Coverage for pySDC/implementations/problem_classes/nonlinear_ODE_1.py: 95%

41 statements  

« prev     ^ index     » next       coverage.py v7.6.7, created at 2024-11-16 14:51 +0000

1import numpy as np 

2 

3from pySDC.core.errors import ProblemError 

4from pySDC.core.problem import Problem 

5from pySDC.implementations.datatype_classes.mesh import mesh 

6 

7 

8# noinspection PyUnusedLocal 

9class nonlinear_ODE_1(Problem): 

10 r""" 

11 This class implements a simple nonlinear ODE with a singularity in the derivative, taken from 

12 https://www.osti.gov/servlets/purl/6111421 (Problem E-4). For :math:`0 \leq t \leq 5`, the problem is 

13 given by 

14 

15 .. math:: 

16 \frac{du(t)}{dt} = \sqrt{1 - u(t)} 

17 

18 with initial condition :math:`u(0) = 0`. The exact solution is 

19 

20 .. math:: 

21 u(t) = t - \frac{t^2}{4}. 

22 

23 Parameters 

24 ---------- 

25 u0 : float, optional 

26 Initial condition. 

27 newton_maxiter : float, optional 

28 Maximum number of iterations for Newton's method. 

29 newton_tol : float, optional 

30 Tolerance for Newton's method to terminate. 

31 stop_at_nan : bool, optional 

32 Indicates that Newton solver has to stop if ``nan`` values arise. 

33 

34 Attributes 

35 ---------- 

36 newton_itercount : int 

37 Counts the Newton iterations. 

38 newton_ncalls : int 

39 Counts calls of Newton method. 

40 """ 

41 

42 dtype_u = mesh 

43 dtype_f = mesh 

44 

45 def __init__(self, u0=0.0, newton_maxiter=200, newton_tol=5e-11, stop_at_nan=True): 

46 nvars = 1 

47 super().__init__((nvars, None, np.dtype('float64'))) 

48 self._makeAttributeAndRegister( 

49 'u0', 'newton_maxiter', 'newton_tol', 'stop_at_nan', localVars=locals(), readOnly=True 

50 ) 

51 

52 self.newton_itercount = 0 

53 self.newton_ncalls = 0 

54 

55 def u_exact(self, t): 

56 r""" 

57 Routine to compute the exact solution at time :math:`t`. 

58 

59 Parameters 

60 ---------- 

61 t : float 

62 Time of the exact solution. 

63 

64 Returns 

65 ------- 

66 me : dtype_u 

67 The exact solution. 

68 """ 

69 

70 me = self.dtype_u(self.init) 

71 me[:] = t - t**2 / 4 

72 return me 

73 

74 def eval_f(self, u, t): 

75 """ 

76 Routine to evaluate the right-hand side of the problem. 

77 

78 Parameters 

79 ---------- 

80 u : dtype_u 

81 Current values of the numerical solution. 

82 t : float 

83 Current time of the numerical solution is computed (not used here). 

84 

85 Returns 

86 ------- 

87 f : dtype_f 

88 The right-hand side of the problem (one component). 

89 """ 

90 

91 f = self.dtype_f(self.init) 

92 f[:] = np.sqrt(1 - u) 

93 return f 

94 

95 def solve_system(self, rhs, dt, u0, t): 

96 """ 

97 Simple Newton solver for the nonlinear equation 

98 

99 Parameters 

100 ---------- 

101 rhs : dtype_f 

102 Right-hand side for the nonlinear system. 

103 dt : float 

104 Abbrev. for the node-to-node stepsize (or any other factor required). 

105 u0 : dtype_u 

106 Initial guess for the iterative solver. 

107 t : float 

108 Current time (e.g. for time-dependent BCs). 

109 

110 Returns 

111 ------- 

112 u : dtype_u 

113 The solution as mesh. 

114 """ 

115 # create new mesh object from u0 and set initial values for iteration 

116 u = self.dtype_u(u0) 

117 

118 # start newton iteration 

119 n = 0 

120 res = 99 

121 while n < self.newton_maxiter: 

122 # form the function g with g(u) = 0 

123 g = u - dt * np.sqrt(1 - u) - rhs 

124 

125 # if g is close to 0, then we are done 

126 res = np.linalg.norm(g, np.inf) 

127 if res < self.newton_tol or np.isnan(res): 

128 break 

129 

130 # assemble dg/du 

131 dg = 1 - (-dt) / (2 * np.sqrt(1 - u)) 

132 # newton update: u1 = u0 - g/dg 

133 u -= 1.0 / dg * g 

134 

135 # increase iteration count 

136 n += 1 

137 

138 self.newton_itercount += 1 

139 

140 if np.isnan(res) and self.stop_at_nan: 

141 raise ProblemError('Newton got nan after %i iterations, aborting...' % n) 

142 elif np.isnan(res): 

143 self.logger.warning('Newton got nan after %i iterations...' % n) 

144 

145 if n == self.newton_maxiter: 

146 self.logger.warning('Newton did not converge after %i iterations, error is %s' % (n, res)) 

147 

148 self.newton_ncalls += 1 

149 

150 return u