You are on page 1of 15

Powder Technology 116 Ž2001.

246–260
www.elsevier.comrlocaterpowtec

Numerical studies of bubble formation dynamics in gas–liquid–solid


fluidization at high pressures
Y. Li, G.Q. Yang, J.P. Zhang, L.-S. Fan )
Department of Chemical Engineering, The Ohio State UniÕersity, 140 West 19th AÕenue, Columbus, OH 43210, USA
Received 31 March 2000; received in revised form 6 September 2000; accepted 20 September 2000

Abstract

A discrete phase simulation ŽDPS. is conducted to investigate multi-bubble formation dynamics in gas–liquid–solid fluidization
systems. A numerical technique based on computational fluid dynamics ŽCFD. with the discrete particle method ŽDPM. and volume
tracking represented by the volume-of-fluid ŽVOF. method is developed and employed for the simulation. A bubble-induced force ŽBIF.
model, a continuum surface force ŽCSF. model, and Newton’s third law are applied to account for the couplings of particle–bubble,
bubble–liquid and particle–liquid interactions, respectively. A close-distance interactive effect between colliding particles is considered in
the formulation of the particle–particle collision model. Two-dimensional simulations of the behavior of bubble formation from
multi-orifices in liquids and liquid–solid suspensions are conducted at high pressures up to 19.4 MPa under constant gas flow conditions.
Experiments are also conducted in this study to quantify the bubble formation behavior from a single orifice under comparable gas flow
conditions. The study indicates that the liquid flow dynamics induced by adjacent bubbles and bubble wake significantly affects the
multi-bubble formation process. The simulation results on the initial bubble size and bubble formation time under various pressures and
solids hold-up are found to be in good agreement with those obtained experimentally as well as those predicted based on the analytical
model developed earlier by the authors. q 2001 Elsevier Science B.V. All rights reserved.

Keywords: Gas–liquid–solid fluidization; Discrete phase simulation; Bubble formation; Bubble dynamics; Volume-of-fluid method; Dispersed particle
method

1. Introduction have significant effects on the behaviors of these pro-


cesses.
Bubble columns and three-phase fluidization systems The phenomena of bubble formation through the ori-
have been extensively applied in industry for chemical and fices connected to a gas chamber can be described in terms
petrochemical processing such as resid hydrotreating, of the dimensionless capacitance number Nc defined as
methanol synthesis, Fischer–Tropsch synthesis and ben- 4Vc g r lrp D 02 P w14,32x. Accordingly, bubbles are formed
zene hydrogenation w6x. The operation of these systems for mainly under two distinct conditions: constant flow condi-
such applications is commonly conducted under high pres- tions when Nc F 1, and variable flow conditions when
sures. Extensive studies have indicated significant effects Nc ) 1 w8x. The effects of pressure or gas density on the
of pressure and solids hold-up on the flow characteristics initial bubble size in the liquid have been extensively
in bubble columns and slurry bubble columns w11,12, examined. It is found that an increase in system pressure or
20,38,42x, especially on the gas hold-up and the bubble gas density significantly reduces the initial bubble size
size. The size of bubbles in gas–liquid–solid fluidized under variable flow conditions w15,33,39,43x. Under con-
systems is closely associated with the dynamics of bubble stant flow conditions, however, the effect of pressure on
formation, bubble coalescence and bubble breakup pro- the initial bubble size is insignificant w11,19x. There are
cesses. The system pressure and particle concentration relatively few studies concerning the effect of particle
presence on the initial bubble size. Luo et al. w19x and
Yang et al. w42x investigated the mechanism of bubble
)
Corresponding author. Tel.: q1-614-292-4935; fax: q1-614-292-
formation in liquid–solid suspensions at elevated pres-
3769. sures. It is found that the presence of particles increases
E-mail address: fan.1@osu.edu ŽL.-S. Fan.. the initial bubble size for both constant flow and variable

0032-5910r01r$ - see front matter q 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 0 3 2 - 5 9 1 0 Ž 0 0 . 0 0 3 9 3 - 4
Y. Li et al.r Powder Technology 116 (2001) 246–260 247

flow conditions, similar to the findings of Massimilla et al. lowed to be Astair-stepB aligned with more than one mesh
w23x, which was obtained at the ambient pressure. coordinate within a given cell in an improved piecewise
These experimental studies on bubble formation in liq- constantrstair-step VOF method w10x. In the piecewise
uids and liquid–solid suspensions were described in vari- linear method, the interfaces are reconstructed by the lines
ous empirical correlations and analytical models in the defined by a slope and intercept within each volume
literature. Marmur and Rubin w22x developed a mechanistic fraction control cell w44x, which provides a more accurate
model to account for the bubble formation process from a approximation.
single orifice submerged in an inviscid liquid. Pinczewski The VOF method is employed in commercial CFD
w27x and Terasaka and Tsuge w30x used the modified codes, and has been widely used in simulations of liquid–
Rayleigh equation to predict the bubble growth rate, for- solid systems. However, little was reported in the literature
mation time and pressure fluctuation in the chamber under regarding the CFD study of bubble formation in three-phase
different pressures. Luo et al. w19x and Yang et al. w42x fluidized systems, especially at elevated pressures. Man-
developed analytical models to account for the initial asseh et al. w21x simulated single bubble formation in water
bubble size in high-pressure liquid–solid suspensions un- at ambient pressure by using the VOF method. Li et al.
der various bubble formation conditions. The models took w17x and Zhang et al. w45x used the modified VOF method
into account various forces induced by particles, such as to examine the pressure effect on the rising behavior of a
the suspension inertial force and particle–bubble collision gas bubble, which is initially at rest inside the system and
force. Most of these models are based on the spherical is of circular shape. In the present study, numerical analy-
symmetry assumption, and their extension to nonspherical ses of the bubble formation from multiple orifices under
bubble formation is unsuitable. Moreover, due to the lim- different pressures are conducted based on the discrete
ited range of experimental parameters employed, most of phase simulation ŽDPS. method developed earlier by Li et
the correlations and analytical models proposed are re- al. w18x. The effects of pressure and particle concentration
stricted in their applicability. on the characteristics of bubble formation, such as the
Computational fluid dynamics ŽCFD. has offered a initial bubble size and bubble formation time under con-
viable approach to the study of local flow structures and stant gas flow conditions, are examined based on the
transient flow phenomena. To date, the gas–bubble-related computation. The rising behavior of a gas bubble immedi-
flows in bubble columns or slurry bubble columns have ately after its detachment from the orifice is also studied.
been simulated principally by three main numerical ap- The effects of the interactions between the leading bubble
proaches: the Eulerian–Eulerian multi-fluid model and the trailing bubbles, as well as that between two
w25,29,31,36,40x, the Eulerian–Lagrangian model w5,16,34x, adjacent bubbles, on the bubble formation behavior are
and the interface tracking method w35x. Since the emphasized. The effects of dynamic fluid flow induced by
Eulerian–Eulerian multi-fluid model treats the gas bubble adjacent bubbles and bubble wakes are also investigated.
as a pseudo-continuum phase and the Eulerian–Lagrangian The computation results are compared with the experi-
model treats the gas bubble as a non-deformable spherical ments, as well as with those calculated based on a reported
particle, both of these models are inappropriate for describ- analytical model.
ing deformable bubble behavior during the formation pro-
cess.
In the interface tracking method, two approaches can be
followed: the front-tracking method based on the La- 2. Discrete phase simulation method
grangian grid system and the volume-tracking method
based on the Eulerian framework. The front-tracking In order to investigate the characteristics of bubble
method constructs a mesh system at the interface and formation in liquid–solid suspensions and the behavior of
tracks the movement of the bubble surface w35x. The entire individual phases, a discrete phase simulation method is
Lagrangian grid mesh system moves with the interface. adopted for the numerical simulations. Specifically, the
The volume-tracking method, on the other hand, recon- Eulerian volume-averaged method, the Lagrangian discrete
structs the interface using marker particles w9x or some particle method ŽDPM. and the volume-of-fluid method
special functions such as fractional volume-of-fluid ŽVOF. are employed in this study to describe the motion of liquid,
w10x. The VOF method constructs the interface as part of particles and gas bubbles, respectively.
the solution based on the same grid system and is flexible
and efficient for describing the changes in the topology of
the gas–liquid interface. 2.1. GoÕerning equations for continuous phases
Two kinds of VOF techniques, the piecewise constant
and piecewise linear methods, are applied to reconstruct For the bubble formation process, liquid is the continu-
the free surfaces. Interfaces within each cell are assumed ous phase outside the gas bubble while gas is the continu-
to be lines aligned with one of local coordinates in the ous phase inside the gas bubble. The volume-averaged
piecewise constant method w26x, or are additionally al- Navier–Stokes equations considering the presence of parti-
248 Y. Li et al.r Powder Technology 116 (2001) 246–260

cles can be obtained for the continuous phases as Ža. 2.2. Gas bubble interface tracking
continuous phase continuity equation:
Based on the definition of F Ž x,t . given in Section 2.1,
E´ f
q = P Ž ´f U . s 0 Ž 1. the movement of the gas bubble interface can be tracked
Et based on the distributions of F Ž x,t . w10x. The gas–liquid
Žb. continuous phase momentum equation: interface exists in the computational cells where F Ž x,t .
lies between 0 and 1. From the values of F Ž x,t . in
EŽ ´f U . neighboring cells, the size and shape of gas bubbles can be
rf q r f= P Ž ´ f UU . constructed. The advection equation for F Ž x,t . is given
Et by:
s y=p q = P Ž ´ f t . q ´ f r f g q Fpf q F bl Ž 2. EF Ž x ,t .
q Ž U P = . F Ž x ,t . s 0 Ž 6.
where U is the velocity vector; ´ f is the hold-up of the Et
continuous fluid Ži.e., for the liquid–solid mixture outside The piecewise linear interface calculation ŽPLIC.
the gas bubbles ´ f stands for the liquid hold-up, ´ l , and method w44x is applied to reconstruct the bubble-free sur-
´ l q ´s s 1; for the gas phase with solid particles inside faces with a multidimensional algorithm to determine the
the gas bubbles ´ f stands for the gas hold-up, ´g , and slope of each reconstructed line. In order to simulate the
´g q ´s s 1.; r f is the density; p is the scalar pressure; t gas flow field inside the bubble as well as the liquid flow
is the viscous stress tensor; Fpf is the force acting on the field, a sharp variation in properties from one phase to the
continuous phase by the solid particles; and F bl is the other on the bubble interface is replaced by a smooth
interaction force between liquid and gas bubbles. variation from one phase to the other within a finite
Since the DPS method does not assume the particle interface thickness. The mixture properties are calculated,
phase as a pseudo-continuum phase, the effect of solids using a volume fraction weighted average method, as:
concentration is taken into account by the hold-up ´ f in
Eq. Ž2.. Therefore, the expression of the viscous stress r s r l P F Ž x ,t . q rg P Ž 1 y F Ž x ,t . .
tensor for the liquid is identical to that for a single phase
m s m l P F Ž x ,t . q mg P Ž 1 y F Ž x ,t . . Ž 7.
flow, i.e.,
T
t s 2 m S s m Ž =U . q Ž =U . Ž 3.
2.3. Particle goÕerning equations
Based on Newton’s third law of motion, the total forces
acting on particles yield a reaction force on the fluid. Thus, 2.3.1. Particle motion equations
the momentum transfer from particles to the continuous Theoretically, if the relative velocity between the liquid
phase is taken into account by the particle–fluid interac- and solid phases is zero, the liquid–solid suspension can
tion force term Fpf in Eq. Ž2.: be treated as one pseudo-homogenous medium. However,
this is not the case in the system of this study. The relative
Ý Ffpk motion between the liquid and solid phases is exhibited in
Fpf s y , x pk g V i j Ž 4.
DVi kj various bubble–particle interactive phenomena Že.g., parti-
cle penetrating bubble.. Therefore, it is necessary to study
where the subscript ij defines the location of a computa- the behavior of the solid phase separately by using the
tional cell; V and DV are the domain and volume of this Lagrangian trajectory method.
cell, respectively; x pk is the location of particle k; and Ffpk The motion of a particle in the flow field can be
is the fluid–particle interaction force acting on an individ- described in Lagrangian coordinates with its origin set at
ual particle, which will be discussed in Section 2.3.1. the center of the moving particle. Translation and rotation
The bubble–liquid interaction force, F bl , is obtained by are considered for the particle motion. The particle rota-
a continuum surface force ŽCSF. model w2x tional acceleration or deceleration is assumed to be negligi-
ble. However, the change of particle rotation due to colli-
F bl Ž x ,t . s "sk Ž x ,t . =F Ž x ,t . Ž 5. sion is considered as will be discussed in Section 2.3.2.
The motion of a single particle without collision is
where the plus sign is used for the liquid phase and the governed by Newton’s second law:
minus sign for the gas phase; s is the surface tension of
the bubble interface; k is the curvature of the gas-free d xp
surface; and F Ž x,t . is a scalar volume fraction function of s Up Ž 8.
dt
fluid, where F Ž x,t . s 1 in the liquid or liquid–solid mix-
ture, 0 - F Ž x,t . - 1 at the gas bubble surface, and F Ž x,t . dUp
mp s m p g y Vp r g q F bp q Ffp Ž 9.
s 0 in the gas bubble. dt
Y. Li et al.r Powder Technology 116 (2001) 246–260 249

where x p and Vp are the particle position and volume, force depends strongly on the local liquid hold-up in the
respectively, and the four terms on the right-hand side of vicinity of the particle under consideration. The effective
Eq. Ž9. are the gravity force, the buoyancy force, the drag coefficient can be obtained by the product of the drag
bubble-induced force on the particle and the fluid–particle coefficient for an isolated particle and a correction factor
interaction force. as given by w37x:
When particles approach the gas–liquid interface, the
surface tension force acts on the particles through the CXD s C D ´y4
f
.7
Ž 13 .
liquid film. Since the size of computational cell is larger where C D is a function of the particle Reynolds number,
than the thickness of the gas–liquid interface film, a Re p . For rigid spherical particles, the drag coefficient C D
bubble-induced force model is applied to the particle: can be estimated by the following equations w28x:
F bp s Vp sk Ž x ,t . =F Ž x ,t . Ž 10 . ° 24 Ž 1 q 0.15Re 0.687
. Re p - 1000
s~ Re
p
If the particle overcomes this bubble-induced force, the CD p Ž 14 .
particle would penetrate the bubble surface. If there is a ¢0.44 Re p G 1000
significant number of particles penetrating bubbles and
particle sizes are large enough, the bubble instability would The added mass force accounts for the resistance of the
clearly yield a reduction in the bubble size. However, in fluid mass that is moving at the same acceleration as the
the bubble formation process, the amount of the particles particle. For a spherical particle, the volume of the added
that actually penetrate is a handful compared to the num- mass is equal to one-half of the particle volume, Vp , so
ber of particles in the bulk and the bubble is able to that:
recover its original shape upon particle penetration. Al- 1 d
though it is difficult to observe the particle-penetrating FAM s r Vp Ž U y Up . Ž 15 .
phenomenon in the bubble formation process, this phe- 2 dt
nomenon has been experimentally observed under the free The Basset force induced by the particle acceleration or
bubbling conditions. More detailed accounts of the parti- deceleration in the liquid can be expressed by w24x:
cle–bubble penetration have been given in Chen and Fan
w3,4x. The particle-penetrating phenomenon manifests the t d Ž U y Up .
FBA s 3pm d p H0 K Ž t y t . dt Ž 16 .
validity of the discrete phase simulation method in ac- dt
counting for the interaction between the continuous liquid
where K Ž t y t . in Eq. Ž16. is given as
and discrete particles. The interaction between the particle
and gas bubble is represented by the source term F bp in
Eq. Ž9.. The interaction affects the VOF solutions through
° p Ž tyt .n
~
1r4

K Ž tyt . s
the liquid flow field. The balance between this force and ¢ r 2
p
other forces determines whether the particle could pene-
trate the gas bubble and whether the bubble could recover
its original shape. 1 Ž U q np y n .
3
2
¶•
1r2 y2

q p Ž tyt . ,
Due to the small particle size, the Saffman, Magnus and 2 rp n f H3 Ž Re . ß
pressure-gradient-related forces are ignored, and therefore,
the fluid–particle interaction force, Ffp , includes the drag f H Ž Re . s 0.75 q 0.105Re; Re s Ud prn . Ž 17 .
force, the added mass force and the Basset force. That is:
where n is the kinematic viscosity of the fluid, and U is
Ffp s FD q FAM q FBA Ž 11 . the mean stream velocity.

Note that the pressure gradient force is not included in


Eq. Ž11. as, for consistency w41x, this force has been 2.3.2. Particle–particle collision analysis
accounted for in the momentum equation of the continuous In this study, it is assumed that the collisions between
phase as given in Eq. Ž2.. The drag force acting on a spherical particles are binary and quasi-instantaneous, and
suspended particle is proportional to the relative velocity there is a sequence of collisions during each time step. An
between the phases and has the following form: equation similar to that in molecular dynamic simulation
w1x is used to locate the minimum flight time of particles
1 before any collision.
FD s CXD r A < U y Up < Ž U y Up . Ž 12 . Differing from the gas–solid flow case, the liquid inter-
2
facial effect is important when two particles move close to
where A is the cross-sectional area of the particle to the each other in liquid–solid systems, especially when the
direction of the incoming flow; and CXD is the effective distance between two particles is less than 0.1d p . Thus, the
drag coefficient. In the liquid–solid suspension, the drag close-distance interaction ŽCDI. model developed by the
250 Y. Li et al.r Powder Technology 116 (2001) 246–260

authors is used to calculate the particle contact velocity The particle collision is important in simulating bubble
upon collision. The model considers the strong damping formation behavior, especially for high solids hold-up
effect due to the liquid film prior to particle collision. conditions. The simulation without considering particle
Details of the CDI model are described in Zhang et al. collision leads to inappropriate non-uniformity of particle
w46x. distribution in the flow field and hence false flow field
Based on above assumptions, the kinematic analysis can information. The numerical instability may also occur as a
be conducted to obtain the velocities of particles after result of inappropriate particle accumulation in a computa-
collision. Assuming that the collision occurs between parti- tional cell. In the present model, only the binary-particle
cles a and b, the normal components after collision can be collision mechanism is considered, which limits the model
given by the definition of the restitution coefficient and the to low solids hold-up conditions Žless than 30–40 vol.%..
momentum conservation equation: For higher solids hold-up cases, the multi-particle collision
mechanism needs to be considered.
° X
UaN y UbN
X

se
~ UbN y UaN Ž 18 .
¢m U a
N N
a q m b Ub s m aUa
N
X
q m b UbN
X 3. Simulation conditions

where U N is the normal velocity of the particle Ž a or b . at 3.1. CVD-2 code


the contact point before or after Žwith superscript X . the
collision. The numerical simulations of bubble formation in liq-
Neglecting the effect of particle torsion during the uid–solid suspensions are conducted in this study by using
collision, the simplified Mindlin’s contact theory account- the two-dimensional code CVD-2 developed by the au-
ing for particle–particle sliding or sticking contact is ap- thors. This code is developed by modifying the program
plied to obtain the tangential components after the colli- Ripple w13x, intended for incompressible flows with free
sion. If the incident angle, defined as the ratio of the surfaces, to account for compressible flows and internal
particle–particle relative velocity in the tangential direc- flows inside a gas bubble, and by programming the La-
tion to that in the normal direction, is less than the critical grangian formulation of the particle phase flow and pro-
angle Ž a cr s tany1 Ž2 f k ., where f k is the friction coeffi- gramming the couplings among individual phases. The
cient., the sticking collision occurs: flowchart of the computation using the code CVD-2 is
given in Fig. 1. The simulations are carried out using the
X X
UaT s UbT Ž 19 . Cray T-94 supercomputer at the Ohio State Supercomputer
Center; about 1.2 h of CPU time are required to compute
Otherwise, the sliding collision occurs, in which w7x: the flow behavior occurring in 1 s of real time.
X X
Ž UaT y UbT . y Ž UaT y UbT . s 2 f k Ž UaN y UbN . Ž 20 .
where U T is the tangential velocity of particle Ž a or b . at
the contact point.
Eq. Ž19. or Ž20., together with the momentum conserva-
tion equation, Eq. Ž21., gives rise to the tangential veloci-
ties:
X X
m aUaT q m b UbT s m aUaT q m b UbT Ž 21 .
As mentioned in Section 2.3.1, the collision induces the
change of particle rotation. The angular velocities after the
collision are determined by:
X
Ia Ž v Xa y v a . s m a Ž UaT y UaT . P r a
½ Ib Ž v Xb y v b . s m b Ž UbT y UbT . P r b
X Ž 22 .

where v is the angular velocity of the particle Ž a or b .,


and I is the moment of inertia defined by I s Ž2 m p rp2 .r5.
The tangential velocities of the particle centers are
given as:
X X
UacT s UaT y v Xa r a
½ X
UbcT s UbT y v Xb r b
X Ž 23 .
Fig. 1. The flowchart of the program CVD-2.
Y. Li et al.r Powder Technology 116 (2001) 246–260 251

3.2. Boundary and initial conditions

In the simulations, an 80 Žwidth. = 50 mm Žheight.


domain is used with two orifices at the bottom, an exit at
the top and two side walls as boundary conditions. The
liquid initially fills the domain to a certain height. For the
case with particle presence, 1000 glass beads are randomly
positioned in the liquid. Particles then settle under gravity
in the liquid medium with an evenly distributed inlet
velocity. At a certain liquid inlet velocity, an almost
uniform solids hold-up of 22.3% is reached. A second
simulation case with a solids concentration of 37.9% is
conducted by using 1700 particles instead of 1000 parti-
cles. At the beginning of the simulation, gas is injected
Fig. 2. Comparison of initial bubble sizes obtained in 2-D and 3-D
into the liquid or liquid–solid suspension through two columns Ž P s 0.1 MPa, ´ p s 0.0..
orifices.

3.3. Selection of parameters level and the effect of liquid flow on the bubble formation
process is negligibly small. The liquid and particles used in
In the simulation, Paratherm NF heat transfer fluid, the experiments are identical to those employed in the
nitrogen, and glass beads of 0.8 mm in diameter and 2,500 numerical simulations. The system pressure and solids
kgrm3 in density are used as liquid, gas, and solid phases, concentration vary in the ranges of 0.1–10.3 MPa and
respectively. The physical properties of the fluids under 0–22 vol.%, respectively. A high-speed video camera Ž240
the simulation conditions are given in Table 1. Bubble framesrs. is used to obtain the images of bubbles emerg-
formation in liquids and liquid–solid suspensions is nu- ing from the orifice in the liquid, and an optic fiber probe
merically studied for three pressure conditions: 0.1, 6.6 is used to detect bubbles in the slurry. The bubbling
and 19.4 MPa, and for three solids hold-up conditions: frequency and initial bubble size are obtained based on the
0.0,0.223 and 0.379, with a computation time step of probe signals and photographs. The details of the high-
5 = 10y6 s. The terminal velocity of the particle in the pressure column, the experimental setup and the optic fiber
liquid is about 2.5 cmrs at ambient pressure and 1.4 cmrs probe system are given by Luo et al. w19x and Yang et al.
at a high pressure of 19.4 MPa. The restitution coefficient w42x.
used in the collision analysis is 0.9 and the frictional To verify the validity of the comparison between 2-D
coefficient is 0.3. These values are selected based on the simulations and 3-D experiments, bubble formation experi-
previous numerical and experimental studies using the ments are also conducted in a two-dimensional column.
same particles w46x. The gas inlet velocity is 14.0 cmrs. The cross-section of the 2-D column is rectangular with
The diameter of each orifice is 6 mm. The simulations are dimensions of 1.2 = 20 cm. Two orifice sizes Ž1.6 and 6
performed with 80 = 50 grids. mm in diameter. and two liquids Žwater and Paratherm NF
heat transfer fluid. are used in the 2-D experiments. The
3.4. Experiments 3-D experimental data for the 1.6-mm orifice in the
Paratherm NF heat transfer fluid are from our previous
In order to verify the simulation results, bubble forma- study w19x. The comparison of initial bubble sizes obtained
tion experiments from a single orifice at elevated pressures in the 2-D and 3-D columns is shown in Fig. 2. It can be
are carried out in a high-pressure slurry bubble column of seen that initial bubble sizes measured in the 2-D and 3-D
1.38 m in height and 0.102 m I.D. Nitrogen is injected into columns are almost identical for both orifices and liquids,
the liquid–solid medium through a single orifice of 6.0 indicating that the bubble formation process in the 2-D and
mm I.D. The particles are suspended by a uniform liquid 3-D columns is comparable. Therefore, it is reasonable to
flow. The liquid flow is maintained at a relatively low compare 3-D experiments with 2-D simulation results in
this study.
Table 1
Physical properties of the gas and liquid used in the simulations ŽT s
308C.
4. Results and discussion
Pressure ŽMPa. 0.1 6.6 19.4
Gas density Žkgrm3 . 1.1 73.9 215.6 4.1. Pressure effect on bubble formation
Liquid density Žkgrm3 . 868 877 895
Liquid viscosity ŽPa s. 0.0232 0.032 0.0399
Surface tension ŽNrm. 0.0292 0.0254 0.0233 The two-bubble formation process in the liquid with the
initial bed height of 30 mm is numerically simulated for
252 Y. Li et al.r Powder Technology 116 (2001) 246–260

three pressures of 0.1, 6.6 and 19.4 MPa at 308C. In order


to isolate the pressure effect on the bubble formation
behavior, the two orifices are intentionally set far apart to
avoid interaction between bubbles. The distance between
the two orifices in this case is 35 mm. It is shown in Fig. 3
that the two bubbles formed in this simulation are identical
and symmetric, which indicates that the distance between
the two orifices is large enough so that the interaction
between the two bubbles is negligible. The distance of 35
mm is also applied in the next two cases to isolate the
effects of particle concentration and bubble wake structure
on the bubble formation behavior. The simulated bubble
formation process at 6.6 MPa is shown in Fig. 3, indicating
two typical formation stages: bubble expansion and bubble
detachment. It can be observed in Fig. 3 that two gas
bubbles formed on the orifices expand to a certain volume
first ŽFig. 3a., and then bubbles begin to detach from the
orifices with narrow necks beneath them ŽFig. 3b.. Two
bubbles separate from the orifices ŽFig. 3c., then rise
rectilinearly, and then formation of new bubbles starts
ŽFig. 3d..
This bubble formation process is validated by the exper-
imental observations described in Section 3.4. A sequence
of bubble images showing the process of bubble formation
in liquids at elevated pressure is shown in Fig. 4. It can be
seen that a bubble growing at the orifice remains roughly
spherical until it detaches from the orifice. The pho-
tographs also show the bubble formation process compris-
ing expansion ŽFig. 4a., detachment ŽFig. 4b., separation
ŽFig. 4c. and rising ŽFig. 4d., which is in qualitative
agreement with the simulation results shown in Fig. 3.
Figs. 5 and 6 compare the simulation results at 6.6 and
19.4 MPa with those at ambient pressure. It can be seen
that the size of bubbles formed from the orifice at 0.1 MPa
is similar to that at 6.6 and 19.4 MPa ŽFig. 5., which
indicates that the effect of pressure on the initial bubble
size under constant flow conditions is not significant.
Although an increase in pressure increases the inlet gas
momentum and decreases the bubble surface tension, which
promotes the detachment of the bubble, the increase in
pressure also increases the liquid viscosity and decreases
the buoyancy force, which impedes the detachment of the
bubble. Therefore, the overall effect of pressure on bubble
formation is insignificant. This finding agrees with the
experimental findings of Luo at el. w19x. Since the simula-
tions at 0.1, 6.6 and 19.4 MPa are conducted under
constant flow conditions with the gas inlet velocity of 14.0
cmrs, the bubble formation times at different pressures are
also comparable ŽFig. 6. due to similar initial bubble sizes. Fig. 3. Simulation results of multi-bubble formation in liquids Ž ´ p s 0.0,
The simulated initial bubble sizes and the bubble forma- P s6.6 MPa..
tion time are in quantitative agreement with the experimen-
tal findings.
The simulation results are also compared with the pre- tion model by taking into consideration various forces
dictions based on the analytical model proposed by Luo et acting on the bubble, especially the forces induced by
al. w19x and Yang et al. w42x. This analytical model is particles, such as the suspension inertial force and the
developed based on the two-stage spherical bubble forma- particle–bubble collision force. The analytical model has
Y. Li et al.r Powder Technology 116 (2001) 246–260 253

Fig. 4. A sequence of bubble images showing the process of bubble formation in liquids under high pressures Ž P s 6.6 MPa, U0 s 14.0 cmrs, D 0 s 0.6
cm..

been verified by a large amount of experimental data under sion ŽFig. 7., which indicates that the presence of particles
various pressure and gas injection conditions. Therefore, prolongs the bubble formation process. This is because the
the calculated results of the analytical model serve as a presence of particles increases the inertial force acting on
benchmark for comparison purposes with this study’s sim- the forming bubbles, which impedes the bubble detach-
ulation results. The predictions by the analytical model are ment from the orifice. These results agree well with the
also shown in Figs. 5 and 6, and they agree with the experimental measurements and the predictions by the
simulation results well. analytical model ŽFig. 8..
Due to the longer formation time in the liquid–solid
4.2. Particle effect on bubble formation suspension under constant flow conditions, the initial bub-
ble size in the liquid–solid suspension is larger than that in
In order to study the particle effect on bubble forma- the pure liquid. As shown in Fig. 9, the variation of the
tion, the same case as the one given above for 6.6 MPa is solids hold-up from 0 to 22.3% yields an increase of about
simulated with 1000 and 1700 particles suspended in the 14% in bubble diameter, which is in good agreement with
liquid. The solids hold-ups in the liquid–solid suspension the experimental findings and the predictions by the ana-
are 22.3% and 37.9%, respectively. lytical model. However, as the solids hold-up increases
The simulation results are shown in Fig. 7, which also further to 37.9%, the effect of the solids hold-up on the
demonstrates the typical formation process of expansion bubble size becomes insignificant.
ŽFig. 7a and b., detachment ŽFig. 7c. and separation ŽFig.
7d.. The same time series are chosen as those in Fig. 3. It 4.3. Dynamic effect of bubble wake
can be noted that the bubble formation time is 75 ms in the The bubble wake structure beneath a single rising bub-
pure liquid ŽFig. 3. and 120 ms in the liquid–solid suspen- ble was numerically studied by Li et al. w17x. They con-

Fig. 5. Pressure effect on the size of bubbles formed in liquids Ž D 0 s 0.6 Fig. 6. Pressure effect on bubble formation time in liquids Ž D 0 s 0.6 cm,
cm, U0 s14.0 cmrs, ´ p s 0.0.. U0 s14.0 cmrs, ´ p s 0.0..
254 Y. Li et al.r Powder Technology 116 (2001) 246–260

Fig. 8. Effect of solids hold-up on the bubble formation time Ž D 0 s 0.6


cm, U0 s14.0 cmrs, P s6.6 MPa..

firmed that the DSP method could accurately simulate the


characteristics of the bubble wake structure. Therefore, this
method is used as a basis for the study of the effect of the
leading bubble wake on the trailing bubble in the bubble
formation process.
Since bubble formation is a dynamic process, numerical
simulations are performed to examine the continuous for-
mation process involving a series of bubbles from the
orifices at 6.6 MPa. As shown in Fig. 10, after the first two
leading bubbles separate, the formation process for the
trailing bubbles follows on. The typical stages of expan-
sion, detachment and separation of the trailing bubbles can
also be observed in Fig. 10b–d.
It is noted that the trailing bubbles are smaller than the
leading bubbles. Compared with the leading bubbles, the
size of the trailing bubbles in Fig. 10 is reduced by 40%.
The difference in size between two bubbles in the same
sequence is due to the dynamic characteristics of the flow
in the column. The corresponding flow field vectors of
Fig. 10 are shown in Fig. 11, in which the symmetric
dual-wake structures can be observed beneath the leading
bubbles. The dynamic fluid flow induced by the two
leading bubbles yields a strong fluid shear effect on the
trailing bubbles, rendering early detachment of the trailing
bubbles.

Fig. 7. Simulation results of multi-bubble formation in liquid–solid Fig. 9. Effect of solids hold-up on the initial bubble size Ž D 0 s 0.6 cm,
suspensions Ž ´ p s 0.223, P s6.6 MPa.. U0 s14.0 cmrs, P s6.6 MPa..
Y. Li et al.r Powder Technology 116 (2001) 246–260 255

Fig. 10. Simulation results with bubble wake effect Ž P s6.6 MPa, Fig. 11. Velocity vector field of multi-bubble formation corresponding to
Lor s 35 mm.. Fig. 10 Ž P s6.6 MPa, Lor s 35 mm..
256 Y. Li et al.r Powder Technology 116 (2001) 246–260

Based on the simulation, it is also noted that the 4.4. Effect of distance between orifices
fluctuations of flow in the liquid phase or liquid–solid
suspension significantly increase the instability of bubble In all previous simulations, bubbles are formed from
formation. The size of the bubble formed in the dynamic two orifices that are situated far apart Ž35 mm., and thus
fluid is found to be generally smaller than that of the there is little interaction between these two bubbles as they
bubble formed in the quasi-still fluid. rise in the column. Bubble interaction takes place when the

Fig. 12. Simulation results with bubble–bubble interaction Ž P s 6.6 MPa, L or s 20 mm..
Y. Li et al.r Powder Technology 116 (2001) 246–260 257

two orifices are situated closely. To study the interaction is conducted at 6.6 MPa. The simulation results are shown
of two bubbles from different orifices, a numerical simula- in Fig. 12. As can be seen from the figure, due to the
tion of two-bubble formation in a column with an orifice closeness of the orifices, the fluctuations of flow field
separation distance of 20 mm and a bed height of 42 mm induced by one forming bubble from one orifice have an

Fig. 13. Velocity vector field of multi-bubble formation corresponding to Fig. 12 Ž P s 6.6 MPa, L or s 20 mm..
258 Y. Li et al.r Powder Technology 116 (2001) 246–260

effect on the forming bubble from the other orifice. This is larger than Dcr , the interaction between two orifices
interactive effect yields varied bubble shapes, and an alter- would not take place. The size of bubbles emerging from a
nate detachment pattern of bubbles from two orifices ŽFig. single orifice without the interaction can be predicted by
12b–f.. In this simulation, initially the bubble from each the analytical model developed earlier by Luo et al. w19x
orifice rises rectilinearly ŽFig. 12a.; after detachment, how- and Yang et al. w42x. Thus, the critical distance Dcr can be
ever, the bubbles tend to move to the centerline, rise in estimated for a given orifice size and inlet gas velocity.
zigzag fashion, and break up at the free surface ŽFig. Under the simulation conditions considered in this study,
12d–f.. The corresponding flow field vectors are shown in the initial bubble size D is in the range of 8–10 mm as
Fig. 13, which explains the interactive behavior of the shown in Fig. 9. Therefore, the critical distance between
liquid phase. It is seen in the figure that due to the two orifices would be about 24–30 mm. From the simula-
closeness of the orifices, two forming bubbles induce a tion results, it can be seen that no nozzle interaction is
high liquid velocity in the area near the centerline. The observed when the distance between two orifices, L or , is
high velocity thus results in a lower pressure in this area 35 mm ŽFigs. 3, 7 and 10.. On the other hand, when Lor is
and drafts bubbles toward the centerline. In other words, reduced to 20 mm, nozzle interaction occurs ŽFig. 12.. The
bubble and bubble wake interaction leads to a zigzag current simulation results verify the above analytical crite-
bubble rising path, and induces a complicated wake flow rion. Further refinement and experimental verification of
field, which, in turn, affects the bubble formation behavior. the analytical criterion are in progress.
Numerical tests of the single bubble formation in a
domain as shown in Fig.14a indicate that the wall effect
cannot be ignored when Dw , the distance between the 5. Concluding remarks
bubble edge and the wall, is smaller than the bubble
diameter Ž D .. That is, when the width of the domain is Numerical simulations are performed to study the bub-
larger than three times the bubble diameter, the wall effect ble formation behavior in liquids and liquid–solid suspen-
on bubble formation can be ignored. Based on this numeri- sions at various pressures. Numerical studies indicate that
cal finding, a simple analytical criterion could be devel- pressure has a negligible effect on bubble formation under
oped to determine the critical distance between orifices the conditions of constant gas flow through orifices. On
beyond which the bubble interaction is negligible. the other hand, the bubble formation behavior varies sig-
For a multi-bubble formation case as shown in Fig. 14b, nificantly with the presence of particles. Specifically, the
it can be assumed that there is a symmetric plane between bubble formation time and bubble size are greater in
the two orifices, and the nozzle interaction effect is signifi- liquid–solid suspensions than those in pure liquids at given
cant when Ds , the distance between the bubble edge and gas and liquid velocities. The simulation also reveals the
the symmetric plane, is smaller than D. Thus, the critical dynamic characteristics of the multi-bubble formation pro-
distance between two orifices for negligible interaction, cess. It is confirmed that the leading bubble affects the
Dcr , would be 3 D. When the distance between two orifices formation process of the trailing bubble through the bubble
wake. It is also found that the behavior of bubble forma-
tion from multi-orifices is different from that from a single
orifice. The multi-bubble formation process is strongly
influenced by the complicated bubble wake flows induced
by the adjacent bubbles. The simulated bubble formation
behavior is shown to be in conformity with the experimen-
tal observations. Based on the simulation studies, the
critical distance between two orifices for negligible inter-
action is found to be three times the bubble diameter.

Notation
A Cross-sectional area of particle
CD Drag coefficient
CXD Modified drag coefficient
D Initial bubble diameter
Ds Distance between the bubble edge and the sym-
metric plane
Dw Distance between the bubble edge and the wall
Dcr Critical distance between two orifices for negligi-
ble interaction
Do Orifice diameter
Fig. 14. Schematic of the analytical criterion for nozzle interaction effect. d Diameter
Y. Li et al.r Powder Technology 116 (2001) 246–260 259

e Restitution coefficient Superscripts


X
F Volume fraction of fluid Value after particle–particle collision
F Force k Particle index
fH Correction function N Normal direction
fk Friction coefficient T Tangential direction
g Gravity acceleration
I Moment of inertia
Lor Distance between two orifices
m Mass Acknowledgements
Nc Dimensionless capacitance number
p Scalar pressure
This work was supported, in part, by the NSF Grant
P System pressure
CTS-9906591, the U.S. Department of Energy Grant
r Radius
DEFC22-95 PC 95051 with a Cooperative Agreement with
Re Reynolds number
Air Products and Chemicals, and the Ohio Supercomputer
S Rate-of-strain tensor
Center. The helpful assistance of Mr. Z. Cui on the
t Time
experimental work shown in Fig. 2 and the constructive
U Velocity vector
comments of Dr. C. Chen on the simulation techniques and
U Mean stream velocity; particle velocity
results are greatly acknowledged.
U0 Orifice gas velocity
V Volume
Vc Chamber volume
Õ Velocity References
x Position vector
w1x M.P. Allen, D.J. Tildesley, Computer Simulation of Liquids, Oxford
Science Publications, New York, 1987.
w2x J.U. Brackbill, D.B. Kothe, C. Zemach, A continuum method for
Greek letters modeling surface tension, J. Comp. Phys. 100 Ž1992. 335.
´ Hold-up w3x Y.-M. Chen, L.-S. Fan, Bubble breakage mechanisms due to colli-
k Curvature of free surface sion with a particle in a liquid medium, Chem. Eng. Sci. 44 Ž1989.
117.
m Dynamic viscosity w4x Y.-M. Chen, L.-S. Fan, Bubble breakage due to particle collision in
n Kinematic viscosity a liquid medium: particle wettability effects, Chem. Eng. Sci. 44
r Density Ž1989. 2762.
s Surface tension w5x E. Delnoij, J.A.M. Kuipers, W.P.M. van Swaaij, Dynamic simula-
t Viscous stress tensor; time tion of gas–liquid two-phase flow: effect of column aspect ratio on
the flow structure, Chem. Eng. Sci. 52 Ž1997. 3759.
V Computation domain w6 x L.-S. Fan, Gas – Liquid – Solid Fluidization Engineering,
v Angular velocity Butterworth-Heinemann, Boston, 1989.
a Angle w7x L.-S. Fan, C. Zhu, Principles of Gas–Solid Flows, Cambridge Press,
New York, 1998.
w8x L.-S. Fan, G.Q. Yang, D.J. Lee, K. Tsuchiya, X. Luo, Some aspects
of high-pressure phenomena of bubbles in liquids and liquid–solid
Subscripts suspensions, Chem. Eng. Sci. 54 Ž1999. 4681.
AM Added mass w9x F.H. Harlow, J.E. Welch, Numerical calculation of time-dependent
a Particle index viscous incompressible flow of fluid with free surface, Phys. Fluids
ac Center of particle a 8 Ž1965. 2182.
w10x C.W. Hirt, B.D. Nichols, Volume of fluid ŽVOF. method for the
BA Basset
dynamics of free boundaries, J. Comp. Phys. 39 Ž1981. 201.
b Particle index; bubble w11x K. Idogawa, K. Ikeda, T. Fukuda, S. Morooka, Behavior of bubbles
bc Center of particle b of the air–water system in a column under high pressure, Int. Chem.
bl Bubble–liquid interaction Eng. 26 Ž1986. 468.
bp Bubble–particle interaction w12x J.R. Inga, Scaleup and scaledown of slurry reactors: a new method-
cr Critical point ology, PhD Thesis, Univ. of Pittsburgh, PA, 1997.
w13x B.K. Kothe, R.C. Mjolsness, M.D. Torrey, RIPPLE: a computer
D Drag program for incompressible flows with free surfaces, AIAA J. 30
f Continuous phase Ž1992. 2694.
fp Fluid–particle interaction w14x R. Kumar, N.R. Kuloor, The formation of bubbles and drops, Adv.
g Gas phase Chem. Eng. 8 Ž1970. 255.
w15x R.D. LaNauze, I.J. Harris, Gas bubble formation at elevated system
ij Cell indices
pressures, Trans. Inst. Chem. Eng. 52 Ž1974. 337.
l Liquid phase w16x A. Lapin, A. Lobbert, Numerical simulation of the dynamics of two
p Particle phase phase gas–liquid flows in bubble columns, Chem. Eng. Sci. 49
pf Particle–fluid interaction Ž1994. 3661.
260 Y. Li et al.r Powder Technology 116 (2001) 246–260

w17x Y. Li, J. Zhang, L.-S. Fan, Discrete-phase simulation of single w32x H. Tsuge, S. Hibino, Bubble formation from an orifice submerged in
bubble rise behavior at elevated pressures in a bubble column, liquids, Chem. Eng. Commun. 22 Ž1983. 63.
Chem. Eng. Sci. 55 Ž2000. 4597. w33x H. Tsuge, Y. Nakajima, K. Terasaka, Behavior of bubbles formed
w18x Y. Li, J. Zhang, L.-S. Fan, Numerical simulation of gas–liquid–solid from a submerged orifice under high system pressure, Chem. Eng.
fluidization systems using a combined CFD-VOF-DPM method: Sci. 47 Ž1992. 3273.
bubble wake behavior, Chem. Eng. Sci. 54 Ž1999. 5101. w34x J.A. Trapp, G.A. Mortensen, A discrete particle model for bubble
w19x X. Luo, G.Q. Yang, D.J. Lee, L.-S. Fan, Single bubble formation in slug two phase flow, J. Comput. Phys. 107 Ž1993. 367.
high pressure liquid–solid suspensions, Powder Technol. 100 Ž1998. w35x S.O. Unverdi, G. Tryggvason, A front-tracking method for viscous,
103. incompressible, multi-fluid flows, J. Comp. Phys. 100 Ž1992. 25.
w20x X. Luo, D.J. Lee, R. Lau, G.Q. Yang, L.-S. Fan, Maximum stable w36x C. Webb, F. Que, P.R. Senior, Dynamic simulation of gas–liquid
bubble size and gas hold-up in high-pressure slurry bubble columns, dispersion behavior in a 2-D bubble column using a graphics
AIChE J. 45 Ž1999. 665. mini-supercomputer, Chem. Eng. Sci. 47 Ž1992. 3305.
w21x R. Manasseh, S. Yoshida, M. Rudman, Bubble formation processes w37x C.Y. Wen, Y.H. Yu, Mechanics of fluidization, Chem. Eng. Prog.
and bubble acoustic signals,ICMF ’98, 3rd Int. Conf. on Mul. Flow, Symp. Ser. 62 Ž1966. 100.
Lyon, France, June 8–12, 1998. w38x P.M. Wilkinson, Physical aspects and scale-up of high-pressure
w22x A. Marmur, E. Rubin, A theoretical model for bubble formation at bubble columns, PhD Thesis, Univ. of Groningen, Groningen, The
an orifice submerged in an inviscid liquid, Chem. Eng. Sci. 31 Netherlands, 1991.
Ž1976. 453. w39x P.M. Wilkinson, L.L. van Dierendonck, A theoretical model for the
w23x L. Massimilla, A. Solimando, E. Squillace, Gas dispersion in solid– influence of gas properties and pressure on single-bubble formation
liquid fluidized beds, Br. Chem. Eng. 6 Ž1961. 232. at an orifice, Chem. Eng. Sci. 49 Ž1994. 1429.
w24x R. Mei, R.J. Adrian, Flow past a sphere with an oscillation in the w40x Y. Wu, D. Gidaspow, Hydrodynamic simulation of methanol synthe-
free-stream and unsteady drag at finite Reynolds number, J. Fluid sis in gas–liquid slurry bubble column reactors, Chem. Eng. Sci. 55
Mech. 237 Ž1992. 323. Ž2000. 573.
w25x D. Mitra-Majumdar, B. Farouk, Y.T. Shah, Hydrodynamic modeling w41x B.H. Xu, A.B. Yu, Authors reply to the comments of B.P.B.
of three-phase flows through a vertical column, Chem. Eng. Sci. 52 Hoomans, J.A.M. Kuipers, W.J. Briels, and W.P.M van Swaaij,
Ž1997. 4485. Chem. Eng. Sci. 53 Ž1998. 2646.
w26x W.F. Noh, P.R. Woodward, SLIC ŽSimple Line Interface Method., w42x G.Q. Yang, X. Luo, R. Lau, L.-S. Fan, Bubble formation in high
in: A.I. van de Vooren, P.J. Zandbergen ŽEds.., Lecture Notes in pressure liquid–solid suspensions with plenum pressure fluctuation,
Physics, vol. 59, 1976, p. 330. AIChE J. 46 Ž2000. 2162.
w27x W.V. Pinczewski, The formation and growth of bubbles at a sub- w43x D.H. Yoo, K. Terasaka, H. Tsuge, Behavior of bubble formation at
merged orifice, Chem. Eng. Sci. 36 Ž1981. 405. elevated pressure, J. Chem. Eng. Jpn. 31 Ž1998. 76.
w28x P.N. Rowe, G.A. Henwood, Drag forces in a hydraulic model of a w44x D.L. Youngs, Time-dependent multi-material flow with large fluid
fluidized bed—Part I, Trans. Inst. Chem. Eng. 39 Ž1961. 43. distortion, in: K.W. Morton, M.J. Baines ŽEds.., Numerical Methods
w29x A. Sokolichin, G. Eigenberger, Gas–liquid flow in bubble columns for Fluid Dynamics, Academic Press, New York, 1982, p. 273.
and loop reactor: Part I. Detailed modeling and numerical simula- w45x J. Zhang, Y. Li, L.-S. Fan, Numerical studies of bubble and particle
tion, Chem. Eng. Sci. 49 Ž1994. 5735. dynamics in a three-phase fluidized bed at elevated pressures, Pow-
w30x K. Terasaka, H. Tsuge, Bubble formation under constant-flow condi- der Technol. 112 Ž2000. 46.
tions, Chem. Eng. Sci. 48 Ž1993. 3417. w46x J. Zhang, L.S. Fan, C. Zhu, R. Pfeffer, D. Qi, Dynamic behavior of
w31x R. Torvik, H.F. Svendsen, Modeling of slurry reactors, a fundamen- collinear collision of elastic spheres in viscous fluids, Powder Tech-
tal approach, Chem. Eng. Sci. 45 Ž1990. 2325. nol. 106 Ž1999. 98.

You might also like