def get_potential2(params, syst_pars):
V_r = params["V_r"]
V_l = params["V_l"]
x0 = params["x0"]
sigma = params["sigma"]
V_bottom = lambda x, V_0: ( # noqa: E731
common.gaussian(x, V_0, x0, sigma) if x > x0 else V_0)
V_top = lambda x: ( # noqa: E731
V_r + common.gaussian(x, V_l - V_r, x0, sigma) if x > x0 else V_l)
z0 = -syst_pars["r1"]
z1 = syst_pars["r1"]
return lambda x, z, V_0: (V_bottom(x, V_0) * (z1 - z) + V_top(x) *
(z - z0)) / (z1 - z0)
def es_mutate_steps(steps, p):
steps_size = len(steps)
new_steps = []
t1 = np.sqrt(2 * steps_size) ** -1
t2 = np.sqrt(2 * np.sqrt(steps_size)) ** -1
for idx,step in enumerate(steps):
new_step = step * np.exp( (t1 * common.gaussian(0,1))
+ (t2 * common.gaussian(0,1)) )
sigma = new_step * p
if sigma > 2:
sigma = sigma % 2
sigma = 1 - sigma
new_steps.append(sigma)
return new_steps
def es_mutate_steps(steps, p):
steps_size = len(steps)
new_steps = []
t1 = np.sqrt(2 * steps_size)**-1
t2 = np.sqrt(2 * np.sqrt(steps_size))**-1
for idx, step in enumerate(steps):
new_step = step * np.exp((t1 * common.gaussian(0, 1)) +
(t2 * common.gaussian(0, 1)))
sigma = new_step * p
if sigma > 2:
sigma = sigma % 2
sigma = 1 - sigma
new_steps.append(sigma)
return new_steps
def main():
# abc = remove_duplicate_pts(common.rect_to_tri_mesh(*common.twist_mesh(10)))
pts, rects = getattr(common, which + '_mesh')(n)
pts, tris = common.rect_to_tri_mesh(pts, rects)
pts, tris = remove_duplicate_pts((pts, tris))
# pts, tris = refine((pts, tris))
# pts, tris = refine_to_size((pts, tris), 0.02)[0]
# np.save('mesh.npy', (pts, tris))
# plt.triplot(pts[:,0], pts[:,2], tris)
# plt.show()
# plt.triplot(pts[:,0], pts[:,1], tris)
# plt.show()
n_tris = tris.shape[0]
tri_pts = pts[tris].reshape((-1, 3))
dist = np.linalg.norm(tri_pts - gauss_center, axis=1)
strike_slip = common.gaussian(*gauss_params, dist)
slip = np.zeros((strike_slip.shape[0], 3))
slip[:, 0] = strike_slip
common_tectosaur.plot_tectosaur((pts, tris), slip[:, 0].reshape(-1, 3),
'inputslip', folder)
slip_to_traction = common_tectosaur.get_slip_to_traction(
(pts, tris), common_tectosaur.tectosaur_cfg)
traction = slip_to_traction(slip)
sxy = traction.reshape(-1, 3, 3)[:, :, 0]
syz = traction.reshape(-1, 3, 3)[:, :, 2]
common_tectosaur.plot_tectosaur((pts, tris), sxy, 'sxy', folder)
common_tectosaur.plot_tectosaur((pts, tris), syz, 'syz', folder)
os.system(f'google-chrome {folder}/*.pdf')
def get_slip_field(pts, tris):
n_tris = tris.shape[0]
tri_pts = pts[tris].reshape((-1, 3))
dist = np.linalg.norm(tri_pts - gauss_center, axis=1)
strike_slip = common.gaussian(*gauss_params, dist)
slip = np.zeros((strike_slip.shape[0], 3))
slip[:, 0] = strike_slip
return slip.flatten()
def main():
pts, rects = common.planar_mesh(n)
pts, tris = common.build_weird_tri_mesh(pts, rects)
if rotate:
pts = pts.dot(np.array([[0, 0, -1], [0, 1, 0], [1, 0, 0]]))
n_tris = tris.shape[0]
n_rects = rects.shape[0]
assert(3 * n_rects == n_tris)
tri_pts = pts[tris]
rect_pts = pts[rects]
tri_centers = np.mean(tri_pts, axis = 1)
tri_dist = np.linalg.norm(tri_centers, axis = 1)
rect_centers = np.mean(rect_pts, axis = 1)
rect_dist = np.linalg.norm(rect_centers, axis = 1)
# plt.figure(figsize = (15,15))
# plt.triplot(pts[:,0], pts[:,2], tris)
# plt.plot(rect_centers[:,0], rect_centers[:,2], '*')
# plt.savefig(os.path.join(folder,'rect_setup.pdf'))
if tri_srces:
strike_slip = common.gaussian(*gauss_params, tri_dist)
slip_tris = np.zeros((strike_slip.shape[0], 3))
slip_tris[:,0] = strike_slip
else:
strike_slip = common.gaussian(*gauss_params, rect_dist)
slip = np.zeros((strike_slip.shape[0], 3))
slip[:,0] = strike_slip
slip_tris = common.dofs_to_tris(slip)
# common.plot_tris((pts, tris), slip_tris[:,0], 'inputslip', folder)
strain, stress = common.eval_tris(rect_centers, tri_pts, slip_tris, sm, nu)
stress_tris = common.dofs_to_tris(stress)
common.plot_tris((pts, tris), stress_tris[:,3], 'rectsxy', folder)
common.plot_tris((pts, tris), stress_tris[:,5], 'rectsyz', folder)
plt.show()
def es_mutate_individual(individual, steps, minmax):
data = individual[0]
new_data = []
mutated_idx = random.randint(0,len(data))
for idx,value in enumerate(data):
if mutated_idx == idx:
new_value = value + (steps[idx] * common.gaussian(0,1))
else:
new_value = value
new_data.append(common.round(new_value))
return (new_data, steps)
def es_mutate_individual(individual, steps, minmax):
data = individual[0]
new_data = []
mutated_idx = random.randint(0, len(data))
for idx, value in enumerate(data):
if mutated_idx == idx:
new_value = value + (steps[idx] * common.gaussian(0, 1))
else:
new_value = value
new_data.append(common.round(new_value))
return (new_data, steps)
def es_mutate_population(population, steps, minmax):
data = population[0]
new_data = []
#mutation_idx = random.randint(0, len(population))
for idx,value in enumerate(data):
new_value = value + (steps[idx] * common.gaussian(0,1))
# Make sure mutated value no over passes max limit
# We're not using negative values
new_value = new_value % minmax[1]
new_data.append(common.round(abs(new_value)))
return (new_data, steps)
def es_mutate_population(population, steps, minmax):
data = population[0]
new_data = []
#mutation_idx = random.randint(0, len(population))
for idx, value in enumerate(data):
new_value = value + (steps[idx] * common.gaussian(0, 1))
# Make sure mutated value no over passes max limit
# We're not using negative values
new_value = new_value % minmax[1]
new_data.append(common.round(abs(new_value)))
return (new_data, steps)
def main():
pts, rects = common.bend_mesh(n)
m = common.rect_to_tri_mesh(pts, rects)
n_tris = m[1].shape[0]
tri_pts = m[0][m[1]]
tri_centers = np.mean(tri_pts, axis = 1)
dist = np.linalg.norm(tri_centers - gauss_center, axis = 1)
strike_slip = common.gaussian(*gauss_params, dist)
slip = np.zeros((strike_slip.shape[0], 3))
slip[:,0] = strike_slip
strain, stress = common.eval_tris(tri_centers, tri_pts, slip, sm, nu)
common.plot_tris(m, stress[:,3], 'trisxy', folder)
common.plot_tris(m, stress[:,5], 'trisyz', folder)
def main():
pts, rects = common.bend_mesh(n)
pts, tris = common.build_weird_tri_mesh(pts, rects)
n_tris = tris.shape[0]
n_rects = rects.shape[0]
assert (3 * n_rects == n_tris)
tri_pts = pts[tris]
rect_pts = pts[rects]
tri_centers = np.mean(tri_pts, axis=1)
tri_dist = np.linalg.norm(tri_centers - gauss_center, axis=1)
rect_centers = np.mean(rect_pts, axis=1)
rect_dist = np.linalg.norm(rect_centers - gauss_center, axis=1)
slip = np.zeros((rect_centers.shape[0], 3))
slip[:, 0] = common.gaussian(*gauss_params, rect_dist)
def xyz_to_tde(vec, rect_pts):
rect_normals = normals(rect_pts)
dip_vec = np.zeros_like(rect_normals)
dip_vec[:, 2] = 1.0
strike_vec = np.cross(rect_normals, dip_vec)
def proj(V, b):
numer = np.sum(V * b, axis=1)[:, np.newaxis] * b
denom = np.linalg.norm(b, axis=1)[:, np.newaxis]**2
return numer / denom
out = np.empty_like(vec)
out[:, 0] = np.linalg.norm(proj(vec, strike_vec), axis=1)
out[:, 1] = np.linalg.norm(proj(vec, dip_vec), axis=1)
out[:, 2] = np.linalg.norm(proj(vec, rect_normals), axis=1)
return out
slip_tde = xyz_to_tde(slip, rect_pts)
slip_tris = common.dofs_to_tris(slip_tde)
strain, stress = common.eval_tris(rect_centers, tri_pts, slip_tris, sm, nu)
stress_tris = common.dofs_to_tris(stress)
common.plot_tris((pts, tris), stress_tris[:, 3], 'rectsxy', folder)
common.plot_tris((pts, tris), stress_tris[:, 5], 'rectsyz', folder)
def main():
pts, rects = common.planar_mesh(n)
m = common.rect_to_tri_mesh(pts, rects)
n_tris = m[1].shape[0]
tri_pts = m[0][m[1]]
tri_centers = np.mean(tri_pts, axis=1)
plt.figure(figsize=(15, 15))
plt.triplot(m[0][:, 0], m[0][:, 2], m[1])
plt.plot(tri_centers[:, 0], tri_centers[:, 2], '*')
plt.savefig(os.path.join(folder, 'tri_setup.pdf'))
# plt.show()
dist = np.linalg.norm(tri_centers, axis=1)
strike_slip = common.gaussian(*gauss_params, dist)
slip = np.zeros((strike_slip.shape[0], 3))
slip[:, 0] = strike_slip
strain, stress = common.eval_tris(tri_centers, tri_pts, slip, sm, nu)
common.plot_tris(m, stress[:, 3], 'trisxy', folder)
common.plot_tris(m, stress[:, 5], 'trisyz', folder)
def input_process_fn(_x):
return gaussian(_x,
is_training=True,
mean=0,
stddev=normilize(noise, 255))
def V_barrier(x, V_barrier_height, V_barrier_mu, V_barrier_sigma):
return common.gaussian(x=x,
a=V_barrier_height,
mu=V_barrier_mu,
sigma=V_barrier_sigma)
def add_noise(imgs, noise):
noise = common.normalize(noise)
return common.gaussian(imgs, is_training=True, mean=0, stddev=noise)