This action might not be possible to undo. Are you sure you want to continue?
Modelling the bubble size distribution in
gasliquid reactors
with QMOM implemented with a new
correction algorithm
Miriam Petitti
A.Nasuti, D.L.Marchisio, M.Vanni, G. Baldi,
N. Mancini, F. Podenzani, A. Bennardo
Dept. Materials Science and Chemical Engineering, Politecnico di Torino(Italy)
ENI R&M, S.DonatoMilanese (Italy)
miriam.petitti@polito.it
2
Outline
oGasliquid systems modeling: challenges and issues
oSimulation of a gasliquid stirred tank (airwater):
oApproach
oDrag force evaluation
oPopulation Balance Model
oQMOM and correction algorithm
oResults:
oGas volume fraction
oBubble Size Distribution (BSD)
oConclusion and possible next steps
3
Gasliquid systems modeling: challenges and issues
o The simulation of polydispersed multiphase systems (i.e.,
constituted by droplets, particles and bubbles) is a complex
problem
o Gasliquid systems are the hardest to describe because fluid
particles have no fixed shape or size:
o Small bubbles à spherical
o Medium bubbles à ellipsoidal
o Big bubbles à spherical cup
o This has a tremendous impact on the interaction between the
liquid (continuous phase) and the bubbles (disperse phase)
for:
o Momentum phase transfer
o Mass phase transfer
o Enthalpy phase transfer
4
Phase coupling
Liquid
Bubble – Bubble Bubble
One – Way Coupling
Two – Way Coupling
Four – Way Coupling
Global gas holdup
Drag force
Virtual mass force
Lift force
Bubble breakup
Collisions between bubbles:
Coalescence and
momentum exchange
5
BREAKUP COALESCENCE
Three conceptual steps
Bubbles come in close proximity because of
turbulent eddies
The controlling mechanism is deformations
induced by turbulence
For gas bubbles binary breakage
is the most common event
Approach: Fourway coupling
o The fluid exerts a drag force on the bubbles that is related to
the slip velocity, bubble size, turbulence intensity and bubble
concentration (details in backup slides)
o In addition there are the lift and virtual mass forces
LIQUIDBUBBLE FORCES
6
Approach
o The phases are simulated as interpenetrating continua that
obey transport equation and exchange momentum. Every
phase has its own mass and momentum conservation
equation.
o Phase interaction occurs via interfacial terms (especially
drag). Closeness of the model used for this terms to the
reality determines accuracy of the model.
o Most of the interfacial terms depend on particle size and
most dispersed flows are not monodispersed, so
knowledge of representative bubble diameter is essential.
Eulerian Multifluid Model (Fluent)
7
Approach
o Expressions of the drag term tailored specifically for
bubbles have been introduced as UDF
o A solver for the population balanceequation has been
included in the code to evaluate the bubble size
distribution in every point of the system. The solver is
based on the QMOM method, which allows accurate
estimation of the integral properties of the BSD with
moderate computational load (6 additional transport
equations)
o Proper expressions for bubble breakage and coalescence
rates have been tested and fitted to experimental data sets
8
Drag force evaluation
Usually in stirred reactors BSD is included in the range of
ellipsoidal bubbles and is not so wide to cause a significant
change in the bubble terminal velocity, according to
Mendelsons’slaw.
9
Drag force evaluation
The drag coefficient is evaluated from bubble terminal velocity: the d
32
of the local BSD is considered and a unique valuefor bubble terminal
velocity is used in the calculation.
The drag forceis evaluated on the basis of the local d
32
and the local slip
velocity between the continuous and dispersed phases.
( )
2
32
3
4
∞
−
·
U
d
C
l
g l
D
ρ
ρ ρ
holdup <5%: about 13 cm/s (turbulence)
holdup >5% :about 8.5 cm/s (turbulence+effect near bubbles)
The damping effect of turbulence and the effect of high gas holdup
are considered for the terminal velocity evaluation:
10
Population BalanceEquation
Its solution describes the evolution of bubbles size distribution:
Number
Density
Function
( )
( ) ( ) ( ) ( ) ( ) ( )
( )
( )
( )
( )
( ) ( ) ( ) ( )
1 3 1 3
3 3 3 3
0 0
, ,
, , , , , , ,
1
, , , , , , , , , ,
2
d
d
d
d d d d d d d d d
L
L
d d d d d d d d d d d d
n L x t
U n L x t g b L n x t d g L n L x t
t
h L n L x t n x t d n L x t h L n x t d
λ λ λ λ
λ λ λ λ λ λ λ λ
∞
∞
∂
1
+∇⋅ · − +
¸ ]
∂
+ − − −
∫
∫ ∫
B
b
D
b
B
c
D
c
birth due tobreakage death due tobreakage
birth due tocoalescence death due tocoalescence
11
Kernels (or frequencies) quantifies the rate with which bubbles
COALESCENCE and BREAK.
Several expressions are available in the literature but in this work the
following have been used (Laakkonen et al., 2006):
Coalescence and breakup rates
( )
1/3
1 2 3
2/3 5/3
1/3 4/3
erfc
c
d
c d
c d d
g L C C C
L
L
µ σ
ε
ρ ε
ρ ρ ε
¸ _
· +
¸ ,
BREAKUP KERNEL
( ) ( )
( )
( )
1/ 2
2
1/ 3 2/ 3 2/ 3
7
, ,
d d d d d d d d
h L C L L L λ ε λ λ η λ · + +
( )
4
8
2
, exp
c c d d
d d
d d
L
L C
L
µ ρ ε λ
η λ
σ λ
¸ _
¸ _
· −
+
¸ ,
¸ ,
C
7
=0.88
(from theory)
COALESCENCE KERNEL
Binary breakage, C
1
from fitting
12
The PBE is solved in terms of the first k =2N moments of the Number Density
Function (NDF) n(L):
0,1, 2,3, 4,5 k·
( ) ( )
ˆ
k
d k d d k d k
m u m S
t
ρ ρ ρ
∂
+∇ ·
∂
The closure on the moments source terms S
k
, concerning bubble coalescence and
breakup, is obtained by resorting to a quadratureapproximation :
QuadratureMethod of Moments
( )
∫
+∞
≡
0
dL L L n m
k
k
where the moments m
k
refer to bubble size L:
L
2
n(L)
L
L
1
w
1
w
2
L
3
w
3
N =3
( ) ( ) ( )
i
N
i
i
L f w dL L f L n
∑
∫
·
+∞
≈
1
0
weights nodes
the first 6 moments
are tracked
( ) ( ) ( ) ( ) ( )
/3
( ) 3 3
1 1 1 1 1 1
1
, ,
2
d d d d d d
N N N N N N
k
k k k k
k i i i i i i i j i j i j i i i j j
i i i j i j
dm
S g L wb L g L w w w L L h L L L w h L L w
dt
· · · · · ·
· · − + + −
∑ ∑ ∑ ∑ ∑ ∑
B
bk
B
ck D
ck
D
bk
13
QMOM and Moments corruption
The population balance is coupled with the multifluid model through the
moment transport equations. The moments can be considered as scalars that are
convected, mixed and diffused. However they are linked by mathematical
relationships that assure the physical existence of the underlying distribution.
The independent transport of the moments does not preserve these relationships,
that are altered by the discretization schemes used by the CFD code, and can
therefore create a set of corrupted (and not valid) moments.
Non  validmoment set
Residuals
I nstabilityproblems
Negative moments, relationships altered
14
QMOM and Moments validity test
The validity of the moment set is checked by asserting the positivity of the
determinants of HankelHadamard (Shohat e Tamarkin, 1943):
1
1 2 1
,
1 2
...
...
0
...
n n n l
n n n l
nl
n l n l n l
m m m
m m m
m m m
+ +
+ + + +
+ + + +
∆ · ≥
M M M M
for n =0,1 and l ≥ 0
These conditions for the first few moments {m
0
, m
1
, m
2
, m
3
}are equivalent to the
convexity check of the curve ln(m
k
) in function of k:
2
2 1
0
k k k
mm m
− −
− ≥
k
ln (m
k
)
k
ln (m
k
)
valid nonvalid
15
QMOM and Moments correction
Adjust that moment which
after adjustment maximizes
smoothness through minimization
of a vector related to third order
differences
Initial moments set
Formsecond order
differences in ln(m
k
)
Satisfies convexity
Formthird order differencesin
ln(m
k
): verctor a
Determinek* for which
vector ais minimized
Adjust mk* toget new
moment sequence
check m4 and m5
Stop: moments
ok
Minimum Square
Gradient Algorithm
(McGraw 2006)
16
QMOM and Moments correction
1) I n certain cases the iterative algorithmenters an infinite loop and
does not manageto correct the moment sequence.
In thesecasesthe moment set is rebuilt through two lognormal
distributions whoseparametersare obtained fromthe first 4 moments
of the original set that must becorrected (Wright 2007)
NDF(1) fromm
0
, m
2
, m
3
NDF(2) fromm
0
, m
1
, m
3
2 2
exp
2
k T
k
m N k
σ
µ
¸ _
· +
¸ ,
( )
( )
2
2
ln
exp
2
2
L T
L
n L N
L
µ
σ
σ π
1
− −
1
1
¸ ]
· ⋅
NDF: moments:
Moments evaluated as averages of the moments of the two distributions
2) Control on gas volume fraction, particularly useful at higher gassing
rates: if gas vof <10
3
sources =0; m
k
=0; d
32
=db
inlet
Without this changeat higher gassing rates it is observed theformation
of zoneswith m
5
<0, tending to spread in neighbouring zones
17
Effects of moments correction
Negative
values!
A
0
≅  10
4
B
10
3
0
10
3
0
m
2
without correction m
2
with correction
Residuals
No instability
problemswhen
correctionapplied,
alsowithhigh
underrelaxation
factors
Residuals
I nstability
problems
18
Effects of moments correction
d
32
without correction
d
32
with correction
The correction algorithmdoes not alter the physical behaviour of the system,
the bubblesizedistribution
“Lucky” simulation with low
numerical diffusion for the moments
19
Standard Rushton turbine
Baffles and blades width: 1.03 cm
Shaft diameter: 3.3 cm
Disk diameter: 13.6 cm
Turbine diameter: 21cm
Disk width: 1.06 cm
Sparger position: z =10.5 cm; d =3.3 cm
Sparger width d =15 mm
Reactor volume: 194 liter
Porous sparger
Reactor configuration  1
rotating
fluid
baffles
shaft
impeller
sparger
Computational domain on half reactor
à N
cells
: 226776
20
Ring sparger (12 holes 2 mm of diameter)
Computational domain on half reactor
à N
cells
: 226776
Standard Rushton turbine
Baffles and blades width: 1.03 cm
Shaft diameter: 3.3 cm
Disk diameter: 13.6 cm
Turbine diameter: 21cm
Disk width: 1.06 cm
Sparger position: z =10.5 cm; d =3.3 cm
Sparger width d =15 mm
Reactor configuration  2
Reactor volume: 194 liter
sparger
zona rotante
baffles
albero
impeller
sparger
zona rotante
baffles baffles
albero
impeller
rotating
fluid
shaft
impeller
sparger
baffles
21
Resultsfluid dynamics
The fluid dynamics predicted by the model wasalso tested for a
standard geometry stirred reactor of 15.4 liter, agitated by a Rushton
turbine, in termsof:
Velocity profiles
Fluid dynamicsregime transitions
Gas cavity structuretransitions
 Power dissipated
The agreement with experimental data and/or
empirical correlations is good
22
Results – gas distribution
Reactor
configuration 1
GAS VOLUME FRACTI ON
23
0.00
0.02
0.04
0.06
0.08
0 0.4 0.8 1.2
gas flow rate, vvm
g
l
o
b
a
l
g
a
s
h
o
l
d

u
p
Results – Gas holdup
N =390 rpm
Reactor configuration 2
T
R
A
S
I
T
I
O
N
A
L
R
E
G
I
M
E Experimental
data
Simulation
24
8 different operating conditions for reactor configuration 1 and 7
different conditions for reactor configuration 2 were considered
resulting in the following best fit values:
C
1
=6 (from fitting) and C
7
=0.88 (from theory)
Results Bubble size distribution
G
a
s
f
l
o
w
r
a
t
e
,
v
v
m
Stirring rate, RPM
Reactor
configuration 1
Reactor
configuration 2
25
0
1
2
3
4
0 2 4 6 8
R2
2.90
2.77
R4
3.25
3.19
R8
2.43
2.90
R9
2.23
2.71
R12
3.34
3.30
Simulation results
Experimental data
Reactor configuration 1
N =250 RPM; 0.093 vvm
0
1
2
3
4
0 2 4 6 8
0
1
2
3
4
0 2 4 6 8
0
1
2
3
4
0 2 4 6 8
0
1
2
3
4
0 2 4 6 8
B
S
D
Mean bubblesize, mm
d, mm
R2 R4
R8
R9
R12
Results – Bubble size distribution
26
R2_A
2.24
2.86
R2_B
2.90
2.77
R2_C
2.25
2.74
R2_D
2.03
2.47
R2_E
2.92
3.37
Simulation results
Experimental data
0
1
2
3
4
0 2 4 6 8
0
1
2
3
4
0 2 4 6 8
0
1
2
3
4
0 2 4 6 8
0
1
2
3
4
0 2 4 6 8
0
1
2
3
4
0 2 4 6 8
R2_E R2_D
R2_C
R2_B
R2_A
B
u
b
b
l
e
v
o
l
u
m
e
d
e
n
s
i
t
y
d, mm
d32, mm
Reactor configuration 1 – N =250 RPM; 0.093 vvm
Results – Bubble size distribution
27
Results – Bubble size distribution
N =250 RPM; 0.052 vvm
N =157 RPM; 0.052 vvm
R2
R8
R9
R4
R12
R2
R8
R9
R4
R12
Reactor configuration 1
N =250 RPM; 0.072 vvm
Experimental
Simulation (45° )
Simulation (135°)
Simulation (45° )
Simulation (135°)
28
Results – Bubble size distribution
Reactor configuration 1
Experimental
Simulation
N =157 RPM; 0.052 vvm N =250 RPM; 0.052 vvm
N =250 RPM; 0.072 vvm
29
Results – Bubble size distribution
experimental
simulation
experimental
simulation
Reactor configuration 2 – N =390 RPM; 0.7 vvm
measurement points
d
3
2
,
m
m
30
Reactor configuration 2 – N =390 RPM
Results – Bubble size distribution
0.000
0.002
0.004
0.006
0.008
0 0.2 0.4 0.6 0.8 1 1.2
gas flow rate, vvm
m
e
a
n
b
u
b
b
l
e
s
i
z
e
,
m
Experimental data (derived
through emp. correlation)
Simulation
31
• Some solid guidelines for the simulation of gasliquid stirrer tanks in
realistic conditions have been identified and validated.
•Great attention must be paid to the calculation of the drag force(this
can be calculated via the bubble terminal velocity taking into account the
effect of turbulence and of the other bubbles for very high holdups)
• The algorithm for the calculation of the bubble size distribution (QMOM
+correction algorithm) was found to be stableunder very different
operating conditions and reactor configurations.
•The coalescence and breakage kernels and the drag law are able to
describe with satisfactory accuracy both the global hold up, the bubble
size distribution and the mean bubble size.
•The simulation settings can be used to describe the behavior of the
stirrer tank under different operating conditions ranging from very low
gas holdup (0.5 %) to high gas holdup (7 % for the airwater system)
Conclusions
32
n Application to mass transfer in gasliquid systems
Mass transfer in gasliquid systems is very sensitive to the bubble size
distribution, because of the effect of interfacial area and because the
transfer coefficient depends on gasliquid slip velocity (which in turn
depends on bubble size). An important point is that there may be a
difference between the size range of bubbles that control fluid dynamics
(usually the larger ones) and those that determine interfacial area (the
smaller ones). This point could be captured well by proper use of the
population balance method.
n Gasliquid chemical reactions
Introduction of chemical reaction paths and interfacial mass transfer
models for process reactions in multicomponent twophase systems.
Probable need to consider heat generation/transfer effects as well.
Possible next steps
33
Thankyouverymuchfor your
kindattention
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.