Powder Technology 131 (2003) 166 184

www.elsevier.com/locate/powtec

The flow patterns and stresses on the wall in a symmetric

louvered-wall moving granular filter bed
C.S. Chou a,*, W.F. Lo a, J. Smid b, J.T. Kuo c, S.S. Hsiau d
a

Department of Mechanical Engineering, National Pingtung University of Science and Technology, Pingtung 91207, Taiwan, ROC
b
Department of Mechanical Engineering, Czech Technical University, 16607 Prague 6, Czech Republic
c
Department of Mechanical Engineering, National Taiwan University, Taipei 10617, Taiwan, ROC
d
Department of Mechanical Engineering, National Central University, Chung-Li 32054, Taiwan, ROC
Received 11 December 2001; received in revised form 25 November 2002; accepted 5 December 2002

Abstract
The flow patterns and stresses on the wall in a symmetric two-dimensional louvered-wall moving granular filter bed were investigated.
The static wall stress distributions produced by the granular solids were measured and compared with the theoretical prediction using the
differential slice and Runge Kutta (order four) methods. The variations in the dynamic wall stresses with time in a moving granular filter bed
were obtained. In addition, the effect of the louver angle upon the flow patterns and wall stresses was investigated. Four different flow
regions were observed in a moving granular filter bed. As the angle of the louver decreases, the quasi-stagnant zone area adjacent to the side
wall becomes smaller and the static normal stress acting on the convergent section of the side wall becomes larger. The magnitude of the
static normal stress acting on the convergent section is approximately 10 times as large as that acting on the vertical section. When the normal
stress measured by pressure gauge installed on the upper stage decreases to zero, the normal wall stress measured by pressure gauge installed
on the adjacent lower stage then begins to descend and fluctuate under the static normal wall stress during granular material withdrawal.
Employing the results obtained using stress measurements and image processing, the pressure pulsation phenomenon in a symmetric twodimensional louvered-wall moving granular filter bed may be further understood.
D 2002 Elsevier Science B.V. All rights reserved.
Keywords: Moving granular filter bed; Symmetrical louvered-wall; Pressure pulsation

1. Introduction
Hot gas particulate filtration is a key component of
current combined cycle power generation systems based
on the combustion and gasification of coals such as the
integrated gasification combined cycle (IGCC). In these
processes, the gases obtained from the coal must be
expanded through a gas turbine and the gas cleanup must
be carried out without cooling the gases to protect the
downstream heat exchanger and gas turbine components
from fouling and erosion. At the same time, flue gas
cleaning must meet particulate emission standards of 50
mg/mSTP3, which was set by legislation in the mid-1980s for
new and/or existing coal-fired boilers [1].
* Corresponding author. Tel.: +886-8-7703202x7016; fax: +886-87740142.
E-mail addresses: Cschou@mail.npust.edu.tw (C.S. Chou),
Jtkuo@w3.me.ntu.edu.tw (J.T. Kuo), Sshsiau@cc.ncu.edu.tw (S.S. Hsiau).

Particulate removal from a hot gas stream can be accomplished using cyclones, barrier filters, electrostatic precipitators, granular bed filters or scrubbers [2]. The choice of
which filter to use is a complicated optimization problem
involving emissions, reliability and costs. The most promising alternatives seem to be granular bed filters [3] and
barrier filters.
Ceramic barrier filters use the most advanced hot gas
filtration technology with several systems nearly ready for
commercial use in the 250 400 jC temperature range [4
7]. However, problems encountered during recent pilot and
demonstration-scale tests, particularly at high temperatures
(up to 900 jC), have led to concerns over the future
exploitation of this technology. For this reason, alternative
technologies such as granular bed filters continue to be
investigated and developed [8].
Beds of granular solids have been employed for dust
collection for many years. This subject has gained prominence recently as a possible means of simultaneously

0032-5910/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0032-5910(02)00344-3

C.S. Chou et al. / Powder Technology 131 (2003) 166184

removing fly ash and sulfur dioxide from powerhouse flue

gases at temperatures in excess of 400 jC. Granular bed
filtration can be operated in four modes: fixed bed, intermittently moving bed, continuously moving bed and fluidized bed [9,10].
In the moving bed cross flow operations, the filter bed is
a vertical layer of granular material, held in place by
retaining grids or louvered walls. The gas passes horizontally through the granular layer while filter granules move
downwards and are removed from the bottom of the moving
filter bed (see Fig. 1). Apparently, the shape and configuration of the louvers are design features that need special
attention. Properly designed louvers can prevent a dense
cake from forming on the panel face that may cause
plugging, arching and filter media flow stoppage [11]. These
plugging and arching problems could cause excessive
pressure drops and hamper continuous filter operation.
Consequently, moving granular bed filtration requires a
filter design that puts no restrictions on the flow of the filter
media between the louvered walls.
Knowledge of filter granule velocity fields inside the
convergent channels between the pairs of louvers provides
important information for system design. However, most
related studies have only characterized the moving filter
bed, using the average granular velocity or the mass flow
rate of the filter granules [12 14].
The flow patterns of filter granules and velocity fields in
a granular moving bed with various wall designs were
experimentally studied in [15 18] and numerically studied
in [19 22]. The main finding of these studies was that
stagnant and quasi-stagnant zones could exist in the regions
adjacent to the louvers. The influences of the louver angles
upon the velocity profiles were also discussed. Additionally,

Fig. 1. Illustration of the granular filter bed [19].

167

Fig. 2. The schematic drawing of the experimental apparatus.

Chou et al. [20] declared that louver efficiency F/L in terms

angle of repose ar and louver angle aL was determined using
F/L = sin(aL)/cos(ar), where L was the louver length and F
was the span of the free surface. The louver efficiency was
found to increase with an increase in both the angle of
repose ar and the louver angle aL.
Roberts [23] explained the behavior of bulk solids
during storage and flow is dominated by the Coulomb
frictional properties of the bulk solid itself as well as the
wall friction angle between the bulk solid and boundary
wall of a hopper. When granular flow occurs in a hopper, a
slip stick type pulsating motion generally characterizes the
flow. Whether the corresponding induced dynamic loads are
observable depends on the degree of severity in the flow
pulsation and on the natural frequencies of the bin and
supporting structure.
Although, in the past, many theoretical, numerical and
experimental methods were conducted to explore the stress
and flow behavior of bulk solids in storage silos, less
attention has been paid to the pulsating wall stresses in a
moving granular filter bed. This research studied the flow
patterns and wall stresses in three kinds of 2-D symmetrical
filter beds. Filter granules were moved between the two
vertical louvered walls of the filter with no interstitial fluid
flow relative to the solids.
In this research work, the method developed by Chou et
al. [24] was employed to investigate the flow pattern and
stresses on the wall in a two-dimensional moving granular
filter bed. A pressure gauge for measuring the normal and
shear stresses of granular solids acting on a sensing plate
was used. The striking feature of this pressure gauge is that
measurements are made of both shear and normal stress.
The flow pattern histories of granular solids in a twodimensional moving granular filter bed were recorded using

168

C.S. Chou et al. / Powder Technology 131 (2003) 166184

a digital camcorder. The static wall stress distributions

produced by the granular solids were measured and compared with the theoretical prediction using the differential
slice and Runge Kutta (order four) methods. The variations
in the dynamic wall stresses with time in a moving granular
filter bed were determined. In addition, the effect of the
louver angle upon the flow patterns and wall stresses was
investigated.

2. Experimental apparatus and procedures

2.1. Two-dimensional moving granular filter bed
A two-dimensional moving granular filter bed was set up
to observe the flow patterns of 6-mm plastic spheres and to
measure the wall stress during centric discharge. A schematic drawing of the moving granular filter bed is shown in

Fig. 3. The layout of the pressure gauges and detailed dimensions of the granular filter bed for three tests.

C.S. Chou et al. / Powder Technology 131 (2003) 166184

Fig. 2. This granular bed consisted of a layer of particulate

material sandwiched between two transparent acrylic panels
with louver-like side walls. The granules were fed into a
vertical channel from a hopper at the top that had a
rectangular discharge slot with the same cross-sectional
dimensions as the vertical channel. A granular solid flow
was induced and controlled by a moving belt underneath the
discharge slot.

169

The height of the granular bed was 1500 mm, the width
was 455 mm and the depth was 124 mm. The granular bed
width to depth ratio was kept above 3:1 to promote twodimensional behavior. The louvers angle could be adjusted
by placing a wedge shape steel plate between the louver and
the vertical section of the side wall (see Fig. 2). Each of the
side walls had six circular holes, reserved for installing
pressure gauges. The layout of the pressure gauges and

Fig. 4. The dimensions of the pressure gauges. (a) Cylindrical type, (b) hexahedron type.

170

C.S. Chou et al. / Powder Technology 131 (2003) 166184

Fig. 4 (continued ).

detailed dimensions of the granular filter bed for the three

tests are shown in Fig. 3.

gauges were attached (see Fig. 5). This ring comprises the
basic dynamometric element of the pressure gauge. The
wiring and locations of the semiconductor strain gauges on

2.2. Two-directional pressure gauge

Pressure gauges for normal and shear stress measurements were installed on the side wall of a granular filter bed.
This pressure gauge with semiconductor strain gauges is
simple and suitable for measuring the static and dynamic
pressures of granular solids acting on silo walls. It enables
simultaneous and independent measurements of both stress
vector components. Two kinds of pressure gauges were used
in this research work. The first, which is a cylinder, was
installed on the vertical section of the side wall. The
effective area of the sensing plate for this pressure gauge
was 23 cm2 (see Fig. 4(a)). The second, which is a
hexahedron, was installed on the convergent section of the
side wall. The effective area of the sensing plate for this
pressure gauge was 1.98 cm2 (see Fig. 4(b)).
The normal stress effect pn and shear stress effect pt for
the granular solid layers were transmitted by the sensing
plate to a steel ring to which the semiconductor strain

Fig. 5. The schematic drawing of two-directional pressure gauge.

C.S. Chou et al. / Powder Technology 131 (2003) 166184

171

and
yit vit kti xit :

The subscripts n and t stand for the normal and tangential

directions, respectively. The superscript i means the ith
pressure gauge, y is the output voltage, v is the initial
voltage, x is the stress due to the weight and k is the slope
of the calibration straight line. Fig. 7(a) (b) demonstrates
the typical results of normal and shear stress measurement
calibrations for one of the pressure gauges used in this
research work, respectively. At the beginning of the calibration, the initial voltages for normal and shear stress

Fig. 6. The schematic drawing of the calibration system for pressure gauges.
(a) For normal stress, (b) for shear stress.

the steel ring were made so that the shear stress effect pt was
completely eliminated during the normal stress pn measurement. The shear stress pt measurement is not affected by the
simultaneously acting normal pressure pn.
In this research work, the standard weights were used to
calibrate the pressure gauges. The schematic drawings for
normal stress measurement calibration and shear stress
measurement calibration are shown in Fig. 6(a) (b), respectively. For each pressure gauge, the shear stress measurement calibration was carried out using the following
procedures: (1) a pressure gauge was installed on the bench
and the sensing plate was placed flush to the surface of the
bench; (2) a screw driven into the sensing plate was
wrapped with a cord; (3) the other end of the cord was
used to tie a rod welded to a base; (4) two electrical wires
for transmitting the output voltage of the pressure gauges
were connected to a home made connector; (5) the different
weights were placed on the base one by one and the output
voltage was recorded.
This pressure gauge exhibits linear calibration characteristics (i.e., the linear relationship between the stress and the
output voltage). For each pressure gauge, the calibration
straight lines for normal and tangential stresses were determined, using, respectively,
yin vin kni xin ;

Fig. 7. The calibration of the stress measurement. (a) For normal stress, (b)
for shear stress.

172

C.S. Chou et al. / Powder Technology 131 (2003) 166184

Table 1
Test conditions

Louver angle (j)

Louver length (mm)
Louver spacing (mm)
Louver width (mm)
Mass flow rate (kg/s)

2.3. Experimental procedures and measuring systems

Test 1

Test 2

Test 3

50
200
363
455
0.0915

45
200
363
455
0.09

40
200
363
455
0.091

measurements are recognized. During the normal stress

measurement calibration, the linear relationship between
the normal stress and the output voltage is shown in Fig.
7(a), but the output voltage of the shear stress measurement
maintains the same value. In contrast, during the shear stress
measurement calibration, the linear relationship between the
shear stress and the output voltage is shown in Fig. 7(b), but
the output voltage of the normal stress measurement maintains the same value.
The sensitive surfaces of the pressure gauges were placed
flush to the inner surface of the filter bed side walls. At each
pressure gauge, the normal and tangential stresses were
determined using

pin

Yni  Vni
;
kni

Yti  Vti
;
kti

and
pit

in which p is the stress produced by the granular material, Y

is the output voltage and V is the initial voltage before
pouring the granular material into the channel.
ACI 313-97 introduces an angle of friction between the
stored material and wall (or hopper) surface [25]. This
equation, which describes a coefficient of wall friction l,
is given by
tan/

Pt
l
Pn

Both the static and dynamic wall friction angles were

determined using this experimental apparatus. The wall
friction angle is one of the important parameters employed
to design a flow corrective insert for eliminating the
stagnant zones in a bin [26,27]. Consequently, in this
research work, the wall stress measurements provide the
key information to properly design a flow corrective insert
for eliminating the stagnant zones in a granular filter bed
and increase the effectiveness of the filter.

The pressure gauges, which could measure normal wall

pressures and tangential wall shear stresses simultaneously,
were calibrated before installation on the side walls. The
cylindrical pressure gauges installed on the vertical section
of the left side wall are marked L1 to L3 and the hexahedron
pressure gauges installed on the convergent section of the
left side wall are marked LS1 to LS3 (see Fig. 3). The
cylindrical pressure gauges installed on the vertical section
of the right side wall are marked R1 to R3 and the
hexahedron pressure gauges installed on the convergent
section of the right side wall are marked RS1 to RS3 (see
Fig. 3). Table 1 lists the test conditions for the three
experiments.
A granular material (6-mm diameter PE spheres with
density of 964 kg/m3) was employed in these experiments.
Sphere packing was characterized by the bulk density
measurements. Two bulk densities were measured: a
poured bulk density of 582 kg/m3 (porosity 0.396) and a
tapped bulk density of 600.5 kg/m3 (porosity 0.377). Both
densities were measured in a graduated glass cylinder. The
friction angles for the above-mentioned granular material
were obtained using a Jenike shear tester and are listed in
Table 2.
Before pouring the granular material into the granular
filter bed, the initial voltage for each pressure gauge
mounted on the side wall was recorded. To record the flow
development of the black colored granules, the black
colored PE spheres were filled in the third louver section
(from the discharge slot). Green colored PE spheres were
filled in the other louver sections.
After filling the bed with granular material, the static
normal and tangential stresses of the granular bed acting on
each pressure gauge were measured. The flow rate was
controlled, and the dynamic normal and tangential stresses
acting on each pressure gauge were measured during
material withdrawal. The mass flow rate measurements
were made by continuous collection of the discharged
granules in a tarred bucket. An electronic balance was used
to weigh the full buckets. The mass flow rate for three tests
was controlled and maintained at 0.09 0.1 kg/s during all
experiments.
A schematic drawing of the stress measuring equipment
and video imaging system is shown in Fig. 8. A home
made connector was used for the convenience of connecting the pressure gauge to the data acquisition card (see

Table 2
Friction angles
Granular
material

Particle
particle

Particle
side steel
wall

Particle
transparent
acrylic wall

PE

26j

8j

12j

C.S. Chou et al. / Powder Technology 131 (2003) 166184

173

Fig. 8. The schematic drawing of the stress measurement and video imaging systems.

Fig. 8). The output voltage from the pressure gauges was
amplified using a power supply (TopWard-6303D). A data
acquisition card (ADVANTECH PCL-818HG) was
employed to convert the analog signals into digital signals.

Computer software (VisiDAQ 3.1) was employed to process the data. At the same time, a digital camcorder (SONY
DCR-TRV310) was used to record the development of the
black colored granule flow until no more granules were

Fig. 9. A differential trapezoid slice in a wedge hopper.

174

C.S. Chou et al. / Powder Technology 131 (2003) 166184

left in the filter granular bed. An image grabber (MotoDV

IEEE-1394) was used to convert the flow images from the
recorded tape into computer graphic files. Computer software (PhotoImpact 7) was employed to edit the flow
images.

3. Theoretical static wall stress

Employing the differential slice method of Chou and
Chen [28] and the Runge Kutta method, which the wall
stresses produced by the granular solids in a two-dimensional wedge hopper were numerically calculated, the static
wall stresses in a moving granular filter bed were determined. Considering a differential trapezoid slice of material
in the convergent section of the granular filter bed shown in
Fig. 9, the static force balance in the vertical direction is
given by
w w dwbdh
qb g
2
dh
dh
Pv wb 2bSw1 cosa
2bPw sina
cosa
cosa
w w dwdh
2Sw2
2

Pv dPv w dwb

Where Sw1 is the tangential stress acting on the left and

right side walls, Sw2 is the tangential stress acting on the
front and rear walls, Pw is the normal stress acting on the
left and right side walls and Pv is the vertical stress acting
on the element slice. In addition, w, b, dh, a and qb
represent the width of the element slice, the thickness of
the element slice, the height of the element slice, the louver
angle and the bulk solid density.
By neglecting the high order terms (e.g., dwdh and
dPvdw), substituting w = w0 + 2htana and dw = 2dhtana
into Eq. (6), and assuming Sw1 = lw1K1Pv, Sw2 = lw2K1Pv
and Pw = K1Pv, Eq. (6) then becomes
Fig. 10. The static normal wall stress distribution in the granular filter bed.
(a) For Tests 1 (louver angle: 50j), (b) for Test 3 (louver angle: 40j).

2Pv bdhtana w0 bdPv 2hbdPv tana qb gw0 bdh

2qb ghbdhtana 2K1 Pv bdhtana 2lw1 K1 Pv bdh
2w0 lw2 K1 Pv dh 4hlw2 K1 Pv dhtana
7

and
K1

Where lw1 is the side-wall friction coefficient, lw2 is the

front- and rear-wall friction coefficient and K1 for the vertical
and convergent sections of the granular filter bed are
respectively given by [29]


p uw

K1 tan
4
2


8

tana
tanuw tana

Where uw is the louvered-wall friction angle.

In terms of K2 = w0b, K3 = 2btana, K4 = 2lw1b, K5 =
2w0Pw2, K6 = 4lw2tana, K7 = K1K3 + K1K4 + K1K5  K3 and
K8 = K1K6, Eq. (7) then becomes
dPv K7 K8 h

P v  qb g
K2 K3 h
dh

10

C.S. Chou et al. / Powder Technology 131 (2003) 166184

175

Fig. 11. The flow history of the black colored granules in the two-dimensional moving granular filter bed under Test 1 (louver angle: 50j). Frames 1 18, time
interval 25 s; frame 18, time 425 s.

176

C.S. Chou et al. / Powder Technology 131 (2003) 166184

Fig. 12. The flow history of the black colored granules in the two-dimensional moving granular filter bed under Test 2 (louver angle: 45j). Frames 1 18, time
interval 25 s; frame 18, time 425 s.

C.S. Chou et al. / Powder Technology 131 (2003) 166184

177

Fig. 13. The flow history of the black colored granules in the two-dimensional moving granular filter bed under Test 3 (louver angle: 40j). Frames 1 18, time
interval 25 s; frame 18, time 425 s.

178

C.S. Chou et al. / Powder Technology 131 (2003) 166184

4. Results and discussion

4.1. Static normal wall stress distribution in the granular
filter bed

Fig. 14. The schematic drawing of the four flow regions [20].

Eq. (10) is a first-order differential equation for the vertical

stress Pv, which can be numerically solved using the Runge
Kutta (order four) method [30]. The normal stress acting on
the louvered-wall can be determined using Pw = K1Pv and
Eqs. (8) and (9).

Fig. 10(a) (b) shows the static normal wall stress distribution in the granular filter bed under Tests 1 (louver
angle: 50j) and Test 3 (louver angle: 40j), respectively. In
each frame, the solid line and dashed line represent the
theoretical static normal stress distributions with surcharge
and under zero surcharge, respectively. In addition, cross,
diamond, square, circle, asterisk and saltire represent the
static normal stress measured by pressure gauges L3, LS3,
L2, LS2, L1 and LS1, respectively.
In general, at the vertical section of the louvered-wall,
the static normal stress measured by the pressure gauge
agreed well with theoretical prediction obtained using the
differential slice method under zero surcharge (see Fig.
10). For example, at the vertical section of the third stage,
the theoretical and experimental static normal stresses
under Test 3 (louver angle: 40j) were 0.68 and 0.6 kPa,
respectively.
In contrast, at the convergent section of the louveredwall, the static normal stress measured by the pressure

Fig. 15. At 125 s, a comparison of the flow status between Tests 1, 2 and 3. (a) For Test 1 (louver angle: 50j), (b) for Test 2 (louver angle: 45j), (c) for Test 3
(louver angle: 40j).

C.S. Chou et al. / Powder Technology 131 (2003) 166184

gauge was closer to the theoretical prediction obtained using

differential slice method with surcharge. For example, at the
convergent section of the first stage, the theoretical and
experimental static normal stresses under Test 1 (louver
angle: 50j) were 2.44 and 3.29 kPa, respectively.
4.2. Flow patterns in the moving granular filter bed
Figs. 11 13 show the flow history of the black colored
granules in the granular filter bed under Tests 1 (louver
angle: 50j), Test 2 (louver angle: 45j) and Test 3 (louver
angle: 40j), respectively. Each figure has 18 frames. Frame
1 shows the initial status of the granular bed. In Figs. 11
13, the time period for each frame is 25 s. The time for
Frame 18 is 425 s. Kuo et al. [15] demonstrated the flow
patterns for symmetrical moving granular filter beds that
louver angles were 40j, 30j and 15j, respectively. The
experimental results reported here provide additional information to the workers in this field.
The purpose of Tests 1, 2 and 3 was to demonstrate the
effect of louver angle on the development of the quasistagnant zone and the wall stresses. The flow pattern results

179

reported here and those of Refs. [15,20] are completely

alike. For example: four different flow regions were
observed: (1) a quasi-stagnant zone (Q-SZ) adjacent to the
louvered-wall; (2) a transition region (TR) between the
quasi-stagnant zone and a central flowing core; (3) a central
flowing core (CFC) with a plug flow; (4) left and right free
surface regions (FSR). A schematic drawing of the four flow
regions is shown in Fig. 14.
A new bed structure and porosity are formed in the quasistagnant zone as the filter granules flow out from the upper
pair of louvers and fill the lower louver section. The quasistagnant zone area becomes larger as the angle of the louver
increases and diminishes as time increases. The results from
Figs. 11 13, which are under Tests 1, 2 and 3, respectively,
explain the effect of the angle of the convergent louver upon
the development of quasi-stagnant zones. In addition, at 125
s from the beginning of the outflow, a comparison of the
flow status in the two-dimensional moving granular filter
bed among Tests 1, 2 and 3 is shown in Fig. 15(a) (c).
Frames (a), (b) and (c) are for Test 1 (louver angle: 50j),
Test 2 (louver angle: 45j) and Test 3 (louver angle: 40j),
respectively.

Fig. 16. The variations in dynamic normal and shear stresses with time for Test 1 (louver angle: 50j). The top panel is for pressure gauge RS3; the center panel
is for pressure gauge RS2; the bottom panel is for pressure gauge RS1.

180

C.S. Chou et al. / Powder Technology 131 (2003) 166184

There is a cascading granular transport in the transition

region at different stages in the granular bed and, consequently, granules flow from the shear zone of the upper
stage into the transition region of the lower stage (see
Frames 2 to 6 in Figs. 11 13). Additionally, the boundary
between the central plug flow core region and the transition
region is a slightly convex curve (see Fig. 15).
Because of the existence of free surfaces in the moving
filter bed, the granular flows expand when the granules are
just leaving the upper stage exit. A convergent granular bed
ensues due to the convergent section of the filter louver (see
Frames 2 to 6 in Figs. 11 13). Unlike the granular flows in
a bin-hopper system, the granular flows in a granular
moving filter bed are affected by the upper and lower
louvered-wall systems.
After bringing the free surface at the top of granular bed
into direct contact with the stagnant material, the boundary
of the stagnant material was diminished due to granule
erosion (see Frames 6 to 8 in Figs. 11 13). The free surface
at the top of the granular bed demonstrated avalanche
behavior.

4.3. Dynamic response of stresses on the wall

4.3.1. At the convergent section of the side wall
Figs. 16 and 17 demonstrate the dynamic response of
normal and shear wall stresses acting on the convergent
section of the right side wall for Test 1 (louver angle: 50j)
and Test 3 (louver angle: 40j), respectively. Two horizontal
dotted lines in each panel in Figs. 16 and 17 represent the
static normal and shear stresses, respectively. In general, for
pressure gauges RS1, RS2 and RS3, the dynamic responses
of the normal stress and shear stress have the same trend. In
addition, the shear stress value is always smaller than the
normal stress value.
In top panel of Fig. 16, the normal wall stress measured by
pressure gauge RS3 fluctuates about the static normal wall
stress (3.54 kPa) between 0 and 200 s. During this period, the
granules were emptied out of the fourth stage from the bottom
(see Frames 1 to 9 in Fig. 11). The stress recorded in the top
panel of Fig. 16 shows that a peak in both the normal and
shear stresses appeared simultaneously at 200 s. The reason
for this sudden stress increase was an avalanche of granules

Fig. 17. The variations in dynamic normal and shear stresses with time for Test 3 (louver angle: 40j). The top panel is for pressure gauge RS3; the center panel
is for pressure gauge RS2; the bottom panel is for pressure gauge RS1.

C.S. Chou et al. / Powder Technology 131 (2003) 166184

sliding along the free surface. This avalanche consequently

impacted on the stagnant material (see Frames 9 to 11 in Fig.
11). An increase in the normal and shear stresses measured by
pressure gage RS3 was then registered.
Between 200 and 275 s, the normal wall stress measured
by pressure gauge RS3 fluctuates under the static normal
wall stress and decreases to zero. At the same time, the
granules were emptied out of the third stage from the
bottom, where pressure gauge RS3 was installed (see
Frames 9 to 12 in Fig. 11).
In general, the dynamic responses of the normal stress
measured by pressure gauges RS1 and RS2 have the very
same trend as that measured by pressure gauge RS3 (see
Fig. 16). When the normal stress measured by pressure
gauge installed on the upper stage (e.g., RS3) decreases to
zero, the normal wall stress measured by pressure gauge
installed on the adjacent lower stage (e.g., RS2) begins to
descend and fluctuate under the static normal wall stress
(see Fig. 16). In addition, the magnitude of the static
normal stress measured by pressure gauges RS1, RS2
and RS3 for Test 1 (louver angle 50j) were of the same
order.

181

The dynamic responses of the wall stresses shown in Fig.

17 (for Test 3: louver angle 40j) have the same trend as that
shown in Fig. 16 (for Test 1: louver angle 50j). However,
the static normal stress acting on the convergent section of
the side wall becomes larger as the angle louver decreases.
For example, the static normal stresses measured by pressure gauge RS2 for Tests 1, 2 and 3 are 3.83, 4.17 and 5.85
kPa, respectively. The reason for this static normal stress
increase was that the volume of the convergent section
increases as the angle of the louver decreases under the
same louver length condition.
4.3.2. At the vertical section of the side wall
Figs. 18 and 19 demonstrate the dynamic response of
normal and shear wall stresses acting on the vertical section
of the left side wall for Test 1 (louver angle: 50j) and Test 3
(louver angle: 40j), respectively. In general, the magnitude
of the static normal stress acting on the convergent section is
approximately 10 times as large as that acting on the vertical
section.
The stress recorded in the top panel of Fig. 18 shows that
the normal wall stress measured by pressure gauge L3

Fig. 18. The variations in dynamic normal and shear stresses with time for Test 1 (louver angle: 50j). The top panel is for pressure gauge L3; the center panel is
for pressure gauge L2; the bottom panel is for pressure gauge L1.

182

C.S. Chou et al. / Powder Technology 131 (2003) 166184

Fig. 19. The variations in dynamic normal and shear stresses with time for Test 3 (louver angle: 40j). The top panel is for pressure gauge L3; the center panel is
for pressure gauge L2; the bottom panel is for pressure gauge L1.

fluctuates above the static normal wall stress (0.42 kPa)

between 125 and 200 s. During this period, the free surface
at the top of the granular bed moved downward from the
fourth stage to the third stage (see Frames 6 to 9 in Fig. 11).
The granular flows expand when the granules are just
leaving the upper stage exit. Consequently, the flow expansion and the avalanche behavior at the top of the free surface
significantly affected the magnitude of the wall stress
produced by the granules during material withdrawal. The
stress recorded in the center panel of Fig. 18 shows that the
normal wall stress measured by pressure gauge L2 fluctuates
under the static normal wall stress (0.56 kPa) during
material withdrawal.
In contrast, the stress recorded in the bottom panel of
Fig. 18 shows that, between 5 and 135 s, the normal wall
stress measured by pressure gauge L1 fluctuates about,
under and above the static normal wall stress (0.5 kPa) in
sequence. Between 135 and 265 s, the normal stress
measured by pressure gauge L1 has the same trend as that
measured by pressure gauge L1 between 5 and 135 s. At
265 s from the beginning of the outflow, only one upper
stage (i.e., second stage) is filled with granules above the

first stage, where pressure gauge L1 is installed (see

Frames 11 to 12 in Fig. 11). The normal wall stress
measured by pressure gauge L1 then fluctuates under the
static normal wall stress during the remaining discharge
time. In general, the height of the granular bed above the
pressure gauge significantly affected the magnitude of the
wall stress produced by the granules during material
withdrawal. Magnitudes of the static normal stress meas-

Table 3
Maximum dynamic, mean dynamic and static stresses for Test 1 (louver
angle: 50j)
Pressure
gauge

Normal

Shear

Max.
dynamic
stress

Mean
dynamic
stress

Static
stress

Max.
dynamic
stress

Mean
dynamic
stress

Static
stress

L1
L2
L3
RS1
RS2
RS3

0.62
0.63
0.68
6.1
7.3
7.5

0.45
0.48
0.51
3.5
4.1
3.7

0.42
0.56
0.5
3.54
3.83
3.29

0.19
0.18
0.01
3.1
3.8
3.3

0.13
0.08
0.01
0.8
0.8
0.6

0.15
0.15
0.01
1.11
1.01
0.26

C.S. Chou et al. / Powder Technology 131 (2003) 166184

Table 4
Maximum dynamic, mean dynamic and static stresses for Test 3 (louver
angle: 40j)
Pressure
gauge

Normal
Max.
dynamic
stress

Mean
dynamic
stress

Static
stress

Shear
Max.
dynamic
stress

Mean
dynamic
stress

Static
stress

L1
L2
L3
RS1
RS2
RS3

0.79
0.83
0.62
9.8
9.8
8.9

0.48
0.51
0.48
5.1
4.3
3.8

0.6
0.59
0.5
4.83
5.85
4.28

0.26
0.21
0.01
2.8
3.1
2.8

0.17
0.1
0.01
1.2
0.8
0.85

0.19
0.13
0.01
0.72
1.25
0.23

ured by pressure gauges L1, L2 and L3 for Test 1 (louver

angle: 50j) were of the same order approximately.
The records of stresses measured by pressure gauges L3
and L2 in the top and center panels in Fig. 19 for Test 3
(louver angle: 40j) showed a stress peak at the beginning of
material withdrawal. The reason for this sudden stress
increase was that at the beginning of the outflow, a flowing
core was formed above the discharge slot. Consequently, the
wall stress levels started increasing possibly due to the
gradual re-compaction of the material next to the walls in
the third and second stages of the filter granular bed.
A sudden stress decrease ensues because the dilated
material comes into contact with the walls. Because the size
of the flowing core for Test 1 (louver angle: 50j) is smaller
than that for Test 3 (louver angle: 40j), a sudden stress
increase and decrease at the beginning of the outflow is not
significant (see Fig. 18). Except for a sudden stress increase
and decrease at the beginning of the outflow, in general, the
normal stresses measured by pressure gauges L1, L2 and L3
for Test 3 (louver angle: 40j) fluctuate under the static
normal stresses, respectively, during the remaining discharge time.
This stress pulsation is explained as due to the existence
of the flowing core and quasi-stagnant zone. The boundary
between them is a shear plane where periodical shear
failures take place. The shear failure causes a stress pulse
that is transmitted through the quasi-stagnant zone to the
side wall of the filter bed. Maximum dynamic normal and
shear stresses, mean dynamic normal and shear stresses and
static normal and shear stresses measured by pressure
gauges L1, L2, L3, RS1, RS2 and RS3 for Test 1 (louver
angle: 50j) and Test 3 (louver angle: 40j) are listed in
Tables 3 and 4, respectively.

183

gauge used in this research work is that measurements

were made of both shear and normal stress. The angle of
the louver influences the granular bed flow in the filter
channel. The quasi-stagnant zone area became larger as the
angle of the louver was increased. Four different flow
regions were observed in a two-dimensional moving granular filter bed.
The static wall stress distributions produced by the
granular solids were measured and compared with the
theoretical prediction using the differential slice and
Runge Kutta (order four) methods. The effect of louver
angle upon the static wall stress was investigated. The
static normal stress acting on the convergent section of the
side wall becomes larger as the angle of the louver
decreases. However, for a fixed louver angle (e.g., 50j),
the magnitude of the static wall normal stress measured by
pressure gauges installed on the convergent section at each
stage (e.g., RS1, RS2 and RS3) were of the same order. In
addition, for a fixed louver angle, the magnitude of the
static normal wall stress measured by pressure gauges
installed on the vertical section at each stage (e.g., L1,
L2 and L3) were also of the same order. In general, the
magnitude of the static normal stress acting on the convergent section was approximately 10 times as large as that
acting on the vertical section.
The dynamic responses of the normal and shear stresses
acting on the vertical and convergent sections of side walls
were observed. In general, the height of the granular bed
above the pressure gauge significantly affected the magnitude of the wall stress produced by the granules during
material withdrawal. The flow expansion, when the granules
were just leaving the upper stage exit, and the avalanche
behavior at the top of the free surface significantly affected
the magnitude of the wall stress produced by the granules
during material withdrawal.
A pulsation phenomenon and inhomogeneous flows,
which are partially associated with instantaneous and
intermittent shear localization, were observed during material withdrawal. The filling method effect on the stress
variation and the effect of the insert shape and placement
on the stress variation, as well as the flow pattern in a
steady flow, are subjects worthy of future study. For this
purpose, a system of circulating granules should be
installed in the existing experimental apparatus in the near
future.

Acknowledgements
5. Conclusions
The flow patterns and stresses on the wall in a twodimensional moving granular filter bed were investigated.
Filter granules were moved between the two vertical
louvered walls of the filter with no interstitial fluid flow
relative to the solids. The striking feature of the pressure

The authors gratefully acknowledge the financial support

from the National Science Council of the R.O.C. for this
work through projects NSC 90-2211-E-020-006 and NSC
90-2625-Z-020-001. In addition, the authors gratefully
acknowledge the financial support from the Sun-CrownKing Taro Ice City for this work through project SCK 91NPUST-001.

184

C.S. Chou et al. / Powder Technology 131 (2003) 166184

References
[1] J.H. Ludwig, P.W. Spaite, Chemical Engineering Progress 63 (6)
(1967) 82.
[2] J.P. Mustonen, S.J. Bossart, M.W. Durner, Proc. of Int. Conf. on
Fluidized Bed Combustion, Montreal, Canada, vol. 1, 1991, p. 475.
[3] S. Hozumi, J. Fukui, K. Kodama, K. Kamei, T. Mori, N. Izumi, S.
Kusada, Proc. Symp. Gas Cleaning at High Temperature, Pergamon,
Oxford, 1986, p. 385.
[4] R. Chang, J. Sawyer, H. Lips, R. Bedick, R. Dellefield, Proc. Symp.
Gas Cleaning at High Temperature, Pergamon, Oxford, 1986, p. 177.
[5] K. Higashi, Journal of the Society of Powder Technology, Japan 36
(1999) 378.
[6] S. Ito, T. Tanaka, S. Kawamura, Advanced Powder Technology 8 (1)
(1997) 53.
[7] T. Nakayama, H. Matsuda, S. Ito, H. Shirai, M. Kobayashi, T. Tanaka,
Karyoku Genshiryoku Hatsuden 43 (1992) 210.
[8] J.R. Longanbach, . Proc. of 23rd Int. Technical Conf. on Coal Utilization and Fuel System, Clearwater, FL, USA, 1998, p. 69.
[9] J.T. Kuo, J. Smid, S.S. Hsiau, C.S. Chou, Proceedings of the National
Science Council, Republic of China (A) 22 (1) (1998) 17.
[10] J.P.K. Seville, R. Clift, in: J.P.K. Seville (Ed.), Gas Cleaning Demanding Applications, Blackie Academic and Professional, London, 1997,
pp. 170 171, Chapter 9.
[11] A.M. Presser, J.C. Alexander, Proc. of 3rd Symposium on Transfer
and Utilization of Particulate Control Technology, 1982, p. 26.
[12] Avco Co., Final Report, Contract No. PH-86-67-51, Phase Ill, AYATD0107-67-RR, U.S. Department of Commerce, Publication PB 185561,
pp. 8, 48, 67, 1969.
[13] R.G. Reese, Tappi (Journal of the Technical Association of the Pulp
and Paper Industry) 60 (3) (1977) 109.

[14] K. Ishikawa, N. Kawamata, K. Kamei, in: R. Clift, J.P.K. Seville

(Eds.), Gas Cleaning at High Temperature, Backie Academic and
Professional, London, 1993, pp. 419 435.
[15] J.T. Kuo, J. Smid, S.S. Hsiau, C.Y. Wang, C.S. Chou, Filtration and
Separation 35 (6) (1998) 529.
[16] S.S. Hsiau, J. Smid, C.Y. Wang, J.T. Kuo, C.S. Chou, Chemical
Engineering Science 54 (3) (1998) 293.
[17] S.S. Hsiau, J. Smid, F.H. Tsai, J.T. Kuo, C.S. Chou, Chemical Engineering Sciences 55 (1999) 4481.
[18] S.S. Hsiau, J. Smid, F.H. Tsai, J.T. Kuo, C.S. Chou, Powder Technology 144 (2000) 205.
[19] C.S. Chou, C.Y. Tseng, J. Smid, J.T. Kuo, S.S. Hsiau, Powder Technology 110 (2000) 239.
[20] C.S. Chou, C.Y. Tseng, J. Smid, J.T. Kuo, S.S. Hsiau, Advanced
Powder Technology 11 (3) (2000) 279.
[21] C.S. Chou, C.C. Tseng, S.S. Hsiau, H.H. Tsai, J. Smid, J.T. Kuo,
Proceedings of the National Science Council, Republic of China
(A) 24 (3) 2000, p. 226.
[22] C.S. Chou, H.C. Lai, J. Smid, J.T. Kuo, S.S. Hsiau, Transaction of the
Filtration Society 1 (2) (2001) 50.
[23] A.W. Roberts, Bulk Solids Handling 16 (1) (1996) 59.
[24] C.S. Chou, Y.C. Chuang, J. Smid, S.S. Hsiau, J.T. Kuo, Advanced
Powder Technology 13 (1) (2002) 1.
[25] American Concrete Institute, ACI 313R-97 (1998).
[26] J.R. Johanson, Powder Technology 1 (1967/1968) 328.
[27] J.R. Johanson, Bulk Solids Handling 2 (3) (1982) 495.
[28] C.S. Chou, R.Y. Chen, Advanced Powder Technology (2002) (in
press).
[29] P.A. Shamlou, Handling of Bulk Solids, Butterworth, London, 1998.
[30] R.L. Burden, J.D. Faires, Numerical Analysis, 4th ed., PWS-KENT
Publishing, Boston, 1989.