cmn.1 computational continuum mechanicscfdlab.utk.edu/finite/graduate/public/pdfs/cmnvu551w.pdf ·...

31
CMn.1 Computational Continuum Mechanics GWS h /TWS h FE implementation widely applicable n -D computational mechanics forms illustrate structural mechanics heat/mass transfer electromagnetics fluid mechanics mechanical vibrations coupled PDE systems matrix PDEs constraints non-linearity

Upload: others

Post on 25-Jan-2021

0 views

Category:

Documents


0 download

TRANSCRIPT

  • CMn.1 Computational Continuum Mechanics

    GWSh /TWSh FE implementation widely applicable n-D computational mechanics forms illustrate

    structural mechanics heat/mass transfer electromagnetics fluid mechanics mechanical vibrations

    coupled PDE systems matrix PDEs constraints non-linearity

  • CMn.2 Structural Mechanics, Conservation Forms

    Newton’s law, statics

    σ ij ≡ stress tensor = σ x τ xy τ xz σ y τ yz(sym) σ z

    Principal of virtual work l inear Hooke’s law, matr ix vector form

    DΕ: Π = 1 2 ε T E ε - u T B d vol - u

    T T d surf ∂Ω

    Ω

    T = surface traction, ε = strain matrix, u = displacement vector

    surfiTiuvoljBjuijjivoleE ddd0 ,d2/1d:D ∫Ω∂

    −∫Ω

    −−∫∫Ω

    =∫Ω

    ≡∏ φεε σ

    , Bi = body force

    DP:

    0ij ii

    Bxσ∂

    + =∂

  • CMn.3 Structural Mechanics, Formulations

    F i r s t F E a p p l i c a t i o n i n s t r u c t u r e s , v i a Ω ⇒ Ω h = ∑ e Ω e

    w h e n e x t r e m i z e d , p r o d u c e s a l g e b r a i c s y s t e m G W S N d e v e l o p m e n t , p l a n e s t r e s s / s t r a i n , n = 2

    D P :

    ∂σ x

    ∂ x + ∂τ xy

    ∂ y + B x = 0

    ∂σ y

    ∂ y + ∂τ xy

    ∂ x + B y = 0

    form ulation ingredients:

    ∑ ∫ ∫Ω Ω∂⎟⎟⎠⎞

    ⎜⎜

    ⎛−−≡ΠΠ=Π≈Π

    e e e

    TTTee

    h dσ}{}{d}{}{}]{[}{21, TuBuεEε τ

    y

    x

    T T

    y

    x

    undeflected geom etry exaggerated deflected state

    { } { }xyyxTi

    j

    j

    iij x

    uuu

    yxvyxuyx

    γ,ε,εε:formmatrix

    2/1ε:kinematics

    ˆ),(ˆ),(),(:ntdisplaceme

    =

    ⎟⎟⎠

    ⎞⎜⎜⎝

    ∂+

    ∂∂

    =

    += jiu

  • CMn.4 Linear Elasticity, Constitutive Model

    Constitutive model relates stress and strain tensor form: σij = λεkk +2Gεij matrix form: {σ} = [E] {ε} Hooke’s law, l inear, homogeneous media

    { } { }

    { } { } [ ]

    )σσ(νσ2/)1(00

    0101

    )21()1(E,,,

    ,,:strainplane

    yxz

    Txyyx

    Txyyx

    +=

    ⎥⎥⎥

    ⎢⎢⎢

    −−

    −−≡≡

    ννν

    νν

    ννγεεε

    τσσσ

    E

    { } { }

    { } { } [ ]

    )ν1/()εε(νε2/)ν1(00

    01ν0ν1

    ν1E,γ,ε,εε

    τ,σ,σσ:

    2

    stressplane

    −+−=

    ⎥⎥⎥

    ⎢⎢⎢

    −−

    ≡≡

    yxz

    Txyyx

    Txyyx

    E

  • CMn.5 DP for Plane Elasticity

    Substitute definitions into DP

    plane stress: L u = ∇

    2u + 1 + v 1 - v

    ∂ x ∂ u

    ∂ x + ∂

    v ∂ y

    + B xG

    = 0

    L v = ∇ 2v + 1 + v

    1 - v ∂

    ∂ y ∂ u

    ∂ x + ∂

    v ∂ y

    + B yG

    = 0

    plane strain: L u = ∇ 2u + 1

    1 - 2 v ∂

    ∂ x ∂ u

    ∂ x + ∂

    v ∂ y

    + B xG

    = 0

    L v = ∇ 2v + 1

    1 - 2 v ∂

    ∂ y ∂ u

    ∂ x + ∂

    v ∂ y

    + B yG

    = 0

    material propert ies : DP + Hooke produces coupled laplacian PDE system

    ∴ BCs required on total i ty of ∂Ω

    for displacement field: u(x, y) ≡ u(x, y)i + v(x, y)j

    ν ≡ Poisson rat io E = Youngs modulus G ≡ shear modulus = E / (2(1 + ν )

  • CMn.6 Vector DP for Plane Elasticity

    Contributions to laplacian embedded in second terms

    directional diffusion: k v = 2/ 1 - ν , plane stress2 1 - ν / 1 - 2 ν , plane strain

    plane stress/strain directional diffusion form: b ≡ B/G

    L u = k v ∂

    2u ∂ x 2

    + ∂ 2u

    ∂ y 2 + k v - 1

    ∂ 2v

    ∂ x ∂ y + b x = 0

    L v = k v ∂

    2v ∂ y

    2 + ∂

    2v ∂ x

    2 + k v - 1 ∂

    2u ∂ x ∂ y

    + b y = 0

    Vector form identifies divergence constraint for g(ν) = ( 1 + ν)/(1 - ν) or 1/(1 - 2ν) DP+ Hooke : L (u) = - ∇2u - g(ν) ∇ (∇•u) - b = 0

  • CMn.7 Matrix DP For Plane Elasticity

    How to handle ∇(∇⋅u) ⇒ matrix DP

    :D =+∂

    ∂i

    j

    ij Bx

    P

    2

    0

    0

    ][

    =⎥⎥⎥⎥⎥⎥⎥

    ⎢⎢⎢⎢⎢⎢⎢

    ∂∂

    ∂∂

    ∂∂

    ∂∂

    =

    nxy

    y

    xD

    DP+Hooke matrix PDE

    DP+Hooke : [D]T [E][D]{u} + {B} = {0}

    { } [ ]{ } { } [ ]{ }uDE == εandεσfor

    { }0BDP =+ }{}σ{][:D T ⇒

  • CMn.8 GWSN, Plane Elasticity Matrix DP

    For any approximation u N ≡ ∑ Ψ α (x)U α

    )(,0d}{}]{][[][)(GWSββ

    xBuDEDx ΨΩ

    ≡⎟⎠⎞⎜

    ⎝⎛ −−Ψ≡ ∫ allforNTN τ

    Green-Gauss theorem symmetr izes

    ∫∫ ∫

    Ω∂≡Ψ−

    Ω ΩΨ−Ψ=

    0d σ}]{][[][β

    d}{βd}]{][[]β[GWS

    NT

    NTN

    uDEN

    BuDED ττ

    Neumann BC surface integra l TT ][ˆ][ DnN •≡

    t e r m in t r o d u c e s s u r f a c e t r a c t io n s T D E e x t r e m u m wi l l p r o vid e t h e m a t r ix f o r m

  • CMn.9 DE Virtual Work Extremum, Plane Elasticity

    A u g m e n t i n g Π f o r i n i t i a l s t r e s s / s t r a i n , p o i n t l o a d s

    Π = 1 2 ε T E ε - ε T E ε 0 + ε T σ 0 d v o l

    Ω

    - u T B dvo l -Ω

    u T T s dsurf - U T P∂Ω

    E v a l u a t e i n t e g r a l a s s u m o v e r F E d o m a i n s Ω e E x t r e m i z e w . r . t n o n - c o n s t r a i n e d F E n o d a l D O F

    }{}]{[}0{DOF

    RUK =⇒≡⎟⎟

    ⎜⎜

    ∂Π∂ h

    ( ) ( ) ( )∑ ∫∫∑ ⎥⎦⎤

    ⎢⎣⎡ ⋅−⋅−⋅=Π=Π≈Π

    Ω∂∩Ω∂ΩΩ e

    eh surfvol

    eee

    dd

    w h ere : [K ] = g lo b a l “s tiffn ess” m atrix {R } = g lo b a l “ lo ad v ec to r”

    D E :

  • CMn.10 DE Virtual Work, FE Stiffness Matrix

    I m p l e m e n t i n g D E e x t r e m u m v i a F E b a s i s o n Ω e

    approximation: ue(x,y) ≡ {Nk}T{U}e

    matrix form: u e ≡ u v e = U V e

    { } { }{ } { } ⎥⎦

    ⎤⎢⎣

    ⎡T

    k

    Tk

    NN

    00

    kinematics: {ε} ≡ [D] {u} S t i f f n e s s m a t r i x c o n t r i b u t i o n t o Π h

    eeV

    UT

    kN

    TkN

    xy

    yx

    exy

    yx

    }]{[}{}0{

    }0{}{0

    0}]{[

    γ

    εε

    UBUD ≡⎭⎬⎫

    ⎩⎨⎧

    ⎥⎥⎥

    ⎢⎢⎢

    ⎥⎥⎥⎥

    ⎢⎢⎢⎢

    ∂∂

    ∂∂

    ≡≡

    ⎪⎪⎭

    ⎪⎪⎬

    ⎪⎪⎩

    ⎪⎪⎨

    { } [ ]{ }

    { } [ ] [ ][ ] { }

    { } [ ] { } [ ] matrixstiffnesselement,21

    dd21

    ddεε21

    ≡⋅⋅⋅+=

    ⋅⋅⋅+=

    ⋅⋅⋅+≡Π

    Ω

    Ω

    eeeTe

    e

    TTe

    Te

    yx

    yx

    e

    e

    kUkU

    UBEBU

    E

  • CMn.11 DE Virtual Work, Plane Elasticity FE Extremum, BCs

    E x t r e m iz a t io n o f Π e f o r D O F n o t B C - c o n s t r a in e d L o a d m a t r ix e x t r e m u m

    C o m p a r in g D P a n d D E e x p o s e s N e u m a n n B C fo r G W S h a s

    ∫ Ω∂ ⎭⎬⎫

    ⎩⎨⎧

    ⇒⎭⎬⎫

    ⎩⎨⎧

    ⎥⎥⎥

    ⎢⎢⎢

    ⎥⎥⎦

    ⎢⎢⎣

    e eeT

    kN

    TkN

    kNkN

    }TY]{200A[}TX]{200A[

    dσTYTX

    }{0

    0}{}{0

    0}{

    k e U e{ } =∂

    Π∂

    e

    e

    U

    ( ) eSβ =Ψ∫ ⋅

    { } { } { } [ ] [ ]{ } [ ] { }( ){ }[ ] { }[ ]{ }

    { }[ ] { }[ ]{ } { }PTg

    BEBrrR

    ++

    +

    −==

    ∫∫

    Ω∂∩Ω∂

    Ω

    Ω

    surfNN

    volNN

    vol

    Tkk

    Tkk

    Te

    Teee

    e

    e

    e

    d

    dσε,S 00

    w h e r e : { P } = p o i n t l o a d s a t D O F { T } = s u r f a c e t r a c t i o n n o d a l f o r c e

  • CMn.12 Structures GWSh FE Template Essence

    With traction BC, FE implementation GWSh yields

    GWS h = S e {WS} e ≡ {0} {WS} e = [STIFF] e {U} e – {b} e

    [STIFF] e template essence

    ∫Ω≡ eyxe

    Tee dd]][[][]STIFF[ BEB

    {WS} e = CONST e DET e - 1 [ETLI ] e T [B2LKE ] [ETKI ] e

    Load vector template essence

    ∫Ω ⎟⎠⎞⎜⎝⎛ +−≡ eyxe

    TNNeTee

    Tee dd}ρ]{}}][{[{}0ε{][}0ε]{[][}b{ gBEB

    ∫ Ω∂∩Ω∂ ++ e eeTNN }δ{dσ}]{}}][{[{ PT

    {WS}e = CONSTe [ETLI]eT [B2L0E] {EPS0}e + CONSTe [ETLI]eT {SIGL0}e + P{δ}e + DETe [B200]{RHOG}e + DETe [A200]{T}e

  • 1 2

    3

    Ωe

    (X1, Y1)e (X2, Y2)e

    ζ3

    ζ1

    ζ2

    x

    y (X3, Y3)e

    For {N1(ζ)} spanning triangle Ωe , index notation

    e

    i

    ⎥⎥⎥

    ⎢⎢⎢

    −−−−−−

    =

    1221

    3113

    2332

    α

    XXYYXXYYXXYY

    ζ

    ⎥⎥⎥

    ⎢⎢⎢

    ⎥⎥⎥

    ⎢⎢⎢

    ∂∂∂∂∂∂

    ∂∂≡ TNT

    TTN

    xxx

    x

    e }1{}0{

    }0{}1{

    1/2/2/0

    01/][B

    via chain rule, ∂/∂xi =∂/∂ζα(∂ζα/∂xi)e,

    [ ]

    0α1 {δ } {0}1 α[ ] 0 α2det {0} {δ }αα2 α1

    0 0 011 21 311 10 0 0 det ET12 22 32det

    12 22 32 11 21 31 3 6

    T T

    e T Tee

    KI eee

    e

    ζ

    ζ

    ζ ζ

    ζ ζ ζ

    ζ ζ ζ

    ζ ζ ζ ζ ζ ζ

    ⎡ ⎤⎡ ⎤⎢ ⎥⎢ ⎥⎢ ⎥= ⎢ ⎥⎢ ⎥⎢ ⎥⎢ ⎥ ⎣ ⎦

    ⎣ ⎦

    ⎡ ⎤⎢ ⎥

    −⎢ ⎥≡ ≡⎢ ⎥⎢ ⎥⎣ ⎦ ×

    B

    and recalling ∂{N1}/∂ζα = {δα}

    CMn.13 GWSh k = 1 FE Basis Implementation

  • CMn.14 GWSh k = 1 Implementation, Stiffness Matrix

    For [B]e ⇒ [ETKI]e , k = 1 element “stiffness matrix”, plane stress

    e

    v

    e

    e

    ⎥⎥⎥⎥⎥⎥

    ⎢⎢⎢⎢⎢⎢

    ⎟⎠⎞

    ⎜⎝⎛ −

    ⎥⎥⎥⎥⎥⎥⎥⎥

    ⎢⎢⎢⎢⎢⎢⎢⎢

    −=

    21

    31ζ21ζ11ζ32ζ22ζ12ζ

    32ζ22ζ12ζ31νζ21νζ11νζ32νζ22νζ12νζ31ζ21ζ11ζ

    31ζ32ζ021ζ22ζ011ζ12ζ032ζ031ζ22ζ021ζ12ζ011ζ

    )2ν1(2

    E1det

    )(

    [ ] [ ] [ ][ ] [ ] [ ][ ]

    [ ]e

    e

    eTeeee

    Te

    KIv

    vv

    v

    yx

    KILIyx

    e

    e

    ET2/)1(00

    0101

    ζζ0ζζ0ζζ0ζ0ζζ0ζζ0ζ

    )1(det

    ddE

    ETB2LKETDET CONSTddSTIFF

    3132

    2122

    1112

    3231

    2221

    1211

    22

    1

    ⎥⎥⎥

    ⎢⎢⎢

    ⎥⎥⎥⎥⎥⎥⎥⎥

    ⎢⎢⎢⎢⎢⎢⎢⎢

    −=

    =≡

    Ω

    ΩEBEB

  • CMn.15 GWSh k = 1, Stiffness Matrix Resolution

    R e s o l u t i o n e

    ee

    ee ]SHEAR[)1(4

    E1det]NORML[

    )21(2

    E1det]STIFF[

    νν +

    −+

    −=

    e

    sym

    e

    ⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥

    ⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢

    =

    232ζ32ζ22ζ32ζ12ζ31ζ32νζ21ζ32νζ32ζ11νζ

    222ζ22ζ12ζ31ζ22νζ21ζ22νζ22ζ11νζ

    212ζ31ζ12νζ21ζ12νζ12ζ11νζ

    231ζ31ζ21ζ31ζ11ζ

    )(221ζ21ζ11ζ

    211ζ

    ]NORML[

    e

    sym

    e

    ⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥

    ⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢

    =

    231ζ31ζ21ζ31ζ11ζ32ζ31νζ22ζ31νζ12ζ31νζ

    221ζ21ζ11ζ32ζ21νζ22ζ21νζ12ζ21νζ

    211ζ32ζ11νζ22ζ11νζ12ζ11νζ

    232ζ22ζ32ζ32ζ12ζ

    )(222ζ22ζ12ζ

    212ζ

    ]SHEAR[

  • CMn.16 GWSh, [DIFF]e ⇔ [STIFF]e Comparison

    The {N1(ζ )} [STIFF] e is ordered for {Q}e ≡ {U1, U2, U3, V1, V2, V3}e T

    extract ing common mult ipl ier

    [ ] [ ][ ] [ ] ⎥⎦

    ⎤⎢⎣

    ⎡+

    ⎟⎠⎞

    ⎜⎝⎛ +

    −+=

    eeee

    e

    eee

    e

    STIFFVVSTIFFVUSTIFFUVSTIFFUU

    det)ν1(2E

    ]SHEAR[21]NORML[

    ν11

    det)ν1(2E]STIFF[

    Comparing to scalar heat transfer construction

    ee

    ee QQ

    k]DIFF[

    det2]DIFF[ =

    common divisor : 2det e

    diffusion coeff ic ient :⎟⎟

    ⎜⎜

    −+⇔

    21,

    11

    ν1E

    vke

    ν < 1 , E ≈ O (10 6) ⇒ e las t ic i ty is very dif fus ive!

  • CMn.17 GWSh, [DIFF]e ⇔ [STIFF]e Comparison

    The metric data [ETLI] e play the fundamental role

    for heat t ransfer

    e = k e

    2 det e

    ζ 112 + ζ 12

    2 , ζ 11 ζ 21

    + ζ 12ζ 22 , ζ 11 ζ 31 + ζ 12

    ζ 32

    ζ 21 ζ 11

    + ζ 22 ζ 12

    , ζ 21 2 + ζ 22

    2 , ζ 21 ζ 31 + ζ 22

    ζ 32

    ζ 31 ζ 11

    + ζ 32 ζ 12

    , ζ 31ζ 21 + ζ 32ζ 22 , ζ 312 + ζ 32

    2 e

    [DIFFQQ]e

    for plane s t ress , representat ive se l f coupl ing term

    ee

    e

    ⎥⎥⎥⎥

    ⎢⎢⎢⎢

    +−+−

    +−

    +=

    ........................................,2ζζ)ν1(ζζ

    ........................................,2ζζ)ν1(ζζ

    ........................................,2ζ)1(ζ

    det)ν1(2E]STIFFUU[

    3212

    22122111

    2

    12

    2

    11

    3111

    ν

    c ross-coupling terms possess s imilar regular i ty

  • CMn.18 GWSh k = 1 Implementation, Load Vector

    Body force, surface traction and point load contributions

    [ ] yxee

    TNNe dd}ρ{}1{}1{}b{ g∫Ω ⎥⎦⎤⎢⎣⎡=

    ee

    ⎭⎬⎫

    ⎩⎨⎧

    ⎥⎦

    ⎤⎢⎣

    ⎡=

    RHOGYRHOGX

    ]200B[]0[]0[]200B[

    6

    det

    { }δPYPX

    TYTX

    ]200A[]0[]0[]200A[

    3

    det

    eee

    ⎟⎟⎠

    ⎞⎜⎜⎝

    ⎛+

    ⎭⎬⎫

    ⎩⎨⎧

    ⎥⎦

    ⎤⎢⎣

    ⎡+

    element-independent matrices [B200], [A200] same as defined in HTn

    ∫Ω ++ e eeeT

    kNkN∂}δ{dσ}]{}}][{[{ PT

  • CMn.19 Plane Stress: Plate with a Hole

    GWSh {N1} basis implementation

    meshing, Ωh von Mises stress

    x displacement, uh y displacement, vh

  • F lu id -th erm a l-stru c tu ra l sy stem d es ig n u ses “ C F D ”

    fo r R ey n o ld s-a v e ra g ed , tu rb u len t, u n s te ad y in c o m p re ss ib le m e an f lo w

    D M : ∇ ⋅u = 0

    D P : ( ) 0guuuu =Θ+∇+⋅∇−∇+∇⋅+∂∂ − ˆ

    ReGrReEu 2

    1 tvpt

    D E : ( ) 0u =+Θ∇+⋅∇−Θ∇⋅+∂Θ∂

    Θ−− sv

    tt

    t1PrPe 1

    D E ′ :

    D E m ′ :

    ( ) 0uTu =−∇+∇+⋅∇−∇⋅+∂∂ −− εPrPe 11 kvk

    tk t

    t

    ( ) 0uTu =−∇+∇⋅∇−∇⋅+∂∂ − k

    kv

    tt

    t /εCεCεPrCεε 22ε

    w h ere :

    n o n -D g ro u p s a re d e fin e d in tex t C h ap te r C M n

    a lg e b ra ic R ey n o ld s s tre ss m o d e l

    ( )εC

    ...)(32

    2kv

    vk

    t

    Tt

    μ≡

    +∇+∇+−=− uuδT

    CMn.20 Navier-Stokes, Computational Fluid Dynamics

  • NS turbulent flow conservation law PDEs

    initial-value explicitly non-linear! multiple DOF/node! as given, are ill-posed! intrinsically unstable for Re >> 1

    Commercial CFD codes contain required capabilities

    have 15-20 year development history are based on FD, FV and/or FE discrete methods learning curve is steep! chances for error are numerous output interpretation requires color graphics & animations

    CMn.21 Navier-Stokes, CFD, Algorithm Issues

  • For n = 2: kuku ˆωˆψ ⋅×∇≡×∇= and

    DM: yidenticall0ˆψ =×∇⋅∇=⋅∇ ku

    :Dˆ Pk ×∇⋅ 0ωReωˆψω 21 =∇−∇⋅×∇+ −kt

    kinematics: ψˆˆψˆω 2−∇=⋅×∇×∇=⋅×∇= kkku

    Steady-state N-S PDEs + BCs:1 2

    2

    ˆ(ω) Re ω ψ ω 0(ψ) ψ ω 0

    −= − ∇ + ∇ × ⋅∇ =

    = −∇ − =

    kLL

    )ω,ψ(ωˆconstantψψ:

    0)ψω,(ˆ:calculusviaψ,ω),(:

    wall

    out

    inininin

    w

    w

    f

    xy

    =∇⋅==Ω∂

    =∇⋅Ω∂⇒Ω∂

    n

    nu

    CMn.23 Streamfuction-Vorticity Navier-Stokes

  • Consider unsteady isothermal laminar NS

    DM: ∇⋅u = 0

    DP: !onnone,onPDEs4

    0ReEu 21

    ppt

    uuuuu

    ⇒=∇−∇+∇⋅+ −

    Mathematically, DM is a constraint on solutions to DP

    theories to enforce the constraint include

    pseudo-compressibility: 0βD 1 =⋅∇+⇒ − utpM

    pressure projection: ε 0, itera tive lyh h∇ ⋅ ⇒ >u

    vector field theory: Ψu ×∇=guaranteesD M

    ∇×DP eliminates pressure appearance

    ⇒ for n = 2, produces streamfunction-vorticity formulation

    CMn.22 Incompressible N-S, Well-Posedness

  • Galerkin weak statements:

    for

    thus:

    α αα

    ( , ) ψ ( ) , {ω,ψ} , {OMG,PSI}N T Tq x y Q q Qα≡ = =∑ x

    N N 1 2 N N Nβ βΩ Ω

    1β α α β γ γ α αΩ

    N Nβ β α α β α αΩ Ω

    ˆGWS (ω) ψ (x)L(ω )dτ ψ Re ω ψ k ω dτ

    Re ψ ψ OMG ψ ψ PSI ψ OMG dτ BC 0, β,α

    GWS ( ) ψ L(ψ )dτ [ ψ ψ PSI ψ ψ OMG ] dτ BC 0, β,αψ

    ⎡ ⎤= = − ∇ + ∇× ⋅∇⎣ ⎦⎡ ⎤= ∇ ⋅∇ + ∇× ⋅∇ + = ∀⎣ ⎦

    = = ∇ ⋅∇ − + = ∀

    ∫ ∫∫∫ ∫

    GWSN ⇒ GWSh = Se{WS}e ≡ {0}

    1

    1w

    w

    {WS (ω )} Re [B2 ] {OMG} {PSI} [B3 0 ] {OMG}

    Re [A200] { (ψ ,ω , U }

    {WS(ψ )} [B2 ]{PSI} [B200] {OMG} [A200] {U }

    h Te e e e e e

    h he w e

    he e e e e e

    KK K K

    f

    KK

    = +

    +

    = − +

    CMn.24 GWSh, Streamfunction-Vorticity NS, n = 2

  • s

    n

    Ωh

    ∂Ωwall

    y

    x

    Ωe

    ^n

    ωkinvolvesωforGWS ∇⋅×∇ ˆψhh

    Vorticity Robin BC generated via TS on Ωh:

    yxxy ∂∂

    ∂∂

    −∂∂

    ∂∂

    =∇⋅×∇ωψωψωˆψk

    [ ]

    ee

    TT

    e yxxNN

    xN

    xNN

    yNKK

    e

    ]B3X0Y[]B3Y0X[

    dd}{}{}{}{}{}{0B3

    −⇒

    ⎥⎦

    ⎤⎢⎣

    ⎡∂

    ∂∂

    ∂−

    ∂∂

    ∂∂

    =∴ ∫Ω

    wall

    ωdψd0ωψψωψ 2

    2

    2

    2

    2

    22

    Ω∂

    −=⇒=−∂∂

    −∂∂−

    =−∇−nns

    TS:

    6/)d/dω(2/ωUψ

    )(6d

    ψd2d

    ψdddψψ)(ψ

    3w

    2www

    43

    w3

    32

    w2

    2

    ww

    nnnn

    nOnn

    nn

    nn

    n

    Δ−Δ−Δ+=

    Δ+Δ

    +Δ+=Δ

    BC: ( ) 0)ψψ)(/6()U)(/6(ω)/3(-ωˆω w1w3w2 =−Δ−Δ+Δ∇⋅= +nnnn

    V(y)

    CMn.25 GWSh Details, Streamfunction-Vorticity NS

  • GWSh ⇒ global Newton statement

    ( ) ( )eepeepp QQQQ }F{S}δ{]JAC[S}F{}δ]{JAC[ 11 −=⇔−= ++

    Template pseudo code for {FQ}e

    {WS}e ≡ (const) (avg)e{dist}e(metric;det)e[matrix]{Q or data}e

    }ψ,(UOMG]{200A)[1}(){)(Re,3(}OMG]{0B3-03B)[1;2E1E}(PSI){)((

    }OMG]{2B)[1;EE}(){)((Re}{FOMG

    ww1

    -1

    Δ+Δ−+

    −+−=

    − fnKJJKKJ

    JKKIJIe

    }U]{200A)[1}(){)((}OMG]{200B)[1}(){)((

    }PSI]{2B)[1;EE}(){)((}{FPSI

    w+−+

    −= JKKIJIe

    CMn.26 Newton {FQ}e Template, (ωh, ψh) GWSh

  • Newton jacobian formed via differentiation

    ee

    ee Q

    Q⎥⎦

    ⎤⎢⎣

    ⎡Ω

    ΩΩΩ=

    ∂∂

    ≡]Jψψ[]Jψ[]ψJ[]J[

    }{}F{]JAC[

    Jacobian template pseudo-code

    ]][2B)[1;EE}(){)((}PSI{}FPS{]Jψψ[

    ]][200B)[1}(){)(1(}OMG{}FPSI{]Jψ[

    ]][200A)[1}(){)(Re,6(]][0B3-03B)[1;2E1E}(OMG){)((}PSI{

    }FOMG{]ψJ[

    ]][200A)[1}(){)(Re,3(

    ]][03B03B)[1;2E1E}(PSI){)((]][2B)[1;EE}(){)((Re}OMG{}FOMG{]J[

    31

    1

    1

    JKKIJII

    nJKKJKJ

    n

    KJJKKJJKKIJI

    e

    ee

    e

    ee

    e

    ee

    e

    ee

    −=∂

    ∂≡

    −=∂∂

    ≡Ω

    Δ−−=∂

    ∂≡Ω

    Δ−

    −−+−=∂

    ∂≡ΩΩ

    CMn.27 Newton Jacobian Template, (ωh, ψh) GWSh

  • G W S h + N ew ton ⇒ com p u tab le m atrix sta tem en t

    S tep -w ell d iffu ser b en ch m ark

    ⎟⎟

    ⎜⎜

    ⎭⎬⎫

    ⎩⎨⎧

    −=⎭⎬⎫

    ⎩⎨⎧

    ⎥⎦

    ⎤⎢⎣

    ⎡Ω

    ΩΩΩ + p

    e

    p

    eee FPSI

    FOMGδPSIδOMG

    JψψJψψJJ

    S1

    0

    2

    4

    6

    8

    10

    12

    14

    0 100 200 300 400 500 600 700

    R e

    L/s

    C om pu tedE xperim en tal, A rm aly, et. a l (19 83 )

    R e = 100

    R e = 600

    V alidation

    CMn.28 Incompressible N-S, Step Wall Diffuser

  • CMn.29 Mechanical Vibrations, Continuum Systems

    Lagrangian viewpoint, mechanical continuum

    E-L:

    22 E21,ρ21 xuVuT ==

    for Lagrangian L ≡ T – V = data) ,,( uu ∇f , and dM = 0 = dE

    0)(

    :equationLEd =∇∂

    ∂⋅∇−⎟

    ⎠⎞

    ⎜⎝⎛

    ∂∂

    ∂∂

    −⇔uu

    P LLt

    2( ) 0, E ρxxu u c u c= − = =L

    Longitudinal oscillations of a bar, n = 1

  • Lagrangian

    L = T – V = data) ,,( uu ∇f

    0:dL-E =⎟⎠⎞

    ⎜⎝⎛

    ∂∇∂

    ⋅∇−⎟⎠⎞

    ⎜⎝⎛

    ∂∂

    ∂∂

    uuP LL

    t

    Transverse vibrations of a plate, n = 2

    0)υE,(:d 22

    =∇⋅∇−∂∂ yf

    tyP

    Sound propagation in free air, n = 3

    Tcpctp γR,0:d 222

    2

    ==∇−∂∂P

    Normal mode solution tietq ω)(),( xQx =

    eigenmodes 0QQ =−∇⋅∇− 2ωf

    homogenous BCs ii ni QQ ⇒≤≤⇒ ,1,ωω

    22

    CMn.30 Mechanical Vibrations, n-D Continuum Systems

  • T ransverse vibrations of a plate

    0),(:BCs

    0)υ,(:d 22

    =

    =∇⋅∇−∂∂

    ty

    yEfty

    bx

    P

    normal mode solution tiety ω)(Q),( xx =

    GWSh for eigenmodes [ ] }0{}{]MASS[ω]STIFF[ 2 =− Qhomogeneous BCs det([MASS]-1 [STIFF] - ω i2[I]) = {0}

    G W S h normal mode so lutions , ω ih = 45 , 71 , 99 for i = 7 , 12 ,

    CMn.31 Transverse Vibrations, L-Shaped Membrane