Shear Flow
Shear Flow
Shear Flow
net/publication/284880522
CITATIONS READS
6 6,166
2 authors:
All content following this page was uploaded by Yu Feng on 19 January 2016.
Fig. 2 Deaths for the ten leading causes of deaths in all ages of the United States
in 2010 [2]
A generally more accurate sub-critical expression (being within ~Basset can be neglected if qf qp or the Stokes relaxation time
F
6% of experimental data) was given by Clift and Gauvin [49] as: is relatively small compared to the time duration from 0 to t.
Lift forces. The total lift coefficient can be defined as [54]:
24 0:42
CD ¼ 1 þ 0:15Re0:687
p þ Rep < 2 (13)
Rep 42500 FL
1 þ 1:16 CL ¼ (20)
Rep 1 2
vp pdp2
vf ~
qf ~
8
Buoyancy. The buoyancy force executed on the particle The total lift force is composed of forces due to different physical
immersed in the fluid can be expressed as: mechanisms.
~Buoyancy ¼ qf qp Vp ~
F g (14) (1) Saffman Force.
The Saffman Force is a lift force due to local flow velocity
gradients (i.e., shear flows). For small particles translating
Pressure gradient force. The effect of the local pressure
in a linear unbounded shear field, the expression for the
distribution gives rise to a force in the direction of the pressure
~p can be written Saffman force’s magnitude can be expressed as follows for
gradient. Therefore, the pressure gradient force F
Rep < 1:
as [50,51]:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
~Pressure ¼ p d3 rp
F (15) Klf 2 1 @ ~ vf ~ vp
6 p F~Saffman ¼ ~vf ~
vp dp
4 f @y
The pressure gradient force is significant only in high-Reynolds
vf ~
@ ~ vp
number flows. sign (21)
@y
Virtual mass force. The virtual mass force is a result of the
accelerating fluid surrounding the accelerating particle. It has a K ¼ 1.615 is a constant determined based on a numerical
tendency to keep the particle from being accelerated in any direc- integration for creeping flows at low shear rates [55]. The
tion. The origin of the force is that the fluid will gain kinetic ~Saffman is either in positive y-direction or nega-
direction of F
energy at the expense of the work done by an accelerating sub- tive y-direction. For higher Rep values or near-wall regions,
merged particle. It can be expressed as: the expression of the Saffman force needs to be modified.
For creeping flows at low shear rates, the magnitude ratio
F~VM ¼ 1 qf Vp d~vf d~
vp
(16) between F ~Magnus and F ~D can be written as [53]:
2 dt dt
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
F
~Saffman 1:615dp 1 @ ~ vf ~ vp
of which the magnitude is one half of the inertia force for the fluid RSaffman ¼ ¼ (22)
with the same volume of the particle. It is worth noting that F ~vm FD 3p f @y
can be neglected if the relative acceleration, i.e., jðd~ vf =dtÞ
ðd~vp =dtÞj is small. Expressions of the virtual mass force were It can be neglected when the shear rate is very small or Rep
proposed for cylinders and ellipsoids by Loewenberg [52]. is very small.
(2) Magnus Force.
Basset force. The Basset force is due to the lagging boundary
The Magnus force is a rotation-induced lift force acting on
layer development with changing relative velocity when particles
the particle. The expression for such a force is:
move through a fluid. It accounts for viscous effects and addresses
the temporal delay in boundary layer development as the relative
velocity changes with time. It can be rigorously derived from the ~Magnus ¼ 1 pdp2 qf x
F ~ ~vf ~
vp (23)
motion of a single accelerating sphere in the Stokes regime in a 8
quiescent fluid as:
where x~ is the angular velocity vector of the particle.
0 1 ~Magnus and F~D can be written
ð t d~
vf d~
vp The magnitude ratio between F
3 p ffiffiffiffiffiffiffiffiffiffiffi
~Basset ¼ d 2 pqf lf @ ds ffi B ds C as [53]:
F pffiffiffiffiffiffiffiffiffi
ffi Ads (17)
2 p t s F 2
0 ~Magnus dp qf
RMagnus ¼ ¼ jx
~j (24)
FD 24 lf
The Basset force accounts for the effects of past acceleration on
the resistance. In the expression, (–s) represents the time elapsed Therefore, for small particles (e. g., nanoparticles) the Mag-
since past acceleration from 0 to t [53]. Therefore, this force is nus force can be neglected.
due to the diffusion of the vorticity around spherical particles and (3) Wall-Induced Lift Force.
decays as t1/2, which is typical of diffusion processes. It is also The hydrodynamic force on a particle moving in a linear
called the history term/force [43]. shear flow close to a wall is of fundamental significance in
For Rep >1, i.e., beyond the Stokes regime, F ~Basset can be modi- fluid dynamics. The effect of the wall is the strongest when
fied by introducing a correction factor. The magnitude ratio the particle is in contact, while the wall effect decays
between F ~D can be written as [53]:
~Basset and F rapidly with distance from the wall [56]. Specifically, the
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi wall-induced lift force is due to two possible mechanisms
F [46]: (i) the presence of a wall near a particle will break up
~Basset 18 qf ss
RBasset ¼ ¼ (18) the axisymmetry of the wake vorticity field, which results
FD p qp t in an effective lift force on the particle directed away from
the wall; (ii) flow relative to the particle will accelerate
where ss is the Stokes relaxation time defined as: faster in the gap between the particle and the wall rather
8
>
< 1 0:3314a12 exp Rep þ 0:3314a12 ;
FL Rep 40
¼ 10 (33)
Fsaffman >
: 1
0:0524 a Rep 2 ; 100 > Rep > 40
Kurose and Komori [54] proposed new correlations for the lift Experimental and numerically studies of the near-wall and non-
coefficient CL in linear shear flow, considering rotational spherical creeping flow effects on lift forces are ongoing.
particles (1 < Rep < 500):
3.2 Nonspherical Particle Dynamics Theory. Fibers are
CL Rep ; a ; x ¼ K0 a0:9 þ K1 a1:1 particles that have one dimension significantly longer than its other
þ K2 þ K3 a þ K4 a2:0 þ K5 a9:5 x (34) two dimensions [62]. Numerous studies have demonstrated that the
fiber aspect ratios as well as fiber durability are critical factors
where K0 to K5 are constants changing with Rep which can be involved in pathogenicity. Therefore, it is important to accurately
found in the paper by Kurose and Komori [54] Equation (34) is describe the orientation effect of fibers caused by their rotations.
suitable for L > 2dp. Zeng [45] proposed a CL-expression for a Other than a fiber’s translational equation, i.e., Newton’s Second
particle touching the wall (L ¼ 0.5dp) as follow: law, Eulerian rotational equations must be introduced and solved in
order to predict fiber orientation when transported in shear flows.
CL Rep ; a ; x ¼ K0 a0:9 þ K1 a1:1
Complete Numerical Simulation (CNS) Methods. Computa-
þ K2 þ K3 a þ K4 a2:0 þ K5 a9:5 x (35) tional studies of gas-solid flows go back to the very beginning
2=3 3 3
1=3 2 b
As p 6Vp 2
W¼ ¼ ¼ (42)
Ap Ap 1 þ 2b
2=3 1=3
p1=3 ðpr 2 hÞ ðrh2 Þ
Wc ¼ ¼ (43)
2pr ðr þ hÞ 2ðr þ hÞ
pffiffiffiffiffiffiffi
2 3 ab2 D ¼ expð1:4681 þ 12:2584W 20:7322W2 þ 15:8855W3 Þ
We ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! (44)
(47d)
b2 a þ a2 b2
a þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ln
a2 b2 b
Such a correlation is used by ANSYS FLUENT 14.0 as a default for
the nonspherical drag law. However, Gabitto and Tsouris [72]
Other parameters (see Table 2) which can describe the shape fac- stated that Haider and Levenspiel’s correlation showed relatively
tors of nonspherical particles were summarized by Gabitto and poor accuracy for particles with W < 0.67 and fitted for spheres,
Tsouris [72]. isometric solids, and disks with a 5.8% root-mean-square (RMS)
Additionally, crosswise sphericity W? is defined as the ratio deviation.
between the cross-sectional area of the volume equivalent sphere Ganser [76] proposed an expression for nonspherical particle
and the projected cross-sectional area of the considered particle drag coefficient in the form of:
perpendicular to the flow [73]. Furthermore, lengthwise sphericity
Wk is defined as the ratio between the cross-sectional area of the 24 h 0:6567 i
volume equivalent sphere and the difference between half the CD ¼ 1 þ 0:1118 Rep K1 K2
surface area and the mean longitudinal (i.e. parallel to the flow Rep K1 K2
direction) projected cross-sectional area of the considered 0:4305
þ (48)
particle. 1 þ 3305= Rep K1 K2
Analytical solutions for the drag coefficient of a particle
only exist for spheres and spheroids [73] in creeping flow (i.e.,
Rep 1). For higher particle Reynolds numbers or more compli- where
cated flow field, drag coefficients can only be determined by proj 1
experiments or numerical simulations. In creeping flow, drag K1 ¼ dp;eff vol
= 3dp;eff þ 2= 3W0:5 (48a)
coefficients for all bodies decrease inversely proportional to the
particle Reynolds number Rep. 0:5743
A simple correlation of CD was proposed by Leith [74] for non- K2 ¼ 101:8148ð log WÞ (48b)
spherical objects in creeping flow as:
proj
8 1 16 1 dp;eff is the equal projected area circle diameter normal to the flow
CD ¼ pffiffiffiffiffiffiffi þ pffiffiffiffi (45) direction:
Rep W? Rep W
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
proj
Based on experimental curve fitting, Haider and Levenspiel dp;eff ¼ 4Ap;n =p (49)
[75] proposed the correlation:
vol
dp;eff is the equal-sphere volume diameter. Chhabra et al. [77]
24 C compared CD correlations for nonspherical particles with experi-
CD ¼ 1 þ A ReBp þ (46)
Rep D mental data (0.09 < W < 1 and 1 e-4 < Rep < 5e5). It was indi-
1þ cated that the correlation proposed by Ganser [76] is more
Rep
accurate (the mean error is 16%) than Haider and Levenspiel [75]
where and three other correlations [78–80].
A more realistic drag coefficient correlation for cylindrical par-
ticles in an arbitrary direction was proposed by Fan et al. [81]:
A ¼ exp 2:3288 6:4581 W þ 2:4486 W2 (47a)
2 (55)
g dp3 qp qf
Ar ¼ (51)
l2f Equation (55) accounts for the particle orientation over the entire
range of Reynolds numbers [73].
Fan et al.’s [81] correlation apparently has an accuracy of 9.4% (ii) Lift Coefficient Corrections for Non-Spherical Particles.
under the conditions of aspect ratio 4 <b <50, particle density The theoretical and empirical predictions for the lift coefficient
1125 kg/m3 < qp < 8000 kg/m3, and Rep < 40. are very limited compared to the information available for drag
Tran-Cong et al. [82] introduced a surface-equivalent sphere coefficients. The usual assumption has been to assume that the
diameter: lift is proportional to the drag and that the dependence with orien-
qffiffiffiffiffiffiffiffiffiffiffiffiffi tation is given by the so-called “cross-flow principle” as suggested
area
dp;eff ¼ 4Ap =p (52) by Hoerner [83]:
CL
where Ap is the particle surface area. They also proposed a particle ¼ sin2 ai cos ai (56)
circularity (or surface sphericity) c: CD
area
pdp;eff where the incidence angle ai between the flow direction and the
c¼ (53) long axis of the fiber-like particle is shown in Fig. 5.
Pp
A more “accurate” correlation for the ratio between CL and CD
was proposed by Mando and Rosendahl [84] as follows:
where Pp is the projected perimeter of the particle in its direction
of motion. The drag coefficient for nonspherical particles is then CL sin2 ai cos ai
area vol ¼ 30 < Rep < 1500 (57)
given for 0.15 < Rep < 1500, 0.4 < c < 1.0, 0.8 < ðdp;eff =dp;eff Þ
CD 0:65 þ 40Re0:72
< 1.5 as: p
!2 !0:687 3 However, Mao [85] summarized that the experimental data of the
area area
24 dp;eff 0:15
41 þ pffiffiffi
dp;eff
5
CD ¼ Rep lift force is not sound and hence there is no generally acceptable
vol
Rep dp;eff vol
c dp;eff correlation for the lift coefficient.
! Corrections of forces for nonspherical particles can improve the
area 2
dp;eff accuracy of numerical simulation results when using the effective
0:42 vol diameter method. However, in order to completely capture the
dp;eff
þ 2 !1:16 3 (54) orientation effect on the transportation and deposition of non-
area spherical particles, additional equations, i.e., the Eulerian equa-
pffiffiffi4 4
dp;eff
5
c 1 þ 4:25 10 vol
Rep tions for rotational motion of the rigid body, are necessary. Thus,
dp;eff
the Euler–Lagrangian with Euler rotational equation method is
now introduced.
Loth [47] took into account the orientation effect of spheroid (1) Euler–Lagrangian with Euler Rotational Equation Method
particles by measuring CD in three axisymmetric axial directions. (EL–ER).
Due to the linearity of the drag in creeping flow conditions, the This approach is a good compromise between numerical
drag force can be obtained based on a simple combination of the simulation accuracy and computational cost for simulating
individual components in three axisymmetric axial directions for nonspherical particle transport and deposition. Based on the
a spheroid particle in an arbitrary direction towards the flow. Euler–Lagrange method, the Eulerian rotational equations for
However, for higher Rep numbers, the linear assumption may not nonspherical particles are solved with user-enhanced programs
be validated. Loth [47] also provided Stokes correction factors for [86]. It is worth emphasizing that the EL–ER method employed in
spheroids of different aspect ratios in different directions. the present study is a one-way coupled method which may not be
Based on empirical data for fixed and freely falling particles, able to accurately predict dense particle suspensions due to the
Hoelzer and Sommerfeld [73] came up with a CD expression for lack of particle-particle interaction representation (see Sec. 4).
nonspherical particles: The details of this method are shown as follows:
In order to build the computational model for ellipsoidal parti-
cle and fiber transport and deposition in different flow fields, the
Euler–Lagrange method is employed. Three different Cartesian
coordinates are introduced as well as Euler’s quaternions to
describe the particle dynamics. The governing equations of the
continuous phase as well as the translation equations for ellipsoi-
dal particles are in the global coordinate frame, while rotational
equations for ellipsoidal particles are in a body-fixed coordinate
frame (see Figs. 6 and 7). Transformation matrices between differ-
ent coordinates in terms of Euler’s quaternions are introduced as
well.
The transformation from a given Cartesian coordinate system to
another one can be carried out by means of three successive
rotations performed in a specific sequence [87]. The Euler angles
(u, h, w) are then defined as the three successive angles of rota-
tion. The rotation sequence is as follows (see Fig. 6):
!
in which !
x ¼ ðx; y; zÞ, x0 ¼ ðx0 ; y0 ; z0 Þ and A¼[aij] is the transfor-
Fig. 6 The rotations defining the Euler angles [87] mation matrix which can be calculated as:
2 32 3
(b) Rotating ngf about the n-axis counterclockwise by an angle cos w sin w 0 1 0 0
h to produce another intermediate coordinate, i.e., the 6 76 7
A ¼ 4 sin w cos w 0 54 0 cos h sin h 5
n0 g0 f0 -axes. 0 0 1 0 sin h cos h
(c) Rotating n0 g0 f0 -axes counterclockwise by an angle w about 2 3
the f0 -axis to produce the desired x0 y 0 z 0 -system. cos / sin / 0
6 7
Therefore the coordinate transformation between axes x0 y 0 z 0 4 sin / cos / 0 5 (59)
and axes xyz can be expressed as: 0 0 1
!0
x ¼A!
x (58) Simplifying Eq. (59) yields:
2 3
cos w cos / cos h sin / sin w cos w sin / þ cos h cos / sin w sin w sin h
A ¼ 4 sin w cos / cos h sin / cos w sin w sin / þ cos h cos / cos w cos w sin h 5 (60)
sin h sin / sin h cos / cos h
/þw h
In order to avoid singularities [19], the transformation matrix A g ¼ cos cos (63d)
2 2
has to be rewritten by introducing Euler’s quaternions [87]. The
Euler quaternions (e1, e2, e3, g) are defined as: where (e1, e2, e3, g) also satisfies:
/w h
e1 ¼ cos sin (63a) e21 þ e22 þ e23 þ g2 ¼ 1 (64)
2 2
1 1
r!
vf ¼ 0 (69)
g ¼ 6 ð1 þ a11 þ a22 þ a33 Þ2 (66a)
2 !
Dv
vf þ qf !
¼ rp þ lf r2 !
f
1 qf g (70)
e1 ¼ ða23 a32 Þ (66b) Dt
4g
1 DT
e2 ¼ ða31 a13 Þ (66c) qf cp ¼ kf r2 T þ U (71)
4g Dt
1 In Eq. (71), U is the dissipation function. For Newtonian fluids, U
e3 ¼ ða12 a21 Þ (66d)
4g can be expressed as:
@uf 2 @vf 2 @wf 2 @uf @vf 2
(2) For g ¼ 0: U ¼ lf 2 þ2 þ2 þ þ
@x @y @z @y @x
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ a11 @wf @vf 2 @uf @wf 2
e1 ¼ 6 (67a) þ þ þ þ
2 @y @z @z @x
a12
e2 ¼ (67b) 2 @uf @vf @wf 2
2e1 lf þ þ (72)
a23 3 @x @y @z
e3 ¼ (67c)
2e2
Clearly, for isothermal fluid flow only Eqs. (69) and (70) are
necessary.
The Euler quaternions will be used instead of Euler angles for
modeling nonspherical particle kinematics, indicating the particle 3.4 Ellipsoidal Particle Equations. As shown, for spherical
orientation in each time step. particles, only translation equations are needed for tracking each
As shown in Figs. 7 and 8, three different coordinate systems of the spheres. However, due to the anisotropic shapes of non-
were introduced for fiber kinematics modeling: spherical particles, the orientation has a strong impact on forces
(1) Space-fixed frame xyz: The inertial coordinate. executed on the particle [88]. Thus, as discussed, in order to cor-
(2) Body-fixed frame x0 y 0 z 0 : The particle coordinate system rectly predict forces acting on ellipsoidal and fiber-like particles
with its origin being at the particle mass center and its axes leading to accurate particle trajectories, Euler rotation equations
being the principal axis. are necessary for tracking nonspherical particles combined with
(3) Co-moving frame x00 y00 z00 : The origin coinciding with that translation equations. Based on existing papers [19,51,84,89,90],
of the body-fixed frame x0 y 0 z 0 and its axes being parallel to nonspherical particle equations can be formulated as follows.
the corresponding axes of the space-fixed frame xyz. Translational Equation (in the Global xyz-Frame)
Specifically, there is only relative rotation motion between
x0 y0 z0 and x00 y00 z00 , while there is only relative translation motion d~
vp ~ ~L þ F
~BM þ F
~g þ F
~other
mp ¼ FD þ F (73)
between x00 y00 z00 and xyz. Hence, according to Eq. (58), the trans- dt
formation between coordinate x0 y0 z0 and x00 y00 z00 can be rewritten
as: ~D is the drag force, F
where mp is the mass of the particle, F ~L is the
~BM is the Brownian motion induced force, F
lift force, F ~g is the
!0 ! gravity, and F ~other are other forces which may need to be
x ¼ A x00 (68) considered.
Drag force. Specifically, for ellipsoidal particles in Stokes
flow, the drag force is:
~D ¼ lf pap ½K !
F vf !
vp (74)
½K ¼ A1 ½K 0 A (75)
0 2 3
K11 0 0
0
½K ¼ 4 0 0
K22 0 5 (76)
Fig. 8 Euler angles between coordinate x0 y0 z0 and coordinate 0
x00 y00 z00 0 0 K33
Ellipsoidal particle shape jii value for Major axis parallel to flow direction jii value for Minor axis parallel to flow direction
Oblate exact (b < 1) ð4=3Þb1=3 1 b2 ð8=3Þb1=3 b2 1
" # " #
1 2b2 cos1 b 3 2b2 cos1 b
pffiffiffiffiffiffiffiffiffiffiffiffiffi þb b pffiffiffiffiffiffiffiffiffiffiffiffiffi
1 b2 1 b2
Oblate approximate (0.25 < b < 1) 4 b 1=3 3 2b 1=3
þ b þ b
5 5 5 5
Disk (b < 0.25) 8 1=3 16 1=3
b b
3p 9p
Prolate exact (b > 1) ð4=3Þb1=3 b2 1 ð8=3Þb1=3 b2 1
2 pffiffiffiffiffiffiffiffiffiffiffiffiffi 3 2 p ffiffiffiffiffiffiffiffiffiffiffiffiffi 3
2b2 1 ln b þ b2 1 2b2 3 ln b þ b2 1
4 pffiffiffiffiffiffiffiffiffiffiffiffiffi 5b 4 pffiffiffiffiffiffiffiffiffiffiffiffiffi 5þb
b2 1 b2 1
Prolate approximate (6 > b > 1) 4 b 1=3 3 2b 1=3
þ b þ b
5 5 5 5
Needle Prolate exact (b > 6) ð2=3Þb2=3 ð4=3Þb2=3
lnð2bÞ 1=2 lnð2bÞ 1=2
1 @vf ;i @vf ;j
Dij ¼ þ (81)
2 @xj @xi Fig. 9 Saffman lift force for a particle in linear shear flow
Shear velocity gradient Related values of i and j Related lift force transformation matrix [B]ij
2 3
@vf ;x =@y i ¼ 1, j ¼ 2 1 0 0
½B 12 ¼ 4 0 1 05
0 0 1
2 3
@vf ;x =@z i ¼ 1, j ¼ 3 1 0 0
½ B 13 ¼ 4 0 0 1 5
0 1 0
2 3
@vf ;y =@x i ¼ 2, j ¼ 1 0 1 0
½ B 21 ¼ 4 1 0 0 5
0 0 1
2 3
@vf ;y =@z i ¼ 2, j ¼ 3 0 0 1
½B 23 ¼ 4 1 0 05
0 1 0
2 3
@vf ;z =@x i ¼ 3, j ¼ 1 0 1 0
½B 31 ¼ 4 0 0 15
1 0 0
2 3
@vf ;z =@y i ¼ 3, j¼2 0 0 1
½ B 32 ¼ 4 0 1 0 5
1 0 0
(84) jB T
P0;z0 ¼ 6plf dst;k (88b)
Cc
Here, i and j denote the three directions of the global coor-
dinates (i, j ¼ 1, 2, 3). The lift force transformation matrix In which dst;k and dst;? denote the Stokes diameters for an ellipsoi-
½ B ij according to different velocity gradients is shown in dal particle oriented parallel and perpendicular to the main drag
~L for ellipsoidal particles
Table 4 . Thus, the total lift force F direction, respectively, and can be expressed as:
in a general flow field can be expressed as:
4 2
X
3 b 1
~L ¼
F ~L ði; jÞ
F ði 6¼ jÞ (85) dst;k ¼ deff;v " 3 # (89a)
qffiffiffiffiffiffiffiffiffiffiffiffiffi
i;j¼1 2b2 1 2
pffiffiffiffiffiffiffiffiffiffiffiffiffi ln b þ b 1 b
b2 1
where F ~L ði; jÞ can be obtained using Eq. (84). Such a
method for calculating the lift force acting on ellipsoidal 8 2
b 1
particles can be employed to solve for the fluid-particle dst;? ¼ deff;v " 3 # (89b)
qffiffiffiffiffiffiffiffiffiffiffiffiffi
dynamics in human respiratory systems with their complex 2b2 3 2
airflow fields. pffiffiffiffiffiffiffiffiffiffiffiffiffi ln b þ b 1 þ b
b2 1
(2) Drew et al. [95] and Auton [96]
For a general form of the Saffman force, Drew et al. [95]
provided an expression for F ~L in linear shear flow in the Euler Rotation Equations (in body-fixed x0 y0 z0 -frame)
form of:
dxx0
Ix0 xy0 xz0 Iy0 Iz0 ¼ Tx0 (90a)
! dt
FL ¼ CL Vp qf !vp !
vf r !
vf (86)
dxy0
Iy0 xz0 xx0 ðIz0 Ix0 Þ ¼ Ty0 (90b)
dt
For spherical particles, the lift coefficient CL is equal to
0.50003 [95]. However, for nonspherical particles, the mag- dxz0
Iz 0 xx0 xy0 Ix0 Iy0 ¼ Tz0 (90c)
nitude of CL has to be measured experimentally or deter- dt
mined numerically.
Here, Ix0 ; Iy0 ; Iz0 are particle
moments
of inertia about the princi-
Brownian motion induced force. As outlined, for spherical par- pal axes x0 , y0 , and z0 ; xx0 ; xy0 ; xz0 are particle angular velocities
ticles, FBM;i encapsulates the components of the Brownian-motion with respect to the principal axes x0 , y0 , and z0 ; and Tx0 ; Ty0 ; Tz0
induced force in the i-th direction, which can be modeled as a are torques acting on the particle with respect to the principal axes
Gaussian white noise process (see Eq. (25)). x0 , y0 and z0 . For ellipsoidal particles, Ix0 ; Iy0 ; Iz0 can be written as:
For ellipsoidal particles with aspect ratio b, the Brownian
motion induced force in three principal directions, i.e., x0 y0 z0 -axes,
1 þ b2 a2p
can be expressed as [57,97]: Ix 0 ¼ Iy 0 ¼ mp (91a,b)
5
rffiffiffiffiffiffiffiffiffiffiffi
pP0;i0 2a2p
FBM;i0 ¼ fi (87) Iz0 ¼ mp (91c)
Dt 5
!
2
Wx0 x0 Wx0 y0 Wx0 z0
3 d!
x ¼ A1 d x0 (100b)
h !T i
6 7 1
½W x0 y0 z0 ¼ 4 Wy0 x0 Wy0 y0 Wy0 z0 5 ¼ rv!p rvp (94) substituting Eqs. (100a) and (100b) to Eq. (99) yields:
2 x0 y0 z0
Wz0 x0 Wz0 y0 Wz0 z0
! !
dv0p ¼ A ½G xyz A1 d x0 (101)
Specifically,
Relationship between Angular Velocities and Euler’s
1 @vp;z0 @vp;y0 Quaternions. The angular velocity components were defined in
Dz 0 y 0 ¼ þ (95a)
2 @y0 @z0 the body-fixed x0 y0 z0 -frame:
1 @vp;x0 @vp;z0 dw d/
Dx 0 z 0 ¼ þ (95b) xx 0 ¼ þ cos h (102a)
2 @z0 @x0 dt dt
dh d/
1 @vp;z0 @vp;y0 xy0 ¼ cos w þ sin h sin w (102b)
Wz0 y0 ¼ (95c) dt dt
2 @y0 @z0 d/ dh
xz 0 ¼ sin h cos w sin w (102c)
dt dt
1 @vp;x0 @vp;z0
Wx0 z0 ¼ (95d)
2 @z0 @x0 Substituting Eqs. (67a)–(67c) into Eqs. (102a)–(102c) yields:
2 3
1 @vp;y0 @vp;x0 de1
Wy0 x0 ¼ (95e) 6 dt 7
2 @x0 @y0 6 7 2 3
6 de2 7 gxx0 e3 xy0 þ e2 xz0
6 7
Additionally, in Eqs. (92a)–(92c), a0, b0, and c0 are given as: 6 dt 7 1 6 6 e3 xx0 þ gxy0 e1 xz0 7
7
6 7
6 de3 7 ¼ 2 6
4 e2 xx0 þ e1 xy0 þ gxz0 5
7 (103)
" pffiffiffiffiffiffiffiffiffiffiffiffiffi# 6 7
b2 b b b2 1 6 dt 7
6 7 e1 xx0 e2 xy0 e3 xz0
a0 ¼ b0 ¼ 2 þ ln pffiffiffiffiffiffiffiffiffiffiffiffiffi (96a) 4 dg 5
b 1 2 b2 1 3=2 b þ b2 1
dt
" pffiffiffiffiffiffiffiffiffiffiffiffiffi#
2 b b b2 1 Equation (103) is used for updating Euler’s quaternions during
c0 ¼ 2 3=2 ln pffiffiffiffiffiffiffiffiffiffiffiffiffi (96b)
b 1 b2 1 b þ b2 1 each time step after the angular velocities have been updated.
Velocity Gradient Transformation. In Eqs. (93) and (94), flow 3.5 Two-Fluid Euler-Euler Methods. Euler–Euler methods
velocity gradients in the body-fixed frame can be obtained by solve two sets of algebraic conservation equations for two differ-
transforming the velocity gradients from the global frame. The ent fluids simultaneously for each node in the field [44]. The con-
velocity gradient transformation [G] from the global frame xyz to cept of phasic volume fraction relies on continuous functions of
the body-fixed frame x0 y0 z0 reads: space and time where the sum of volume fractions of each
phase is equal to one. Among many Euler–Euler methods, e.g.,
the Volume-of-Fluid (VOF) model or the Homogeneous Mixture
½G x0 y0 z0 ¼ A ½G xyz A1 (97) model [42], the two-fluid model is widely used in simulating
DownloadedViewFrom:
publicationhttp://biomechanical.asmedigitalcollection.asme.org/
stats on 10/30/2013 Terms of Use: http://asme.org/terms