Sunteți pe pagina 1din 14

129

Analysis of Initial Bubble Acceleration


Using the Level-Set Method
M. Dominik and K. W. Cassel*
Mechanical, Materials, and Aerospace Engineering Department, Illinois Institute of
Technology, Chicago, IL 60616, USA

Received: 10 December 2014, Accepted: 30 April 2015

Abstract
A new axisymmetric level-set code for incompressible, multiphase flows using the
vorticity-streamfunction formulation has been developed. The level-set method is well
suited to treating multiphase flows having complex interface shapes that may undergo
topological changes, such as merging and splitting of bubbles. The initial acceleration
of a single air bub- ble in water has been analyzed and found to be approximately 3.3g,
not 2g, which is the predicted value from an added mass analysis based on potential-
flow theory.

1. INTRODUCTION
Gas bubbles contacting liquids is an important and common operation in the chemical process
industry, petrochemical industry, as well as steel making and mineral processing. In applications
such as absorption, distillation, and froth flotation, the interaction of two phases occurs through
bubbling of gas into the liquid, and the equipment is designed based on knowledge of the bubble
rising physics (acceleration, shape, terminal velocity, etc.) and the hydrodynamic parameters
suitable for the desired performance (Kulkarni and Joshi [1] and Hua and Lou [2]).
For an excellent review of the fluid mechanics of bubble rise, see Magnaudet and Eames [3].
When a gas bubble accelerates in a liquid, some of the surrounding liquid must also accelerate,
effectively increasing the mass of the bubble. This has been called added mass and is an inertial,
not viscous, effect. Added mass is normally derived using Bernoulli’s equation and unsteady
potential-flow theory or using the velocities from potential flow and determining the rate of change
of the kinetic energy of the entire flow (Panton [4]). The derivation assumes that creeping flow
and steady-state, potential flow applies, which in many situations are poor assumptions. Such an
analysis is provided in Section 2, where it is shown that the initial bubble rise is predicted to have
an acceleration of 2g.
Solutions of the full Navier-Stokes equations do not require the added mass to be accounted for
explicitly as all viscous and inertial effects are already accounted for implicitly. Here, the level-set
method is used to compute the rise of an axisymmetric bubble, and the numerical results show that
the initial acceleration is greater than 3g, not the 2g predicted by potential-flow theory.

2. POTENTIAL-FLOW ANALYSIS
In this section, the theoretical initial acceleration is determined using potential- flow theory and the
added mass concept. In Section 2.1, the Euler-Lagrange equation of motion for the gas bubble is
used to obtain a general expression for the initial acceleration. Potential-flow theory is used in
Section 2.2 along with the added mass concept to estimate the initial acceleration.

2.1. Euler-Lagrange Equation


The Lagrangian is the difference between the kinetic energy and potential energy. The kinetic
energy of a rising spherical bubble is

(1)

where mg is the mass of the gas bubble, ml is the mass of the surrounding liquid whose motion is
*Corresponding author: E-mail: cassel@iit.edu
130 Analysis of Initial Bubble Acceleration Using the Level-Set Method

induced by the moving bubble, i.e. the added mass, and U (t) is the velocity of the bubble. Also, r0
is the radius of the spherical gas bubble, and ρg and ρl are the densities of the gas and liquid,
respectively. Finally, f is the mass fraction of the added mass reflecting that added mass is
proportional to the mass of the fluid displaced by the body. The potential energy is

(2)

where z is the vertical coordinate. Assuming no viscous drag, which initially is a good
approximation as the velocity is low, the Lagrangian is

(3)

Noting that z˙ = U is the velocity of the bubble, the Euler-Lagrange equation is

(4)

Substituting the Lagrangian (3) into the Euler-Lagrange equation (4) yields

(5)

where the factor 4 πρl r03 f is the added mass. If ρl  ρg, then the initial acceleration of the bubble is
3

(6)

It then remains to determine the mass fraction f in the added mass in order to determine the
acceleration.

2.2. Added Mass


When considering an object, such as a bubble, to be a particle in a dynamics sense moving through
a fluid, it is necessary to account for the inertial effects of the added mass owing to the fact that the
object must push the fluid out of the way as it moves. “Qualitatively, the idea of virtual mass is a
familiar one. For example, let a light paddle be dipped into still water and then suddenly given a
rapid acceleration broadside. It is a matter of common experience that the apparent inertia (i.e.
resistance to acceleration) of the paddle is greatly increased by the water around it” (Birkhoff [5]).
“In 1826, Bessel reported experiments with a spherical pendulum moving in air and water, and he
found that to account for the observed behavior by the conventional equations of motion, a so-
called ‘added mass’ MA , which was proportional to the mass of the fluid displaced by the body,
would have to be added to that of the sphere” (Torobin and Gauvin [6]). In 1883, the “idea of virtual
mass was first given exact mathematical interpretation by Green and Stokes” (Birkhoff [5]). The
following is the classical derivation of the manner in which added mass has been calculated for a
sphere in an inviscid fluid, such that potential theory may be utilized (see, for example, Basset [7]).
The velocity potential for flow about a sphere of radius r0 moving with velocity U (see, for
example, Panton [4]) is

(7)

Journal of Computational Multiphase Flows


M. Dominik and K. W. Cassel 131

Note that this velocity potential assumes that the bubble is moving with con- stant speed; however,
it is accelerating in the present context, i.e. U = U (t).
The unsteady Bernoulli equation in terms of the velocity potential Φ is

(8)

Accounting for the hydrostatic pressure according to

(9)

setting C (t) = 0, and solving for the pressure p* yields

(10)

To calculate the vertical force on the bubble from p*, consider

(11)

where S is the surface of the sphere, and

(12)

and

(13)

Thus, the vertical force on the sphere is

(14)

At the surface of the sphere, from equation (7)

(15)

and

(16)

Substituting (15) and (16) into the Bernoulli equation (10) and then into (14) gives the vertical force
as

Volume 7 · Number 3 · 2015


132 Analysis of Initial Bubble Acceleration Using the Level-Set Method

(17)

or upon integration

(18)

Hence, the vertical force on the spherical bubble is one half the mass of a sphere of the displaced
liquid times the bubble’s acceleration.
Consider Newton’s second law applied to the spherical gas bubble as follows

(19)

where the left-hand side is the mass times acceleration of the gas bubble, and the terms on the right-
hand side are due to buoyancy and added mass, respectively. Substituting equation (18) yields

(20)

With ρg  ρl, the above equation becomes

(21)

or solving for the initial acceleration gives

(22)

Note that drag has been neglected due to the small velocities during initial acceleration.
The above results indicate that the initial acceleration for t  1 is predicted to be 2g.
Comparing with equation (6) reveals that the mass fraction is f = 1 . This classical analysis includes
2
the following assumptions: 1) the fluid motion around the bubble is inviscid, i.e. there is no viscous
drag for t  1; 2) the steady-state velocity potential applies during the initial acceleration; 3) the
density of the gas is much smaller than that of the liquid, i.e. ρg  ρl; and 4) the bubble remains
spherical. Are these assumptions valid when the bubble is accelerating from rest? Torobin and
Gauvin [6] state that the “added mass concept is shown to be completely inadequate and
theoretically unsound.” They conclude that potential flow theory can only be used at the beginning
of the rectilinear acceleration. Brennen [8] states that the viscous effects cause com- plications in
the flow around the body. “The essence of the complications is that in certain regimes of flow the
viscous process of flow separation and vor- tex shedding cause radical modifications to the forces
expected on the basis of simple addition of fluid inertial and fluid drag forces.” To quantify how
inac- curate this classical result is, numerical calculations using the level-set method are conducted
to determine the influence of density difference, viscosity, surface tension, and bubble size.

Journal of Computational Multiphase Flows


M. Dominik and K. W. Cassel 133

3. LEVEL-SET METHOD
Numerical solutions of axisymmetric bubble rise are now considered in order to compare the actual
initial bubble rise, including viscous effects, with that predicted in the previous section.
The level-set method was introduced by Stanley Osher and J. A. Sethian in the 1980s. It is a
computational technique for tracking moving interfaces by relying on an implicit representation of
the interface whose equation of motion is numerically approximated using schemes built from
those for hyperbolic con- servation laws (Sethian and Smereka [9]). The level-set method is
advantageous because instead of tracking the interface explicitly and applying the equations
directly at the interface, the level set is calculated throughout the entire domain such that the
interface is simply the zero contour of the level-set function, and the physical values and constants
are determined from the corresponding signed distance function. Besides being robust, i.e. easily
handling changing topology, such as merging, etc., level-set methods allow one to accurately
represent inter- facial quantities, such as the interfacial normal and curvature. Unlike interfaces in
the volume-of-fluid (VOF) method, where the transition from one fluid to the next takes place over
one grid cell, the level-set function transitions smoothly across the liquid-gas interface; therefore,
standard discretizations of the normal and the curvature using the level-set function can be as
accurate as needed (e.g. second-order accurate or higher) [10]. The primary shortcoming of the
level-set method is that mass (volume) is not always conserved, and the algorithm needs to check
and correct for this problem.
The level set φ(t, x, y, z) is a field function computed throughout the domain, and the interface
between phases is given by the φ = 0 level set. The respective phases then correspond to positive
and negative φ on either side of the interface. A distance function is defined as the minimum
distance from the interface, and it satisfies ∇φ = 1. The signed distance function requirement
helps maintain the level set, φ, to be the distance from the interface.
Implementation of the level-set method in an unsteady CFD code consists of the following: 1)
the level set determines which fluid exists and what the fluid properties are at each grid point in the
domain, 2) the Navier-Stokes momentum equation is solved with these properties to determine the
velocities and pressures at the new time values, and 3) the new velocities are used to advect the
level set. The above process is then repeated to move the fluids forward in time.

4. NAVIER-STOKES EQUATIONS FOR TWO- PHASE FLOW


The Navier-Stokes momentum equation is a vector equation, which for two- dimensional and
axisymmetric cases, consists of at least two equations and two dependent variables. To simplify the
formulation, the curl of the Navier-Stokes momentum equation is taken to obtain the vorticity-
transport equation, which is one scalar equation with the pressure being eliminated. The
streamfunction can be determined from a Poisson equation if the vorticity is known. Then the
streamfunction can be used to determine the velocities. The velocities are then used to advect the
level set. The level set is used not only to determine the physical properties of the fluid (density
and viscosity), but it is also used to determine the curvature of the interfaces for calculation of
surface tension.

4.1. Primitive-Variables Formulation


Incompressible flow is assumed, which is a good approximation as long as the fluid velocities are
much smaller than the speed of sound. Including the sur- face tension force, the dimensional form
of the incompressible Navier-Stokes momentum equation is

(23)

where ρgn̄z is the body force due to gravity in the z-direction, σ is the surface tension at the liquid-
gas interface, κ̂ is the curvature, and δ̂ (φ) is the Dirac delta function.
Using the method from Sussman [11], the Dirac delta function, δ̂, is replaced with the Heaviside
function, Ĥ , such that

Volume 7 · Number 3 · 2015


134 Analysis of Initial Bubble Acceleration Using the Level-Set Method

(24)

Then

(25)

and

(26)

such that

(27)

Substituting into the surface tension term of the Navier-Stokes momentum equa- tion (23) produces

(28)

Using the initial bubble radius Rb as the characteristic length and W0 as the characteristic velocity
(to be determined), the Navier-Stokes equations are nondimensionalized according to

(29)

(30)

where as before properties with subscript l indicate the liquid phase, e.g. ρl , µl , etc. Substituting
into equation (28) and simplifying gives

(31)

To make the gravity term unity, define the characteristic velocity by

(32)

which leads to the Reynolds (Re) and Bond (Bo) numbers being defined by

(33)

Then equation (31) gives the non-dimensional momentum equation in the form

Journal of Computational Multiphase Flows


M. Dominik and K. W. Cassel 135

(34)

4.2. Axisymmetric Vorticity-Streamfunction Formulation


For application to the spherical bubble, the momentum equation is rewritten in axisymmetric form.
The gradient expression in cylindrical coordinates is

(35)

Using u, θ˙, and w as the velocities in the n̄r, n̄ θ, and n̄ z directions and assuming axisymmetric
flow, such that θ˙ = 0 and all derivatives with respect to θ vanish, the incompressible Navier-Stokes
momentum equations in cylindrical coordinates become

(36)

(37)

The continuity equation in cylindrical coordinates for axisymmetric flow is

(38)

Now let us rewrite the Navier-Stokes equations in the vorticity-streamfunction formulation. The
vorticity is the curl of the velocity vector, but for axisymmetric flow, there is only one component

(39)

Note that whereas the densities and viscosities are constant in each phase, and the partial
derivatives of these properties are zero, as reflected in the above equation, such terms must be
included near the interface. These terms will be restored later when needed. Taking ∂/∂z of the
radial momentum equation (36), subtracting ∂/∂r of the axial momentum equation (37), and
dividing by ρ/ρl leads to the vorticity-transport equation

(40)

To determine the velocities, the streamfunction ψ is used, which is defined by

Volume 7 · Number 3 · 2015


136 Analysis of Initial Bubble Acceleration Using the Level-Set Method

(41)

Combining the above equations with the definition of vorticity gives the Poisson equation for
streamfunction in cylindrical coordinates as

(42)

The level set φ(t, r, z) is advected by the velocities from the Navier-Stokes solution according to

(43)

which in this case is

(44)

The level set φ(τ, r, z) = 0 corresponds to the gas-liquid interface, and φ < 0 in the gas and φ > 0
in the liquid.

4.3. Numerical Methods


Because the vorticity-transport equation for two-phase flow (40) cannot nor- mally be solved
analytically, it is solved numerically. The vorticity values are then used to calculate the
streamfunction at each time from equation (42), and the streamfunction is used to calculate the
velocities using equation (41). With the velocities, the level set is advected using an upwind
procedure for equation (44). This iterative process is carried out at each time step to advance the
solution forward in time. All discretizations are second-order accurate except for the upwind
scheme used to advect the level-set variable, which is first-order accurate as is typical of level-set
methods.
As the level set φ is advected, it drifts from being a signed distance func- tion, thereby causing
the gradient to become larger or smaller than the signed distance function, for which ∇φ = 1. This
drift causes numerical errors that can be mitigated by reinitialization [12]. The reinitialization is
carried out by comparing the absolute value of the gradient of the level set to unity and the
difference is used to adjust the level set back to being a signed distance function [13]. The level-
set reinitialization is repeated until the new distance function converges.
The primary disadvantage of the level-set method is the propensity for bub- bles to lose mass
and volume owing to numerical errors at the interface. Indeed, this was found to be an issue in the
present calculations. Mass/volume conser- vation was ensured by calculating the volume of the
bubble at each time step and comparing it to the original volume. If it had lost more than 0.5% of
its volume, a small amount was subtracted from the level set to restore the proper volume.
There are situations for which the effects of a small computational domain may adversely affect
the results because the boundary conditions have a large effect on the results. In these cases, it is
beneficial to modify the scheme near the boundaries to produce ‘far-field’ boundary conditions that
allow for the stretching of the grid to produce a larger domain with the same number of grid points
[14].
The density and viscosity jump is smoothed out across the interface between phases with a
regularized Heaviside function [15] to avoid numerical instabilities. The interface thickness ∈ is
normally given in terms of grid size as α∆x. α is typically an integer plus one half, and our trials
showed that α = 3.5 gives good results. The vorticity is moved forward in time with a Factored
Alternating- Direction-Implicit (ADI) method (see, for example, Obabko [16]).

Journal of Computational Multiphase Flows


M. Dominik and K. W. Cassel 137

The form of the axisymmetric vorticity-transport equation used for the alternating-direction-
implicit algorithm (ADI) is

(45)

where for r not equal to zero, the coefficients are

(46)

and

(47)

It should be pointed out that the last term in S(τ, r, z) must be evaluated in the form given, and the
ωr/r term cannot be included in the Q(τ, r, z) term or the program blows up near r equal to zero
[17]. Special care has to be taken for the axisymmetric case as r → 0 owing to the 1/r terms;
specifically, l’Hôpital’s rule is used to modify S(τ, r, z) as r → 0.
A second-order accurate geometric multigrid algorithm is used to calculate the streamfunction
from the known vorticity. The streamfunction is then used to determine the velocities. In
cylindrical coordinates, the Poisson equation for the streamfunction (42) is

(48)

Thus, the general form for use in the multigrid algorithm is

(49)

where for r not equal to zero, the coefficients become

(50)

As with the vorticity-transport equation, the axisymmetric streamfunction equation must address
the 1/r terms as r → 0 using l’Hôpital’s rule.
For additional details regarding the numerical methods used, the r → 0 modifications,
verification of grid-independence, and the extensive code validation through comparison with
experimental studies and other computational investigations, see Dominik [17]. Based on internal
consistency checks and the comparisons with other experimental and computational results, the
values of the numerical parameters follow these guidelines:
1. Use α = 3.5.
2. Use ∆r ≤ 0.04.
3. ∆t should be less than one half of the ∆r divided by the maximum velocity.
4. The bottom of the bubble should be more than a half of a bubble radius from the bottom
of the domain at t = 0.
5. The top of the bubble should end more than two bubble radii from the top of the domain.

Volume 7 · Number 3 · 2015


138 Analysis of Initial Bubble Acceleration Using the Level-Set Method

6. The width of the domain should be at least 4.25 bubble radii wide.
The present axisymmetric results were obtained with ∆r = ∆z = 0.02, with a grid having 257
points in the radial direction and 1, 025 points in the vertical direction.

5. LEVEL-SET RESULTS
Although the focus of this investigation is on initial bubble acceleration from rest, Figure 1 shows the
vertical velocity of the bubble versus time for a single air bubble starting from rest in water to
illustrate the overall behavior. Results are shown for four bubble sizes and two surface tensions (note
that σ = 73 g/s2 corresponds to air bubbles in water). Two of the cases have clearly reached a
terminal velocity in the time shown. These correspond to Rb = 0.10 cm (red dots) and Rb = 0.02 cm
(blue dots), both with σ = 73 g/s2. The other two cases, having Rb = 0.50 cm (radius) with σ =
73 g/s2 (yellow dots) and Rb = 0.08 cm with σ = 0 (green dots), oscillate around a mean velocity
after the initial period of acceleration. Observe that it is the largest of the bubbles and that with zero
surface tension that oscillate.
The same cases shown in Figure 1 are repeated in Figure 2 on a log-log plot of vertical
displacement versus time. At times between 10–6 and 10–3, most of the cases are near the best-fit

Figure 1: Vertical velocity versus time for various bubble sizes and surface tensions.
The Rb = 0.50 cm with σ = 73 g/s2 (yellow dots), the Rb = 0.08 cm with σ = 0 (green
dots), and the other two sets of points Rb = 0.10 cm (red dots) and Rb = 0.02 cm (blue
dots), both with σ = 73 g/s2 .

Figure 2: Log-log plot of the vertical dispacement versus time immediately after t = 0 for
the same cases as in Figure 1. The blue line is z = 1.65gt2 (a = 3.3g), and the red line
is z = gt2 (a = 2g).
Journal of Computational Multiphase Flows
M. Dominik and K. W. Cassel 139

line a = 3.3g (z = 1 at2), which is well above the 2g acceleration predicted by potential theory.
2
Observe that the larger bubbles adhere to the 3.3g for a longer period of time, because smaller
bubbles reach terminal velocity earlier in time. Figure 3 shows the zero level set and streamlines
for an air bubble in water with initial radius Rb = 0.08 cm for a series of times. As the bubble
accelerates, it remains nearly spherical for this size and surface tension, and both the gas inside and
liquid immediately outside accelerate along with the bubble.

Figure 3: Streamlines (red lines) and φ = 0 interface (black line) for a Rb 0.08 cm radius
air bubble in water with σ = 73 g/s2 for a series of times.

Volume 7 · Number 3 · 2015


140 Analysis of Initial Bubble Acceleration Using the Level-Set Method

Results for the same case, but without surface tension, i.e. σ = 0, are shown in Figure 4. As the
bubble accelerates, observe that it is no longer spherical after a short period of time, but rather
develops a cap shape. This shape results in additional drag and a substantially slower terminal
velocity (see Table 1); however, the initial acceleration is the same as the bubbles remain nearly
spherical during the initial acceleration phase for all cases considered.

Table 1: Initial acceleration between 10–6 and 10–4 seconds and terminal velocity for
gas bubbles in liquid with a change in the surface tension, viscosity, or liquid density. The
gas density is 0.0012 g/cm3 .

Walters and Davidson [18, 19] analyzed two- and three-dimensional bubbles. Their analysis and
experimental results show that the initial acceleration for spherical bubbles is 2g for bubble radii
greater than or equal to 0.3 cm, and the bubbles become cap shaped. The present results are for
bubbles with radii less than or equal to 0.5 cm, for which most are nearly spherical. The first time
data point in the Walters and Davidson paper is 0.008 seconds, but Figure 2 shows the trajectories
moving away from the 3.3g line by this time. The experimental results of Zhang, Yang, and Mao
[20] has a smallest time of 0.01 seconds. Once again, Figure 2 shows the points moving away from
the 3.3g line by this time. Consequently, there is limited data on the initial acceleration of a sphere
in a viscous liquid, and the data available starts when the 3.3g line is no longer valid.
Table 1 lists accelerations and terminal aspect ratios and velocities for gas bubbles in liquid for
a range of radius, surface tension, viscosity, and liquid density. As the liquid viscosity decreases,
the initial acceleration increases; however, the total change is less than 14%. Because the bubble
initially remains spherical, the surface tension does not have a significant effect on the initial
acceleration. For example, although the terminal aspect ratio is quite different, the two cases with
and without surface tension have the same acceleration from 10–6 to 10–3 seconds. Observe also
that the initial acceleration is the least when the liquid density is the lowest, and the higher the
density the higher the initial acceleration.

Journal of Computational Multiphase Flows


M. Dominik and K. W. Cassel 141

6. CONCLUSIONS
In summary, a new level-set code has been developed for incompressible, multiphase flows using
the vorticity-streamfunction formulation in axisymmetric cases. Using the level-set approach with
the alternating-direction-implicit algorithm (ADI) to calculate the vorticity and the multigrid solver
to calculate the streamfunction work well to duplicate previous experimental and numerical results
for both bubble shape and terminal velocity, thereby validating the code and helping to determine
the numerical parameters that should be used.

Figure 4: Streamlines (red lines) and φ = 0 interface (black line) for a Rb = 0.08 cm
radius air bubble in water with σ = 0 g/s2 for a series of times.

Volume 7 · Number 3 · 2015


142 Analysis of Initial Bubble Acceleration Using the Level-Set Method

For most values of the relevant parameters the average initial acceleration of a bubble in liquid
is found to be between 3.0 and 3.7g, which is significantly more than the 2g predicted by potential
flow. The initial acceleration is also found to increase as the liquid density increases but is
insensitive to changes in liquid viscosity. The historic added mass analysis assumes that the bubble
is inviscid, creeping flow, the density and viscosity of the liquid do not affect the results, and the
steady state velocity potential applies, which is not true for a bubble accelerating in a viscous liquid
with inertia, as has been shown by these results. Note that although these computations have been
performed assuming axisymmetry, even high-Reynolds-number bubbles, which may experience
three- dimensional motions as they rise, are likely to be approximated well during their initial linear
acceleration phase.

REFERENCES
[1] Kulkarni, A.A., and Joshi, J.B., “Bubble Formation and Bubble Rise Velocity in Gas Liquid Systems: A Review,”
Industrial Engineering Chemical Research, 44, pp. 5873–5931, 2005.
[2] Hua, J. and Lou, J., “Numerical Simulation of Bubble Rising in Viscous Liquid,” Journal of Computation Physics,
222, pp. 769–795, 2007.
[3] Magnaudet, J. and Eames, I., “The Motion of High-Reynolds-Number Bubbles in Inhomogeneous Flows,” Annual
Review of Fluid Mechanics, 32, pp.659–708, 2000.
[4] Panton, R.L., Incompressible Flow, John Wiley and Sons, 2005.
[5] Birkoff, G., Hydrodynamics, A Study in Logic, Fact, and Similitude, Dover Publications, Inc., 1955.
[6] Torobin, L.B. and Gauvin, W.H., “Fundamental Aspects of Solid-Gas Flow, Part III. Accelerated Motion of a Particle
in a Fluid,” The Canadian Journal of Chemical Engineering, 37, pp. 224–236, 1959.
[7] Basset, A.B., A Treatise on Hydrodynamics, Dover Publications, Inc., 1961. [8] Brennen, C. E., “A Review of Added
Mass and Fluid Inertial Forces,” Naval Civil Engineering Laboratory, Report Number CR 82.010, 1982.
[9] Sethian, J.A., and Smereka, P., “Level-Set Methods for Fluid Interfaces,” Annual Review of Fluid Mechanics, 35(1),
pp. 341–372, 2003.
[10] Prosperetti, A. and Tryggvason, G., Computational Methods for Multiphase Flow, Cambridge University Press,
2009.
[11] Sussman, M., A Level-Set Approach for Computing Solutions to Incompressible Two-Phase Flow, Ph.D. Thesis,
UCLA, 1994.
[12] Osher, S. and Fedkiw, R., Level-Set Methods and Dynamic Implicit Surfaces, Springer-Verlag, 2003.
[13] Sussman, M., Smereka, P. and Osher, S., “A Level-Set Approach for Computing Solutions to Incompressible Two-
Phase Flow,” Journal of Computation Physics, 114, pp. 146–159, 1994.
[14] Sussman, M. and Smereka, P., “Axisymmetric Free Boundary Problems,” Journal of Fluid Mechanics, 341, pp.
269–294, 1997.
[15] Chang, Y.C., Hou, T.Y., Merriman, B., and Osher, S., “A Level-Set Formulation of Eulerian Interface Capturing
Methods for Incompressible Fluid Flows,” Journal of Computation Physics, 124, pp. 449–464, 1996.
[16] Obabko, A. V., Navier-Stokes Solutions of Unsteady Separation Induced by a Vortex – Comparison with Theory and
Influence of a Moving Wall, Ph.D. Thesis, Illinois Institute of Technology, 2001.
[17] Dominik, M., Two-dimensional and Axisymmetric Bubble Rise Using the Level-Set Method, Ph.D. Thesis, Illinois
Institute of Technology, 2013.
[18] Walters, J.K., and Davidson, J.F., “The Initial Motion of a Gas Bubble Formed in an Inviscid Liquid, Part 1. The
Two-dimensional Bubble,” Journal of Fluid Mechanics, 12, pp. 408–416, 1962.
[19] Walters, J.K. and Davidson, J.F., “The Initial Motion of a Gas Bubble Formed in an Inviscid Liquid, Part 2. The
Three-dimensional Bubble and the Toroidal Bubble,” Journal of Fluid Mechanics, 17, pp. 321–336, 1963.
[20] Zhang, L., Yang, C. and Mao, Z-S., “Unsteady Motion of a Single Bubble in a Highly Viscous Liquid and Empirical
Correlation of Drag Coefficient,” Chemical Engineering Science, 63, pp. 2009–2106, 2008.

Journal of Computational Multiphase Flows

S-ar putea să vă placă și