def test_6(self):
# Two half spaces elastic model
vp1, vp2 = 5.72, 2.87
vs1, vs2 = 2.93, 1.61
ro1, ro2 = 2.86, 2.14
vs1_a = 1.8
vs1_b = 3.5
dx = 0.4
nx = int((vs1_b - vs1_a) / dx)
angle = 30.
mode, rid = 'PS', 2
for i in range(nx):
vs1 = vs1_a + (i + 1) * dx
r1, r2, r3, r4 = elapar_hs2ratio(vp1, vs1, ro1, vp2, vs2, ro2)
self.cmp2m(r1, r2, r3, r4, angle, mode, rid)
def test_1(self):
# Two half spaces elastic model
vp1, vp2 = 3.0, 2.0
vs1, vs2 = 1.5, 1.0
ro1, ro2 = 2.3, 2.0
vp2_a = 1.2
vp2_b = 2.5
dx = 0.1
nx = int((vp2_b - vp2_a) / dx)
angle = 20.
mode, rid = 'PP', 1
for i in range(nx):
vp2 = vp2_a + (i + 1) * dx
r1, r2, r3, r4 = elapar_hs2ratio(vp1, vs1, ro1, vp2, vs2, ro2)
self.cmp2m(r1, r2, r3, r4, angle, mode, rid)
def test_8(self):
# Two half spaces elastic model
vp1, vp2 = 5.72, 2.87
vs1, vs2 = 2.93, 1.61
ro1, ro2 = 2.86, 2.14
ro2_a = 0.9
ro2_b = 2.8
dx = 0.5
nx = int((ro2_b - ro2_a) / dx)
angle = 30.
mode, rid = 'PS', 4
for i in range(nx):
ro2 = ro2_a + (i + 1) * dx
r1, r2, r3, r4 = elapar_hs2ratio(vp1, vs1, ro1, vp2, vs2, ro2)
self.cmp2m(r1, r2, r3, r4, angle, mode, rid)
def test_pp_clean(self):
# Two half spaces elastic model
# vp1, vp2 = 3.0, 2.0
# vs1, vs2 = 1.5, 1.0
# ro1, ro2 = 2.3, 2.0
vp1, vp2 = 4.0, 2.0
vs1, vs2 = 2.0, 1.0
ro1, ro2 = 2.4, 2.0
# Change parameterization
r1, r2, r3, r4 = elapar_hs2ratio(vp1, vs1, ro1, vp2, vs2, ro2)
# Define angles
angles = np.arange(1, 60, 6)
# Calculate the reflection amplitude or b in Ax=b
m = len(angles)
rpp = np.zeros(m)
rps = np.zeros(m)
for i in range(m):
angle = angles[i]
amp, pha = rpp_cer1977(r1, r2, r3, r4, angle)
rpp[i] = amp
amp, pha = rps_cer1977(r1, r2, r3, r4, angle, amp_type='abs')
rps[i] = amp
# print("Target model:", r1, r2, r3, r4)
r1_ini = 2.4 / 4.0
r2_ini = 2.2 / 4.0
r3_ini = 1.3 / 4.0
r4_ini = 1.6 / 2.4
x_ini = (r1_ini, r2_ini, r3_ini, r4_ini)
# print("Initial model:", x_ini)
for i in range(5):
x_new = cer1itr(angles, rpp, x_ini, rps=rps)
# print("Updated", x_new)
x_ini = x_new
self.assertLessEqual(r1 - x_new[0], 0.001)
self.assertLessEqual(r2 - x_new[1], 0.001)
self.assertLessEqual(r3 - x_new[2], 0.001)
self.assertLessEqual(r4 - x_new[3], 0.001)
def test_2(self):
# Two half spaces elastic model
vp1, vp2 = 5.72, 2.87
vs1, vs2 = 2.93, 1.61
ro1, ro2 = 2.86, 2.14
# Change parameterization
r1, r2, r3, r4 = elapar_hs2ratio(vp1, vs1, ro1, vp2, vs2, ro2)
# Define angles
# angles = np.arange(0, 90, 1)
# for angle in angles:
# amp, pha = rps_cer1977(r1, r2, r3, r4, angle, amp_type='abs')
# amp, pha = rps_cer1977(r1, r2, r3, r4, angle, amp_type='real')
# print("ang,amp,pha =", angle, amp, pha)
angle = 10
amp, pha = rps_cer1977(r1, r2, r3, r4, angle, amp_type='real')
amp_truth = -0.14823324
err = amp_truth - amp
self.assertLessEqual(err, 0.001)
def test_1(self):
# Two half spaces elastic model
vp1, vp2 = 3.0, 2.0
vs1, vs2 = 1.5, 1.0
ro1, ro2 = 2.3, 2.0
# vp1, vp2 = 2.0, 4.0
# vs1, vs2 = 0.88, 1.54
# ro1, ro2 = 2.0, 2.3
# Change parameterization
r1, r2, r3, r4 = elapar_hs2ratio(vp1, vs1, ro1, vp2, vs2, ro2)
# Define angles
# angles = np.arange(0, 90, 1)
# for angle in angles:
# amp, pha = rpp_cer1977(r1, r2, r3, r4, angle, amp_type='abs')
# amp, pha = rpp_cer1977(r1, r2, r3, r4, angle, amp_type='real')
# print("ang,amp,pha =", angle, amp, pha)
angle = 0
amp, pha = rpp_cer1977(r1, r2, r3, r4, angle, amp_type='real')
amp_truth = -0.266055
err = amp_truth - amp
self.assertLessEqual(err, 0.001)
def modeling(model, inc_angles, equation, reflection):
"""
Unified API for GUI call.
Parameters
----------
model : tuple
Two half-space elastic model (vp1, vs1, ro1, vp2, vs2, ro2)
inc_angles : str
Incident angles in degrees, either comma separated values, or
1-60(2) means from 1 to 60 with step 2.
equation : str
modeling equation, 'linear', 'quadratic', 'zoeppritz'
reflection : str
reflection type, 'PP', 'PS'
Returns
-------
rc : array
amplitude and phase of the reflection coefficients at the angles.
The array shape is mx3 of columns: incident angle, amplitude, phase.
"""
# Change parameterization
vp1, vs1, ro1, vp2, vs2, ro2 = model
ro_rd, vp_rd, vs_rd, vs_vp_ratio = \
elapar_hs2delta(vp1, vs1, ro1, vp2, vs2, ro2)
r1, r2, r3, r4 = elapar_hs2ratio(vp1, vs1, ro1, vp2, vs2, ro2)
if '-' in inc_angles:
a, b = inc_angles.split('-')
c, d = b.split('(')
e, f = d.split(')')
a1 = float(a)
a2 = float(c)
ad = float(e)
angles = np.arange(a1, a2, ad)
else:
angles = np.array([float(a) for a in inc_angles.split(',')])
ave_angles = inc2ave_angle(angles, vp_rd)
m = len(angles)
a, p = np.zeros(m), np.zeros(m)
if reflection == 'PP':
if equation == 'linear':
a = aki1980(vs_vp_ratio, ro_rd, vp_rd, vs_rd, ave_angles)
return np.vstack((angles, a, p)).T # mx3 array
elif equation == 'quadratic':
a = wang1999(vs_vp_ratio, ro_rd, vp_rd, vs_rd, ave_angles)
return np.vstack((angles, a, p)).T # mx3 array
elif equation == 'zoeppritz':
for i in range(m):
angle = angles[i]
amp, pha = rpp_cer1977(r1, r2, r3, r4, angle)
a[i], p[i] = amp, pha
return np.vstack((angles, a, p)).T # mx3 array
else:
raise NotImplementedError
elif reflection == 'PS':
if equation == 'linear':
raise NotImplementedError
elif equation == 'quadratic':
raise NotImplementedError
elif equation == 'zoeppritz':
for i in range(m):
angle = angles[i]
amp, pha = rps_cer1977(r1, r2, r3, r4, angle)
a[i], p[i] = amp, pha
return np.vstack((angles, a, p)).T # mx3 array
else:
raise NotImplementedError
else:
raise NotImplementedError
def test_pp_noise(self):
# Two half spaces elastic model
vp1, vp2 = 4.0, 2.0
vs1, vs2 = 2.0, 1.0
ro1, ro2 = 2.4, 2.0
# Change parameterization
r1, r2, r3, r4 = elapar_hs2ratio(vp1, vs1, ro1, vp2, vs2, ro2)
# Define angles
angles = np.arange(1, 60, 1)
# Calculate the reflection amplitude or b in Ax=b
m = len(angles)
rpp = np.zeros(m)
rps = np.zeros(m)
for i in range(m):
angle = angles[i]
amp, pha = rpp_cer1977(r1, r2, r3, r4, angle)
rpp[i] = amp
amp, pha = rps_cer1977(r1, r2, r3, r4, angle, amp_type='abs')
rps[i] = amp
# Add noise to data
# ramp = np.max(rpp) - np.min(rpp)
# mu, sigma = 0, 0.05 * ramp # mean and standard deviation
# noise = np.random.normal(mu, sigma, m)
# hard code noise to make test result consistent
noise = np.array([
-3.40745756e-03,
4.41326891e-03,
-1.84969329e-02,
-8.57665267e-03,
-4.64722728e-03,
1.81164323e-02,
2.35764041e-03,
8.66820650e-03,
2.61862025e-03,
5.60835320e-03,
-1.32386200e-02,
-1.13868325e-02,
-3.85411636e-03,
8.30156732e-04,
6.08364262e-03,
-1.13107829e-02,
7.51568819e-03,
8.32391400e-03,
7.18915187e-03,
2.48970883e-03,
1.42114394e-02,
2.45652884e-04,
-4.69414374e-03,
4.60964000e-03,
1.43935631e-02,
-5.88788401e-03,
3.13041871e-03,
-6.68177919e-04,
-6.20489672e-03,
-1.68069368e-04,
-1.78392131e-02,
8.38724551e-04,
1.30622636e-03,
-9.83497743e-03,
-1.17627106e-02,
-1.62056738e-02,
4.62611536e-03,
1.48628494e-02,
-1.24973356e-02,
-1.01725440e-02,
7.38562227e-03,
9.21933387e-03,
-6.69923701e-03,
6.42089408e-03,
-4.77129595e-03,
2.33900064e-03,
3.29402557e-05,
9.54770479e-04,
-1.49280387e-02,
-6.65381602e-03,
-1.58004300e-02,
-7.08064272e-03,
5.65539007e-04,
-2.76684435e-03,
-5.60120257e-03,
8.84405490e-03,
-3.24883460e-03,
5.64724034e-03,
-9.45532624e-03,
])
rpp_noisy = rpp + noise
# ramp = np.max(rps) - np.min(rps)
# mu, sigma = 0, 0.05 * ramp # mean and standard deviation
# noise = np.random.normal(mu, sigma, m)
# hard code noise to make test result consistent
noise = np.array([
-0.0309984,
0.00092359,
-0.00770345,
-0.03662312,
0.00336188,
0.00583431,
-0.02101242,
-0.0248055,
-0.00333648,
0.02492424,
-0.00099495,
0.00944948,
-0.00325943,
0.01934984,
-0.00704765,
0.01490579,
0.00779604,
0.02183828,
-0.00405295,
-0.01820525,
-0.00446887,
0.01793082,
0.03251096,
0.0026122,
0.01377384,
-0.01452418,
0.02901279,
-0.00881719,
0.02308159,
0.01260138,
-0.00522267,
0.00769085,
0.02171298,
-0.01478435,
0.01349567,
-0.00778548,
-0.01922285,
-0.01798599,
-0.02126122,
-0.00327526,
0.01550364,
0.00130878,
0.00680895,
0.02670106,
-0.05456112,
0.02081972,
0.02333233,
0.03656901,
0.01069452,
-0.01197574,
0.02639394,
0.01850353,
0.0232636,
-0.00037154,
-0.01148699,
0.03056004,
0.006255,
0.01079065,
0.02806546,
])
rps_noisy = rps + noise
# print("Target model:", r1, r2, r3, r4)
r1_ini = 2.4 / 4.0
r2_ini = 2.2 / 4.0
r3_ini = 1.3 / 4.0
r4_ini = 1.6 / 2.4
x_ini = (r1_ini, r2_ini, r3_ini, r4_ini)
# print("Initial model:", x_ini)
for i in range(10):
# x_new = cer1itr(angles, rpp_noisy, x_ini, rps=None) # fails
x_new = cer1itr(angles, rpp_noisy, x_ini, rps=rps_noisy)
# print("Updated", i, x_new)
x_ini = x_new
self.assertLessEqual(0.54935233 - x_new[0], 0.001)
self.assertLessEqual(0.48925867 - x_new[1], 0.001)
self.assertLessEqual(0.25932287 - x_new[2], 0.001)
self.assertLessEqual(0.76031868 - x_new[3], 0.001)