Upload
eustacco
View
213
Download
0
Embed Size (px)
Citation preview
8/22/2019 ATOMIZAO_phd1
1/16
Chapter 1
INTRODUCTION
Laminar liquid jets possess several attractive advantages over othertypes of apparatus for fundamental studies of the mechanism of gasabsorption. To obtain a valid experimental test of unsteady statediffusion theory in a flow system, it is imperative that the fluiddynamics of the system be known accurately.
L. E. Scriven & R. L. Pigford (1959)
Liquid-liquid systems are important to workers in many areas of engineer-
ing, physics, and chemistry. A few examples of current interest are liquid-liquid ex-
traction equipment (Treybal, 1963, 1980; Jeffreys, 1987; Rousseau, 1987, Tsouris
and Tavlarides, 1990), liquid-liquid jets (Richards et al., 1993, 1994a, 1994b),
space applications in propulsion systems, life support, and storage (Kim et al.,
1994), coating flows (Ruschak, 1985; Christodoulou and Scriven, 1989), oil-water
mixtures in pipeline flow (Joseph et al., 1984), extrusion of polymers (Mavridis
et al., 1987), secondary oil recovery (Weidner and Schwartz, 1991), interfacial
tension measurements (Li and Fu, 1992), liquid bridges (Russo and Steen, 1989;
Slobozhanin and Perales, 1993), static menisci (Finn, 1986; Cuvelier and Schulkes,
1990), and interfacial rheology (Edwards et al., 1991).
In this dissertation we seek an improved understanding of the fluid me-
chanics underlying liquid-liquid extraction, a widely used unit operation in which
solutes dissolved in a liquid solution are separated by contact with another, gen-
erally immiscible, liquid. If different solutes in the original solution distribute
themselves differently between the two phases, a certain degree of separation will
develop, and this may be enhanced by multiple contacts in staged operation. If
1
8/22/2019 ATOMIZAO_phd1
2/16
2
the liquids are left in contact long enough, mass transfer between the two phases
causes the solute distribution to approach an equilibrium condition. Extractors
can be constructed in many forms, which can be run batch or continuous, e.g.,
mixer-settlers, plate columns, packed columns, as well as more exotic designs.
Many articles and books have been written on the subject, and the details of cur-
rent design techniques are well documented in, for example, Treybal (1963, 1980),
Jeffreys (1987), and Rousseau (1987).
The efficiency and stability of the separation depend on the underlying pro-
cesses taking place in the stage. These are the fluid flow field, the thermodynamic
equilibria, and the mass transfer within and between phases. The research pro-
posed here focuses on the fluid mechanics alone, as once the flow field is known,the mass transfer and thermodynamic effects can, in principle, be added to com-
plete the analysis. The fluid mechanics of the extractor are dependent on the type
of device considered. If it is a sieve-tray extractor, the light fluid rises through
the continuous phase in trains of drops. If the extractor is a mixer-settler, the
dispersed phase is broken into drops by the energy of an impeller, with turbulent
flow commonly found in practice.
1.1 Four Distinct Levels of Analysis
Various levels of analysis can be applied to liquid-liquid extractors, de-
pending on the information sought, and can be classified as corresponding to the
discrete macroscopic (phenomenological), pseudo-continuum, continuum, or the
molecular level. Each level of analysis can be used to answer different questions
to different degrees of accuracy. Most plant design work is typically performed at
the macroscopic level, where little information is required on the internal details
of the system. Equipment design is at the pseudo-continuum level where empir-
ical information is needed to determine appropriate values of model parameters.
This dissertation is not concerned with questions at the macroscopic level such as
the ones related to the overall performance of the extractor, where most of the
8/22/2019 ATOMIZAO_phd1
3/16
3
design work has occurred in the past. On the opposite end of the scales of length,
questions addressed at the molecular level are mostly concerned with the physical
properties of the components of the system and the boundary conditions, which
again are fairly well characterized. It remains to examine the extractor at the two
remaining levels.
A more detailed level of analysis than the macroscopic level, sometimes
referred to as the pseudo-continuum level, provides a more elaborate description of
the internal structure of the system and the transport within it. For example, mass
transfer coefficients are often used in equipment design (Treybal, 1980). However,
such descriptions still involve spatial averaging (e.g., over a swarm of droplets or
over random column packing) or temporal averaging (in turbulent flow), so thatquantities such as velocities and concentrations do not represent the true values
at a given point in the system. Further, the transport properties (e.g., turbulent
viscosities) are not true state properties, and their process dependence again makes
empirical information necessary.
These limitations are obviated at the continuum level, where velocities and
concentrations are local quantities and transport properties such as viscosities,
densities, surface tensions, and diffusivities, are true state variables. The obvious
problem is that most actual liquid-liquid contactors are much too complex in
their geometry and operation to permit a full continuum description, even with
the computational capabilities available today. In this dissertation, therefore, we
seek a middle ground, by analyzing at the continuum level idealized liquid-liquid
systems pertinent to actual extractors. We expect these analyses to enhance our
understanding of more realistic systems in two ways:
i. Among the systems we examine are ones that represent key elements
of one or more types of actual contactors; for example, a liquid-liquid jet is
representative of the situation above an orifice on a sieve tray.
ii. The parametric dependence of the behavior seen in the ideal model
systems is expected to be related to that in more complex systems; dimensionless
8/22/2019 ATOMIZAO_phd1
4/16
4
parameters that determine the size and the specific surface area of liquid-liquid
jets and drops are investigated.
The object is to model the velocity and pressure fields as well as the interface
shape and location in a two-phase system that has one or more free surfaces.
Various simplified situations can be envisioned. The ones chosen for this work are
the liquid-liquid jet, and the drops formed before and during jetting. Two regions
are considered for the liquid-liquid jets: the steady region near the nozzle, and the
entire region from the nozzle to the breakup of the jet into drops.
1.2 Background
As noted above, most previous analysis and design work on liquid-liquid
contactors has been performed at the macroscopic or the pseudo-continuum level.
Results of such studies, and procedures based on them, may be found in, for
instance, Treybal (1963, 1980), Perry and Green (1984), Jeffreys (1987), and
Rousseau (1987). More detailed analyses at the pseudo-continuum level have
been undertaken using other approaches. For example, Guimaraes et al. (1988,
1990) used the population balance approach to model the effects of drop breakage
and coalescence on the hydrodynamics and mass transfer efficiency of liquid-
liquid continuous-flow stirred tanks at steady-state. However, even in these more
elaborate models it is necessary to estimate parameter values based on empirical
data. It is only at the continuum level, where transport properties are true state
variables, that this problem can be overcome. However, as noted earlier, the
analysis of only relatively idealized systems is possible at the continuum level, even
with the computational capabilities available today. A variety of such systems have
been studied in the past:
Isolated drops. Analytical models of single drops in idealized situations have
a long history. For example, the oscillation of a stationary drop can be described
exactly in the inviscid case (Lamb, 1932), and it has been studied more recently
for small Reynolds numbers by Lundgren and Mansour (1988). More complex
8/22/2019 ATOMIZAO_phd1
5/16
5
systems have been examined either numerically or experimentally or both. Oliver
and Chung (1987) investigated the flow of a fluid sphere translating in another fluid
by solving numerically the steady-state Navier-Stokes equations inside and outside
the sphere using a streamfunction formulation. They obtained numerically the
velocity and pressure distributions inside and outside the sphere. Surface tension
acts to keep the drop spherical, but at high shear rates the drop will deform, and
eventually break. Isolated drops have been experimentally investigated during
breakup and coalescence (Clift et al., 1978; Ashgriz and Poo, 1990). The slender
body theory of Acrivos and Lo (1978) can be used for breakup at non-zero Reynolds
numbers, but requires that the drop or bubble that is breaking up must be
elongated. Stone and Leal (1989) have numerically investigated the dynamics ofdrop deformation and breakup at low Reynolds numbers; Bentley and Leal (1986)
studied these effects experimentally. Basaran et al. (1989) studied small amplitude
drop oscillations in liquid-liquid systems both experimentally and theoretically
and Jeelani and Hartland (1991) investigated the collision of two oscillating liquid
drops.
Liquid-liquid jet. When one liquid is injected into another, a jet may
be formed. That this occurs in sieve plate columns has motivated its study in
apparatuses specifically designed for mass transfer and surface tension experiments
with a single jet (Skelland and Huang, 1977, 1979; Skelland and Walker, 1989).
Meister and Scheele (Meister, 1966; Meister and Scheele, 1967, 1969a, 1969b;
Scheele and Meister, 1968) worked to develop an understanding of the jet and
drop formation based on experiments obtained with 15 liquid-liquid systems. They
reported theoretical results based on linear stability analysis for jet breakup and
drop formation, and overall force balances for drop formation below jetting. It was
noted by Meister (1966) that mass transfer coefficients in extraction columns are
generally reported as overall coefficients on a volumetric basis KySs because of the
difficulty of separating the mass transfer coefficient Ky from the specific surface
Ss. To obtain better predictions of the volumetric mass transfer coefficient, and
8/22/2019 ATOMIZAO_phd1
6/16
6
hence of the mass transfer rate, better predictions of the specific surface formed in
the injection process are required. More accurate estimates of the mass transfer
coefficients can be obtained once the velocity and concentration fields in the two
phases are known. For an example of mass transfer coefficients calculated using
the concentration and velocity profiles in a liquid jet falling in (and absorbing)
carbon dioxide, see Scriven and Pigford (1959). Richards and Scheele (1985)
measured velocity profiles in liquid jets of xylene injected upwards into water
and found good agreement of the boundary layer theory (Yu and Scheele, 1975;
Gospodinov et al., 1979) with jet radius measurements, but poor agreement with
velocity measurements. Even today, the liquid-liquid jet and drop system is still
not completely understood.Liquid jets into air have been studied extensively (Scriven and Pigford,
1959; Vrentas and Vrentas, 1982), with current research focusing on jet breakup
(Mansour and Lundgren, 1990). Current limitations of jet breakup calculations,
such as those of Mansour and Lundgren (1990), who used a boundary element
method (BEM), are that they are generally inviscid and can only be carried
out only as far as the pinch point, at which time the Lagrangian mesh is highly
distorted. The Eulerian volume of fluid (VOF) technique discussed in section 1.3
can overcome both of these limitations; an example of a simple free liquid cylinder
capillary breakup modeled by this method is given by Kothe et al. (1991).
Rayleigh-Taylor problem. An even simpler situation is the case of a heavier
fluid above a lighter fluid in a gravitational field. It was first investigated in the
linear stability analysis limit by Rayleigh (1883) and by Taylor (1950) and has
since then been the subject of numerous studies. Linear stability analysis can be
used to show when this situation is unstable (Harlow and Welch, 1966; Drazin
and Reid, 1987), but this solution holds only for small interface deformations.
The subsequent evolution of the interface and the flow field is more interesting
for our purposes. This has been examined in both two and three dimensions.
The problem was studied numerically by Daly (1967, 1969a) using a Marker
8/22/2019 ATOMIZAO_phd1
7/16
7
and Cell (MAC) type method. Tan (1986) reexamined this unstable situation
assuming incompressible, inviscid and irrotational fluid flow in a bounded three-
dimensional domain. Steady solutions to this problem exist and were derived
in approximate form by using bifurcation theory. It was shown that the surface
develops into a bubbles-and-spikes configuration that can be stable to infinitesimal
disturbances. Tryggvason and Arefs (1983, 1985) numerical investigation of the
Rayleigh-Taylor flow (renamed Taylor-Saffman for Hele-Shaw-cell flow) found that
the evolution of the resulting fingers was affected only by the viscosity ratio. More
recently, Tryggvason and Unverdi (1990) have performed numerical studies in three
dimensions using a front-tracking technique. Both theory and experiment for the
three-dimensional Rayleigh-Taylor instability were studied by Jacobs and Catton(1988).
Kelvin-Helmholtz problem. This problem is a generalization of the previous
one in that the fluids are moving relative to each other. Even when the heavy fluid
is below the light fluid, instability can occur due to shear, as can be shown by a
linear stability analysis (Drazin and Reid, 1987). Surface tension and viscosity
tend to stabilize the flow. Linear stability analyses have also been applied to
more complex situations, e.g., to the Couette flow of two fluids between rotating
concentric cylinders (Renardy and Joseph, 1985a, 1985b), to two-fluid pipe flow
(Joseph et al., 1984), and more recently to air-water flow (Bontozoglou and
Hanratty, 1990). For larger interface deformations, few (numerical) results appear
to have been reported. An example is the Kelvin-Helmholtz analog of the Hele-
Shaw cell studied by Pozrikidis and Higdon (1985), who found different initial
perturbations to result in different final rollups of the finite vortex sheet. However,
many of these and other analyses of the Kelvin-Helmholtz problem are limited to
inviscid, linear regimes, or to the Hele-Shaw approximation (see also Weidner and
Schwartz, 1991).
The Kelvin-Helmholtz problem for two fluids has also been studied exper-
imentally by Thorpe (1968, 1969), using a long rectangular tube containing two
8/22/2019 ATOMIZAO_phd1
8/16
8
immiscible fluids. When the tube was tilted away from the horizontal, a uniformly
accelerating flow was produced, with shear at the interface leading to instability
characterized by growing waves. Although these phenomena had already attracted
considerable attention by the end of the last century, a comprehensive parametric
investigation of the effects of different parameters (Reynolds number, Weber num-
ber, etc.) on the development of the mixing of the two phases is still lacking. This
particular example is an inherently unsteady one that is pertinent to the startup
of stirred tank contactors, and may help to shed light on the phenomenon of phase
inversion (Quinn and Sigloh, 1963; Selker and Sleicher, 1965).
Despite the fact that these problems are ones that have been extensively
studied, a significant part of the effort has been devoted to developing solutionmethods, and not much attention has been paid to examining parametric depen-
dence of the solutions. This is a key aspect affecting the relevance of the idealized
continuum problems to more complex systems, as similarities can be expected
at least in the qualitative effects of different dimensionless parameters. Several
different kinds of parameters are important in a real contactor: (i) physical prop-
erty parameters, e.g., density ratio, viscosity ratio; (ii) geometric parameters, e.g.,
aspect ratio; (iii) hydrodynamic parameters, e.g., Reynolds numbers, Weber num-
bers. Each of the idealized problems outlined above is characterized by a smaller
number of parameters, e.g., the classic Rayleigh-Taylor problem contains only
physical property parameters, but by imposing a finite vertical or lateral size, geo-
metric parameters are added. Similarly, the transition from the Rayleigh-Taylor to
the Kelvin-Helmholtz problem introduces one or two hydrodynamic parameters.
The liquid-liquid jet and drop formation problem includes all of the kinds
of parameters seen in the simpler systems already noted. Thus it is particularly
relevant to liquid-liquid contactors. Also, since it is not completely understood at
the present time, it has been chosen as a model problem for this dissertation. The
systematic study of parametric dependence of solutions to idealized problems is
what we expect to shed light on the behavior of more complex systems.
8/22/2019 ATOMIZAO_phd1
9/16
9
1.3 Numerical Methods for Multiphase Flow
The simulation of the dynamics of isolated drops, liquid-liquid jets, and
Rayleigh-Taylor and Kelvin-Helmholtz involve the solution of the continuity and
momentum equations for the two fluids with specified boundary conditions. The
description of the free surface offers a particular challenge, which can be dealt
with numerically from either the Eulerian or Lagrangian view. The Eulerian mesh
remains fixed in space with the flow moving through it, while the Lagrangian mesh
is convected with the flow. The major drawback of the Lagrangian approach, e.g.,
Ramaswamys (1990) study of free surface liquid sloshing in a container, is that
only simple nonintersecting interfaces can be represented due to a limitation on the
amount of mesh distortion allowed. The finite element method (FEM) has beenused successfully in calculating steady free jet flows (Reddy and Tanner, 1978;
Omodei, 1980; Georgiou et al., 1988; Cuvelier and Schulkes, 1990). Kheshgi and
Scriven (1984) have reviewed the finite element analysis of unsteady free surface
flows, while the works by Baker (1983) and Zienkiewicz (1988) provide a general
reference for the FEM method. As can be seen by examining these studies, it may
be possible to use the Lagrangian FEM to solve the specific steady liquid-liquid
jet problem considered in this dissertation. However, we are further interested in
predicting jet breakup and subsequent drop formation, past the point of necking
and through pinch-off of drops (with possible subsequent drop coalescence), which
precludes the FEM since it cannot, at the present time, be used to model such
complex breaking and reforming of interfaces.
Two basic Eulerian approaches have been formulated to track the interface
and have been reviewed by Hyman (1984) and Unverdi (1990), namely front
tracking and volume tracking. Front tracking uses the technique of characterizing
the interface by computational elements, such as a string of particles, which are
convected with the flow (Daly 1967, 1969a, 1969b). Tryggvason and Unverdi
(1990) used an additional grid with finite thickness that was convected with the
flow. Volume tracking involves using markers (MAC (marker and cell) method,
8/22/2019 ATOMIZAO_phd1
10/16
10
Harlow and Welch, 1966) or a marker function convected by the flow (VOF
(volume of fluid) method, Hirt and Nichols, 1981). Yeung (1982) has presented a
general review of the numerical methods used in free-surface flows, while general
references for computational fluid dynamics (CFD) are Peyret and Taylor (1983)
and Anderson et al. (1984).
A completely different approach is the use of lattice-gas automata (Stock-
man et al., 1990), which are collections of discrete particles constrained to move on
fixed geometric lattices. This approach allows complex interface shapes to evolve
in time, and can be implemented on special parallel architectures. Initially only
fluids with the same density could be accommodated with this method, and an
extremely large number of cells had to be used to obtain accurate results compa-rable to the explicit differential equation methods. More recently, formulation of
lattice-Boltzmann techniques has circumvented most of these deficiencies (Succi
et al. 1991).
Volume of fluid method. The major incentive for using the VOF method is
that the types of problems that can be solved involve highly complex free surface
flows. Reasonable accuracy is attainable and yet the method is relatively simply
implemented. The basic algorithm is available in a two-fluid code called SOLA-
VOF (Nichols et al., 1980; Hirt and Nichols, 1981), and part of our work has been
devoted to modifying this algorithm to adapt it to the problems of interest, which
involve transient free surface flows with two immiscible fluids. More recently,
Kothe et al. (1991) introduced a one-fluid code, RIPPLE, which incorporates
various improvements to the one-fluid VOF algorithm.
The VOF formulation assumes that the flow in each phase is unsteady,
viscous, and incompressible. The equations of motion and continuity are solved in
a manner similar to that used by Nichols et al. (1980) with appropriate boundary
conditions for no-slip, free-slip, continuative, periodic, and contact angle, in two-
dimensional Cartesian or axisymmetric coordinates. The interface surface forces
are incorporated as accelerations in the momentum equations rather than as
8/22/2019 ATOMIZAO_phd1
11/16
11
boundary conditions using a marker function field, defined as F = 1 for fluid
2 and F = 0 for fluid 1. This VOF function F is obtained by solving a kinematic
relation. The discontinuous density and viscosity fields are obtained from linear
interpolations using the F function. In our work, we have also included in the
algorithm the calculation of the streamfunction, which can be obtained from a
Poisson equation.
The SOLA-VOF code is well suited for high Reynolds number flows, includ-
ing those involving free surfaces. Among the latter, however, it is better suited for
gas-liquid than for liquid-liquid systems, and relaxing this limitation has been an
important part of our efforts. A one-fluid program, RIPPLE, which implements
the Continuum Surface Force (CSF) algorithm and incorporates various improve-ments in the SOLA-VOF algorithm, has been introduced by Kothe et al. (1991)
and Brackbill, et al. (1992).
1.4 Dissertation Objectives
The goal of this dissertation is to model aspects of complex liquid-liquid flow
behavior, implemented within a robust and efficient code to simulate steady-state
and transient liquid-liquid flows, including high Reynolds number ones. This code
is applied to the realistic simulation of aspects of the complex fluid mechanical
behavior in order to develop quantitative insight into the underlying processes
involved, such as drop and jet formation, size, shape, and breakup. Of course, it is
unavoidable that a limited range of the parameters has been investigated, dictated
primarily by the availability of reliable experimental data.
The two principal barriers to further progress in the area of liquid-liquid
separations are perceived to be the lack of a fundamental understanding of the
fluid mechanical phenomena associated with the development and destruction of
free surfaces separating two immiscible liquid phases, and the lack of a robust
and efficient numerical code that would enable the reliable simulation of key test
flow cases. This dissertation aims to contribute towards the resolution of both the
8/22/2019 ATOMIZAO_phd1
12/16
12
above issues, physical and computational.
First, we have focused our research efforts towards the improvement of the
VOF numerical technique that we consider to be the best approach to the numeri-
cal solution of free surface problems that involve two immiscible liquid phases, high
Reynolds number laminar flows and complex, time-dependent interfaces. Consid-
erable progress has recently been made in developing VOF codes for 2-D and 3-D
free surface flows involving liquid-air or liquid-vacuum interfaces (RIPPLE and
FLOW-3D (developed by Flow Science, Los Alamos, NM, Hirt, 1988) are two
examples), and although these indicate the appropriateness of the technique for
complex free surface problems, no commercially available code exists for liquid-
liquid high Reynolds numbers free surface problems beyond the original SOLA-VOF (Hirt and Nichols, 1981). POLYFLOW (Crochet, 1987), one of the best
commercial codes based on FEM that can perform multiple phase calculations,
is best suited for low Reynolds number flows as, for example, the low Reynolds
number free jet die swell problem (Hill and Chenier, 1984). This code, as do other
codes, uses the method of spines for representation of the free surface, so it can
only represent simple, nonintersecting, single valued interfaces. A spine is a line
segment connecting two interface nodes and each interior node must belong to a
single spine. Thus, there is a constraint on the type of meshes that can be formed
with the use of spines. The SOLA-VOF code represented the starting point for our
computational code development. However, due to the age of the original code (it
was written before 1980), substantial further improvement was necessary in order
to transform it to a versatile (robust) and efficient research tool for complex high
Reynolds number free surface flows.
Second, we have contributed to the knowledge base on liquid-liquid interface
issues by investigating a few selected test cases, using the code developed as
discussed above. Specifically, we have examined a liquid jet injected into another
immiscible liquid. As explained in section 1.2, this is a critical problem for the
understanding of liquid-liquid separations. However, and in contrast to the much
8/22/2019 ATOMIZAO_phd1
13/16
13
more widely studied liquid-air jet problem, little information is available on the
flow characteristics associated with the jet and the corresponding jet breakup
problem. Most of the available information comes from the previous experimental
work of Meister and Scheele (Meister, 1966; Meister and Scheele, 1967, 1969a,
1969b; Scheele and Meister, 1968) and Richards and Scheele (1985), whereas the
theoretical work, limited to boundary-layer solutions of the problem, seems to
be in considerable disagreement with the experimental data for the established
velocity profiles, despite correctly capturing the jet shape. We have used our
simulations to predict the velocity profiles in an effort to validate the approach
and simultaneously to enable us to do a systematic parametric search. We also
evaluated numerically the breakup conditions as well as the subsequent non-linearevolution of the drops, as a function of time. Thus, this research is expected to have
a significant impact on the way complex liquid-liquid contactors are designed in
the future, not to mention a plethora of other important technological applications
(such as, for example, ink jet printing studied by Croucher and Hair, 1989 or
low-gravity flows in space vehicles, Ostrach, 1982), where the understanding of
interface development is of paramount importance.
1.5 Dissertation Outline
The chapters of this dissertation are organized in the following manner. In
chapter 2 the governing equations that are used throughout this work are derived
from a fundamental viewpoint both to serve as a rigorous basis for the dissertation
and to indicate where future extensions may evolve. Here the Volume of Fluid
(VOF) function is introduced in conjunction with the novel use of the Heaviside
step function, which is an invaluable analytical tool to describe the interface.
The solutions to these problems involve the solution of the equations of
motion and continuity for two fluids with specified boundary conditions. The free
surface offers a particular challenge. During the course of this work we have found
difficulty in producing numerically stable free surface results with the SOLA-VOF
8/22/2019 ATOMIZAO_phd1
14/16
14
surface tension algorithm. In chapter 3 we have combined the two-fluid capability
of the SOLA-VOF algorithm with the free surface implementation of the CSF
algorithm to investigate problems that involve transient 2-D free surface flows
with two immiscible fluids with possibly high Reynolds numbers, and found the
results to be much more numerically stable. The numerical method used to solve
the finite difference equations for the velocity and pressure and the advection of
the free surface is described. The surface force calculation is detailed as well as the
boundary conditions used in the free surface problems. Numerical stability issues
are also discussed. Several test problems that have known analytical or numerical
solutions were used to test the accuracy and validate the modified algorithm and
are detailed in section 3.8. The results of these problems indicate reasonableaccuracy, with some difficulty encountered in the capillary tube at small contact
angles and very low Reynolds numbers (Re < 1). It was also found that the
viscosity could not be considered spatially constant across the free surface as it was
in the original SOLA-VOF code, since it was found that the normal stresses were
not equal at the free surface in the problem of two-phase cocurrent flow between
parallel plates. Thus, the code was rewritten in terms of a variable viscosity.
In chapter 4 an efficient and robust method is presented for solving the
Young-Laplace equation that describes the shape of the meniscus in a vertical
cylinder for a constrained liquid volume. This serves as a test problem for future
work and as an introduction to the static free-surface equation. An overview
of this field is covered by Finn (1986). The method of solution proposed here
explicitly incorporates the constraint in the equation, transforming the two-point
boundary value problem into an initial value one with the constraint determining
the unknown centerline height. This allows rapid determination of the solution
families, which are characterized by only centerline height and meniscus arclength.
The method can be generalized to other axisymmetric systems such as spheres,
cones, and ellipsoids with or without axial rotation.
In chapter 5 we investigate the steady liquid-liquid jet. When a liquid
8/22/2019 ATOMIZAO_phd1
15/16
15
is injected into another liquid a jet may be formed. This occurs in liquid-
liquid extraction plate columns as well as apparatus specifically designed for mass
transfer experiments with a single jet. The axisymmetric steady-state laminar flow
of a Newtonian liquid jet injected vertically into another immiscible Newtonian
liquid is investigated for various Reynolds numbers. The steady-state solution
was calculated by solving the axisymmetric transient equations of motion and
continuity using a numerical scheme based on the Volume of Fluid (VOF) method
combined with the new Continuum Surface Force (CSF) algorithm. The analysis
takes into account pressure, viscous, inertial, gravitational, and surface tension
forces.
The liquid-liquid jets eventually break up due to the increasing amplitudeof disturbance waves on their surface. In chapter 6 we investigate the length of the
resulting jets, which is dependent on many factors. Meister and Scheele (1969a)
have so far provided the most complete picture of liquid-liquid jet breakup. The
axisymmetric, dynamic breakup of a Newtonian liquid jet injected vertically into
another immiscible Newtonian liquid at various Reynolds numbers is investigated
in chapter 6. The full transient from jet startup to breakup into drops was
simulated numerically by solving the time-dependent axisymmetric equations of
motion and continuity using the code developed in this dissertation. The algorithm
has been further refined here based on its performance on transient problems
such as the solution of the free liquid-liquid capillary jet breakup problem. The
simulation results are compared with previous experimental measurements of jet
length under conditions where all forces, i.e., viscous, inertial, buoyancy, and
surface tension, are important.
The size of the drops produced before and during jetting is important from
an industrial standpoint due to the creation of large, new surface area. In chapter 7
the formation of drops by the vertical injection of a Newtonian liquid into another
stationary immiscible Newtonian liquid, at low to high Reynolds numbers, before
and after jet formation, is investigated. The full transient from before startup
8/22/2019 ATOMIZAO_phd1
16/16
16
to breakup into drops was simulated numerically by solving the time-dependent
axisymmetric equations of motion and continuity. The numerical simulations are
compared with experiments and previous simplified analyses of Meister and Scheele
(1968, 1969b) based on drop formation before and after jetting. Prediction of the
sizes of these drops over wide ranges of Reynolds numbers is another important
result from this work, where only approximate theories have existed up until this
time.
Solutions to these difficult problems shed light on the distortion of interfaces
into various shapes and the physical variables that are important in extractors from
a hydrodynamic viewpoint.