Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
j =1
a
ij
f (x
n1
+ c
j
h, Y
j
) , (2a)
y
n
= y
n1
+ h
s
j =1
b
j
f (x
n1
+ c
j
h, Y
j
). (2b)
Y
i
represent the internal stage value,
y
n
represent the update of y at the n
th
step.
y
0
y
n
x
0
x
n
| | |
Y
< >
h
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Runge-Kutta methods
We can display (2a) and (2b) by a Butcher tableau of the form
c A
b
T
,
c is the vector of abscissae, Matrix A are the coecients, Vector b
represents the quadrature of weight indication.
Where
c
i
=
s
j =1
a
ij
, i = 1, 2, . . . , s. (3)
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Runge-Kutta methods
1
Divided into two main types according to the style of the
matrix A
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Runge-Kutta methods
1
Divided into two main types according to the style of the
matrix A
Explicit methods - Where the matrix A is strictly lower
triangular.
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Runge-Kutta methods
1
Divided into two main types according to the style of the
matrix A
Explicit methods - Where the matrix A is strictly lower
triangular.
Implicit methods:-
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Runge-Kutta methods
1
Divided into two main types according to the style of the
matrix A
Explicit methods - Where the matrix A is strictly lower
triangular.
Implicit methods:-
Fully implicit - Matrix A is not lower triangular,
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Runge-Kutta methods
1
Divided into two main types according to the style of the
matrix A
Explicit methods - Where the matrix A is strictly lower
triangular.
Implicit methods:-
Fully implicit - Matrix A is not lower triangular,
Semi-implicit - Matrix A is lower triangular with at least one
non-zero diagonal element,
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Runge-Kutta methods
1
Divided into two main types according to the style of the
matrix A
Explicit methods - Where the matrix A is strictly lower
triangular.
Implicit methods:-
Fully implicit - Matrix A is not lower triangular,
Semi-implicit - Matrix A is lower triangular with at least one
non-zero diagonal element,
Diagonal implicit - Matrix A is a lower triangular with all
elements are equal and non-zero,
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Runge-Kutta methods
1
Divided into two main types according to the style of the
matrix A
Explicit methods - Where the matrix A is strictly lower
triangular.
Implicit methods:-
Fully implicit - Matrix A is not lower triangular,
Semi-implicit - Matrix A is lower triangular with at least one
non-zero diagonal element,
Diagonal implicit - Matrix A is a lower triangular with all
elements are equal and non-zero,
Singly implicit - Matrix A is a non-singular matrix with single
eigenvalue.
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Explicit Methods
1
Example of explicit methods
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Explicit Methods
1
Example of explicit methods
Euler method
0 0
1
,
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Explicit Methods
1
Example of explicit methods
Euler method
0 0
1
,
Midpoint method
0 0 0
1
2
1
2
0
0 1
,
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Explicit Methods
1
Example of explicit methods
Euler method
0 0
1
,
Midpoint method
0 0 0
1
2
1
2
0
0 1
,
Trapezoidal rule
0 0 0
1 1 0
1
2
1
2
.
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Implicit Methods
1
3 Family of implicit methods
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Implicit Methods
1
3 Family of implicit methods
Radau methods - Implicit Euler
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Implicit Methods
1
3 Family of implicit methods
Radau methods - Implicit Euler
Gauss methods - Implicit midpoint
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Implicit Methods
1
3 Family of implicit methods
Radau methods - Implicit Euler
Gauss methods - Implicit midpoint
Lobatto IIIA - Implicit Trapezoidal
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Implicit Methods
1
3 Family of implicit methods
Radau methods - Implicit Euler
Gauss methods - Implicit midpoint
Lobatto IIIA - Implicit Trapezoidal
2
Example of implicit methods
Table: Family of Implicit Method
1 1
1
1
2
1
2
1
1
1
2
1
2
1
2
1
2
Implicit Euler Implicit Midpoint Rule Implicit Trapezoidal Rule
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Explicit vs Implicit
Table: Explicit vs Implicit
Subject Explicit Implicit
Implementation Very easy Expensive and dicult
Sti problem Cannot solve Can solve
Stability Poor Better Stability
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Explicit vs Implicit:Implementation
Example: Lets y
0
= 1
Explicit Euler:
y
1
= y
0
+ hf (y
0
)
Implicit Midpoint rule:
Y = y
0
+
h
2
f (x
0
+
1
2
h, Y)
y
n
= y
0
+ hf (Y)
*Hard to solve nonlinear equations for implicit midpoint rule.
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Explicit vs Implicit:Stiness
By applying Dahlquist test problem to any methods,
y
= 1000y + sin(x)
By applying implicit Runge-Kutta methods, the problem can be
solved. Explicit methods also can solve the problem, but we need
to use small h which will have lot of iterations.
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Example of implementation of Holsapple, Iyer and
Domans problem using explicit method with h=0.01
5 4.5 4 3.5 3 2.5 2
14
12
10
8
6
4
2
0
Log of stepsize
L
o
g
o
f
E
r
r
o
r
Holsapple, Iyer & Doman problem for explicit method
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Example of implementation of Holsapple, Iyer and
Domans problem using Implicit method with h=0.01
3.6 3.4 3.2 3 2.8 2.6 2.4 2.2 2
11
10.5
10
9.5
9
8.5
8
7.5
Log of stepsize
L
o
g
o
f
E
r
r
o
r
Implicit RK method for Hosaple
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Explicit vs Implicit:Stability
Applying Dahlquist test equation on an s-stage Runge-Kutta
methods yields,
Y = ey
0
+ hAF(Y),
= ey
0
+ hqAY,
= (I zA)
1
ey
0
,
y
n
= y
0
+ hb
T
F(Y),
= y
0
+ hb
T
(qY),
= y
0
+ zb
T
(I zA)
1
ey
0
,
= (1 zb
T
(I zA)
1
)ey
0
,
y
n
= R(z)y
0
,
where e is a vector of unit 1. The stability function is therefore
dened by
R(z) = 1 zb
T
(I zA)
1
e.
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Explicit vs Implicit:Stability
For explicit Euler methods, stability function is
R(z) = 1 zb
T
(I zA)
1
e = 1 + z
It is show polynomial function and does not satisfy A-stable
condition where R(z) 1.
For implicit methods, stability function is
R(z) = 1 zb
T
(I zA)
1
e =
1
1 z
Implicit Euler method show rational function and will satisfy
A-stable condition.
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Extrapolation
1
Consider numerical integration of an initial value problem over
the time interval from t = t
0
to t = t
n
. If h is the constant
step size used, then the number of steps taken is given by
n =
t
n
h
.
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Extrapolation
1
Consider numerical integration of an initial value problem over
the time interval from t = t
0
to t = t
n
. If h is the constant
step size used, then the number of steps taken is given by
n =
t
n
h
.
2
Let y
h
(t
n
) denote the numerical approximation of a given
method at time t
n
using stepsize h and yh
2
(t
n
) the numerical
solution at time t
n
using stepsize
h
2
.
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Extrapolation
1
Consider numerical integration of an initial value problem over
the time interval from t = t
0
to t = t
n
. If h is the constant
step size used, then the number of steps taken is given by
n =
t
n
h
.
2
Let y
h
(t
n
) denote the numerical approximation of a given
method at time t
n
using stepsize h and yh
2
(t
n
) the numerical
solution at time t
n
using stepsize
h
2
.
3
Extrapolation means apply the linear combination of y
h
(t
n
)
and yh
2
(t
n
) to eliminate error term.
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Extrapolation
Extrapolation formula is given as
2
p
yh
2
(t
n
) y
h
(t
n
)
2
p
1
(4)
as extrapolation formula for rst level.
By applying extrapolation, 2
nd
methods resulted order 3, if the
methods is unsymmetrical methods and order 4 if it is symmetry.
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Chapter 3:Methodology
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Chapter 4: Numerical Analysis
1
Some studies about explicit RK methods: Accuracy and
eciency graph.
Amira binti Ismail A STUDY OF HIGHER ORDER IMPLICIT RUNGE-KUTTA METHODS
Chapter 4: Numerical Analysis
1
Some studies about explicit RK methods: Accuracy and
eciency graph.
2
Implementation and extrapolation of implicit method for
Prothero Robinsons problem:
y