u0, v0)
        
    p = probes.array()
    if MPI.rank(mpi_comm_world()) == 0:
        assert round(p[0,0] - 2.5, 7) == 0
        assert round(p[0,4] - 0.625, 7) == 0
コード例 #5
0
def test_StatisticsProbes_segregated_2D(V2):
    u0 = interpolate(Expression('x[0]', degree=1), V2)
    v0 = interpolate(Expression('x[1]', degree=1), V2)
    x = array([[0.5, 0.25], [0.4, 0.4], [0.3, 0.3]])
    probes = StatisticsProbes(x.flatten(), V2, True)

    for i in range(5):
        probes(u0, v0)

    p = probes.array()
    if MPI.rank(MPI.comm_world) == 0:
        assert round(p[0,0] - 2.5, 7) == 0
        assert round(p[0,4] - 0.625, 7) == 0
コード例 #6
0
def test_StatisticsProbes_segregated_3D(V3):
    u0 = interpolate(Expression('x[0]', degree=1), V3)
    v0 = interpolate(Expression('x[1]', degree=1), V3)
    w0 = interpolate(Expression('x[2]', degree=1), V3)
    x = array([[0.5, 0.25, 0.25], [0.4, 0.4, 0.4], [0.3, 0.3, 0.3]])
    probes = StatisticsProbes(x.flatten(), V3, True)

    for i in range(5):
        probes(u0, v0, w0)

    p = probes.array()
    if MPI.rank(MPI.comm_world) == 0:
        assert round(p[0,0] - 2.5, 7) == 0
        assert round(p[0,4] - 0.3125, 7) == 0
コード例 #7
0
def pre_solve_hook(mesh, V, **NS_namespace):
    # Normals and facets to compute flux at inlet and outlet
    normal = FacetNormal(mesh)
    Inlet = AutoSubDomain(inlet)
    Outlet = AutoSubDomain(outlet)
    Walls = AutoSubDomain(walls)
    Centerline = AutoSubDomain(centerline)
    facets = FacetFunction('size_t', mesh, 0)
    Inlet.mark(facets, 1)
    Outlet.mark(facets, 2)
    Walls.mark(facets, 3)
    Centerline.mark(facets, 4)

    z_senterline = linspace(-0.18269, 0.320, 1000)
    x = array([[i, 0.0] for i in z_senterline])
    senterline = StatisticsProbes(x.flatten(), V)

    return dict(uv=Function(V), senterline=senterline, facets=facets,
                normal=normal)
コード例 #8
0
def pre_solve_hook(u_, Lx, D, V, Q, mesh, restart_folder, velocity_degree, AssignedVectorFunction, newfolder, MPI, **NS_namespace):
    """Called prior to time loop"""
    if MPI.rank(MPI.comm_world) == 0:
        makedirs(path.join(newfolder, "Probes"))

    # set up probes
    eval_dict = {}
    # probe aloing y-direction from the centerline, we select middle point in terms of x-direction
    probe_points = np.linspace((Lx/2, 0, 0), (Lx/2, D/2, 0), 100)

    # Store points file in checkpoint
    if MPI.rank(MPI.comm_world) == 0:
        probe_points.dump(path.join(newfolder, "Probes", "points"))

    # setup StatisticsProbes for velocity and probes for pressure
    eval_dict["u_probes"] = StatisticsProbes(probe_points.flatten(), V, True)
    # eval_dict["p_probes"] = Probes(probe_points.flatten(), Q)

    uv = AssignedVectorFunction(u_)
    return dict(eval_dict=eval_dict, uv=uv)
コード例 #9
0
ファイル: interpolate.py プロジェクト: daveb-dev/flowtools
class Interpolation:
    def __init__(self,
                 mesh_in,
                 filename_in,
                 folder_out=None,
                 p_filename_in=None,
                 compute_stress=False):
        self.comm = mpi4py.MPI.COMM_WORLD
        self.mpi_rank = self.comm.Get_rank()
        self.mpi_size = self.comm.Get_size()

        self.h5fmesh_str = mesh_in
        self.h5fu_str = filename_in
        self.h5fp_str = p_filename_in
        self.folder_out = folder_out

        if folder_out is not None and \
           not os.path.exists(folder_out) and self.mpi_rank == 0:
            os.makedirs(folder_out)

        # Form compiler options
        df.parameters["form_compiler"]["optimize"] = True
        df.parameters["form_compiler"]["cpp_optimize"] = True
        df.parameters["form_compiler"]["representation"] = "quadrature"
        df.parameters["std_out_all_processes"] = False

        df.info("Loading mesh")
        self._load_mesh()

        df.info("Initializing elements and function spaces")
        E = df.FiniteElement("Lagrange", self.mesh.ufl_cell(), 1)
        self.V = df.FunctionSpace(self.mesh, E)

        df.info("Initializing coordinates-to-node map")
        self._initialize_geometry()

        self.compute_stress = compute_stress

        if self.compute_stress:
            df.info("Preparing to compute stresses")
            # self.Vv = df.VectorFunctionSpace(self.mesh, "DG", 0)
            self.Vv = df.VectorFunctionSpace(self.mesh, "CG", 1)
            # self.V2 = df.FunctionSpace(self.mesh, "CG", 2)
            self.grad_u = dict()
            for dim in xrange(3):
                self.grad_u[dim] = df.Function(self.Vv)

    def _make_dof_coords(self):
        dofmap = self.V.dofmap()
        my_first, my_last = dofmap.ownership_range()
        x = self.V.tabulate_dof_coordinates().reshape((-1, 3))
        unowned = dofmap.local_to_global_unowned()
        dofs = filter(
            lambda dof: dofmap.local_to_global_index(dof) not in unowned,
            xrange(my_last - my_first))
        self.x = x[dofs]

    def _make_xdict(self, x_data):
        if self.mpi_rank == 0:
            self.xdict = dict([(tuple(x_list), i)
                               for i, x_list in enumerate(x_data.tolist())])
        else:
            self.xdict = None
        self.xdict = self.comm.bcast(self.xdict, root=0)

    def _set_val(self, f, f_data):
        vec = f.vector()
        values = vec.get_local()
        values[:] = [
            f_data[self.xdict[tuple(x_val)]] for x_val in self.x.tolist()
        ]
        vec.set_local(values)
        vec.apply('insert')

    def _load_mesh(self):
        """ Reads the mesh from the timeseries file """
        self.mesh = df.Mesh()
        h5fmesh = df.HDF5File(self.mesh.mpi_comm(), self.h5fmesh_str, "r")
        h5fmesh.read(self.mesh, "/mesh", False)
        h5fmesh.close()

    def _initialize_geometry(self):
        """ Initialize from timeseries file """
        self._make_dof_coords()
        with h5py.File(self.h5fu_str, "r") as h5fu:
            self.x_data = np.array(h5fu.get("Mesh/0/mesh/geometry"))
        self._make_xdict(self.x_data)
        self.u = dict()
        for dim in xrange(3):
            self.u[dim] = df.Function(self.V)
        if self.h5fp_str is not None:
            self.p = df.Function(self.V)
        self.u_data = np.zeros_like(self.x_data)
        if self.h5fp_str is not None:
            self.p_data = np.zeros((len(self.x_data), 1))

    def set_probes(self, ptsfilename):
        df.info("Reading probe points")
        self.ptsfilename = ptsfilename
        with h5py.File(self.ptsfilename, "r") as h5fi:
            pts = np.array(h5fi["x"])
        self.set_probe_points(pts)

    def set_probe_points(self, pts):
        self.probes = StatisticsProbes(pts.flatten(), self.V, True)
        if self.h5fp_str is not None:
            self.probes_p = StatisticsProbes(pts.flatten(), self.V, False)
        if self.compute_stress:
            self.probes_grad_u = dict()
            for dim in xrange(3):
                self.probes_grad_u[dim] = StatisticsProbes(
                    pts.flatten(), self.Vv)

    def update(self, step):
        """ Update fields """
        with h5py.File(self.h5fu_str, "r") as h5fu:
            if step == "-1" or step == -1:
                steps = [int(key) for key in h5fu["VisualisationVector"]]
                self.stepstr = str(np.max(steps))
            else:
                self.stepstr = step
            if "VisualisationVector/" + self.stepstr not in h5fu:
                return False
            self.u_data[:, :] = np.array(
                h5fu.get("VisualisationVector/" + self.stepstr))
        for dim in xrange(3):
            df.info("Setting u[" + str(dim) + "]")
            self._set_val(self.u[dim], self.u_data[:, dim])
        if self.h5fp_str is not None:
            with h5py.File(self.h5fp_str, "r") as h5fp:
                if "VisualisationVector/" + self.stepstr not in h5fp:
                    return False
                self.p_data[:] = np.array(
                    h5fp.get("VisualisationVector/" + self.stepstr))
                df.info("Setting p")
                self._set_val(self.p, self.p_data[:])
        if self.compute_stress:
            for dim in xrange(3):
                df.info("Computing grad(u[" + str(dim) + "])")
                # df.info(" --> step 1")
                # u2 = df.project(self.u[dim], self.V2,
                #                 solver_type="gmres",
                #                 preconditioner_type="amg",
                #                 form_compiler_parameters={"optimize": True})
                # df.info(" --> step 2")
                # self.grad_u[dim].assign(
                #    df.project(
                #         df.nabla_grad(u2), self.Vv,
                #         solver_type="gmres",
                #         preconditioner_type="amg",
                #         form_compiler_parameters={"optimize": True}))
                self.grad_u[dim].assign(
                    df.project(df.nabla_grad(self.u[dim]),
                               self.Vv,
                               solver_type="gmres",
                               preconditioner_type="amg",
                               form_compiler_parameters={"optimize": True}))
        return True

    def probe(self):
        self.probes.clear()
        df.info("Probing!")
        self.probes(self.u[0], self.u[1], self.u[2])
        if self.h5fp_str is not None:
            self.probes_p(self.p)
        if self.compute_stress:
            for dim in xrange(3):
                self.probes_grad_u[dim](self.grad_u[dim])

    def dump(self, add=0):
        data_mat = self.probes.array(1)
        if self.h5fp_str is not None:
            data_mat_p = self.probes_p.array(1)
        if self.compute_stress:
            data_mat_grad_u = dict()
            for dim in xrange(3):
                data_mat_grad_u[dim] = self.probes_grad_u[dim].array(1)
        if add == 0:
            stepstrout = self.stepstr
        else:
            stepstrout = str(int(self.stepstr) + add)
        if self.mpi_rank == 0:
            df.info("Saving!")
            with h5py.File(
                    os.path.join(self.folder_out, "u_" + stepstrout + ".h5"),
                    "w") as h5f_out:
                h5f_out.create_dataset("/u/" + stepstrout,
                                       data=data_mat[:, :3])
                if self.h5fp_str is not None:
                    h5f_out.create_dataset("/p/" + stepstrout,
                                           data=data_mat_p[:, 0])
                if self.compute_stress:
                    h5f_out.create_dataset("/grad_u/" + stepstrout,
                                           data=np.hstack(
                                               (data_mat_grad_u[0][:, :3],
                                                data_mat_grad_u[1][:, :3],
                                                data_mat_grad_u[2][:, :3])))

    def get_dsets(self, id_first, id_last):
        with h5py.File(self.h5fu_str, "r") as h5fu:
            dsets_u = set(
                [int(key) for key in h5fu["VisualisationVector/"].keys()])
        if self.h5fp_str is not None:
            with h5py.File(self.h5fp_str, "r") as h5fp:
                dsets_p = set(
                    [int(key) for key in h5fp["VisualisationVector/"].keys()])
            dsets = [str(key) for key in sorted(list(dsets_u & dsets_p))]
        else:
            dsets = [str(key) for key in sorted(list(dsets_u))]
        if id_last >= len(dsets):
            id_last = len(dsets) - 2
        if id_first > id_last:
            id_first = id_last
        dsets_loc = dsets[id_first:id_last]
        dsets_loc.append(dsets[id_last])
        return dsets_loc
コード例 #10
0
            radius.append(r_1)
        else:
            radius.append(r_0)

    # Container for all StatisticsProbes
    eval_dict = {}
    key_u = "slice_u_%s"

    eps = 1e-8
    n_slice = 200
    for i in range(len(z)):
        # Set up dict for the slices
        u_ = key_u % z[i]
        slices_points = linspace(-radius[i], radius[i], n_slice)
        points = array([[x, 0, z[i]] for x in slices_points])
        eval_dict[u_] = StatisticsProbes(points.flatten(), Pv, True)

        # Store points
        points.dump(path.join(newfolder, "Stats", "Points", "slice_%s" % z[i]))

    # Setup probes in the centerline and at the wall
    N = 10000
    z_senterline = linspace(start + eps, stop - eps, N)
    eval_senter = array([[0.0, 0.0, i] for i in z_senterline])
    eval_senter.dump(path.join(newfolder, "Stats", "Points", "senterline"))
    eval_wall = []

    cone_length = 0.022685
    for z_ in z_senterline:
        # first and last cylinder
        if z_ < -0.062685 or z_ > 0.0: