def exercise8():
file_name = 'q8_3.npz'
f = open(file_name, 'rb')
npzfile = np.load(f)
a = npzfile['a']
b = npzfile['b']
d = npzfile['d']
xc = npzfile['xc']
b_measured = npzfile['F']
# Create source points
n = xc.shape[0]
xs = chebyshev(a, b, n) # Source points
lambda_select = 10**(-4)
print(a)
print(b)
print(d)
print(xc.shape[0])
# Calculate A-matrix from Legendre-Gauss quadrature
xq, w = legendre(a, b, 2 * n**2)
a_matrix = fredholm_lhs(d, xc, xs, xq, w, kernel)
identity_matrix = np.identity(a_matrix.shape[0])
a_matrix_transposed = np.transpose(a_matrix)
left_side = np.matmul(a_matrix_transposed,
a_matrix) + (lambda_select * identity_matrix)
right_side = np.matmul(a_matrix_transposed, b_measured)
rho_solution = np.linalg.solve(left_side, right_side)
plt.figure("measurement-8-3")
plt.title("Solution to measurement 3")
plt.plot(xc, rho_solution, label="Numeric, $\lambda = 10^{-4}$")
plt.xlabel('xc')
plt.ylabel('Rho(x)')
plt.legend()
plt.grid(True)
plt.show()
def exercise7():
# Orders
Nc = 30 # Number of collocation points
Ns = 30
Nq = Nc**2 # Number of panels
# Limits
a = 0
b = 1
# Constants
omega = 3 * np.pi
gamma = -2
# Depths
d2 = 0.25
d3 = 2.5
# Create chebyshev points
xc = chebyshev(a, b, Nc) # Collocation points
xs = chebyshev(a, b, Nc) # Source points
rho_analytic = rho(omega, gamma, xs)
# Analytic disturbed b-vector
b_analytic_disturbed2 = fredholm_rhs(a, b, omega, gamma, d2,
xc) * (1 + epsilon(xc))
b_analytic_disturbed3 = fredholm_rhs(a, b, omega, gamma, d3,
xc) * (1 + epsilon(xc))
xq, w = legendre(a, b, Nq)
lambda_vector = lambda_list(16)
identity_matrix = np.identity(Nc)
rho_vector2 = np.zeros(lambda_vector.shape[0])
rho_vector3 = np.zeros(lambda_vector.shape[0])
# Calculate A-matrix from Legendre-Gauss quadrature
a_matrix2 = fredholm_lhs(d2, xc, xs, xq, w, kernel)
a_matrix_transposed2 = np.transpose(a_matrix2)
a_matrix3 = fredholm_lhs(d3, xc, xs, xq, w, kernel)
a_matrix_transposed3 = np.transpose(a_matrix3)
for j in range(lambda_vector.shape[0]):
# Calculate 2
left_side = np.matmul(a_matrix_transposed2,
a_matrix2) + (lambda_vector[j] * identity_matrix)
right_side = np.matmul(a_matrix_transposed2, b_analytic_disturbed2)
rho_vector2[j] = np.amax(
np.absolute(
np.subtract(np.linalg.solve(left_side, right_side),
rho_analytic)))
# Calculate 3
left_side = np.matmul(a_matrix_transposed3,
a_matrix3) + (lambda_vector[j] * identity_matrix)
right_side = np.matmul(a_matrix_transposed3, b_analytic_disturbed3)
rho_vector3[j] = np.amax(
np.absolute(
np.subtract(np.linalg.solve(left_side, right_side),
rho_analytic)))
plt.figure("tikhonov")
plt.title("Rho error")
plt.loglog(lambda_vector, rho_vector2, label="Numeric, $d = 0.25$")
plt.loglog(lambda_vector, rho_vector3, label="Numeric, $d = 2.5$")
plt.xlabel('Lambda')
plt.ylabel('Rho(x) error')
plt.legend()
plt.grid(True)
plt.show()
def exercise5():
# Orders
Nc = 30 # Number of collocation points
Nq = Nc**2 # Number of panels
# Limits
a = 0
b = 1
# Constants
omega = 3 * np.pi
gamma = -2
# Depths
d1 = 0.025
d2 = 0.25
d3 = 2.5
# Error
error1 = np.zeros(Nc - 5)
error2 = np.zeros(Nc - 5)
error3 = np.zeros(Nc - 5)
axis = np.arange(5, Nc, 1)
for i in range(5, Nc, 1):
# Print each iteration
print(i)
# Create points
xc = chebyshev(a, b, i) # Collocation points
xs = chebyshev(a, b, i) # Source points
# Calculate true rho
rho_analytic = rho(omega, gamma, xs)
# Calculate A-matrix from Legendre-Gauss quadrature
xq, w = legendre(a, b, Nq)
a_matrix_legendre1 = fredholm_lhs(d1, xc, xs, xq, w, kernel)
a_matrix_legendre2 = fredholm_lhs(d2, xc, xs, xq, w, kernel)
a_matrix_legendre3 = fredholm_lhs(d3, xc, xs, xq, w, kernel)
# Find rho by linear algebra
rho_numeric1 = np.linalg.solve(
a_matrix_legendre1, fredholm_rhs(a, b, omega, gamma, d1, xc))
rho_numeric2 = np.linalg.solve(
a_matrix_legendre2, fredholm_rhs(a, b, omega, gamma, d2, xc))
rho_numeric3 = np.linalg.solve(
a_matrix_legendre3, fredholm_rhs(a, b, omega, gamma, d3, xc))
error1[i - 5] = np.amax(
np.absolute(np.subtract(rho_numeric1, rho_analytic)))
error2[i - 5] = np.amax(
np.absolute(np.subtract(rho_numeric2, rho_analytic)))
error3[i - 5] = np.amax(
np.absolute(np.subtract(rho_numeric3, rho_analytic)))
plt.figure("inverse-error")
plt.title("Numerical solution for Rho")
plt.semilogy(axis, error1, label="$d = 0.025$")
plt.semilogy(axis, error2, label="$d = 0.25$")
plt.semilogy(axis, error3, label="$d = 2.5$")
plt.xlabel('Nc')
plt.ylabel('Error')
plt.legend()
plt.grid(True)
plt.show()
def exercise6():
# Orders
Nc = 30 # Number of collocation points
Nq = Nc**2 # Number of panels
# Limits
a = 0
b = 1
# Constants
omega = 3 * np.pi
gamma = -2
# Depths
d1 = 0.025
d2 = 0.25
d3 = 2.5
# Create chebyshev points
xc = chebyshev(a, b, Nc) # Collocation points
xs = chebyshev(a, b, Nc) # Source points
"""
# Plot 1 (d = 0.025)
b_analytic1 = fredholm_rhs(a, b, omega, gamma, d1, xc)
b_analytic_disturbed1 = fredholm_rhs(a, b, omega, gamma, d1, xc) * (1 + epsilon(xc))
plt.figure("inverse-error")
plt.title("Analytic with and without random noise")
plt.plot(xc, b_analytic1, label="Analytic, $d = 0.025$")
plt.plot(xc, b_analytic_disturbed1, label="Analytic disturbed, $d = 0.025$")
plt.xlabel('x')
plt.ylabel('F(x)')
plt.legend()
plt.grid(True)
plt.show()
# Plot 2 (d = 0.25)
b_analytic2 = fredholm_rhs(a, b, omega, gamma, d2, xc)
b_analytic_disturbed2 = fredholm_rhs(a, b, omega, gamma, d2, xc) * (1 + epsilon(xc))
plt.figure("inverse-error")
plt.title("Analytic with and without random noise")
plt.plot(xc, b_analytic2, label="Analytic, $d = 0.25$")
plt.plot(xc, b_analytic_disturbed2, label="Analytic disturbed, $d = 0.25$")
plt.xlabel('x')
plt.ylabel('F(x)')
plt.legend()
plt.grid(True)
plt.show()
# Plot 3 (d = 2.5)
b_analytic3 = fredholm_rhs(a, b, omega, gamma, d3, xc)
b_analytic_disturbed3 = fredholm_rhs(a, b, omega, gamma, d3, xc) * (1 + epsilon(xc))
plt.figure("inverse-error")
plt.title("Analytic with and without random noise")
plt.plot(xc, b_analytic3, label="Analytic, $d = 2.5$")
plt.plot(xc, b_analytic_disturbed3, label="Analytic disturbed, $d = 2.5$")
plt.xlabel('x')
plt.ylabel('F(x)')
plt.legend()
plt.grid(True)
plt.show()
"""
# Rho analytic
rho_analytic = rho(omega, gamma, xs)
xq, w = legendre(a, b, Nq)
# Plot 4
a_matrix_legendre1 = fredholm_lhs(d1, xc, xs, xq, w, kernel)
rho_numeric1 = np.linalg.solve(a_matrix_legendre1,
fredholm_rhs(a, b, omega, gamma, d1, xc))
rho_numeric_disturbed1 = np.linalg.solve(
a_matrix_legendre1,
fredholm_rhs(a, b, omega, gamma, d1, xc) * (1 + epsilon(xc)))
plt.figure("inverse-error")
plt.title("Rho")
plt.plot(xc, rho_analytic, label="Analytic, $d = 0.025$")
plt.plot(xc, rho_numeric1, label="Numeric, $d = 0.025$")
plt.plot(xc,
rho_numeric_disturbed1,
label="Numeric disturbed, $d = 0.025$")
plt.xlabel('x')
plt.ylabel('Rho(x)')
plt.legend()
plt.grid(True)
plt.show()
# Plot 5
a_matrix_legendre2 = fredholm_lhs(d2, xc, xs, xq, w, kernel)
rho_numeric2 = np.linalg.solve(a_matrix_legendre2,
fredholm_rhs(a, b, omega, gamma, d2, xc))
rho_numeric_disturbed2 = np.linalg.solve(
a_matrix_legendre2,
fredholm_rhs(a, b, omega, gamma, d2, xc) * (1 + epsilon(xc)))
plt.figure("inverse-error")
plt.title("Rho")
plt.plot(xc, rho_analytic, label="Analytic, $d = 0.25$")
plt.plot(xc, rho_numeric2, label="Numeric, $d = 0.25$")
plt.plot(xc, rho_numeric_disturbed2, label="Numeric disturbed, $d = 0.25$")
plt.yscale('symlog')
plt.xlabel('x')
plt.ylabel('Rho(x)')
plt.legend()
plt.grid(True)
plt.show()
# Plot 6
a_matrix_legendre3 = fredholm_lhs(d3, xc, xs, xq, w, kernel)
rho_numeric3 = np.linalg.solve(a_matrix_legendre3,
fredholm_rhs(a, b, omega, gamma, d3, xc))
rho_numeric_disturbed3 = np.linalg.solve(
a_matrix_legendre3,
fredholm_rhs(a, b, omega, gamma, d3, xc) * (1 + epsilon(xc)))
plt.figure("inverse-error")
plt.title("Rho")
plt.plot(xc, rho_analytic, label="Analytic, $d = 2.5$")
plt.plot(xc, rho_numeric3, label="Numeric, $d = 2.5$")
plt.plot(xc, rho_numeric_disturbed3, label="Numeric disturbed, $d = 2.5$")
plt.yscale('symlog')
plt.xlabel('x')
plt.ylabel('Rho(x)')
plt.legend()
plt.grid(True)
plt.show()
def exercise4():
# Orders
Nc = 40 # Number of collocation points
Ns = Nc # Number of source points
Nq = 400 # Number of panels
# Limits
a = 0
b = 1
# Constants
omega = 3 * np.pi
gamma = -2
# Depths
d = 0.025
# Create points
xc = chebyshev(a, b, Nc) # Collocation points
xs = chebyshev(a, b, Ns) # Source points
# Calculate rho vector
rho_vector = rho(omega, gamma, xs)
# Calculate A-matrix from Legendre-Gauss quadrature
xq, w = legendre(a, b, Nq)
a_matrix_legendre = fredholm_lhs(d, xc, xs, xq, w, kernel)
# Calculate numerical B-vector
b_numeric_legendre = np.dot(a_matrix_legendre, rho_vector)
# Calculate analytic B-vector
b_analytic = fredholm_rhs(a, b, omega, gamma, d, xc)
# Plot integration
plt.figure("integration-plot")
plt.title("Analytic and numerical integration")
plt.plot(xc, b_analytic, label="Analytic")
plt.plot(xc, b_numeric_legendre, label="Legendre")
plt.xlabel('x')
plt.ylabel('F(x)')
plt.legend()
plt.grid(True)
# Get error by # of panels
error_legendre = np.zeros(Nq - 2)
error_axis = np.arange(1, Nq - 1, 1)
for i in range(1, Nq - 1, 1):
# Print each iteration
print(i)
# Calculate A-matrix from Legendre-Gauss quadrature
xq, w = legendre(a, b, i)
a_matrix_legendre = fredholm_lhs(d, xc, xs, xq, w, kernel)
# Calculate numerical B-vector
b_numeric_legendre = np.dot(a_matrix_legendre, rho_vector)
# Calculate error
error_legendre[i - 1] = np.amax(
np.absolute(np.subtract(b_analytic, b_numeric_legendre)))
plt.figure("error-plot")
plt.title("Legendre-Gauss integration error")
plt.semilogy(error_axis, error_legendre, label="Legendre error")
plt.xlabel('Nq')
plt.ylabel('Max integration error')
plt.legend()
plt.grid(True)
plt.show()