File size: 9,302 Bytes
c902ee7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
"""An interface-aligned impedance step with an exact acoustic modal solution.

The finite analytic modal initial field is fixed in physical coordinates,
independently of the finite-volume grid.
"""
from functools import lru_cache
from pathlib import Path
import json
import numpy as np
from scipy import sparse, linalg, special
from numpy.polynomial.legendre import leggauss

HERE = Path(__file__).resolve().parent
CFG = json.loads((HERE / "config.json").read_text())


def _gauss_interval(order, a, b):
    x, w = leggauss(order)
    return (a + b) / 2 + (b - a) * x / 2, w * (b - a) / 2


def x_modes(x, maximum=None):
    """Beta-orthonormal x modes; beta=1 on the left and 1/8 on the right."""
    maximum = CFG["max_x_mode"] if maximum is None else maximum
    m = np.arange(maximum + 1)
    norm2 = np.where(m == 0, 9 / 16, np.where(m % 2 == 0, 9 / 32, 9 / 4))
    scale = np.where((np.asarray(x)[:, None] > .5) & (m[None, :] % 2 == 1), 8., 1.)
    return scale * np.cos(np.pi * np.asarray(x)[:, None] * m[None, :]) / np.sqrt(norm2)


def y_modes(y, maximum=None):
    maximum = CFG["max_y_mode"] if maximum is None else maximum
    modes = np.arange(maximum + 1)
    return np.cos(np.pi * np.asarray(y)[:, None] * modes[None, :]) * np.where(modes == 0, 1., np.sqrt(2.))


def gaussian_integrals(edges, centre):
    z = (np.asarray(edges) - centre) / (np.sqrt(2.) * CFG["sigma"])
    return np.sqrt(np.pi / 2) * CFG["sigma"] * np.diff(special.erf(z))


@lru_cache(maxsize=4)
def analytic_data(order=None):
    order = CFG["quadrature_order"] if order is None else order
    xl, wl = _gauss_interval(order, 0., .5)
    xr, wr = _gauss_interval(order, .5, 1.)
    x = np.r_[xl, xr]
    wx = np.r_[wl, wr]
    beta = np.r_[np.ones(order), np.full(order, 1 / 8)]
    y, wy = _gauss_interval(order, 0., 1.)
    X = x_modes(x)
    Y = y_modes(y)
    coefficients = []
    for cx, cy in CFG["source_centres"]:
        gx = np.exp(-.5 * ((x - cx) / CFG["sigma"]) ** 2)
        gy = np.exp(-.5 * ((y - cy) / CFG["sigma"]) ** 2)
        aa = np.outer(X.T @ (wx * beta * gx), Y.T @ (wy * gy))
        aa[0, 0] = 0.
        aa *= np.sqrt(2 * CFG["initial_energy"]) / np.linalg.norm(aa)
        coefficients.append(aa)
    coefficients = np.stack(coefficients, axis=-1)
    receivers = []
    for cx, cy in CFG["receiver_centres"]:
        gx = np.exp(-.5 * ((x - cx) / CFG["sigma"]) ** 2)
        gy = np.exp(-.5 * ((y - cy) / CFG["sigma"]) ** 2)
        nx = gaussian_integrals([0., 1.], cx)[0]
        ny = gaussian_integrals([0., 1.], cy)[0]
        receivers.append(np.outer(X.T @ (wx * gx) / nx, Y.T @ (wy * gy) / ny))
    m = np.arange(CFG["max_x_mode"] + 1)
    ll = np.arange(CFG["max_y_mode"] + 1)
    omega = np.pi * np.sqrt(m[:, None] ** 2 + ll[None, :] ** 2)
    return {"coefficients": coefficients, "receiver_modes": np.stack(receivers, axis=-1),
            "omega": omega, "x_gram": X.T @ ((wx * beta)[:, None] * X),
            "y_gram": Y.T @ (wy[:, None] * Y)}


def averaged_modes(n):
    if n % 2:
        raise ValueError("An even grid is required for interface alignment")
    centres = (np.arange(n) + .5) / n
    m = np.arange(CFG["max_x_mode"] + 1)
    ll = np.arange(CFG["max_y_mode"] + 1)
    return (x_modes(centres) * np.sinc(m / (2 * n))[None, :],
            y_modes(centres) * np.sinc(ll / (2 * n))[None, :])


def exact_receiver(times, order=None):
    data = analytic_data(order)
    modal = np.cos(np.asarray(times)[:, None, None] * data["omega"][None, :, :])
    return np.einsum("tml,mls,mlr->trs", modal, data["coefficients"],
                     data["receiver_modes"], optimize=True)


def exact_pressure_norm_squared(times):
    """Continuous integral of beta*p(t)^2, separately for the three sources."""
    data = analytic_data()
    phase = np.cos(np.asarray(times)[:, None, None] * data["omega"][None, :, :])
    return np.einsum("tml,mls->ts", phase ** 2, data["coefficients"] ** 2, optimize=True)


def cell_projection_norm_squared(n, times):
    """Norm of the exact cell averages (n exceeds the largest modal index)."""
    if n <= max(CFG["max_x_mode"], CFG["max_y_mode"]):
        return np.sum(exact_mass_weighted_pressure(n, times) ** 2, axis=1)
    data = analytic_data()
    m = np.arange(CFG["max_x_mode"] + 1)
    ll = np.arange(CFG["max_y_mode"] + 1)
    filter2 = np.sinc(m / (2 * n))[:, None] ** 2 * np.sinc(ll / (2 * n))[None, :] ** 2
    phase = np.cos(np.asarray(times)[:, None, None] * data["omega"][None, :, :])
    return np.einsum("tml,mls,ml->ts", phase ** 2, data["coefficients"] ** 2, filter2, optimize=True)


def exact_mass_weighted_pressure(n, times):
    data = analytic_data()
    X, Y = averaged_modes(n)
    modal = np.cos(np.asarray(times)[:, None, None] * data["omega"][None, :, :])[:, :, :, None]
    modal = modal * data["coefficients"][None, :, :, :]
    pressure = np.einsum("im,tmls,jl->tijs", X, modal, Y, optimize=True).reshape(len(times), n * n, 3)
    mass = np.repeat(np.where((np.arange(n) + .5) / n < .5, 1., 1 / 8), n) / n ** 2
    return pressure * np.sqrt(mass)[None, :, None]


def exact_solution(n, times):
    return {"time": np.asarray(times), "mass_weighted_pressure": exact_mass_weighted_pressure(n, times),
            "receiver_pressure": exact_receiver(times)}


def build_sparse(n):
    if n % 2:
        raise ValueError("An even grid is required for interface alignment")
    x = (np.arange(n) + .5) / n
    X, Y = np.meshgrid(x, x, indexing="ij")
    rho = np.where(X < .5, 1., 8.)
    bulk = rho.copy()
    index = np.arange(n * n).reshape(n, n)
    a = np.r_[index[:-1, :].ravel(), index[:, :-1].ravel()]
    b = np.r_[index[1:, :].ravel(), index[:, 1:].ravel()]
    faces = np.arange(len(a))
    face_rho = .5 * (rho.ravel()[a] + rho.ravel()[b])
    incidence = sparse.coo_matrix((np.r_[np.ones(len(a)), -np.ones(len(a))],
                                  (np.r_[a, b], np.r_[faces, faces])), shape=(n * n, len(a))).tocsr()
    mass = (1 / (n * n * bulk)).ravel()
    C = sparse.diags(1 / np.sqrt(mass)) @ incidence @ sparse.diags(1 / np.sqrt(face_rho))
    J = sparse.bmat([[None, -C], [C.T, None]], format="csr")
    return X, Y, rho, bulk, mass, C, J


def source_receiver(n, mass):
    B = exact_mass_weighted_pressure(n, [0.])[0]
    edges = np.linspace(0., 1., n + 1)
    windows = []
    for cx, cy in CFG["receiver_centres"]:
        gx = gaussian_integrals(edges, cx)
        gy = gaussian_integrals(edges, cy)
        weights = np.outer(gx / gx.sum(), gy / gy.sum()).ravel()
        windows.append(weights / np.sqrt(mass))
    return B, np.stack(windows, axis=-1)


def assembly(n, need_stiffness=True):
    X, Y, rho, bulk, mass, C, J = build_sparse(n)
    B, R = source_receiver(n, mass)
    A = (C @ C.T).tocsr() if need_stiffness else None
    return mass, A, J, B, R, rho, bulk


def separable_setup(n):
    """Complete n^2 eigensystem from two n-dimensional tridiagonal problems.
    
    Setup includes the physical initial field and receiver construction.
    """
    xx = (np.arange(n) + .5) / n
    bulk_x = np.where(xx < .5, 1., 8.)
    face_rho = .5 * (bulk_x[:-1] + bulk_x[1:])
    inv_face = 1 / face_rho
    dx = n * n * bulk_x * np.r_[inv_face[0], inv_face[:-1] + inv_face[1:], inv_face[-1]]
    ex = -n * n * np.sqrt(bulk_x[:-1] * bulk_x[1:]) * inv_face
    dy = np.r_[1., np.full(n - 2, 2.), 1.] * n * n
    ey = np.full(n - 1, -n * n, dtype=float)
    lx, Vx = linalg.eigh_tridiagonal(dx, ex)
    ly, Vy = linalg.eigh_tridiagonal(dy, ey)
    lx[np.abs(lx) < 1e-8] = 0.
    ly[np.abs(ly) < 1e-8] = 0.
    mass = np.repeat(1 / bulk_x, n) / n ** 2
    B, R = source_receiver(n, mass)
    source_modes = np.einsum("im,ijs,jl->mls", Vx, B.reshape(n, n, 3), Vy, optimize=True)
    receiver_modes = np.einsum("im,ijr,jl->mlr", Vx, R.reshape(n, n, 4), Vy, optimize=True)
    return {"x_eigenvalues": lx, "y_eigenvalues": ly, "x_eigenvectors": Vx,
            "y_eigenvectors": Vy, "initial_modal_coefficients": source_modes,
            "receiver_modes": receiver_modes, "mass": mass,
            "omega": np.sqrt(lx[:, None] + ly[None, :])}


def separable_receiver(model, times):
    n = len(model["x_eigenvalues"])
    weights = np.einsum("mlr,mls->mlrs", model["receiver_modes"], model["initial_modal_coefficients"]).reshape(n * n, 12)
    result = np.empty((len(times), 4, 3))
    for start in range(0, len(times), 32):
        tt = np.asarray(times)[start:start + 32]
        phase = np.cos(tt[:, None] * model["omega"].ravel()[None, :])
        result[start:start + len(tt)] = (phase @ weights).reshape(len(tt), 4, 3)
    return result


def separable_pressure(model, times):
    n = len(model["x_eigenvalues"])
    modal = np.cos(np.asarray(times)[:, None, None] * model["omega"][None, :, :])[:, :, :, None]
    modal = modal * model["initial_modal_coefficients"][None, :, :, :]
    return np.einsum("im,tmls,jl->tijs", model["x_eigenvectors"], modal,
                     model["y_eigenvectors"], optimize=True).reshape(len(times), n * n, 3)


def trace_only(n, times):
    return separable_receiver(separable_setup(n), times)


def separable_full(n, times):
    model = separable_setup(n)
    return {"time": np.asarray(times), "mass_weighted_pressure": separable_pressure(model, times),
            "receiver_pressure": separable_receiver(model, times), **model}