LoopStructural 1.0.9__cp36-cp36m-win32.whl

LoopStructural/interpolators/structured_tetra.py

@@ -0,0 +1,681 @@
+"""
+Tetmesh based on cartesian grid for piecewise linear interpolation
+"""
+import logging
+
+import numpy as np
+from LoopStructural.interpolators.cython.dsi_helper import cg, constant_norm, fold_cg
+
+from LoopStructural.utils import getLogger
+logger = getLogger(__name__)
+
+class TetMesh:
+    """
+
+    """
+    def __init__(self, origin = [0,0,0], nsteps = [10,10,10], step_vector = [1,1,1]):
+        self.origin = np.array(origin)
+        self.step_vector = np.array(step_vector)
+        self.nsteps = np.array(nsteps)
+        self.n_nodes = self.nsteps[0]*self.nsteps[1]*self.nsteps[2]
+
+        self.nsteps_cells = self.nsteps - 1
+        self.n_cell_x = self.nsteps[0] - 1
+        self.n_cell_y = self.nsteps[1] - 1
+        self.n_cell_z = self.nsteps[2] - 1
+        self.n_cells = self.n_cell_x * self.n_cell_y * self.n_cell_z
+        self.maximum = origin+self.nsteps*self.step_vector
+        self.tetra_mask_even = np.array([
+            [7,1,2,4],
+            [6,2,4,7],
+            [5,1,4,7],
+            [0,1,2,4],
+            [3,1,2,7]
+        ])
+
+        self.tetra_mask = np.array([
+            [0,6,5,3],
+            [7,3,5,6],
+            [4,0,5,6],
+            [2,0,3,6],
+            [1,0,3,5]
+        ])
+        self.ntetra = self.n_cells * 5
+        self.n_elements = self.ntetra
+        self.cg = None
+
+    @property
+    def nodes(self):
+        """
+        Gets the nodes of the mesh as a property rather than using a function, accessible as a property! Python magic!
+
+        Returns
+        -------
+        nodes : np.array((N,3))
+            Fortran ordered
+        """
+        max = self.origin + self.nsteps_cells * self.step_vector
+        x = np.linspace(self.origin[0], max[0], self.nsteps[0])
+        y = np.linspace(self.origin[1], max[1], self.nsteps[1])
+        z = np.linspace(self.origin[2], max[2], self.nsteps[2])
+        xx, yy, zz = np.meshgrid(x, y, z, indexing='ij')
+        return np.array([xx.flatten(order='F'), yy.flatten(order='F'),
+                               zz.flatten(order='F')]).T
+
+    def barycentre(self, elements = None):
+        """
+        Return the barycentres of all tetrahedrons or of specified tetras using
+        global index
+
+        Parameters
+        ----------
+        elements - numpy array
+            global index
+
+        Returns
+        -------
+
+        """
+        if elements is None:
+            elements = np.arange(0,self.ntetra)
+        tetra = self.get_elements()[elements]
+        barycentre = np.sum(self.nodes[tetra][:, :, :],
+                                 axis=1) / 4.
+        return barycentre
+
+    def evaluate_value(self, pos, property_array):
+        """
+        Evaluate value of interpolant
+
+        Parameters
+        ----------
+        pos - numpy array
+            locations
+        prop - string
+            property name
+
+        Returns
+        -------
+
+        """
+        values = np.zeros(pos.shape[0])
+        values[:] = np.nan
+        vertices, c, tetras, inside = self.get_tetra_for_location(pos)
+        values[inside] = np.sum(c[inside,:]*property_array[tetras[inside,:]],axis=1)
+        return values
+
+    def evaluate_gradient(self, pos, property_array):
+        """
+        Evaluate the gradient of an interpolant at the locations
+
+        Parameters
+        ----------
+        pos - numpy array
+            locations
+        prop - string
+            property to evaluate
+
+
+        Returns
+        -------
+
+        """
+        values = np.zeros(pos.shape)
+        values[:] = np.nan
+        vertices, element_gradients, tetras, inside = self.get_tetra_gradient_for_location(pos)
+        #grads = np.zeros(tetras.shape)
+        values[inside,:] = (element_gradients[inside,:,:]*property_array[tetras[inside,None,:]]).sum(2)
+        length = np.sum(values[inside,:],axis=1)
+        # values[inside,:] /= length[:,None]
+        return values
+
+    def inside(self, pos):
+        inside = np.ones(pos.shape[0]).astype(bool)
+        for i in range(3):
+            inside *= pos[:, i] > self.origin[None, i]
+            inside *= pos[:, i] < self.origin[None, i] + \
+                      self.step_vector[None, i] * self.nsteps_cells[None, i]
+        return inside
+
+    def get_tetra_for_location(self, pos):
+        """
+        Determine the tetrahedron from a numpy array of points
+
+        Parameters
+        ----------
+        pos : np.array
+
+
+
+        Returns
+        -------
+
+        """
+        pos = np.array(pos)
+        inside = self.inside(pos)
+        # initialise array for tetrahedron vertices
+        vertices = np.zeros((5, 4, pos.shape[0], 3))
+        vertices[:] = np.nan
+        # get cell indexes
+        c_xi, c_yi, c_zi = self.position_to_cell_index(pos)
+        
+        # determine if using +ve or -ve mask
+        even_mask = (c_xi + c_yi + c_zi) % 2 == 0
+        # get cell corners
+        xi, yi, zi = self.cell_corner_indexes(c_xi, c_yi, c_zi)  # global_index_to_node_index(gi)
+        # convert to node locations
+        nodes = self.node_indexes_to_position(xi, yi, zi).T
+
+        vertices[:, :, even_mask, :] = nodes[:, even_mask, :][self.tetra_mask_even, :, :]
+        vertices[:, :, ~even_mask, :] = nodes[:, ~even_mask, :][self.tetra_mask, :, :]
+        # changing order to points, tetra, nodes, coord
+        vertices = vertices.swapaxes(0, 2)
+        vertices = vertices.swapaxes(1, 2)
+        # use scalar triple product to calculate barycentric coords
+
+        vap = pos[:, None, :] - vertices[:, :, 0, :]
+        vbp = pos[:, None, :] - vertices[:, :, 1, :]
+        #         # vcp = p - points[:, 2, :]
+        #         # vdp = p - points[:, 3, :]
+        vab = vertices[:, :, 1, :] - vertices[:, :, 0, :]
+        vac = vertices[:, :, 2, :] - vertices[:, :, 0, :]
+        vad = vertices[:, :, 3, :] - vertices[:, :, 0, :]
+        vbc = vertices[:, :, 2, :] - vertices[:, :, 1, :]
+        vbd = vertices[:, :, 3, :] - vertices[:, :, 1, :]
+        va = np.einsum('ikj, ikj->ik', vbp, np.cross(vbd, vbc, axisa=2, axisb=2)) / 6.
+        vb = np.einsum('ikj, ikj->ik', vap, np.cross(vac, vad, axisa=2, axisb=2)) / 6.
+        vc = np.einsum('ikj, ikj->ik', vap, np.cross(vad, vab, axisa=2, axisb=2)) / 6.
+        vd = np.einsum('ikj, ikj->ik', vap, np.cross(vab, vac, axisa=2, axisb=2)) / 6.
+        v = np.einsum('ikj, ikj->ik', vab, np.cross(vac, vad, axisa=2, axisb=2)) / 6.
+        c = np.zeros((va.shape[0], va.shape[1], 4))
+        c[:, :, 0] = va / v
+        c[:, :, 1] = vb / v
+        c[:, :, 2] = vc / v
+        c[:, :, 3] = vd / v
+
+        # if all coords are +ve then point is inside cell
+        mask = np.all(c > 0, axis=2)
+
+        inside = np.logical_and(inside,np.any(mask,axis=1))
+        # get cell corners
+        xi, yi, zi = self.cell_corner_indexes(c_xi, c_yi, c_zi)
+        #create mask to see which cells are even
+        even_mask = (c_xi + c_yi + c_zi) % 2 == 0
+        # create global node index list
+        # print('nsteps',self.nsteps, 'nsteps_cells', self.nsteps_cells)
+        # print('x',np.min(c_xi),np.max(c_xi),np.min(xi),np.max(xi))
+        # print('y',np.min(c_yi),np.max(c_yi),np.min(yi),np.max(yi))
+        # print('z',np.min(c_zi),np.max(c_zi),np.min(zi),np.max(zi))
+
+        gi = xi + yi * self.nsteps[0] + zi * self.nsteps[0] * self.nsteps[1]
+        # container for tetras
+        tetras = np.zeros((xi.shape[0], 5, 4)).astype(int)
+
+        tetras[even_mask, :, :] = gi[even_mask, :][:, self.tetra_mask_even]
+        tetras[~even_mask, :, :] = gi[~even_mask, :][:, self.tetra_mask]
+        inside = np.logical_and(inside,self.inside(pos))
+        vertices_return = np.zeros((pos.shape[0],4,3))
+        vertices_return[:] = np.nan
+        # set all masks not inside to False
+        mask[~inside,:] = False
+        vertices_return[inside,:,:] = vertices[mask,:,:]#[mask,:,:]#[inside,:,:]
+        c_return = np.zeros((pos.shape[0],4))
+        c_return[:] = np.nan
+        c_return[inside] = c[mask]
+        tetra_return = np.zeros((pos.shape[0],4)).astype(int)
+        tetra_return[:] = -1
+        tetra_return[inside,:] = tetras[mask,:]
+        return vertices_return, c_return, tetra_return, inside
+
+    def get_constant_gradient(self, region, direction=None):
+        """
+        Get the constant gradient for the specified nodes
+
+        Parameters
+        ----------
+        region : np.array(dtype=bool)
+            mask of nodes to calculate cg for
+
+        Returns
+        -------
+
+        """
+        """
+        Add the constant gradient regularisation to the system
+
+        Parameters
+        ----------
+        w (double) - weighting of the cg parameter
+
+        Returns
+        -------
+
+        """
+        if direction is not None:
+            print('using cg direction')
+            logger.info("Running constant gradient")
+            elements_gradients = self.get_element_gradients(np.arange(self.ntetra))
+            if elements_gradients.shape[0] != direction.shape[0]:
+                logger.error('Cannot add directional CG, vector field is not the correct length')
+                return
+            region = region.astype('int64')
+
+            neighbours = self.get_neighbours()
+            elements = self.get_elements()
+            idc, c, ncons = fold_cg(elements_gradients, direction, neighbours.astype('int64'), elements.astype('int64'), self.nodes)
+
+            idc = np.array(idc[:ncons, :])
+            c = np.array(c[:ncons, :])
+            B = np.zeros(c.shape[0])
+            return c, idc, B
+        if self.cg is None:
+            logger.info("Running constant gradient")
+            elements_gradients = self.get_element_gradients(np.arange(self.ntetra))
+            region = region.astype('int64')
+
+            neighbours = self.get_neighbours()
+            elements = self.get_elements()
+            idc, c, ncons = cg(elements_gradients, neighbours.astype('int64'), elements.astype('int64'), self.nodes,
+                               region.astype('int64'))
+
+            idc = np.array(idc[:ncons, :])
+            c = np.array(c[:ncons, :])
+            B = np.zeros(c.shape[0])
+            self.cg = (c,idc,B)
+        return self.cg[0], self.cg[1], self.cg[2]
+    def get_constant_norm(self, region):
+        """
+        Get the constant gradient for the specified nodes
+
+        Parameters
+        ----------
+        region : np.array(dtype=bool)
+            mask of nodes to calculate cg for
+
+        Returns
+        -------
+
+        """
+        
+        logger.info("Running constant gradient")
+        elements_gradients = self.get_element_gradients(np.arange(self.ntetra))
+        region = region.astype('int64')
+
+        neighbours = self.get_neighbours()
+        elements = self.get_elements()
+        idc, c, ncons = constant_norm(elements_gradients, neighbours.astype('int64'), elements.astype('int64'), self.nodes,
+                            region.astype('int64'))
+
+        idc = np.array(idc[:ncons, :])
+        c = np.array(c[:ncons, :])
+        B = np.zeros(c.shape[0])
+        
+        return  c,idc,B
+    def get_elements(self):
+        """
+        Get a numpy array of all of the elements in the mesh
+
+        Returns
+        -------
+        numpy array elements
+
+        """
+
+        x = np.arange(0, self.n_cell_x)
+        y = np.arange(0, self.n_cell_y)
+        z = np.arange(0, self.n_cell_z)
+
+        c_xi, c_yi, c_zi = np.meshgrid(x, y, z,indexing='ij')
+        c_xi = c_xi.flatten(order='F')
+        c_yi = c_yi.flatten(order='F')
+        c_zi = c_zi.flatten(order='F')
+        # get cell corners
+        xi, yi, zi = self.cell_corner_indexes(c_xi, c_yi, c_zi)
+        even_mask = (c_xi + c_yi + c_zi) % 2 == 0
+        gi = xi + yi * self.nsteps[0] + zi * self.nsteps[0] * self.nsteps[1]
+        tetras = np.zeros((c_xi.shape[0], 5, 4)).astype('int64')
+        tetras[even_mask, :, :] = gi[even_mask, :][:, self.tetra_mask_even]
+        tetras[~even_mask, :, :] = gi[~even_mask, :][:, self.tetra_mask]
+
+        return tetras.reshape((tetras.shape[0]*tetras.shape[1],tetras.shape[2]))
+
+    def get_element_gradients(self, elements = None):
+        """
+        Get the gradients of all tetras
+
+        Parameters
+        ----------
+        elements
+
+        Returns
+        -------
+
+        """
+        if elements is None:
+            elements = np.arange(0,self.ntetra)
+        x = np.arange(0, self.n_cell_x)
+        y = np.arange(0, self.n_cell_y)
+        z = np.arange(0, self.n_cell_z)
+
+        c_xi, c_yi, c_zi = np.meshgrid(x, y, z, indexing='ij')
+        c_xi = c_xi.flatten(order='F')
+        c_yi = c_yi.flatten(order='F')
+        c_zi = c_zi.flatten(order='F')
+        even_mask = (c_xi + c_yi + c_zi) % 2 == 0
+        # get cell corners
+        xi, yi, zi = self.cell_corner_indexes(c_xi, c_yi, c_zi)  # global_index_to_node_index(gi)
+        # convert to node locations
+        nodes = self.node_indexes_to_position(xi, yi, zi).T
+
+        points = np.zeros((5, 4, self.n_cells, 3))
+        points[:, :, even_mask, :] = nodes[:, even_mask, :][self.tetra_mask_even, :, :]
+        points[:, :, ~even_mask, :] = nodes[:, ~even_mask, :][self.tetra_mask, :, :]
+
+        # changing order to points, tetra, nodes, coord
+        points = points.swapaxes(0, 2)
+        points = points.swapaxes(1, 2)
+
+        ps = points.reshape(points.shape[0] * points.shape[1], points.shape[2], points.shape[3])
+
+        m = np.array(
+            [[(ps[:, 1, 0] - ps[:, 0, 0]), (ps[:, 1, 1] - ps[:, 0, 1]),
+              (ps[:, 1, 2] - ps[:, 0, 2])],
+             [(ps[:, 2, 0] - ps[:, 0, 0]), (ps[:, 2, 1] - ps[:, 0, 1]),
+              (ps[:, 2, 2] - ps[:, 0, 2])],
+             [(ps[:, 3, 0] - ps[:, 0, 0]), (ps[:, 3, 1] - ps[:, 0, 1]),
+              (ps[:, 3, 2] - ps[:, 0, 2])]])
+        I = np.array(
+            [[-1., 1., 0., 0.],
+             [-1., 0., 1., 0.],
+             [-1., 0., 0., 1.]])
+        m = np.swapaxes(m, 0, 2)
+        element_gradients = np.linalg.inv(m)
+
+        element_gradients = element_gradients.swapaxes(1, 2)
+        element_gradients = element_gradients @ I
+
+        return element_gradients[elements,:,:]
+
+    def get_tetra_gradient_for_location(self, pos):
+        """
+        Get the gradient of the tetra for a location
+
+        Parameters
+        ----------
+        pos
+
+        Returns
+        -------
+
+        """
+        vertices, bc, tetras, inside = self.get_tetra_for_location(pos)
+        ps = vertices
+        m = np.array(
+            [[(ps[:, 1, 0] - ps[:, 0, 0]), (ps[:, 1, 1] - ps[:, 0, 1]),(ps[:, 1, 2] - ps[:, 0, 2])],
+             [(ps[:, 2, 0] - ps[:, 0, 0]), (ps[:, 2, 1] - ps[:, 0, 1]),(ps[:, 2, 2] - ps[:, 0, 2])],
+             [(ps[:, 3, 0] - ps[:, 0, 0]), (ps[:, 3, 1] - ps[:, 0, 1]),(ps[:, 3, 2] - ps[:, 0, 2])]])
+        I = np.array(
+            [[-1., 1., 0., 0.],
+             [-1., 0., 1., 0.],
+             [-1., 0., 0., 1.]])
+        m = np.swapaxes(m, 0, 2)
+        element_gradients = np.linalg.inv(m)
+
+        element_gradients = element_gradients.swapaxes(1, 2)
+        element_gradients = element_gradients @ I
+        return vertices, element_gradients, tetras, inside
+
+
+
+    def global_node_indicies(self, indexes):
+        """
+        Convert from node indexes to global node index
+
+        Parameters
+        ----------
+        indexes
+
+        Returns
+        -------
+
+        """
+        indexes = np.array(indexes).swapaxes(0, 2)
+        return indexes[:, :, 0] + self.nsteps[None, None, 0] \
+               * indexes[:, :, 1] + self.nsteps[None, None, 0] * \
+               self.nsteps[None, None, 1] * indexes[:, :, 2]
+
+    def global_cell_indicies(self, indexes):
+        """
+        Convert from cell indexes to global cell index
+
+        Parameters
+        ----------
+        indexes
+
+        Returns
+        -------
+
+        """
+        indexes = np.array(indexes).swapaxes(0, 2)
+        return indexes[:, :, 0] + self.nsteps_cells[None, None, 0] \
+               * indexes[:, :, 1] + self.nsteps_cells[None, None, 0] * \
+               self.nsteps_cells[None, None, 1] * indexes[:, :, 2]
+
+    def cell_corner_indexes(self, x_cell_index, y_cell_index, z_cell_index):
+        """
+        Returns the indexes of the corners of a cell given its location xi,
+        yi, zi
+
+        Parameters
+        ----------
+        x_cell_index
+        y_cell_index
+        z_cell_index
+
+        Returns
+        -------
+
+        """
+        x_cell_index = np.array(x_cell_index)
+        y_cell_index = np.array(y_cell_index)
+        z_cell_index = np.array(z_cell_index)
+
+        xcorner = np.array([0, 1, 0, 1, 0, 1, 0, 1])
+        ycorner = np.array([0, 0, 1, 1, 0, 0, 1, 1])
+        zcorner = np.array([0, 0, 0, 0, 1, 1, 1, 1])
+        xcorners = x_cell_index[:, None] + xcorner[None, :]
+        ycorners = y_cell_index[:, None] + ycorner[None, :]
+        zcorners = z_cell_index[:, None] + zcorner[None, :]
+        return xcorners, ycorners, zcorners
+
+    def position_to_cell_corners(self, pos):
+        """
+        Find the nodes that belong to a cell which contains a point
+
+        Parameters
+        ----------
+        pos
+
+        Returns
+        -------
+
+        """
+        inside = self.inside(pos)
+        ix, iy, iz = self.position_to_cell_index(pos)
+        cornersx, cornersy, cornersz = self.cell_corner_indexes(ix, iy, iz)
+        globalidx = self.global_cell_indicies(
+            np.dstack([cornersx, cornersy, cornersz]).T)
+        return globalidx, inside
+
+    def position_to_cell_index(self, pos):
+        """
+        Find which cell a point is in
+
+        Parameters
+        ----------
+        pos
+
+        Returns
+        -------
+
+        """
+        ix = pos[:, 0] - self.origin[None, 0]
+        iy = pos[:, 1] - self.origin[None, 1]
+        iz = pos[:, 2] - self.origin[None, 2]
+        ix = ix // self.step_vector[None, 0]
+        iy = iy // self.step_vector[None, 1]
+        iz = iz // self.step_vector[None, 2]
+        return ix.astype(int), iy.astype(int), iz.astype(int)
+
+    def node_indexes_to_position(self, xindex, yindex, zindex):
+        """
+        Get the xyz position from the node coordinates
+
+        Parameters
+        ----------
+        xindex
+        yindex
+        zindex
+
+        Returns
+        -------
+
+        """
+        x = self.origin[0] + self.step_vector[0] * xindex
+        y = self.origin[1] + self.step_vector[1] * yindex
+        z = self.origin[2] + self.step_vector[2] * zindex
+
+        return np.array([x, y, z])
+
+    def global_index_to_node_index(self, global_index):
+        """
+        Convert from global indexes to xi,yi,zi
+
+        Parameters
+        ----------
+        global_index
+
+        Returns
+        -------
+
+        """
+        # determine the ijk indices for the global index.
+        # remainder when dividing by nx = i
+        # remained when dividing modulus of nx by ny is j
+        x_index = global_index % self.nsteps[0, None]
+        y_index = global_index // self.nsteps[0, None] % \
+                  self.nsteps[1, None]
+        z_index = global_index // self.nsteps[0, None] // \
+                  self.nsteps[1, None]
+        return x_index, y_index, z_index
+
+    def global_index_to_cell_index(self, global_index):
+        """
+        Convert from global indexes to xi,yi,zi
+
+        Parameters
+        ----------
+        global_index
+
+        Returns
+        -------
+
+        """
+        # determine the ijk indices for the global index.
+        # remainder when dividing by nx = i
+        # remained when dividing modulus of nx by ny is j
+
+        x_index = global_index % self.nsteps_cells[0, None]
+        y_index = global_index // self.nsteps_cells[0, None] % \
+                  self.nsteps_cells[1, None]
+        z_index = global_index // self.nsteps_cells[0, None] // \
+                  self.nsteps_cells[1, None]
+        return x_index, y_index, z_index
+
+    def get_neighbours(self):
+        """
+        This function goes through all of the elements in the mesh and assembles a numpy array
+        with the neighbours for each element
+
+        Returns
+        -------
+
+        """
+        # elements = self.get_elements()
+        # neighbours = np.zeros((self.ntetra,4)).astype('int64')
+        # neighbours[:] = -1
+        # tetra_neighbours(elements,neighbours)
+        # return neighbours
+        tetra_index = np.arange(0, self.ntetra)
+        neighbours = np.zeros((self.ntetra, 4)).astype('int64')
+        neighbours[:] = -9999
+        neighbours[tetra_index%5 == 0,:] = tetra_index[tetra_index%5 == 0,None]  \
+                                           + np.arange(1,5)[None,:] # first tetra is the centre one so all of its neighbours are in the same cell
+        neighbours[tetra_index % 5 != 0, 0] = np.tile(
+            tetra_index[tetra_index % 5 == 0], (4, 1)).flatten(
+            order='F')  # add first tetra to other neighbours
+
+        # now create masks for the different tetra indexes
+        one_mask = tetra_index % 5 == 1
+        two_mask = tetra_index % 5 == 2
+        three_mask = tetra_index % 5 == 3
+        four_mask = tetra_index % 5 == 4
+
+        # create masks for whether cell is odd or even
+        odd_mask = np.sum(self.global_index_to_cell_index(tetra_index // 5),
+                          axis=0) % 2 == 1
+        odd_mask = ~odd_mask.astype(bool)
+
+        # apply masks to
+        masks = []
+        masks.append([np.logical_and(one_mask, odd_mask),
+                      np.array([[-1, 0, 0, 1], [0, 1, 0, 2], [0, 0, 1, 3]])])
+        masks.append([np.logical_and(two_mask, odd_mask),
+                      np.array([[1, 0, 0, 2], [0, -1, 0, 1], [0, 0, 1, 4]])])
+        masks.append([np.logical_and(three_mask, odd_mask),
+                      np.array([[-1, 0, 0, 4], [0, -1, 0, 3], [0, 0, -1, 2]])])
+        masks.append([np.logical_and(four_mask, odd_mask),
+                      np.array([[1, 0, 0, 3], [0, 1, 0, 4], [0, 0, -1, 1]])])
+
+        masks.append([np.logical_and(one_mask, ~odd_mask),
+                      np.array([[1, 0, 0, 1], [0, 1, 0, 2], [0, 0, 1, 4]])])
+        masks.append([np.logical_and(two_mask, ~odd_mask),
+                      np.array([[-1, 0, 0, 2], [0, -1, 0, 1], [0, 0, 1, 3]])])
+        masks.append([np.logical_and(three_mask, ~odd_mask),
+                      np.array([[-1, 0, 0, 4], [0, 1, 0, 3], [0, 0, -1, 1]])])
+        masks.append([np.logical_and(four_mask, ~odd_mask),
+                      np.array([[1, 0, 0, 3], [0, -1, 0, 4], [0, 0, -1, 2]])])
+
+        for m in masks:
+            logic = m[0]
+            mask = m[1]
+            c_xi, c_yi, c_zi = self.global_index_to_cell_index(
+                tetra_index[logic] // 5)
+            # mask = np.array([[1,0,0,4],[0,0,-1,2],[0,1,0,3],[0,0,0,0]])
+            neigh_cell = np.zeros((c_xi.shape[0], 3, 3)).astype(int)
+            neigh_cell[:, :, 0] = c_xi[:, None] + mask[:, 0]
+            neigh_cell[:, :, 1] = c_yi[:, None] + mask[:, 1]
+            neigh_cell[:, :, 2] = c_zi[:, None] + mask[:, 2]
+            inside = neigh_cell[:, :, 0] >= 0
+            inside = np.logical_and(inside, neigh_cell[:, :, 1] >= 0)
+            inside = np.logical_and(inside, neigh_cell[:, :, 2] >= 0)
+            inside = np.logical_and(inside,
+                                    neigh_cell[:, :, 0] < self.n_cell_x)
+            inside = np.logical_and(inside,
+                                    neigh_cell[:, :, 1] < self.n_cell_y)
+            inside = np.logical_and(inside,
+                                    neigh_cell[:, :, 2] < self.n_cell_z)
+
+            global_neighbour_idx = np.zeros((c_xi.shape[0], 4)).astype(int)
+            global_neighbour_idx[:] = -1
+            global_neighbour_idx = (neigh_cell[:, :, 0] + neigh_cell[:, :, 1] *
+                                    self.n_cell_x + neigh_cell[:, :, 2] *
+                                    self.n_cell_x * self.n_cell_y) * 5 + mask[:, 3]
+
+            global_neighbour_idx[~inside] = -1
+            neighbours[logic, 1:] = global_neighbour_idx
+
+        return neighbours
+
+
+