Esquemas Para Ec. de Convedascción

download Esquemas Para Ec. de Convedascción

of 8

Transcript of Esquemas Para Ec. de Convedascción

  • 8/18/2019 Esquemas Para Ec. de Convedascción

    1/8

    Computational Fluid Dynamics 

    Wave Equation

    Grétar Tryggvason Spring 2013 

    http://www.nd.edu/~gtryggva/CFD-Course/  

    Computational Fluid Dynamics 

    ! 2

     f 

    ! t 2  " c

    2 ! 2

     f 

    !  x2= 0

    First write the equation as a system of first order equations 

    u =!  f 

    ! t ;   v =

    !  f 

    !  x;

     

    ! u

    ! t " c

    2 ! v

    !  x = 0

    ! v

    ! t "

    ! u

    !  x = 0

    yielding 

    Introduce 

    from the pde 

    since ! 

    2 f 

    ! t !  x=

    ! 2 f 

    !  x! t 

    Wave equation 

    Computational Fluid Dynamics 

    ! u

    ! t " c

    2 ! v

    !  x = 0

    ! v

    ! t "

    ! u

    !  x = 0

    ! u

    ! t 

    ! v

    ! t 

    "

    #

    $$$$

    %

    &

    ''''

    +  0   (c2

    (1 0

    "

    #$

    %

    &'

    ! u

    !  x 

    ! v

    !  x 

    "

    #

    $$$$

    %

    &

    ''''

    =  0

    det   AT  !" I( ) = det

      !"    !1

    !c2 !" 

    #

    $%%

    &

    '((=   " 

    2 !  c2( ) = 0

     

    !"  = ±c

    !  =1

    2ab ±   b2

    " 4ac( )

     

    b = 0;   a =1;   c =!c2

    We can also use 

    with 

    To find the characteristics 

    Wave equation 

    Computational Fluid Dynamics 

    t  

     x 

    Two characteristic lines dt 

    dx= +

    1

    c;  dt 

    dx= !

    1

    c

     

    dt 

    dx= +

    1

    c

     

    dt 

    dx= !

    1

    c

    Wave equation 

    Computational Fluid Dynamics 

    l1

    ! u

    ! t "  c2

     ! v

    !  x=  0

    # $ %

      & ' ( 

    +l2

    ! v

    ! t  "

     ! u

    !  x =  0

    # $ %

      & ' ( 

    !"    !1

    !c2 !" 

    #

    $%%

    &

    '((

    l1

    l2

    #

    $

    %%

    &

    '

    ((= 0

     

    !"  =   cTo find the solution we needto find the eigenvectors 

    Take l1=1 

    !c l1!  l

    2= 0   "   l

    2=!cFor  ! = +c

     

    +c l1!  l

    2= 0   "   l

    2= +cFor  !  ="c

    Wave equation 

    Computational Fluid Dynamics 

    1 ! u

    ! t " c2

     ! v

    !  x = 0

    #  $  % 

    &  '  (  " c

    ! v

    ! t " ! u

    !  x = 0

    #  $  % 

    &  '  (   Add the

    equations 

    l1=1   l

    2=!cFor  !  = +c

    For  !  ="c

     

    l1=1   l

    2= +c

     

    ! u

    ! t +  c

    ! u

    !  x 

    "  

    #  $ 

    %  

    &  '  ( c

    ! v

    ! t +  c

    ! v

    !  x 

    "  

    #  $  

    %  

    &  '   = 0

     

    du

    dt ! c

     dv

    dt = 0 on

      dx

    dt = +c

     

    du

    dt + c

     dv

    dt = 0 on

      dx

    dt = !c

    Relationbetween the

    total derivative

    on thecharacteristic Similarly: 

    Wave equation 

  • 8/18/2019 Esquemas Para Ec. de Convedascción

    2/8

    Computational Fluid Dynamics 

    For constant c 

    du

    dt 

    ! c dv

    dt 

    = 0 on  dx

    dt 

    = +c

     

    du

    dt + c

     dv

    dt = 0 on

      dx

    dt = !c

     

    dr1

    dt = 0 on

      dx

    dt = +c   where   r

    1= u ! cv

     

    dr2

    dt = 0 on

      dx

    dt = !c   where   r

    2= u + cv

    r1 and r2 are called the Rieman invariants  

    Wave equation 

    Computational Fluid Dynamics 

    The general solution can

    therefore be written as: 

     f x, t ( ) = r1   x ! ct ( ) + r2   x + ct ( )

     

    r1  x( ) =

    !  f 

    ! t " c

    !  f 

    !  x

    $ % & 

    ' ( t =0

    r2  x( ) =

    !  f 

    ! t + c

    !  f 

    !  x

    $ % & 

    ' ( t =0

    where 

    Can also be verified by direct substitution  

    Wave equation 

    Computational Fluid Dynamics 

    Domain ofInfluence 

    Domain ofDependence 

    t  

     x 

    Two characteristic lines dt 

    dx= +

    1

    c;  dt 

    dx= !

    1

    c

     

    dt 

    dx= +

    1

    c

    dt 

    dx= !

    1

    c

    Wave equation-general 

    Computational Fluid Dynamics 

    Ill-posedproblems

    Computational Fluid Dynamics 

    Ill-posed Problems 

    ! 2 f 

    ! t 2  = "

    ! 2 f 

    !  x2

    Consider the initial value problem: 

    ! 2 f 

    ! t 2  +

    ! 2 f 

    !  x2  =  0

    This is simply Laplace’s equation

     

    which has a solution if " f / "t  or  f  are given onthe boundaries 

    Computational Fluid Dynamics 

    Here, however, this equation appeared as an initialvalue problem, where the only boundary conditionsavailable are at t  = 0. Since this is a second order

    equation we will need two conditions, which we may

    assume are that f  and " f / " x are specified at t  =0. 

    Ill-posed Problems 

  • 8/18/2019 Esquemas Para Ec. de Convedascción

    3/8

    Computational Fluid Dynamics 

    Look for solutions of the type: 

     f  = ak (t )eikx

     

    d 2a

    dt 2

    = k 2a

    Substitute into: ! 

    2 f 

    ! t 2  +

    ! 2 f 

    !  x2  =  0

    to get: 

    Ill-posed Problems 

     f x,t ( ) =   ak (t )eikx

    !

    The general solution can be written as:  

    where the a’s dependon the initial conditions 

    Computational Fluid Dynamics 

    )0(;)0(   ikaa 

    d 2a

    dt 2

    =

     k 

    2

    ak 

    determined by initial conditions

    General solution 

    ak (t ) =  Ae

    kt + Be

    !kt 

     B A,

    Therefore:  !")(t a as  !"t 

    Ill-posed Problem 

    Ill-posed Problems 

    Generally, both A and B are non-zero 

    Computational Fluid Dynamics 

    Ill-posed Problems 

    Long wave with short wave perturbations  

    Computational Fluid Dynamics 

    ! f 

    !t = D

     !2 f 

    ! x2;  D  < 0

    has solutions with unbounded growth rate for high wavenumber modes and is therefore an ill-posed problem 

    Similarly, it can be shown that the diffusionequation with a negative diffusion coefficient 

    Ill-posed Problems 

    Computational Fluid Dynamics 

    Ill-posed problems generally appear when the initial orboundary data and the equation type do not match. 

    Frequently arise because small but important higher ordereffects have been neglected 

    Ill-posedness generally manifests itself in the exponential

    growth of small perturbations so that the solution does not

    “depend continuously on the initial data” 

    Inviscid vortex sheet rollup, multiphase flow models andsome viscoelastic constitutive models are examples ofproblems that exhibit ill-posedness.

    Ill-posed Problems 

    Computational Fluid Dynamics 

    Classical Methodsfor Hyperbolic

    Equations

  • 8/18/2019 Esquemas Para Ec. de Convedascción

    4/8

    Computational Fluid Dynamics 

    ! 2 f 

    ! t 2  " c 2 ! 

    2 f 

    !  x2= 0

     

    ! u

    ! t " c

    2 ! v

    !  x = 0

    ! v

    ! t "

    ! u

    !  x = 0

    The wave equation: 

    Write as: 

    In general: 

    ! u

    ! t 

    ! v

    ! t 

    "

    #

    $$$$

    %

    &

    ''''

    +

    a11

      a12

    a21

      a22

    "

    #

    $$

    %

    &

    ''

    ! u

    !  x 

    ! v

    !  x 

    "

    #

    $$$$

    %

    &

    ''''

    =  0

    Most of the issues involved can be addressed by examining: 

    !  f 

    ! t + U 

    !  f 

    !  x= 0

    Methods for Advection 

    Computational Fluid Dynamics 

    Forward in Time,Centered in Space

    (FTCS) andUpwind

    Computational Fluid Dynamics 

    We will start by examining the linear advection equation:  

    !  f 

    ! t + U 

    !  f 

    !  x=  0

    The characteristic for this equation are: 

    dx

    dt =U ;

      df 

    dt =  0;

    Showing that theinitial conditions are

    simply advected by a

    constant velocity U  

    t  

     f  

     f  

     x 

    Methods for Advection 

    Computational Fluid Dynamics 

    A simple forward in time, centered in spacediscretization yields 

    !  f 

    ! t + U 

    !  f 

    !  x=  0

     f  j n+1

    =   f  j n! "t 

    2hU ( f  j +1

    n!   f  j !1

    n)

     j-1 j j+1 

    n+1 

    Methods for Advection 

    Computational Fluid Dynamics 

    This scheme is O(#t, h2) accurate, but a stabilityanalysis shows that the error grows as 

    ! n+1

    ! n  =1"  i

    U #t 

    2hsin kh

    Since the amplificationfactor has the form 1+i()

    the absolute value of this

    complex number is alwayslarger than unity and the

    method is unconditionallyunstable for this case. 

    iU !t 

    2hsinkh

    1

    ! n+1

    ! n

    Methods for Advection 

    Computational Fluid Dynamics 

    A simple forward in time but “upwind” in spacediscretization yields 

    !  f 

    ! t + U 

    !  f 

    !  x=  0

     f  j n+1

    =   f  j n! " t 

    hU ( f  j 

    n!   f  j !1

    n)

     j-1 j  

    n+1  This schemeis O(#t, h)

    accurate.

    Another scheme for 

    Flow direction 

    Methods for Advection 

  • 8/18/2019 Esquemas Para Ec. de Convedascción

    5/8

    Computational Fluid Dynamics 

    To examine the stability we use the von Neuman’s method:

     

    !  j 

    n+1" !  j 

    n

    #t +U 

    h(!  j 

    n" !  j "1

    n)  = 0

    !  j n=  ! 

    neikx j 

    ! n+1" ! 

    n

    #t +U 

     ! n

    h(1" e

    " ikh)= 0

    The evolution of the error is governed by: 

    Write the error as:

     

    ! n+1

    ! n

    =1"U #t 

    h(1" e

    "ikh)

     

    G =! n+1

    ! n

    =1" # (1" e"ikh

    ),   # =U $t 

    h

    Amplification factor 

    G =1! " + " e! ikh

    Or:  G

  • 8/18/2019 Esquemas Para Ec. de Convedascción

    6/8

    Computational Fluid Dynamics 

    Finite Volume point of view: 

     x j -1/2  x j +1/2 

     f  j !1 f  j 

     x 

     f 

     f  j +1

     

     f  j n+1

    =   f  j n! "t 

    h(F  j +1/ 2

    n! F  j !1/ 2

    n) =   f  j 

    n! "t 

    hU ( f  j 

    n!  f  j !1

    n)

     

    F  j +1/ 2=Uf  j n

     

    F  j !1/ 2=Uf  j !1n

    Methods for Advection 

    Computational Fluid Dynamics 

    Generalized Upwind Scheme (for both U > 0 and U < 0 ) 

     f  j n+1

    =   f  j n!U "t 

    h( f  j 

    n!   f  j !1

    n ),   U > 0

     

     f  j 

    n+1=   f  j 

    n!U "t 

    h( f  j +1

    n!  f  j 

    n),   U < 0

    Define:  U +

    =1

    2U + U ( ),   U ! =

    1

    2U ! U ( )

    The two cases can be combined into a single expression: 

     f  j 

    n+1=   f  j 

    n! "t 

    hU +( f  j 

    n!  f  j !1

    n)+ U 

    !( f  j +1

    n!  f  j 

    n)[ ]

    Or, substituting !+

    U U    ,

     

     f  j n+1

    =   f  j n!U 

     "t 

    2h( f  j +1

    n!  f   j !1

    n)+

    U "t 

    2h( f  j +1

    n! 2 f  j 

    n+   f  j !1

    n)

    central difference + numerical viscosity   Dnum   =U h

    2

    Computational Fluid Dynamics 

    Other First Order Schemes 

    Computational Fluid Dynamics 

    Implicit (Backward Euler) Method 

    - Unconditionally stable - 1st order in time, 2nd order in space - Forms a tri-diagonal matrix (Thomas algorithm) 

    ( )   02

    1

    1

    1

    1

    1

    =!+"

    !+

    !

    +

    +

    +

    n

     j

    n

     j

    n

     j

    n

     j f   f  

    h

     f   f  

    n

     j

    n

     j

    n

     j

    n

     j  f  

    t  f  

    h

    U  f  

    t  f  

    h

    !="

    !+

      +

    "

    ++

    +

    1

    2

    1

    2

    1

    1

    11

    1

     j

    n

     j j

    n

     j j

    n

     j j  C  f  b f  d  f  a   =++

      +

    !

    ++

    +

    1

    1

    11

    1

    Methods for Advection 

    Computational Fluid Dynamics 

    Lax-Fredrichs method 

    - stable for $ < 1 

    - 1st order in time, 1nd order in space - Conditionally consistent 

     f  j n+1

    !  1

    2  f  j +1

    n+   f  j !1

    n( )"t 

    +U 

    2h f  j +1

    n!   f  j !1

    n( ) =  0

    ( )  xxx xx

      f  Uh

     f  Uh

      2

    2

    13

    1

    2! ! 

    ! "+# 

    $ % 

    &' (  "

    Error term: 

    !   =U "t 

    h

    h

    ! =

    h2

    U "t 

    Methods for Advection 

    Computational Fluid Dynamics 

    Second Order Schemes 

  • 8/18/2019 Esquemas Para Ec. de Convedascción

    7/8

    Computational Fluid Dynamics 

    )(2

    2

    11

    t Ot 

     f   f  

     f    n

     j

    n

     j!+

    !

    "=

    #

    #  "+

    Leap Frog Method 

    The simplest stable second-order accurate (in time) method: 

    Modified equation 

    ( )n j

    n

     j

    n

     j

    n

     j  f   f  h

    t U  f   f  

    11

    11

    !+

    !+!

    "!=

    ( )   !+!="

    "+

    "

    " xxx

     f  Uh

     x

     f  U 

     f  1

    6

    2

    2

    - Stable for 1

    - Dispersive (no dissipation) – error will not damp out 

    - Initial conditions at two t ime levels- Oscillatory solution in time (alternating)

    !  =U "t 

    h

    Methods for Advection 

    Computational Fluid Dynamics 

    !+!

    "

    "+

    !

    "

    "+!

    "

    "+=!+

    62)()(

    3

    3

    32

    2

    2 t 

     f  t 

     f  t 

     f  t  f  t t  f  

    Lax-Wendroff’s Method (LW-I) 

    First expand the solution in time 

    Then use the original equation to rewrite the time derivatives 

     x

     f  U 

     f  

    !

    !"=

    !

    !

    2

    2

    2

    2

    2

     x

     f  U 

     f  

     xU 

     x

     f  U 

    t t 

     f  

    t t 

     f  

    !!

    =

    !!

    !!

    "=# $ % 

    &' ( 

    !!

    !!

    "=# $ % 

    &' ( !!

    !!

    =

    !!

    Methods for Advection 

    Computational Fluid Dynamics 

    )(2

    )()(   32

    2

    22 t O

     x

     f  U t 

     x

     f  U t  f  t t  f     !+

    !

    "

    "+!

    "

    "#=!+

    Substituting 

    Using central differences for the spatial derivatives  

     f  j 

    n+1=   f  j 

    n!U "t 

    2h f 

     j +1

    n!  f  j !1

    n( ) +U 2"t 2

    2h2

      f  j +1

    n! 2 f  j 

    n+   f  j !1

    n( )

    2nd order accurate in space and time 

    Stable for  1<!

    h

    t U 

    Methods for Advection 

    Computational Fluid Dynamics 

    Two-Step Lax-Wendroff’s Method (LW-II) 

    LW-I into two steps:  

    For the linear equations, LW-II is identical to LW-I 

    02/

    2/)(11

    2/1

    2/1=

    !+

    "

    +!++

    +

    +

    h

     f   f  U 

     f   f   f    n

     j

    n

     j

    n

     j

    n

     j

    n

     j

    0

    2/1

    2/1

    2/1

    2/1

    1

    =

    !+

    "

    !  +

    !

    +

    +

    +

    h

     f   f  U 

     f   f    n

     j

    n

     j

    n

     j

    n

     j

    Step 1 (Lax) 

    Step 2 (Leapfrog) 

    - Stable for 1/  

    - Second order accurate in time and space

    Methods for Advection 

    Computational Fluid Dynamics 

    MacCormack Method 

    Similar to LW-II, without

    ( )n j

    n

     j

    n

     j

     j  f   f  h

    t U  f   f     !

    "!=

    +1

    ( )!"#$%

    & '('+= '+   t 

     j

     j

     j

    n

     j

    n

     j  f   f  

    ht U  f   f   f  

    1

    1

    2

    1

    Predictor 

    Corrector 

    - A fractional step method  - Predictor: forward differencing - Corrector: backward differencing 

    - For linear problems, accuracy and stability

    properties are identical to LW-I.

    2/1,2/1   !+   j  j

    Methods for Advection 

    Computational Fluid Dynamics 

    Second-Order Upwind Method 

    Warming and Beam (1975) – Upwind for both steps 

    ( )n j

    n

     j

    n

     j

     j  f   f  h

    t U  f   f  

    1!!

    "!=

     

     f  j

    n+1=

    1

    2 f 

     j

    n+  f 

     j

    t !U  "t h

     f  j

    t !  f  j!1

    t ( ) !U  "t h

     f  j

    n ! 2 f  j!1

    n+  f 

     j!2

    n( )#$% &

    '(

    Predictor  Corrector 

    Combining the two: 

     f  j

    n+1=   f 

     j

    n! "    f 

     j

    n!  f 

     j!1

    n( ) +1

    2" ("  ! 1)   f 

     j

    n! 2 f 

     j!1

    n+  f 

     j!2

    n( )

    - Stable if - Second-order accurate in time and space

    20   !! " 

    Methods for Advection 

  • 8/18/2019 Esquemas Para Ec. de Convedascción

    8/8

    Computational Fluid Dynamics 

    The one-step Lax-Wendroff is not easily extended tonon-linear or multi-dimensional problems. The splitversion is. 

    In the Lax-Wendroff and the MacCormack methodsthe spatial and the temporal discretization are notindependent. 

    Other methods have been developed where the time

    integration is independent of the spatial discretization,

    such as the Beam-Warming and various Runge-Kuttamethods 

    Methods for Advection 

    Computational Fluid Dynamics 

    FTCSUnconditionally

    Unstable

    UpwindStable for

     

    ImplicitUnconditionally

    Stable

    Lax-Friedrichs Conditionallyconsistent

    Stable for

    0=+ xt 

      Uf   f  

    02

    11

    1

    =

    !+

    "

    !!+

    +

    h

     f   f  U 

     f   f     n jn

     j

    n

     j

    n

     j ( )  xxx xx   f  Uh

     f  U 

    t    222

    2162

    ! +"#"

    01

    1

    =

    !+

    "

    !!

    +

    h

     f   f  U 

     f   f     n jn

     j

    n

     j

    n

     j

    ( )

    ( )  xxx

     xx

     f  Uh

     f  Uh

    1326

    12

    2

    2

    +!!

    !

    " " 

    1!" 

    ( )  xxx xx   f  Uh

     f  Uh

      2

    2

    13

    1

    2! ! 

    ! "+# 

    $ % 

    &' (  "

    ( )0

    2

    1

    1

    1

    1

    1

    =!

    +"

    !  +

    !

    +

    +

    +

    h

     f   f  U 

     f   f     n j

    n

     j

    n

     j

    n

     j xxx xx   f  t U Uh f  

    t U !"

    #$%

    &'+(

    '  232

    2

    3

    1

    6

    1

    2

    ( )0

    2

    2/

    11

    11

    1

    =!

    +

    "

    +!

    !+

    !+

    +

    h

     f   f  U 

     f   f   f  

    n

     j

    n

     j

    n

     j

    n

     j

    n

     j

    1!" 

    Summary 

    Computational Fluid Dynamics 

    Leap FrogStable for

     

    Lax-Wendroff IStable for

     

    Lax-Wendroff II

    Same as LW-IStable for

     

    MacCormack

    Same as LW-IStable for

     

    0=+ xt 

      Uf   f  

    022

    11

    11

    =

    !+

    "

    !!+

    !+

    h

     f   f  U 

     f   f     n j

    n

     j

    n

     j

    n

     j ( )  xxx f  Uh

    16

    2

    2

    !" 

    ( )

    ( )   xxxx

     xxx

     f  Uh

     f  Uh

    2

    3

    2

    2

    18

    16

    ! ! 

    ""

    ""

    1!" 

    1!" 

    1!" 

    ( )0

    2

    2

    2

    2

    1122

    11

    1

    =+!

    "!

    !+

    "

    !

    !+

    !+

    +

    h

     f   f   f  t U 

    h

     f   f  U 

     f   f  

    n

     j

    n

     j

    n

     j

    n

     j

    n

     j

    n

     j

    n

     j

    02/

    2/)(11

    2/1

    2/1=

    !+

    "

    +!++

    +

    +

    h

     f   f  U 

     f   f   f     n j

    n

     j

    n

     j

    n

     j

    n

     j

    0

    2/1

    2/1

    2/1

    2/1

    1

    =

    !+

    "

    !  +

    !

    +

    +

    +

    h

     f   f  U 

     f   f     n j

    n

     j

    n

     j

    n

     j 1!" 

    ( )0

    1

    =!

    +"

    !+

    h

     f   f  U 

     f   f     n j

    n

     j

    n

     j

     j

    0

    2/1

    1

    =

    !

    +"

    +!!

    +

    h

     f   f  

    U t 

     f   f   f     t  j

     j

     j

    n

     j

    n

     j

    Summary 

    Computational Fluid Dynamics 

    Beam-WarmingStable for

     

    QUICKStable for

     

    ENO

    WENO

    0=+ xt 

      Uf   f  

    ( )   022

    2

    43

    212

    2

    21

    1

    =+!"

    !

    +!+

    "

    !

    !!

    !!

    +

    n

     j

    n

     j

    n

     j

    n

     j

    n

     j

    n

     j

    n

     j

    n

     j

     f   f   f  h

    t U 

    h

     f   f   f  U 

     f   f  

    20   !! " 

    ( )( )

    ( ) ( )   xxxx

     xxx

     f  Uh

     f  Uh

    ! ! 

    ! ! 

    """

    ""

    218

    216

    2

    3

    2

     f  j n+1!  f  j 

    n

    "t +

    U 3 f  j 

    n+ 6 f  j !1

    n!  f  j !2

    n( ) !   3 f  j +1n+ 6 f  j 

    n!  f  j !1

    n( )8h

    1!" 

    A large number of (conditionally) stable andaccurate methods exists for hyperbolic equationswith smooth solutions

    And Many More! 

    Summary