Sie sind auf Seite 1von 15

Chemical Engineering Science 59 (2004) 4637 4651

www.elsevier.com/locate/ces

Pneumatic transport of granular materials through a 90 bend


Lai Yeng Leea , Tai Yong Queka , Rensheng Dengb , Madhumita B. Raya , Chi-Hwa Wanga, b,
a Department of Chemical and Biomolecular Engineering, National University of Singapore, 4 Engineering Drive 4, Singapore 117576, Singapore
b Singapore-MIT Alliance, E4-04-10, 4 Engineering Drive 3, Singapore 117576, Singapore

Received 12 January 2004; received in revised form 26 June 2004; accepted 7 July 2004

Abstract
In the present study, a pneumatic conveying system incorporating a 90 bend is investigated. This study employs the use of three
non-invasive instruments to measure solids concentration and velocity distribution determination in the pneumatic conveying system.
They are namely the electrical capacitance tomography (ECT), particle image velocimetry and phase doppler particle analyzer. Pressure
transducers were also used to monitor the pressure drop characteristics along the post-bend vertical pipe region. Two different classes of
granular materials, polypropylene beads (2600 m, Geldart class D) and glass beads (500 m, Geldart class B), were used to investigate the
differences in the ow characteristics for granular particles of various Geldart classes. The experimental results show a constant frequency
pulsating ow for polypropylene beads in the dense-phase ow regime. This is illustrated by the visualization, ECT and pressure drop
data. For dilute-phase ow regime, both polypropylene and glass beads show a continuous annulus ow structure. Numerical simulation
using the EulerEuler method was also conducted using computational uid dynamics and the uid and particle ow characteristics were
compared with the experimental data obtained in the present study.
2004 Elsevier Ltd. All rights reserved.
Keywords: Pneumatic conveying; Multiphase ow; Bend; Granular materials; Simulation

1. Introduction
Pneumatic conveying is an important process in the food
and pharmaceutical industry for transportation of granular
particles. The transport phenomenon of the conveying process in gassolid system is not fully understood despite
numerous studies, both experimental and numerical, have
been conducted on different pneumatic conveying systems
to characterize the ow proles of the solids in the pipes
of different sizes and for different pipe bends. These studies helped to optimize the pneumatic conveying process and
to assess the different methods of monitoring the conveying
systems. The gassolid two-phase ow in a vertical pipe is
heterogeneous by nature and locally unsteady. As the solids
mass loading increases, particles may come together to form
groups such as sheets, streamers or clusters. Some of the
particle groups may even experience back-ow or slipping
Corresponding author. Tel.: +1-65-6874-5079; fax: +1-65-6779-1936.

E-mail address: chewch@nus.edu.sg (C.H. Wang).


0009-2509/$ - see front matter 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2004.07.007

near the pipe wall (Rautiainen et al., 1999). Several studies


are reported on different kinds of vertical pneumatic conveying systems (Rautiainen et al., 1999; Yilmaz and Levy,
1998, 2001; Dyakowski et al., 2000; Van de Wall and Soo,
1994, 1998; Van den Moortel et al., 1997; Plumpe et al.,
1993; Zhu et al., 2003). These studies report various uctuations and clustering of particles in the vertical pipe. Table 1
summarizes the parameters investigated in some of the research. Most of these researches were conducted for smaller
sized particles falling under class A of Geldart classication.
The Geldart classication of particles provides a guideline
describing the ease of uidization of particles and its ease
of handling in pneumatic conveying (Klinzing et al., 1997).
Yilmaz and Levy (1998, 2001) measured solids velocity
and mass concentration using a ber optic probe in two different 90 pipe bends with bend radius to pipe diameter ratio
of 1.5 and 3.0. Their investigations suggested a continuous
rope-like structure that formed within the elbow and disintegrated further downstream into large and discontinuous
clusters. A continuous stream of particles was observed near

4638

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

Table 1
Operating parameters used in pneumatic conveying studies in the literature
Reference

Rautiainen et al. (1999)


Yilmaz and Levy (1998, 2001)
Dyakowski et al. (2000)
Van de Wall and Soo (1994, 1998)
Van den Moortel et al. (1997)
Plumpe et al. (1993)

dt a

Solid particles

Ug b

Gs c

(mm)

Material

Dp ( m)

(m s1 )

(kgm2 s1 )

192
154, 203
150
127

Glass beads (spherical)


Pulverized coal
Powders
Glass beads (spherical)

A
A
A
A

Glass beads (spherical)


Glass particles

3.513
1529
4 .8
1015
20

8, 10

0141

30121

200
51

64
7076

50
4260
120
85105

A
A

Geldart class

a Pipe internal diameter (d ).


t
b Supercial air velocity (U ).
g
c Solids mass loading (G ).
s

the post-bend region, while the rope dissolves into loose,


agglomerated particles further downstream. This ow phenomenon did not seem to change with the radius of curvature
of the bend (Yilmaz and Levy, 2001). The EulerLagrangian
numerical model tested by Yilmaz and Levy (2001) showed
good agreement with their experimental results. They also
showed that secondary ows induced by the 90 bends
were responsible for the dispersion of the particles after the
bend. However, their simulations did not show any signicant changes in the differences in the rope dispersion with
different bend radii, in contrast to the observation that secondary ows are caused by pressure gradients between the
inner and outer walls of a bend, and they are different for
different curvatures of the bend. Furthermore, particle interactions within the bend in terms of particlewall impacts
were shown to be less dependent on bend radius of curvature than other factors, such as conveying velocity (Li and
Shen, 1995). Thus, the phenomena and detailed structure of
roping and clustering induced by the bend in the post-bend
region are exciting areas of research but are not fully investigated in the literature.
Monitoring and control of pneumatic conveying systems are important to ensure reliable transport of materials
(Klinzing et al., 1997). One of the common instruments for
monitoring gassolids ow systems is transducer, which is
useful for determining uctuations and ow condition in
the pipe. Tomography technique has been applied for nonintrusive measurements of solids concentration and velocity
characterization for pneumatic conveying systems (Rao
et al., 2001; Zhu et al., 2003) and is useful in dense-phase
conveying systems. In this study, distributions of solids
concentration for different ow regimes in a vertical pipe
following a 90 bend were determined and analyzed using
the electrical capacitance tomography (ECT) technique.
Laser techniques such as particle image velocimetry (PIV)
and phase doppler particle analyzer (PDPA) were also used
to investigate the velocity and velocity distribution of the
500 m glass beads in the post-bend vertical pipe region. In
addition, numerical simulations using computational uid

dynamics (CFD) were conducted, and the experimental


data were compared with the numerical simulations. An
EulerEuler two-phase ow model was adopted to simulate
the dense- and dilute-phase solids ow in the vertical bend.

2. Methodology
2.1. Materials
A schematic diagram of the experimental setup is shown
in Fig. 1. Two types of granular solids, polypropylene and
glass beads, were used and their physical properties are summarized in Table 2. The test section consists of a vertical
0.75m
4

7
1.5m

6
5
3

0.65m

1m
2

2
3m

Key:

1.
2.
3.
4.

Solids feeder
5.
Acrylic pipe (In blue)
6.
Glass pipe (in green)
7.
Hose pipe (in red)
8.
Pressure transducer
ECT (connected to analyzer)
Plane of measurement for PIV and PDPA

Filter tank
Rotameter
Control valve
Vortex blower

Fig. 1. Schematic diagram of the pneumatic conveying facility.

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651


Table 2
Physical properties of particles
Material

Mean diameter SD
(m)

Geldart class

Density
(kg m3 )

Polypropylene
Glass beads

2800 500
500 50

D
B

1123
1755

glass pipe of 1.5 m (internal diameter = 0.05 m) connected


to two acrylic pipes of lengths 0.65 and 0.75 m (internal diameter = 0.05 m) at its top and bottom, respectively. The
glass pipe was specially chosen as the vertical test section
for ECT, PDPA and PIV measurements because of its rigidity and transparency over acrylic or other ordinary plastic
piping material. ECT measurements are very sensitive to
any deformation in the section of the pipe where the ECT
sensors are mounted and the higher rigidity of a glass pipe
provides better resistance to deformation. The use of PDPA
and PIV requires penetration of laser beam in the pipe, and
transparency of the pipe is necessary for these measurements. The upper portion of the vertical pipe is connected
to the vortex blower by a 4.5 m long-reinforced hose (internal diameter = 0.032 m) through a lter tank. The lower
portion is connected to a 3 m horizontal conveying section
(internal diameter = 0.050 m) by a 90 L-shaped bend with
zero bend radius.
2.2. Methods
Air and solids were fed into the pneumatic conveying
system through the inlet at the horizontal conveying section. The solids were conveyed up the vertical test pipe
through the 90 bend and are subsequently collected in the
lter tank. The experiments were conducted for conveying
of polypropylene beads with solids mass loading (Gs ) of
15.8, 31.1 and 46.3 kg m2 s1 . For each mass ux, experiments were conducted at several supercial gas velocities
(Ug ) ranging from 10.617.0 m s1 . For glass beads, experiments were conducted with solids mass loading of 15.8 and
29.0 kg m2 s1 in the same range of air supercial velocity.
The system was left to run for a steady stream of air before
the introduction of solids. Several non-invasive instruments
were utilized in this study to obtain the solid-phase ow
characteristics measurements in the post-bend vertical pipe
section.
2.2.1. Pressure transducers
Two pressure taps were mounted at locations of 0.65 and
2.35 m, respectively, from the 90 bend to install a differential pressure transducer (Model-DT1400-1UD-125 Stellar). Differential pressure measurements were obtained between the two pressure taps and the data were acquired using
DataVIEW (Version 1.1, Cumming, Georgia). The sampling
rate was set at 200 Hz and the pressure data were acquired
for 30 s.

4639

2.2.2. ECT
Two 12-electrode ECT sensors were mounted on the
glass pipe at 1.64 and 2.14 m, respectively, from the bend.
ECT data were acquired using ECT32 software and a dataacquisition module (Process Tomography Ltd, Wilmslow,
Cheshire, UK). The sampling rate was set at 40 Hz and the
data were acquired for 30 s for each experimental run. s
(x, y, z, t), dened as the local volume fraction of the solids
was obtained from post-processing of ECT data for each
instant of time (t) using the simultaneous iterative reconstruction technique (SIRT) given by Su et al. (2000). x and
y denote the Cartesian coordinates in the cross-sectional
plane and were normalized using the pipe diameter as the
characteristic length, the axial coordinate z denotes the location of the ECT electrodes. Particle volume fraction is
expressed as s , while the (t ), known as the time-averaged
particle volume fraction values were generated by averaging s (x, y, z, t) over a time period T (in this case, the
rst 25 s of data acquired for each run):

1 T
t (x, y, z) =
s (x, y, z, t) dt.
(1)
T 0
The instantaneous value of the cross-sectional average particle volume fraction, s (z, t), is dened as
 
1
s (z, t) =
s (x, y, z, t) dx dy
(2)
A
and the time-average value of s (z, t) is denoted by (z)

 
1 T
1
(z) =
s (z, t) dt
 t (x, y, z) dx dy.
T 0
A
(3)
2.2.3. PIV
Measurements using a PowerView 2D PIV system (TSI
Corporation, USA) were conducted at the vertical pipe region 1 m downstream from the bend. The laser light sheet
generated by the Laser PulseTM Solo Mini Dual Nd:YAG
laser was introduced into the glass pipe and two snapshots
of the lighted particles were taken by a PowerViewTM 4M
2K2K camera for a very small time interval (dT =100 s).
The cross-correlation yields the distance traveled by the particles from the rst snapshot to the second and the velocity
distribution of particles in the plane can be determined by
dividing by the time interval. In this study, the variation of
velocity in the post-bend region in the axial direction of the
horizontal pipe was measured and the results were compared
with PDPA measurements and simulated numerical results.
2.2.4. PDPA
Some measurements using PDPA (XMT204-2.2, TSI Corporation, USA) were taken to compare with the velocity
measurements obtained by PIV under the same ow conditions for dilute-phase ow. The laser wavelength used was
514 nm. The phase doppler method is based upon the principles of light-scattering interferometry. The PDPA is able to

4640

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

measure the local velocity of each particle and the particle


size distributions accurately. The signal analyzer and postprocessing software (Aerometrics Real-Time Signal Analyzer, DataVIEW) could also record the number density of
the measurement volume online. The PDPA measurements
were taken at three radial points. For each measurement location, at least 1000 particles were sampled.

Outlet

g
2m

2.2.5. Numerical simulation (methods and boundary


conditions)
Commercial CFD package FLUENT 6.0 was used to
simulate the pneumatic conveying system. The air and
solid ows were modeled using conservation equations of
Anderson and Jackson (1967). The volume-averaged approach was used for the conservation equations (refer to
Appendix A). The phase-coupled SIMPLE (PC-SIMPLE)
algorithm (Vasquez and Ivanov, 2000) was used for the
pressurevelocity coupling. The velocities for both solid and
gas phases were solved from the coupled differential equations using the block algebraic multi-grid scheme (Weiss
et al., 1999) with the governing equations solved sequentially in an implicit, unsteady fashion. A total of 30,000
structured hexagonal grids were used in the geometry with
size ranges from 3.8108 to 3.2107 m3 . Grid independence studies were performed to show that increasing the
number of grids does not result in noticeable changes to the
simulation results (Quek, 2003). Time steps of 1 106 nd
1 104 s were used for simulation of polypropylene and
glass beads, respectively. Details of the model construction
can be found in Appendix A.
The calculation domain consists of a horizontal inlet pipe
(internal diameter = 0.05 m) of length 1 m, a 90 elbow oriented vertically, followed by a vertical pipe (internal diameter = 0.05 m) of height 2 m (as shown in Fig. 2a). For illustration, all ow quantities were either averaged over a plane
or taken along the centerline in a plane. Qualitative contour
results are presented in the vertical (xy) symmetry plane
of the geometry (as shown in Fig. 2b).
3. Results and discussion
There is a marked difference in the ow characteristics of
polypropylene and glass beads. This can be observed from
experimental results of system uctuations, solids concentration and velocity distribution in the vertical and horizontal lines. Some of the experimental observations were also
obtained from the simulation conducted for the system.
3.1. Flow regimes and uctuations in the pneumatic
conveying system
The ow characteristics of polypropylene beads differ
from glass beads signicantly. The ow characteristics observed in the experiments for the polypropylene may be divided into two distinct types: dense- and dilute-phase ow

Plane B

1m

y
Inlet
Origin, 0

0.05 m

(a)

Plane A

Symmetry planes A &


B showing the
qualitative feature of
particle distributions

Plane centerline

z
Plane B

Plane A

(b)
Fig. 2. Geometry for numerical simulation based on Fig. 1. Pipe diameter
= 0.05 m, and 1 m in the horizontal section before the bend, 2 m in
the vertical section after. All numerical results presented either as a
cross-sectional average or along the centerline of the plane. Qualitative
pictures were taken in the vertical symmetry planes A& B.

depending on whether the pressure drop decreased or increased with increasing supercial gas velocity (Rao et al.,
2001). This can be seen from the differences in pressure gradient prole, ow pattern, pressure and solids concentration
uctuation and solids concentration distribution as described
in the following section.
The pressure gradient prole of the system is useful for
identication of the type of ow in the pneumatic conveying system. The Zenz state diagram (Zenz, 1949; Herbreteau
and Bouard, 2000) obtained by plotting pressure drop with
supercial velocity for various solids mass loading is useful
for identication of ow regime in a pneumatic conveying
system. In the region where pressure drop decreases with increasing supercial gas velocity, the system is operating in
the dense-phase regime with non-homogenous and unsteady
ow in the conveying lines. As supercial velocity of gas

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

P/L, Pa/m

800.0
-2 -1

700.0

Gs = 15.8 kgm s

600.0

Gs = 31.1 kgm-2s-1

500.0

Gs = 46.3 kgm-2s-1

400.0
300.0
200.0
100.0
0.0
8.0

10.0

12.0

(a)

14.0

16.0

18.0

20.0

Ug, ms-1
150.0
140.0

P/L, Pa/m

130.0
120.0
110.0
100.0
90.0

Gs = 15.8 kgm-2s-1

80.0

Gs = 29.0 kgm-2s-1

70.0
60.0
8.0

(b)

10.0

12.0

14.0

16.0

18.0

20.0

Ug, ms-1

Fig. 3. Pressure gradient (Pa m1 ) against supercial gas velocity


(m s1 ) in the vertical pipe (post-bend section) for two-phase ow (a)
air-polypropylene beads; (b) air-glass beads. Gs indicated in the legend
refer to the solid mass ux (kg m2 s1 ).

increases further, the pressure drop begins to increase with


gas supercial velocity and the ow becomes steady and
homogenous (dilute-phase ow). The ow characteristics of
the two classes of solids are illustrated by pressure gradient prole plot for polypropylene beads and glass beads in
Figs. 3a and b, respectively.
Fig. 3a shows the pressure gradient in the vertical pipe
generally decreased with increasing supercial gas velocity (a dense-phase conveying characteristics). In this regime,
the solids were conveyed in alternating pulses of high and
low solids concentration as shown in Figs. 4a and b. The
solids formed clusters at the horizontal pipe region and the
particles were conveyed in moving dunes as illustrated in
Fig. 4c. In this case, cluster formation is determined by a
qualitative observation of the differences in solids concentration and solids distribution in adjacent sections of the
conveying line. The formation of moving dunes in horizontal conveying systems for the Geldart class D beads was
also reported in the study by Zhu (2003). Dilute-phase ow
for polypropylene beads were obtained for the lowest solids
mass loading near the maximum supercial gas velocity
used for the experiments. In this regime, the solids ow was
continuous, steady and homogeneous.

4641

Conversely, as shown in Fig. 3b, the pressure gradient for


glass beads increased with increasing supercial gas velocity for the same range of supercial gas velocity used in
polypropylene beads experiments. This is an indication that
the system was operating in a dilute phase whereby the ow
of particles was continuous. No roping or clustering was observed both in the horizontal and post-bend vertical section
of the pneumatic conveying system. Unlike the conveying
of polypropylene beads, a transition from dense- to dilutephase ow cannot be achieved. For the range of supercial
gas velocity and mass loading of glass beads used in this
study, only a dilute-phase ow regime was obtained. Reducing the supercial gas velocity below 10 m s1 caused the
glass beads to accumulate at the inlet of the conveying section and the conveying rate was less than the feeding rate.
Therefore, the experiments for glass beads were conducted
with supercial velocities higher than 10 m s1 to ensure a
fully developed steady state ow. In that range, the glass
beads move along the horizontal conveying section in a continuous stream. Small clusters of solids like moving dunes
ow were not noticeable. In the vertical post-bend region,
the glass beads were conveyed in a homogeneous, continuous and dispersed stream. The differences in the pressure
gradient proles for the two classes of particles illustrate the
marked differences in the ow characteristics.
Pressure uctuations in the conveying of solids are useful for ow pattern identication (Dhodapkar and Klinzing,
1993). The pulsating ow pattern for the dense-phase conveying of polypropylene beads was demonstrated by the
measurements of the uctuation of pressure gradient along
the vertical pipe with time, as shown in Fig. 5a. This pulsating ow pattern was also observed in the cross-section average solids volume ratio measurements (s ) made using ECT
(Fig. 5b). As Figs. 5a and b show, the uctuations have a regular period of about 2 s (i.e. 0.5 Hz). In order to investigate
the uctuation frequency for this dense-phase ow, power
spectral density (PSD) functions for pressure drop and volume ratio were used to analyze the pressure and ECT data
and the results are illustrated in Figs. 6a and b, respectively.
Sampling rates for pressure and ECT measurements were
set at 200 and 40 Hz, respectively, therefore the peaks obtained for power spectra of pressure uctuations were more
distinct (Fig. 6a). The power spectra obtained from both sets
of data are similar and show a dominant peak at approximately 0.5 Hz. This is consistent with the uctuation frequency of approximately 2 s as observed from Figs. 5a and
b. The power spectra also indicate secondary peaks present
at approximately 1 and 1.5 Hz. Further experiments showed
that at low supercial velocity, the pressure uctuation is
predominantly near 0.5 Hz. As supercial velocity increases
and the ow becomes dilute, the peak near 0.5 Hz is less
predominant and uctuations extend to 1015 Hz regime.
The ECT data obtained experimentally were reconstructed
to obtain the solids distribution prole for the pneumatic
conveying system. The ECT visualization results for dense
ow of polypropylene beads and dilute ow of glass beads

4642

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

Fig. 4. Photograph of conveying of polypropylene beads for Gs = 31.1 kg m2 s1 , Ug = 10.6 m s1 . (a) Pulse of high concentration in vertical pipe;
(b) low solids volume fraction, dispersed ow in vertical pipe after a few seconds and (c) moving dunes formed in horizontal conveying section (clusters
of solids as shown in circle drawn).

are shown in Figs. 7a and b, respectively. Figs. 7a and b


show the solids volume fraction distribution with 1 s interval between two successive frames. Fig. 7a showed consistent results with the uctuations analysis. As the ECT data
indicate, during the dense-phase ow, the solids show an alternating coreannulus/annulus structure. There was a high
concentration of solids in both the core and annular region
of the pipe at one time and the core disappears leaving only
an annular structure 1 s later. When the system was operating in the dilute phase, the solids tend to be in the annular
region. As shown in Fig. 7b, the solids were seen to consistently form an annular structure for all the 25 frames shown.
From the numerical simulation, differences in the pneumatic conveying of the two classes of particles were also observed. Fig. 8 shows the solids concentration distribution in
the cross-section of vertical pipe at 0.9 m downstream from
the bend for conveying of polypropylene beads in the densephase ow regime. Fig. 8a shows a time when there was low
concentration of solids moving through, while Fig. 8b shows
a high concentration solids core at a short time following
Figure 8a. The time difference between Figs. 8a and b is
0.11 s. Figs. 8c and d show the solid concentration prole

at the same location but at a few seconds later than the time
for Figs. 8a and b. The time difference between Figs. 8c and
d is about 0.17 s. This is in agreement with the experimental
results of alternating high and low solids concentration with
time. However, the frequency of uctuation is different from
the experimental results. Figs. 8e and f show the solids concentration at a distance of 2 m from the bend, with a time
difference of 0.12 s. The variation in solids ow structure
decreases with the increasing distance from the bend. On
the other hand, numerical results for the conveying of glass
beads showed a steady consistent solids volume ratio distribution in the vertical pipe at a plane 1 m downstream of the
pipe bend. This is consistent with the experimental observation of a non-uctuating system. The cross-sectional volume
fraction distribution results at the planes y = 0.9 and 2.0 m
from the bend are shown in Figs. 9a and b, respectively.
Though the numerical results presented in Figs. 8ad
showed some qualitative similarity to Fig. 7a, the solid uctuation frequency, and thus the cluster formation, which may
be attributed mainly to the particleparticle interaction, appears to be different. The uctuation period was about 2 s for
the ECT experiments, whereas a period of only about 0.1 s

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

4643

700.0
600.0
P/L, Pa/m

500.0
400.0
300.0
200.0
100.0
0.0

10

(a)

15

20

25

15

20

25

t, s
0.06
0.05

0.04
0.03
0.02
0.01
0
(b)

10
t, s

Fig. 5. Fluctuations in pneumatic conveying of polypropylene beads in


the vertical pipe for Gs = 31.1 kg m2 s1 , Ug = 11.9 m s1 (a) pressure
gradient in vertical pipe (Pa m1 ) against time (s) and (b) solids volume
fraction (s ) in a single plane in the vertical pipe against time (s).

was recorded for the simulations as illustrated in Fig. 10a.


The apparently more rapid succession and less-concentrated
clusters could be attributed to their differences in the mechanism of cluster formation. The uctuations observed in the
ECT experiments could be mainly due to the inlet air/solids
ow rate variations, and in the simulation model, the inlet
air/solids ow rates were assumed to be constant. On the
other hand, the simulation results for glass beads indicated
the absence of high uctuations of solids concentration as
observed in polypropylene beads simulation, which is shown
in Fig. 10b.
The differences in solids distribution for the two types
of conveying system were also observed in the simulation
results. Fig. 11 shows the simulation results for solids concentration distribution of particles from various sections of
the vertical conveying line. Though the numerical simulation was carried out in a time-dependent state, the qualitative features of the solid ow did not vary signicantly with
time. The corner of the bend showed deposition of solids
(Fig. 11a) as observed in the experiments. These solids,
which were trapped permanently in the corner, affected the
particleparticle collisions and rebound of the solids moving through the bend. In the solids clusters formed in a bend,
such particleparticle collisions are frequently encountered
and they are believed to preserve solids size and shape better
than collisions between unlike materials, i.e., between solids

Fig. 6. Power spectral density (PSD) analysis of uctuations for the


pneumatic conveying of polypropylene beads in the vertical pipe.
Gs = 31.1 kg m2 s1 , Ug = 11.9 m s1 (a) PSD of pressure data and (b)
PSD of solids concentration data.

and wall. Some cluster like pattern could be seen directly in


the region above the bend, where the mass of solids started
to form groups with higher concentration. These higher concentration groups separate into distinct clusters at a further
distance from the bend, producing a pulsing effect as shown
in Fig. 11b. This can be seen to occur within 1 m (20 diameters distance) from the bend and similar pulses were observed in the experiments shown in Figs. 4a and b where the
actual pulsing ow in the experiments is recorded using a
conventional video camera. Fig. 12 shows the numerical results for pneumatic conveying of glass beads under the same
supercial velocity and comparable solids mass loading with
polypropylene. For smaller solids (glass beads), a continuous particle ow was observed along the vertical pipe. There
were no signicant variations in the cross-sectional solids
concentration distribution for different planes in the vertical
pipe and at different times. Therefore, there was no roping
or clustering of solids as observed in the experimental and
numerical results for polypropylene. This is consistent with
the experimental observation of continuous steady ow of
glass beads as stated above. Table 3 shows the time-averaged
particle concentration at the core for ECT in comparison
with those obtained from simulations. A large change in the
core concentration was observed for different supercial gas

4644

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

Fig. 9. Solid-phase volume fraction contours at (a) y =0.9 m; (b) y =2.0 m


from the bend at time t = 8 s. Gs = 29.0 kg m2 s1 , Ug = 10.6 m s1 .
A similar prole is observed which indicates a steady ow stream in the
vertical pipe region.

0.04

0.03
Fig. 7. (a) ECT images obtained for pneumatic conveying of polypropylene
in the vertical pipe. Gs = 31.1 kg m2 s1 , Ug = 11.9 m s1 and (b) ECT
images obtained for pneumatic conveying of glass beads in the vertical
pipe. Gs = 29.0 kg m2 s1 , Ug = 10.6 m s1 .

0.02

0.01
x-sectional average s at 0.9m
x-sectional average s at 2.0m

0.00
4.5

5.0

5.5

(a)

6.0

t, s
0.004
0.0035
0.003

0.0025
0.002
0.0015
0.001
x-sectional average s at 0.9 m

0.0005

x-sectional average s at 2.0 m

0
5

Fig. 8. Simulation results for polypropylene beads conveying. (a)(d)


Solid phase volume fraction contours in the vertical post-bend region,
in the horizontal plane at y = 0.9 m from the bend, taken at different
times Gs =31.1 kg m2 s1 , Ug =11.9 m s1 ; (e)(f) Solid-phase volume
fraction contours in the horizontal plane at y =2.0 m from the bend, taken
at different times Gs = 31.1 kg m2 s1 , Ug = 11.9 m s1 .

velocities in ECT measurements, but simulation results


showed only small changes. This may be attributed to the
inability of the model to capture the moving dunes formation at constant frequency in the horizontal pipe section of
the system.

(b)

5.5

6.5

7.5

t, s

Fig. 10. Numerical solutions illustrating the differences in uctuations for


the two systems. Solids volume fraction plane-averaged values (s ), in the
planes at distances of 0.9 and 2.0 m in the vertical section from the bend
for (a) polypropylene beads, Gs = 31.1 kg m2 s1 , Ug = 11.9 m s1 ; (b)
glass-beads, Gs = 29.0 kg m2 s1 , Ug = 10.6 m s1 .

3.2. Distribution of solids velocity in the vertical pipe


The volume ratio data obtained simultaneously for the two
ECT planes may be processed using the cross-correlation
function to determine the pattern velocity of the solids in

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

Fig. 11. Predicted polypropylene solids volume fraction (s ) in the symmetry plane of the entire geometry. High solid concentration in the sharp
bend corner, with solid clusters above it and disperse rope at the top
of the vertical section. Enlarged sections show the clustering and roping
regions of the vertical pipe Gs = 31.1 kg m2 s1 , Ug = 11.9 m s1 .

Table 3
Comparison of experimental time-averaged solid volume fraction of
polypropylene at the core and polypropylene pattern velocity from ECT
with simulation results
Supercial gas
velocity (m s1 )

ECT

Simulation

s

s

ECT pattern
velocity (m s1 )

Simulation
us (m s1 )

10.6
11.9
16.0

0.136
0.048
< 0.01

0.0719
0.0590
0.0300

3.08
3.33
3.33

1.11
1.63
3.48

Gs = 31.1 kg s1 .

the pipe. The cross-correlation coefcient, C(d), where d


denotes the delay time, was then computed as

1 T
C(d) =
(s (z1 , t) (z1 ))(s (z2 , t + d)
T 0
(z2 )) dt.
(4)
Here, z1 and z2 refer to upstream and downstream planes,
respectively, and d was taken to be positive. The dominant pattern propagation velocity, V , was estimated from

4645

Fig. 12. Predicted glass beads solids volume fraction (s ) in the symmetry plane of the entire geometry. Accumulation of solids observed in the
sharp bend corner. Continuous ow in the horizontal and vertical sections. No clustering or roping was observed for the vertical pipe region.
Gs = 29.0 kg m2 s1 , Ug = 10.6 m s1 .

V = L/D , where L is the axial distance between the two


ECT sensors and D is the value of d at which C(d) assumes the largest value. In this study, the cross-sectional
averaged volume ratios obtained from the two ECT planes
were processed using the cross-correlation function and resulting C(d) values were normalized with the maximum
C(d) value. The time at which the peak value for a plot
C(d) with time occurs is taken as the time delay for estimate
of pattern velocity. V estimated in this manner is clearly
based on cross-sectional averages.
Table 3 shows the pattern velocity obtained from ECT data
for polypropylene in comparison with simulated solid-phase
velocities with varying supercial gas velocities. The experimental solid-phase velocities were calculated by crosscorrelating the temporal signals about the cross-section averaged solids concentrations at the planes y = 1.64 and 2.14 m
after the pipe bend. The experimental values and the numerical results were within the same order of magnitude, showing qualitative consistency of the measurements. However,
the ECT data indicate insignicant change in pattern velocity with increasing inlet velocity, while the simulation model
shows higher sensitivity to varying inlet gas velocities.

4646

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

Normalized solids velocity (us /us,max)

Pipe wall

(a)
6

Solids velocity, us (m/s)

1
0.8
0.6
0.4
Numerical Results

0.2
0
-25

Pipe wall

1.2

PIV results

-20

-15

-10
-5
0
5
10
15
Position from pipe center (mm)

20

25

Fig. 14. Comparison of the normalized velocity prole for glass beads
from PIV measurements and numerical simulation.

2
10.6 m/s
11.9 m/s
13.2 m/s
17.0 m/s

Pipe wall

Pipe wall

0
-25

-20

-15

(b)

-10

-5

10

15

20

25

Position from pipe center (mm)

Fig. 13. Experimental results from PIV measurements: (a) Example of


a snapshot taken using the PIV system and (b) velocity (us ) prole of
the glass beads at position 1 m from the pipe bend for four different
supercial gas velocities.

Table 4
Time-averaged solids velocity of glass beads at different radial positions
in the vertical pipe
r/R

PIV, us
(m s1 )

PDPA, us
(m s1 )

Simulation, us
(m s1 )

0.6
0.25
0

3.064
3.132
3.264

2.737
2.426
4.336

4.806
5.215
5.704

Gs = 29.0 kg s1 , Ug = 10.6 m s1 . R is the radius of the pipe.

For glass beads, the solids velocity in the vertical postbend section could not be accurately determined from crosscorrelation of ECT volume ratio data due to the very dilute
ow of the system. Therefore, laser techniques namely the
PDPA and PIV were employed to determine the solids velocity of glass beads. Figs. 13a and b show a snapshot of
the image taken using PIV for a supercial gas velocity of
10.6 m s1 and the cross-section distribution of solids velocity for four different air ow rates, respectively. Table 4
shows the comparison of solids velocity for glass beads using PIV, PDPA and simulation. The qualitative results for the
velocity distribution were quite consistent with a very low
(close to zero) solids velocity just next to the pipe wall and
a maximum velocity close to the center of the pipe. The average solids velocities obtained were also fairly consistent.
The normalized velocity proles for both PIV measurements
and numerical simulation are shown in Fig. 14. Simulation

results showed a more distinct maximum velocity off the


pipe center while the measurements from PIV demonstrate
a atter velocity prole along the radius of the pipe. The
asymmetric nature of the velocity prole in the numerical
results indicates more signicant sensitivity of the simulation to the effects of the pipe bend then actual experiments.
In the actual experimental runs, the particleparticle interactions may enhance the mixing of solids in the post-bend
region which may lead to a more symmetrical ow as observed in ECT and PIV measurements of solids concentration and velocity.
3.3. Numerical results for the horizontal conveying section
Simulation data for pneumatic conveying of the class D
particles (polypropylene beads) were compared with the experimental data obtained from the literature. Fig. 15a shows
the simulated gas-phase axial velocity, plotted along the
centerline of the planes at distances of 12, 16 and 18 diameters from the pipe inlet (Fig. 2b), respectively (lines AC in
Fig. 15(a)). The simulated results were compared with the
experimental data of Tsuji and Morikawa (1982), conducted
under the experimental conditions similar to those used
in the present numerical simulation (shown by line D in
Fig. 15(a)). The almost identical velocity proles at different distances indicate that the airow is fully developed before reaching the bend. This is consistent with Carpinlioglu
and Gundogdu (1998), who measured fully developed twophase ow for 3 < x/D < 4, using wheat particles with a
mean diameter of 825 m. It was noted that the particle size,
shape and the Reynolds number have signicant effect on
this development distance.
Fig. 15a also shows that the maximum velocity value
is found closer to the top of the pipe, and this asymmetric prole is also qualitatively similar to those found by
previous investigators (Tsuji and Morikawa, 1982; Huber

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

1.0

1.0
0.8

0.8

A
B
C
D

0.6

A
B
C

0.6
y/D

y/D

4647

0.4

0.4

0.2
0.0

0.2
0.0

0.2

0.4

0.6

(a)

0.8
1.0
ug/Ug

1.2

1.4

1.6
0.0
0.0

1.0

0.1

0.2

0.3

0.4

(u'g2)1/2/Ug, (u's2)1/2/Ug, us/Ug

0.8

Fig. 16. Predicted (A) axial solid velocity (us /Ug ), (B) solid-phase

A
B
C

0.6

axial turbulence ((us2 )0.5 /Ug ) and (C) gas-phase axial turbulence


y/D

((ug2 )0.5 /Ug ) values. Values taken at the vertical line through the center of
the pipe at a distance of 0.9 m from the pipe inlet, Gs = 31.1 kg m2 s1 ,
Ug = 11.9 m s1 .

0.4
0.2
0.0
0.00
(b)

0.01

0.02

0.03

0.04

0.05

0.06

Solid volume fraction, s

Fig. 15. (a) Normalized gas-phase axial velocity (centerline values) from
numerical simulations, at distances of 12 (A), 16 (B) and 18 (C) diameters (x = 0.5, 0.8&0.9 m) measured from the inlet. Gs = 31.1 kg m2 s1 ,
Ug =11.9 m s1 . Curve D presents values taken from Tsuji and Morikawa
(1982), with 0.2 mm diameter particles through a 0.0305 m diameter pipe,
Gs = 18.4 kg m2 s1 , Ug = 6 m s1 and (b) predicted solid volume fraction (s ) for 2.8 mm polypropylene beads in the horizontal pipe section,
along the centerline in the plane located at the different distances from
inlet, x/D=10 (A), 16 (B) and 18 (C), Gs =kg m2 s1 , Ug =11.9 m s1 .

and Sommerfeld, 1994, 1998; Zhu et al., 2003; Levy and


Mason, 1998; Bilirgen and Levy, 2001). This is attributed to
the loss of gas-phase momentum because of the higher particle concentration near the bottom of the pipe than the top
(Fig. 15b). Distortion by the bend in front was also cited as
a reason for the parabolic proles (Huber and Sommerfeld,
1994, 1998).
The high solid volume fraction (s ) at the bottom of the
horizontal pipe wall, shown in Fig. 15(b), was caused mainly
by gravitational settling of solids (Tsuji and Morikawa,
1982; Huber and Sommerfeld, 1994, 1998; Zhu et al., 2003;
Levy and Mason, 1998; Bilirgen and Levy, 2001). However,
this was aggravated by the more intense particleparticle
and particlewall collisions at the bottom of the pipe, resulting in greater loss of solids momentum as compared to the
upper portion of the pipe (Huber and Sommerfeld, 1994). In
addition to inelastic particle collisions, viscous dissipation
of the gas phase also played a role in the momentum loss
(Huber and Sommerfeld, 1994).

Similar parabolic prole could be seen in the simulation


results of particle velocity (lled circles) data as shown in
Fig. 16. Fig. 16 also shows the axial velocity uctuations of
the both gas and solids phases (open circle for solids velocity,
and the lled triangles for gas velocity). More uctuations
were seen in the gas velocity. Such similarities in velocity
uctuations were also observed for the fully developed twophase ow by previous studies (Tsuji and Morikawa, 1982;
Huber and Sommerfeld, 1994, 1998; Zhu et al., 2003). Minimum velocity uctuations in both phases coincide at an
off center region, as observed experimentally by Huber and
Sommerfeld (1994, 1998) while high values were observed
at the wall regions. This was attributed to the response of
the solids to the high gas turbulence near the walls (Huber
and Sommerfeld, 1994). Another reason for these observed
proles was due to the particlewall collisions occurring
in the near wall regions (Tsuji and Morikawa, 1982; Huber
and Sommerfeld, 1994, 1998). However, gas-phase turbulence at the pipe bottom wall was relatively lower than that
at the pipe top wall. This was due to the higher solids concentration at the bottom wall (Fig. 15b), leading to higher
gas-phase turbulence suppression (Huber and Sommerfeld,
1994). Numerical simulation scheme adopted in this work
seemed to predict the gas and solid velocity proles
reasonably well.
3.4. Numerical results near the bend
The characteristics of the ow phenomena of the granular
solids in the vertical section are mainly due to the presence
of the 90 bend upstream of the vertical pipe. The cluster
formation could be attributed to the inelastic collision within
the bend, as well as the impact angle. Particles traveling
in the horizontal pipe towards the bend can be expected to

4648

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

collide at or near an angle of 90 . This would severely dissipate some of the kinetic energy while redirecting particles
from the horizontal to the vertical pipe. Particles would then
have to decelerate, rebound and change ow direction, and
re-accelerate upwards. Particles colliding within the bend
do not transfer its kinetic energy upwards immediately, but
would rebound backwards and collide with solids arriving
later. This would cause a net loss of momentum due to inelastic collisions for the whole group of solids arriving at
the bend within a short period of time. These groups would
then be accelerated upwards as a cluster due to the dissipation of granular kinetic energy.
At a further distance, near the outlet of the vertical pipe,
a continuous stream of solids of lower concentration can be
seen moving rapidly up (Fig. 11c). The solids ow structure
evolution in the vertical section can be understood as four
continuous stages: cluster formation (Fig. 11a), cluster ow
(Fig. 11b), dispersion/rope formation (Fig. 11c) and rope
ow (Fig. 11d).
One point to note is that a sharp bend has been used
in this work as compared to the smooth bend used in the
work of Yilmaz and Levy (1998, 2001). The rope formation
from the disintegration of the clusters was not seen in the
experiments of Yilmaz and Levy (1998, 2001), where an
opposite phenomenon of dissipation of a continuous rope
into discontinuous clusters occurred for smooth bends.
Previous studies (Yilmaz and Levy, 1998, 2001; Huber and
Sommerfeld, 1994, 1998; Levy and Mason, 1998; Bilirgen
and Levy, 2001) reveal that secondary velocities in the gas
phase and turbulent ow are responsible for the dispersion
of particles in the vertical post-bend region. For the present
study, only the turbulent local mixing can be used to explain
particle dispersion, as secondary velocities induced by the
bend does not seem to have observable effect on particle
dispersion, which usually occurs at a short distance after the
bend. However, solid clusters are found to be formed within
1 m from the bend, and the dispersion of the clusters in a
rope occur only after about 20 diameter distances from the
bend (> 1 m).

4. Conclusions
Distinct differences in the ow characteristics of different
granular materials were observed in this study. The singleplane ECT data at various axial locations of the conveying
pipe determine the non-uniformities in the cross-sectional
solids concentration distribution for the vertical bend. ECT
images for the vertical conveying of polypropylene in a
dense-phase ow show a low-frequency pulsing ow with
an alternating coreannulus and annulus structure. This pulsating ow pattern was also observed in the simulation results. The coreannulus structure is a result of the immature
moving dunes formed at the horizontal conveying section.
At higher supercial gas velocities, the ow becomes more
dilute and the time-averaged solids concentration at the core

is found to decrease. In dilute phase, no pulsing ow was


observed in the vertical conveying section, and no particles
are found in the core and the ow has an annulus structure.
The formation of moving dunes in the horizontal conveying section is an inevitable phenomenon for conveying of
granular materials with similar properties as polypropylene
beads. This type of ow behavior in the horizontal pipe has
effects on the ow pattern in the vertical post-bend region.
The conveying of class B particles (glass beads) shows
a continuous ow with less uctuation in solids concentration in the vertical pipe region. This is reported by both the
simulated results and experimental ECT measurements. It is
important to note that granular materials with similar physical properties to the glass beads used (Class B) have to be
conveyed in the dilute phase. Conveying at low supercial
gas velocities will lead to unsteady ow of materials with
feeding rate higher than conveying rate and problems such
as choking of pipes may arise.
Numerical solutions for the ow characteristics of the
uid phase in the horizontal section compares well with
the experimental results obtained from the literature reports.
The EulerEuler model for the two-phase ow simulates the
cluster formation in the vertical pipe after the bend reasonably well, although dune formation in the horizontal section
could not be predicted without considering uctuations in
ow conditions at the inlet. While core structures in the solid
ow were seen in the post-bend region in experiments and
simulations, their uctuation intervals differ by nearly 20
times (0.5 vs.10 Hz). The cross-section averages of the solid
volume fraction in the clusters predicted by the numerical
simulation are of the same order as ECT measurements.

Notations
CD
ds
ess
g0,ss
Gs
kg
ks
Ksg
n
Res
ug
us
Ug
Vg
Vs
V

interphase drag coefcient


diameter of solid particle, length
interparticle collision coefcient, energy after
collision and before collision
radial distribution function
solids mass loading, kg m2 s1
gas-phase turbulent kinetic energy, m2 s2
solid-phase turbulent kinetic energy, m2 s2
interphase momentum exchange, from solid-togas phase
number density of solid particles, number of particles per unit volume
Reynolds number
gas-phase velocity, m s1
solid-phase velocity, m s1
supercial gas velocity, m s1
phase-weighted gas velocity, m s1
phase-weighted solid velocity, m s1
solids pattern velocity, m s1

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

Greek letters

and for the solid, it is:

g
s
s
s
g
g
s


s = s s ( us + uTs ) + s (s 23 s ) us I .

s

gas-phase volume fraction


solid-phase volume fraction
solid bulkviscosity, kg m1 s1
solid shear viscosity, kg m1 s1
gas-viscosity, kg m1 s1
gas-phase density, kg m3
solid-phase density, kg m3
angle between mean particle velocity and mean
gas velocity
granular temperature, m2 s2

Acknowledgements
This study has been supported by the National University
of Singapore under the Grant No. R-279-000-095-112. We
thank Prof. Reginald Beng Hee Tan and Dr. S. Madhusudana
Rao for many helpful discussions on the project.
Appendix A. Equations applied in FLUENT simulation
of the pneumatic conveying system
The equations of mass and momentum in general forms
are:
For the gas phase:

*
(g g ) + (g g ug ) = 0,
*t
*
(g g ug ) + (g g ug ug )
*t
us ug )
= g pg + g + Ksg (

+g g Flift,g + g g g.

(A.1)

*
(s s ) + (s s us ) = 0,
*t
*
(s s us ) + (s s us us )
*t
ug us )
= s pg ps + s + Kgs (
+s s Flift,s + s s g,

(A.3)

(A.4)

where t is the time, g and s are gas and solids densities,


respectively, the g and s are the volume fractions of the
air and solid phases, respectively, Flift,g and Flift,s are the
lift forces (Flift,g = Flift,s ), g is the gravitational acceleration, pg is the pressure gradient of the gas phase, ps is
the solid pressure gradient, Kgs and Ksg are the interphase
exchange coefcients, g and s are the stress-strain tensors
of the gas and solid phases, respectively. For the gas it is
written as

g = g g ( ug + uTg ) 23 g g ug I

bulk viscosities and I is the unit tensor. The solids pressure


is composed of a kinetic term and a second term due to
particle collisions:
ps = s s s + 2s (1 + ess )2s g0,ss s ,

(A.5)

(A.7)

where ess is the coefcient of restitution for particle collisions, set at 0.9, g0,ss is the radial distribution function, and
s is the granular temperature. The radial distribution function is modeled as proposed by Ogawa et al. (1980).

 1 1

3
s
g0,ss = 1
(A.8)
s,max
The maximum packing limit, s,max has been set to 0.65 in
this simulation.
The interphase exchange model follows that of Yang and
Yu (1966),


s g g us ug  2.65
3
Ksg = CD
g
,
(A.9)
4
ds
where
CD =

24
[1 + 0.15(g Res )0.687 ]
g Res

(A.10)

and the relative Reynolds number, Res is dened by

g ds |
us ug |
g

(A.11)

with Ksg = Kgs .


The lift force is computed from Drew and Lahey (1993).
Flift,s = 0.5s g |
us ug | ( us ).

For the solid phase:

(A.6)

Here g is the gas viscosity, s and s are the solid shear and

Res =

(A.2)

4649

(A.12)

The granular temperature is proportional to the kinetic


energy of the random motion of the particles. The transport
equation derived from kinetic theory (Ding and Gidaspow,
1990) can be written as:


3 *
( s s ) + (s s us s )
2 *t s
= (ps I + s ) : us + (ks s )
2 )g
12(1 ess
0,ss

s 2s 3/2

s 3Kgs s ,
ds 

(A.13)

where ks is the thermal conductivity of pseudo-thermal


energy.
The solids bulk viscosity, s accounts for the resistance
of the solid particles to compression and expansion. It takes
the following form according to Lun et al. (1984),
 1/2
s
4
s = s s ds g0,ss (1 + ess )
.
(A.14)
3


4650

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

The shear viscosity of the particles is calculated from the


particle momentum exchange due to collisions (Gidaspow
et al., 1992; Syamlal et al., 1993) and translation (Gidaspow
et al., 1992). These two are added to give the solid shear
viscosity as
 1/2
s
4
s = s s ds g0,ss (1 + ess )
5


10s ds s 
+
96s (1 + ess ) g0,ss

2
4
1 + g0,ss s (1 + ess ) .
5

gs =

and
(A.15)

*
(s s ks ) + (s s Vs ks )
*t


t,s
ks + (s Gk,s s s s )
= s
k
t,g

+Kgs (2kg Csg ks ) Kgs ( V g V s )


g
g g
t,s

+Kgs ( V g V s )
s
(A.16)
s s
*
(s s s ) + (s s Vs s )
*t


t,s
s
s + (1.42s Gk,s 1.68s s s )
= s

ks
s
+1.2 (2kg Csg ks )
ks
t,g
s
g
1.2 Kgs (Vg Vs )
ks
g g
t,s
s
+1.2 Kgs (Vg Vs )
s
(A.17)
ks
s s
Here Vg and Vs are the phase-weighted velocities, and
Gk,s is the generation of turbulent kinetic energy, and is
expressed as

*us,j
*x i
(i and j represent the x, y directions, respectively)
(A.18)


Gk,s = s us,i us,j

and

t,s

k2
= 0.09s s
s

(A.20)

with

The model used also accounts for the interphase turbulence transfer. Additional transport equations for the turbulence and its effects on the respective phases are calculated.
Phase weighted averaging is used in averaging the uctuating turbulence quantities.
For each phase, two additional transport equations are
solved. For the solid phase, for example:

and Csg can be approximated as


gs
Csg = 2
1 +
gs

0.135Kgs ks

g g s (g /s + 0.5) 1 + 2 (1.8 1.35 cos2 )
(A.21)

ug us |
2 0.135|
.

1
3
ks2

For the interphase momentum transfer, the turbulent drag


terms in Eqs. (A.3) and (A.5) are modeled as follows:

Ksg (
us ug ) = Ksg ( V s V g ) + Ksg


Dg
Ds

s
g ,
0.67s
0.67g

(A.23)

where Dg and Ds are diffusivities and are calculated directly from the expressions given below (Simonin and
Viollet, 1990):


b +
gs
2
Dg = ks

gs F,gs
3
1 +
gs


b +
gs
2
2
+
(A.24)
kg b
ks F,gs ,
3
3
1 +
gs


1
g
+ 0.5
,
s


g
1
= g g Kgs
+ 0.5 ,
s

b = 1.5

F,gs

(A.24a)
(A.25)

Ds can be expressed as above with the terms whose subscripts are interchanged accordingly.
A no-slip condition is applied to both phases at the wall
surface, with wall roughness height (Hs ) of 1 105 m. A
modied wall function is used to account for roughness:


0.548up k 1/2
yp k 1/2
1
B,
=
(A.26)
ln 5.373
w /
0.42

where up and yp are the mean velocity and the distance from
the wall, respectively. Hs is the roughness height described
below.
The roughness function B is based on the formulas proposed by Cebeci and Bradshaw (1977), and its computation
depends on the roughness regime (Hs+ ).
Hs+ = 0.548

(A.19)

(A.22)

For

Hs k 1/2
,


Hs+ < 3,

B = 0,

(A.27)
(A.28)

L.Y. Lee et al. / Chemical Engineering Science 59 (2004) 4637 4651

3 < Hs+ < 70,


 +

1
Hs 2.25
+
B =
+ 0.5Hs
0.42
87.75



sin 0.4258 ln Hs+ 0.811 ,

For

For

Hs+ > 70,

B =

1
ln(1 + 0.5Hs+ ).
0.42

(A.29)
(A.30)

References
Anderson, T.B., Jackson, R., 1967. A uid mechanical description of
uidized beds. Industrial & Engineering Chemistry Fundamentals 6,
527534.
Bilirgen, H., Levy, E.K., 2001. Mixing and dispersion of particle ropes in
lean phase pneumatic conveying. Powder Technology 119, 134152.
Carpinlioglu, M.O., Gundogdu, M.Y., 1998. An experimental approach
for the determination of development length in particulate ows.
International Journal of Multiphase Flow 24 (2), 347353.
Cebeci, T., Bradshaw, P., 1977. Momentum Transfer in Boundary Layers.
Hemisphere Publishing Corporation, New York.
Dhodapkar, S.V., Klinzing, G.E., 1993. Pressure uctuations in pneumatic
conveying systems. Powder Technology 74, 179195.
Ding, J., Gidaspow, D., 1990. A bubbling uidization model using kinetic
theory of granular ow. A.I.Ch.E. Journal 36 (4), 523538.
Drew, D.A., Lahey, R.T., 1993. Particulate Two-Phase Flow. ButterworthHeinemann, Boston.
Dyakowski, T., Jeanmeure, L.F.C., Jaworski, A.J., 2000. Applications of
electrical tomography for gassolids and liquidsolids owsa review.
Powder Technology 112, 174192.
Gidaspow, D., Bezburuah, R., Ding, J., 1992. Hydrodynamics of
circulating uidized beds kinetic theory approach. In: Fluidization VII,
Proceedings of the Seventh Engineering Foundation Conference on
Fluidization, pp. 7582.
Herbreteau, C., Bouard, R., 2000. Experimental study of parameters
which inuence the energy minimum in horizontal gassolid conveying.
Powder Technology 213, 213220.
Huber, N., Sommerfeld, M., 1994. Characterization of the cross-sectional
particle concentration distribution in pneumatic conveying systems.
Powder Technology 79, 191210.
Huber, N., Sommerfeld, M., 1998. Modeling and numerical calculation of
dilute-phase pneumatic conveying in pipe systems. Powder Technology
99, 90101.
Klinzing, G.E., Marcus, R.D., Risk, F., Leung, L.S., 1997. Pneumatic
Conveying of Solids: A Theoretical and Practical Approach. second
ed. Powder Technology Series, Chapman & Hill, London.
Levy, A., Mason, D.J., 1998. The effect of a bend on the particle crosssection concentration and segregation in pneumatic conveying systems.
Powder Technology 98, 95103.
Li, X., Shen, H.H., 1995. Dilute phase pneumatic transport of granular
materials in a pipe bend. Gas-Particle Flows, ASME FED- vol. 228.
Lun, C.K.K., Savage, S.B., Jeffrey, D.J., Chepurniy, N., 1984. Kinetic
theories for granular ow: inelastic particles in Couette ow and
slightly inelastic particles in a general ow eld. Journal of Fluid
Mechanics 140, 223256.

4651

Ogawa, S., Umemura, A., Oshima, N., 1980. On the equation of fully
uidized granular materials. Journal of Applied Mathematics and
Physics 31, 483.
Plumpe, J.G., Zhu, C., Soo, S.L., 1993. Measurement of uctuations in
motion of particles in a dense gassolid suspension in vertical pipe
ow. Powder Technology 77, 209214.
Quek, T.Y., 2003. M. Eng. Dissertation. Department of Chemical and
Biomolecular Engineering, National University of Singapore.
Rao, S.M., Zhu, K.W., Wang, C.H., Sundaresan, S., 2001. Electrical
capacitance tomography measurements on the pneumatic conveying of
solids. Industrial & Engineering Chemistry Research 40, 42164226.
Rautiainen, A., Stewart, G., Poikolainen, V., Sarkomaa, P., 1999.
An experimental study of vertical pneumatic conveying. Powder
Technology 104, 139150.
Simonin, C., Viollet, P.L., 1990. Predictions of an oxygen droplet
pulverization in a compressible subsonic coowing hydrogen ow.
Numerical Methods for Multiphase Flows FED91, 6582.
Su, B.L., Zhang, Y.H., Peng, L.H., Yao, D.Y., Zhang, B.F., 2000. The
use of simultaneous iterative reconstruction technique for electrical
capacitance tomography. Chemical Engineering Journal 77, 3741.
Syamlal, M., Rogers, W., OBrien, T.J., 1993. MFIX Documentation: vol.
1, Theory Guide. National Technical Information Service, Springeld,
VA. (DOE/METC-9411004, NTIS/DE9400087).
Tsuji, Y., Morikawa, Y., 1982. LDV measurements of an airsolid twophase ow in a horizontal pipe. Journal of Fluid Mechanics 120,
385409.
Van de Wall, R.E., Soo, S.L., 1994. Measurement of particle cloud density
and velocity using laser devices. Powder Technology 81, 269278.
Van de Wall, R.E., Soo, S.L., 1998. Relative motion between phases of
a particulate suspension. Powder Technology 95, 153163.
Van den Moortel, T., Santini, R., Tadrist, L., Pantaloni, J., 1997.
Experimental study of the particle ow in a circulating uidized
bed using a Phase Doppler Particle Analyzer: a new post-processing
data algorithm. International Journal of Multiphase Flow 23 (6),
11891209.
Vasquez, S.A., Ivanov, V.A., 2000. A phase coupled method for solving
multiphase problems on unstructured meshes. In: Proceedings of
ASME FEDSM00: ASME 2000 Fluids Engineering Division Summer
Meeting, Boston, June 2000.
Weiss, J.M., Maruszewski, J.P., Smith, W.A., 1999. Implicit solution
of preconditioned NavierStokes equations using algebraic multigrid.
AIAA Journal 37 (1), 2936.
Yang, W.C., Yu, Y.H., 1966. Mechanics of uidization. Chemical
Engineering Program Symposium Series 62, 100111.
Yilmaz, A., Levy, E.K., 1998. Roping phenomena in pulverized coal
conveying lines. Powder Technology 95, 4348.
Yilmaz, A., Levy, E.K., 2001. Formation and dispersion of ropes in
pneumatic conveying. Powder Technology 114, 168185.
Zenz, F.A., 1949. Two-phase uid-solid ow. Industrial and Engineering
Chemistry 41 (12), 28012806.
Zhu, K., 2003. Pneumatic conveying of granular solids. Ph.D. Dissertation,
Department of Chemical and Environmental Engineering, National
University of Singapore.
Zhu, K., Rao, S.M., Wang, C.H., Sundaresan, S., 2003. Electrical
capacitance tomography measurements on the vertical and inclined
pneumatic conveying of granular solids. Chemical Engineering Science
58 (18), 42254245.

Das könnte Ihnen auch gefallen