lec13 (1)

# lec13 (1) - Introduction to Simulation - Lecture 13...

This preview shows pages 1–11. Sign up to view the full content.

Introduction to Simulation - Lecture 13 Convergence of Multistep Methods Jacob White Thanks to Deepak Ramaswamy, Michal Rewienski, and Karen Veroy

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Outline Small Timestep issues for Multistep Methods Local truncation error Selecting coefficients. Nonconverging methods. Stability + Consistency implies convergence Next Time Investigate Large Timestep Issues Absolute Stability for two time-scale examples. Oscillators.
Multistep Methods General Notation Basic Equations () ( () , () ) d xt f xt ut dt = Nonlinear Differential Equation: k-Step Multistep Approach: 00 ˆˆ , kk lj jj l j xtf x u t αβ −− == =∆ ∑∑ Solution at discrete points Time discretization Multistep coefficients 2 ˆ l x l t 1 l t 2 l t 3 l t lk t ˆ l x 1 ˆ l x ˆ x

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Multistep Methods Common Algorithms Basic Equations () 00 ˆˆ , kk lj jj l j xtf x u t αβ −− == =∆ ∑∑ Multistep Equation: ( ) ( ) ( ) ( ) ( ) 11 1 , ll xt tf xt ut ≈+ FE Discrete Equation: ( ) ( ) 1 ˆ , l l xx t f xu t −= 01 0 1 1, 0, 1 k α β = −= = Forward-Euler Approximation: Multistep Coefficients: BE Discrete Equation: Multistep Coefficients: 0 1 0 k = ( ) 1 ˆ , l l t f x u t Trap Discrete Equation: ( ) 1 ˆ ˆ ,, 2 l l t x x fxut fx ut + 0 1 , 22 k αα = = Multistep Coefficients:
Multistep Methods Definitions and Observations Basic Equations () 00 ˆˆ , kk lj jj l j xtf x u t αβ −− == =∆ ∑∑ Multistep Equation: 0 1) If 0 the multistep method is implicit β≠ 2) A step multistep method uses previous ' and ' x s f s 0 3) A normalization is needed, 1 is common α= 4) A -step method has 2 1 free coefficients + How does one pick good coefficients? Want the highest accuracy

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Multistep Methods Simplified Problem for Analysis • Nonlinear Analysis has many unrevealing subtleties • Scalar is equivalent to vector for multistep methods. Why such a simple Test Problem? Le t () Ey t x t = 1 00 ˆˆ kk lj jj y tE A E y αβ −− == =∆ ∑∑ 1 n yt y λ ⎡⎤ ⎢⎥ ⎣⎦ % Decoupled Equations d xt A dt = multistep discretization xtA x () ( ) 0 , 0 d vt vt v v dt λλ ^ Scalar ODE:
Multistep Methods Simplified Problem for Analysis () ( ) 0 () , 0 d vt vt v v dt λλ == ^ Must Consid ALL er Scalar ODE: 00 ˆˆ kk lj jj αβλ −− =∆ ∑∑ Scalar Multistep formula: λ∈ ^ Growing Solutions Decaying Solutions O s c i l l a t i o n s ( ) Im λ ( ) Re

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Multistep Methods Convergence Analysis Convergence Definition Definition: A multistep method for solving initial value problems on [0,T] is said to be convergent if given any initial condition ( ) 0, ˆ max 0 as t 0 l T l t vv l t ⎡⎤ ⎢⎥ ⎣⎦ ∆→ exact v t ˆ computed with 2 l v ˆ computed with t l v