aboutsummaryrefslogtreecommitdiff
path: root/content/numerik_dgl.tex
blob: 0f573e4ca47420ef0de38413f7cca42263ca2ece (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
\section*{Existenzsatz von Peano}

Sei $f : G \subset \R \times \R^n \to \R^n$ stg., $G$ Gebiet.

Dann $\forall (\tilde x,\tilde y) \in G \exists$ Lösung $y'=f(x,y)$ im Gebiet.

\section*{Existenz- und Eindeutigkeit}

Sei $f : S \to \R^n$ stg. auf Steifen $S := [a,b] \times \R^n$ und $f$ erfülle die Lipschitz-Bedingung:

\vspace*{-2mm}
$$\|f(x,y)-f(x,\tilde y)\| \leq L \|y-\tilde y\|, \ L > 0$$

Dann existiert für das AWP genau eine Lösung in $\mathcal{C}([a,b],\R^n)$ für jedes Element in $S$.

\section*{Einzelschrittverfahren}

Sei $f : [a,b] \times \R^n \to \R^n, \ y(x_0)=y_0 \in \R^n$ AWP.

Ein \emph{Einschrittverfahren} ist Vorschrift:

\vspace*{-4mm}
$$\eta_0 = y_0, \ \eta_{k+1} = \eta_k + h \cdot \Phi(x_k, \eta_k, h), \ x_{k+1} = x_k + h$$

Für \emph{Verfahrensfunktion} $\Phi : [a,b] \times \R^n \times \R \to \R^n$.

\spacing

Die \emph{Näherungslösung} $\eta_k$ ist eine \emph{Gitterfunktion}.

\vspace*{-4mm}
$$\eta_k : \{x \in [x_0,b] | x = x_0 + i \cdot h, \ i \in \N_0 \} \to \R^n$$

\subsection*{Explizites Eulerverfahren}

$$\Phi(x,y,h) := f(x,y) \text{ d.h. Butcher-Schema: } \begin{array}{c|c}
0 & 0 \\
\hline
  & 1
\end{array}$$

\subsection*{Implizites Eulerverfahren}

$$\Phi(x,y,h) := f(x+h, g(x,y,h)), \ g = y + h \cdot f(x+h, g)$$

Mit Butcher-Schema: $\begin{array}{c|c}
1 & 1 \\
\hline
  & 1
\end{array}$


\subsection*{Konsistenz}

Sei $z$ mit $z'(x) = f(x,z(x))$ die exakte AWP Lösung.

Der \emph{lokale Diskretisierungsfehler} in $(x,y)$:

$$\tau(x,y,h) := \frac{z(x+h)-y}{h} - \Phi(x,y,h)$$

Ein ESV ist brauchbar, wenn $\lim_{h \to 0} \tau(x,y,h) = 0$ bzw. $\lim_{h \to 0} \Phi(x,y,h) = f(x,y)$.

\subsubsection*{Konsistenzordnung}

ESV mit $\Phi$ ist \emph{konsistent mit Ordnung $p$}, falls:

\vspace*{-2mm}
$$\tau(x,y,h) \in \mathcal{O}(h^p) \text{ für } h \to 0$$

Hierzu ist eine Taylorentwicklung von $z$ hilfreich.

Beide Eulerverfahren haben Ordnung $1$.

\subsection*{Allgemeiner Ansatz für Ordnung $2$}

$$\Phi(x,y,h) := a_1 \cdot f(x,y) + a_2 \cdot f(x+p_1 h, y+p_2 h \cdot f(x,y))$$

für Konstanten $a_1, a_2, p_1, p_2 \in \R$.

\spacing

Bedingungen: $a_1 + a_2 = 1, \ a_2 p_1 = \frac{1}{2}, \ a_2 p_2 = \frac{1}{2}$

\subsubsection*{Verfahren von Heun}

$$\Phi(x,y,h) := \frac{1}{2}(f(x,y) + f(x+h, y+h \cdot f(x,y)))$$

\subsubsection*{Verfahren von Runge}

$$\Phi(x,y,h) := f(x + \frac{h}{2}, y + \frac{h}{2} f(x,y) )$$

\subsubsection*{Implizite Trapezregel}

\vspace*{-2mm}
\begin{align*}
	\Phi(x,y,h) &:= \frac{1}{2} (f(x,y) + f(x+h, g(x,y,h)) \\
	g(x,y,h) &:= y + \frac{h}{2} (f(x,y) + f(x+h,g(x,y,h))
\end{align*}

\subsection*{Konvergenz}

Der \emph{globale Diskretisierungsfehler} für $x \in [a,b]$:

\vspace*{-4mm}
$$e(x,h_n) := \eta(x,h_n) - y(x), \ h_n=h_n(x)=\frac{x-x_0}{n}, n \in \N_0$$

Ein ESV ist \emph{konvergent}, falls:

\vspace*{-4mm}
$$\forall x \in [a,b], \text{hinr. glatte } f : \lim_{n \to \infty} e(x,h_n) = 0$$

\section*{Autonomisierung}

$$\eta := \begin{pmatrix}x \\ y\end{pmatrix} \in \R^{n+1}, \ \widehat{f} : \R^{n+1} \to \R^{n+1}, \eta \mapsto \begin{pmatrix}1 \\ f(x,y)\end{pmatrix}$$

AWP $\eta'=\widehat{f}(\eta)$ mit Bedingung: $\eta(0) = \begin{pmatrix}x_0 \\ y_0\end{pmatrix}$

\subsection*{Invarianz gegen Autonomisierung}

ESV $\Phi$ ist \emph{invariant gegen Autonomisierung}, wenn:

$$\widehat{\Phi}_1(\eta,h)=1, \ \Phi(x,y,h) = \widehat{\Phi}_2(\eta,h), \ \eta = \begin{pmatrix}x \\ y\end{pmatrix}$$

Wobei $\widehat{\Phi} = \begin{pmatrix}\widehat{\Phi}_1, \widehat{\Phi}_2\end{pmatrix}$ die Anwendung von $\Phi$ auf das autonomisierte System ist.

\section*{Explizite Runge-Kutta-Verfahren}

Verfahrensfunktion $\Phi$ eines $s$-stufigen RKV:

\vspace*{-4mm}
\begin{align*}
\Phi(x,y,h) &:= b_1 k_1 + b_2 k_2 + \cdots + b_s k_s \\
k_i &:= f(x+c_i h, y + h \sum_{j=1}^{i-1} a_{i,j} k_j)
\end{align*}
\vspace*{-8mm}

\subsection*{Butcher-Schema}

Darstellung der Koeffizienten $b_i, c_i$ und $a_{i,j}$:

$$\begin{array}{c|c}
c & A \\
\hline
  & b^\intercal
\end{array}$$

Hierbei ist $A$ strikte untere Dreiecksmatrix.

\subsection*{Konsistentsbedingung mit Ordnung 1}

Konsistent mit Ordnung 1 gdw. $\displaystyle\sum_{i=1}^s b_i = 1$

Für die Ordnung $p$ eines $s$-stufigen RKV: $p \leq s$

\subsection*{Invarianzbedingung}

RKV ist invariant gegen Autonomisierung gdw. es konsistent und $c_i$ die $i$-te Zeilensumme von $A$ ist:

$$\sum_{i=1}^s b_i = 1 \text{ und } \sum_{j=1}^{i-1} a_{i,j} = c_i \text{ für } i=1,\dots,s$$

Somit genügt $(b, A)$ zur Definition von gegenüber Autonomisierung invarianter RKV.

\subsection*{Konsistenzbedingung für Ordnung $2$}

$$\sum_{i=1}^s b_i = 1 , \sum_{i=1}^s b_i c_i = \frac{1}{2}$$

\subsection*{Konsistenzbedingung für Ordnung $3$}

$$\sum_{i=1}^s b_i = 1 , \sum_{i=1}^s b_i c_i = \frac{1}{2} , \sum_{i=1}^s b_i c_i^2 = \frac{1}{3} , \sum_{i,j=1}^s b_i a_{i,j} c_j = \frac{1}{6}$$

\section*{Explizite Extrapolationsverfahren}

\section*{Mehrschrittverfahren}

\section*{Partielle Differentialgleichungen}

\subsection*{Finite Differenzen}