|
| 1 | +# |
| 2 | +# Test for the operator class |
| 3 | +# |
| 4 | +import pybamm |
| 5 | +from tests import ( |
| 6 | + get_mesh_for_testing, |
| 7 | + get_p2d_mesh_for_testing, |
| 8 | + get_1p1d_mesh_for_testing, |
| 9 | +) |
| 10 | + |
| 11 | +import numpy as np |
| 12 | +from scipy.sparse import kron, eye |
| 13 | +import unittest |
| 14 | + |
| 15 | + |
| 16 | +def errors(pts, function, extrap): |
| 17 | + |
| 18 | + domain = "test" |
| 19 | + x = pybamm.SpatialVariable("x", domain=domain) |
| 20 | + geometry = { |
| 21 | + domain: {"primary": {x: {"min": pybamm.Scalar(0), "max": pybamm.Scalar(1)}}} |
| 22 | + } |
| 23 | + submesh_types = {domain: pybamm.MeshGenerator(pybamm.Uniform1DSubMesh)} |
| 24 | + var_pts = {x: pts} |
| 25 | + mesh = pybamm.Mesh(geometry, submesh_types, var_pts) |
| 26 | + |
| 27 | + spatial_methods = {"test": pybamm.FiniteVolume} |
| 28 | + disc = pybamm.Discretisation(mesh, spatial_methods) |
| 29 | + |
| 30 | + disc.spatial_methods["test"].extrapolation = extrap |
| 31 | + var = pybamm.Variable("var", domain="test") |
| 32 | + left_extrap = pybamm.BoundaryValue(var, "left") |
| 33 | + right_extrap = pybamm.BoundaryValue(var, "right") |
| 34 | + |
| 35 | + submesh = mesh["test"] |
| 36 | + y, l_true, r_true = function(submesh[0].nodes) |
| 37 | + |
| 38 | + disc.set_variable_slices([var]) |
| 39 | + left_extrap_processed = disc.process_symbol(left_extrap) |
| 40 | + right_extrap_processed = disc.process_symbol(right_extrap) |
| 41 | + |
| 42 | + l_error = np.abs(l_true - left_extrap_processed.evaluate(None, y)) |
| 43 | + r_error = np.abs(r_true - right_extrap_processed.evaluate(None, y)) |
| 44 | + |
| 45 | + return l_error, r_error |
| 46 | + |
| 47 | + |
| 48 | +def get_errors(function, extrap, pts): |
| 49 | + |
| 50 | + l_errors = np.zeros(pts.shape) |
| 51 | + r_errors = np.zeros(pts.shape) |
| 52 | + |
| 53 | + for i, pt in enumerate(pts): |
| 54 | + l_errors[i], r_errors[i] = errors(pt, function, extrap) |
| 55 | + |
| 56 | + return l_errors, r_errors |
| 57 | + |
| 58 | + |
| 59 | +class TestExtrapolation(unittest.TestCase): |
| 60 | + def test_quadratic_convergence(self): |
| 61 | + |
| 62 | + # all tests are performed on x in [0, 1] |
| 63 | + |
| 64 | + def x_squared(x): |
| 65 | + y = x ** 2 |
| 66 | + l_true = 0 |
| 67 | + r_true = 1 |
| 68 | + return y, l_true, r_true |
| 69 | + |
| 70 | + pts = 10 ** np.arange(1, 6, 1) |
| 71 | + dx = 1 / pts |
| 72 | + l_errors_lin, r_errors_lin = get_errors(x_squared, "linear", pts) |
| 73 | + l_errors_quad, r_errors_quad = get_errors(x_squared, "quadratic", pts) |
| 74 | + |
| 75 | + l_lin_rates = np.log(l_errors_lin[:-1] / l_errors_lin[1:]) / np.log( |
| 76 | + dx[:-1] / dx[1:] |
| 77 | + ) |
| 78 | + |
| 79 | + r_lin_rates = np.log(r_errors_lin[:-1] / r_errors_lin[1:]) / np.log( |
| 80 | + dx[:-1] / dx[1:] |
| 81 | + ) |
| 82 | + |
| 83 | + np.testing.assert_array_almost_equal(l_lin_rates, 2) |
| 84 | + np.testing.assert_array_almost_equal(r_lin_rates, 2) |
| 85 | + |
| 86 | + # check quadratic is equal up to machine precision |
| 87 | + np.testing.assert_array_almost_equal(l_errors_quad, 0, decimal=14) |
| 88 | + np.testing.assert_array_almost_equal(r_errors_quad, 0, decimal=14) |
| 89 | + |
| 90 | + def x_cubed(x): |
| 91 | + y = x ** 3 |
| 92 | + l_true = 0 |
| 93 | + r_true = 1 |
| 94 | + return y, l_true, r_true |
| 95 | + |
| 96 | + l_errors_lin, r_errors_lin = get_errors(x_squared, "linear", pts) |
| 97 | + |
| 98 | + l_lin_rates = np.log(l_errors_lin[:-1] / l_errors_lin[1:]) / np.log( |
| 99 | + dx[:-1] / dx[1:] |
| 100 | + ) |
| 101 | + |
| 102 | + r_lin_rates = np.log(r_errors_lin[:-1] / r_errors_lin[1:]) / np.log( |
| 103 | + dx[:-1] / dx[1:] |
| 104 | + ) |
| 105 | + |
| 106 | + np.testing.assert_array_almost_equal(l_lin_rates, 2) |
| 107 | + np.testing.assert_array_almost_equal(r_lin_rates, 2) |
| 108 | + |
| 109 | + # quadratic case |
| 110 | + pts = 5 ** np.arange(1, 7, 1) |
| 111 | + dx = 1 / pts |
| 112 | + l_errors_quad, r_errors_quad = get_errors(x_cubed, "quadratic", pts) |
| 113 | + |
| 114 | + l_quad_rates = np.log(l_errors_quad[:-1] / l_errors_quad[1:]) / np.log( |
| 115 | + dx[:-1] / dx[1:] |
| 116 | + ) |
| 117 | + |
| 118 | + r_quad_rates = np.log(r_errors_quad[:-1] / r_errors_quad[1:]) / np.log( |
| 119 | + dx[:-1] / dx[1:] |
| 120 | + ) |
| 121 | + |
| 122 | + np.testing.assert_array_almost_equal(l_quad_rates, 3) |
| 123 | + np.testing.assert_array_almost_equal(r_quad_rates, 3, decimal=3) |
| 124 | + |
| 125 | + |
| 126 | +if __name__ == "__main__": |
| 127 | + print("Add -v for more debug output") |
| 128 | + import sys |
| 129 | + |
| 130 | + if "-v" in sys.argv: |
| 131 | + debug = True |
| 132 | + pybamm.settings.debug_mode = True |
| 133 | + unittest.main() |
| 134 | + |
0 commit comments