东南大学研究生-数值分析上机题(2023)Python 6 常微分方程数值解法
常微分方程初值问题数值解
6.1 题目
- 编制RK4方法的通用程序;
- 编制AB4方法的通用程序(由RK4提供初值);
- 编制AB4-AM4预测校正方法通用程序(由RK4提供初值);
- 编制带改进的AB4-AM4预测校正方法通用程序(由RK4提供初值);
- 对于初值问题
{ y ′ = − x 2 y 2 , 0 ≤ x ≤ 1.5 , y ( 0 ) = 3 \begin{cases} y'=-x^{2}y^{2}, & 0\leq x \leq 1.5,\\ y(0)=3 & \\ \end{cases} {y′=−x2y2,y(0)=30≤x≤1.5,
取步长 h = 0.1 h=0.1 h=0.1,应用(1)-(4)中的四种方法进行计算,并将计算结果和精确解 y ( x ) = 3 / ( 1 + x 3 ) y(x)=3/(1+x^3) y(x)=3/(1+x3)作比较; - 通过本上机题,你能得到哪些结论?
6.2 Python源程序
# 定义一阶微分方程
def y_fxy(x, y):
return - (x ** 2) * (y ** 2)
# 定义一阶微分方程的精确解 函数
def y(x):
return 3 / (1 + x ** 3)
# RK4
def rk4(y0, h_, x0, xi):
N = int((xi - x0) / h_) # 运算的次数
x = [x0]
y_pdt = [y0] # 近似解
y_real = [y0]
err_list = [y0-y0]
for i in range(N):
k1 = y_fxy(x[-1], y_pdt[-1])
k2 = y_fxy(x[-1] + 1 / 2 * h_, y_pdt[-1] + 1 / 2 * h_ * k1)
k3 = y_fxy(x[-1] + 1 / 2 * h_, y_pdt[-1] + 1 / 2 * h_ * k2)
k4 = y_fxy(x[-1] + h_, y_pdt[-1] + h_ * k3)
y_pdt.append(y_pdt[-1] + h_ / 6 * (k1 + 2 * k2 + 2 * k3 + k4))
x.append(x[-1]+h_)
y_real.append(y(x[-1]))
err_list.append(y_real[-1] - y_pdt[-1])
return x, y_pdt, y_real, err_list
# AB4
def ab4(y0, h_, x0, xi):
N = int((xi - x0) / h_) # 运算的次数
x, y_pdt, y_real, err_list = rk4(y0, h_, x0, x0 + 3 * h_) # y0给定 y1,y2,y3由RK4得出
for i in range(3, N):
y_pdt.append(y_pdt[-1] + h_ / 24 * \
(55 * y_fxy(x[-1], y_pdt[-1]) - 59 * y_fxy(x[-2], y_pdt[-2]) + \
37 * y_fxy(x[-3], y_pdt[-3]) - 9 * y_fxy(x[-4], y_pdt[-4])))
x.append(x[-1]+h_)
y_real.append(y(x[-1]))
err_list.append(y_real[-1] - y_pdt[-1])
return x, y_pdt, y_real, err_list
# AB4_AM4预测算法
def ab4_am4(y0, h_, x0, xi):
N = int((xi - x0) / h_) # 运算的次数
x, y_pdt, y_real, err_list = rk4(y0, h_, x0, x0 + 3 * h_) # y0给定 y1,y2,y3由RK4得出
for i in range(3, N):
y_pdt.append(y_pdt[-1] + h_ / 24 * \
(55 * y_fxy(x[-1], y_pdt[-1]) - 59 * y_fxy(x[-2], y_pdt[-2]) + \
37 * y_fxy(x[-3], y_pdt[-3]) - 9 * y_fxy(x[-4], y_pdt[-4])))
x.append(x[-1] + h_)
y_pdt[-1] = y_pdt[-2] + h_ / 24 * \
(9 * y_fxy(x[-1], y_pdt[-1]) + 19 * y_fxy(x[-2], y_pdt[-2]) - \
5 * y_fxy(x[-3], y_pdt[-3]) + y_fxy(x[-4], y_pdt[-4]))
y_real.append(y(x[-1]))
err_list.append(y_real[-1] - y_pdt[-1])
return x, y_pdt, y_real, err_list
# 改进的AB4_AM4预测算法
def plus_ab4_am4(y0, h_, x0, xi):
N = int((xi - x0) / h_) # 运算的次数
x, y_pdt, y_real, err_list = rk4(y0, h_, x0, x0 + 3 * h_) # y0给定 y1,y2,y3由RK4得出
for i in range(3, N):
y_pdt.append(y_pdt[-1] + h_ / 24 * \
(55 * y_fxy(x[-1], y_pdt[-1]) - 59 * y_fxy(x[-2], y_pdt[-2]) + \
37 * y_fxy(x[-3], y_pdt[-3]) - 9 * y_fxy(x[-4], y_pdt[-4])))
x.append(x[-1] + h_)
y_c = y_pdt[-2] + h_ / 24 * \
(9 * y_fxy(x[-1], y_pdt[-1]) + 19 * y_fxy(x[-2], y_pdt[-2]) - \
5 * y_fxy(x[-3], y_pdt[-3]) + y_fxy(x[-4], y_pdt[-4]))
y_pdt[-1] = 251 / 270 * y_c + 19 / 270 * y_pdt[-1]
y_real.append(y(x[-1]))
err_list.append(y_real[-1] - y_pdt[-1])
return x, y_pdt, y_real, err_list
def display(x, y_pdt, y_real, err_list, h_, x0, xi):
N = int((xi - x0) / h_) # 运算的次数
print("i xi yi y(xi) y(xi)-yi")
for i in range(N):
print("{:d} {:.2f} {:.8f} {:.8f} {:.8f}".format\
(i+1, x[i+1], y_pdt[i+1], y_real[i+1], err_list[i+1]))
if __name__ == '__main__':
y_0 = 3 # 初值
h = 0.1 # 步长
x_0 = 0 # 区间左端点
x_i = 1.5 # 区间右端点
X, Y_pdt, Y_real, Error = rk4(y_0, h, x_0, x_i)
print("RK4:")
display(X, Y_pdt, Y_real, Error, h, x_0, x_i)
print("RK4整体截断误差:{:.8f}".format(max(list(map(abs, Error)))))
X, Y_pdt, Y_real, Error = ab4(y_0, h, x_0, x_i)
print("AB4:")
display(X, Y_pdt, Y_real, Error, h, x_0, x_i)
print("AB4整体截断误差:{:.8f}".format(max(list(map(abs, Error)))))
X, Y_pdt, Y_real, Error = ab4_am4(y_0, h, x_0, x_i)
print("AB4-AM4预测校正:")
display(X, Y_pdt, Y_real, Error, h, x_0, x_i)
print("AB4-AM4预测矫正整体截断误差:{:.8f}".format(max(list(map(abs, Error)))))
X, Y_pdt, Y_real, Error = plus_ab4_am4(y_0, h, x_0, x_i)
print("改进的AB4-AM4预测校正:")
display(X, Y_pdt, Y_real, Error, h, x_0, x_i)
print("改进的AB4-AM4预测矫正整体截断误差:{:.8f}".format(max(list(map(abs, Error)))))
6.3 程序运行结果
RK4:
RK4:
i xi yi y(xi) y(xi)-yi
1 0.10 2.99700281 2.99700300 0.00000019
2 0.20 2.97619008 2.97619048 0.00000039
3 0.30 2.92112875 2.92112950 0.00000076
4 0.40 2.81954726 2.81954887 0.00000161
5 0.50 2.66666349 2.66666667 0.00000318
6 0.60 2.46710026 2.46710526 0.00000501
7 0.70 2.23379914 2.23380491 0.00000577
8 0.80 1.98412285 1.98412698 0.00000413
9 0.90 1.73510711 1.73510700 -0.00000012
10 1.00 1.50000581 1.50000000 -0.00000581
11 1.10 1.28701259 1.28700129 -0.00001131
12 1.20 1.09972217 1.09970674 -0.00001542
13 1.30 0.93839746 0.93837973 -0.00001773
14 1.40 0.80130043 0.80128205 -0.00001838
15 1.50 0.68573209 0.68571429 -0.00001780
RK4整体截断误差:0.00001838
AB4:
AB4:
i xi yi y(xi) y(xi)-yi
1 0.10 2.99700281 2.99700300 0.00000019
2 0.20 2.97619008 2.97619048 0.00000039
3 0.30 2.92112875 2.92112950 0.00000076
4 0.40 2.81838926 2.81954887 0.00115961
5 0.50 2.66467247 2.66666667 0.00199420
6 0.60 2.46520263 2.46710526 0.00190263
7 0.70 2.23307895 2.23380491 0.00072596
8 0.80 1.98495058 1.98412698 -0.00082359
9 0.90 1.73704329 1.73510700 -0.00193629
10 1.00 1.50219455 1.50000000 -0.00219455
11 1.10 1.28876344 1.28700129 -0.00176216
12 1.20 1.10072420 1.09970674 -0.00101746
13 1.30 0.93871050 0.93837973 -0.00033077
14 1.40 0.80113495 0.80128205 0.00014710
15 1.50 0.68533458 0.68571429 0.00037971
AB4整体截断误差:0.00219455
AB4-AM4预测校正:
AB4-AM4预测校正:
i xi yi y(xi) y(xi)-yi
1 0.10 2.99700281 2.99700300 0.00000019
2 0.20 2.97619008 2.97619048 0.00000039
3 0.30 2.92112875 2.92112950 0.00000076
4 0.40 2.81967843 2.81954887 -0.00012956
5 0.50 2.66687598 2.66666667 -0.00020932
6 0.60 2.46725176 2.46710526 -0.00014650
7 0.70 2.23373141 2.23380491 0.00007350
8 0.80 1.98378670 1.98412698 0.00034028
9 0.90 1.73460744 1.73510700 0.00049956
10 1.00 1.49951594 1.50000000 0.00048406
11 1.10 1.28665714 1.28700129 0.00034415
12 1.20 1.09953315 1.09970674 0.00017360
13 1.30 0.93834252 0.93837973 0.00003721
14 1.40 0.80132737 0.80128205 -0.00004532
15 1.50 0.68579611 0.68571429 -0.00008183
AB4-AM4预测校正整体截断误差:0.00049956
改进的AB4-AM4预测校正:
改进的AB4-AM4预测校正:
i xi yi y(xi) y(xi)-yi
1 0.10 2.99700281 2.99700300 0.00000019
2 0.20 2.97619008 2.97619048 0.00000039
3 0.30 2.92112875 2.92112950 0.00000076
4 0.40 2.81958771 2.81954887 -0.00003884
5 0.50 2.66671285 2.66666667 -0.00004619
6 0.60 2.46709703 2.46710526 0.00000823
7 0.70 2.23368249 2.23380491 0.00012242
8 0.80 1.98388468 1.98412698 0.00024230
9 0.90 1.73480801 1.73510700 0.00029899
10 1.00 1.49973191 1.50000000 0.00026809
11 1.10 1.28682068 1.28700129 0.00018061
12 1.20 1.09962178 1.09970674 0.00008496
13 1.30 0.93836732 0.93837973 0.00001242
14 1.40 0.80131135 0.80128205 -0.00002930
15 1.50 0.68576045 0.68571429 -0.00004616
改进的AB4-AM4预测校正整体截断误差:0.00029899
6.4 总结感悟
- 根据数值分析理论推导的结果,RK4、AB4、AB4-AM4预测校正具有4阶精度,而改进的AB4-AM4预测校正具有5阶精度,但是对于该问题来说,比较四种常微分方程数值解法在 [ 0.1 , 1.5 ] [0.1,1.5] [0.1,1.5]上的整体截断误差,则是RK4<改进的AB4-AM4预测校正<AB4-AM4预测校正<AB4,RK4(单步法)的精度要比多步法(AB4、AB4-AM4预测校正、改进的AB4-AM4预测校正)的精度更高;
- 要根据不同的问题选择合适的数值解法,公式的精度越高不代表实际的求解精度越高;
- 常微分方程的数值解法是广泛应用的方法,在以后的工程实践与科研之中会有更多的应用.