Tundish Operation of a Two Fluid Model Using Two Scale k-ε ...
Coarse-Grid Simulation of Riser Flows · 2020. 8. 13. · Original two-fluid model and constitutive...
Transcript of Coarse-Grid Simulation of Riser Flows · 2020. 8. 13. · Original two-fluid model and constitutive...
-
1
Coarse-Grid Simulation of Riser Flows
Yesim Igci & S. SundaresanPrinceton University
NETL 2009 Workshop on Multiphase Flow ScienceMorgantown, WV
April 22, 2009
• DOE – UCR program• ExxonMobil Research & Engg. Co.
Art Andrews, Peter Loezos, Kapil Agrawal S. Pannala, M. Syamlal, T. O’Brien, S. Benyahia, R. BreaultNick Jones, Alvin Chen, Tim Healy, Rob Johnson, Rutton Patel
-
2
Roadmap for dilute gas-particle flows
Report on Workshop on Multiphase Flow Research, Morgantown, WV, June 6-7, 2006
Near-term (by 2009)• B3: Develop coarse-grained (filtered) two-fluid models
Mid-term (by 2012)•B3: Develop procedures to coarsen models for reactive flows
Connection to roadmap• Performed highly resolved simulations of a kinetic theory
based two-fluid model for gas-particle flow in various test domains
• Filtered the results to learn about constitutive relations for the filtered two-fluid model
• Verified the fidelity of the filtered model. Validation remains to be done.
-
Characteristics of flows in turbulent fluidized beds & fast fluidized beds
• Persistent density and velocity fluctuationsWide range of length and time scales
• Identifiable macroscopic inhomogeneous structures
• (Kinetic theory based) two-fluid models are able to capture the above characteristics
But, they require high resolution and large computational times
• Do we really want to resolve all the fine structures?If we do not, then how should we modify the two-fluid model?
3
-
What we expect based on experimental data
Solution of discretized form of the kinetic theory based two-fluid model
What we get
30 m tall
76 cm channel width
75μm particles
2 cm grid
Gas vel = 6 m/sGas vel = 6 m/sSolids flux = 220 kg/m2.sSolids flux = 220 kg/m2.s
2-D simulations
Solids fraction field
Red – high
Blue -low
4
-
What we get in narrow channels with fine grids
Solution of discretized form of the kinetic theory based two-fluid model
What we get in wide channels and coarse grids
But in smaller channels and with finer grid resolution, kinetic theory based model does produce the right kind of spatial and temporal variations
5
-
Multiphase flow computations via two-fluid models
Density contour showing particle-
rich streamersIndividual
particles in gas
Reaction engineering need:Tools to probe macro-scale reactive flow features directly
All the constitutive models for the two-fluid models arefor nearly homogeneous
mixtures
Overview of the coarse-graining (filtering)
6
-
Modeling challenge for gas-particle flows
Original two-fluid model and constitutive relations
Significant advances in the past three decades
Filtered two-fluid modelModified constitutive relations
forhydrodynamic termsspecies and energy
dispersioninterphase heat and mass
transfer rateseven modified reaction rate
expressions!
Develop models that allow us to focus on large-scale flow structures, without ignoring the possible consequence of the smaller scale structures.
7
-
8
First Step in development of Filtered Model
Density contour showing particle-
rich streamers
Individual particles in gas
Approach: Probe details of mesoscale structures and develop effective
coarse-grained equations
8Igci, et al., AIChE J., 54, 1431 (2008)
-
Filter “data” generated through highly resolved simulations of two-fluid models
Snapshot of particle volume
fraction fields obtained in highly
resolved simulations of gas-
particle flows. Squares illustrate
regions (i.e. filters) of over which
averaging over the cells is
performed.
9Igci, et al., AIChE J., 54, 1431 (2008)
The filtered drag coefficient, particle phase pressure and viscosity are now functions of particle volume fraction and filter size.
-
Wall correction to the filtered closures
2-D Kinetic theory based simulations
Height: 500 cm
Width:
The inlet gas superficial velocity: 93 cm/s.
The inlet particle phase superficial velocity is 2.38 cm/s.
The inlet particle phase volume fraction is 0.07.
Partial-slip BC for particle phase and free-slip BC for gas phase
20 cm, 30 cm, 50 cm
Grid size: 0.25 cm (0.514 dimensionless units)
75 75 μμmm particles in airparticles in air
10
-
The filtered drag coefficient is noticeably different in the core and the wall regions
0 10 20 30 40 50
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
Dimensionless horizontal location
Sca
led
filte
red
drag
coe
ffici
ent
Domain width: 102.80Domain width: 61.86Domain width: 41.12
The wall correction for the filtered drag coefficient is independent of channel width.
The same conclusion applies for filtered particle phase pressure and viscosity.
, with wall correction, scaled *
( , r)
( )filtered s
filtered
filtered s
β φβ
β φ=
filter size
*
,( ) ( , )filtered s filtered periodic s Frβ φ β φ Δ=
Dimensionless filter size:
filter size
filter size2
1
t
gV FrΔ
Δ=
Dimensionless distance:
distance
distance2
1
t
gV FrΔ
Δ=
Filter size: 2.056
Grid size: 0.514
Wall CoreDimensionless distance
11
-
10 20 30 40 500.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Dimensionless distance
Sca
led
filter
ed d
rag
coef
ficie
nt
Filter size: 4.112Filter size: 2.056
The wall correction for the filtered drag coefficient is nearly independent of filter size.
The same conclusion applies for filtered particle phase pressure and viscosity.
, with wall correction, scaled *
( , r)
( )filtered s
filtered
filtered s
β φβ
β φ=
filter size
*
,( ) ( , )filtered s filtered periodic s Frβ φ β φ Δ=
Dimensionless filter size:
Dimensionless distance:
filter size
filter size2
1
t
gV FrΔ
Δ=
distance
distance2
1
t
gV FrΔ
Δ=
Wall Core
Grid size: 0.514
The filtered drag coefficient is noticeably different in the core and the wall regions
12
-
Verification of the filtered model
Compare macroscopic featuresCompare macroscopic features
Original two-fluid model
Filtered two-fluid model
Solve a test problem using the original two-fluid model equations
Solve the same test problem using the filtered two-fluid model equations
1-D Linear Stability Analysis of filtered two-fluid model equations
13
-
Verification of the filtered model
Compare predictions of kinetic theory and filtered models
Height: 500 cm
Width: 30 cm
The inlet gas superficial velocity: 93 cm/s.
The inlet particle phase superficial velocity is 2.38 cm/s.
The inlet particle phase volume fraction is 0.07.
Free-slip boundary conditions
• Kinetic theory model: Grid size: 0.25 cm (0.514 dimensionless units)• Filtered model with closures corresponding to a filter size of 2 cm
75 75 μμmm particles in airparticles in air
14
-
Solution of discretized form of the filtered two-fluid model
15
0 100 200 300 400 5000
0.05
0.1
0.15
0.2
0.25
0.3
Elevation (cm)
Ave
rage
par
ticle
pha
se v
olum
e fra
ctio
n
Filtered model-- 1 cm gridFiltered model -- 0.5 cm grid
Filtered model with closures corresponding to a filter size of 2 cm
0 5 10 15 20 25 30
0.1
0.15
0.2
0.25
0.3
Distance (cm)P
artic
le p
hase
vol
ume
fract
ion
Filtered model -- 1cm gridFiltered mode -- 0.5 cm grid
Nearly grid‐size independent time‐averaged profiles when grid size is less than or equal to one‐half of the filter size.
-
Solution of discretized form of the filtered two-fluid model
16• Kinetic theory model: Grid size: 0.25 cm (0.514 dimensionless units)• Filtered model with closures corresponding to a filter size of 2 cm
-
Compare predictions of filtered models with different filter sizes
Height: 500 cm
Width: 50 cm
The inlet gas superficial velocity: 93 cm/s.
The inlet particle phase superficial velocity is 2.38 cm/s.
The inlet particle phase volume fraction is 0.07.
Free-slip boundary conditions
Verification of the filtered models
• Kinetic theory model: Grid size: 0.25 cm (0.514 dimensionless units)• Filtered model with closures corresponding to a filter sizes of 2 and 4 cm
75 75 μμmm particles in airparticles in air
17
-
Solution of discretized form of the filtered two-fluid models
18
Filtered model with closures corresponding to filter sizes of 2 and 4 cm yield nearly the same time‐averaged results as long as grid size is less
than equal to 0.5 (filter size)
0 100 200 300 400 5000
0.05
0.1
0.15
0.2
0.25
Elevation (cm)
Ave
rage
par
ticle
pha
se v
olum
e fra
ctio
n
Filter size: 4cm; Grid size: 2cmFilter size: 4cm; Grid size: 1cmFilter size: 2cm; Grid size:1cm
0 10 20 30 40 50
0.1
0.15
0.2
0.25
0.3
Distance (cm)P
artic
le p
hase
vol
ume
fract
ion
Filter size: 4cm; Grid size: 2cmFilter size: 4cm; Grid size: 1cmFilter size: 2cm; Grid size: 1cm
-
Solution of discretized form of the filtered two-fluid models
19Filtered model with closures corresponding to filter sizes of 2 and 4 cm
10 20 30 40-140
-120
-100
-80
-60
-40
-20
0
20
40
Distance (cm)
Par
ticle
pha
se m
ass
flux
(g /
cm2
s)
Filter size: 4cm; Grid size: 2cmFilter size: 4cm; Grid size: 1cmFilter size: 2cm; Grid size: 1cm
Elevation = 3m
-
20
Computational times
• Assumption: Irrespective of the process device size, accurate integration of the kinetic theory based model will require grids as small as 0.25 cm (or smaller) for 75 μm particles.
• 2D simulations in the 30 cm x 500 cm channel:• 2 cm filtered model with 1 cm grid size ~ 40 times faster than
the kinetic theory model• Had it been done in 3D, the ratio would be ~ 100
• With a 2 cm filter size, the filter volume (3D) ~ 8 cm3 ; the grid volume (3D) ~ 1 cm3
• Large scale devices: grid volume (3D) ~ 100 - 1000 cm3 • The filtered model is expected to be ~ 105 – 106 times faster
than the original model!• Thus, the filtered model converts a virtually impossible
problem to a manageable problem!
-
21
Summary
• Performed highly resolved simulations of a kinetic theory based two-fluid model for gas-particle flow in various test domains
• Filtered the results to learn about constitutive relations for the filtered two-fluid model
• Verified the fidelity of the filtered model
• Validation remains to be completed – in progress.• Filtered species and energy balance equations (and rate
of chemical reactions) remain to be developed: future research project.
-
Extra slides
22
-
The filtered particle phase pressure is noticeably different in the core and the wall regions.
The wall correction for the filtered particle phase pressure is nearly independent of filter size.
, , with wall correction, , scaled *
( , r)
( )s filtered s
s filtered
s s
PP
P
φ
φ=
Filter size
*
, ,( ) ( , )s s s filtered periodic sP P Frφ φ Δ=
filter size
filter size2
1
t
gV FrΔ
Δ=
distance
distance2
1
t
gV FrΔ
Δ=
Wall Core10 20 30 40 50
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Dimensionless distance
Sca
led
filter
ed p
artic
le p
hase
pre
ssur
e
Filter size: 2.056Filter size: 4.112
23
-
The filtered particle phase viscosity is noticeably different in the core and the wall regions.
, , with wall correction, , scaled *
( , r)
( )s filtered s
s filtered
s s
μ φμ
μ φ=
Filter size
*
, ,( ) 1.15 ( , )s s s filtered periodic s Frμ φ μ φ Δ= ×
filter size
filter size2
1
t
gV FrΔ
Δ=
distance
distance2
1
t
gV FrΔ
Δ=
The wall correction for the filtered particle phase viscosity isnearly independent of filter size.
Wall Core10 20 30 40 50
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Dimensionless distance
Sca
led
filter
ed p
artic
le p
hase
vis
cosi
ty
Filter size: 2.056Filter size: 4.112
24
-
Two-fluid model equations
( ) ( )
f f f f f f f f f f ftρ φ ρ φ φ ρ φ∂ +∇ ⋅ = − ∇ ⋅ − +
∂u u u f gσ
( ) ( ) 0ts s
s s s
ρ φρ φ
∂+∇ ⋅ =
∂u
( ) ( ) 0tf f
f f f
ρ φρ φ
∂+∇ ⋅ =
∂u
Solids
Fluid
Solids
Fluid
inertia solid phase stress interphaseinteraction
gravityeffective
buoyancy
( ) ( )
s s s s s s s s s f s stρ φ ρ φ φ ρ φ∂ +∇⋅ = −∇ ⋅ − ∇ ⋅ + +
∂u u u f gσ σ
1s fφ φ+ =
25
-
Filtered continuity equations
0
0
0
s s s f f f s
s s s s s s s s s
f f f f f f f f f
φ φ φ φ φ φ φ
φ φ φ
φ φ φ
′ ′ ′= + = + =
′ ′= + = =
′ ′= + = =
u u u u u u
u u u u u u
( ) ( ) 0ts s
s s s
ρ φρ φ
∂+∇ ⋅ =
∂u ( ) ( ) 0
tf f
f f f
ρ φρ φ
∂+∇ ⋅ =
∂u
All filtered quantities are functions of
space and time
( ) ( )( ) ( )
0t
0t
s ss s s
f ff f f
ρ φρ φ
ρ φρ φ
∂+∇ ⋅ =
∂
∂+∇ ⋅ =
∂
u
u
26
-
Filtered particle phase momentum balance
( ) ( )stress due to
sub-filter scalefluctuations
effective sub-filter scale
s s s s s s s s s s s s s f
s f drag
pt
p
ρ φ ρ φ ρ φ φ
φ
⎛ ⎞⎜ ⎟
∂ ⎜ ⎟′ ′+∇ ⋅ = −∇ ⋅ + − ∇⎜ ⎟∂⎜ ⎟⎜ ⎟⎝ ⎠
′ ′− ∇ +
14243u u u u u
f
σ
fluid-particleinteraction force
s sρ φ+1442443g
( ) ( )
s s s s s s s s s f drag s sptρ φ ρ φ φ ρ φ∂ +∇ ⋅ = −∇⋅ − ∇ + +
∂u u u f gσ
Upon filtering,
27
-
Filtered fluid phase momentum balance
( ) ( )stress due to
sub-filter scalefluctuations
effective sub-filter scaleflu
f f f f f f f f f f f f f
f f drag
pt
p
ρ φ ρ φ ρ φ φ
φ
⎛ ⎞⎜ ⎟
∂ ⎜ ⎟′ ′+∇ ⋅ = −∇ ⋅ − ∇⎜ ⎟∂⎜ ⎟⎜ ⎟⎝ ⎠
′ ′− ∇ −
14243u u u u u
f
id-particleinteraction force
f fρ φ+1442443g
( ) ( )
f f f f f f f f f drag f fptρ φ ρ φ φ ρ φ∂ +∇ ⋅ = − ∇ − +
∂u u u f g
Upon filtering,
28
-
Sub-filter scale correlations
stress due to stress due tosub-filter scale sub-filter scale
fluctuations fluctuations
effective sub-filter scalefluid-pa
s s s s f f f f
f f dragp
ρ φ ρ φ
φ
′ ′ ′ ′
′ ′− ∇ −
14243 14243u u u u
feffective sub-filter scale
rticle fluid-particleinteraction force interaction force
appearing in the fluid appearing in the particlephase equation phase equation
s f dragpφ
⎛ ⎞⎜⎜⎜
′ ′= − − ∇ +⎜⎜⎜⎜⎝ ⎠
1442443 1442443f
⎟⎟⎟⎟⎟⎟⎟
~ ,
, then
s f
s s f f
s s s s f f f f
If ρ φ ρ φ
ρ φ ρ φ
′ ′
>>
′ ′ ′ ′>>
u u
u u u u
So, only two terms to focus on
29
-
30
Sub-filter scale correlations
{
( )
fromkinetictheory
, ,2
s eff s effs s s s s
s f f sd efg fra
p
p
μ
β
ρ φ
φ
′ ′ + = +
′ ′− ∇ + = −
Iu u S
f u u
σ
sφAll the effective quantities will depend on the choice of filter size, and so on; such filter size dependence is present in LES of turbulent flows as well.
2
1
t
gV FrΔ
Δ=
Scaled filter size
-
0 0.05 0.1 0.15 0.2 0.250
0.1
0.2
0.3
0.4
0.5
particle-phase volume fraction
dim
ensi
onle
ss fi
ltere
d dr
ag c
oeffi
cien
t
Filtered drag coefficient decreases as filter size increases for both 2-D and 3-D
2
1
t
gV FrΔ
Δ=
eff t
s
Vg
βρ
Example: 75 μm; 1500 kg/m3; domain size = 8 cm
1 2 1cmFr Δ
= ⇒ Δ =
2-D
0.257
0.514
1.028
2.056
Domain size = 16 x 16
64 x 64 grids
16 x 16 x 16
0 0.05 0.1 0.15 0.2 0.250
0.1
0.2
0.3
0.4
0.5
particle-phase volume fraction
dim
ensi
onle
ss fi
ltere
d dr
ag c
oeffi
cien
t
64 x 64 x 64
3-D
31
-
0 0.05 0.1 0.15 0.2 0.25 0.3 0.350
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
particle-phase volume fraction
dim
ensi
onle
ss fi
ltere
d dr
ag c
oeffi
cien
t
Filter Size = 4.112
resolution = 256x256resolution = 512x512resolution = 1024x1024
Dependence of the filtered drag coefficient on resolution (2-D)
Domain size
=132x132 (dim.less)
= 64 cm x 64 cm
Filter size
= 4 (dim.less)
= 2 cm
Corresponds to 1/8 cm
32
-
0 0.05 0.1 0.15 0.2 0.25 0.3 0.350
0.1
0.2
0.3
0.4
0.5
0.6
particle-phase volume fraction
dim
ensi
onle
ss fi
ltere
d dr
ag c
oeffi
cien
t
Filter Size = 4.112
Domain size=32.896x32.896; resolution = 128x128Domain size=131.584x131.584; resolution=512x512
Filtered drag coefficient is independent of domain size (2-D)
Red: Domain size
= 32 x 32 (res: 128 x 128)
Green: Domain size = 128 x 128 (res: 512 x 512)
(dimensionless)
33
-
Filtered drag coefficient is independent of domain size (3-D)
0 0.05 0.1 0.15 0.2 0.25 0.3 0.350
0.1
0.2
0.3
0.4
0.5
particle-phase volume fraction
dim
ensi
onle
ss fi
ltere
d dr
ag c
oeffi
cien
tFilter Size = 2.056
Domain size=8.224³; resolution=32³Domain size=16.448³; resolution=64³
Blue: Domain size
= 8 x 8 x 8 (res: 32 x 32 x 32)
Green: Domain size = 16 x 16 x 16
(res: 64 x 64 x 64)
(dimensionless)
34
-
0 0.05 0.1 0.15 0.2 0.250
0.02
0.04
0.06
0.08
0.1
0.12
0.14
particle-phase volume fraction
dim
ensi
onle
ss fi
ltere
d pa
rticl
e-ph
ase
pres
sure
0 0.05 0.1 0.15 0.2 0.250
0.02
0.04
0.06
0.08
0.1
0.12
0.14
particle-phase volume fraction
dim
ensi
onle
ss fi
ltere
d pa
rticl
e-ph
ase
pres
sure
2
1
t
gV FrΔ
Δ=
,2
s eff
s t
pVρ
1 2 1cmFr Δ
= ⇒ Δ =
2-D
0.257
0.514
1.028
2.056
Domain size: 16 x 16
64 x 64 grids
16 x 16 x 16
64 x 64 x 64
3-D
Example: 75 μm; 1500 kg/m3; domain size = 8 cm
Filtered particle phase pressure increases as filter size increases for both 2-D and 3-D
35
-
Filtered particle phase pressure increases as filter size increases for both 2-D and 3-D
• The effective pressure in 3-D is smaller than that in 2-D (smaller by 40%)
• The effective pressure increases nearly linearly with filter size
• Effective pressure is by and large due to the mesoscale velocityfluctuations once the filter size exceeds 1 cm for FCC particles.
• This means that it may not be necessary to consider a filtered granular energy equation, if we use a large enough filter size
{ , ,fromkinetictheory
2 s s s s s s eff s effpρ φ μ′ ′ + = +Iu u Sσ
36
-
0 0.05 0.1 0.15 0.2 0.250
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
particle-phase volume fraction
dim
ensi
onle
ss fi
ltere
d pa
rticl
e-ph
ase
visc
osity
0 0.05 0.1 0.15 0.2 0.250
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
particle-phase volume fraction
dim
ensi
onle
ss fi
ltere
d pa
rticl
e-ph
ase
visc
osity
2
1
t
gV FrΔ
Δ=
,3
s eff
s t
gV
μρ
1 2 1cmFr Δ
= ⇒ Δ =
2-D0.257
0.514
1.028
2.056
Domain size = 16 x 16
64 x 64 grids
16 x 16 x 16
64 x 64 x 64
3-D
Example: 75 μm; 1500 kg/m3; domain size = 8 cm
Filtered particle phase viscosity increases as filter size increases for both 2-D and 3-D
37
-
Filtered particle phase viscosity increases as filter size increases for both 2-D and 3-D
• The effective viscosity in 3-D and 2-D are comparable
• The effective viscosity ~ (filter size)1.5–1.6
• Effective viscosity is by and large due to the mesoscalevelocity fluctuations once the filter size exceeds 1 cm for FCC particles.
• This means that it may not be necessary to consider a filtered granular energy equation
{ , ,fromkinetictheory
2 s s s s s s eff s effpρ φ μ′ ′ + = +Iu u Sσ
38
Coarse-Grid Simulation of Riser FlowsRoadmap for dilute gas-particle flowsCharacteristics of flows in turbulent fluidized beds & fast fluidized bedsWall correction to the filtered closuresThe filtered drag coefficient is noticeably different in the core and the wall regionsThe filtered drag coefficient is noticeably different in the core and the wall regionsVerification of the filtered modelVerification of the filtered modelVerification of the filtered modelsComputational timesSummaryExtra slidesThe filtered particle phase pressure is noticeably different in the core and the wall regions.The filtered particle phase viscosity is noticeably different in the core and the wall regions.Two-fluid model equationsFiltered continuity equationsFiltered particle phase momentum balanceFiltered fluid phase momentum balanceSub-filter scale correlationsSub-filter scale correlations