seistorch/examples/sthelse/utils.py at dev · StarrMoonn/seistorch · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
import torch, tqdm
import numpy as np
import matplotlib.pyplot as plt

def imshow(data, vmin=None, vmax=None, cmap=None, figsize=(10, 10), savepath=None):
    plt.figure(figsize=figsize)
    plt.imshow(data, vmin=vmin, vmax=vmax, cmap=cmap, aspect="auto")
    plt.colorbar()
    if savepath:
        plt.savefig(savepath)
    plt.show()

def generate_mesh(mshape, dh=1):
    '''Generates a flattened grid of (x,y,...) coordinates in a range of -1 to 1.
    sidelen: int
    dim: int'''
    tensors_for_meshgrid = []
    for size in mshape:
        tensors_for_meshgrid.append(torch.linspace(-1, 1, steps=size))
        # tensors_for_meshgrid.append(torch.linspace(0, size*dh/1000, steps=size))
    mgrid = torch.stack(torch.meshgrid(*tensors_for_meshgrid, indexing='ij'), dim=-1)
    mgrid = mgrid.reshape(-1, len(mshape))
    return mgrid

def show_gathers(rec, size=3, figsize=(8, 5)):
    randno = np.random.randint(0, rec.shape[0], size=size)
    fig,axes=plt.subplots(1, randno.shape[0], figsize=figsize)
    for i, ax in enumerate(axes):
        vmin,vmax=np.percentile(rec[i], [1, 99])
        kwargs=dict(vmin=vmin, vmax=vmax, cmap="seismic", aspect="auto")
        ax.imshow(rec[randno[i]], **kwargs)
        ax.set_title(f"shot {randno[i]}")
    plt.tight_layout()
    plt.show()

def showgeom(vel, src_loc, rec_loc, figsize=(10, 10)):
    plt.figure(figsize=figsize)
    plt.imshow(vel, vmin=vel.min(), vmax=vel.max(), cmap="seismic", aspect="auto")
    plt.colorbar()
    plt.scatter(*zip(*src_loc), c="r", marker="v", s=100, label="src")
    plt.scatter(*zip(*rec_loc), c="b", marker="^", s=10, label="rec")
    plt.legend()
    plt.show()

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# configure
kernel = torch.tensor([[[[0.0, 1.0, 0.0], [1.0, -4.0, 1.0], [0.0, 1.0, 0.0]]]]).to(device)

def ricker(t, f=10):
    r = (1 - 2 * (np.pi * f * t) ** 2) * np.exp(-(np.pi * f * t) ** 2)
    return torch.from_numpy(r).float().to(device)

def laplace(u, h):
    return torch.nn.functional.conv2d(u, kernel, padding=1) / (h ** 2)

def step(u_pre, u_now, c=1.5, dt=0.001, h=10./1000., b=None):

    # With boundary condition
    u_next = torch.mul((dt**-2 + b * dt**-1).pow(-1),
                (2 / dt**2 * u_now - torch.mul((dt**-2 - b * dt**-1), u_pre)
                + torch.mul(c.pow(2), laplace(u_now, h)))
                )
    # u_next = ((dt**-2 + b * dt**-1).pow(-1) *
    #             (2 / dt**2 * u_now - (dt**-2 - b * dt**-1) * u_pre
    #             + c.pow(2) * laplace(u_now, h)))
    # Without boundary condition
    # u_next = 2 * u_now - u_pre + (c * dt) ** 2 * laplace(u_now, h)
    return u_next

def forward(wave, c, b, src_list, domain, dt, h, dev, recz=0, pmln=50):
    nt = wave.shape[0]
    nz, nx = domain[0], domain[1]
    nshots = len(src_list)
    u_pre = torch.zeros(nshots, 1, *domain, device=dev)
    u_now = torch.zeros(nshots, 1, *domain, device=dev)
    rec = torch.zeros(nshots, nt, nx-2*pmln, device=dev)
    b = b.unsqueeze(0).to(dev)
    c = c.unsqueeze(0)
    shots = torch.arange(nshots, device=dev)
    srcx, srcz = zip(*src_list)

    h = torch.Tensor([h]).to(dev)
    dt = torch.Tensor([dt]).to(dev)

    source_mask = torch.zeros_like(u_now)
    source_mask[shots, :, srcz, srcx] = 1

    for it in range(nt):
        u_now += source_mask * wave[it]
        u_next = step(u_pre, u_now, c, dt, h, b)
        u_pre, u_now = u_now, u_next
        rec[:,it, :] = u_now[:, 0, recz, pmln:-pmln]

    return rec

def generate_pml_coefficients_2d(domain_shape, N=50, B=100., multiple=False):
    Nx, Ny = domain_shape

    R = 10**(-((np.log10(N)-1)/np.log10(2))-3)
    #d0 = -(order+1)*cp/(2*abs_N)*np.log(R) # Origin
    R = 1e-6; order = 2; cp = 1000.
    d0 = (1.5*cp/N)*np.log10(R**-1)
    d_vals = d0 * torch.linspace(0.0, 1.0, N + 1) ** order
    d_vals = torch.flip(d_vals, [0])

    d_x = torch.zeros(Ny, Nx)
    d_y = torch.zeros(Ny, Nx)

    if N > 0:
        d_x[0:N + 1, :] = d_vals.repeat(Nx, 1).transpose(0, 1)
        d_x[(Ny - N - 1):Ny, :] = torch.flip(d_vals, [0]).repeat(Nx, 1).transpose(0, 1)
        if not multiple:
            d_y[:, 0:N + 1] = d_vals.repeat(Ny, 1)
        d_y[:, (Nx - N - 1):Nx] = torch.flip(d_vals, [0]).repeat(Ny, 1)

    _d = torch.sqrt(d_x ** 2 + d_y ** 2).transpose(0, 1)
    _d = _corners(domain_shape, N, _d, d_x.T, d_y.T, multiple)

    return _d

def _corners(domain_shape, abs_N, d, dx, dy, multiple=False):
    Nx, Ny = domain_shape
    for j in range(Ny):
        for i in range(Nx):
            # Left-Top
            if not multiple:
                if i < abs_N+1 and j< abs_N+1:
                    if i < j: d[i,j] = dy[i,j]
                    else: d[i,j] = dx[i,j]
            # Left-Bottom
            if i > (Nx-abs_N-2) and j < abs_N+1:
                if i + j < Nx: d[i,j] = dx[i,j]
                else: d[i,j] = dy[i,j]
            # Right-Bottom
            if i > (Nx-abs_N-2) and j > (Ny-abs_N-2):
                if i - j > Nx-Ny: d[i,j] = dy[i,j]
                else: d[i,j] = dx[i,j]
            # Right-Top
            if not multiple:
                if i < abs_N+1 and j> (Ny-abs_N-2):
                    if i + j < Ny: d[i,j] = dy[i,j]
                    else: d[i,j] = dx[i,j]

    return d