Title: Comparison of different time discretization methods explicitimplicit
1Comparison of different time discretization
methodsexplicit-implicit
- SIWIR-Seminar, 12.01.2009, Jana Hutter
2Contents
- Definition Initial value problem, Time
discretization - Intuitive try Euler-procedure
- Explicit one-step procedures
- Implicit one-step procedures
- Example Crank-Nicolson procedure
- Comparision with regard to convergence and
stability
3Initial value problem
- Initial value problem for a system of ordinary
differential equations of first degree
4Time Discretization
We define a grid in the following way
5Intuitive and easy try
Now all values y can be calculated
successively -gt This is the explicit Euler
procedure (Polygonzug-Verfahren) (Which looks
good on the first look, but .)
6Evaluation of a procedure I
- Errors made by the approximation from one step in
time to the next one should be very small - Consistence
- The local errors shouldnt be accumulated
Stability
7Evaluation of a procedure II
- y exact solution of the initial value problem
- yh approximation on the grid
- Convergence
- Consistence
8Back to the explicit Euler procedure
- It can be shown that Euler has a convergence and
consistence degree of order 1, which is really
bad! - -gt Back on the road for better procedures
9Next try general explicit one-step procedures
Example Runge-Kutta methods
10Results
Degree h2
Degree h
11Problem Stiff ODEs I
- general problem with explicit methods
- - perhaps higher degree of convergence than
Euler (Heun h2) - -but big problems with so called stiff
differential equations (Real values of negative
Eigenvalues very different)
12Problem Stiff ODEs II
- Example
- Solution
- Explicit/Implicit Euler method
- -gtBlackboard
- Better
- Methods which offer qualitative right solution
independent from - the value of h (so called A-Stability)
13Example
14Implicit one-step procedures
Difference to explicit methods To calculate
yj1, not only the already known values
y0,y1,,yi are used, but also the (unknown) value
yi1! -gt Ín every step a linear equation system
has to be solved Example implicit Euler-method
15Implicit Examples
The implicit Euler method can be written in the
following way (with Lamda0)
With Lamda0.5, we get the implicit
Crank-Nicolson method
16Overview
- Implicit methods
- Examples
- implicit Euler, Cranc-Nicolson
- -Solving of a linear equations system
- stable
- Explicit methods
- Examples
- explicit Euler, Heun
- easy to calculate
- - Not A-stable