Finite Element Methods Lecture 1 - KTH

6 downloads 201 Views 1MB Size Report
2D1260: Finite Element Methods. Lecturer: Johan ... 2. project presentation (Mon 12 Dec/Tue 13 Dec) ... Space-time FEM, stabilization, Convection-Diffusion. 9.
Finite Element Methods Lecture 1 Johan Hoffman [email protected]

Johan Hoffman – KTH – p.1

2D1260: Finite Element Methods Lecturer: Johan Hoffman Assistant Professor (Forskarassistent) Research: Adaptive FEM for CFD, turbulence,... [email protected] Teaching Assistant: Erik von Schwerin PhD student (Doktorand) Research: Adaptive FEM, Stochastic DE,... [email protected] www.nada.kth.se/kurser/kth/2D1260

Office Hours: Mon 9-10 (JH), Tue 9-10 (ES)

Johan Hoffman – KTH – p.2

2D1260: Finite Element Methods 9 lectures + 6 excercise sessions Book: Computational Differential Equations Examination: 1. project report (Fri 9 Dec) 2. project presentation (Mon 12 Dec/Tue 13 Dec) 3. written exam (Fri 16 Dec) Project starts Week 3 Mon 14 Nov; 2 parts: 1. computer assignement; the same for everyone 2. extension of the first part chosen by the group

Johan Hoffman – KTH – p.3

Overview: Lectures Introduction, DE, Galerkin’s method

2.

BVP, FE basis functions

3.

Abstract problem, Lax-Milgram Theorem

4.

Interpolation, error estimation, adaptivity

5.

FE software; assembly, mapping, quadrature

6.

IVP, stability, -method, space-time FEM

7.

Adaptivity, error estimation, duality

8.

Space-time FEM, stabilization, Convection-Diffusion

9.

Navier-Stokes, adaptivity, functional output



1.

Johan Hoffman – KTH – p.4

 

in



(DE)





Differential Equation (DE)



  





 

: differential operator : solution : source term : domain where (DE) is valid:







  



Ordinary Differential Equation (ODE) : Partial Differential Equation (PDE) : u scalar : scalar DE u vector : system of DE











DE : Equation relating derivatives of a function

Johan Hoffman – KTH – p.5





BC: specifying equation on boundary

of the domain





on





 

in





(BC)







(DE)

 

Boundary conditions











   

 

for all





for all





  

  

Neumann boundary condition:

 

 



Dirichlet boundary condition:

  

Examples:

Johan Hoffman – KTH – p.6





is two



 



This equation smooths the solution: the solution times more differentiable than the source term

  







 

























Ex: Poisson’s equation

Modeling: gravitation, ground water flow, electrostatics,...

Johan Hoffman – KTH – p.7













for







and







for



 



BC:



Dirichlet BC

Johan Hoffman – KTH – p.8













for





and

 





for



 



BC:



 

Dirichlet + Neumann BC

Johan Hoffman – KTH – p.9



 























 

Ex: Heat equation















 

 



































This equation smooth the solution over time:

Modeling: heat conduction, pollution,...

Johan Hoffman – KTH – p.10

Ex: Heat equation

Johan Hoffman – KTH – p.11

Ex: Heat equation Osmosis; diffusion through cell membrane

Johan Hoffman – KTH – p.12



















Ex: Schrödinger equation







is the wave function in the quantum mechanical model of the motion of an electron orbiting around one proton at the origin.

Johan Hoffman – KTH – p.13

 



























Ex: Black-Scholes equation













This equation is close to the heat equation, with the asset playing the role of the spatial dimension price Modeling: option pricing,...

Johan Hoffman – KTH – p.14

Ex: Linear Elasticity



 







and the stress

  



Models the displacement elastic bodies

  

div













div



Cauchy-Navier’s elasticity equations:

for

Johan Hoffman – KTH – p.15







):

 













 



 

















This equation conserves energy (for













Ex: Wave equation

Modeling: wave phenomena, accoustics,...

Johan Hoffman – KTH – p.16

Ex: Wave equation Seismic waves in simulation of California Earthquake

Johan Hoffman – KTH – p.17

 

 



is transported (convected) by the .



  





The solution convection field













 

Ex: Transport equation

Modeling: Pollution,...

Johan Hoffman – KTH – p.18



 



















Ex: Convection-Diffusion-Reaction













  







 







  

 

  

This equation is a combination of transport (convection) by , diffusion with diffusivity , the convection field , of a spieces and reaction with reaction coefficient . Modeling: chemical reactions, pollution,...

Johan Hoffman – KTH – p.19

Ex: Convection-Diffusion-Reaction Chernobyl 1986; simulation by SMHI

Johan Hoffman – KTH – p.20

       



    

 











 







Ex: Maxwell equations

magnetic field,

current density





electric field,

Johan Hoffman – KTH – p.21

Ex: Maxwell equations Magnetic field around a coil.

Johan Hoffman – KTH – p.22



 



















Ex: Stokes equations

velocity and

pressure

Modeling: low velocity flow phenomena

Johan Hoffman – KTH – p.23

Ex: Stokes equations Groundwater flow

Johan Hoffman – KTH – p.24





 

 











 







Ex: Navier-Stokes equations

velocity and

pressure

Modeling: flow phenomena, weather prediction, blood flow,...

Johan Hoffman – KTH – p.25

Ex: Navier-Stokes equations

Johan Hoffman – KTH – p.26

 

Vorticity



Ex: Navier-Stokes equations around wheel.

Johan Hoffman – KTH – p.27

 

Vorticity



Ex: Navier-Stokes equations around a full car.

Johan Hoffman – KTH – p.28

 

Vorticity



Ex: Navier-Stokes equations around a full car.

Johan Hoffman – KTH – p.29

Ex: Navier-Stokes equations Blood flow in artery.

Johan Hoffman – KTH – p.30

Function spaces to DE





 



and

in

)



and all functions



(for all real numbers



  















  





 







  

in a certain class

is a vector space (linear space):







  









A function space



look for approximate solutions of functions: FUNCTION SPACE





Often impossible to find exact solutions

Functional Analysis: linear algebra for function spaces

Johan Hoffman – KTH – p.31





 

 





















 

 



   







  









 



 



 









 



Hölder inequality:

 

  







    









    







-norm:







 



max-norm:

 

orthogonal: 







Cauchy-Schwarz inequality: and

(size of )





  







-norm:







Inner product (scalar product):





Linear Algebra for Function spaces

Johan Hoffman – KTH – p.32

 

)

 













  





















so that the error is small:



Determine















  

















 

 



 

 









in a function space

 

Seek approximate solution with “simple” basis functions:



Approximation methods



Examples of spaces : trigonometric functions (Fourier), polynomials (Newton, Lagrange), piecewise polynomials (Finite Element Methods)

Johan Hoffman – KTH – p.33



















 

 









such that





the function space

is orthogonal to all













 



 







:



 











  



The residual of the exact solution functions:









The RESIDUAL: is orthogonal to all testfunctions GALERKIN ORTHOGONALITY













  















Galerkin method: Find



Galerkin method

Johan Hoffman – KTH – p.34

  



  

















 









, 













 





























 





 







for 





linear system of equations









  



Solve for

















  











 

















 





  

























































Galerkin method

Johan Hoffman – KTH – p.35

Galerkin method spectral method (trigonomatric b.f.), p-method (global polynomial b.f.), h-method (piecewise polynomial b.f.),...

FEM: piecewise polynomial functions on a mesh in 1d

Johan Hoffman – KTH – p.36

Galerkin method

FEM: computational mesh in 3d (www.geuz.org/gmsh)

Johan Hoffman – KTH – p.37

Finite Element Methods FEM: piecewise polynomials 1. basis functions are almost orthogonal (local support) 2. basis functions are simple to differentiate and integrate 3. applicable to general geometry ϕ 1

i−1

ϕ

i

ϕ

i+1

x x i−2

x i−1

xi

x i+1

x i+2

Johan Hoffman – KTH – p.38





 







Dirichlet Boundary conditions:























 



  

Example

boundary nodes h i=x i−xi−1 interior nodes I i=(x i−1,x i) x x2

x i−1

xi

xM+1 =1







x1



x 0=0

continuous piecewise linear functions on a mesh

Johan Hoffman – KTH – p.39









 

 









 







  

 

 





  

 

  





   









 







for  









  



,

   



 









:







Integration by parts



 







  

 

 



nodal basis for





 

  













 









 























Example

Variational form:

weak form of DE

Johan Hoffman – KTH – p.40











 







 



















  





  





 





 





  





for

  



, Vector 





Linear system of equations for :





  





  







 







  















  





  









 

 







  

 

 





  

 

  











Galerkin (cG(1)): Find









Matrix 

















 

 







 

 







Example

:

Johan Hoffman – KTH – p.41

Goal of FEM course Construct Finite Element Methods for general DE (ODE and PDE). Analyze stability of DE, and the corresponding FEM solution. Analyze the error in the FEM approximation. Construct adaptive FEM methods. Learn the structure of FEM software.

Johan Hoffman – KTH – p.42