You are on page 1of 14

Powder Technology 109 2000.

2740
www.elsevier.comrlocaterpowtec

Numerical simulation of metallic solid bridging particles in a fluidized


bed at high temperature
Kenya Kuwagi ) , Takafumi Mikami, Masayuki Horio
Department of Chemical Engineering, Tokyo Uniersity of Agriculture and Technology, Koganei, Tokyo 184-8588, Japan
Accepted 20 September 1999

Abstract
High temperature fluidization of iron particles was investigated by numerical simulation based on the discrete element method DEM.
as a case study for metallic bridging. A model was developed for metallic solid bridging by surface diffusion mechanism including the
effect of surface roughness. The simulated fluidization behavior was highly time dependent, which is completely different from our
previous results for liquid bridging particles wT. Mikami, H. Kamiya, M. Horio, Chem. Eng. Sci. 53 1998. 1927.x. Both the amplitude of
pressure fluctuation and the absolute value of bed pressure drop decreased with time. These tendencies agreed well with the experimental
data of Mikami et al. wT. Mikami, H. Kamiya, M. Horio, Powder Techonol. 89 1996. 231.x although temperature and bed size were
different. Size and shape of agglomerates were much different for different surface roughness models. Hypha-shaped agglomerates were
more dominant in the case of the lowest cohesiveness three-microcontact-point model.. The size of agglomerates grown on the wall was
largest for the largest cohesiveness and smallest for the smallest cohesiveness. q 2000 Elsevier Science S.A. All rights reserved.
Keywords: Numerical simulation; Discrete element method; Surface diffusion; Iron particles; Agglomerate; Defluidization

1. Introduction
Metallic solid bridging is a well-recognized factor dominant in metallic fluidized bed processes such as iron oxide
reduction and silicon CVD if they are conducted in a
fluidized bed. Mikami et al. w1x carried out an experimental
investigation for a typical case of iron particles at high
temperatures and concluded that the neck growth of solid
bridging and the cohesion force between iron particles can
be predicted by the surface diffusion model of Kuczynskis
w2x sintering models. Although investigations into metallic
solid bridging fluidized beds have been conducted for
process developments w35x, and defluidization models
were recently proposed by Iwadate and Horio w6x and
Knight et al. w7x, no attempt has ever been made to
formulate the whole mechanism of agglomerating fluidization behavior of bridging metallic particles. Recently,
Mikami et al. w8x developed a simulation code SAFIRE.
based on the discrete element method DEM., considering
cohesion force by liquid bridging, and they successfully
simulated agglomerating fluidized bed behavior of wet
)

Corresponding author. Tel.: q81-42-388-7067; fax: q81-42-3863303; e-mail: quwagi@cc.tuat.ac.jp

particles. Iwadate and Horio w9x extended such simulation


into fluidized particles with van der Waals force interaction. However, the sintering phenomenon is quite time
dependent because sintering necks between contacting particles grow rather rapidly with time. Accordingly, the
fluidization behavior of solid bridging particles is completely different from that of liquid bridging particles or
group C particles of Geldart w10x classification. In this
paper, a solid bridging model is developed and applied for
numerical simulation of high temperature fluidization of
iron particles. Since agglomerating fluidization behavior
depends on the properties of agglomerates, attention is
focused also on the size and shape of agglomerates formed
in the bed.

2. Theoretical analysis
2.1. Definition of time
Three kinds of time are used in the present model. The
first is the absolute time, t, starting from the beginning of
computation. The second is the time of neck growth, t neck ,
corresponding to each neck. When a neck is broken in a
collision, t neck for the neck is reset to zero. The third is the

0032-5910r00r$ - see front matter q 2000 Elsevier Science S.A. All rights reserved.
PII: S 0 0 3 2 - 5 9 1 0 9 9 . 0 0 2 2 4 - 7

K. Kuwagi et al.r Powder Technology 109 (2000) 2740

28

time, t c , for collision analysis having its origin at any


instance of our interest.
2.2. Goerning equations

x neck s

The following locally averaged equations of Anderson


and Jackson w11x are used as governing equations for gas
phase. These are the same as those used in the previous
DEM simulations e.g., Refs. w8,12x..

Et

E ui .
E xi

s 0.

1.

Momentum balance:

rf

E ui .
Et

q rf

E ui u j .
E xj

s y

Ep
E xi

y fi .

2.

is the voidage, p the fluid pressure, f i the fluidparticle


interaction force as shown in Appendix A. The SIMPLE
Semi-Implicit Method for Pressure-Linked Equations.
scheme of Patanker w13x is adopted to solve the pressure
field of fluid.
2.2.2. Particle motion
Equations of motion for each particle are as
follows.Translational motion:
m

d
dt

56gd 4
k BT

1r7
3

Ds a t neck

surface diffusion. ,
5.

x neck s

10gd 3
k BT

1r5
2

Dv a t neck

volume diffusion. ,
6.

2.2.1. Fluid phase


Continuity equation:

between neck growth time t neck and neck radius x neck for
both the surface diffusion and the volume diffusion cases:

s ymg q Fpi q Fn q Ft q Fc ,

3.

where a is the curvature radius, Ds is the surface diffusion


coefficient, Dv is the volume diffusion coefficient, k B is
the Boltzmann constant, T is temperature, g is the surface
tension, and d is the lattice constant.
It has been said that surface diffusion is the main
mechanism that plays in sintering of iron powder at lower
temperatures - 1323 K. w14x. The surface diffusivity of
iron is given by Matsumura w14x by:
Ds s D 0,s exp yEsrRT . ,

7a .

where frequency factor, D 0,s , and activation energy of


volume diffusion, Es , are given as follows:
D 0,s s 2.4 m2rs,
Es s 2.42 = 10 5 Jrmol

T - 1180 K : for a y Fe . ,
7b .

D 0,s s 5.2 = 10y2 m2rs,


Es s 2.21 = 10 5 Jrmol

1180 K - T : for g y Fe . .
7c .

Rotational motion:
I

dv
dt

s Ft rp ,

4.

where, m is the particle mass, Fpi is fluidparticle interaction force acting on a particle, g is gravity acceleration, Fn
is the normal component of the soft sphere contact interaction, Ft is the tangential component of the soft sphere
contact interaction, Fc is cohesive interaction force, I is
moment of particle inertia, v is angular velocity, and rp is
the radius of a particle. Eq. 4. is the corrected angular
momentum equation of Mikami et al. w8x i.e., the righthand side of Eq. 4. was written as < Ft < rp by mistake.. Fpi ,
Fn and Ft are defined in Appendix A. The expression for
Fc , the solid bridging force in the present investigation, is
derived in Section 2.3.3.
2.3. Model for solid bridging force
2.3.1. Mechanism of sintering
The surface diffusion and the volume diffusion mechanisms are the most significant ones in sintering of metal
powders. Kuczynski w2x derived the following relationships

2.3.2. Experimental results by Mikami et al. [1]


Experimental data of solid bridging particles by Mikami
et al. w1x are shown in Fig. 1. Typical SEM images of a
contact point are shown in Fig. 1a., where sintered necks
between particles can be observed. These samples were
prepared by keeping steel shots at 923 or 1123 K for 1 h.
In Fig. 1b., it is shown how heat treatment temperature
affected the neck diameter. The upper solid line in Fig.
1b. shows the neck diameter calculated from the Kuczynskis surface diffusion model, Eq. 5.. In the high temperature region T ) 973 K., the observed neck diameters
agreed well with those calculated. However, in the low
temperature region, the observed neck diameters were
much smaller than those calculated for d p s 200 mm.
Here, the curvature radius for surface roughness of iron
particles was estimated from SEM photographs as about
10 mm. The lower line in Fig. 1b. was calculated from the
surface diffusion model for a s 10 mm. The neck diameters, thus estimated from the neck growth behavior, agreed
fairly well with the visually observed ones. From practical
viewpoints, the surface roughness appears to be an important parameter for estimating the defluidization behavior.

K. Kuwagi et al.r Powder Technology 109 (2000) 2740

29

Fig. 1. Experimental data of solid bridging particles w1x.

2.3.3. Assumptions for calculations


For the present model formulation, the following assumptions are made.

1. The spring constant of the solid bridge, for both


repulsion and attraction, is the same as that of bulk materials.

K. Kuwagi et al.r Powder Technology 109 (2000) 2740

30

Fig. 2. Schematic of surface roughness and magnification of a contact point.

2. A neck is formed between any particles in contact.


3. The neck radius increases with contact time following the Kuczynskis surface diffusion model Eq. 5...
Accordingly, the solid bridging force is expressed as follows:
2
Fc s p x neck
sneck .

8.

4. In particle collision calculation, softened Hookean


spring constant k s 800 Nrm. is adopted due to computation efficiency. For dry particles as well as liquid bridged
particles the spring constant was found not significantly
affecting over the range of k s 80080 000 Nrm w15x.
However, in the case of solid bridging, the neck size at the
end of collision needs to be calculated based on the real
time for collision, i.e., the one for a nonmodified Hertzian
spring, so that the judgement whether the neck is broken
after collision or not is not affected by the adoption of the
softened Hookean spring for computation economy. In
order to compromise them, the neck growth rate during the
deformationrcollision period is adjusted so that the neck
can have the same size at the end of a softened Hookean
collision as that at the end of a nonsoftened Hertzian
collision. This is expressed by the following equation:
t neck s t neck ,0 q tq
neck ,

9.

where tq
neck is an increment of the modified neck growth
time from the beginning of a collision now taking place to
calculate the neck size given by:
tq
neck s t c

td ,Hertz
td ,Hooke

10 .

The collision time from Hertz and Hooke models in the


above equation, td,Hertz w16x and td,Hooke , are given by:
td ,Hertz s 2.44

td ,Hooke s p

1r5

m2

a 2 r

11 .

12 .

where r is the relative velocity at the beginning of


collisionrdeformation and:

a'

E dp
3 1 y n 2 .

13 .

Accordingly, the size of a neck that had the size x neck,0 at


t neck s t neck,0 , from Eq. 5., can be expressed by:
x neck t . s x neck ,0

t neck

/
t neck ,0

1r7

14 .

When a deformation period is over, i.e., t c s t c1 , the neck


is examined if it can still survive. If it can, the additional
time to compensate the above artificial retardation should
be introduced. At contact time t c s t c1 , t neck is set to return
to the time of ordinary speed. Accordingly:
t neck s t neck ,0 q tc1 .

15 .

However, in the present computation, this correction was


not executed. This implies the neck size might be 1.4
times, at most, and the rupture force was two times,
underestimated..
5. Since surface roughness is important in estimating
neck diameter as discussed in Section 2.3.2, surface roughness is approximately taken into account by introducing

K. Kuwagi et al.r Powder Technology 109 (2000) 2740

spherical subgrains into the particle surface as illustrated in


Fig. 2. The radius of a grain, rg , is used as the curvature
radius a in Eq. 5.. For steel shots, Mikami et al. w1x
obtained 10 mm for the curvature radius of surface roughness and 6 mm for apparent diameter of roughness. When
surface roughness exists, a macroscopic contact point may
be broken down to multiple microcontact points. The
number of microcontact points between particles with surface roughness at one contact point is either 3 Case 2:
three-microcontact-point model., or 9 Case 3: nine-microcontact-point model. as shown in the plan of Fig. 2. The
present computation covers only the earliest part of sintering and accordingly further complications, such as the
merge of microcontact points into a larger single contact
points, were not attempted. The cohesive force due to solid
bridging depends on the summation of the cross-sectional
area of necks of the microcontact points. Neck cross-sectional areas calculated by Kuczynskis Eq. 5. are shown
in Fig. 3 for the three cases. The tensile strengths of a
contact point at neck growth time, t neck , for the three cases
are given as follows:

Fc s psneck

56gd 4
k BT

2r7
3

Ds a t neck

Case 1: Smooth surface. ,


Fc s 3psneck

56gd 4
k BT

Ds

16 .
2r7

t neck

10

2.3.4. Determination of sintered (true) agglomerates


To analyze the bed structure at one instant, agglomerates are defined as clusters of primary particles connected
by necks strong enough to survive the whole deformation
and elastic repulsion, which the relevant particles are
experiencing at a certain instant. Accordingly, taking one
primary particle arbitrarily and checking all contact points
successively for all connected particles, an agglomerate
can finally be determined. However, from assumption 2 in
Section 2.3.3, a neck exists in all particle-to-particle contact points but some weak ones of them may be broken
almost immediately. Also, by the effect of macroscopic
bed motion, any already existing necks, not too weak ones,
can be recompressed or restrained and severed again.
Therefore, let us introduce a simplified analysis to judge if
a neck can exist after the rebounce of two particles having
deformation x 0 i.e., the overlapping of two spheres.,
relative velocity 0 and the sum of forces F0 at an
instant t c s 0 we are looking at. The overlapping length x
of two particles satisfies the following momentum balance
equation:
m

' Fc1

s 0.417Fc1

d2 x

dx

qh

d t c2

d tc

q kx s S F ,

Fc s 9psneck

56gd 4
k BT

Ds

/
10

17 .

x s exp y

2r7

t neck

s 1.25Fc1

Case 3: nine y microcontacty point model. ,

F0
k
h

18 .

where a is the curvature radius for Case 1, 100 mm. As


can be seen from the above three expressions, Case 3 nine
microcontact points. gives the largest strength, then Case 1
smooth surface. and Case 2 three microcontact points.
gives the smallest.
6. Neck breakage takes place if either of the forces
acting on the contact point in normal or tangential direction satisfies the following rupture conditions:

21 .

where h is damping coefficient, F is the sum of compression forces acting on two contacting particles. Assuming that F remains constant during an ongoing contact
i.e., F s F0 ., the general solution of Eq. 21. is given
as follows:

Case 2: three y microcontacty point model. ,

31

sy

2m

tc

/ (C q C
2
1

sin

(4 mk y h
2m

tc q b

/
22 .

2m

exp y

2m

= C12 q C22 sin

q exp y

=cos

2
2

2m

tc

(4 mk y h

2m

tc q b

(4 mk y h
/ (C q C 2 m

(4 mk y h
2m

tc

2
1

2
2

tc q b ,

23 .

< Fn < G sneck A neck ,

19 .

b ' tany1 C1rC2 . ,

< Ft < G tneck A neck ,

20 .

where C1 and C2 are arbitrary constants.


Suppose the displacement x 0 and the relative velocity
0 at present t c s 0. are known. From x s x 0 and s 0
at t c s 0, C1 and C2 are determined as:

where A neck in Eqs. 19. and 20. is the sum of the


cross-sectional areas of all microcontact points in each
contact point.
7. Friction coefficient of two particles in contact is
assumed to be infinity so that a contact point can be fixed.

C1 s x 0 y

F0
k

C2 s

2 m 0 q h C1

(4 mk y h

24 .

K. Kuwagi et al.r Powder Technology 109 (2000) 2740

32

Fig. 3. Neck growth by the surface diffusion mechanism w2x.

The time tUc at which the overlap becomes zero can be


approximately determined i.e., assuming expyhr
2 m. tUc . ( 1. from Eq. 22. by substituting x s 0 at t c s tUc
and solving it for tUc :

exp y
q

2m

F0
k

tUc s

tUc

/(

C12 q C22 sin

2m

tUc q b

s 0,

4 mk y h

tm s

/
25 .

2m

(4 mk y h

yb y siny1

yt d - tUc - t d

tU1 is the time where the two particles are supposed to have
come into contact and tU2 is the time when they will
separate. From tU1 and tU2 , the time t m when the fictitious
of the two overlap reaches its maximum is calculated by:

C12 q C22

.,

tU1 q tU2
2

27 .

Now, the maximum repulsion force Frep is obtained as:


h
Frep s kx max s k exp y
t
C12 q C22
2m m

=sin

26 .

where td is duration of contact.


From Eq. 26., we have two solutions for tUc , one
negative tU1 and the other positive tU2 , within ytd - tUc - td .

/(

(4 mk y h
2m

t m q b q S F.

Accordingly, the neck strength can be judged by the


following criterion:
1. If Frep ) Fc , the neck breaks even if the two particles
remain in physical contact.

Table 1
Computation conditions
Particles: steel shot
Number of particles
Particle density
Particle diameter
Young modulus
Poisson ratio
Tensile strength
Normal, sneck
Tangential, tneck
Lattice constant
Boltzman constant
Surface energy

Gas:H 2 :N2 s 3:1


7000
7800 kgrm3
200 mm
8.0 = 10 10 Nrm2
0.28
4.0 = 10 6 Pa
1.6 = 10 7 Pa
2.89 = 10y10 m
1.38 = 10y2 3 JrK
1.72 Nrm

28 .

Temperature
Viscosity
Density
Collision parameters
Restitution coefficient
Friction coefficient
Spring constant
Computational grid parameters
Number of fluid cells
Time step
Fluidized bed parameters
Column size
Distributor
Number of nozzles
Diameter of a nozzle

1273 K
2.29 = 10y5 Pa s
9.67 = 10y2 kgrm3
0.9
`
800 Nrm
21 = 82
1.00 = 10y6 s
0.0153 = 0.06 m
3
0.73 mm

K. Kuwagi et al.r Powder Technology 109 (2000) 2740

2. If Frep - Fc , the neck remains after the rebounce and


the two particles are sintered to remain in the same
agglomerate.
A growth of particles connected by both physical contacts and sintered necks is called a superficial agglomerate and a group of particles connected only by sintered

33

necks is called a true agglomerate in this paper if necessary.


2.4. Computation conditions
Computation condition is depicted in Table 1. Corresponding to the experimental condition w1x, the particle

Fig. 4. Pressure drop of bridging iron.

34
K. Kuwagi et al.r Powder Technology 109 (2000) 2740

Fig. 5. Snapshots showing fluidization behavior of solid bridging particles without surface roughness.

K. Kuwagi et al.r Powder Technology 109 (2000) 2740

Fig. 6. Snapshots showing fluidization behavior of solid bridging particles with surface roughness curvature radiuss10 mm, three microcontact points..
35

36

K. Kuwagi et al.r Powder Technology 109 (2000) 2740

diameter chosen was 200 mm. The bed was fluidized at


superficial gas velocity, u 0 s 0.26 mrs and the bed aspect
ratio LrD was 1.2. It took about 1.5 h for a bed to
defluidize by sintering at 773 K w1x. To save computation
time, the temperature in the present computation was
artificially set at a very high temperature 1273 K so that
sintering could take place much earlier. Sintering was
numerically allowed to start after the first 0.5 s, i.e., after a
free bubbling condition was achieved. For numerical determination of agglomerate size and shape, all particle-to-particle contact points were sampled for x 0 and 0 and Frep
were estimated. Then the criterion defined in Section 2.3.4
was applied to construct agglomerate images.

3. Results and discussions


3.1. Fluidization behaior
The calculated bed pressure drop is shown in Fig. 4a..
Both amplitude of pressure fluctuation and absolute value
of the pressure drop decreased with time. The former
decrease corresponds to the decrease in the effective bed
height due to sintering in the bed bottom region and the
latter corresponds to gas flow through channels where
pressure drop is less than that through a dense bed. The

calculated result for Case 1 without surface roughness.


shows the same tendency as that for Case 3 with surface
roughness. although defluidization took place earlier in
Case 3. These two tendencies, i.e., decreasing amplitude of
pressure fluctuation and decreasing absolute value of pressure drop, can be seen in the experimental result Fig.
4b.. obtained for particles of the same density and diameter as the computed except for temperature and the bed
size. In the experiment, the temperature was much lower
and the bed height was much higher than the computing
condition, and accordingly, the pressure fluctuation due to
bubbling was much longer and the pressure decreasing rate
was much smaller.
Fig. 5 shows snapshots of the fluidization behavior of
solid bridging particles for Case 1 without surface roughness. and Fig. 6 for Case 2 three-microcontact-point
model.. These fluidization behaviors are completely different from that of liquid bridging particles w8x. In the snapshots of Fig. 5, agglomerating fluidization can be found in
the upper region of the bed. In the bottom region between
nozzles, the dead zones are much larger than that of
noncohesive particles w8x. In the snapshot at t s 3.25 s
roughly, the lower half of bed was sintered. On the other
hand, the dead zones between nozzles were smallest, channels were shortest and the sintering was weakest in Case 2
three-microcontact-point model..

Fig. 7. Agglomerates or dead zones. grown on the wall t s 1.21 s..

K. Kuwagi et al.r Powder Technology 109 (2000) 2740


Table 2
Number of contact points
Case

Surface

Smooth

Total contact points


physicalqsintered.
Physical contact
points

7635

Three microcontact points


2262

Nine microcontact points


7883

108 1.41%.

580 25.6%.

157 1.99%.

Fig. 7 shows the dead zone, i.e., agglomerates adhered


to the wall for the three cases at t s 1.21 s. The dead
zones growing between nozzles at the bottom in Case 1
were much larger than those in Case 2. A large agglomerate is hanging from the left wall in Case 1 at this moment,

37

which should have been broken afterwards at least partly


by bubbles as can be seen in Fig. 5. In Case 3, the
agglomerates adhered to the side walls were more stable
and larger than in other cases. These are in accordance
with the order of cohesive force shown in Fig. 3.
3.2. Shape and size of agglomerates
Table 2 shows the number of contacting points. The
number of total contact points is small for the case of
three-microcontact-point model and the number of physical contact points is larger than others. This is understood
to be caused by the weakest cohesive force.
Fig. 8 shows cumulative size distributions of agglomerates at t s 1.21 s, i.e., a. cumulative over size in number

Fig. 8. Cumulative number and weight distributions of agglomerates.

K. Kuwagi et al.r Powder Technology 109 (2000) 2740

38

Fig. 9. Examples of floating agglomerates sampled at t s 1.21 s.

of primary particles in each agglomerate and b. in weight.


From Fig. 8a., the total numbers of agglomerates are 164
for Case 1 without surface roughness., 248 for Case 2
three-microcontact-point model. and 237 for Case 3
nine-microcontact-point model.. In Fig. 8b., solid lines
are for true agglomerates defined in Section 2.3.4 and
broken lines are for superficial agglomerates, i.e., those
contacting superficially by physical contact. Two tendencies can be seen in these figures. There are many large
agglomerates in Cases 1 and 3 and many small agglomerates in Case 2. This tendency again agrees with the order
of strength of cohesive force. Agglomerates containing
more than 100 particles were mostly found in the lower
part of the bed and many of them were sintered to the wall.
On the other hand, most of the agglomerates containing
less than 100 primary particles were fluidized. Sizes of
true agglomerates were of course smaller than superficial
agglomerates. In the case of three-microcontact-point

model the case of the weakest bonding., the difference


between true and superficial agglomerate sizes was larger
than other two cases, indicating there were more contact
points to be broken after collision.
Typical shapes of agglomerates floating in the fluidized
zone were sampled at t s 1.21 s and shown in Fig. 9.
Hypha-shaped agglomerates can be seen in Cases 1 and 2.
On the other hand, agglomerates were rather in a lump in
Case 3.
Average bond number, i.e., the number of neighbor
particles sintered to a particle, per a particle in agglomerates consisting of 3 to 100 particles, is shown in Table 3
for the three cases. Average bond number for Case 3 was
larger than those for the other two. This also corresponds
to the evidence that in Case 3 there were many lumpy
agglomerates. It is known that hypha-shaped clusters are
formed for very fine nanoparticles that have a strong
cohesive force. Nevertheless, hypha-shaped agglomerates
were dominant in the case of the weakest cohesive force
Case 2..

Table 3
Average number of contacting particles
Case

Surface
Contact points

Smooth

Average number
of bonds per a
primary particle

1.67

Rough
Three microcontact points
1.66

Rough
Nine microcontact points
1.81

4. Conclusions
The DEM code SAFIRE. developed for simulation of
agglomerating fluidized beds based on the soft sphere
interaction model was extended to the case of metallic
solid bridging particles. Surface roughness was considered
by introducing small grains on the surface of a particle. By

K. Kuwagi et al.r Powder Technology 109 (2000) 2740

taking into account the cohesiveness of particles at high


temperatures by surface diffusion sintering, the fluidization
behavior of metallic particles was successfully demonstrated in terms of decreasing pressure fluctuation and
upward growing channels between dead zones.
Furthermore, size and shape of agglomerates were investigated. A sintered agglomerate at an instance is defined
as cluster of primary particles where all bridges contained
in it can remain even when on-going collisions relax. To
judge if a neck is remaining or parting after relaxation, the
cohesive force and the repulsion force for each neck were
compared. The sizes and shape of agglomerates including
dead zones depended on the cohesive force. Many large
agglomerates as well as strong dead zones were observed
in the case of strong cohesive force. Hypha-shaped agglomerates were found most in the cases of the three-microcontact-point model and the smooth surface model. On
the other hand, agglomerates showed a lump shape with
few hypha-shaped agglomerates for the case of the ninemicrocontact-point model, although the cohesive force is
almost the same as in the case of a smooth surface.

5. Nomenclature
A neck
a
Ds
D 0,s
dp
E
Es
Fc
Fn
Fpi
Frep
Ft
fi
g
I
k
kB
m
R
rg
rp
T
t
tc
tUc

Neck cross-sectional area m2 .


Curvature radius of surface roughness m.
Surface diffusion coefficient m2rs.
Frequency factor m2rs.
Particle diameter m.
Youngs modulus Nrm2 .
Activation energy of surface diffusion
Jrmol.
Cohesive force solid bridging force. N.
Normal soft sphere interaction at contact
N.
Fluidparticle interaction force acting on
a particle N.
Maximum repulsion force N.
Tangential soft sphere interaction at contact N.
Fluidparticle interaction force acting on
unit volumeNrm3 .
Gravity acceleration mrs 2 .
Moment of particle inertia kg m2rs.
Spring constant Nrm.
Boltzmann constant JrK.
Particle mass kg.
Gas constant Jrmol K.
Curvature radius of a grain m.
Radius of a particle m.
Temperature K.
Time s.
Time for collision analysis s.
Time at which particles overlap becomes
zero s.

t c1

39

x
x neck

Termination time of particle deformation


s.
Duration of collision estimated by Hertz
theory s.
Duration of collision estimated by Hooke
theory s.
Neck growth time s.
Modified neck growth time increment s.
Superficial gas velocity mrs.
Minimum fluidization velocity mrs.
Gas velocity mrs.
Particle velocity mrs.
The relative velocity at the beginning of
collisionrdeformation mrs.
Displacement by collision m.
Neck radius m.

Greek letters
a
DP
d

h
m
g
mf
n
rf
rp
sneck
tneck
v

' E d p r31 y n 2 .
Bed pressure drop Pa.
Lattice constant m.
Voidage y.
Damping coefficient y.
Friction coefficient y.
Surface tension Nrm.
Gas viscosity Pas.
Poissons ratio y.
Gas density kgrm3 .
Particle density kgrm3 .
Tensile strength of a neck Pa.
Shear strength of a neck Pa.
Angular velocity 1rs.

td,Hertz
td,Hooke
t neck
tq
neck
u0
u mf
u

Appendix A
Constitutive equations for fluidparticle and particle
particle interactions are as follows.
A.1. Fluidparticle interaction
For a cell being in a dense condition - 0.8., the
well-known pressure drop equation of Ergun w17x is used
to estimate the force acting on unit volume, f i :
f i s 150

2
1 y . mf u y .

q 1.75 1 y .

d p2

rf u y . < u y <
dp

29 .

where mf is the gas viscosity, u is gas velocity and is


the average particle velocity in a fluid cell. The force, Fpi ,
acting on a single particle in the fluid cell, is obtained by
taking into account the effective buoyancy force. The
effective buoyancy force is caused by the pressure gradient

K. Kuwagi et al.r Powder Technology 109 (2000) 2740

40

of the gas phase. In the present calculations, the pressure


gradient is estimated as yf ir , which can be obtained by
assuming in Eq. 2. that the acceleration term the sum of
the two terms in the left-hand side. for the gas element is
negligible. Then we can write:
Fpi s

ft

dp
y

Vp (

fi

30 .

n
where n is the number of particles in a fluid cell and Vp is
the volume of a particle. For a cell being in the dilute
condition ) 0.8., the modified Stokes-type single particle drag force Wen and Yu w18x. is used for the force Fpi,l
acting on particle l in a cell i:
p
Fpi ,l s CXD r f 2 u y l . < u y l < d p2 ,
31 .
8
n

dx

CXD s y4 .65 C D ,
CD s

24

Ft s m < Fn <

d xt
d tc

xt

if < Ft < F m < Fn < ,

if < Ft < ) m < Fn < .

< xt <

35 .
36 .

The damping coefficient, h , is determined in terms of the


restitution coefficient, e. The duration of collision contact,
td , is obtained approximately by solving a simple equation
of motion for a spring and mass system as follows:
2

h s 2g 'km , g s

ln e .

2
ln e . q p 2

37 .

References

Re - 1000,

Re G 1000,

33 .

where Re s r f d p < u y l <.rmf , d p is the particle diameter, C D is the drag coefficient for an isolated single
particle, CXD is the modified drag coefficient, and l is the
velocity of particle l. The summation of the force acting
on all particles l s 1n. in a fluid cell i is adopted for
Fpi,l in the fluid dynamic calculation.

In the SAFIRE model for the particle collision, we used


a simple model incorporating a Hookes simple linear
spring and a dashpot for both normal and tangential components following Tsuji et al. w12x assuming Coulombs
law of friction.
The normal contact force of soft sphere interaction is
given by:
d xn
d tc

w1x
w2x
w3x
w4x
w5x
w6x
w7x
w8x
w9x

A.2. Particleparticle collision interaction

Fn s k n D x n y hn

Ft s k t D x t y ht

32 .

l q 0.15Re 0.687 .

Re
C D s 0.44

And the tangential contact force of soft sphere interaction is given by:

34 .

w10x
w11x
w12x
w13x
w14x
w15x
w16x
w17x
w18x

T. Mikami, H. Kamiya, M. Horio, Powder Technol. 89 1996. 231.


G.C. Kuczynski, Trans. ASME 185 1949. 169.
B.G. Langston, F.M. Stephans Jr., J. Met. 12 1960. 312.
H. Schenck, W. Wenzel, H.D. Butzmann, Arch. Eisenhuttenwes. 33
1962. 211.
V.W. Wenzel, F.R. Block, E. Wortberg, Arch. Eisenhuttenwes. 43
1972. 805.
Y. Iwadate, M. Horio, Powder Technol. 100 1998. 223.
P.C. Knight, H. Kamiya, M. Horio, J.P.K. Seville, Chem. Eng. Sci.,
in press.
T. Mikami, H. Kamiya, M. Horio, Chem. Eng. Sci. 53 1998. 1927.
Y. Iwadate, M. Horio, in: L.S. Fan, T. Knowlton Eds.., Fluidization
IX, AIChE, 1998, p. 293.
D. Geldart, Powder Technol. 7 1973. 285.
T.B. Anderson, R. Jackson, Ind. Eng. Chem. Fundam. 6 1967. 527.
Y. Tsuji, T. Kawaguchi, T. Tanaka, Powder Technol. 77 1993. 79.
S.V. Patanker, Numerical Heat Transfer and Fluid Flow, Hemisphere, New York, 1980.
G. Matsumura, Acta. Metall. 19 1971. 851.
T. Mikami, PhD thesis, Tokyo University of Agriculture and Technology, 1998, p. 62.
S.P. Timoshenko, J.N. Goodier, Theory of Elasticity, McGraw-Hill,
Singapore, 1970, p. 421.
S. Ergun, Chem. Eng. Prog. 48 1952. 89.
C.Y. Wen, Y.H. Yu, Chem. Eng. Prog., Symp. Ser. 62 1966. 100.

You might also like