1-s2.0-S0009250903002148-mainext

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

i An update to this article is included at the end

Chemical Engineering Science 58 (2003) 3489 – 3498


www.elsevier.com/locate/ces

Modelling the motion of cylindrical particles in a nonuniform ow


Chungen Yin∗ , Lasse Rosendahl, Sren Knudsen Kr, Henrik Srensen
Institute of Energy Technology, Aalborg University, Pon 101, Aalborg East DK-9220, Denmark

Received 28 August 2002; received in revised form 29 January 2003; accepted 22 April 2003

Abstract

The models currently used in computational uid dynamics codes to predict solid fuel combustion rely on a spherical shape assumption.
Cylinders and disks represent a much better geometrical approximation to the shape of bio-fuels such as straws and woods chips. A sphere
gives an extreme in terms of the volume-to-surface-area ratio, which impacts both motion and reaction of a particle. For a nonspherical
particle, an additional lift force becomes important, and generally hydrodynamic forces introduce a torque on the particle as the centre
of pressure does not coincide with the centre of mass. Therefore, rotation of a nonspherical particle needs to be considered. This paper
derives a model for tracking nonspherical particles in a nonuniform ow eld, which is validated by a preliminary experimental study:
the calculated results agree well with measurements in both translation and rotation aspects. The model allows to take into account shape
details of nonspherical particles so that both the motion and the chemical reaction of particles can be modelled more reasonably. The
ultimate goal of such a study is to simulate ow and combustion in biomass-red furnaces using nonspherical particle tracking model
instead of traditional sphere assumption, and thus improve the design of biomass-red boilers.
? 2003 Elsevier Ltd. All rights reserved.

Keywords: Particle; Multiphase ow; Modelling; Hydrodynamics; Biomass combustion; Nonspherical

1. Introduction of nonspherical particles, especially particle deposition in


two-phase ows, e.g., Blaser (2002), Brenner (1964), Fan
Biomass combustion for power generation is now gain- and Ahmadi (2000), Gallily and Cohen (1979), Gradon,
ing more and more attention, but biomass particles dier Grzybowski, and Pilacinski (1988), Higashitani, Inada, and
a lot from traditional pulverized coal particles, which are Ochi (1991). Most of the existing studies are for tiny and
normally approximated as spherical particles in combustion nonreacting particles.
simulation. A biomass particle is generally nonspherical, and This study is conducted with such an ultimate goal that we
looks more like a cylinder or a disk. For example, typical are able to simulate ow and combustion in biomass-red
straw particles used for combustion can be approximated as boilers using nonspherical particle tracking model instead of
cylinders with a mean diameter of about 2 mm and mean traditional spherical particle assumption, so that both motion
length of about 20 mm (maximum length of ∼ 150 mm for history and combustion history of nonspherical particles can
individual particles). The sphere assumption gives an ex- be predicted more reasonably. As the rst stage, a model
treme in terms of the volume-to-surface-area ratio, which for nonspherical particles tracking in a simple nonuniform
impacts both the motion and the reaction rates of particles. two-phase ow is derived in this paper, in which forces, in-
Comparing with a spherical particle, a nonspherical parti- cluding inertia, drag, gravity, lift, a force due to pressure
cle will also be subjected to an additional lift, and a torque gradient and virtual-mass force, are taken into account in
created by hydrodynamic forces, which act on its centre particle force balance; two major torques, torque due to non-
of pressure located some distance away from its centre of coincident centre of mass and centre of pressure (on the
mass. Some studies, theoretical, experimental and applied, latter hydrodynamic forces act) and torque due to the resis-
have been successfully performed for predicting motions tance on a rotating body, are considered in particle rotation
motion. Based on the model, a computer code is developed,
∗ Corresponding author. Tel.: +45-9635-9248; fax: +45-9815-1411. in which, particle rotation and orientation are determined by
E-mail address: chy@iet.auc.dk (C. Yin). including the shape details of nonspherical particles; while

0009-2509/03/$ - see front matter ? 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0009-2509(03)00214-8
3490 C. Yin et al. / Chemical Engineering Science 58 (2003) 3489 – 3498

particle position and translation velocity are still determined


by the point-particle assumption, however they are orienta-
tion dependent. The coupling between particle motion and
ow eld is taken into account with a standard method in
any CFD code. For simplicity, a cylindrical particle is as-
sumed in this paper.
To validate the model, the motion of a cylindrical PVC
particle in originally stagnant water is studied experimen-
tally, where the undisturbed stagnant water becomes nonuni-
form under the eect of the settling of the big PVC particle.
The modelled motion agrees well with the experiments,
in both translation and rotation aspects, which indicates an
encouraging potential to extend the model for practical
applications.

Fig. 1. A cylindrical particle and the coordinate systems.

2. Coordinate systems

Both translation and rotation motions are included in non-


spherical particle tracking, and normally they are described
in dierent coordinate systems: with translation motion in
inertial frame; while with rotation motion in particle frame.
So, the dierent coordinate frames and the transformation
between them are rstly introduced.
Fig. 1 shows the schematic of a cylindrical particle
moving in a general ow eld and the corresponding
coordinate systems. x̃ = [x; y; z] is the inertial coordi-

nates; x = [x ; y ; z  ] is the particle coordinate system
with its origin being at the particle mass centre and

its axes being the principal axis; and x = [x ; y ; z  ]
is the co-moving coordinate with its origin coincid- Fig. 2. Denition sketch of the Euler angles.
ing with that of the particle frame and its axes be-
ing parallel to the corresponding axes of the inertial The time dependence of the Euler angles is related to the
frame. angular velocity according to the equations
The transformation between the co-moving and the par-  
d  
ticle frame coordinates is given by Goldstein (1980), and !x cos − !y sin
 dt 
widely used in nonspherical particle tracking, e.g., by Blaser    
 
(2002), Fan and Ahmadi (2000), and Gallily and Cohen  d    (!x sin + !y cos )= sin   ;
 = (3)
(1979),  dt      

  d
d  !z − cos 
→ → dt

x = A · x ; (1) dt
where, t is time, and [!x ; !y ; !z ] is angular velocities of
where A=[aij ] is the transformation matrix, whose elements the particle with respect to the particle axes.
represent the direction cosines of the particle axes in the Due to the inevitable singularity in determining the change
inertial frame, and can be expressed in terms of Euler angles rate of Euler angle , a transformation matrix in terms of
(; ; ). The Euler angles, as shown in Fig. 2, are dened
following the x-convention of Goldstein:

 
cos cos  − cos  sin  sin cos sin  + cos  cos  sin sin sin 
 
 
A =  −sin cos  − cos  sin  cos −sin sin  + cos  cos  cos cos sin   : (2)
 
sin  sin  −sin  cos  cos 
C. Yin et al. / Chemical Engineering Science 58 (2003) 3489 – 3498 3491

Euler parameters, given by Hughes (1986), is used in this respect to the particle axes and can be expressed as, Ix =
study, Iy = 14 mp a2 + 13 mp b2 , and Iz = 12 mp a2 for a cylinder,
  where a is its radius, and b is its half length (aspect ratio
1 − 2(22 + 32 ) 2(1 2 + 3 ) 2(1 3 − 2 )
   = b=a). [Tx ; Ty ; Tz ] are torques acting on the particle.
A=  2(2 1 − 3 ) 1 − 2(32 + 12 ) 2(2 3 + 1 )  ; Here, it should be emphasized that, in the above equations,
translation motions are expressed in the inertial frame; while
2(3 1 + 2 ) 2(3 2 − 1 ) 1 − 2(12 + 22 ) rotation motions are written in the particle frame.
(4)

where 1 ; 2 ; 3 and  are Euler parameters, and they are 4. Forces on the particle
related to Euler angles by
−  −  4.1. Particle force balance
1 = cos sin ; 2 = sin sin ; (5)
2 2 2 2
The equation governing particle velocity ṽ in a nonuni-
+  +  form ow eld ũ was derived by Maxey and Riley (1983),
3 = sin cos ;  = cos cos : (6)
2 2 2 2
dṽ
In this work, Euler angles are used for assigning the initial p V = 6a1 (ũ − ṽ) + V (p − f )g̃
dt
particle orientations. The corresponding initial Euler param- inertia
drag gravity
eters are obtained using Eqs. (5) and (6), and then used to
evaluate the initial transformation matrix. In the subsequent Dũ 1 d
+ f V + 2 f V (ũ − ṽ)
calculations, the time histories of Euler parameters are cal- Dt dt
culated according to Eq. (7), pressure virtual mass
 d  gradient
1
   
 dt  t
(d=d )(ṽ − ũ)
  !x − 3 !y + 2 !z −6a21  d : (12)
 d2    ((t − )=f )1=2
 
 1 3 !x + !y − 1 !z 
0
 dt  
 =  : (7)
  2 
Basset history term
 d3   − !
2 x  +  !
1 y  + ! z 

  In the above equation, we neglect the Faxen correction term,
 dt 
  −1 !x − 2 !y − 3 !z which becomes signicant only in the event of large curva-
d
ture in the velocity prole. f and  are the density and vis-
dt cosity of the surrounding uid; p and V represent the den-
sity and volume of the particle; and ũ = [ux ; uy ; uz ] describes
the uid undisturbed velocity vector at the point occupied
3. Basic equation of motion
by the particle’s centre of mass, with respect to the inertial
frame. In Eq. (12), the derivative d=dt is used to denote a
The equation of motion of a nonspherical particle moving
time derivative following the moving particle, so that
in an arbitrary ow is given by Gallily and Cohen (1979) and  
used to study wall deposition of small ellipsoids in turbulent dui @ui @ui
= + vj (13)
ows (Fan & Ahmadi, 2000), dt @t @xj
Translation motion is the time derivative of uid velocity component ui . The
dṽ derivative D=Dt is used by contrast to denote the time deriva-
mp = F̃: (8) tives following a uid element, and
dt  
Dui @ui @ui
Rotation motion = + uj (14)
Dt @t @xj
d!x
Ix  − !y !z (Iy − Iz ) = Tx ; (9) is the uid acceleration as observed at the instantaneous cen-
dt
tre of the particle. The dierent terms in Eq. (12) represent
d!y in order, the force needed to accelerate the particle, viscous
Iy − !z !x (Iz − Ix ) = Ty ; (10)
dt Stokes drag, gravity, a pressure gradient force accounting
for the acceleration of the displaced uid, “virtual mass”
d!z
Iz − !x !y (Ix − Iy ) = Tz : (11) force, and Basset history term.
dt The order-of-magnitude estimates presented by Lazaro
In these equations, mp is the particle mass, ṽ = [vx ; vy ; vz ] and Lasheras (1989) indicate that, for small heavy particles
is the translation velocity vector of the particle mass-centre in the dilute regime, drag and inertia eects dominate over
in the inertial frame, and F̃ = [Fx ; Fy ; Fz ] are forces act- those of the pressure eld, the virtual mass, and the particle’s
ing on the particle. [Ix ; Iy ; Iz ] are moments of inertia with history. According to this, only the terms describing inertia,
3492 C. Yin et al. / Chemical Engineering Science 58 (2003) 3489 – 3498

CD is the drag coecient. Chhabra, Agarwal, and Sinha


(1999) collected about 1900 individual experimental data
of drag coecient on nonspherical particles from 19 inde-
pendent studies, covering wide ranges of particle shapes:
cylinders, needles, cones, prisms, discs, cubes and so on,
and encompassing wide ranges of physical and kinematics
conditions as: sphericity of 0.09 –1 and Reynolds number of
10−4 –5 × 105 ; examined critically the performances of ve
commonly used methods with those data, including Haider
and Levenspiel (1989) and Ganser (1993); and found that
the best method was that proposed by Ganser (1993): the
resulting overall mean error is about 16%. The method of
Ganser (1993) is therefore used here,
CD 24
= [1 + 0:1118(Rep K1 K2 )0:6567 ]
Fig. 3. Incidence angle, aerodynamic forces and centre of pressure. K2 Rep K1 K2
drag and gravity are retained in the study of the accumulation 0:4305
+ ; (18)
and dispersion of heavy particles in forced mixing layers 1 + 3305=(Rep K1 K2 )
(Martin & Meiburg, 1994; Raju & Meiburg, 1995).
where, coecient K1 = (dn =(3d) + 2=(3 0:5 ))−1 ; coecient
In this study, since our nal purpose is to track the non- )0:5743
spherical biomass particle in biomass-red furnaces and the K2 = 101:8148(−log
 ; dn is the equal projected area circle
PVC particle used for experiment is a “light” particle, the diameter, dn = 4Se 1 =. Rep is particle Reynolds number,
“virtual mass” force is retained, as well as the force arising dened on the basis of the relative velocity between parti-
due to the pressure gradient in the uid. Additionally, since cle and the surrounding uids and equal-volume sphere di-
we are tracking nonspherical particles while not spheres, a ameter; is particle sphericity, and can be calculated as
lift should also be taken into account, as stated above. There- = s=S; s is the surface of a sphere having the same vol-
fore, the governing equation for evolution of the particle’s ume as the particle and S is the actual surface area of the
velocity then becomes nonspherical particle.
dṽ
p V = F̃D + V (p − f )g̃ + F̃ PG + F̃ VM + F̃ L ; (15) 4.3. Lift force
dt
where, F̃D ; F̃ PG ; F̃ VM and F̃ L represent drag, force due to A lift force can be directly related to a drag force; how-
the pressure gradient in the uid, vitual-mass force and lift, ever, it is still dicult to assign a specic value to the con-
respectively. The forces, F̃ PG and F̃ VM , are determined in the stant of proportionality. Here, a lift formation is used, in
same way as shown in Eq. (12); while F̃D and F̃ L , governed →

by particle’s orientation (e.g., Blaser, 2002; Hoerner, 1965), which the particle major axis ( z  ) is also used to character-
will be introduced below. ize the force.
In this model, except Faxen correction term and Basset Considering that the lift force is orthogonal to the relative
history term, some other terms, such as thermophoretic force velocity (ũ −ṽ) and lies in the plane dened by the particle

and Brownian force are also neglected, as well as Saman major axis direction ( z  ) and the relative velocity and that
lift thanks to the negligible rotation with respect to the major the lift must be invariant under a 180◦ rotation of the particle

axis of particle, i.e., z  -axis in Fig. 1. It might be acceptable major axis z  and vanishes if = 0 or , we express the lift
i
to exclude these forces in the particle force balance of the force as
gas–solid ow in a biomass-red furnace. →
1 z  · (ũ − ṽ) → 
F̃ L = CL f Se 2 z × (ũ − ṽ) × (ũ − ṽ);
4.2. Drag force 2 |ũ − ṽ|
(19)
The drag force is commonly expressed over the entire where, Se 2 is the particle area normal to the direction of the
Reynolds number spectra as lift force and is calculated by
F̃D = 12 CD f Se 1 |ũ − ṽ|(ũ − ṽ); (16) Se 2 = a2 (sin2 i + (4=)2 cos2 i )0:5 ; (20)
where Se 1 is the particle area normal to the direction of the CL is the lift coecient. In this study, it is determined in
drag force. It changes with the incidence angle ( i ) between the way that the ratio of lift to drag meets the relationship
relative velocity (ũ − ṽ) and particle major axis direction (Hoerner, 1965)

( z  ), as shown in Fig. 3, and is determined by |F̃ L |
= |sin2 ( i ) cos ( i )|: (21)
Se 1 = a2 (cos2 i + (4=)2 sin2 i )1=2 ; (17) |F̃D |
C. Yin et al. / Chemical Engineering Science 58 (2003) 3489 – 3498 3493

5. Torques on the particle uid due to particle rotation is (!f − !p )l, where !f is
used to take into account the uid undisturbed vorticity at
Two contributions to the torque acting on the particle are the point occupied by the particle’s centre of mass, with
taken into account in this model: one is caused by nonco- respect to the particle frame, which might be related with
incident centre of mass and centre of pressure, on the latter the local velocity gradient; l is the distance from the centre
hydrodynamic forces act; and the other is due to the resis- of rotation. The resulting torque is thus expressed as
tance on a rotating body.
L=2 L=2
T2 = 2 FD dl = CD f (!f − !p )2 l2 Dl dl: (25)
5.1. Torque due to hydrodynamic forces 0 0

As the centre of pressure does not coincide with the centre In a simple nonuniform ow eld, this torque can be sim-
of mass at nonzero incidence angles, hydrodynamic forces plied. For example, the component with respect to x -axis,
described above, which act at the centre of pressure rather as shown in Fig. 4, can be expressed as
than at the centre of mass, will give rise to a torque on the 1
particle. T2; x = CD f × 2a × !x2 × (2b)4 (26)
64
The centre of pressure is located at a distance behind the
leading edge about 0:25×chordlength in a fairly wide range by using D = 2a, L = 2b, and !p = !x for those notations
of incidence angles (approximately 0 –15◦ ) for a symmetric in Eq. (25). In the same way, the other two components can
at plate (Hoerner, 1965). Generally following this idea, and also be written. However, when shear or vorticity are present
also taking incidence angle, i , and particle aspect ratio, , in the ow eld, this torque will appear in a more complex
into account, a distance between the centre of pressure and form. Dierent with the torque due to hydrodynamic forces,
the centre of mass of a cylinder might be proposed in this this torque is dened directly in the particle frame.
paper, as
xcp = 0:25b(1 − e3(1−) )|cos3 i |; (22)
which is similar in form with that in Rosendahl (2000). 6. Computational procedure
Then, the torque can be expressed as

Based on the above model, a computer code was devel-
T1 = (xcp z  ) × (F̃D + F̃ L + F̃ PG + F̃ VM ): (23) oped to solve the translation and rotation of nonspherical
particles in a simple nonuniform 3D ow eld. Here, all the
Here, it should be stated this torque is expressed in the dimensional equations are used, and can be converted into
inertial frame. The torque calculated in the inertial frame nondimensional forms in case of need. The computational
should be transformed to that in the particle frame, using algorithm mainly consists of the following steps:
T1 = AT1 : (24) (1) solve ow equation in a standard way as described in
any computational uid dynamics code. In this study,
5.2. Torque due to resistance i.e., the Navier-Stokes equation;
(2) set initialization condition of particle;
If the particle has an angular velocity, !, with respect to (3) calculate all the forces acting on the particle, as listed
an axis, it will introduce a torque due to the resistance on in Eq. (15);
the rotating body. This torque will always act to reduce the (4) calculate the two torques acting on the particle, as
angular velocity, as shown schematically in Fig. 4. shown in Eqs. (24) and (25), respectively;
This torque contribution can be obtained by integration (5) solve the basic equation of motion, Eqs. (8)–(11); and
along the particle length. The local relative velocity of the update particle position and orientation;
(6) repeat Steps (3)–(4) until the particle arrives outlet
boundary. In this study, i.e., trapped at the bottom wall
of the water tank;
(7) couple the particle iteration to the ow calculation; then
return to step (1) and continue the procedure until the
desired terminating condition is reached.

In the procedure, steps (1) and (7) are standard meth-


ods and not introduced in this paper; while steps (2) – (6)
are special process for cylindrical particles in a nonuniform
ow, which are quite dierent with traditional spherical par-
Fig. 4. Torque due to the resistance. ticles, and described in details as above.
3494 C. Yin et al. / Chemical Engineering Science 58 (2003) 3489 – 3498

Table 1
A PVC cylindrical particle used in experiments

Material 2a (mm) 2b (mm) d (mm) p (kg=m3 )

Cylinders (sphericity = 0:593; aspect ratio  = 9:242)

PVC 5.41 50 13.00 1366

Fig. 6. The calculated nonuniform ow eld: (a) Flow at the X -middle
plane; (b) Flow at the Y -middle plane.

Fig. 5. A measured trajectory of the PVC particle in originally stagnant


water.

as seen in Fig. 7(a), which implies a correct expression of


7. Results and discussions the hydrodynamic forces.
The calculated y-position, shown in Fig. 7(b), is not
A preliminary experimental study, including several re- plotted against the measurements here, because the repeated
peated measurements for one PVC particle, was carried out measurements themselves show a big discrepancy in the lat-
to validate the above model and the code. As shown in eral trajectories. However, one can see the calculated lateral
Table 1, a PVC particle ( ≈ 10) was used as the study trajectory agrees qualitatively with the observation, as
object, and water was used as the uid (p =f = 1:36). The shown in Fig. 5. Dierent with spherical particles, the lat-
experimental setup consists of a digital video camera JVC eral movement of the cylindrical particle is also important.
GR-DVL9800, a vacuum ejector, a particle dispenser (xed To quantitatively verify the lateral motion of the cylin-
initial orientation 0 = 60◦ , 0 = 0 = 0; and without any drical particle, more experiments need to be planned to
initial release velocity), two 500 W halogen lamps, and a achieve measured lateral trajectories, as well repeatable as
rectangular fall tube (l×w ×h=0:475 m ×0:48 m ×1:5 m), its vertical trajectory.
as described in details by SHrensen and Rosendahl (2001). The translation velocities of the particle, calculated
The motion of the PVC particle, including both transla- with the model, are shown in Fig. 8, from which the
tion and rotation, was tracked and recorded experimentally. terminal velocity of the PVC particle is estimated to be
Fig. 5 shows a sample of the measured trajectory of the par- 0:1265 m=s.
ticle settling in originally stagnant water (the tube is about Fig. 9 shows the evolution of the particle Reynolds num-
half-full of water). ber and the resulting drag coecient, with the estimated
The calculations are conducted on the basis of the same mean values about 1372 and 2.49, respectively. The parti-
particle, tube and initial release condition. The resulting cle Reynolds number in the experiments is close to some
nonuniform ow eld is shown in Fig. 6: the settling of the straw particles of a certain size in real furnaces and thus, the
big PVC particle has obvious eect on the ow eld. The possible behaviour of the straw particles of such a size in a
calculated motions of the PVC particle in this nonuniform real furnace might be partly reected by these experiments
ow eld are plotted against the measurements. thanks to similar particle Reynolds number, particle aspect
ratio. However, the straw particle Reynolds numbers in a
7.1. Translation motion of the PVC particle real furnace covers a huge range, depending largely on straw
particle size distribution. For example, for a big straw par-
Fig. 7 shows the translation motion of the particle. The ticle of equal-volume sphere diameter d ≈ 13 mm in a real
calculated z-position agrees very well with the measurement, gas/straw co-ring furnace of capacity about 400 MW, the
C. Yin et al. / Chemical Engineering Science 58 (2003) 3489 – 3498 3495

Fig. 7. Position of the PVC particle during settling: (a) Vertical position, Z; (b) Lateral position, Y .

Fig. 8. Calculated translation velocities of the PVC particle during settling.

average particle Reynolds number is found to be Rep ≈ 4100 sphere diameter d ≈ 0:16 mm, the average particle Reynolds
over the initial 5 m path and Rep ≈ 1100 over the full fur- number is much lower: Rep ≈ 8 over the initial 5 m path and
nace height; while for a tiny straw particle of equal-volume Rep ≈ 0:7 over the full furnace height.
3496 C. Yin et al. / Chemical Engineering Science 58 (2003) 3489 – 3498

Fig. 9. Calculated particle Reynolds number and drag coecient during settling.

Fig. 10. Calculated rotation of the PVC particle during settling: (a) Angular velocity with respect to dierent particle axes; (b) contributions of the two
dierent torques.
C. Yin et al. / Chemical Engineering Science 58 (2003) 3489 – 3498 3497

Fig. 11. Incidence angle and the orientation of the particle against horizontal plane.

7.2. Rotation motion of the PVC particle number of measurements is needed for possible quantitative
error analysis of the experimental data themselves, then to
The calculated rotation of the PVC particle is plotted in further validate the model.
Fig. 10, including angular velocity and acceleration. In this In short, for the current measurements the model is found
case, the rotation with respect to x -axis of the particle (paral- capable of tracking correctly the PVC particle in a simple
lel to the inertial x-axis in Fig. 5) takes the leading role. The nonuniform ow eld, in both the translation and the rotation
angular velocity oscillates around 0 and has a stable uctua- aspects. With more experiments carried out, the model will
tion with a mean peak about 1:5 s−1 . With respect to another be further validated and ned.
minor axis (y -axis), there is still some rotation, which is
caused by the nonuniform ow eld, but in a less important
role. The rotation with respect to the major axis of the par- 8. Conclusions
ticle can be ignored at all in this ow. Fig. 10(b) shows the
angular acceleration caused by both the two torques. It is In this paper, a model is successfully derived to track the
found that the torque due to resistance on a rotating particle motion of nonspherical particles in a non-uniform ow eld,
also plays an important role in uids of high densities. including translation and rotation aspects. The model holds
Fig. 11 shows the resulting particle orientation history for both high- and low-speed ows. The good agreement of
due to rotation, including the incidence angle and the angle preliminary experiments and the calculated results indicates
between particle major axis and the horizontal plane, from an encouraging potential toward tracking nonspherical parti-
which one can see the calculated orientation history agrees cles in biomass-red furnaces. The model does not increase
well with both the observations and the measurements. Here, signicantly the computational cost.
it has to be stated that, the two angles in Fig. 11 do not nec- However, this model still needs to be further rened for
essarily sum up to 90◦ , since the relative velocity vector is general and practical applications. For example, some other
not always parallel to the inertial Z-axis during the particle’s forces, such as Faxen correction terms and Basset history
settling. The incidence angle oscillates around i = 90◦ . It terms, might be taken into account in some cases when the
can be understood since this point is in fact a stable equilib- nonuniform ow eld around the particle is relatively com-
rium in this kind of ow. When the incidence angle deviates plex, e.g., largely curved; the collisions between particles
from i = 90◦ , the hydrodynamic forces act as a resistance also need to be included when volume fraction of particles
to reduce the deviation; while any deviation from i = 0◦ becomes larger.
will accelerate the rotation away from the point under the
action of hydrodynamic forces, so point i = 0◦ is an unsta-
ble equilibrium state. Notation
The calculated angle between the particle major axis
(z  -axis) and the inertial y-axis also agree well with the a radius of cylinder, m
measured one: the particle’s oscillation frequency is very A = [aij ] transformation matrix
well reproduced; although the amplitude is a little lower at b half-length of cylinder, m
some peaks than the measured one. A collection of a large CD drag coecient
3498 C. Yin et al. / Chemical Engineering Science 58 (2003) 3489 – 3498

CL coecients in evaluating lift ; ; Euler angles


force 0 ; 0 ; 0 initial orientation of particle
d equal-volume sphere diameter, m particle sphericity (=s=S)
dn equal projected area circle diam-  viscosity of the surrounding uid,
eter, m kg=(m s)
F̃ = [Fx ; Fy ; Fz ] force acting on the particle, N f density of the surrounding uid,
F̃D drag force, N kg=m3
* p particle density, kg=m3
F lift force, N
L !x ; !y ; !z angular velocities of particle with
F̃ PG force due to the pressure gradient respect to the particle axes, 1/s
in the uid, N
F̃ VM vitual-mass force, N
Ix ; Iy ; Iz moments of inertia of the parti- References
cle with respect to particle axes,
Blaser, S. (2002). Forces on the surface of small ellipsoidal particles
kg m2 immersed in a linear ow eld. Chemical Engineering Science, 57,
K1 ; K2 two coecients in evaluating CD 515–526.
of Ganser (1993) Brenner, H. (1964). The Stokes resistance of an arbitrary particle—IV
mp particle mass, kg arbitrary elds of ow. Chemical Engineering Science, 19, 703–727.
Rep particle Reynolds number, Rep = Chhabra, R. P., Agarwal, L., & Sinha, N. K. (1999). Drag on non–
spherical particles: An evaluation of available methods. Powder
d|ũ − ṽ|= Technology, 101, 288–295.
s surface of a sphere having the Fan, F. G., & Ahmadi, G. (2000). Wall deposition of small ellipsoids
same volume as the particle, m2 from turbulent air ows—a Brownian dynamics simulation. Journal
S actual surface area of the non- Aerosol Science, 31(10), 1205–1229.
spherical particle, m2 Gallily, I., & Cohen, A. H. (1979). On the orderly nature of the motion of
nonspherical aerosol particles II. Inertial collision between a spherical
Se 1 projected area of particle normal large droplet and an axially symmetrical elongated particle. Journal
to drag force, m2 of Colloid and Interface Science, 68(2), 338–356.
Se 2 projected area of particle normal Ganser, G. H. (1993). A rational approach to drag prediction of spherical
to lift force, m2 and nonspherical particles. Powder Technology, 77, 143–152.
t time, s Goldstein, H. (1980). Classical mechanics (2nd ed.). New York:
Addison-Wesley.
T1 = [T1; x ; T1; y ; T1; z ] torque due to hydrodynamic Gradon, L., Grzybowski, P., & Pilacinski, W. (1988). Analysis of motion
forces, N m and deposition of brous particles on a single lter element. Chemical
T2 = [T2; x ; T2; y ; T2; z ] torque due to resistance on a ro- Engineering Science, 43(6), 1253–1259.
tating body, N m Haider, A., & Levenspiel, O. (1989). Drag coecient and terminal
→ velocity of spherical and nonspherical particles. Powder Technology,
u = [ux ; uy ; uz ] undisturbed gas velocity at the 58, 63–70.
particle mass centre, m/s Higashitani, K., Inada, N., & Ochi, T. (1991). Floc breakup along
ṽ = [vx ; vy ; vz ] translational velocity of the par- centerline of contractile ow to orice. Colloids and Surfaces, 56,
ticle mass in the inertial frame, 13–23.
m/s Hoerner, S. F. (1965). Fluid–dynamics drag. New York: Hoerner Fluid
Dynamics (A book: 20 chapters totally. Published by the author).
V particle volume, m3 Hughes, P. C. (1986). Spacecraft attitude dynamics. New York: Wiley.

x = [x; y; z] inertial coordinate system Lazaro, B. J., & Lasheras, J. C. (1989). Particle dispersion in a turbulent,
→ plane, free shear layer. Physics of Fluids A, 1, 1035–1044.
x = [x ; y ; z  ]

particle coordinate system Martin, J. E., & Meiburg, E. (1994). The accumulation and dispersion

x = [x ; y ; z  ]

co-moving coordinate system of heavy particles in forced two-dimensional mixing layer. I. The
fundamental and subharmonic cases. Physics of Fluids, 6, 1116–1132.
xcp distance between pressure-centre Maxey, M. R., & Riley, J. J. (1983). Equation of motion for a
and mass-centre of the particle, small rigid sphere in a nonuniform ow. Physics of Fluids, 26,
m 883–889.
Raju, N., & Meiburg, E. (1995). The accumulation and dispersion of
heavy particles in forced two-dimensional mixing layer. Part 2: The
Greek letters
eect of gravity. Physics of Fluids, 7, 1241–1264.
Rosendahl, L. (2000). Using a multi-parameter particle shape description
i incidence angle between (ũ − ṽ) to predict the motion of non-spherical particle shapes in swirling ow.
→ Applied Mathematical Modelling, 24, 11–25.
and particle major axis ( z  ) SHrensen, H., & Rosendahl, L. (2001). On the motion of discrete cylinders
 particle aspect ratio (=b=a) in a stagnant uid. Proceedings of the fourth international conference
1 ; 2 ; 3 ;  Euler parameters on multiphase ow—ICMF2001, New Orleans, USA (pp. 1–9).
Update
Chemical Engineering Science
Volume 66, Issue 1, 1 January 2011, Page 117

DOI: https://doi.org/10.1016/j.ces.2010.10.021
Chemical Engineering Science 66 (2011) 117

Contents lists available at ScienceDirect

Chemical Engineering Science


journal homepage: www.elsevier.com/locate/ces

Corrigendum

Corrigendum to ‘‘Modelling the motion of cylindrical particles in a nonuniform


flow’’ [Chem. Eng. Sci. 58 (2003) 3489–3498]
Chungen Yin n, Lasse Rosendahl, Søren Knudsen Kær, Henrik Sørensen
Department of Energy Technology, Aalborg University, 9220 Aalborg East, Denmark

a r t i c l e in f o

Article history:
Received 30 September 2010
Accepted 9 October 2010

The authors inform that there is an incomplete equation in the published article.
Eq. (23)
! ! ! ! ! !
T1 ¼ ðxcp zu Þ  ð F D þ F L þ F PG þ F VM Þ ð23Þ
As expressed in the source code developed for this article, the above equation should be completed with an extra-condition, as follows:
! ! ! ! ! !
T1 ¼ ðxcp zu Þ  ð F D þ F L þ F PG þ F VM Þ ð23Þ
! ! ! !0
If cos a 4 0, then T1 will change its sign, i.e., T1 ¼ ð1ÞUT1 . Here a is the angle between the orientation of the particle axis, z , and the net
! ! ! ! !
hydrodynamic force, F  F D þ F L þ F PG þ F VM .
The first author thanks Prof. Luis Adriano Oliveira from Departamento de Engenharia Mecânica, Universidade de Coimbra, Portugal and
his co-workers, A. Gameiro, B.R. Baliga, C.X. Viegas and X. Viegas, for their valuable discussions and informing me this mistake. To offset the
inconvenience induced by this mistake, a user-defined function (UDF) version of this model applicable to commercial CFD package
(FLUENT) will be available upon readers’ request.

DOI of original article: 10.1016/S0009-2509(03)00214-8


n
Corresponding author. Tel.: + 45 30622577; fax: + 45 98151411.
E-mail address: chy@et.aau.dk (C. Yin).

0009-2509/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2010.10.021

You might also like