6066 Ind. Eng. Chem. Res.

2007, 46, 6066-6083

Pneumatic Transport of Granular Materials in an Inclined Conveying Pipe:

Comparison of Computational Fluid Dynamics-Discrete Element Method
(CFD-DEM), Electrical Capacitance Tomography (ECT), and Particle Image
Velocimetry (PIV) Results
Yan Zhang, Eldin Wee Chuan Lim, and Chi-Hwa Wang*
Department of Chemical and Biomolecular Engineering, National UniVersity of Singapore,
4 Engineering DriVe 4, Singapore, 117576

This paper presents three flow patternssdispersed flow, reverse flow, and half-ring flowsin the post-bend
region of a 45° inclined pneumatic conveying pipe. Solid concentration and velocity distribution were measured
using electrical capacitance tomography (ECT), particle image velocimetry (PIV), and a high-speed camera.
The axial velocity of particles and solid transverse motion in the pipe cross section were obtained using a
modified cross-correlation method from the ECT data for the three flow patterns. The axial velocity profile
over the pipe cross section was compared with data obtained from the PIV system. In particular, emphasis
has been placed on the characterization of the reverse-flow pattern, where negative values of solid velocities
were observed. By analyzing the images captured with the high-speed camera, three distinct regions in the
solid phase of the reverse-flow regime may be discerned. A dense region with high solid concentration formed
next to the bottom wall of the inclined conveying pipe, while a dilute region with a relatively low solid
concentration formed near the center of the pipe. Reverse flow was observed to occur predominantly in the
dense region and a transition region between the two. Finally, the dynamic forces acting on single particles
were analyzed for the three flow patterns. This analysis showed that the phenomena of reverse-flow and
half-ring flow structure formation may be attributed to the effects of electrostatics. This finding has also been
validated with results of numerical simulations performed in the present study.

1. Introduction layer were imaged using electrical capacitance tomography

Pneumatic conveying systems are widely used for the
transportation of granular material in various industries. Re-
search work performed in this field can be categorized according
to different geometries of the conveying section used in the
transportation process. These include pneumatic conveying
through horizontal, vertical, and inclined pipes and about a bend.
Many studies have been conducted on gas-solid flow in
horizontal and vertical pipes, whereas relatively fewer studies
are available on solid transport in inclined pipes. These are
mainly focused on pressure drop data through the conveying
pipes. For example, Levy et al.1 developed an analytical model
for gas-solid suspension flow through an inclined pipe section,
which predicted the ratio of the total pressure drop in an inclined
pipe to that of a horizontal pipe and agreed with the experimental
data satisfactorily; Hirota et al.2 examined the effect of me-
chanical properties and the angle of an inclined pipe on the
pressure drop for the pneumatic conveying of fine powders.
Other studies of granular flow focused on the different flow
patterns arising from different physical conditions and operating
parameters imposed. Zhu et al.3 used computational fluid
dynamics (CFD) simulations to investigate the pneumatic
conveying of granular solids through an inclined pipe at different
inclinations and reported results on the influence of model
parameters, inclination angle, and feeding conditions on the flow
patterns. Experiments have also been performed to study solid
flow patterns in pneumatic conveying through vertical and
inclined pipes and various flow patterns were observed.4 In their
work, regimes such as eroding dune flow and flow over a settled
* To whom correspondence should be addressed. Tel.: 65-6516-
5079. Fax: 65-6779-1936. E-mail: chewch@nus.edu.sg.
10.1021/ie061304i CCC: $37.00 © 2007 American Chemical Society
Published on Web 05/11/2007
(ECT), and some flow regimes with downflow were observed
visually. However, detailed analyses were not given. The study
on downflow or reverse flow has not drawn much attention
among researchers. Among the limited literature on this subject,
most have concentrated on heat-transfer investigations in liquid
systems. Maré et al.5 investigated the laminar flow of water in
an inclined isothermal tube numerically and experimentally
using three-dimensional elliptical model and particle image
velocimetry (PIV), respectively, and concluded that the com-
bination of the Reynolds number (Re), Grashof number, and
Prandtl number determined the phenomena of reverse flow. The
objective of this work was to study three flow patternss
dispersed, reverse, and half-ring flowsobserved in a 45° inclined
conveying pipe. The methods applied for capturing solid flow
patterns and velocity profiles include ECT, PIV, and a high-
speed camera. The study on flow patterns is a fundamental
concept in pneumatic conveying systems and is helpful toward
understanding real-life problems, which may arise during actual
operations. The appearance of the variety of solid flow patterns
observed in laboratory experiments or actual industrial opera-
tions are normally attributed to the physical conditions used.
However, some peculiar clustering behaviors may not be
accounted for purely by hydrodynamics effects alone, such as
the formation of ring structures in vertical and inclined
pneumatic conveying. This paper attempts to analyze such flow
behaviors of solid particles in an inclined pneumatic conveying
pipe through the consideration of the effects of electrostatic
charge generation. This continues from previous research
studies6 on the electrostatics of granular flow in horizontal and
vertical pneumatic conveying systems and provides further
 Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007 6067

Figure 1. Schematic of the pneumatic conveying experiment facility. Legend: 1, air control valve; 2, dryers; 3, rotameter; 4, hopper; 5, solids feed valve;
6, rotary valve feeder; 7, feed control valve; 8, computer; 9, DAM; 10, ECT plane 1; 11, ECT plane 2; 12, plane of PIV measurements; 13, plane of
measurements for the high-speed camera; 14, measurements for induced current; 15, measurements for particle charge; 16, pressure transducer sensor 1; and
17, pressure transducer sensor 2.

Table 1. Experimental Conditions pipes: one is positioned at a distance of 3.41 m, and the lower
parameter value one is positioned at a distance of 4.12 m. A 45° inclined pipe
air flow rate 1600 L/min, 1100 L/min, 1000 L/min
(4-cm ID) was connected to the two horizontal pipes. The entire
solids feed valve 75% opening conveying loop was made of poly(vinyl chloride) (PVC) pipes.
roll speed of rotary valve 25 rpm 2.2. Electrical Capacitance Tomography (ECT). ECT is a
pipe material poly(vinyl chloride), PVC noninvasive technique for measuring and displaying the con-
pipe diameter (inner) 40.0 mm centration distribution of a mixture of two dielectric fluids. In
pipe thickness 5.0 mm
particles conveying style cycle this experiment, two sets of 12-electrode (10 cm long) ECT
particle material polypropylene, PP sensors (labeled as “10” and “11” in Figure 1) were arranged
particle size 2.80 mm on the inclined pipe at distances of 1.45 and 2.18 m from a
particle density 1123 kg/m3 bend connecting this inclined riser to a horizontal duct at the
bottom. A data acquisition module (DAM) (Process Tomogra-
insights into the relationship between solids flow behavior and phy, Wilmslow, Cheshire, U.K.) was connected to the ECT
electrostatic effects. sensors to gather capacitance data. The ECT was sampled at
40 Hz, corresponding to a time period of 0.025 s for each frame.
The resolution of each frame was 1024 (32 × 32) pixels. The
2. Experimental Section
ECT system was calibrated for the lower and upper permittivity
2.1. Experimental Setup. The experimental facility, which bounds: air was used for the lower bound (with 0 as the
was modified from a previous study4,7 and used to perform the reference value) and solids were used as the upper bound (with
measurements, is shown schematically in Figure 1; the experi- 1 as the reference value).
mental conditions are listed in Table 1. Air from the compressor By post-processing the ECT data at each instant of time t,
was metered and sent to the rotary air lock feeder (General using the simultaneous iterative reconstruction technique (SIRT)
Resources Corp., Hopkins, MN), where it entrained the particles. described by Su et al.,8 the particle concentration in each pixel
The rotary feeder equipped with eight pockets and a vent at the of one plane, R(x,y,z,t), was obtained. Here, as shown in Figure
body of the rotary valve, providing the passage of the air 1 by the viewing section A-A, x and y denote Cartesian
leakage, was operated at a fixed speed of 25 rpm. The test coordinates in the cross-sectional plane, made dimensionless
section consisted of two 4-cm-inner-diameter (ID) horizontal using the pipe diameter as the characteristic length, and z is the
6068 Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007
axial coordinate denoting the location of the ECT electrodes.
The solid velocity can be calculated based on cross-correlation
of twin plane data, which was described by Hua et al.,9 Rao et
al.,7 and Zhu et al.4 The time-averaged particle concentration
R
j t was found by averaging R(x,y,z,t) over a time period T (30
s in this case):
R
j t(x,y,z) )
1
T
∫0T R(x,y,z,t) dt
The instantaneous value of the cross-sectional average particle
concentration, R
j s(z,t), is defined as
j s(z,t) )
R
1
A
∫∫ R(x,y,z,t) dx dy
j s(z,t) is denoted by 〈R〉(z):
and the time-averaged value of R
〈R〉(z) )
1
T
∫0T Rjs(z,t) dt ≡ A1 ∫∫ Rj t(x,y,z) dx dy
The correlation coefficient, C(d), was then computed as
C(d) )
1
T
∫0 T
j s(z1,t) - 〈R〉(z1)) (R
(R j s(z2,t + d) - 〈R〉(z2)) dt
Here, z1 and z2 refer to upstream and downstream planes,
respectively; d denotes the delay time. The dominant pattern
propagation velocity (V*) was estimated from V* ) L/D,
where L ) z2 - z1 (L ) 2.18-1.45 ) 0.73 m in the current
work), is the axial distance between the two ECT sensors and
D is the value of d at which C(d) assumes the largest value. V*
estimated in this manner is clearly based on cross-sectional
averages. One can also define pixel-pixel cross-correlation
function as
c(x,y,d) )
1
T
∫0T (R(x,y,z1,t) -
R
j t(x,y,z1)) (R(x,y,z2,t + d) - R
j t(x,y,z2)) dt (5)
The dominant pattern propagation velocity (V(x,y)) was esti-
mated from eq 6:
L 3.1. Discrete Element Method. The translational and rota-
V(x,y) )
D(x,y)
where D(x,y) is the value of d at which c(x,y,d) assumes the
maximum value (for that x and y).
One limitation of the classical cross-correlation method arises
from a basic assumption that the particle motion between the
sensors is parallel to the pipe axis, i.e., perpendicular to the
sensor plane. However, the actual movements are quite complex,
and the trajectories of particles do not remain parallel to the
pipe axis. Thus, it is necessary to introduce a new method,10
namely, the best-correlated pixel method, which does not require
the assumption about the solid trajectories within a sensor
volume and will be shown in section 4.1.1.
2.3. Particle Image Velocimetry (PIV). PIV is generally
applied to exhibit the magnitudes and vectors of particle
velocities and afford exact values of velocities. Measurements
using a PowerView 2D PIV system (TSI, Shoreview, MN) were
conducted at the inclined pipe region 1.83 m downstream from
the bend. The laser light sheet generated by the Laser Pulse
Solo Mini Dual Nd:YAG laser was introduced from the top of
the wall/sidewall to lighten the granules at the center plane
(labeled as “12” in Figure 1). Images of the granules were
captured by a PowerView 4M (2K × 2K) camera in the
direction perpendicular to the lightsheet and transmitted to the
computer for processing. The cross-correlation yields the
distance traveled by the granules in a small time interval (dT )
50-58000 µs) from the first snapshot to the second and the
velocity distribution of granules in the plane can be determined
by dividing by the time interval. In this study, the variation of
velocity in the axial direction of the 45° inclined pipe was
(1) measured.
2.4. Electrostatic Measurements. During the pneumatic
conveying process, frictional contacts between the solid particles
and pipe wall generate electrostatic charges. With proper pipe
connections, current induced on the surface of the pipe wall
can be measured as a function of time using a digital
(2) electrometer (ADVANTEST R8252, Tokyo, Japan). The mea-
surement section is labeled “14” in Figure 1 while details of
the measurement methods have been described previously by
Yao et al.6 In addition, the particle charge density was measured
(3) using a Faraday cage at the horizontal segment6 (labeled “15”
in Figure 1).
3. Numerical Method
The method of combining the discrete element method (DEM)
(4) with CFD is well-established in the literature for numerical
studies of various types of solid-fluid systems. Recently, it has
also been applied successfully to reproduce the various flow
regimes observed in pneumatic conveying of granular materials
through vertical and horizontal pipes.11 Lim and co-workers12
have also described, in detail, a method of incorporating an
electrostatic force model into the CFD-DEM model for
numerical studies of pneumatic conveying processes with
electrostatic effects. It was shown that the eroding dunes regime
observed in previous experimental studies using inclined
pneumatic conveying pipes could be reproduced computation-
ally. Reversed flow of particles was observed in a dense region
close to the bottom wall of the conveying pipe and forward
flow in a more dilute region in the space above. The same
methodology has been applied in the present study to comple-
ment the experimental investigations conducted. A brief outline
of the main component of the CFD-DEM model is provided
in the following section.
(6) tional motions of individual solid particles are governed by
Newton’s laws of motion:
dvi N
mi )
dt j)1
∑
(fc,ij + fd,ij) + mig + ff,i + fE,i (7)
dωi N
Ii
dt
) ∑
j)1
Tij (8)
where mi and Vi are the mass and velocity of particle i,
respectively; N is the number of particles in contact with this
particle; fc,ij and fd,ij are the contact and viscous contact damping
forces, respectively; ff,i is the fluid drag force due to an inter-
stitial fluid; fE,i is the electrostatic force; Ii is the moment of
inertia of particle i; ωi is its angular velocity; and Tij is the torque
arising from contact forces, which will cause the particle to
rotate.
Contact and viscous contact damping forces must be calcu-
lated using force-displacement models, which relate such forces
to the relative positions, velocities, and angular velocities of
the colliding particles. In the present work, the linear force-
displacement model was implemented for the calculation of

these collision forces. The normal (fcn,ij, fdn,ij) and tangential (fct,ij,
fdt,ij) components of the contact and damping forces are
calculated according to the following equations:
fcn,ij ) -κn,iδn,ij
fct,ij ) -κt,iδt,ij
fdn,ij ) -ηn,i (Vr‚ni)ni
fdt,ij ) -ηt,i (Vr‚ti)ti + (ωi × Ri - ωj × Rj)
where fcn,ij, fdn,ij and fct,ij, fdt,ij are the normal and tangential
components of the contact and viscous contact damping forces,
respectively; κn,i, δn,ij, ni, ηn,i and κt,i, δt,ij, ti, ηt,i are the spring
constants, displacements between particles, unit vectors, and
viscous contact damping coefficients in the normal and tan-
gential directions, respectively; Vr is the relative velocity between
particles; and Ri and Rj are the radius vector (from particle center
to a contact point) for particles i and j, respectively. If |fct,ij| >
|fcn,ij| tan φ, then “slippage” between the two contacting surfaces
is simulated by a Coulomb-type friction law,
|fct,ij| ) |fcn,ij| tan φ
where tan φ is analogous to the coefficient of friction.
3.2. Fluid Drag Force. In a multiphase system such as the
gas-solid pneumatic conveying system considered in this study,
interactions between the two phases take the form of fluid drag
forces on the solid particles exerted by the interstitial fluid and
arise from velocity differences between the two phases. In this
study, the model due to Di Felice,13 which is applicable over a
wide range of particle Reynolds numbers, was used to evaluate
the fluid drag force:
ff,i ) ff0,i-(χ+1)
i (13)
ff0,i ) 0.5cd0,iFf πRi2i2|ui - Vi|(ui - Vi)
[ ]
(1.5 - log10 Rep,i)2
χ ) 3.7 - 0.65 exp -
2 between particles i and j, and ni the unit normal vector in the
( )
4.8 2
cd0,i ) 0.63 +
Rep,i0.5
2Ff Rii|ui - Vi|
Rep,i )
µf
where ff0,i is the fluid drag force on particle i in the absence of
other particles, χ is an empirical parameter, i is the local average
porosity in the vicinity of particle i, cd0,i is the drag coefficient,
Rep,i is the Reynolds number based on particle diameter, Ff is
the fluid density, µf is the fluid viscosity, and ui is the fluid
velocity of the computational cell in which particle i is located.
3.3. Computational Fluid Dynamics. The motion of the
continuum gas phase is governed by the Navier-Stokes
equations with an additional source term in the momentum
equation to represent the reaction force acting on the fluid by
the particles:
∂
+ ∇‚(u) ) 0
∂t
∂(Ff u)
+ ∇‚(Ff uu) ) -∇P + ∇‚(µf ∇u) + Ff g - F
∂t
(19)
Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007 6069
where u is the velocity vector,  the local average porosity, P
the fluid pressure, and F the source term due to fluid-particle
interaction.
The computational domain was divided into uniform grid
(9) cells, and all quantities such as velocities and pressure were
assumed constant over each cell. The source term F for a
(10) particular computational cell was calculated by summing the
(11) fluid drag forces on all particles present within that cell:
(12) n
∑
i)1
ff,i
F) (20)
∆V
where ∆V is the volume of a computational cell and n is the
number of particles present in the cell.
3.4. Electrostatic Effects. During pneumatic conveying, solid
particles gain electrostatic charges as a result of repeated
collisions and impacts against other particles and with the walls
of the conveying pipe. The total electrostatic force acting on
each particle may then be written as the sum of electrostatic
forces due to charges carried by other particles and the pipe
walls:
fE,i ) fEp,i + fEw,i (21)
where fEp,i and fEw,i are the electrostatic forces due to other
charged particles and the pipe walls acting on particle i,
respectively.
The electrostatic force arising from charges carried by other
particles may be calculated by assuming each particle to be a
constant point charge:
N
Q2
fEp,i ) ∑ ni (22)
j)1 4πorij2
j*i
(14)
where Q is the constant charge assumed to be carried by all
particles, o the permittivity of free space, rij the distance
(15)
direction of the line joining the two particle centers.
(16) In the present work, a dimensionless quantity Λ depicting
the ratio of the electrostatic force arising from the charged pipe
wall to the gravitational force exerted on each particle, fEw,i/
(17) mig, is defined. In other words,
fEw,i ) Λmi g (23)
4. Results and Discussion
4.1. Flow Patterns and Velocities for Particle Transport
in a 45° Inclined Conveying Pipe. Depending on the different
air flow rates, three distinct phenomenasnamely, dispersed flow,
reverse flow, and half-ring flowsfor particle transport in a 45°
inclined conveying pipe was observed and schematically il-
lustrated in Figures 2a, 2b, and 2c, respectively. When the air
flow rate is high, particle flow is dilute and particles move in
the direction of gas flow as shown in Figure 2a, the dispersed
flow pattern. When air flow rate is decreased, some particles
near the pipe center moved forward, while most of the particles
(18) formed layers at the bottom of the pipe and slid downward,
which is called reverse flow (shown in Figure 2b). When the
air flow rate was further decreased, particles formed a layer on
the pipe wall, which resulted in an annular structure. This flow
pattern is called half-ring flow (shown in Figure 2c). An
6070 Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007
Figure 2. Schematic of particle flow in three different flow patterns in pneumatic conveying: (a) dispersed flow; (b) reverse flow; and (c) half-ring flow,
Panel (d) shows unsteady reverse flow with pulsating wave.
interesting unsteady flow patternsreverse flow with pulsating
wave on the surfacesis also discussed in section 4.2.4.
4.1.1. Dispersed Flow Pattern. Figure 3 shows the images
of particles concentration captured by ECT on plane 1 and plane
2, respectively. When the air flow rate is 1600 L/min, the particle
concentration is generally low and particles do not concentrate
on the pipe wall. As such, it forms the dispersed flow pattern.
Figure 4 shows the result of velocity calculated by the best-
correlated pixel method proposed by Mosorov and colleagues.10
The Mosorov method is modified from the classical cross-
correlation method (section 2.2) but is not limited by the basic
assumption of the latter that the solid’s motion between the
sensors is parallel to the pipe axis and perpendicular to the sensor
plane. This method is to calculate cross-correlation between a
pixel from the first plane and a number of pixels from the second
plane. Therefore, the essential difference between the classical
cross-correlation method and the best-correlated pixel method
is the choice of pixels in the second plane. In the former, only
corresponding pixels are correlated; however, in the latter, the
pixels from the second plane are chosen from the corresponding
pixel and those in its neighbors (not just the nearest neighbors).
Equation 24 describes the function in detail:
c(x,y,i,j,d) )
1
T
∫0T (R(x,y,z1,t) - Rt(x,y,z1))(R(i,j,z2,t + d) -
Rt(i,j,z2)) dt (24)
where (i,j) are dimensionless coordinates of the pixel in plane
2, R(x,y,z1,t) and R(i,j,z2,t + d) are the values associated with
pixel (x,y) from the image at time t obtained from plane 1 and
pixel (i,j) from the image at time (t + d) obtained separately
from plane 2. Therefore, similar to the classical cross-correlation
method, when c(x,y,i,j,d) reaches its maximum, d is D(x,y) and,
according to eq 6, the axial solid velocity of each pixel, Vz-
(x,y), can be obtained. It is observed that velocities are high
over the entire pipe (see Figure 4).
Furthermore, the coordinates (i, j) on plane 2 (nondimen-
sionalized by the pipe diameter as the characteristic length scale)

Figure 3. Images of particle concentration (R) at twin planes captured by ECT for the dispersed-flow pattern (air flow rate ) 1600 L/min).
Figure 4. Particle axial velocity on pipe cross-section correlated from ECT
data for the dispersed-flow pattern (air flow rate ) 1600 L/min).
can be obtained, according to the best-correlated pixel method.
If i ) x, j ) y, it shows that the particles maintain their direc-
tion during the transport from plane 1 to plane 2; otherwise, it
reflects the direction change of particles. Therefore, in such case,
the actual velocity for particle transport between planes 1 and
plane 2 is
()
Vx
B
V ) Vy
Vz
Here, Vx, Vy, and Vz are the magnitudes of the velocity
components in the three perpendicular directions, respectively.
Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007 6071
The lateral velocity can be defined as the projection of B
V on
the pipe cross section and is described as
2Rx(x - i)2 + (y - j)2
Vl(x,y) ) (26)
D(x,y)
where R is the pipe radius. The average lateral velocity for the
dispersed flow pattern was calculated to be ∼8 × 10-4 m/s and
appeared negligible, as compared to the axial velocity, as will
be shown in Figure 6.
Figure 5 shows the images of particle transport captured by
a high-speed video camera (FastCam PCI, Photron, CA) at a
frame rate of 250 fps. The motion of particles through the section
of the conveying pipe could be visualized in slow motion after
all the required frames were captured with the high-speed video
camera. During this process, each particle could be tracked
unambiguously from one frame to another, because the average
distance traveled by each particle in the time interval between
two consecutive frames was less than the diameter of the
particle. This also allowed a specific particle to be identified in
any two frames that are not necessarily consecutive to each
other, such as those presented in Figure 5. The inclined pipes
are positioned with the pipe axis at an angle of 45° relative to
the direction of gravity, as indicated in Figure 5. The direction
of air flow is from right to left. Similar illustration is also applied
later to Figures 7 and 11. In Figure 5a, most particles are
transported in regions close to the bottom pipe wall and the
overall space among particles is fairly large, thus resulting in a
dilute solid phase. For illustration purpose, one particle was
(25) chosen as an example and traced over several images. The
positions of this particle in two images are marked in circles,
as shown in Figure 5a. Generally, particle motion is observed
to follow the gas flow direction. Subsequently, 10 particles were
chosen randomly from 1000 pictures over a period of 4 s and
6072 Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007

Figure 5. Images of particle transport captured by a high-speed video camera: (a) dispersed flow (time interval between two successive pictures ) 0.008
s, air flow rate ) 1600 L/min); (b) reverse flow (time interval between two successive pictures ) 0.072 s, air flow rate ) 1100 L/min); (c) half-ring flow
(time interval between two successive pictures ) 0.096 s, air flow rate ) 1000 L/min); (d) reverse flow with pulsating wave (time interval between two
successive pictures ) 0.176 s, air flow rate ) 1100 L/min, PP with an antistatic agent).

the average velocity of particles was computed as the distance
between the two positions in selected images divided by the
duration between these two images. The average particle velocity
value obtained was 1.59 ( 0.41 m/s.
Figure 6 displays the velocity profile of particles at different
positions along the pipe diameter. The data were obtained
from ECT and PIV experiments, respectively. The figure shows
that velocity profiles obtained from these two different meth-
odologies exhibited similar trends. However, there are some
quantitative differences, probably because of the different
extents of experimental errors associated with the two instru-
ments. Here, velocities are normalized with respect to the
maximum value of velocity observed within the pipe cross
section. The normalized velocities are high near the pipe center
but decrease toward the pipe wall. Such profile was observed
in both side view (Figure 6a) and top view (Figure 6b). This
was also predicted by CFD simulation of pneumatic conveying
of granular solids in an inclined pipe.3 It may also be observed
from the velocity profiles in both side and top views in Figure
6 that lateral velocities are much lower than those of axial
velocities, implying that axial movement is primary for such
flow pattern, thereby validating the results obtained by eq 26
from the ECT data.
Figure 7 shows the side view image from the top to the
bottom of the pipe of the dispersed flow pattern captured by
PIV, where the vector arrows verify the observation that the
particle flow directions are consistent with the air flow.
4.1.2. Reverse Flow Pattern. When the air flow rate is
decreased to 1100 L/min, it is observed in the experiment that
some particles near or above the pipe center moved forward
with fairly high velocities, whereas most particles were deposited
and formed layers at the bottom of the pipe. These slid backward
in a direction opposite to that of the general flow direction at
relatively slow velocities. These are represented by the red areas
in Figure 8.
Figure 9 presents the resulting velocity obtained with the best-
correlated pixel method for the reverse-flow pattern. In this
figure, the entire solid phase is divided into three parts. It is
observed that velocities are relatively high and positive near
and above the center of the pipe, where the corresponding
coordinate along the y-axis are 0.35-1, whereas those at the
bottom (y ) 0-0.18) are very small and negative. These two
areas are defined as dilute and dense regions, respectively.
Particles at the interface between these regions (referenced as
the transition region with y ) 0.18-0.35) flowed backward with
relatively higher negative velocities. The lateral velocities
calculated from eq 26 also show three distinct velocity scales,
increasing from 10-6 m/s, to 10-5 m/s, to 10-4 m/s, from the
dense region to the dilute region, respectively.
The limitation of the cross-correlation method is that it is
difficult to get exact values when data are the same on two
correlation planes, because the principle of correlation is based
on variation of the data on two planes. For the dense region,
the concentrations of particles measured by two ECT sensors

Figure 6. Diametrical distribution of particle axial velocity for the
dispersed-flow pattern (air flow rate ) 1600 L/min): (a) side view and (b)
top view.
are at the maximum values continuously. As a result, the
velocities in this section are obtained from the images captured
by a high-speed video camera, as shown in Figure 5b. It can be
observed that most of granules are accumulated at the bottom
of the pipe; however, a small amount is transported in the core
region and at the top of the pipe. This is consistent with the
solid fraction profile shown in Figure 8. In the dense region,
particles at the bottom of the pipe seem almost stationary and
the velocity obtained is only about -0.0020 ( 0.00072 m/s.
The path traced by the particle is not shown, because it is not
easily interpolated between the two images. In the transition
region, some particles close to the core of the pipe are active
and are referenced as transition particles. This is probably due
to the shearing effects between the dilute and dense regions.
The locations of the particle labeled with the symbol “A” in
the three images indicate that it did not move ahead. Instead, it
seemed to have jumped over some particles in a direction
opposite to the air flow. The average velocity is -0.05 ( 0.02
m/s. However, in the dilute region, particles such as that
labeled as “B”, moved ahead, with an average velocity of
0.62 ( 0.26 m/s.
Correspondingly, the normalized particle velocity profiles
obtained from both ECT and PIV and presented in Figure 10
demonstrate that three regions exist in the solid phase. Positive
velocities in the dilute region are generally forward flow. In
contrast, in the dense and transition region, negative velocities
indicate reverse flow. In particular, the change in velocity profile
Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007 6073
Figure 7. Velocity vectors for the dispersed-flow pattern by PIV measure-
ments (air flow rate ) 1600 L/min).
within the transition region revealed that reverse flow is
significant in this zone. Figure 10 also shows that lateral
velocities are much weaker in the dilute region, as compared
to axial velocities, and, hence, can be neglected. However, in
the transition and dense regions, the magnitudes of lateral and
axial velocities are observed to be comparable. These findings
further verify the three levels of lateral velocities observed from
ECT data.
Arrows in the vector plot of Figure 11 are pointing toward
two completely opposite directions. This finding demonstrates
that some particles are floating in the dilute region and moving
forward, but some particles, especially those in the transition
region, exhibit a backflow behavior. From the magnified zone,
particles in the transition and dense regions also resulted in some
wave structures in the transverse direction.
4.1.3. Half-ring Flow Pattern. When the air flow rate is
further reduced to 1000 L/min, particles deposit at the bottom
of the pipe and are stagnant initially. Later, particles concentrated
on the pipe wall and gradually formed an annulus structure,
while those at the bottom continued to remain stagnant, as shown
in Figure 12. The thickness of this ring is roughly one layer of
particles, and the solid concentration at the top of the pipe seems
to be smaller than that at the bottom, because of the action of
gravity.
Figure 13 shows that velocity is zero at the bottom and
relatively high at the top of the pipe wall, presenting an
incomplete ring structure in the pipe cross-section. Correspond-
ingly, the lateral velocities calculated according to eq 26 for
the dense and dilute regions have magnitudes of 10-4 and 10-3
m/s, respectively. This shows that lateral motion is not negligible
in the dilute region, while transverse motion becomes dominant
in the dense region.
6074 Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007
Figure 8. Images of particle concentration (R) at twin planes captured by ECT for the reverse flow (air flow rate ) 1100 L/min).
Figure 9. Particle axial velocity on pipe cross-section correlated from ECT
data for the reverse-flow pattern (air flow rate ) 1100 L/min).
Figure 5c shows that there is half-ring formation around the
pipe wall. The ring of particles seemed to be suspended in the
air or moving with lateral velocities. From the interspaces among
the particles adhered on the pipe, it is observed that a small
quantity of particles in the pipe core still moved ahead in the
dilute region, and the average velocity is estimated to be 0.23
( 0.06 m/s, using the previous method (particle B only moved
a short distance during a time interval of 0.096 s). On the other
hand, in the dense region, some particles settled on the bottom
of the pipe (for instance, particle A appears almost motionless
in the three frames).
Because of the peculiar nature of the half-ring flow pattern,
normalized particle velocity distributions near the pipe surface
and in the core region of the pipe are presented separately in
Figure 14. This is shown in the sketch map of the pipe cross-
section (Figure 2c), where both types of data are obtained from
the correlation result from ECT measurements. From the surface
velocity profile, it may be seen that the corresponding calculated
velocity data points fall on the zero axes, demonstrating the
stagnant ring flow, except that some positive values near the
top indicate a small amount of forward flow in this area. As
for the inner core region, combined with the measurement from
high-speed video camera, the velocities at the bottom of the
pipe are zero, because of the stagnant ring structure in the dense
region. Some peak values indicated forward motion of a few
particles in the dilute core region. Furthermore, a hollow core
area around the center of the pipe is observed, emphasizing the
existence of the half-ring structure.
However, it may be noted at this point that velocity
measurements using PIV is difficult for this particular type of
flow pattern, because of blockage of the laser sheet from the
system by particles adhered on the pipe walls. Because of the
relatively large size of the particles used throughout this study,
images of particles could be captured by the PIV camera, as
long as the particles were slightly illuminated by the laser. In
the dispersed- and reverse-flow regimes, the laser from the PIV
systems was able to penetrate through the relatively dilute layer
of solid particles near the pipe walls to illuminate those near
the center of the conveying pipe. As such, particle velocities
near the pipe center could be obtained. On the other hand, the
solid concentration near the pipe walls was very high in the
ring-flow regime and the PIV laser was completely blocked by
this layer of particles near the walls. Thus, the particle velocity
profile was unavailable for this flow regime.
To validate the reliability of the various methods applied for
velocity measurements in this study, the values of axial velocity
for the three flow patterns measured by the three different

Figure 10. Diametrical distribution of particle axial velocity for the reverse flow (air flow rate ) 1100 L/min).
Figure 11. Velocity vectors for the reverse-flow pattern from PIV
measurements (air flow rate ) 1100 L/min).
methods are listed in Table 2. The velocities in the dense regions
are fairly similar; however, in the dilute regions, data obtained
via ECT and PIV seemed to contain a larger amount of noise
and, thus, a wider range of values may be observed.
4.2. Electrostatic and Dynamic Analysis for Three Flow
Patterns. Generally, electrostatic effects are responsible for
Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007 6075
many anomalous behaviors in granular flow systems. It is
believed that the formation of the various flow patterns described
previously in pneumatic conveying through inclined pipes is
related to electrostatic charge generation resulting from friction,
collision, or rolling between particles and the pipe walls during
the conveying process. To investigate the relationship between
electrostatic effects and flow behaviors, the electrostatic charge
generation characteristics for each of the aforementioned flow
patterns will be discussed in the following sections.
4.2.1. Electrostatic Characters for Three Flow Patterns.
In a previous work,6 electrostatic characteristics for three flow
patterns have been studied by measuring the induced current
on the pipe wall using an electrometer, when granular polypro-
pylene (PP) were transported in horizontal and vertical pneu-
matic conveying pipes (made of PVC). Accordingly, the charge
accumulated on the pipe wall can be calculated from eq 27:
q) ∫0T I dt (27)
where I is the induced current measured from the electrometer
and T is the time period.
Recently, Yao et al.14 proposed the concepts of electrostatic
equilibrium and averaged current:
hI )
1
T
∫0T I dt (28)
In their opinion, the averaged current fluctuated before reaching
a constant value over a period of time (namely, the charging
time, Tc). The periods where fluctuations occur and a steady
value is reached are called the charging and electrostatic
equilibrium states, respectively. After the state of electrostatic
equilibrium is attained, the amount of charge residing on the
particles and the pipe wall would remain more or less constant
with time. The procedure of exploring whether the average
current achieves electrostatic equilibrium has been specified in
a previous study by Yao et al.14 An equilibrium state was
deemed to have been attained when the ratio of variation in
6076 Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007
Figure 12. Images of particle concentration (R) at twin planes captured by ECT for the half-ring flow (air flow rate ) 1000 L/min).
Figure 13. Particle axial velocity on pipe cross-section correlated from
ECT data for the half-ring-flow pattern (air flow rate ) 1000 L/min).
average current to the average current at the final state was
<10%. A fixed ratio is used in the present work to facilitate
comparisons among cases corresponding to different experi-
mental conditions. The charging process for three flow patterns
is described in Figure 15, where significant fluctuations in the
current were observed until the charging time Tc was reached.
In the earlier experiment, a sensitivity analysis of the definition
of Tc on the actual state of electrostatic equilibrium has been
investigated qualitatively, to verify its validity. The same
observation, with regard to current fluctuations, was obtained
when the ratio was chosen to be >90%, thus verifying the
adequacy of using Tc as an indication of the establishment of a
state of electrostatic equilibrium. Therefore, the amount of
charge accumulated on the pipe wall when equilibrium is
reached are calculated for each of the three flow patterns,
according to eq 27, by substituting T with the charging time
Tc, and the results are listed in Table 3.
4.2.2. Simplified Electrostatic Field. A simplified electro-
static field12 is selected to describe a two-dimensional transverse
section of pipe wall in accord with the numerical work and to
perform the electrostatic analysis in this section. In comparison
with the axial characteristic length scale of the pipe, the particle
is sufficiently small such that charges on the pipe wall may be
taken to be uniformly distributed. The charges on the wall are
assumed to be infinitesimal point charges, dq, and every particle
is taken as a point charge Q (assumption 1). By integrating the
point charges over the entire wall, the electric field intensity
due to charged pipe wall may be represented as follows:
E) ∫ 4π
1 dq
0 r
2
)
λ
∫π/2 cos θ dθ ) 2πλ 0l
2π0l 0
(29)
Here, 0 is the permittivity constant (8.85 × 10-12 C2/(N m2)
under vacuum); q is the total equilibrium charge on pipe wall
obtained from integration of the induced current, by eq 27; λ is
the linear charge density along pipe wall, which is defined as
the equilibrium charge on the pipe divided by the length of
induced current measurement; r is the distance from a point
charge on the pipe (dq) to a point charge (i.e., a charged particle)
(Q); l is the vertical distance from the pipe wall to a point charge
(i.e., a charged particle); and θ is the angle between r and l.
4.2.3. Dynamic Analysis for Single Particle on Pipe Wall.
In the aforementioned model for dynamic force analysis,
electrostatic interactions between particles were not taken into
consideration (assumption 2). This assumption has been verified
in a previous experimental study,14 where the electrostatic forces
exerted on each particle by other particles due to the pipe wall
were determined to be <0.1% and could be neglected. A CFD-
 Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007 6077

Figure 14. Diametrical distribution of particle axial velocity for the half-ring-flow pattern (air flow rate ) 1000 L/min).

Table 2. Comparison of Axial Particle Velocities by Different Experimental Methods

Particle Velocity (m/s)
flow air flow rate flow
pattern (L/min) region ECT PIV high-speed camera
disperse 1600 average 3.3 × 10-1-1.6 × 100 1.1 × 100-3.9 × 100 1.2 × 100-2.0 × 100
reverse 1100 dilute 0.7 × 10-1-3.9 × 10-1 3.8 × 10-1-9.0 × 10-1 3.6 × 10-1-8.8 × 10-1
reverse 1100 transition (-8.3 × 10-2)-0.9 × 10-2 (-3.4 × 10-2)-(-1.2 × 10-2) (-7.7 × 10-2)-(-3.1 × 10-2)
reverse 1100 dense (-22 × 10-3)-(-5.6 × 10-3) (-4.4 × 10-3)-0.6 × 10-3 (-2.7 × 10-3)-(-1.3 × 10-3)
half-ring 1000 dilute 0.4 × 10-1∼2.6 × 10-1 N.A.a 1.7 × 10-1-2.9 × 10-1
half-ring 1000 dense 0 N.A.a 0
a Velocity measured by PIV is not applicable, because laser sheet is blocked by the annular structure of particles on the pipe wall.

DEM work12 also predicted that the average electrostatic force where µg is the gas viscosity, and b
ug and b
up are the translational
acting on each particle in similar systems was only 10-9 N and, velocity vectors of the gas and particle, respectively.
thus, confirmed that electrostatic effects due to charged particles (ii) Electrostatic force is assumed to be mainly contributed
are negligible, as compared with other types of forces. Evidence by the lower pipe wall in this simplified model, FE:
for this claim is that particle agglomeration is seldom observed
in the present operation. On the contrary, clustering is prevalent FE ) EQ (34)
for fine powders. This might be because of the fact that the
electrostatic interactions among fine particles are significant, where Q is the electrical charge on each particle and can be
because of the small size and weight. obtained from multiplying the charge-mass ratio (charge den-
Summarizing the aforementioned discussion, the major forces sity)6 by the mass of a single particle.
that act on each particle therefore include the following (iii) Friction between particle and pipe wall: the product of
components (besides gravity): normal force and coefficient of friction, which is estimated to
(i) Aerodynamic drag force, B FD:15 be ∼0.56 from an internal angle of friction experiment.16
Table 3 displays the comparison of forces on a single particle
πFgDp2 for three flow patterns, where the particle velocity shown is
B
FD ) bg - b
cD(u bg - b
u p)|u u p| (30) the averaged value obtained from a high-speed video camera.
8 For the dispersed-flow pattern, the magnitude of the electrostatic
where Fg is the gas density, Dp the particle diameter, and cD force, on average, is only 10-6 N and, thus, can be neglected;
the drag coefficient: whereas the calculated aerodynamic drag force is more than 4
times and 300 times that of gravity and electrostatic force,

cD )
24
Rep
(1 + 0.15Rep0.687) )
24
f
Rep D( ) (for Rep e 1000)
(31)
respectively, showing that the primary driving force is indeed
the drag force, which is also responsible for transporting the
particles upward.
For the reverse-flow pattern, the aerodynamic drag force in
cD ) 0.44 (for Rep > 1000) (32) the dilute region was more than 10-4 N. This is still the
FgDp|u
bg - b
u p| dominant force, as compared to gravity and electrostatic forces.
Rep ) (33) This implies that particles would also move upward in this
µg region. In the dense region, aerodynamic drag force decreased
6078 Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007

Figure 15. Averaged current on pipe wall for three flow patterns (AF ) air flow rate; Tc ) charging time).

Table 3. Comparison of Forces on a Single Particle for Three Flow Patterns

Disperse Flow Reverse Flow Half-ring Flow
average dilute transition dense dilute dense
parameter flow region flow region flow region flow region low region flow region
air flow rate (L/min) 1600 1100 1000
particle velocity, b
upa (m/s) 1.6 6.2 × 10-1 -5.4 × 10-2 -2.0 × 10-3 2.3 × 10-1 0
equilibrium charge on the pipe, qb (C) -3.8 × 10-8 -1.0 × 10-6 -7.1 × 10-7
charge per particle, Qc (C) 1.6 × 10-11 2.6 × 10-11 3.0 × 10-11
gravity (N) 1.3 × 10-4 1.3 × 10-4 1.3 × 10-4
drag force, B
FDd (N) 6.2 × 10-4 5.0 × 10-4 4.1 × 10-4 2.5 × 10-4 4.3 × 10-4 2.1 × 10-4
electrostatic force, FEe (N) 1.8 × 10-6 8.0 × 10-5 2.3 × 10-4 1.2 × 10-3 6.5 × 10-5 9.3 × 10-4
friction on the bottom of the pipe (N) N.A. N.A. N.A. 6.9 × 10-4 N.A. 5.7 × 10-4
FD:G:FE 345:71:1 6.2:1.6:1 3.1:1:1.8 2.0:1:9.1 6.6:1.9:1 1.6:1:7.4
bp) is the averaged value and obtained from a high-speed camera. b Equilibrium charge on pipe (q) is calculated from eq 27, where T
a Particle velocity (u

is the charging time Tc. c Charge per particle (Q) is obtained from multiplying the charge density by the mass of a single particle. d Drag force (F BD) is
calculated from eq 30, where cD ) 0.44. e Electrostatic force (FE) is calculated from eq 34.

to the same range as gravity and could not provide enough
driving force to transport particles upward, so that particles are
likely to slide downward, because of the pull of gravity.
However, the extents of electrostatic force and associated friction
were high (∼10-3 N). It is probable that the electrostatic force
causes particles to stick to the pipe wall and friction limits the
reverse speed to almost zero. However, in the transition region,
the magnitudes of the three types of forces were almost in the
same ranges, indicating that particles should move either upward
or downward. Some particles may be attracted and approach
the dense region by electrostatic force. These particles are close
to, but do not touch, the pipe wall and flow reversely; however,
the reverse velocities in this region were greater than that of
the bottom particles, because of the absence of friction for such
suspended particles. Therefore, in this region, the track of
particle motion was more wavelike, as shown in Figure 2b. On
the other hand, some particles may still move ahead but also
may collide with particles moving in the opposite direction, thus
resulting in such particles turning around. Generally, the results
seem to confirm that the reverse-flow pattern is most significant
in the transition region.
In addition, the formation of a transition zone may be
interpreted to indicate the presence of shearing instabilities at
the interface between the dilute and dense regions. Particles in
the dilute region move at higher velocities than those in the
dense region. This relative velocity between the two regions
causes significant shearing effects in a transition region between
the two. Particles in this transition region are more unstable
than those in the dense region. Furthermore, in the transition
region, aerodynamic drag forces are smaller than those in the
dilute region and electrostatic forces are smaller than those in
the dense region. This resulted in the particularly negative
velocities of solids within the transition region observed in the
present study. Similarly, according to the report by Goldfarb et
al.,17 a transition region formed at the interface between a steady
state and a wavy state formed by two shearing granular flows
in their experiment involving two streams of identical grains
flowing on an inclined chute downstream of a splitter plate.
 Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007 6079

Figure 16. Comparison of electrostatic characters on the bottom of the pipe and the top of the pipe for the half-ring-flow pattern (air flow rate ) 1000
L/min): (a) induced current measurement section, (b) averaged current on the pipe wall, and (c) wall charge from integration of the induced current.

Table 4. Electrostatic Forces on Single Particle in the Entire Reverse Area of Inclined Pneumatic Conveying (Air Flow Rate ) 1100 L/min)
Value
parameter PP without Larostat-519 PP with Larostat-519
particle velocity (m/s) (-5.4 × 10-2)-(-2.0 × 10-3) (-3.3 × 10-1)-(-1.7 × 10-1)
equilibrium charge on pipe (C) -1.0 × 10-6 -5.2 × 10-8
charge per particle (C) 2.6 × 10-11 3.0 × 10-12
electrostatic force (N) 2.3 × 10-4-1.2 × 10-3 1.3 × 10-6-6.7 × 10-6
friction on the bottom of pipe (N) 6.9 × 10-4 5.35 × 10-5
measured Λ valuea 1.8-9 0.01-0.05
a Λ ) FE/G.

As for the half-ring-flow pattern, the ratio of the three types
of forces and the observations in the two regions were both
similar to those of the reverse-flow pattern. As such, the details
of the analysis will not be repeated here. Furthermore, particle
velocity was much lower, because of a smaller air flow rate in
the dilute region and particles had a tendency to remain
stationary and form a ring structure on the pipe wall in the dense
region and part of the dilute region, probably because of the
force balance described previously; thus, the ring structure here
was relatively stable during the entire transport process.
In addition, a further experiment was performed to clarify
the reason why a ring or half-ring structure formed. This
measurement is modified from the induced current test and is
shown in Figure 16a, where the test section is divided into two
parts and the experimental data can be obtained from the top
and bottom of the pipe, respectively. The results of average
currents calculated from eq 28 were shown in Figure 16b, in
which equilibrium times, i.e., charging times (Tc) can be read
as fairly similar values for two cases. Consequently, the induced
currents were integrated with time to obtain the charge on the
pipe wall (Figure 16c), according to eq 27. It is observed that
the wall charge at the top of the pipe is greater than that at the
bottom of the pipe during the entire experiment, as well as at
the moment of equilibrium time. Therefore, the magnitude of
6080 Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007

Figure 17. Results of reverse-flow pattern with pulsating wave (air flow rate ) 1100 L/min): (a) normalized signal from two pressure transducers against
time (4 s); (b) particle concentration (R) captured from two ECT planes against time (4 s); (c) power spectral density of ECT data in plane 1; and (d) power
spectral density of ECT data in plane 2.

electrostatic force at the top of the pipe is high (up to 10-3 N);
thus, such a value would be large enough to overcome the
gravity and attract particles to stick on the upper part of the
pipe wall.
4.2.4. Validation of Control Experiments. The unique
feature of the reverse-flow phenomena, especially the velocity
difference between the transition and dense region, is attributed
to the electrostatic effect, as analyzed previously. Thus, it is
anticipated that particles at the bottom of the pipe would slide
downward with higher speed and, thus, the transition region
would disappear, when the electric force was reduced. To verify
this postulation, control experiments, which isolate the electro-
static effect on a system by holding constant all variables but
the one under observation, were performed. A commercially
available antistatic agent, Larostat-519 powder,18 has been
demonstrated to be a suitable means to reduce electrostatic
effects effectively in a pneumatic conveying system.6 Larostat-
519 is a non-inflammable white powder with a bulk density of
520 kg/m3; it is composed of ∼60% soyadimethylethyl am-
monium and ∼40% ethasulfate amorphous silica. In this
experiment, 0.5% (by weight) of Larostat-519 powder was
mixed with PP granules and other conditions were fixed to be
the same as those for reverse flow. In comparison with previous
cases, where no such agent was used, the equilibrium charge
on the pipe wall and on each particle obtained in the presence
of the Larostat-519 powder was smaller, by ∼2 orders of

Figure 18. ECT images of particle concentration (R) at test plane 2 when the air flow rate is 1000 L/min: (a) PP without an antistatic agent and (b) PP with
an antistatic agent (Larostat-519 powder).
magnitude and 1 order of magnitude, respectively, as shown in
Table 4. Correspondingly, a drastic drop in electric force acting
on each particle at the bottom of the pipe, as well as frictional
forces, were observed. Under the same air flow rate (1100
L/min), such small aerodynamic drag force cannot afford enough
power to move particles upward, as mentioned in the last section.
Consequently, particles on the bottom of the pipe slid downward
with higher velocities, compared to the situation without an
antistatic agent, because the small electric force and associated
friction would not restrict particles in this region. Furthermore,
bottom particles would drive other particles in the dense region
with almost the same velocities. Thus, particles in the entire
dense region flowed in the reverse direction.
This process was captured by a high-speed video camera, as
shown in Figure 5d. The range of the dilute region in these
three images is basically the same as that in Figure 5b; however,
no obvious transition region is observed. This is because
particles A on the bottom of the pipe and particle B at the
interface of the dense and dilute regions are moving with almost
the same speed (-0.25 ( 0.078 m/s).
Furthermore, during the experimental process of reverse-flow
pattern without the influence of electrostatic forces achieved
Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007 6081
by adding the antistatic agent, it was observed that particles
did not slide downward in a smooth and steady manner. Instead,
from the observation made in the present study, the speed of
the reversing particles showed periodic fluctuations and solids
moved downward in alternating pulses of high and low
concentration, as shown in Figure 2d. The slow motion captured
by the high-speed video camera presented this phenomenon in
detail: initially, a stream of solids flowed down at a relatively
high speed from downstream to upstream along the conveying
pipe; then, the reverse speed of some particles decreased; these
particles blocked those flowing back from downstream and
accumulated on the upstream side of the pipe wall, making the
solids stream intermittent. (This continued until friction between
the particles and the wall or among the particles was unable to
support the gravity of the group of particles.) The process was
then repeated.
In conclusion, the behavior of solids motion in the dense
region exhibits a pulsating type of movement. Such flow pattern
can be identified by the pressure fluctuations in the conveying
of solids.19 Therefore, an additional measurement was performed
to obtain the pressure data using two pressure transducers (Gems
2200BGA1002A3UA, Basingstoke, England) at distances of
6082 Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007

Table 5. Material Properties and System Parameters for Figures 17c and 17d show that both power spectral density
CFD-DEM Simulations profiles exhibit a dominant peak at ∼1 Hz. This is also consistent
parameter value/comment with the fluctuation period of 1 s, as observed from Figures
shape of particles spherical 17a and 17b.
type of particles polypropylene, PP Another experiment was performed for the purposes of
number of particles 1000 comparison and confirmation, by mixing an antistatic agent into
particle diameter 2.80 mm
particle density 1123 kg/m3 particles at a fixed air flow rate of 1000 L/min. The ring-flow
spring constant, κ 5.0 × 103 N/m pattern then disappeared, as shown in Figure 18.
coefficient of restitution 0.9 4.2.5. Validation of Numerical Results. To verify the
coefficient of friction 0.3
gas density 1.205 kg/m3
aforementioned analysis, the experimental results for reverse
gas viscosity 1.8 × 10-5 N s/m2 flow were compared with a numerical simulation of pneumatic
pipe diameter 40.0 mm conveying of granular solids through a 45° inclined pipe, which
pipe length 1.0 m was achieved using a simple electrostatic field model and DEM
pipe inclination 45° coupled with CFD. The geometry of the pneumatic conveying
computational cell size 4 mm × 4 mm
simulation time step, ∆t 10-7 s system and the type of particles used in the simulations are based
on the experimental work. Material properties and system
parameters are listed in Table 5. The number of particles used
0.85 and 1.85 m from the bend, respectively as labeled “16” was 1000, corresponding to an overall solid concentration of β
and “17” in Figure 1. The pressure data were acquired ) 0.16, where β is defined as the overall volume fraction of
simultaneously for 100 s with the sampling rate set at 100 Hz. particles divided by the volume fraction of particles at maximum
Figure 17 shows the results of the normalized signal from packing, which is generally taken to be 0.64. A dimensionless
the two pressure transducers against time (4 s). It is observed quantity (Λ) that indicates the ratio of electrostatic force arising
that two pressure waveforms from the two sensors are out of from the charged pipe wall to the gravitational force acting on
phase with each other, with each signal having a regular each particle is defined in section 3.4. The specific values used
period of ∼1 s. The time lag (∆t) between the two waves for the various parameters, such as the spring constant and
indicated the time taken for one cluster of particles to slide from coefficients of restitution and friction, have been shown to have
test point 1 to test point 2. The average ∆t value was obtained minimal effects on the flow patterns produced from the CFD-
statistically to be 0.5 ( 0.3 s. The corresponding time lag DEM simulations in a sensitivity analysis performed by Lim et
between two particle concentration waves measured by two sets al.11 In all simulations performed, particles were first allowed
of ECT sensors was also observed to be similar in mag- to settle freely under gravity for 0.5 s and form a packing at
nitude (see Figure 17b). The power spectral densities for both the “bottom” of the pipe before gas flow was initiated. Periodic
waves were obtained by a fast Fourier transform (FFT) of the boundary conditions were applied to the solid phase to simulate
pressure and concentration data, to have a more quantitative an open flow system while a uniform gas velocity profile was
comparison of the dominant frequencies for both waveforms. maintained at the inlet. Particles that were carried out of the

Figure 19. Numerical result of reverse flow [β ) 0.16 (1000 particles)]: (a) enlarged image of one section in pneumatic conveying through a pipe (inlet
gas velocity ) 10 m/s, Λ ) 5.0) and(b) diametrical distribution of particle axial velocity.

conveying pipe by the flowing gas were simulated to re-enter
from the inlet of the pipe with the same velocities and radial
positions.
Figure 19a successfully shows that reverse-flow behavior can
be observed in particles moving along the bottom wall, when
the inlet gas velocity is 10 m/s and Λ ) 5.0. Figure 19b
describes the solid velocity distributions along the pipe diameter
from the bottom of the pipe (y ) 0) to the top of the pipe (y )
1). Here, particle velocities are actual values and not normalized,
to have a good comparison between the two groups of data. In
this figure, the symbols at the bottom right-hand corner represent
the inlet gas velocities and the dimensionless quantity Λ.
Consequently, when electrostatic force was 5 times the mag-
nitude of the gravitational force (Λ ) 5), negative velocities
appeared. This shows that some solids at the bottom of the pipe
may move in a direction opposite to the air flow. In particular,
a peak negative value (-0.48 m/s) in the velocity between the
positions y ) 0.2 and y ) 0.3 corresponds to the transition
region and illustrates a quantitative agreement with the experi-
mental data (-0.077 m/s to -0.031 m/s). However, when the
electrostatic force was absent (Λ ) 0), the velocity profile
showed more significant backflow (-0.55 m/s to -0.38 m/s)
in the dense region. Furthermore, the change from negative
values to positive values from the bottom to the top of the pipe
indicates the gradual disappearance of the transition region.
Therefore, these results demonstrate that the electrostatic force
is an essential determinant of negative solid velocities in the
transition zone of the reverse-flow regime. Generally, the
simulation results show good agreement with the experimental
results and physical observations.
5. Conclusions
The concentration distribution of polypropylene (PP) particles
transported in a 45° inclined pneumatic conveying pipe was
measured using electrical capacitance tomography (ECT). By
applying the best-correlated pixel method, the characteristics
of velocities and directions of particle motions in three different
flow patterns were described and compared with results of
particle image velocimetry (PIV) and high-speed video camera
measurements. It is concluded that when the air flow rate is
high, the entire solid phase is dilute and particles are dispersed
over the entire cross section of the pipe with high velocity,
especially in the pipe center. In contrast, at low air flow rates,
most of the solid particles are deposited at the bottom of the
pipe and some of them are transported in the upper part of the
pipeline, forming the dense and dilute regions, respectively; the
corresponding velocity distribution shows the presence of
reverse flow in a transition region between these two regions.
The reverse-flow pattern is also observed in the computational
fluid dynamics-discrete element method (CFD-DEM) simula-
tion results. When the air flow rate is decreased further, a half-
ring flow structure is formed and the majority of particles
adhered to the pipe wall. The major forces acting on single
particles were analyzed for the three flow patterns, including
the drag force, electrostatic force, gravity, and friction. For the
dispersed-flow pattern, the dilute region of the reverse-flow
pattern, and the half-ring-flow pattern, aerodynamic drag force
was the major driving force transporting particles upward. For
the dense region of the reverse-flow pattern and half-ring-
flow pattern, the electrostatic force predominated and at-
tracted particles onto the pipe wall. For the transition region of
reverse flow, three major forces were in the same ranges and
caused particles to be in a suspended state. In this region,
particles would be drawn by the gravitational force to flow
Ind. Eng. Chem. Res., Vol. 46, No. 19, 2007 6083
downward. Thus, the electrostatic force was essential for the
occurrence of three regions in reverse flow, and this was also
validated by the control experiments and the results of CFD-
DEM simulation.
Acknowledgment
This study has been supported by the SERC (A*STAR), under
Grant No. R-279-000-208-305. The authors are grateful to Dr.
Jun Yao, Dr. Kewu Zhu, Fong Yew Leong, and Lai Yeng Lee
for many helpful discussions on this project.
Literature Cited
(1) Levy, A.; Mooney, T.; Marjanovic, P.; Mason, D. J. A comparison
of analytical and numerical models with experimental data for gas-solid
flow through a straight pipe at different inclinations. Powder Technol. 1997,
93, 253-260.
(2) Hirota, M.; Sogo, Y.; Marutani, T.; Suzuki, M. Effect of mechanical
properties of powder on pneumatic conveying in inclined pipe. Powder
Technol. 2002, 122, 150-155.
(3) Zhu, K;, Wong, C. K.; Rao, S. M.; Wang, C. H. Pneumatic conveying
of granular solids in horizontal and inclined pipes. AIChE J. 2004, 50, 1729-
1745.
(4) Zhu, K.; Rao, S. M.; Wang, C. H.; Sundaresan, S. Electrical
capacitance tomography measurement on vertical and inclined pneumatic
conveying of granular solids. Chem. Eng. Sci. 2003, 58, 4225-4245.
(5) Maré, T.; Voicu, I.; Miriel, J. Numerical and experimental visualiza-
tion of reverse flow in an inclined isothermal tube. Exp. Therm. Fluid Sci.
2005, 30, 9-15.
(6) Yao, J.; Zhang, Y.; Wang, C. H.; Matsusaka, S.; Masuda, H.
Electrostatics of the granular flow in a pneumatic conveying system. Ind.
Eng. Chem. Res. 2004, 43, 7181-7199.
(7) Rao, S. M.; Zhu, K. W.; Wang, C. H.; Sundaresan, S. Electrical
capacitance tomography measurements on the pneumatic conveying of
solids. Ind. Eng. Chem. Res. 2001, 40, 4216-4226.
(8) Su, B. L.; Zhang, Y. H.; Peng, L. H.; Yao, D. Y.; Zhang, B. F. The
use of simultaneous iterative reconstruction technique for electrical
capacitance tomography. Chem. Eng. J. 2000, 77, 37-41.
(9) Hua, J. S.; Wang, C. H. Electrical capacitance tomography measure-
ments of gravity driven granular flows. Ind. Eng. Chem. Res. 1999, 38,
621-630.
(10) Msosorov, V.; Sankowski, D.; Manzurkiewicz, L.; Dyakowski, T.
The ‘best-correlated pixels’ method for solid mass flow measurements using
electrical capacitance tomography. Meas. Sci. Technol. 2002, 13, 1810-
1814.
(11) Lim, E. W. C.; Wang, C. H.; Yu, A. B. Discrete element simulation
for pneumatic conveying of granular material. AIChE J. 2006, 52, 496-
509.
(12) Lim, E. W. C.; Zhang, Y.; Wang, C. H. Effects of an electrostatic
field in pneumatic conveying of granular materials through inclined and
vertical pipes. Chem. Eng. Sci. 2006, 61, 7889-7908.
(13) Di, Felice, R. The voidage function for fluid-particle interaction
systems. Int J. Multiphase Flow 1994, 20, 153-159.
(14) Yao, J.; Zhang, Y.; Wang, H, C.; Liang, Y. C. On the Electrostatic
Equilibrium of Granular Flow in Pneumatic Conveying Systems. AIChE J.
2006, 52, 3775-3793.
(15) Sommerfeld, M. Analysis of collision effects for turbulent gas-
particle flow in a horizontal channel: Part I. particle transport. Int. J.
Multiphase Flow 2003, 29, 675-699.
(16) Zhang, Y.; Wang, C. H. Particle Attrition due to Rotary Valve
Feeder in a Pneumatic Conveying System: Electrostatics and Mechanical
Characteristics. Can. J. Chem. Eng. 2006, 84, 663-679.
(17) Goldfarb, D. J.; Glasser, B. J.; Shinbrot, T. Shear instabilities in
granular flow. Nature 2002, 415, 302-305.
(18) Zhang, Y. F.; Yang, Y.; Arastoopour, H. Electrostatic effect on
the flow behavior of a dilute gas/cohesive particle flow system. AIChE J.
1996, 42, 1590-1599.
(19) Dhodapkar, S. V.; Klinzing, G. E. Pressure fluctuations in pneumatic
conveying systems. Powder Technol. 1993, 74, 179-195.
ReceiVed for reView October 12, 2006
ReVised manuscript receiVed March 23, 2007
Accepted March 30, 2007
IE061304I