1-s2.0-s0009250912006677-main
TRANSCRIPT
![Page 1: 1-s2.0-S0009250912006677-main](https://reader030.vdocument.in/reader030/viewer/2022021223/577ce0461a28ab9e78b2fad9/html5/thumbnails/1.jpg)
7/29/2019 1-s2.0-S0009250912006677-main
http://slidepdf.com/reader/full/1-s20-s0009250912006677-main 1/12
![Page 2: 1-s2.0-S0009250912006677-main](https://reader030.vdocument.in/reader030/viewer/2022021223/577ce0461a28ab9e78b2fad9/html5/thumbnails/2.jpg)
7/29/2019 1-s2.0-S0009250912006677-main
http://slidepdf.com/reader/full/1-s20-s0009250912006677-main 2/12
these multiphase processes are widely applied detailed under-
standing of the relative motion between the particles and the
fluid and their corresponding influence on mass, heat and
momentum transfer [for instance, Joshi et al., 1980, 1981, 2003;
Thakre and Joshi, 1999; Murthy et al., 2007] has been limited for
two reasons. First, non-intrusive experimental observations are
very difficult to undertake, especially of the liquid motion in the
interstices between the particles, droplet or bubbles. Second, the
computational requirements for direct numerical modelling of such a complex flow have been prohibitively large. Recent
advances in both flow visualization techniques and computing
capability have reduced these limitations to such an extent that
detailed investigation of multiphase systems that more closely
reflect actual industrial practices is now possible.
In this study, particle image velocimetry (PIV) has been
utilized to quantify both the particle slip and interstitial liquid
velocity within a solid–liquid fluidized bed in the Reynolds
number range of 51–759. The experimental measurements are
complemented by in-house direct numerical simulations (DNS) to
quantify the influence of the presence of neighbouring particles
on the wake structure and settling velocity of freely falling
particles. The DNS is carried out using the in-house code reported
in Jin et al. (2009), whereby the hydrodynamic force between the
particle and the fluid is resolved without the need for assuming a
given drag law relationship.
1.1. Previous work
Numerous experimental and theoretical studies (e.g.
Richardson and Zaki, 1954; Hanratty and Bandukwala, 1957;
Garside and A1-Dibouni, 1977; Joshi, 1983; Pandit and Joshi,
1998; Gevrin et al., 2008; Reddy and Joshi., 2009; Zivkovic et al.,
2009) have investigated bed expansion characteristics of
solid–liquid fluidized beds; and whilst these studies are useful
they give no real insight into the flow behaviour in the interstices
between the particles. There have been a number of studies that
have attempted to address this issue. Handely et al. (1966)
undertook photographic measurements in beds of 3 mm glassbeads at ReN of 45, 87 and 182 in a 75 mm diameter glass column
and found that the root mean square fluctuating velocity in the
axial direction is about 2.5 times higher than that in the radial
direction. Chen and Kadambi (1990) studied solid–liquid slurry
flows in a horizontal pipe using Laser Doppler Velocimetry (LDV)
and refractive index matching of the solid and liquid phases. They
used silica-gel particles having an average size of 40 mm and 50%
W/W sodium iodide aqueous solution. They were able to measure
the average axial liquid velocity for silica-gel concentrations
between 5 and 50% W/W. Neither Handely et al. (1966) nor
Chen and Kadambi (1990) reported turbulence quantities.
Chen and Fan (1992) applied PIV to three phase gas–liquid–
solid fluidized beds to obtain 49 velocity vectors that provided an
excellent basis for quantifying the mean flow. However, thenumber was well below the $100,000 required for reliable
estimation of turbulence intensity, turbulent kinetic energy,
turbulent energy dissipation rate and structure functions. Haam
et al. (2000) combined PIV measurements with refractive index
matching to obtain velocity information inside solid–liquid flui-
dized beds at solids concentrations much higher than was pre-
viously possible. They measured the axial and radial velocity
components in a fluidized bed of glass beads for ReN values of
1040 and 1550. They found that the axial turbulent intensity of
the fluid increased by up to 70% due to the presence of the
glass beads.
Dijkhuizen et al. (2007) coupled their PIV measurements with
simultaneous measurement of the instantaneous velocity and
granular temperature fields. The PIV algorithm was specifically
optimized for dense granular systems and applied to a fluidized
bed at incipient fluidization conditions into which both single-
and multi-bubbles were injected. They observed that the highest
granular temperature was in the vicinity of the bubble(s), and for
the case of 1.5 mm glass particle bed the granular temperature
(Gidaspow, 1994) was in the range of 0.022–0.069 m2 sÀ2.
Shi (2007) applied PIV to investigate the particle motion and
cluster properties in a gas–solid two-phase flow in a circulating
fluidized bed riser. Visual images and micro-structure of variousclusters were captured. After the boundary of clusters was
determined by the gray level threshold method, clusters were
classified by the distance between particles and the shape and
position of clusters. In addition, the process of cluster formation
and breakup was described, and the sizes of clusters were also
obtained. With the Minimum Quadric Difference cross-correlation
algorithm suitable for high-density particles, the axial velocities
of the particles were obtained in the dilute phase section. Analysis
of the magnitude and distribution of particle axial velocity in the
radial direction showed at most radial cross-sections a parabolic
profile in the upward direction. The magnitude of axial velocity in
the core region was found to be higher than that in the near wall
region of the riser.
Muller et al. (2009) employed simultaneous PIV and Planar
Laser-Induced Fluorescence (PLIF) measuring techniques to
investigate the eruption of both a single and continuous stream
of bubbles in the freeboard region of a fluidized bed. The observed
bubble eruption patterns were in general agreement with the
bubble models published in the literature. Based on the calculated
vorticity of the gas in the freeboard it was found that the bubble
induced turbulence decays rapidly. Stereoscopic PIV measure-
ments of the out-of-plane component of the liquid velocity were
found to be not negligible.
Kashyap et al. (2011) used PIV to obtain laminar and turbulent
properties near the wall in the developing region of circulation of
Geldart B type particles in the riser part of circulating gas–solid
fluidized bed. Instantaneous velocities for the solid phase were
measured simultaneously in the axial and radial directions using
a CCD camera and a coloured rotating transparency. A novelmethod was used to determine axial and radial solid phase
dispersion coefficients using the autocorrelation technique.
The measured laminar and Reynolds stresses, laminar and turbu-
lent granular temperatures, laminar and turbulent dispersion
coefficients and energy spectra all exhibited anisotropy. The total
granular temperatures were in reasonable agreement with the
literature values. However, the axial and radial solid dispersion
coefficients measured near the wall were slightly lower than the
radially averaged values in the literature.
Hernandez-Jimenez et al. (2011) investigated both experimen-
tally and computationally the hydrodynamics of a rectangular,
bubbling air-fluidized bed of 5 mm thickness. The authors used
PIV and Digital Image Analysis (DIA) to obtain quantitative
information on bubble hydrodynamics, dense-phase probability,and time-averaged vertical and horizontal components of the
dense-phase velocity as a function of gas flow rate through the
bed. Simulations were also conducted using an Euler–Euler two-
fluid approach based on two different closure models for the gas–
particle interaction, namely the drag models of (1) Gidaspow
(1994) and (2) Syamlal and O’Brien (1989). Good agreement
between experimental observation and simulation results was
obtained for dense-phase probability, gas bubble diameter rise
velocity. The vertical component of the simulated dense-phase
velocity, however, was found to be nearly an order of magnitude
larger than that obtained from the PIV experiments.
Van Buijtenen et al. (2011) applied PIV, Positron Emission
Particle Tracking (PEPT), and Electrical Capacitance Tomography
(ECT) measuring techniques to quantify the flow in both quasi-2D
R.K. Reddy et al. / Chemical Engineering Science 92 (2013) 1–122
![Page 3: 1-s2.0-S0009250912006677-main](https://reader030.vdocument.in/reader030/viewer/2022021223/577ce0461a28ab9e78b2fad9/html5/thumbnails/3.jpg)
7/29/2019 1-s2.0-S0009250912006677-main
http://slidepdf.com/reader/full/1-s20-s0009250912006677-main 3/12
and 3D spouted fluidized beds. In the pseudo-2D bed the
measurements were able to highlight the appearance of dead
zones in the annulus near the bottom of the bed in the spout-
fluidization regime. Measurements were also made in the jet-in-
fluidized-bed regime beds with spout heights of 0, 20, and 40 mm.
The results from both experimental systems were compared with
full 3D discrete particle modelling (DPM) simulations; with
generally good agreement being achieved except for slight over-
prediction of the velocities in the annular region for the quasi-2Dcase where wall effects in the experimental system are possibly
higher than assumed in the modelling.
In addition to the experimental investigations there have been
many studies involving simulations of multi-particle systems. For
a recent review of the topic, with particular focus on DNS, see
Reddy et al. (2010b). Some more recent studies include:
Deen et al. (2009) have reported DNS results for multi-fluid
flows in which simultaneously moving deformable (drops or
bubbles) and non-deformable (particles) elements are present,
possibly with the additional presence of free surfaces. They utilized
a front tracking (FT) model to circumvent the explicit computation
of the interface curvature. They also used an immersed boundary
(IB) model to incorporate both particle–fluid and particle–particle
interaction via a direct forcing method and a hard sphere Discrete
Particle approach. The capabilities of the hybrid FT-IB model are
demonstrated by the authors with a number of examples in which
complex topological changes in the interface are encountered.
Uhlmann and Pinelli (2009) conducted a DNS study of dilute
turbulent particulate flow in a vertical plane channel, considering
up to 8192 finite-size rigid particles with numerically resolved
phase interfaces. The particle diameter corresponded to approxi-
mately nine wall units, with a terminal Reynolds number of 136.
The bulk fluid upflow was maintained at a Re¼2700 to maintain
solids suspension. Two different density ratios were simulated,
which varied by a factor of 4.5. The corresponding Stokes
numbers for the two particles were O(10) in the near-wall region
and O(1) in the outer flow. The DNS simulations indicated that the
mean flow velocity profile tended towards a concave shape, and
anisotropy for both the turbulence intensity and normal stresses.Large-scale elongated streak-like structures were predicted by the
DNS, with stream-wise dimensions of the order of eight channel
half-widths and cross-stream dimensions of the order of one
channel half-width. There was no evidence for the formation of
particle clusters, which suggested that the large structures were
due to an intrinsic instability of the flow that was triggered by the
presence of the particles.
Xu and Subramaniam (2010) have reported DNS results for
turbulent flow past uniform and clustered configurations of fixed
particle assemblies at the same solid volume fraction. Their
approach was based on a discrete-time, direct-forcing immersed
boundary method that imposes no-slip and no-penetration
boundary conditions on each particle surface. Results are reported
for mean flow Reynolds number of 50 where the ratio of theparticle diameter to the Kolmogorov scale was 5.5. The DNS
confirmed experimental observations that the fluid-phase turbu-
lent kinetic energy was enhanced by clustered configurations
leading to increased levels of anisotropic turbulence.
Tenneti et al. (2011) have reported particle-resolved DNS
results of interphase momentum transfer in flow past fixed
random assemblies of mono-disperse spheres with finite fluid
inertia using a continuum Navier–Stokes solver. This solver is
based on a new formulation called the Particle-resolved Unconta-
minated-fluid Reconcilable Immersed Boundary Method (PURe-
IBM). The principal advantage of this formulation is that the fluid
stress at the particle surface is calculated directly from the flow
solution (velocity and pressure fields), which when integrated over
the surfaces of all particles yields the average fluid–particle force.
The PUReIBM is a consistent numerical method to study gas–solid
flow because it results in a force density on particle surfaces which
is reconcilable with the averaged two-fluid theory. The numerical
convergence and the accuracy of PUReIBM approach were estab-
lished through a comprehensive suite of validation tests. The
normalized average fluid–particle force, F , was obtained as a
function of solid volume and mean flow Reynolds number for
random assemblies of mono-disperse spheres. From the simula-
tions, a simple correlation for F , and hence drag coefficient, wasdeveloped for the particle Re range 100–300. Given that the
simulations were based on a fixed particle assembly, any effects
of mobility of the particles was not included. However, approach is
a good approximation for high Stokes number particles, which are
encountered in most gas–solid flows.
1.2. Research priorities
The great majority of industrial systems are operated under
turbulent conditions, wherein a compendium of eddies (turbulent
structures) of different length and time scales govern the momen-
tum, heat and mass transfer and mixing behaviour. For this reason
it is desirable to more fully understand the formation and
dynamics of these turbulent structures. All the above-mentioned
studies on PIV measurement and DNS computations have provided
an excellent foundation, and with recent advances in experimental
and mathematical techniques there is an opportunity to expand
that knowledge for multi-particle systems. The following list of 13
research priorities is aimed at coupling both experimental and
computational investigations. The priorities include:
1. DNS simulation for creeping flow (ReNo1) around a single
particle in both an infinite liquid and a solid–liquid fluidized
bed. Estimation of skin, form and total friction at all locations
on the surface of the particle. Estimation of total drag
coefficient at different solid volume fractions in the fluidized
bed to quantify the hindrance effect.
2. DNS simulation of flow around a particle in the ReN range of
1–103 to obtain 3D component instantaneous velocities attemporal resolutions up to the Kolmogorov scale. Simulation
of boundary layer separation and also the size, shape and
stability of wakes behind a particle with respect to ReN.
Estimation of skin and form drag on the entire surface of the
particle.
3. DNS simulation of flow around a single particle in the ReNrange of 103–106, capturing the sudden drop in drag coeffi-
cient, C D, at ReN of around 2 Â 105 due to the transition from
laminar to turbulent boundary layer separation.
4. DNS simulation of settling of a particle in a Hele-Shaw cell to
include the wall effect.
5. DNS simulation of flow around two neighbouring particles.
6. DNS simulation of a solid–liquid fluidized bed in the ReN
range of 1–106 to obtain 3D component instantaneous velo-cities at temporal resolutions up to a small scale permissible
by the computational power. Estimation of skin, form and
total friction at all locations on particles within the fluidized
bed. Simulation of boundary layer separation in multi-
particle systems and also understanding of size, shape and
stability of wakes with respect to ReN and solids volume
fraction.
7. DNS simulation of solid–liquid fluidized bed where the momen-
tum transfer is accompanied by mass/heat transfer at the wall.
Quantification of heat and mass transfer coefficients as a function
of ReN and ALutilizing u0u0, u0c 0 and u0T 0 simulated results.
8. Quantification of homogeneity, isotropy and energy spectrum
in solid–liquid dispersions as a function of ReN, AL and
particle diameter and shape.
R.K. Reddy et al. / Chemical Engineering Science 92 (2013) 1–12 3
![Page 4: 1-s2.0-S0009250912006677-main](https://reader030.vdocument.in/reader030/viewer/2022021223/577ce0461a28ab9e78b2fad9/html5/thumbnails/4.jpg)
7/29/2019 1-s2.0-S0009250912006677-main
http://slidepdf.com/reader/full/1-s20-s0009250912006677-main 4/12
9. Development of scaling laws for the inertial, dissipation and
large scale sub-range in solid–liquid dispersions.
10. DNS simulation for estimation of interface forces, such as
drag, lift, virtual mass, and Basset; and subsequent applica-
tion in the closure formulation for RANS and LES simulations
of solid–liquid dispersions.
11. Measurement of instantaneous velocity within the fluidized
bed using techniques such as high speed PIV, at sampling
rates of up to 10 kHz, in combination with refractive indexmatching of the liquid and solid phases. Application of
mathematical techniques, such as discrete wavelet trans-
forms (DWT), continuous wavelet transforms (CWT) and
proper orthogonal decomposition (POD), to identify and
characterize the flow structures in terms of shape, size,
velocity and energy distributions.
12. To understand the mechanism of vorticity generation and
hence the origin of turbulence.
13. Relate momentum, heat and mass transfer to turbulent
structure dynamics.
Priorities 1 and 2 for the authors have already been reported in
Reddy et al. (2010a, 2010b). This study is focused on addressing
priorities 3, 6 and 11. In terms of DNS (Priorities 3 and 6),simulating up to 245 particles (as performed previously) with
Reynolds number range extended to 200. In terms of experimen-
tal development (Priority 11), standardization of PIV measure-
ment in terms of matching of refractive indices of the solid
particles and liquid; achieving sampling frequencies of 2 Hz over
a wide range of Reynolds numbers; and data processing in terms
of mean and turbulence characteristics which can be directly
compared to DNS results.
2. Experimental
2.1. Solid–liquid fluidized bed apparatus
The schematic of experimental apparatus is shown in Fig. 1. It
consisted of an acrylic circular column (1), with an inner diameter
of 50 mm and a height of 300 mm. A 0.5 HP centrifugal pump
(6) was used for pumping the refractive index-matched liquid.
A calming section packed (4) with 3–4 mm glass beads of 0.3 m
height was provided to homogenize the liquid flow before it
reached the liquid distributor. The distributor was a perforated
plate (3) containing 128 holes of 2 mm diameter on a triangular
pitch of 3.1 mm. The circular test section of the fluidized bed was
encased with a square tank (2) which was filled with a refractive
index-matched liquid to allow for Particle Image Velocimetry
(PIV) measurements within the bed.
2.2. Solid–liquid refractive index matching
PIV measurement within opaque fluidized beds is only possible
at solids concentrations of only a few percent by volume. However,
if the refractive index of the particle is matched1 to that of the fluid
then it is possible to take reliable measurements at solids con-
centrations greater than 50% by volume. In this study, different
mixtures of organic liquid and inorganic salts were screened for
matching the refractive index of 1.47 of high precision borosilicate
glass beads (d p¼3 mm and r p¼2230 kg mÀ3). As indicated in
Table 1, from the screening process it was found that Solution 12
(68% turpentine and 32% tetra-hydronaphthalene) provided the
best match for the RI of the borosilicate glass beads as well as being
compatible with the acrylic column. Paraffin oil, with very similar
RI to that of the solution and glass beads, was also added in small
quantities to change the viscosity of the solution and still maintain
the required RI.
2.3. PIV measurements
A schematic of the TSI PIV set-up is shown in Fig. 2. The laser
source was provided by a pulsed Nd:YAG laser that had a pulse
duration of 6 ns and was synchronized with the camera. The time
difference between the two laser pulses was optimized based on
Nyquist criteria. The optics included a combination of cylindrical
and spherical lenses that produced a thin laser sheet of 0.1 mmthickness. Images were captured at a frequency of 2 Hz using a
high-resolution 4 M CCD camera (2048Â 2048 pixels) placed
perpendicular to the laser sheet. The refractive index matched
liquid was seeded with silver-coated hollow glass particles with a
mean diameter equal to 20 mm. The interrogation area was set at
50 Â 100 mm (64 Â 64 pixels, with a 50% overlap) resulting in
approximately 1500 vectors per image.
Post-processing of the captured raw PIV images was under-
taken to determine the velocity vectors. Out-of-plane motion of
the seeding particles and strong local velocity gradients caused
some spurious velocity vectors. Median filtering, with a threshold
value 1.5 times the median of surrounding vectors, was applied to
filter the high spurious vectors. A signal-to-noise ratio of 4 was
applied to filter the low spurious vectors. Parameters like time
Outlet
50
100
300
InletRefractive index
matching liquid
PUMP
Acrylic column
Outer square column
Distributor
Gasket
VentCalming section
Flange
1
Heat
Exchanger
8
2
3
4
5
7
6
9
Fig. 1. Solid–liquid fluidized bed experimental setup (all dimensions are in mm)
[(1) acrylic column; (2) outer square column; (3) distributor; (4) calming section;
(5) flange; (6) pump; (7) refractive index matching liquid; (8) heat Exchanger;
(9) gasket].
1 An excellent review on the subject of RI matching has been written by
Wiederseiner et al. (2011).
R.K. Reddy et al. / Chemical Engineering Science 92 (2013) 1–124
![Page 5: 1-s2.0-S0009250912006677-main](https://reader030.vdocument.in/reader030/viewer/2022021223/577ce0461a28ab9e78b2fad9/html5/thumbnails/5.jpg)
7/29/2019 1-s2.0-S0009250912006677-main
http://slidepdf.com/reader/full/1-s20-s0009250912006677-main 5/12
difference between laser pulses, light sheet thickness and seeding
density were optimized so that spurious vectors remained below
2%. The PIV measurements were performed in darkness to
minimize light contamination from external sources. Additional
information2 pertaining to the PIV measurements can be found in
Deshpande et al. (2009, 2010) and Sathe et al. (2010).
3. Direct numerical simulation
3.1. Simulation set-up
The physical properties of particles and liquid used to simulate
different particle Reynolds numbers within a fluidized bed are the
same as those used in Reddy et al. (2010b) and are listed in
Table 3. Direct numerical modelling was undertaken for 1, 9, 27,
100, 180 and 245 particles contained. For each system, corre-
sponding to a desired particle hold-up, the particles were initially
arranged in a regular lattice. At time, t ¼0, the particles were
allowed to settle through a rectangular domain with width and
depth of 10 and 20 particle diameters, respectively. A moving
reference frame method was employed, resulting in the rectan-gular domain to move downward with a velocity equal to the
average velocity of particle ensemble. At steady state, this situa-
tion is comparable to a fluidized bed where the heavier particles
are suspended by the upward velocity of the liquid. The compar-
ison between the particle and liquid velocity reference conditions
is shown in Fig. 3. For the fluidized bed (A), the liquid rises
upwards with particle settling velocity; whilst for the DNS
domain (B) the volume averaged liquid velocity is zero and the
particle is moving down with its settling velocity. The domain is
sliding downward with the same velocity as the particle. In both
cases, the relative velocity of particles with respect to the walls of
the domain is zero, and the relative velocity of the liquid with
respect to the domain walls is equal to settling velocity of the
particles, in the upward direction. The moving reference framehas two advantages for this application. First, very long settling
times can be simulated by using a domain with finite dimensions
making the simulations computationally affordable; and second,
it is a relatively straightforward to establish a fluidized bed with a
given particle hold-up. The initial conditions for the DNS are
summarized in Table 4. The side faces of the box have been
treated as wall boundary condition. The top face of the box was
treated as a pressure outlet, whilst the bottom face was treated as
the velocity inlet with zero velocity magnitude. The initial particle
positions are shown in Fig. 4.
Table 1
Screening of refractive index matching liquid for PIV experiments.
ID Liquid RI25 1C Remax Remarks
1 90% W/W glycerinþ10% W/W water 1.45 2 RI not exactly matched
2 Sodium idiode solution (55% W/W) 1.475 210 Highly corrosive
3 Paraffin oil light 1.465 23 For low ReN4 Potassium thiocyanate solution (42% W/W) 1.46 200 Harmful and corrosive
5 Ammonium thiocyanate solution (45% W/W) 1.47 200 Harmful and corrosive
6 P-cyamene 1.465 545 Specialty chemical7 Turpentine 1.44 850 RI not matched
8 Be nzene o r methyl benzoate with turpentine 1.47 900 Not compatib le with the acryl ic column (col umn was c racked)
9 30% W/W naphthale neþ70% W/W turpentine 1.46 5 750 Corrosive and not stable , turned in to pale yell ow
10 Turpentineþchloro-naphthalene 1.465 850 Chlorine compounds are not compatible with the acrylic column
11 Turpentineþbenzyl alcohol 1.47 850 Not compatible with the acrylic column (column was cracked)
12 68% Turpentineþ3 2% Tet ra hy dro n ap ht halene 1 .4 67 7 60 RI has b een matched and com patible wit h t he acry lic column
Table 2
PIV experimental conditions.
PIV exp
no.
Paraffin
light oil %
W/Wa
mL
(Pa s)
rL
(kg mÀ3)
RI (–) V S N(m sÀ1)
eL range
(–)
ReN(–)
1 85 0.01010 855 1.462 0.199 0.52–0.71 51
2 63 0.00420 870 1.471 0.266 0.46–0.75 164
3 48 0.00280 880 1.468 0.298 0.56–0.72 276
4 34 0.00193 890 1.469 0.309 0.53–0.73 437
5 18 0.00165 900 1.466 0.317 0.51–0.73 525
6 6.5 0.00142 910 1.470 0.325 0.46–0.66 625
7 0.0 0.00120 915 1.470 0.333 0.49–0.75 759
a Added to liquid 12 listed in Table 1.
Synchronizer
Laser
Solid-liquidfluidized bed
PumpRefractive
indexmatching
liquid
Camera
Fig. 2. Schematic diagram of PIV setup.
2 For a more general discussion on the topic see Tokuhiro et al. (1998), Deen
et al. (1999), Lindken and Merzkirch (1999, 2002) and Broder and Sommerfeld
(2003).
R.K. Reddy et al. / Chemical Engineering Science 92 (2013) 1–12 5
![Page 6: 1-s2.0-S0009250912006677-main](https://reader030.vdocument.in/reader030/viewer/2022021223/577ce0461a28ab9e78b2fad9/html5/thumbnails/6.jpg)
7/29/2019 1-s2.0-S0009250912006677-main
http://slidepdf.com/reader/full/1-s20-s0009250912006677-main 6/12
3.2. Fictitious domain formulation
The DNS simulations were carried out using the fictitious
domain method described previously (see Diaz-Goano et al.,
2003; Veeramani et al., 2007). Briefly, the fluid and particles
occupy domains O1 and O2, respectively, and the relevant
equations of continuity and motion are applied to each domain
as appropriate. In the fictitious domain approach the equations
for the fluid are extended into the particle domain, such that
O¼(O1UO2), resulting in savings in computational time as the
liquid domain no longer needs to be re-meshed each time there is
particle motion. The relevant equations3 are
Du1
Dt
¼ Àr p1 þ 1Re r 2u1 þ
r2,iÀr1
r1G À F ð Þ, r Uu1 ¼ 0 inO ð1Þ
and
DU 1Dt
¼1
V i
Z O2,i
FdO ð2Þ
where
G ¼1Fr e g inO2,i, i ¼ 1,:::,n
0 inO1
(ð3Þ
and
F ¼1Fr e g þ
r1r2,iÀr1
^F inO2,i, i ¼ 1,:::,n
0 inO1
(ð4Þ
and
^F ¼À Du1
Dt þ 1Re r 2u1Àr p1 inO2,i, i ¼ 1,:::,n
0 inO1
(ð5Þ
where F is the interaction force. For collision of a particle with the
plane wall, the interaction force can be replaced by the lubrica-
tion force given by tenCate et al. (2002):
^ F W
i ¼À6pr iU ?m1
r i^h
À r ih
if ^hoh
0 otherwise
(ð6Þ
An operator splitting procedure is applied to the discretization
process, with a rigid body constraint being applied to the particles
in the last time step. Finally, the solver is parallelized using the
PETSc libraries ( Jin et al., 2008, 2009) on a shared memory system.
4. Results and discussion
4.1. PIV measurements
Seven experiments were performed using the refractive index
matched solid–liquid fluidized bed. A summary of the experi-
mental conditions is given in Table 2. Two sample raw images at
ReN of 625 and superficial liquid velocities of 0.046 and 0.116 m/s
are shown in Fig. 5. The post-processing of instantaneous velocityflow field at ReN of 625 and superficial liquid velocity of
0.046 m sÀ1 is shown in Fig. 6. The radial variation of the average
axial velocity, vL, at superficial liquid velocities of 0.046 and
0.116 m sÀ1is shown in Fig. 7. It can be observed that these
velocity profiles are almost flat in the radial direction.
It can be observed that at ReN of 625 and at superficial liquid
velocity, V L, of 0.047 m sÀ1 the axial turbulent intensity was on
average 69% and the radial turbulent intensity was around 42%
(Fig. 8A). An increase in the V L from 0.047 to 0.116 m sÀ1, results in
a decrease in the solid volume fraction of the bed from 0.54 to 0.355.
Therefore the turbulent intensity (ratio of root mean square velocity
to the average velocity) also decreases to about 47% in the axial
direction and around 28% in radial direction (Fig. 8B).
It has been observed that at ReN¼437 and at V L
of 0.047 m sÀ1
the axial turbulent intensity is 59% and radial turbulent intensity is
around 36%. In order to examine the effect of Reynolds number on
the turbulent intensity, the Reynolds number has been varied over
the range 51–759. The variation of average axial and radial
turbulent intensities with respect to Reynolds number is shown in
Fig. 9. At the lowest Reynolds number of 51 it was observed that the
axial turbulent intensity is only 7% whilst the radial turbulent
intensity is around 2%. When the Reynolds number is increased
from 51 to 164 it was observed that the axial and radial turbulent
intensities increase to 19% and 8%, respectively. Further increase in
Reynolds number from 164 to 759 led to an axial turbulent intensity
of 75% and a corresponding radial turbulent intensity of 47%. In the
turbulent regime, at ReN4500 (Clift et al., 1978; Joshi, 1983), it was
found that the axial turbulent intensity is 1.65 times that of the
radial turbulent intensity. This value is lower than the value of 2.5 reported by Handely et al. (1966). The image capture and
processing technique employed here, utilizing a laser light sheet
to clearly illuminate the particle boundary, is arguably more precise
than the photographic measurement of Handely et al. (1966). Based
on the limited number of experimental measurements, the follow-
ing simple relationships are proposed for the relationships between
the turbulence intensity in the axial, v0, and radial, u0, directions and
the solid phase holdup, AS , and slip velocity, V S :
v0
V S ¼ 1:29AS ð7Þ
u0
V S ¼ 0:78AS ð8Þ
Table 3
DNS physical properties.
ReN (–) 1 5 10 30 65 100 200
dP (mm) 15 15 15 15 15 15 15
rp (kg mÀ3) 1120 1120 1120 1120 1120 1120 1120
rL (kg mÀ3) 900 900 900 900 900 900 900
mL (Pa s) 0.5700 0.2210 0.1430 0.0680 0.0390 0.0282 0.0166
VS
VL=0
H
Liquid stationary,
Control volume
moving down
with velocity VS
VS=
VL
H
VL
Control volume
stationary,
Liquid moving
up with velocity
VL
Fig. 3. Schematic of fluidized bed and moving reference frame DNS simulation
framework.
3
See Veeramani et al. (2007) for definition of terms.
R.K. Reddy et al. / Chemical Engineering Science 92 (2013) 1–126
![Page 7: 1-s2.0-S0009250912006677-main](https://reader030.vdocument.in/reader030/viewer/2022021223/577ce0461a28ab9e78b2fad9/html5/thumbnails/7.jpg)
7/29/2019 1-s2.0-S0009250912006677-main
http://slidepdf.com/reader/full/1-s20-s0009250912006677-main 7/12
The radial variation of turbulent kinetic energy at different
Reynolds numbers is shown in Fig. 10. It can be seen that
as Reynolds number increases from 51 to 759, the turbulent
kinetic energy also increases from 2 Â 10À5 to 4 Â 10À3 m2 sÀ2.
Moreover, for each Reynolds number there is practically no
variation in the turbulent kinetic energy profile along the radial
direction, which is similar to that for the turbulent inten-
sity profiles. Both of these observations are further validations of
the homogenous nature of turbulence in the solid–liquid
fluidized bed.
Fig. 4. Initial particle positions for (A) 1, (B) 9, (C) 27, (D) 100, (E) 245 particles [see Fig. 14A for orientation for 180 particles].
Fig. 5. Raw images at ReN¼ 625 for V L¼(A) 0.046 and (B) 0.116 m sÀ1.
Table 4
DNS initial conditions.
Number of particles 1 9 27 100 180 245
dP (mm) 15 15 15 15 15 15
r p (kg mÀ3) 1120 1120 1120 1120 1120 1120
Ln (–) —— 1.2 1.2 1.2 1.2 1.2
Comp. domain dimension (mm) Â (mm) Â (mm) 8 Â 16 Â 8 8 Â 16 Â 8 8 Â 16 Â 8 8 Â 16 Â 8 8 Â 16 Â 8 8 Â 16 Â 8
Number of nodes (–) 1.2Â 106 1.9 Â 106 2.6 Â 106 4.1 Â 106 5.2 Â 106 6.5 Â 106
R.K. Reddy et al. / Chemical Engineering Science 92 (2013) 1–12 7
![Page 8: 1-s2.0-S0009250912006677-main](https://reader030.vdocument.in/reader030/viewer/2022021223/577ce0461a28ab9e78b2fad9/html5/thumbnails/8.jpg)
7/29/2019 1-s2.0-S0009250912006677-main
http://slidepdf.com/reader/full/1-s20-s0009250912006677-main 8/12
4.2. Direct numerical simulations
4.2.1. Wake structure
Fig. 11 shows the wake structure for a particle both in isolation
and with other eight particles present. It can be observed that at
the wake structure of a single particle at ReN¼200 in (A) an
infinite medium; and (B) surrounded by eight other particles at
centre-to-centre distance of 1.2 d p. It can be seen that whilst the
separation angle was similar for both particles the wake length
was reduced by almost 33% due to the presence of other particles.
For a single particle in an infinite medium the axisymmetric
toroidal vortex of the wake has been observed up to ReN¼210.
This same behaviour is exhibited for the isolated particle shown
in Fig. 11A, which is not surprising given that ReN¼200. For the
case of the nine particles this axisymmetry in the wake is broken
as shown in Fig. 11B. The reason for this early instability in the
wake is attributed to the asymmetric velocity gradients aroundthe particle, resulting in centrifugal acceleration in the core of the
toroidal vortex and an increase in the azimuthal velocity. The
axisymmetric toroidal vortex eventually breaks into two counter
rotating vortices. Additional DNS simulations were undertaken to
investigate the influence of particle spacing on the length of the
wake, and particularly to determine the minimum spacing
required so that the wake is no longer influenced by the
surrounding particles. The results (for ReN¼200) are shown in
Fig. 12 for dimensionless wake length, W *, as a function of
dimensionless particle spacing between particles, L*. It can be
seen for very close particle spacing then W *o1; and it steadily
increases to unity at L* equal to 6. Therefore, it can be concluded
that the effect of the surrounding particles can be neglected when
the centre-to-centre particle spacing is more than 6d p.
Axial velocity
(m s1)
Fig. 6. Instantaneous axial velocity flow field at ReN¼625 and V L¼ 0.047 m sÀ1.
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0 0.2 0.4 0.6 0.8 1
1
2
Fig. 7. Average axial velocity vs radial position at ReN¼625 and V L¼ (1) 0.047 and
(2) 0.116 m sÀ1.
0
10
20
30
40
50
60
70
80
0 0.2 0.4 0.6 0.8 1
0
10
20
30
40
50
60
0 0.2 0.4 0.6 0.8 1
Fig. 8. Turbulent intensities vs axial and radial position at ReN¼625 and V L¼
(1) 0.047 and (2) 0.116 m sÀ1.
R.K. Reddy et al. / Chemical Engineering Science 92 (2013) 1–128
![Page 9: 1-s2.0-S0009250912006677-main](https://reader030.vdocument.in/reader030/viewer/2022021223/577ce0461a28ab9e78b2fad9/html5/thumbnails/9.jpg)
7/29/2019 1-s2.0-S0009250912006677-main
http://slidepdf.com/reader/full/1-s20-s0009250912006677-main 9/12
4.2.2. Normalized settling velocity
The simulated normalized settling velocities (shown as solidlines) for 1, 2, 9, 27, 100, 180 and 245 particles are plotted as a
function particle Reynolds number in Fig. 13. It can be seen that
the time averaged settling velocity of the particle in the presence
of other particles decreases with an increase in the number of
particles surrounding it. This is in line with the hindrance effect
observed in particles settling in swarm. In the presence of eight
other particles, at ReN¼1, the normalized average settling velo-
city, V S , is 5.5% less than that for a single particle, V S N, settling in
an infinite liquid. For a system of 27 particles the average settling
velocity was found to be 0.606, which is 39.4% lower than V S N;
whilst for 100 particles V S is only 27% of V S N. The DNS results
show clearly that the particle settling velocity decreases with an
increasing number of particles, or solids volume fraction, AS ,
within the computational domain. Given that a decrease in
particle velocity is a consequence of an increase in the drag
coefficient, the simulation result is consistent with the findings of
Joshi (1983) and Pandit and Joshi (1998) who used an energybalance approach to derive the following relationship for the
particle drag coefficient, C D, as a function of AL:
C DC D1
¼ AÀ4:8L ð9Þ
In creeping flow the drag coefficient in an assemblage of
particle is increased due to (i) increased true fluid velocity within
the interstices between the particles, (ii) increased velocity
gradients resulting from more zero slip boundary condition
surfaces; and (iii) increased length of the fluid flow through the
assemblage of particles. In turbulent flow the same conditions
prevail, with enhanced momentum transfer, and hence higher
drag as the solids concentration is increased.
T U R B U L E N T I N T E N S I T Y ,
u i R
M S
/ ‹ u 2 › x 1 0 0 , ( - )
REYNOLDS NUMBER, Re∞ (-)
0
10
20
30
40
50
60
70
80
0 200 400 600 800
Fig. 9. Average (’) axial and radial (m) turbulent intensities vs ReN atV L¼ 0.047 m sÀ1.
NORMALIZED RADIAL DISTANCE, r/R, (-)
T U R B U L E N T K I
N E T I C
E N E R G Y , k , ( m 2 / s 2 )
0.00001
0.0001
0.001
0.01
0 0.2 0.4 0.6 0.8 1
Fig. 10. Turbulent kinetic energy radial profile vs ReN at V L¼ 0.047 m sÀ1
[ReN¼51(D), 164(J), 276(m), 437(’), 795()].
Fig. 11. Wake of the sphere at ReN¼200 [(A) settling in the infinite medium;(B) Settling in the presence of eight other surrounding particles].
D I M E N S I O N L E
S S W A K E L E N G T H ( - )
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1 2 3 4 5 6
Fig. 12. Wake length vs particle spacing for ReN¼200.
R.K. Reddy et al. / Chemical Engineering Science 92 (2013) 1–12 9
![Page 10: 1-s2.0-S0009250912006677-main](https://reader030.vdocument.in/reader030/viewer/2022021223/577ce0461a28ab9e78b2fad9/html5/thumbnails/10.jpg)
7/29/2019 1-s2.0-S0009250912006677-main
http://slidepdf.com/reader/full/1-s20-s0009250912006677-main 10/12
Also shown in Fig. 13, is the corresponding normalized settling
velocity predictions using the correlation of Richardson–Zaki
(1954) where the liquid void fraction, AL, used for the R–Z
calculation was that obtained from the DNS for the different
number of particles. It can be seen that the DNS simulations are in
good agreement (maximum deviation of 24%) with the R– Z
correlation. Interestingly, the R–Z correlation is based on a very
large number of experimental studies, typically for systems with a
homogeneous dispersion of a very large number of particles
(at least 2000). Whilst the current DNS simulations are for arelatively small number particles. It can be seen from Fig. 13 that
the agreement is increasing with increasing number of particles
where the influence of the wall on the wake of an individual
particle in the bulk assemblage is less. Ideally, more particles
should be included in the simulations. For this study, however, a
64 IBM Power4 processors having 256 GB RAM was required to
satisfactorily resolve the computational domain of 6.5 million
nodes for the 245 particles at ReN¼200. Further increase in either
the numbers of particles or Reynolds numbers will require much
bigger domains. To do this, the code will need to be fully
parallelized using domain decomposition technique in order to
perform simulations within a reasonable time frame.
4.2.3. Particle–particle and particle–wall collisions
The time sequence at 0, 3, 6, 9 and 12 s for the sedimentation
of 180 particles at ReN¼200 is shown in Fig. 14. The images show
that at different time intervals different particles come in contact
with each other, which is qualitatively in agreement with experi-
mental observations. For the simulations in this study a simple
lubrication model was used to account for both the particle–
particle and particle–wall collisions. The lubrication model is
really only valid for low Reynolds numbers, and at best an
approximation at higher Reynolds numbers. Moreover, the model
does not account for the experimentally observed clusters, espe-
cially at higher solids concentrations, where two or more particles
remain in contact with each other for some time. The
incorporation of a collision model that captures all of the
dynamics is an area of on-going DNS research.
4.2.4. Energy dissipation rate
The computed energy dissipation rate, e, was calculated from
the three components of instantaneous velocity and nine velocity
gradients at 12 points in the computational domain using the
expression:
e¼ 2@u1
@ x1
2
þ2@u2
@ x2
2
þ 2@u3
@ x3
2
þ@u1
@ x2
2
þ@u1
@ x3
2
þ@u2
@ x1
2
þ@u2
@ x3
2
þ@u3
@ x1
2
þ@u3
@ x2
2
þ2@u1
@ x2
@u2
@ x1
þ2
@u2
@ x3
@u3
@ x2
þ2
@u3
@ x1
@u01
@ x3
ð10Þ
The computed averaged energy dissipation rate for 245 parti-
cles at ReN¼51, with up to 3500 time steps was found to be
0.303 m2 sÀ3. This value compares favourably with the experi-
mental value of 0.36 m2 sÀ3 obtained by a volume-averaged
energy balance over the refractive index matched fluidized bed
used in this study4 at the same ReN¼51.
The local energy dissipation rate computed from the PIV mea-
surements at the point equivalent of 0.6, 8.7, 0.6 in the computa-tional domain is shown in Fig. 15 as a function of normalized time,
t *. Two peeks can be clearly seen at t *¼34.8 and 43.5 during the
measurement time. The first peak of approximately 4.0 m2 sÀ3
corresponded to the approach of a particle to the measurement
position, whilst the second peak of 5.5 m2 sÀ3 corresponded to the
departure of the particle. The period between these two peaks
corresponded to the time when the particle occupied the sampling
volume and hence a no liquid velocity vectors were measured and a
zero local energy dissipation rate was recorded. The peaks in the
local energy dissipation rate at the surface of the particle represent
an increase of between 10 and 18 times than that of the average
energy dissipation rate of 0.36 m2 s-3.
The flow patterns (including the energy dissipation rate) in the
vicinity of the particle interface and in the bulk have differentroles to play in terms of industrial operations. The flow pattern
near the particle surface governs the heat and mass transfer
characteristics at the particle–fluid interface, whereas the flow
pattern in the bulk governs the solid and liquid phase dispersions.
Typically, high values of particle–fluid heat and mass transfer
coefficients are desired along with low levels of particle and fluid
phase dispersion so that plug flow can be achieved. These two
characteristics are achieved by high and low levels of energy
dissipation rate, respectively. This behaviour is reflected in Fig. 15
as a particle transits through the sampling location, and for this
reason it is perhaps not surprising that fluidized beds are so
widely used for heat and mass transfer applications. However,
there are some limitations. First, the low dissipation rate in the
bulk results in low value of mass and heat transfer coefficients atthe container wall. Second, there may be some practical cases
where complete bulk mixing is desired in the bulk. For achieving
these process objectives an optimum selection of design (includ-
ing column diameter and height) and the operating parameters
(such as particle size and density, superficial liquid velocity, liquid
viscosity and density) is required.
5. Conclusions
Flow visualization experiments were performed using particle
image velocimetry and refractive index matching of the solid and
Fig. 13. Average velocity of the particles: 1, single particle; 2, nine particles; 3, 27
particles; 4, 100 particles; 5, 180 particles; 6, 245 particles, DNS (solid line),
————, Richardson and Zaki (1954) For 1oRe1o200 n ¼ 4:45þ18 d p=DÀ ÁÀ Á
ReÀ0:11
For 200oRe1o500 n ¼ ð4:45ÞReÀ0:11 .
4
PIV experiment 1 as listed in Table 2.
R.K. Reddy et al. / Chemical Engineering Science 92 (2013) 1–1210
![Page 11: 1-s2.0-S0009250912006677-main](https://reader030.vdocument.in/reader030/viewer/2022021223/577ce0461a28ab9e78b2fad9/html5/thumbnails/11.jpg)
7/29/2019 1-s2.0-S0009250912006677-main
http://slidepdf.com/reader/full/1-s20-s0009250912006677-main 11/12
liquid phases to understand the characteristics of turbulence in a
fluidized bed. For the Reynolds number in the range of 51–759
the experimental measurements revealed that the turbulence
intensity was constant in both the radial and axial directions,
thus establishing the homogenous nature system. Complemen-
tary DNS simulations provided increased spatial resolution of the
velocity field that could be obtained by the PIV measurements.
The computational domain, moving in the downward directionwith velocity equal to the averaged velocity of particles in the
swarm, comprised 1, 9, 27, 100, 180 and 245 particles. For each
case, the wake of individual particle was observed to attenuate
with increase in the volume fraction of particles. The averaged
particle slip velocity decreased with increase in the number of
surrounding particles, and compared reasonably with the
Richardson and Zaki (1954) correlation. Finally, the local energy
dissipation rate was computed from the DNS simulation, and forReN¼51, the energy dissipation rate near the surface of the
particle was found to be approximately 18 times the volume
averaged energy dissipation in the fluidized bed. Such a variation
in energy dissipation rate distribution would need to be taken
into consideration when designing fluidized beds given that heat
and mass transfer will be controlled by what is occurring at the
particle surface whilst mixing of the liquid and solid phases will
be controlled by the dissipation rate in the bulk.
Nomenclature
c 0 fluctuating concentration (kmol mÀ3)C DN drag coefficient of a single particle in infinite med-
ium (–)
C D drag coefficient of a single particle in the presence of
neighbouring particles (–)dP diameter of the particle (m)
e g unit vector in the direction of gravity
Fr Froude number (V2S1= gd p) (–)
F dimensionless redefinition of interaction force in the
most convenient form (–)
G representation of gravitational force in Eq. (6)^h distance between particle and wall or particle and
particle (m)
h grid size (m)L* dimensionless particle spacing between particles (–)
pL extension of the ^ pL to the entire domain O (–)ReN Reynolds number based on the particle (d p V sNrL / mL ) (–)
r i dimensionless radius of ith particle (–)
t time (s)
t * normalized time (–)U ? velocity component perpendicular to the wall or
particle (m sÀ1)
ui Velocity of fluid in ith direction (m sÀ1)
uL dimensionless fluid velocity (–)
uL extension of the uL to the entire domain O
u0 radial fluctuation velocity (m sÀ1)
v0 axial fluctuation velocity (m sÀ1)
V i volume of the ith particleVL superficial liquid velocity for fluidization (or) hin-
dered settling velocity of the particle in sedimenta-
tion (m sÀ1)
VS1 terminal settling velocity (m sÀ1)
Vs interstitial velocity (m sÀ1)
W wake length (m)
(W n¼W /d p) before the dimension (–)
Greek symbols
AL volume fraction of fluid
AS volume fraction of solid
Fig. 14. Sedimentation of 180 particles at different time intervals for ReN¼200 [(A) t ¼0; (B) t ¼3; (C) t ¼ 6; (D) t ¼9; (E) t ¼12].
Fig. 15. Local energy dissipation rate vs time for ReN¼51 [measurement taken at
the point (0.6, 8.7, 0.6)].
R.K. Reddy et al. / Chemical Engineering Science 92 (2013) 1–12 11
![Page 12: 1-s2.0-S0009250912006677-main](https://reader030.vdocument.in/reader030/viewer/2022021223/577ce0461a28ab9e78b2fad9/html5/thumbnails/12.jpg)
7/29/2019 1-s2.0-S0009250912006677-main
http://slidepdf.com/reader/full/1-s20-s0009250912006677-main 12/12
r p density of solid particle (kg mÀ3)rS ,i density of ith solid particle (kg mÀ3)
rL density of fluid (kg mÀ3)
OL fluid domain
O entire computational domainOS solid domain
mL molecular viscosity of fluid (kg mÀ1 sÀ1)
e energy dissipation rate (m2/sÀ3)
Subscripts
i particle speciesL liquid phase
S solid phase
References
Broder, D., Sommerfeld, M., 2003. Combined PIV/PTV-measurements for theanalysis of bubble interactions and coalescence in a turbulent flow. Can. J.Chem. Eng. 81, 756–763.
Chen, R.C., Kadambi, J.R., 1990. LDV measurements of solid–liquid flows.Particulate Sci. Technol. 8, 97–109.
Chen, R.C., Fan, L.S., 1992. Particle image velocimetry for characterizing the flowstructure in three-dimensional gas–liquid–solid fluidized beds. Chem. Eng. Sci.47, 3615–3622.
Clift, R., Grace, J.R., Weber, M.E., 1978. Bubbles, Drops and Particles. AcademicPress, New York.
Deen, N.G., Annaland, M.v.S., Kuipers, J.A.M., 2009. Direct numerical simulation of complex multi-fluid flows using a combined front tracking and immersedboundary method. Chem. Eng. Sci. 64, 2186–2201.
Deshpande, S.S., Sathe, M.J., Joshi, J.B., 2009. Evaluation of local turbulent energydissipation rate using PIV in jet loop reactor. Ind. Eng. Chem. Res. 48,5046–5057.
Deshpande, S.S., Tabib, M.V., Joshi, J.B., Ravi Kumar, V., Kulkarni, B.D., 2010.Analysis of flow structures and energy spectra in chemical process equipment.
J. Turbulence 11, 1–39.Diaz-Goano, C., Miniv, P.D., Nandakumar, K., 2003. A fictitious domain/finite
element method for particular flows. J. Comput. Phys. 192, 105–123.
Dijkhuizen, W., Bokkers, G.A., Deen, N.G., Van Sint Annaland, M., Kuipers, J.A.M.,2007. Extension of PIV for measuring granular temperature field in densefluidized beds. AIChE J. 53, 108–118.
Garside, J., A1-Dibouni, M.R., 1977. Velocity-voidage relationships for fluidizationand sedimentation in solid–liquid systems. Ind. Eng. Chem. Process Des. Dev.16, 206–214.
Gevrin, F., Masbernat, O., Simonin, O., 2008. Granular pressure and particlevelocity fluctuations prediction in liquid fluidized beds. Chem. Eng. Sci. 63,2450–2464.
Gidaspow, D., 1994. Multiphase Flow and Fluidization: Continuum and KineticTheory Descriptions. Academic Press, San Diego.
Haam, S.J., Brodkey, R.S., Fort, I., Klaboch, L., Placnik, M., Vanecek, V., 2000.Laser doppler anemometry measurements in an index of refraction matchedcolumn in the presence of dispersed beads. Int. J. Multiphase Flow 26,1401–1418.
Handely, D., Doraisamy, A., Butcher, K.L., Franklin, N.L., 1966. A study of the liquidand particle mechanics in liquid fluidized beds. Trans. Inst. Chem. Eng. 44,T260–T273.
Hanratty, T.J., Bandukwala, A., 1957. Fluidization and sedimentation of spherical
particles. AIChE J. 3, 293–296.Hernandez-Jimenez, F., Sanchez-Delgado, S., Gomez-Garcıa, A., Acosta-Iborra, A.,
2011. Comparison between two-fluid model simulations and particle imageanalysis & velocimetry (PIV) results for a 2D gas–solid fluidized bed. Chem.Eng. Sci. 66, 3753–3772.
Jin, S., Veeramani, C., Minev, P. and Nandakumar, K., 2008. A Parallel Algorithm forthe Direct Numerical Simulation of 3D Inertial Particle Sedimentation. In: 16th
Annual Conference of the CFD Society of Canada. Saskatoon, Canada, 8–10 June2008.
Jin, S., Minev, P., Nandakumar, K., 2009. A scalable parallel algorithm for the directnumerical simulation of 3D incompressible particulate flow. Int. J. Comput.Fluid Dyn. 23, 427–437.
Joshi, J.B., 1983. Solid–liquid fluidized beds: some design aspects. Chem. Eng. Res.Des. 61, 143–161.
Joshi, J.B., Sharma, M.M., Shah, Y.T., Singh, C.P.P., Ally, M., Klinzing, G.E., 1980. Heattransfer in multiphase contactors. Chem. Eng. Commun. 6, 257–271.
Joshi, J.B., Shah., Y.T., 1981. Hydrodynamics and mixing models for bubble columnreactors. Chem. Eng. Commun. 11, 165–199.
Joshi, J.B., Ranade, V.V., 2003. Computational fluid dynamics for designing processequipment: Expectations, current status and path forward. Ind. Eng. Chem.Res. 42, 1115–1128.
Kashyap, M., Chalermsinsuwan, B., Gidaspow, D., 2011. Measuring turbulence in acirculating fluidized bed using PIV techniques. Particuology 9, 572–588.
Lindken, R., Merzkirch, W., 1999. Velocity measurement for liquid and gaseousphase for a system of bubbles rising in water. Exp. Fluids, Supplement:s194–s201.
Lindken, R., Merzkirch, W., 2002. A novel PIV technique for measurements inmultiphase flows and its application to two-phase bubbly flows. Exp. Fluids33, 814–825.
Muller, C.R., Hartung, G., Hult, J., Dennis, J.S., Kaminski, C.F., 2009. Laser diagnosticinvestigation of the bubble eruption patterns in the freeboard of fluidizedbeds: Simultaneous acetone PLIF and stereoscopic PIV measurements. AIChE J.55, 1369–1382.
Murthy, B.N., Ghadge, R.S., Joshi, J.B., 2007. CFD simulations of gas–liquid–solidstirred reactor: Prediction of critical impeller speed for solid suspension.Chem. Eng. Sci. 62, 7184–7195.
Pandit, A.B., Joshi, J.B., 1998. Pressure drop in packed, expanded and fluidized beds,packed columns and static mixers – a unified approach. Rev. Chem. Eng. 14,321–371.
Reddy, R.K., Joshi, J.B., 2009. CFD modelling of solid–liquid fluidized beds of monoand binary particle mixtures. Chem. Eng. Sci. 64, 3641–3658.
Reddy, R.K., Jin, S., Joshi, J.B., Nandakumar, K., Minev, P.D., 2010a. Direct numericalsimulation of free falling sphere in creeping flow. Int. J. Comput. Fluid Dyn. 24,109–120.
Reddy, R.K., Joshi, J.B., Nandakumar, K., Minev, P.D., 2010b. Direct numericalsimulations of wake generated by a freely falling sphere. Chem. Eng. Sci. 65,2159–2171.
Richardson, J.F., Zaki, W.N., 1954. Sedimentation and fluidization: part I. Trans.Inst. Chem. Eng. 32, 35–53.
Sathe, M.J., Thaker, I.H., Strand, T.E., Joshi, J.B., 2010. Advanced PIV/LIF andshadowgraphy system to visualize flow structure in two-phase bubbly flows.Chem. Eng. Sci. 65, 2431–2442.
Shi, H., 2007. Experimental research of flow structure in a gas–solid circulatingfluidized bed riser by PIV. J. Hydrodyn. 19, 712–719.
Syamlal, M., O’Brien, T.J., 1989. Computer simulation of bubbles in a fluidized bed.
AIChE Symp. Ser. 85, 22–31.tenCate, A., Nieuwstad, C.H., Derksen, J.J., van den Akker, H.E.A., 2002.
Particle imaging velocimetry experiments and lattice-Boltzmann simulationson a single sphere settling under gravity. Phys. Fluids 1411, 4012–4025.
Thakre, S.S., Joshi, J.B., 1999. CFD simulation of flow in bubble column reactorsimportance of drag force formulation. Chem. Eng. Sci. 54, 5055–5060.
Tenneti, S., Garg, R., Subramaniam, S., 2011. Drag law for monodisperse gas–solidsystems using particle-resolved direct numerical simulation of flow past fixedassemblies of spheres. Int. J. Multiphase Flow 37, 1072–1092.
Uhlmann, M., Pinelli, A., 2009, Direct numerical simulation of vertical particulatechannel flow in the turbulent regime. In: Proceedings of the 20th InternationalConference on Fluidized Bed Combustion, pp. 83–96.
Van Buijtenen, M.S., Buist, K., Deen, N.G., Kuipers, J.A.M., Leadbeater, T., Parker, D.J.,2011. Numerical and experimental study on spout elevation in spout-fluidizedbeds. AIChE J. 58, 2524–2535.
Veeramani, C., Minev, P.D., Nandakumar, K., 2007. A fictitious domain formulationfor flows with rigid particles: a non-Lagrange multiplier version. J. Comput.Phys. 224, 867–879.
Wiederseiner, S., Andreini, N., Epely-Chauvin, G., Ancey, C., 2011. Refractive-index
and density matching in concentrated particle suspensions: a review. Exp.Fluids 50, 1183–1206.
Xu, Y., Subramaniam, S., 2010. Effect of particle clusters on carrier flow turbulence:a direct numerical simulation study. Flow, Turbulence Combust. 85, 735–761.
Zivkovic, V., Biggs, M.J., Glass, D., Pagliai, P., Buts, A., 2009. Granular temperature ina liquid fluidized bed as revealed by diffusing wave spectroscopy. Chem. Eng.Sci. 64, 1102–1110.
R.K. Reddy et al. / Chemical Engineering Science 92 (2013) 1–1212