import os import h5py import numpy as np import torch import h5py class FsiDataReader(): def __init__(self, location, mu, in_lets_x1=None, in_lets_x2=None,): ''' Data set of fluid solid interaction simulations. At each time step t, the simulataion records 5 variables: velocity_t = [vx_t, vy_t]: velocity in x and y direction, P_t: pressure, displacment_t = [dx_t, dy_t]: displacement in x and y direction. The inital mesh is loaded as self.input_mesh. The mesh is a 2D mesh with 2 columns. The first column is the x coordinate and the second column is the y coordinate. The mesh is time dependent i.e., the mesh changes with time. The mesh at time t is given by mesh_t = self.input_mesh + displacement_t. Parameters ---------- location : str path to the directory containing the data mu : list, optional list mu vlues. The siumulations corresponding to the mu values will be loaded. The values should be one of ['0.1', '0.01', '0.5', '5', '1.0', '10.0'] and slould exactly match the string values given here. The mu='0.5' should not be loaded separately. in_lets_x1, : list, optional list of x1 parameter controlling the inlet boundary condition of the simulation. The values should be one of ['-4.0', '-2.0', '0.0', '2.0', '4.0', '6.0'] and slould exactly match the string values given here. default is None, which loads all the values. in_lets_x2 : list, optional list of x2 parameter controlling the inlet boundary condition of the simulation. The values should be one of ['-4.0', '-2.0', '0.0', '2.0', '4.0', '6.0'] and slould exactly match the string values given here. default is None, which loads all the values. ''' self.location = location self._x1 = ['-4.0', '-2.0', '0.0', '2.0', '4.0', '6.0'] self._x2 = ['-4.0', '-2.0', '0', '2.0', '4.0', '6.0'] self._mu = ['0.5', '5', '1.0', '10.0'] # keeping vx, xy, P, dx,dy self.varable_idices = [0, 1, 3, 4, 5] if mu is not None: # check if mu is _mu else raise error assert set(mu).issubset(set(self._mu)) self._mu = mu if in_lets_x1 is not None: # check if in_lets_x1 is _x1 else raise error assert set(in_lets_x1).issubset(set(self._x1)) self._x1 = in_lets_x1 if in_lets_x2 is not None: # check if in_lets_x2 is _x2 else raise error assert set(in_lets_x2).issubset(set(self._x2)) self._x2 = in_lets_x2 # assert _mu = 0.5 should not be mixed with other mu values assert not('0.5' in self._mu and len(self._mu) > 1), "mu=0.5 should not be mixed with other mu values" self.load_mesh(location) def load_mesh(self, location): if '0.5' in self._mu: x_path = os.path.join(location, 'mu=0.5', 'coord_x.txt') y_path = os.path.join(location, 'mu=0.5', 'coord_y.txt') mesh_x = np.loadtxt(x_path) mesh_y = np.loadtxt(y_path) # create mesh from mesh_x and mesh_y mesh = np.zeros((mesh_x.shape[0], 2)) mesh[:, 0] = mesh_x mesh[:, 1] = mesh_y self.input_mesh = torch.from_numpy(mesh).type(torch.float) else: mesh_h = h5py.File(os.path.join(location, 'mesh.h5'), 'r') mesh = mesh_h['mesh/coordinates'][:] self.input_mesh = torch.from_numpy(mesh).type(torch.float) def _readh5(self, h5f, dtype=torch.float32): a_dset_keys = list(h5f['VisualisationVector'].keys()) size = len(a_dset_keys) readings = [None for i in range(size)] for dset in a_dset_keys: ds_data = (h5f['VisualisationVector'][dset]) if ds_data.dtype == 'float64': csvfmt = '%.18e' elif ds_data.dtype == 'int64': csvfmt = '%.10d' else: csvfmt = '%s' readings[int(dset)] = torch.tensor(np.array(ds_data), dtype=dtype) readings_tensor = torch.stack(readings, dim=0) print(f"Loaded tensor Size: {readings_tensor.shape}") return readings_tensor def get_data(self, mu, x1, x2): if mu not in self._mu: raise ValueError(f"Value of mu must be one of {self._mu}") if x1 not in self._x1 or x2 not in self._x2: raise ValueError( f"Value of is must be one of {self._ivals3} and {self._ivals12} ") path = os.path.join( self.location, 'mu='+str(mu), 'x1='+str(x1), 'x2='+str(x2), 'Visualization') filename = os.path.join(path, 'displacement.h5') h5f = h5py.File(filename, 'r') displacements_tensor = self._readh5(h5f) filename = os.path.join(path, 'pressure.h5') h5f = h5py.File(filename, 'r') pressure_tensor = self._readh5(h5f) filename = os.path.join(path, 'velocity.h5') h5f = h5py.File(filename, 'r') velocity_tensor = self._readh5(h5f) combined = torch.cat([velocity_tensor, pressure_tensor, displacements_tensor], dim=-1)[..., self.varable_idices] # return velocity_tensor, pressure_tensor, displacements_tensor return combined def get_data_txt(self, mu, x1, x2): if mu not in self._mu: raise ValueError(f"Value of mu must be one of {self._mu}") if x1 not in self._x1 or x2 not in self._x2: raise ValueError( f"Value of is must be one of {self._ivals3} and {self._ivals12} ") path = os.path.join( self.location, 'mu='+str(mu), 'x1='+str(x1), 'x2='+str(x2), '1') #try: dis_x = torch.tensor(np.loadtxt(os.path.join(path, 'dis_x.txt'))) dis_y = torch.tensor(np.loadtxt(os.path.join(path, 'dis_y.txt'))) pressure = torch.tensor(np.loadtxt(os.path.join(path, 'pres.txt'))) velocity_x = torch.tensor(np.loadtxt(os.path.join(path, 'vel_x.txt'))) velocity_y = torch.tensor(np.loadtxt(os.path.join(path, 'vel_y.txt'))) # reshape each tensor into 2d by keeping 876 entries in each row dis_x = dis_x.view(-1, 876,1) dis_y = dis_y.view(-1, 876,1) pressure = pressure.view(-1, 876,1) velocity_x = velocity_x.view(-1, 876,1) velocity_y = velocity_y.view(-1, 876,1) combined = torch.cat([velocity_x, velocity_y, pressure, dis_x, dis_y], dim=-1)[..., ] return combined def get_loader(self, batch_size, shuffle=True): data_t0 = [] data_t1 = [] for mu in self._mu: for x1 in self._x1: for x2 in self._x2: try: if mu == '0.5': mu_data = self.get_data_txt(mu, x1, x2) else: mu_data = self.get_data(mu, x1, x2) mu_data_t0 = mu_data[:-1,:,:] mu_data_t1 = mu_data[1:,:,:] data_t0.append(mu_data_t0) data_t1.append(mu_data_t1) except FileNotFoundError as e: print( f"file not found for mu={mu}, x1={x1}, x2={x2}") continue data_t0 = torch.cat(data_t0, dim=0) data_t1 = torch.cat(data_t1, dim=0) tensor_dataset = torch.utils.data.TensorDataset(data_t0, data_t1) data_loader = torch.utils.data.DataLoader(tensor_dataset, batch_size=batch_size, shuffle=shuffle) return data_loader