Vous êtes sur la page 1sur 8

Advances in Engineering Software 42 (2011) 604611

Contents lists available at ScienceDirect

Advances in Engineering Software


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

Binary collision of drops in simple shear ow at nite Reynolds numbers: Geometry and viscosity ratio effects
M. Bayareh a,, S. Mortazavi b
a b

Department of Mechanical Engineering, Islamic Azad University, Lamerd Branch, Lamerd, Iran Department of Mechanical Engineering, Isfahan University of Technology, Isfahan, Iran

a r t i c l e

i n f o

a b s t r a c t
The collision of two equal-size drops in an immiscible phase undergoing a shear ow is simulated over a range of viscosity ratios (k) and different geometries. The full NavierStokes equations are solved by a nite difference/front tracking method. Based on experimental data, different cases were simulated by changing the offset, size of drops, and viscosity ratio. The distance between drop centres along the velocity gradient direction (z) was measured as a function of time. It was found that Dz increases after collision and reaches a new steady-state value after separation. The values of Dz, during the interaction, increases with increasing initial offset. Our results show that the time of approaching of drops at low initial offset is greater than the other cases, but the maximum deformation is the same for equal drop sizes. The deformation decreases with decreasing the size of drops. As the initial offset increases, the drops rotate more quickly and the available contact time for lm drainage decreases. We found that the trajectories of drops in the approaching stage are different owing to the different initial offsets. However, after the drops come into contact, it observed that they follow the same trajectories. As k increases, the drops rotate more slowly, and the point at which the drops separate is delayed. The trajectories of drops become more symmetric with the increased k. 2011 Elsevier Ltd. All rights reserved.

Article history: Received 2 October 2010 Received in revised form 4 October 2010 Accepted 26 April 2011 Available online 18 May 2011 Keywords: Two-phase ow Front tracking Reynolds number Weber number Capillary number Offset Drainage time

1. Introduction The ow of uids composing immiscible liquidliquid mixtures is of great interest in a wide range of research areas, such as foods, polymers, pharmaceuticals and cosmetics. The systematic study of drop deformation and break-up was initiated by Taylor [1]. Taylors experiments revealed the existence of steady rounded and pointed drops, as well as bursting drops of the same type, depending on the viscosity ratio and the capillary number. Cox and Mason [2] have conrmed and extended Taylors results. Most efforts have focused on drop deformation for clean uid systems in which interfacial tension gradients are absent. For these drops, experiments, asymptotic analysis, and numerical simulations have been performed. Rallison [3] studied the time-dependent deformation and burst of a viscous drop in an arbitrary shear ow at zero Reynolds number. Magna and Stone [4] reported the time-dependent interaction between two buoyancy-driven deformable drops in a low Reynolds number ow. Calculations and experiment with initially offset drops showed that the axisymmetric drop conguration was stable for sufciently deformable drops. They introduced three modes for
Corresponding author.
E-mail addresses: mbayareh@iaulamerd.ac.ir (M. Bayareh), saeedm@cc.iut.ac.ir (S. Mortazavi). 0965-9978/$ - see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.advengsoft.2011.04.010

lm drainage between the drops: rapid drainage, uniform drainage and dimple formation. As the separation distance between the two drops decreases, the mode of lm drainage may change from rapid drainage to uniform drainage and eventually a dimple may form. Zhou and Pozrikidis [5] studied the ow of periodic suspension of two-dimensional viscous drops in a channel that was bounded by two parallel plane walls. They found that there exists a critical capillary number below which the suspension exhibits stable periodic motion, and above which the drops elongate and tend to coalesce. Feng et al. [6] reported the results of a two-dimensional nite element simulation of the motion of a circular particle in a Couette and Poiseuille ow. They showed that a neutrally buoyant particle migrates to the centreline in a Couette ow and the stagnation pressure on the particle surface is particularly important in determining the direction of migration. Li et al. [7] studied the motion of two-dimensional, doubly periodic, dilute and concentrated emulsions of liquid drops with constant surface tension in a simple shear ow. Their numerical method is based on a boundary integral formulation. They showed that the shear ow is able to stabilize a concentrated emulsion against the tendency of the drops to become circular and coalesce, thereby allowing for periodic evolution. Loewenberg and Hinch [8] did a three-dimensional simulation of a concentrated emulsion in shear ow, at zero-Reynolds-number and nite capillary numbers. The collision of two

M. Bayareh, S. Mortazavi / Advances in Engineering Software 42 (2011) 604611

605

Nomenclature A area (m2) b minor axis of the deformed drop (m) Ca capillary number D diameter of drop (m) De Taylor deformation parameter Fr force due to surface tension on each element of front (N) h the grid spacing (m) H the height of the channel (m) l major axis of the deformed drop (m) We Weber number x (x, y, z) position in Eulerian coordinate (m) X (X, Y, Z) position in Lagrangian coordinate (m) n unit vector normal to the drop surface P pressure (Pa) R radius of undeformed drop (m) Reb bulk Reynolds number Rep particle Reynolds number s a closed contour (m) t tangent vector to each element (m) t dimensionless time u uid velocity vector (m/s) ub bottom wall velocity (m/s) ut top wall velocity (m/s) Greek letters b indicate two or three-dimensional in NavierStokes equations d delta function dA a surface element (m2) D half of the initial distance between drop centres (m) Ds a short front element (m) Dsl the area of the element (m2) Dt time step (s) /ijk an approximation to the grid value /g /f the interface quantity /g the grid value /l a distance approximation to the front value /f _ c shear rate (1/s) j twice the mean curvature for three-dimensional ows k viscosity ratio l viscosity (N s/m2) l0 viscosity of ambient uid (N s/m2) ld viscosity of drop (N s/m2) q density (kg/m3) q0 density of ambient uid (kg/m3) qd density of drop (kg/m3) r surface tension (N/m) xlijk the weight of grid point ijk

equal-sized drops immersed in an immiscible liquid undergoing a shear ow in a Couette apparatus was investigated by Guido and Simeone [9] over a range of capillary numbers. Trajectories of the drops and their deformation were presented. Mortazavi and Tryggvasson [10] studied the motion of a drop in poiseuille ow. They simulated the motion of many drops at nite Reynolds numbers. Esmaeeli and Tryggvason [11] simulated the motion of two-and three-dimensional buoyant bubbles at nite Reynolds numbers. They showed that the rise Reynolds number is nearly independent of the number of bubbles, the velocity uctuations in the liquid (the Reynolds stresses) increase with the size of the system. Cristini et al. [12] simulated the drop break-up and coalescence by an adaptive mesh algorithm. The surface discretization was fully adaptive, thus providing accurate resolution for a highly deformed drop shapes. Their algorithm was used to study drop break-up in shear ow. Balabel et al. [13] introduced a numerical model based on the level set method for computing the unsteady motion of droplets in a channel. Yoon et al. [14] investigated experimentally the effect of the dispersed to continues-phase viscosity ratio on the ow-induced coalescence of two equal-sized drops with clean interfaces. Effects of inertia on the rheology of a dilute emulsion of drops in shear ow were investigated by Zhao [15] using direct numerical simulation. The drop shape and ow were computed by solving the NavierStokes equations in two phases using front tracking method. Sibillo et al. [16] investigated the deformation and break-up of a drop in an immiscible equiviscous liquid undergoing unbounded shear ow. They showed that wall effects can be exploited to obtain nearly monodisperse emulsions in microconned shear ow. Zhao [17] investigated the drop break-up in dilute Newtonian emulsions in simple shear ow by high-speed microscopy over a wide range of viscosity ratios, focusing on high capillary numbers. He showed that the nal drop size distribution intimately links to drop break-up mechanism, which depends on viscosity ratio and capillary number. Lac and Biesel [18] used a boundary integral formulation to investigate the collision of two identical capsules in simple shear ow. Each capsule consisted of a viscous liquid drop enclosed by

an elastic membrane. The hydrodynamic interaction was characterized by an irreversible cross-ow displacement after the capsules had crossed each other. They showed for sufciently spaced trajectories, the capsules exhibit negative deections which displace them to closer streamlines. Bayareh and Mortazavi [19,20] simulated the migration of a drop and the interaction of two drops in shear ow at nite Reynolds numbers using a nite difference/front tracking method. They showed that the proper dimensionless parameter for the interfacial tension is the capillary number; the interaction between deformable drops increases the cross-ow separation of their centres. At different values of capillary numbers, the deformation of drops are maximum when the drops are pressed against each other and minimum when they are drawn a part. Their results agreed qualitatively with experimental and theoretical data. It is important to understand and control the size and size distribution of the dispersed drops because the macroscopic properties of the emulsion depend on them. The nal size distribution is determined by a balance between ow-induced break-up and coalescence. The majority of numerical simulations are based on the interaction of two deformable drops in a shear ow, the drainage of the thin lm between two colliding drops and the problems of coalescence of two deformable drops. Most of experimental efforts are based on blending studies that analyses the drop size distribution of a blend or a concentrated emulsion. In this article, we present numerical simulations describing the effects of geometry and the viscosity ratio on the motion of a pair of drops under simple shear ow at nite Reynolds numbers. 2. Mathematical formulation The governing equations for the motion of unsteady, viscous, incompressible, immiscible two- uid systems are the Navier Stokes equations in conservative form:

@ qu r:quu rP r:lru ruT r @t

jndb x Xds:
1

606

M. Bayareh, S. Mortazavi / Advances in Engineering Software 42 (2011) 604611

This equation is valid for the whole ow eld, even if the density eld, q, and the viscosity eld, l, change discontinuously. Here u is the uid velocity, p is the pressure, and r is the surface tension coefcient. db is a two- or three-dimensional delta function (for b = 2 and b = 3) respectively. j is the curvature for two-dimensional ows and twice the mean curvature for three-dimensional ows. n is a unit vector normal to the drop surface pointing outside of the drop. x is the position in Eulerian coordinate and X is the position of front in Lagrangian coordinate. Mass conservation is given by

Element i 3 2

@q r qu 0: @t

Fig. 1. Structure of the front.

Both immiscible uids are taken to be incompressible, so that the density of a uid particle remains constant:

Dq 0: Dt
This reduces the mass conservation equation to

r u 0:
Also, the viscosity of each uid is constant:

Dl 0: Dt
3. Numerical method

long edge by four new elements. Similarly, elements are deleted two at a time by collapsing the shortest edge into a point (Fig. 2). The NavierStokes equations are solved on a xed grid, but the surface tension is found on the front. Therefore it is necessary to transfer a quantity that exists at the front to a grid. Since the front represents a d-function, the transfer corresponds to the construction of an approximation to this d-function on the xed grid. The interface quantity, /f , is usually expressed in units per area, but the grid value, /g , should be given in terms of units per volume. For conserving the total value in the smoothing, it is required:

Z
DS

/f sdS

Z
Dv

/g xdv

This is accomplished by writing Various methods have been used to simulate two-phase ows. These methods include: (1) Capture the front directly on a regular, stationary grid. The best known examples are the marker-and-cell (MAC) method, where marker particles are used to identify each uid, the volume-of-uid (VOF) method, where a marker function is used and the level set method, where the uid interface marker is used. (2) Use separate, boundary tted grids for each phase. (3) Lagrangian methods, where the grid follows the uid. (4) Front tracking methods, where a separate front marks the interface but a xed grid, is used for the uid within each phase. In addition to front tracking methods that are applicable to the full Navier Stokes equations, specialized boundary integral methods have been used for both potential and Stokes ows. In general, the interface representation can be explicit (moving mesh) or implicit (xed mesh) or a combination of both. The front-tracking method is combination of xed and moving mesh method. Although an interface grid tracks the interface, the ow is solved on a xed grid. The interface conditions are satised by smoothing the interface discontinuities and interpolating interface forces from the interface grid to the xed grid. The equations presented in the last section are solved by a projection method. To compute the momentum advection, the pressure term, and the viscous forces, we use a xed, regular, staggered MAC grid and discretize the momentum equations using a conservative, second-order centred difference scheme for the spatial variables and an explicit second-order time integration method. For the viscous terms we use standard second-order centred differences with simple averages for the viscosity at points where it is not dened. A front structure is used that consists of points connected by triangular elements (Fig. 1). Both the points and the elements (the front objects) are stored in link lists that contain pointers to the previous object and the next abject in the list. The points only know about their coordinates but the elements know about their corner points and the elements that share their edges. As the front moves, it deforms and stretches. For restructuring the front, if an element is too large, the longest edges are spited into two and replace both this element and the one sharing the

/ijk

X
l

/l xlijk

Dsl
h
3

Here /l is a distance approximation to the front value /f and /ijk is an approximation to the grid value /g . Dsl is the area of element l. xlijk is the weight of grid point ijk with respect to element l. we must require that:

X
ijk

xlijk 1

The number of grid points used in the interpolation depends on the particular weighting function selected. The weighting function is usually written as a product of one-dimensional functions. The weight of the grid point (i, j, k) for smoothing from xp xp ; yp ; zp is given by:

xijk xp dxp ihdyp jhdzp kh

where h is the grid spacing. The simplest interpolation is the volume weighting:

dr

h jr j=h jr j < h 0 jr j P h

10

Fig. 2. The restructuring of a front.

M. Bayareh, S. Mortazavi / Advances in Engineering Software 42 (2011) 604611

607

Peskin [21] suggested:

 dr

1=4h1 cospr=2h jrj < 2h 0 j r j P 3h

11

The uid properties are not advected directly. It is necessary to reset the properties at every time step. To construct a method that sets the density correctly even when two interfaces lay close to each other, the fact is used that the front marks the jump in the density and that this jump is translated into a step gradient on the xed grid. If two interfaces are close to each other, the grid-gradient simply cancel. The gradient can be expressed as:

rq

Dqndx xf dS

12

and the discrete version is:

rhqijk

X
l

Dqxlijk nl Dsl

13

where Dsl is the area of the element. Taking the numerical divergence of the grid-density-gradient results in a numerical approximation to the Laplacian:

r2 q rh rhqijk

14

equation for the pressure. Due to the equality in density between the drop and the ambient uid, a quick Poisson solver (FICHPACK) solves the pressure equation. A detailed of numerical method is given by Tryggvason et al. [22]. Numerical simulations of the collision of two drops in simple shear ow were performed over a range of values of the governing non-dimensional parameters of the ow. These parameters are: (i) the viscosity ratio k = ld =l0 , where l0 and ld are the viscosities of ambient uid and of the drop, respectively, (ii) the density ratio g = qd =q0 , where q0 and qd are the densities of ambient uid and of the drop, respectively, (iii) the Reynolds number (particle and _ R2 =l0 and Reb q0 c _ H2 =l0 , bulk Reynolds numbers): ReP q0 c _ ut ub =H is the shear rate. ut and ub are top and botwhere c tom wall velocities, respectively. R is the radius of the undeformed _ R=r, where r is the interdrop. (iv) The capillary number Ca l0 c facial tension. The capillary number is the ratio of viscous force to _ R3 =r. The surface-tension force, (v) the Weber number We q0 c Weber number is the ratio of inertia force to surface-tension force. It should be pointed out that the Weber number and the capillary _ R=H2 , so only one of number are related by We Reb Ca1=c them should be considered in the present study.It is usual to dene a scalar measure of the drop deformation (the Taylor deformation) by:

The left hand side is approximated by standard centred difference and solving the resulting Poisson equation with the appropriate boundary conditions yields the density eld everywhere. The interface conditions are satised by smoothing the interface discontinuities and interpolating interface forces from the interface grid to the xed grid. The force due to surface tension on each element of front is:

De

lb : lb

18

dFr

Z
DS

where l, b are the major and minor semi-axes of the drop (dened by the largest and smallest distances of the surface from the centre). In addition, the collision or lm drainage time is the time between the points where the centre-to-centre distance is equal to one undeformed drop diameter to the instant of coalescence. 4. Results

rjndS

15

The average surface curvature can be written as:

jn n r n
Then, the force on each element surface is:

16 Z
rA

dFr r

Z
rA

jndA r

n r ndA r

I
S

t ndS

17

where the Stokes theorem has been used to convert the area integral into a line integral along the edges of the element. The integration is over the boundary of each element representing the front. t and n are the tangent and the normal vector to each element, respectively. The cross product is a vector that lies on the surface and is normal to the edge of the element. If the element is at, the net force is zero but if the element is curved, the net force is normal to it when the surface tension is constant. For any incompressible ow it is necessary to solve an elliptic equation. For the velocitypressure formulation this is a Poisson

The reference frame to describe the results is shown in Fig. 3. According to experiments of Yoon et al., initial offset is dened as the shortest distance from the centre of the drop to the inow axis (D) divided by drop radius R. The centre-to-centre distance between drops is 4R as shown in Fig. 3. The coordinate axes are oriented as follows: the x-axis is parallel to ow direction, the y-axis is parallel to the vorticity direction, and the z-axis is parallel to the velocity gradient. The relative trajectory of the two drops is expressed in terms of the differences Dz z2 z1 and Dx x2 x1 , where xi and zi are the centre-of-mass coordinates of the ith drop. The difference Dy between the y-coordinates of the two drop centres is zero. In all plots, Dx and Dz will be made dimensionless by the radius R of the undeformed drops. We will compare our results with experimental results of Guido and Simeone [9] and Yoon et al. [14] and numerical results of Loewenberg and Hinch [8].

Fig. 3. Schematic of the relative trajectory between a pair of deformable interacting drops in shear ow (Offset = D/R).

608

M. Bayareh, S. Mortazavi / Advances in Engineering Software 42 (2011) 604611

0.4 0.35 0.3 0.25

Vx

0.2 0.15 0.1 0.05 0 -5 0


Ca = 0.13 Guido & Simeone - Ca = 0.13

X/R

10

Fig. 4. Relative trajectory of two interacting drops in shear ow (Reb = 10, Ca = 0.05, k = 1).

Fig. 7. The relative velocity between the drops in the x-direction versus Dx/R with k = 1, Reb = 10, and offset = 0.08.

Fig. 4 shows the relative trajectories of two interacting drops computed on 66 34 66 and 34 18 34 grids at Reb = 10. _ . Calculations show little Time is normalized by the shear rate c changes between the two cases. Since the calculation time depends directly on the size of grid, we select the coarse grid for our simulations. Fig. 5 shows ve sequences of two interactions in simple shear ow. The interaction between two drops (frame 2) at three different capillary numbers has been compared in Fig. 6. In Fig. 7 the relative velocity between the centres of mass along the ow direction, indicated as DVx, is plotted as a function of Dx. The relative velocity in the x-direction increases rapidly when the drops start interacting. This shows that drops accelerate while sliding over each other. The nal steady-state value of the relative velocity along the ow direction is higher than the value, owing to the increased distance between the drops along the z-axis. The relative velocity along the velocity gradient DVz is plotted as a function of Dx in Fig. 8. The value of DVz before and after collision is zero, as expected. DVz is positive when the drops are

sliding over each other and the distance between the centres of mass along z-direction increases. It becomes negative when the drops are separating and the increase in Dz is partly recovered. In Fig. 9, Dz is plotted as a function of Dx during approach, collision and separation. The data correspond to sequence depicted in Fig. 5. It can be seen that Dz starts increasing after the drops come into apparent contact (Dx $ 2R), reaches a maximum value, and, after separation, reaches a new steady-state value. The nal value of Dz (which is 1.4 for offset = 0. 2, 1.52 for offset = 0.512, and 1.72 for offset = 0.8) is greater than the value before collision. In the other words, if the drops were made to collide again by reversing the ow direction, Dz would increase further. So, the effect is irreversible, and repeated collisions lead to larger values of Dz until drop interaction becames negligible [9]. This is also in agreement with the numerical simulations of Loewenberg and Hinch [8]. Experimental results of Guido and Simeone [9] was based on k = 1.4 and numerical simulations of Loewenberg and Hinch [8] were performed for viscosity ratio of one. The deformation parameter is shown in Fig. 10 as a function of dimensionless time with viscosity ratio of one and Ca = 0.3. Based

Fig. 5. Sequences (15) showing the interaction between two drops in simple shear ow at Reb = 10, Ca = 0.13, Offset = 0.08, k = 1.

Fig. 6. Compresional quadrant of the interaction between two drops with Reb = 10, t= 2.84, k = 1, Offset = 0.08. (1) Ca = 0.05, (2) Ca = 0.13, (3) Ca = 0.3.

M. Bayareh, S. Mortazavi / Advances in Engineering Software 42 (2011) 604611

609

0.55 0.5 0.45 0.4

Deformation

0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 10


Offset = 0.12 Offset = 0.2 Numerical simulations (Loewenberg & Hinch)

t*

20

Fig. 8. The relative velocity between the drops in the z-direction versus Dx/R with k = 1, Reb = 10, and offset = 0.08.

Fig. 10. The deformation parameter as a function of dimensionless time between interacting drops with Ca = 0. 3, k = 1, Reb = 10, and different offsets.

2.2 2 1.8 1.6

0.2 0.18 0.16 0.14

Deformation
Offset = 0.2 Offset = 0.512 Offset = 0.8 Expriment (Guido & Simeone)

1.4

Z/R

0.12 0.1 0.08 0.06 0.04 0.02 D / H = 0.3 D / H = 0.36

1.2 1 0.8 0.6 0.4 0.2 0

X / R

10

t*

10

15

Fig. 9. Cross-ow separation (velocity gradient direction) versus Dx/R between interacting drops with Ca = 0.13, k = 1, and Reb = 10 (present work) and k = 1.4 (experiment), and different offsets.

Fig. 11. The deformation parameter versus dimensionless time between interacting drops with Ca = 0. 075, Reb = 10, and k = 1 and different size of drops.

on experimental observation of Guido and Simeone [9], deformation of the drops is the same. Deformation slightly increases, and then reaches a maximum, a minimum, a second maximum, and eventually reaches a steady state value which is different from initial stages. Numerical simulations of Loewenberg and Hinch [8] show no difference before and after collision at low-Reynolds numbers. Comparison between the results shows that the time of approaching of drops at smaller initial offset is larger (for current simulations). Following Allan and Mason [23] and Guido and Simeone [9], the minimum value of deformation is lower than the steady-state value before (or after) collision. This can be explained as a result of two processes: (i) relaxation of drop shape once it leaves the compressional axis and (ii) action of the surrounding uid on the drops. At xed initial offsets, as the size of drops increases, the deformation increases as shown in Fig. 11. We see that the approach-collision-separation times of drops are different owing to the different initial sizes. The lm drainage time increases with

increasing size of drops. The dimensionless drainage time is 5.21 for D/H = 0.3 and 5.86 for D/H = 0.36. Fig. 12 shows the trajectories of the drops for different initial offsets with the same capillary number for a viscosity ratio of 1. As the initial offset increases, the drops rotate more quickly and the available time for lm drainage decreases as shown in Fig. 10. Therefore, we should expect that the critical capillary number for coalescence will decrease with increasing offset. This was found in studies of Yang et al. [14], where the viscosity ratio was 0.096. Fig. 13 shows the separation distance versus orientation angle (a) formed by the drop major axis and horizontal direction at two initial offsets. We see that the approaching parts of the trajectories are different owing to different initial offsets. However, after the drops come into contact, it can be seen that they follow the same trajectories of separation distance. This is an agreement with the experimental results of Yoon et al. [14]. They said that they have no explanation for this result.

610

M. Bayareh, S. Mortazavi / Advances in Engineering Software 42 (2011) 604611

2.5

2.6 2.4

2.2 2 1.8

Distance, d / 2R

1.5

Distance, d / 2R
Offset = 0.08 Offset = 0.12 Expriment (Yosang Yoon et al.), Offset = 0.08 Expriment (Yosang Yoon et al.), Offset = 0.12

1.6 1.4 1.2 1 0.8

0.5

0.6 0.4 0.2


20

Viscosity Rayio = 0.19 = 1.2 Expriment (Yosang Yoon et al.), 0.19 Expriment (Yosang Yoon et al.), 1.2

10

15

t*
Fig. 12. Trajectories of drops for different initial offsets, separation distance versus dimensionless time, with Ca = 0.00481, Reb = 10, and k = 1 (present work), and k = 1.2 (experiment).

10

t*

20

30

Fig. 14. Trajectories of drops for the different viscosity ratios at the offset of 0.08, separation distance versus dimensionless time, Reb = 10, Ca = 0.00721 for k = 0.19, and Ca = 0.00417 for k = 1.2.

2.5

2.5 2.25

2 1.75

Distance, d / 2R

1.5

Distance,d/2R
Offset = 0.08 Offset = 0.12 Expriment (Yosang Yoon et al.), Offset = 0.08 Expriment (Yosang Yoon et al.), Offset = 0.12

1.5 1.25 1 0.75

0.5

0.5 0.25 0 10 20 30 40

Viscosity Ratio = 0.19 = 1.2 Expriment (Yosang Yoon et al.), 0.19 Expriment (Yosang Yoon et al.), 1.2

25

50

75

100

Angle (deg.)
Fig. 13. Trajectories of drops for different initial offsets, separation distance versus orientation angle (a), with Ca = 0.00481, Reb = 10, and k = 1 (present work), and k = 1.2 (experiment).

50

60

70

80

90

100

Angle (deg.)
Fig. 15. Trajectories of drops for the different viscosity ratios at the offset of 0.08, separation distance versus orientation angle (a), Reb = 10, Ca = 0.00721 for k = 0.19, and Ca = 0.00417 for k = 1.2.

In Fig. 14 the trajectories of drops (centre-to-centre distance scaled by the undeformed diameter) are plotted for the initial offset of 0.08 for different viscosity ratios. As the viscosity ratio increases, the drops rotate more slowly, and the point at which the drops separate is delayed. The lubrication force in the thin lm gets stronger owing to the reduction of interfacial mobility caused by increase of the viscosity ratio. This force reduces the tendency of drops to come into contact, and the drops rotate to a larger angle before they separate. This results a larger external hydrodynamic force which pulls the drops apart, and overcomes the enhanced lubrication force [14]. Fig. 15 shows the separation distance versus the orientation angle for the different viscosity ratios. It can be seen that these trajectories become more symmetric with the increased viscosity ratio.

5. Concluding remarks The effects of viscosity ratio and geometry including the initial offset and the size of drops on the interaction of two equal-sized drops in simple shear ow was presented using a nite difference/front tracking method. Simulations were performed at nite Reynolds numbers. The deformation, relative velocities, relative trajectories, and lm drainage time were examined by changing the initial offset, size of drops, and the viscosity ratio. We changed the offset and the size of drops based on experimental data. It was found that Dz increases after collision and reaches to a new steady-state value after separation. The values of Dz, during the interaction, increases with the increasing initial offset.

M. Bayareh, S. Mortazavi / Advances in Engineering Software 42 (2011) 604611

611

Our results showed that the time of approaching of drops at low initial offset is greater than for larger initial offsets, but the maximum deformation is the same for equal drop sizes. The deformation decreases with decreasing the size of drops. As the initial offset increases, the time for lm drainage decreases. Also, the approaching parts of the trajectories are different from each other owing to different initial offsets. However, after the drops come into contact, it can be seen that they follow the same trajectories. Also, as the viscosity ratio increases, the point at which the drops separate is delayed and the separation distance versus the orientation angle becomes more symmetric, similar to experimental data. References
[1] Taylor GI. The deformation of emulsions in denable elds of ow. Proc R Soc (London) 1934;A146:50123. [2] Cox RG, Mason SG. Suspended particle in uid ow through tubes. Ann Rev Fluid Mech 1972;3:291308. [3] Rallison JM. The deformation of small viscous drops and bubbles in shear ows. Ann Rev Fluid Mech 1984;16:4566. [4] Magna M, Stone HA. Buoyancy-driven interactions between two deformable viscous drops. J Fluid Mech 1993;256:64783. [5] Zhou H, Pozrikidis C. The ow of suspensions in channels: single les of drops. Phys Fluids A 1993;5(2):31124. [6] Feng J, Hu HH, Joseph DD. Direct simulation of initial value problems for the motion of solid bodies in a Newtonian uid. Part 2. Couette and Poiseuille ows. J Fluid Mech 1994;277:271301. [7] Li X, Zhou H, Pozrikidis C. A numerical study of the shearing motion of emulsions and foams. J Fluid Mech 1995;286:374404. [8] Loewenberg M, Hinch E. Numerical simulation of a concentrated emulsion in shear ow. J Fluid Mech 1996;321:395419. [9] Guido S, Simeone M. Binary collision of drops in simple shear ow by computer-assisted video optical microscopy. J Fluid Mech 1998;357:120.

[10] Mortazavi SS, Tryggvasson G. A numerical study of the motion of drop in poiseuille ow, Part 1: lateral migration of one drop. J Fluid Mech 1999;411:32550. [11] Esmaeeli A, Tryggvason G. Direct numerical simulations of bubbly ows. Part 1. Low Reynolds number arrays. J Fluid Mech 1999;377:31345. [12] Cristini V, Blawzdziewicz J, Loewenberg M. An adaptive mesh algorithm for evolving surface: simulation of drop break-up and coalescence. J Comput. Phys 2001;168:44563. [13] Balabel A, Binninger B, Herrmann M, Peters N. Calculation of droplet deformation by surface tension effects using the Level Set method. J Combust Sci Technol 2002;174:25778. [14] Yoon Y, Borrell M, Park CC, Leal G. Viscosity ratio effects on the coalescence of two equal-sized drops in a two-dimensional linear ow. J Fluid Mech 2005;525:35579. [15] Zhao X. Effects of inertia on the rheology of a dilute emulsion of drops in shear ow. J Rheol 2005;49:137794. [16] Sibillo V, Pasquariello G, Simeone M, Cristini V, Guido S. Drop deformation in micro conned shear ow. Phys Rev Lett 2006;97:24. [17] Zhao X. Drop break-up in dilute Newtonian emulsions in simple shear ow: new drop break-up mechanism. J Rheol 2007;51:36792. [18] Lac E, Biesel D. Pairwise interaction of capsules in simple shear ow: threedimensional effects. Phys Fluid 2008;20:0408016. [19] Bayareh M, Mortazavi S. Numerical simulation of the motion of a single drop in a shear ow at nite Reynolds numbers. Iran J Sci Technol 2009;33:44152. [20] Bayareh M, Mortazavi S. Geometry effects on the interaction of two equalsized drops in simple shear ow at nite Reynolds numbers. In: 5th international conference: computational methods in multiphase ow (WIT), vol. 63; 2009. p. 37988. [21] Peskin CS. Numerical analysis of blood ow in the heart. J Comput Phys 1977;25:22052. [22] Tryggvason G, Bunner B, Esmaeeli A, Juric D, Al-rawahi N, Tauber W, et al. A front-tracking method for the computations of multiphase ow. J Comput Phys 2001;169:70859. [23] Allan RS, Mason SG. Particle motion in sheared suspensions. XIV. Coalescence of liquid drops in electric and shear elds. J Colloid Interface Sci 1962;17:383408.

Vous aimerez peut-être aussi