数值积分和数值微分:
概念 | 释义 | 性质 |
---|---|---|
机械求积 | 若选取 [ a , b ] [a,b] [a,b]上若干点 x k x_k xk,然后用 f ( x k ) f(x_k) f(xk)加权求和得到积分 ∫ a b f ( x ) d x = ∑ k = 0 n a k f ( x k ) \int_a^bf(x)dx = \sum_{k=0}^na_kf(x_k) ∫abf(x)dx=∑k=0nakf(xk),而系数 a k a_k ak只和点的选取有关而不依赖于函数的具体形式,则称这类数值积分方法为机械求积,其中 x k x_k xk为求积节点, a k a_k ak为求积系数 / 权 | |
代数精度 | 如果某个数值积分公式对于次数不超过n的多项式均能准确地成立,但对于n+1次多项式就不能精确成立,则称该求积公式具有n次代数精度 | |
p阶收敛 | 若求积公式 I n I_n In(步长为h),满足 lim h → 0 I − I n h p = C \lim_{h\rightarrow 0}\cfrac{I-I_n}{h^p} = C limh→0hpI−In=C,则称该求积公式p阶收敛 |
设将积分区间
[
a
,
b
]
[a,b]
[a,b]划分为n等份,步长
h
=
b
−
a
n
h = \frac{b-a}{n}
h=nb−a,选取等距节点
x
k
=
a
+
k
h
x_k = a+kh
xk=a+kh,
则可得到插值型求积公式:
I n = ( b − a ) ∑ k = 0 n C k ( n ) f ( x k ) I_n = (b-a)\sum_{k=0}^nC_k^{(n)}f(x_k) In=(b−a)∑k=0nCk(n)f(xk)
其中 C k ( n ) = ( − 1 ) n − k n k ! ( n − k ) ! ∫ 0 n ∏ j = 0 且 j ≠ k n ( x − j ) d x C_k^{(n)} = \cfrac{(-1)^{n-k}}{nk!(n-k)!}\int_{0}^n \prod_{j=0且j \ne k}^n(x-j)dx Ck(n)=nk!(n−k)!(−1)n−k∫0n∏j=0且j=kn(x−j)dx称 C o t e s Cotes Cotes系数
=> 当 n = 2 m n = 2m n=2m时, N e w t o n − C o t e s Newton-Cotes Newton−Cotes公式具有 n + 1 n+1 n+1阶代数精度
=> 当 n = 2 m + 1 n = 2m+1 n=2m+1时, N e w t o n − C o t e s Newton-Cotes Newton−Cotes公式具有 n n n阶代数精度
=> 当 n = 1 n = 1 n=1时, C 0 ( 1 ) = C 1 ( 1 ) = 1 2 C_0^{(1)} = C_1^{(1)} = \frac{1}{2} C0(1)=C1(1)=21, I 1 = T = b − a 2 [ f ( a ) + f ( b ) ] I_1 = T = \frac{b-a}{2}[f(a)+f(b)] I1=T=2b−a[f(a)+f(b)] (梯形公式)
其余项 R T = − b − a 12 ( b − a ) 2 f ′ ′ ( ξ ) R_T = - \frac{b-a}{12}(b-a)^2f''(\xi) RT=−12b−a(b−a)2f′′(ξ)
=> 当 n = 2 n = 2 n=2时, C 0 ( 2 ) = 1 6 , C 1 ( 2 ) = 2 3 , C 2 ( 2 ) = 1 6 C_0^{(2)} = \frac{1}{6}, C_1^{(2)} = \frac{2}{3},C_2^{(2)} = \frac{1}{6} C0(2)=61,C1(2)=32,C2(2)=61, I 2 = S = b − a 6 [ f ( a ) + 4 f ( a + b 2 ) + f ( b ) ] I_2 = S = \frac{b-a}{6}[f(a)+4f(\frac{a+b}{2})+f(b)] I2=S=6b−a[f(a)+4f(2a+b)+f(b)] ( S i m p s o n Simpson Simpson公式)
其余项 R S = − b − a 180 ( b − a 2 ) 4 f ( 4 ) ( ξ ) R_S = - \frac{b-a}{180}(\frac{b-a}{2})^4f^{(4)}(\xi) RS=−180b−a(2b−a)4f(4)(ξ)
=> 当 n = 4 n = 4 n=4时, I 4 = C = b − a 90 [ 7 f ( a ) + 32 f ( x 1 ) + 12 f ( x 2 ) + 32 f ( x 3 ) + 7 f ( b ) ] I_4 = C = \frac{b-a}{90}[7f(a)+32f(x_1)+12f(x_2)+32f(x_3)+7f(b)] I4=C=90b−a[7f(a)+32f(x1)+12f(x2)+32f(x3)+7f(b)] ( C o t e s Cotes Cotes公式)
其余项 R C = − 2 ( b − a ) 945 ( b − a 4 ) 6 f ( 6 ) ( ξ ) R_C = - \frac{2(b-a)}{945}(\frac{b-a}{4})^6f^{(6)}(\xi) RC=−9452(b−a)(4b−a)6f(6)(ξ)
=> 当 n ≥ 8 n\ge 8 n≥8时, C o t e s Cotes Cotes系数有正有负,不够稳定,故通常不采用高阶的 N e w t o n − C o t e s Newton-Cotes Newton−Cotes公式
(设将积分区间 [ a , b ] [a,b] [a,b]划分为n等份,步长 h = b − a n h = \frac{b-a}{n} h=nb−a,选取等距节点 x k = a + k h x_k = a+kh xk=a+kh,先用低阶的 N e w t o n − C o t e s Newton-Cotes Newton−Cotes公式求得每个子区间 [ x k , x k + 1 ] [x_k,x_{k+1}] [xk,xk+1]的积分值 I k I_k Ik,然后求和,用 ∑ k = 0 n − 1 I k \sum_{k=0}^{n-1}I_k ∑k=0n−1Ik作为 I I I的近似值)
=> 复化梯形公式: T n = h 2 [ f ( a ) + f ( b ) + 2 ∑ k = 1 n − 1 f ( x k ) ] T_n = \frac{h}{2}[f(a)+f(b)+2\sum_{k=1}^{n-1}f(x_k)] Tn=2h[f(a)+f(b)+2∑k=1n−1f(xk)]
其余项 R T n = − b − a 12 h 2 f ′ ′ ( ξ ) = 1 n 2 R T R_{T_n} = - \frac{b-a}{12}h^2f''(\xi) = \frac{1}{n^2}R_T RTn=−12b−ah2f′′(ξ)=n21RT
=> 复化 S i m p s o n Simpson Simpson公式: S n = h 6 [ f ( a ) + 4 ∑ k = 0 n − 1 f ( x k + 1 2 ) + f ( b ) + 2 ∑ k = 1 n − 1 f ( x k ) ] S_n = \frac{h}{6}[f(a)+4\sum_{k=0}^{n-1}f(x_{k+\frac{1}{2}})+f(b)+2\sum_{k=1}^{n-1}f(x_k)] Sn=6h[f(a)+4∑k=0n−1f(xk+21)+f(b)+2∑k=1n−1f(xk)]
其余项 R S n = − b − a 180 ( h 2 ) 4 f ( 4 ) ( ξ ) = 1 n 4 R S R_{S_n} = - \frac{b-a}{180}(\frac{h}{2})^4f^{(4)}(\xi) = \frac{1}{n^4}R_S RSn=−180b−a(2h)4f(4)(ξ)=n41RS
=> 复化 C o t e s Cotes Cotes公式: C n = h 90 [ 7 f ( a ) + 32 ∑ k = 0 n − 1 f ( x k + 1 4 ) + 12 ∑ k = 0 n − 1 f ( x k + 1 2 ) + 32 ∑ k = 0 n − 1 f ( x k + 3 4 ) + 7 f ( b ) + 14 ∑ k = 1 n − 1 f ( x k ) ] C_n = \frac{h}{90}[7f(a)+32\sum_{k=0}^{n-1}f(x_{k+\frac{1}{4}})+12\sum_{k=0}^{n-1}f(x_{k+\frac{1}{2}})+32\sum_{k=0}^{n-1}f(x_{k+\frac{3}{4}})+7f(b)+14\sum_{k=1}^{n-1}f(x_k)] Cn=90h[7f(a)+32∑k=0n−1f(xk+41)+12∑k=0n−1f(xk+21)+32∑k=0n−1f(xk+43)+7f(b)+14∑k=1n−1f(xk)]
其余项 R C n = − 2 ( b − a ) 945 ( h 4 ) 6 f ( 6 ) ( ξ ) = 1 n 6 R C R_{C_n} = - \frac{2(b-a)}{945}(\frac{h}{4})^6f^{(6)}(\xi)= \frac{1}{n^6}R_C RCn=−9452(b−a)(4h)6f(6)(ξ)=n61RC
① 梯形公式构造 S i m p s o n Simpson Simpson公式: S n = 4 3 T 2 n − 1 3 T n S_n = \frac{4}{3}T_{2n}-\frac{1}{3}T_{n} Sn=34T2n−31Tn
② S i m p s o n Simpson Simpson公式构造 C o t e s Cotes Cotes公式: C n = 16 15 S 2 n − 1 15 S n C_n = \frac{16}{15}S_{2n}-\frac{1}{15}S_{n} Cn=1516S2n−151Sn
③ C o t e s Cotes Cotes公式构造 R o m b e r g Romberg Romberg公式: R n = 64 63 C 2 n − 1 64 C n R_n = \frac{64}{63}C_{2n}-\frac{1}{64}C_{n} Rn=6364C2n−641Cn
=> R i c h a r d s o n Richardson Richardson外推加速法:
以
{
T
k
0
}
\{T^{0}_{k}\}
{Tk0}表示将区间
[
a
,
b
]
[a,b]
[a,b]二分k次后得到的梯形值序列,
且以
{
T
k
m
}
\{T^{m}_{k}\}
{Tkm}表示经过m次加速后的将区间
[
a
,
b
]
[a,b]
[a,b]二分k次后得到积分值序列,则有以下递推公式:
T k m = 4 m 4 m − 1 T k + 1 m − 1 − 1 4 m T k m − 1 T^m_{k} = \frac{4^m}{4^m-1}T^{m-1}_{k+1}-\frac{1}{4^{m}}T^{m-1}_{k} Tkm=4m−14mTk+1m−1−4m1Tkm−1
设积分区间为 [ − 1 , 1 ] [-1,1] [−1,1],则有 G a u s s − L e g e n d r e Gauss-Legendre Gauss−Legendre公式:
∫ − 1 1 f ( x ) d x = ∑ k = 0 n a k f ( x k ) \int_{-1}^1 f(x)dx = \sum_{k=0}^na_kf(x_k) ∫−11f(x)dx=∑k=0nakf(xk)
其中使得求积公式具有2n+1次代数精度的序列 { x k } \{x_k\} {xk}称为 G a u s s Gauss Gauss点,而该区间上的 G a u s s Gauss Gauss点即是 L e g e n d r e Legendre Legendre多项式 P n + 1 ( x ) P_{n+1}(x) Pn+1(x)的零点
求得 { x k } \{x_k\} {xk}后,系数只需要再满足机械求积公式具有 n n n次代数精度的一般条件:
D n + 1 × ( a 0 , a 1 , . . . , a n ) T = [ b − a , 1 2 ( b 2 − a 2 ) , . . . , 1 n + 1 ( b n + 1 − a n + 1 ) ] T D_{n+1} \times (a_0,a_1,...,a_n)^T = [b-a,\frac{1}{2}(b^2-a^2),...,\frac{1}{n+1}(b^{n+1}-a^{n+1})]^T Dn+1×(a0,a1,...,an)T=[b−a,21(b2−a2),...,n+11(bn+1−an+1)]T
其中: D n + 1 D_{n+1} Dn+1为 { x k } \{x_k\} {xk}构造的n+1阶 V a n d e r m o n d e Vandermonde Vandermonde行列式对应的矩阵, 此处令 b = 1 , a = − 1 b=1,a=-1 b=1,a=−1
=> 对于一般积分区间 [ a , b ] [a,b] [a,b],只需要作变换: x = b − a 2 t + b + a 2 ( t ∈ [ − 1 , 1 ] ) x = \frac{b-a}{2}t+\frac{b+a}{2}(t\in[-1,1]) x=2b−at+2b+a(t∈[−1,1]),此时: ∫ a b f ( x ) d x = b − a 2 ∫ − 1 1 f ( b − a 2 t + b + a 2 ) d t \int_a^b f(x)dx = \frac{b-a}{2} \int_{-1}^1f(\frac{b-a}{2}t+\frac{b+a}{2})dt ∫abf(x)dx=2b−a∫−11f(2b−at+2b+a)dt
=> 两点 G a u s s − L e g e n d r e Gauss-Legendre Gauss−Legendre公式: ∫ − 1 1 f ( x ) d x = f ( − 1 3 ) + f ( 1 3 ) \int_{-1}^1 f(x)dx = f(\frac{-1}{\sqrt 3})+f(\frac{1}{\sqrt 3}) ∫−11f(x)dx=f(3−1)+f(31)
=> 三点 G a u s s − L e g e n d r e Gauss-Legendre Gauss−Legendre公式: ∫ − 1 1 f ( x ) d x = 5 9 f ( − 15 5 ) + 8 9 f ( 0 ) + 5 9 f ( 15 5 ) \int_{-1}^1 f(x)dx = \frac{5}{9}f(\frac{-\sqrt{15}}{5})+\frac{8}{9}f(0)+ \frac{5}{9}f(\frac{\sqrt{15}}{5}) ∫−11f(x)dx=95f(5−15)+98f(0)+95f(515)
=> 积分余项: R ( x ) = f ( 2 n + 2 ) ( 2 n + 2 ) ! ∫ a b ω 2 ( x ) d x R(x) = \cfrac{f^{(2n+2)}}{(2n+2)!}\int_a^b\omega^2(x)dx R(x)=(2n+2)!f(2n+2)∫abω2(x)dx,其中 ω ( x ) = ∏ i = 0 n ( x − x i ) \omega(x) = \prod_{i=0}^n(x-x_i) ω(x)=∏i=0n(x−xi)
=> 对于带权 ρ ( x ) = 1 1 − x 2 \rho(x) = \frac{1}{\sqrt{1-x^2}} ρ(x)=1−x21的积分 ∫ − 1 1 ρ ( x ) f ( x ) d x = ∑ k = 0 n a k f ( x k ) \int_{-1}^1 \rho(x)f(x)dx = \sum_{k=0}^na_kf(x_k) ∫−11ρ(x)f(x)dx=∑k=0nakf(xk),所选取的 G a u s s Gauss Gauss点即为 C h e b y s h e v Chebyshev Chebyshev多项式 T n + 1 ( x ) T_{n+1}(x) Tn+1(x)的零点,系数的求解同理
(已知一系列的观测点 ( x i , y i ) (x_i,y_i) (xi,yi),在给定步长 h h h的情况下,求出观测点处的导数的近似值)
(1)中点公式:
f
′
(
a
)
=
G
(
h
)
=
f
(
a
+
h
)
−
f
(
a
−
h
)
2
h
f'(a) = G(h) = \cfrac{f(a+h)-f(a-h)}{2h}
f′(a)=G(h)=2hf(a+h)−f(a−h), h为待求导点
a
a
a的邻域的半径
(2)两点公式:
f
′
(
x
0
)
=
1
h
[
f
(
x
1
)
−
f
(
x
0
)
]
f'(x_0) = \frac{1}{h}[f(x_1)-f(x_0)]
f′(x0)=h1[f(x1)−f(x0)], 余项:
R
(
x
0
)
=
h
2
f
′
′
(
ξ
)
R(x_0) = \frac{h}{2}f''(\xi)
R(x0)=2hf′′(ξ)
f
′
(
x
1
)
=
1
h
[
f
(
x
1
)
−
f
(
x
0
)
]
f'(x_1) = \frac{1}{h}[f(x_1)-f(x_0)]
f′(x1)=h1[f(x1)−f(x0)], 余项:
R
(
x
1
)
=
−
h
2
f
′
′
(
ξ
)
R(x_1) = -\frac{h}{2}f''(\xi)
R(x1)=−2hf′′(ξ)
其中:
h
=
x
1
−
x
0
h = x_1-x_0
h=x1−x0
(3)三点公式:
f
′
(
x
0
)
=
1
2
h
[
−
3
f
(
x
0
)
+
4
f
(
x
1
)
−
f
(
x
2
)
]
f'(x_0) = \frac{1}{2h}[-3f(x_0)+4f(x_1)-f(x_2)]
f′(x0)=2h1[−3f(x0)+4f(x1)−f(x2)], 余项:
R
(
x
0
)
=
−
h
2
3
f
′
′
′
(
ξ
)
R(x_0) = -\frac{h^2}{3}f'''(\xi)
R(x0)=−3h2f′′′(ξ)
f
′
(
x
1
)
=
1
2
h
[
f
(
x
2
)
−
f
(
x
0
)
]
f'(x_1) = \frac{1}{2h}[f(x_2)-f(x_0)]
f′(x1)=2h1[f(x2)−f(x0)], 余项:
R
(
x
1
)
=
h
2
6
f
′
′
(
ξ
)
R(x_1) = \frac{h^2}{6}f''(\xi)
R(x1)=6h2f′′(ξ)
f
′
(
x
2
)
=
1
2
h
[
f
(
x
0
)
−
4
f
(
x
1
)
+
3
f
(
x
2
)
]
f'(x_2) = \frac{1}{2h}[f(x_0)-4f(x_1)+3f(x_2)]
f′(x2)=2h1[f(x0)−4f(x1)+3f(x2)], 余项:
R
(
x
0
)
=
−
h
2
3
f
′
′
′
(
ξ
)
R(x_0) = -\frac{h^2}{3}f'''(\xi)
R(x0)=−3h2f′′′(ξ)
其中:
h
=
x
2
−
x
1
=
x
1
−
x
0
h = x_2-x_1 = x_1-x_0
h=x2−x1=x1−x0
(4)五点公式:
f
′
(
x
0
)
=
1
12
h
[
−
25
f
(
x
0
)
+
48
f
(
x
1
)
−
36
f
(
x
2
)
+
16
f
(
x
3
)
−
3
f
(
x
4
)
]
f'(x_0) = \frac{1}{12h}[-25f(x_0)+48f(x_1)-36f(x_2)+16f(x_3)-3f(x_4)]
f′(x0)=12h1[−25f(x0)+48f(x1)−36f(x2)+16f(x3)−3f(x4)]
f
′
(
x
1
)
=
1
12
h
[
−
3
f
(
x
0
)
−
10
f
(
x
1
)
+
18
f
(
x
2
)
−
6
f
(
x
3
)
+
f
(
x
4
)
]
f'(x_1) = \frac{1}{12h}[-3f(x_0)-10f(x_1)+18f(x_2)-6f(x_3)+f(x_4)]
f′(x1)=12h1[−3f(x0)−10f(x1)+18f(x2)−6f(x3)+f(x4)]
f
′
(
x
2
)
=
1
12
h
[
f
(
x
0
)
−
8
f
(
x
1
)
+
8
f
(
x
3
)
−
f
(
x
4
)
]
f'(x_2) = \frac{1}{12h}[f(x_0)-8f(x_1)+8f(x_3)-f(x_4)]
f′(x2)=12h1[f(x0)−8f(x1)+8f(x3)−f(x4)]
f
′
(
x
3
)
=
1
12
h
[
−
f
(
x
0
)
+
6
f
(
x
1
)
−
18
f
(
x
2
)
+
10
f
(
x
3
)
+
3
f
(
x
4
)
]
f'(x_3) = \frac{1}{12h}[-f(x_0)+6f(x_1)-18f(x_2)+10f(x_3)+3f(x_4)]
f′(x3)=12h1[−f(x0)+6f(x1)−18f(x2)+10f(x3)+3f(x4)]
f
′
(
x
4
)
=
1
12
h
[
3
f
(
x
0
)
−
16
f
(
x
1
)
+
36
f
(
x
2
)
−
48
f
(
x
3
)
+
25
f
(
x
4
)
]
f'(x_4) = \frac{1}{12h}[3f(x_0)-16f(x_1)+36f(x_2)-48f(x_3)+25f(x_4)]
f′(x4)=12h1[3f(x0)−16f(x1)+36f(x2)−48f(x3)+25f(x4)]
其中:
h
=
x
4
−
x
3
=
x
3
−
x
2
=
x
2
−
x
1
=
x
1
−
x
0
h =x_4-x_3 = x_3-x_2 = x_2-x_1 = x_1-x_0
h=x4−x3=x3−x2=x2−x1=x1−x0