discretization of the convection term...introductory course on the fvm, in an advanced course on cfd...
TRANSCRIPT
FMIA
ISBN 978-3-319-16873-9
Fluid Mechanics and Its ApplicationsFluid Mechanics and Its ApplicationsSeries Editor: A. Thess
F. MoukalledL. ManganiM. Darwish
The Finite Volume Method in Computational Fluid DynamicsAn Advanced Introduction with OpenFOAM® and Matlab®
The Finite Volume Method in Computational Fluid Dynamics
Moukalled · Mangani · Darwish
113
F. Moukalled · L. Mangani · M. Darwish The Finite Volume Method in Computational Fluid Dynamics An Advanced Introduction with OpenFOAM® and Matlab ®
This textbook explores both the theoretical foundation of the Finite Volume Method (FVM) and its applications in Computational Fluid Dynamics (CFD). Readers will discover a thorough explanation of the FVM numerics and algorithms used in the simulation of incompressible and compressible fluid flows, along with a detailed examination of the components needed for the development of a collocated unstructured pressure-based CFD solver. Two particular CFD codes are explored. The first is uFVM, a three-dimensional unstructured pressure-based finite volume academic CFD code, implemented within Matlab®. The second is OpenFOAM®, an open source framework used in the development of a range of CFD programs for the simulation of industrial scale flow problems.
With over 220 figures, numerous examples and more than one hundred exercises on FVM numerics, programming, and applications, this textbook is suitable for use in an introductory course on the FVM, in an advanced course on CFD algorithms, and as a reference for CFD programmers and researchers.
Engineering
9 783319 168739
Discretization of the Convection Term
Chapter 12
Convection Term∂ ρφ( )∂t
+∇⋅ ρvφ( ) = ∇⋅ Γ∇φ( ) +Q∂ ρv( )∂t
+∇⋅ ρvv( ) = ∇⋅τ +B
∂ ρα( )∂t
+∇⋅ ρvα( ) = 0
ΩCρ
Δt′ P P( ) + CρU f
* ′ P ( )f
f∑ − ρ f
* D f ∇ ′ P ( ) f ⋅S f( )f∑ = −
ΩΔt
ρP* − ρP
o( ) + ρ f*U f
*( )f∑
⎡
⎣ ⎢ ⎢
⎤
⎦ ⎥ ⎥ − ′ ρ f ′ v f ⋅S f( )
f∑ − ρ f
* H ′ v [ ] f ⋅S f( )f∑
Transient-like Term
Advection-like Termaccount for compressibility effects
Figure 7. (a) Pressure contours and (b) Mach number distributions along the upper and lower walls for supersonic flow over a bump (Minlet¼ 1.4). Convergencehistory plots of the various algorithms using the (c) single-grid, (d ) prolongation grid, and (e) multigrid methodologies for supersonic flow over a bump (Minlet¼ 1.4).
20
Figure 6. (a) Pressure contours and (b) Mach number distributions along the upper and lower walls for transonic flow over a bump (Minlet¼ 0.675). Convergencehistory plots of the various algorithms using the (c) single-grid, (d ) prolongation grid, and (e) multigrid methodologies for transonic flow over a bump (Minlet¼ 0.675).
18
∂ ρE( )∂t
+∇⋅ ρvE( ) = −∇⋅ vP( ) +∇⋅ v ⋅τ( ) +Q
∂ ραC( )∂t
+∇⋅ ρvαC( ) = 0Fluid 1
Fluid 2
Fluid 3
Fluid 4
The simplicity of the convection term does not extend to its discretization (3 decades of research
and continuing)
Pressure Correction Equation
High Order Schemes
NotationP E
e
WEE
WW
P E
e
WEE
WW
P E
e
W EEWW
vf
vf
C D
f
U DDUU
D C
f
DD U UU
Convection Schemes
ϕC
vf
C DU DDUU
f
ϕf
vf
C DU DDUU
f
ϕCϕf
ϕD
vf
C DU DDUU
f
ϕCϕf
ϕU
vf
C DU DDUU
f
ϕDϕf
ϕU
ϕC
vf
C DU DDUU
f
ϕfϕU
ϕD
UPWIND
CENTRAL
SOU
DOWNWIND
QUICK
€
φ f = φC
€
φ f =φC + φD2
€
φ f =3
2φC −
1
2φU
€
φ f =3
4φC +
3
8φU −
1
8φD
€
φ f = φD
Convection Scheme Formula
UPWIND
CENTRAL
SOU(Second Order Upwind)
QUICK(Quadratic Upwind Interpolation
for Convective Kinematics)DOWNWIND
Convection Schemes
Convection Discretization
�
∇ ⋅ ρvφ( ) = 0
�
ρvφ( ) f ⋅Sff = nb(P )∑ = 0
Sw
P EW
N
S SE
NE
SW
NW
Se
Sn
Ss
(∆x)P
(δx)w (δx)e
(∆y)P
(δy)n
(δy)s
1m
v(2,1)
1m
Φ=1
Φ=1
Φ=1
�
ρeve ⋅Se( )φe + ρwvw ⋅Sw( )φw + ρnvn ⋅Sn( )φn + ρsvs ⋅Ss( )φs = 0
Sw Se
Sn
Ss
EW
N
S
C
NW NE
SW SE
Δy( )C
δ x( )w δ x( )e
δ y( )n
δ y( )s
Δx( )C
ew
n
s
x
y
Second -Order Upwind
�
ρeve ⋅Se( )φe = ρeve ⋅Se( ) 32φP −
12φW
⎛ ⎝ ⎜
⎞ ⎠ ⎟
�
φ f ,SOU = 32φC −
12φU
⎛ ⎝ ⎜
⎞ ⎠ ⎟
�
ρwvw ⋅Sw( )φw = ρwvw ⋅Sw( ) 32φW − 1
2φWW
⎛ ⎝ ⎜
⎞ ⎠ ⎟
�
ρeve ⋅Se( ) 32φP −
12φW
⎛ ⎝ ⎜
⎞ ⎠ ⎟ + ρwvw ⋅Sw( ) 3
2φW − 1
2φWW
⎛ ⎝ ⎜
⎞ ⎠ ⎟ +
ρnvn ⋅Sn( ) 32φP −
12φS
⎛ ⎝ ⎜
⎞ ⎠ ⎟ + ρsvs ⋅Ss( ) 3
2φS −
12φSS
⎛ ⎝ ⎜
⎞ ⎠ ⎟ = 0
�
aPφP + aNφNN=E ,W ,N ,S,WW ,SS,NN ,SS
∑ = 0
�
aE = aN = aEE = aNN = 0�
aW = 32
ρwvw ⋅Sw( ) − 12
ρeve ⋅Se( )
�
aS = 32
ρsvs ⋅Ss( ) − 12
ρnvn ⋅Sn( )
�
aWW = − 12
ρwvw ⋅Sw( )
�
aSS = − 12
ρsvs ⋅Ss( )
�
aP = − aNN=E ,W ,N ,S,EE ,WW ,NN ,SS
∑
Sw Se
Sn
Ss
EW
N
S
C
NW NE
SW SE
Δy( )C
δ x( )w δ x( )e
δ y( )n
δ y( )s
Δx( )C
ew
n
s
x
y
Deferred Correction
�
ρ f vf ⋅Sf( ) φ f ,SOU + φ f ,UPWIND −φ f ,UPWIND( )f = nb(P )∑ = 0
�
ρ f vf ⋅Sf( )φ f ,UPWINDf = nb(P )∑ = − ρ f vf ⋅Sf( ) φ f ,SOU −φ f ,UPWIND( )
f = nb(P )∑
�
aPφP + aNφNN=NB(P )∑ = bP
dc
ρvφSOU( ) f ⋅S ff∑ = 0
Unstructured Grids
P
N1
N2
SfN2
fN2
N3fN3
P
dPN1
fN1
D
C
Vf U
D
C
Vf
U
�
φD −φU = ∇φC ⋅ 2rCD
Unstructured Grid
€
φ f = φC +∇φ f ⋅ rCf
€
φ f = φC + 2∇φC −∇φ f( ) ⋅ rCf
€
φ f = φC +1
2∇φ f +∇φC( ) ⋅ rCf
Convection Scheme Formula
UPWIND no change
CENTRAL
SOU(Second Order Upwind)
QUICK(Quadratic Upwind Interpolation
for Convective Kinematics)DOWNWIND no change
SUDS
Numerical DispersionNumerical Diffusion Numerical Dispersion
NVF-SUDS
STOIC
φC
~
φf
~
0
11/2
3/4
1
UPWIND
DOWNWIND
SOU
SMART
FROMM
1/2
1/4CENTRAL
UPWIND
Computing the Gradient
�
∇φ dVV∫ = φ dS
∂V∫
�
∇φ V = ∇φ dVV∫
�
∇φP = 1VP
φ fS f∂VP
∫
�
∇φP = 1VP
φ fS ff = nb(P )∑
PN1
N4
N3
N2
f1
Sf
f2
f3f4
Gradient at the Cell Face
P
F1
F4
F3
F2
P
F1
F4
F3
F2
�
∇φ f = ∇φ f − ∇φ f ⋅ edc( )edc + φD −φCdCD
edc
Gradient Formulation of HO schemes
�
φ f = φC + φD2
− φD − 2φC + φU8
= φC + φD −φC2
− φD −φC + φU −φC8
= φC + φD −φC2
− φD −φC + φU −φD + φD −φC8
= φC + φD −φC4
− φD −φU8
= φC + 12∇φ f + ∇φC( ) ⋅ rCf
�
φ f = 32φC −
12φU
= φC + φC −φU2
= φC + φD −φU2
− φD −φC2
= φC + 2∇φC −∇φ f( ) ⋅ rCf
�
φ f = φC + φD2
= φC + φD −φC2
= φC + ∇φ f ⋅ rCf
QUICK SOU CENTRAL
Gradient Form of Schemes
€
φ f = φC +∇φ f ⋅ rCf
€
φ f = φC + 2∇φC −∇φ f( ) ⋅ rCf
€
φ f = φC +1
2∇φ f +∇φC( ) ⋅ rCf
Convection Scheme Formula
UPWIND no change
CENTRAL
SOU(Second Order Upwind)
QUICK(Quadratic Upwind Interpolation
for Convective Kinematics)DOWNWIND no change
High Resolution Schemes
φC >1~
φU=0~
φD=1~φf
~
1-Normalization Procedure
φC φDφU
φf
φC φDφU
φf
φC φDφU
φf
φC φDφU
φf
φC=1~
φU=0~
φD=1~
φf~
φC φDφU
φf
φC φDφU
φf
φC =0~
φU=0~
φD=1~
φf~
!φ = φ −φU
φD −φU
vf
vf
C DU DDUU
D CDD U UU
φC φDφU
φf
φC φDφU
φf
φC <0~
φU=0~
φD=1~
φf~
Normalized HO Schemes
!φ f =
φ f −φUφD −φU
!φU = 0
!φD = 1
!φC = φC −φU
φD −φU
FROMMQUICK
φC~
φf~
011/2
3/4
1
SOU
1/2CD 3/8
1/4
DW
UPWIND
€
φ f = φC φ f = φC
€
φ f =φC + φD2
φ f =12φC +
12
€
φ f =3
2φC −
1
2φU
φ f =
32φC
φ f =34φC +
18φD −
38φD
φ f =
34φC +
18
€
φ f = φD φ f = 1
Scheme Formula Normalized
UPWIND
CENTRAL
SOU(Second Order Upwind)
QUICK(Quadratic Upwind
Interpolation for
DOWNWIND
2-CBC Criteria
!φ f =
f !φC( ) continuous
f !φC( )=1 if !φC =1
!φC < f !φC( )<1 0< !φC <1
f !φC( )= 0 if !φC = 0
f !φC( )= !φC if !φC < 0 or !φC >1
⎧
⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪
⎩
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪
φin+1 = H φi−k
n ,φi−k+1n ,…,φi+k
n( )
∂H∂φ j
n ≥ 0 ∀j ∈ i− k,i+ k[ ]
min φNB ,φC( )≤φ f ≤max φNB ,φC( )
PositivenessFlux Limiter (TVD)Boundedness Criteria
φf~
φu~
φC~n
vf0
1
φC~
φf~
0
11/2
3/4
1 DW
UPWIND
φC~
φf~
011/2
3/4
1
UPWIND
DOWNWIND
SOU
SMART
FROMM
1/2
1/4CENTRA
L
!φ f
=12!φ C
+12
!φ f=3
2!φ C
!φ f=!φ C
!φ f = 1
!φ f=34!φ C+38
!φ f=!φ C+13
The Normalized Variable Diagram
φC~
φf~
011/2
3/4
1/2
1/4
MINMOD
φC~
φf~
011/2
3/4
1/2
1/4
OSHER
φC~
φf~
011/2
3/4
1/2
1/4
MUSCL
φC~
φf~
011/2
3/4
1/2
1/4
1 BOUNDED CD
φC~
φf~
011/2
3/4
1/2
1/4
SMART
φC~
φf~
011/2
3/4
1/2
1/4
STOIC
Bounded HR Schemes
φC~
φf~
011/2
3/4
1/2
1/4
OSHER
φC~
φf~
011/2
3/4
1/2
1/4
SMART
φC~
φf~
011/2
3/4
1/2
1/4
MUSCL
φC~
φf~
011/2
3/4
1/2
1/4
MINMOD
φC~
φf~
011/2
3/4
1/2
1/4
STOIC
φC~
φf~
011/2
3/4
1/2
1/4
1 BOUNDED CD
Implementation
P
N1
N2
SfN2
fN2
N3fN3
P
dPN1
fN1
φP ,φN1 ,φN2 ,φN31
φC ←φPφD ←φN1φU ←φD −∇φP ⋅2rPN1
2
!φC = φC −φU
φD −φU
3
φ f ← f !φC( ) φD −φU( ) +φU5
φC~
φf~
011/2
3/4
1/2
1/4
SMART
!φ f = f !φC( )
4
Deferred Correction
!mfφ fSMART
f =nb P( )∑ = !mfφ f
UPWIND
implicit" #$ %$f =nb P( )
∑ + !mf φ fSMART −φ f
UPWIND( )deferred
" #$$$ %$$$f =nb P( )∑
aPφPUPWIND + aFφF
UPWIND
F=NB P( )∑ = bP − !mf φ f
SMART −φ fUPWIND( )
deferred" #$$$ %$$$f =nb P( )
∑
aPφPUPWIND + aFφF
UPWIND
F=NB P( )∑ = bP + bP
dc−SMART
ρ fv fφ f( ) ⋅S f
f =nb P( )∑ = !mfφ f
f =nb P( )∑ = 0
ρ fv fφ fSMART ⋅S f = ρ fv f ⋅S f( )φ f
SMART
= !mfφ fSMART
= !mfφ fUPWIND
treat implicitly" #$ %$
+ !mf φ fSMART −φ f
UPWIND( )treat explicitly
" #$$$ %$$$
SUDS
Numerical DispersionNumerical Diffusion Numerical Dispersion
NVF-SUDS
STOIC
X
φ
0.5 1
0
0.5
1
UPWIND
MINMOD
OSHER
TVD-MUSCL
SUPERBEE
BJ-MUSCL
EXACT
φC
~
φf
~
0
11/2
3/4
1
UPWIND
DOWNWIND
SOU
SMART
FROMM
1/2
1/4CENTRAL
UPWIND
HR Schemes
4.2. Advection of a sinusoidal profile
This problem is similar to the previous one in geo-metry except that a sinusoidal profile is used. The sinu-soidal profile involves steep and smooth regions, as wellas an extremum point, making its simulation much moredemanding than the simple step profile. The profile isgiven as
/ ¼ sin p2max 1" absðy"0:1707Þ
0:1707 ; 0! "! "
06 y6 0:3414
0 otherwise
(
ð28Þ
The problem is depicted in Fig. 5(a), the same meshes asthe step-profile problem were used. Results are shown inFig. 5(b) and (c) for the coarse and fine meshes respec-
tively. As expected all the schemes suffer from a sub-stantial decrease in the numerical extremum, with itsvalue decreasing down to 0.48 for the UPWIND scheme.The SUPERBEE preserves on the coarse mesh more ofthe extremum value (0.83), and experience no loss inextremum on the fine mesh. The BJ-MUSCL, TVD-MUSCL and OSHER schemes results are much betterthan those of the UPWIND scheme and the MINMODscheme, while still experiencing on the coarse and finemeshes a decrease in the extrema down to 0.68 and 0.92respectively.
4.3. Advection of a double-step profile
A double-step profile is imposed at inlet to the squaredomain depicted in Fig. 6(a). The profile is given as
Fig. 5. Convection of a sinusoidal profile: (a) physical domain, (b) / profile at y ¼ 0:8 using coarse grid, (c) / profile at y ¼ 0:8 usingdense grid.
606 M.S. Darwish, F. Moukalled / International Journal of Heat and Mass Transfer 46 (2003) 599–611
4.2. Advection of a sinusoidal profile
This problem is similar to the previous one in geo-metry except that a sinusoidal profile is used. The sinu-soidal profile involves steep and smooth regions, as wellas an extremum point, making its simulation much moredemanding than the simple step profile. The profile isgiven as
/ ¼ sin p2max 1" absðy"0:1707Þ
0:1707 ; 0! "! "
06 y6 0:3414
0 otherwise
(
ð28Þ
The problem is depicted in Fig. 5(a), the same meshes asthe step-profile problem were used. Results are shown inFig. 5(b) and (c) for the coarse and fine meshes respec-
tively. As expected all the schemes suffer from a sub-stantial decrease in the numerical extremum, with itsvalue decreasing down to 0.48 for the UPWIND scheme.The SUPERBEE preserves on the coarse mesh more ofthe extremum value (0.83), and experience no loss inextremum on the fine mesh. The BJ-MUSCL, TVD-MUSCL and OSHER schemes results are much betterthan those of the UPWIND scheme and the MINMODscheme, while still experiencing on the coarse and finemeshes a decrease in the extrema down to 0.68 and 0.92respectively.
4.3. Advection of a double-step profile
A double-step profile is imposed at inlet to the squaredomain depicted in Fig. 6(a). The profile is given as
Fig. 5. Convection of a sinusoidal profile: (a) physical domain, (b) / profile at y ¼ 0:8 using coarse grid, (c) / profile at y ¼ 0:8 usingdense grid.
606 M.S. Darwish, F. Moukalled / International Journal of Heat and Mass Transfer 46 (2003) 599–611