Tutorial 10
Transcription
Tutorial 10
Platzhalter für Bild, Bild auf Titelfolie hinter das Logo einsetzen Introduction to PDEs and Numerical Methods Tutorial 10. Hyperbolic equations– numerical solution Dr. Noemi Friedman, 23.01.2015. Overview of this tutorial Introduction (classification of PDEs, examples of important PDEs) 1D – Poisson equation −𝑢′′ 𝑥 = 𝑓(𝑥) Analytical solution (existence, uniqueness of solution, stability) Finite difference approximation of the Poisson eqaution Eigenvalue problem, Fourier transform Elliptic equations (−𝑎𝑎′′ 𝑥 + 𝑏 𝑥 𝑢′ 𝑥 + 𝑐 𝑥 𝑢 𝑥 = 𝑓 𝑥 ) Analytical solution: existence, and uniqueness of solution, and its bounds FD approximation Consistency, stability, convergence FD approximation of PDEs on non-equidistant grids Higher order FD schemes With Taylor expansion With Lagrange interpolation Compact schemes 23. 01. 2015. | Dr. Noemi Friedman | PDE tutorial | Seite 2 Overview of this tutorial Schemes for the numerical approximation of initial value problems Explicit Euler Implicit (backward) Euler Theta-methods Runge-Kutta methods Heat eqaution with 1d spatial dimension analytical solution (seperation of variables) Spatial and time discretisation, consistency and stability analysis Transport and wave equations with 1d spatial dimension analytical solution numerical soluion Multidimensional problems 23. 01. 2015. | Dr. Noemi Friedman | PDE tutorial | Seite 3 Numerical solution of transport equation – downwind method 𝜕𝑢(𝑥, 𝑡) 𝜕𝜕 +𝑐 =0 𝜕𝑡 𝜕𝜕 Numerical approximation 𝑡 1 𝑐 𝑢 − 𝑢𝑛,𝑗 + 𝑢𝑛,𝑗+1 − 𝑢𝑛,𝑗 = 0 ∆𝑡 𝑛+1,𝑗 ℎ 𝑐∆𝑡 𝑐∆𝑡 𝑖𝑘𝑘 − 𝑒 ℎ ℎ for 𝑐 > 0 unconditionally instable 𝐺 𝑘 =1+ transport direction 𝑐>0 𝑥 23. 01. 2015. | Dr. Noemi Friedman | PDE tutorial | Seite 4 downwind upwind 𝑡 Numerical solution of transport equation – upwind method 𝜕𝑢(𝑥, 𝑡) 𝜕𝜕 +𝑐 =0 𝜕𝑡 𝜕𝜕 Numerical approximation 𝑡 backward differences 1 𝑐 𝑢 − 𝑢𝑛,𝑗 + 𝑢𝑛,𝑗 − 𝑢𝑛,𝑗−1 = 0 ∆𝑡 𝑛+1,𝑗 ℎ 𝐺 𝑘 =1− for 𝑐 > 0 stable when 𝑐∆𝑡 𝑐∆𝑡 𝑖𝑘𝑘 + 𝑒 ℎ ℎ (CFL) transport direction 𝑐>0 𝑥 23. 01. 2015. | Dr. Noemi Friedman | PDE tutorial | Seite 5 downwind upwind 𝑡 Numerical solution of transport equation – NAIVE scheme 𝜕𝑢(𝑥, 𝑡) 𝜕𝜕 +𝑐 =0 𝜕𝑡 𝜕𝜕 𝑡 1 𝑐 𝑢 − 𝑢𝑛,𝑗 + 𝑢 − 𝑢𝑛,𝑗−1 = 0 ∆𝑡 𝑛+1,𝑗 2ℎ 𝑛,𝑗+1 𝐺 𝑘 =1− 𝑐∆𝑡 𝑖 sin(𝑘𝑘) ℎ 𝑥 unconditionally instable 23. 01. 2015. | Dr. Noemi Friedman | PDE tutorial | Seite 6 𝑡 Numerical solution of transport equation –Lax Friedrich’s method 𝜕𝑢(𝑥, 𝑡) 𝜕𝜕 +𝑐 =0 𝜕𝑡 𝜕𝜕 from central differences: 1 𝑢 + 𝑢𝑛,𝑗+1 2 𝑛,𝑗−1 1 𝑐 𝑢𝑛+1,𝑗 − 𝑢𝑛,𝑗 + 𝑢𝑛,𝑗+1 − 𝑢𝑛,𝑗−1 = 0 ∆𝑡 2ℎ 1 1 𝑢 − 𝑢 + 𝑢𝑛,𝑗+1 ∆𝑡 𝑛+1,𝑗 2 𝑛,𝑗−1 𝑐∆𝑡 𝑖 sin(𝑘𝑘) ℎ 2 𝑐∆𝑡 = cos2 (𝑘𝑘) − sin2 (𝑘𝑘) ℎ 𝐺 𝑘 = cos(𝑘𝑘) − 𝐺 𝑘 2 𝑥 𝑐 + 𝑢 − 𝑢𝑛,𝑗−1 = 0 2ℎ 𝑛,𝑗+1 𝐺 𝑘 2 ≤1 𝑐∆𝑡 ≤1 ℎ 23. 01. 2015. | Dr. Noemi Friedman | PDE tutorial | Seite 7 𝑡 Numerical solution of transport equation – Lax-Wendroff method 𝜕𝑢(𝑥, 𝑡) 𝜕𝜕 +𝑐 =0 𝜕𝑡 𝜕𝜕 Approximate second derivate with the three point stencil (spatial discretisation of the heat eq.) from Fridrich’s scheme 𝑥 from central differences 23. 01. 2015. | Dr. Noemi Friedman | PDE tutorial | Seite 8 𝑡 Hints for next assignment ≤1 23. 01. 2015. | Dr. Noemi Friedman | PDE tutorial | Seite 9 Hints for next assignment 1) Get scheme from plugging in the first equation to the second one 2) If you can bring it to the form: then the scheme is conservative 3) If the scheme is consistent 23. 01. 2015. | Dr. Noemi Friedman | PDE tutorial | Seite 10