See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/269043977

A CFD Methodology for Liquid Jet Breakup and Vaporization Predictions of

Compressible Flows

Conference Paper · January 2008

DOI: 10.2514/6.2008-1023

CITATIONS READS
9 1,567

5 authors, including:

Kevin Brinckman Ashvin Hosangadi

Combustion Research and Flow Technology, Inc. Combustion Research and Flow Technology, Inc.
24 PUBLICATIONS 292 CITATIONS 122 PUBLICATIONS 1,258 CITATIONS

SEE PROFILE SEE PROFILE

Vineet Ahuja Sanford Dash

Combustion Research and Flow Technology, Inc. Combustion Research and Flow Technology, Inc.
80 PUBLICATIONS 935 CITATIONS 178 PUBLICATIONS 2,010 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Supercritical CO2 Compressors View project

All content following this page was uploaded by Kevin Brinckman on 23 December 2014.

The user has requested enhancement of the downloaded file.

46th AIAA Aerospace Sciences Meeting and Exhibit AIAA 2008-1023
7 - 10 January 2008, Reno, Nevada

A CFD Methodology for Liquid Jet Breakup and

Vaporization Predictions in Compressible Flows

K.W. Brinckman 1 , A. Hosangadi 2 , V. Ahuja 3 , and S.M. Dash 4

Combustion Research & Flow Technology, Inc. (CRAFT Tech), Pipersville, PA 18947

and

G.M. Feldman 5
Combustion Research & Flow Technology, Inc. (CRAFT Tech), Charlotte, NC 28277

A robust computational fluid dynamics methodology for simulating liquid jet discharge
and breakup in high-speed gas/liquid flows is being developed for use in practical
engineering applications. The proposed approach is cast within a RANS framework and
utilizes a volume-of-fluid type (VOF) methodology to efficiently capture the gas/liquid
interface location. Relevant physics are modeled to predict liquid atomization/vaporization
through a cascading process involving interface surface breakup, primary droplet
formation, and droplet secondary breakup and vaporization. The current VOF approach is
well suited for applications involving liquid jet discharge at lower ambient pressures, such as
liquid fuel venting, gas-turbine fuel injection, or atmospheric bulk-dispense problems, where
the liquid behavior is essentially incompressible making the numerical solution more
difficult in a compressible flow environment. In place of a traditional VOF approach with
different thermodynamic treatments of gas and liquid, a unified, multi-phase
thermodynamic framework is used which is applicable to both the gas and liquid phases.
Density-based fluid dynamic equations are transformed to a “quasi-pressure-based” form,
and preconditioning is used which facilitates integrating the equations with widely disparate
sound speeds. This approach is implemented in the structured grid code CRAFT CFD®, as
well as the multi-element unstructured grid code, CRUNCH CFD®, permitting grid
adaptation to be applied to enhance efficient gas-liquid interface tracking. In order to avoid
resolving the liquid surface breakup numerically, a surface breakup model is applied with
correlations for droplet formation based on local shear and surface tension across the
gas/liquid interface, allowing the size of the droplets generated to vary spatially as well as in
time with the local evolution of the gas/liquid interface. These primary droplets are
transferred to an Eulerian dispersed phase where they are subject to secondary breakup and
vaporization. Several solutions of exemplary problems are presented.

I. Introduction

A methodology to predict the sequence from liquid jet discharge to vaporization in practical engineering
applications requires the integration of a number of physics-based models into a robust numerical framework,
capable of producing high-resolution solutions. The current work presents progress in applying an engineering
model for primary atomization of liquid jet which predicts the local rate of droplet formation and their sizes along a
gas/liquid interface, coupled to dispersed phase models to analyze the entire process from liquid jet
breakup/atomization to droplet vaporization, shown schematically in Figure 1. With a computationally efficient
modeling framework in place, the next step is to complete a validation effort utilizing a “building block” approach in

1
Senior Research Scientist, 6210 Keller’s Church Rd., Member AIAA
2
Principle Scientist, 6210 Keller’s Church Rd., Member AIAA
3
Senior Research Scientist, 6210 Keller’s Church Rd., Member AIAA
4
President & Chief Scientist, 6210 Keller’s Church Rd., Associate Fellow AIAA.
5
Research Scientist, Charlotte, NC 28277, Member AIAA

1
American Institute of Aeronautics and Astronautics
Copyright © 2008 by Copyright @ 2008 by the authors. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
which individual models are evaluated with relevant data, before pursuing an integrated model validation. This
effort is ongoing and complete validation results will be discussed in a future work. Here, we will focus on a
discussion of the modeling framework which provides sufficient computational efficiency to allow analysis of
practical system-level problems, such as missile vent simulations, and gas turbine and scramjet/rocket fuel
injection/combustor analysis.
The simulation of liquid jet injection into a high-speed gas stream requires prediction of the gas/liquid interface
in a high Reynolds number turbulent flow regime. Several methods for interface capturing employ either volume of
fluid (VOF) and/ or level set (LS) methods in which the complete Navier-Stokes equations are solved for both the
liquid jet and gas flow, along with a surface-interface equation. These methods attempt to capture a sharp interface
through the use of the surface-interface equation. In early VOF-like formulations1, governing equations were solved
only for the liquid phase and surface location. This approach is relevant to free-surface flows, such as fuel tank
sloshing, where the gas phase does not impart large shearing forces on the surface and a pressure boundary condition
is adequate. Subsequent VOF formulations considered the motion of both the gas and liquid phases along with a
function to define the surface location (e.g. Ref. 2), and are more suited to considering surface tension effects.
Similarly, LS methods3 seek to implicitly capture the interface, using a level-set function. They differ from VOF
methods, in that the level-set function varies smoothly across the interface, while the volume fraction is
discontinuous.
There are many surface capturing
approaches based on both VOF and LS
formulations. Once the surface is captured,
VOF and LS methods may seek to predict the
surface wave structure and material stripping
process leading to droplet formation. While
there has been some success in applying
surface capturing methods to resolve the
surface waves and predict breakup on a
research level, their application in
engineering simulations is impractical, in that
time-accurate and highly resolved calculation
of the interface evolution is required to
predict and track the surface wave instability.
An ensuing breakup of the liquid surface can
yield a large number of arbitrarily shaped
ligaments and droplets, whose size is well Figure 1. Schematic of liquid jet breakup and vaporization
below the resolution of a practical in a gas crossflow.
computational mesh. Thus, the ability to
resolve the liquid jet surface breakup and resulting droplets, using a surface capturing method, as part of a larger
scale problem, such as fuel injection in a gas turbine combustor, becomes computationally prohibitive. For this
reason, the current approach to engineering simulations of liquid jet breakup and atomization in complex
environments is accomplished through the use of a robust gas/liquid interface capturing methodology, where the
interface cells contain a homogenous gas/liquid mixture, coupled to an established surface breakup and atomization
correlation using local gas/liquid properties. This method avoids the requirement of capturing a sharply defined
interface within a given computational cell, in order to retain the detailed structure of the surface waves.
In the current approach the evolution of the liquid jet and its interactions with surrounding gas flow is captured
as part of a new unified VOF formulation, which allows for generalized fluid thermodynamics (i.e., pure liquid, pure
gas or mixtures in between)4. Droplets are generated by first identifying the location of the gas/liquid interface,
which lies in the two-phase mixture region and may extend across several cells. Correlation are applied in these cells
for primary breakup to strip the liquid phase away and form droplets based on local shear and surface tension across
the interface. This stripping of the liquid phase from the two-phase cells tends to sharpen the interface to a point
where the two-phase region is one cell thick, with pure gas/liquid on either side. The size of the primary droplets
generated will vary spatially as well as in time and is a function of the local evolution of the gas/liquid interface. In
this respect, the current methodology represents a significant improvement over other engineering practices where
droplet size correlations are based on global measures of the shear and surface tension forces, without accounting for
the coupling with the unsteady evolution of the local flow.
Droplets resulting from the primary breakup of the liquid jet are tracked using an Eulerian framework and
dispersed phase models are applied to capture the subsequent sequence from droplet breakup to vaporization. A

2
American Institute of Aeronautics and Astronautics
summary of the gas/liquid and droplet methodology being incorporated into the CRAFT Tech suite of CFD codes is
provided in Table I. This procedure is being used at CRAFT Tech to study droplet generation in fuel vented from
missiles at higher altitudes (see Ref. [5]), and for liquid fuel-injection propulsive problems for both subsonic and
supersonic applications. The models applied for liquid jet breakup and vaporization will be discussed in the
following sections, along with several exemplary problems illustrating their application to various configurations.
Table I. Gas/liquid/droplet methodology

Feature Relevance Status

Bulk Liquid/Gas Capture interface of liquid and its Quasi-pressure-based, preconditioned formulation can
Interactions motion. analyze real-fluid thermodynamics for G/L mixtures.
Primary Breakup Liquid droplets are formed at the Correlations used to provide rate of droplet formulation
gas/liquid interface. and droplet size utilizing resolved interface and local
properties (surface tension and shear forces).
Secondary Larger droplets breakup into Droplet Weber numbers used to predict secondary
Breakup smaller droplets. breakup. TAB methodology being incorporated.
Droplet Droplets can vaporize, based on Nonequilibrium vaporization model incorporated .
Vaporization heating rates and local conditions.
Droplet Vapor can condense on pre- Nonequilibrium condensation model incorporated.
Condensation existing nuclei.
Droplet Droplets can burn in hot zones. Modeling of gas-phase kinetics of fuel vapor available
Combustion for select fuels.

II. Gas/Liquid Methodology Overview

The ability to capture the sharp density gradient between a nearly incompressible liquid and a compressible gas
in an accurate manner, presents a challenge to the robustness of the numerical scheme. CRAFT Tech has
implemented a new compressible, gas/liquid formulation for modeling unsteady, cloud cavitation problems, with
application to turbomachinery equipment, e.g. pumps, hydrofoils, underwater propeller blades, etc.6 Compressible
gas-liquid methodologies tend to be poorly suited to cavitation flow regimes, where apart from the localized gas
bubble, the flow is completely incompressible and numerical stiffness is encountered. In the current formulation,
this is remedied by transforming to a pressure-based set of equations and implementing matrix pre-conditioning that
deals with stiff equation systems that result from integrating through a gas/liquid interface.
In this new formulation, the independent variables are transformed from a density and internal energy system to
a pressure and enthalpy system. Mixtures of “generalized fluids” follow Amagat's law and not Dalton's law. The
matrix Γ defines the transformation of the variables from Q to Qv . The original system of equations is given by:

∂Q ∂ E ∂ F ∂G
+ + + =S+D v (1)
∂t ∂ x ∂ y ∂ z

Q = [ ρ , ρ u, ρ v, ρ w, ρYG , e, ρ k , ρε ]
T
where (2)

Here, ρ represents the mixture density which is related to the liquid (L) and gas phase (G) densities by their
respective mass fraction in a given computational cell by:
1
ρ= (3)
YL YG
+
ρL ρG
The transformed system takes the form:
∂ Qv ∂ E ∂ F ∂ G
Γ + + + = S + Dv (4)
∂t ∂x ∂y ∂z

Qv = [ p, u , v, w, YG , T , k , ε ]
T
where (5)

3
American Institute of Aeronautics and Astronautics
 Thermodynamic derivatives in this formulation are defined in a completely generalized fashion that works with
user defined equations of state with the transformation matrix given by:
⎛ ⎞
⎜ ⎟
⎜ βρ p 0 0 0 ρT ρy G
0 0 ⎟
⎜ u βρ p ρ 0 0 u ρT u ρ yG 0 0 ⎟
⎜ ⎟
⎜ v βρ p 0 ρ 0 v ρT v ρ yG 0 0 ⎟
⎜ ⎟
⎜ wβρ p 0 0 ρ w ρT wρ yG 0 0 ⎟
(6)
Γ=⎜ ⎟
⎜ H βρ p − (1 − ρ hp ) ρ u ρ v ρ w H ρT − (1 − ρ hT ) H ρ yG + ρ hyG 0 0 ⎟
⎜ ⎟
⎜ YG βρ P 0 0 0 YG ρT (Y ρ
G YG + ρδ ij ) 0 0 ⎟
⎜ ⎟
⎜ k βρ p 0 0 0 k ρT k ρ yG ρ 0 ⎟
⎜ εβρ p 0 0 0 ερT ερ y 0 ρ ⎟
⎜ G
⎟
⎜ ⎟
⎝ ⎠
where subscripts P, T, and YG, denote a partial derivative to the respective variable.
For efficient operation over a wide range of Mach numbers, this matrix is preconditioned using the parameter β
to get better-matched eigenvalues that improve convergence and reduce round-off errors. For a comprehensive
discussion on the gas/liquid formulation, solution procedure, and numerics the reader is referred to Ref. [7].
Previous works 4,5,6,7 have shown that the upgraded formulation provides a robust methodology for capturing the
gas/liquid interface for various applications, such as cavitation, liquid venting, and bulk liquid fly-out problems
without limitations on system pressure or required liquid compressibility, and comparisons to data have been good.
Here, we show a comparison of three-dimensional gas/liquid interface predictions to experimental observations from
a fuel tank slosh test, reported by Fisher et al.8. These experiments were conducted in a ground drop test facility at
NASA Marshall Space Flight Center and consisted of dropping a cylindrical fuel tank partially filled with petroleum
ether as the test fluid. Plots of predicted liquid volume fraction on the symmetry plane are compared to photographs
taken in the experimental facility at a sequence of times in Figure 2. Taking into account that the photographs
capture three-dimensional surface effects, while the predictions are taken as symmetry plane contour plots, the
comparison is quite good, as the asymmetry of the time-dependent surface sloshing behavior is captured.

t= -0.1s t = 1.0s

t = 2.0s t = 4.0s
Figure 2. Liquid fuel tank slosh - prediction vs. experimental observation.

4
American Institute of Aeronautics and Astronautics
 III. Droplet Physics Methodology
A capability to simulate a complete liquid jet discharge through vaporization sequence is being developed which
couples a number of physical models to the gas/liquid solution methodology discussed above. First, a breakup model
is applied to strip liquid from the jet in the form of primary droplets, which are tracked with an Eulerian particle
model. The primary droplets are then subject to secondary breakup and vaporization as they travel through the
surrounding gas stream. The models used to implement the droplet physics methodology are discussed below.

A. Primary Breakup
Currently, the most challenging part of simulating a liquid discharge to vaporization sequence is modeling the
primary breakup of the liquid jet. The primary breakup mechanism involves the deformation of the gas-liquid
interface and the formation of surface waves. The stripping of bulk liquid can be resolved by computing the
merging of iso-surfaces of liquid mass fraction in a VOF type calculation, which may be an alternative in describing
breakup in the Rayleigh regime, where the drops are of the order or greater than the original jet diameter. However,
in more general situations, for atomizing flows, the byproducts of primary breakup will be much smaller than the
bulk, and the computational grids required to resolve the surface waves will be prohibitively large. To circumvent
this difficulty, models are used to compute the rate of formation of a spray of droplets and characterize their size.
As an alternative to resolving the surface breakup numerically, CRAFT Tech has investigated the use of a
surface breakup model, in which correlations for droplet formation are applied based on local shear and surface
tension across the gas/liquid interface at that location9. The sizes of the droplets generated vary spatially as well as
in time and are a function of the local evolution of the gas/liquid interface. This represents a significant
improvement over current engineering practices where droplet size correlations are based on global measures of the
shear and surface tension forces and the coupling with the unsteady evolution of the flow is not present. This
approach to the prediction of surface breakup and droplet formation is directly applicable to liquid jet atomization
modeling and provides an alternative for analyzing complex three-dimensional fuel injection problems without the
large computational expense of resolving the gas/liquid interface breakup. The model responds to changes in liquid
viscosity and surface tension, as well as local flow conditions.
Different models have been derived to simulate primary breakup based on the flow and liquid characteristics,
however, many of these models restrict their application to problems where the original shape of the liquid is well
known and a length scale readily identifiable. To predict the onset of breakup for problems with arbitrary shapes, a
general expression for the rate of mass breakup at the surface is better suited. The semi-empirical correlation of
Mayer10 for liquid jet atomization in high velocity gas streams was implemented as formulated in the Coaxial
Injection Combustion Model (CICM) 11. Mayer used an energy balance on a wave surface driven by wind velocity
to derive a wind-maintained wavelength λ, and formulate a droplet formation rate and mean droplet diameter, based
on erosion of the crest of a wave with amplitude comparable to λ. Applying the approach of Mayer10, the CICM
derives a local stripping rate of the form:
2

m b = C A [
(
μ1 ρ gU 2 r ) 1
]3 π D1 (Δz ) (7)
σl ρl

and the drop diameter at this location is computed as:

μ l σl ρl
d p = BA [ 2
]2 / 3 (8)
ρ gU r

The empirical constants, CA and BA, are equal to 0.08 and 120, respectively for liquid oxygen, as derived in Ref.
[11], however, they must be recalibrated for different fluids. The subscript l denotes liquid quantities, g denotes gas
quantities, µ is the viscosity and Ur is the magnitude of the relative velocity between the two phases. The CICM
was originally developed to simulate the breakup of cylindrical jets, for which Dl represents the jet diameter and
πDl(Δz), the surface area of the liquid in a computational cell of length Δz. For our applications, Eq. (7) was
generalized by reformulating the stripping rate as the mass rate of breakup per surface area of the liquid, Sl:
2

′′
m b =
m b
= CA[
(
μ1 ρ gU 2r ) 1
]3 (9)
Sl σl ρl

5
American Institute of Aeronautics and Astronautics
The breakup surface area Sl is based on the respective cell size and interface normal as described in detail by Tonello
et al.9.
Knowing how to compute m b′′ and dp, the issue to be addressed in applying the surface breakup model is to
decide where and when breakup occurs. To that end, and to simulate the physical mechanism responsible for
breakup which is that surface waves grow until they become unstable and break off, surface breakup was modeled as
a rate process, based on the local, computed properties of the captured interface. The strength of this approach is
that the breakup is modeled based on local, dynamic properties, not on a-priori or global considerations.
Therefore, provided a fundamental correlation is used to compute the breakup rate, the physics of breakup will be
represented accurately.
The present model is obtained by analogy with chemical kinetic modeling, where a reaction mechanism is
divided into an induction and a reaction period whose evolution is followed by computing the change in the fraction
of elapsed induction and reaction times 12,13. In the present case, we track the mass available for breakup, and
mb
breakup occurs when enough mass has been accumulated. Defining the fraction of broken mass, f = , where mb
md
is the mass available for breakup and md the mass of a drop, a conservation equation for f is solved as part of the
gas/liquid equation set:
∂ (ρ f ) m b
+ ∇ (u ρ f ) = ρ + fm (10)
∂t md

with m as the rate of mass of leaving the mixture liquid phase, and ρ, u are the mixture density and velocity vector
respectively. With this model, at each step, the fraction of broken mass is computed in each cell and:
ƒ if f < 1, no breakup, f increases according to Eq. (10),
ƒ if f > 1, form Nd = int(f) drops of nominal diameter given by Eq. (8).
This formulation is ideally suited for application with our gas/liquid methodology, in that the local gas/liquid
conditions are calculated for use in the stripping rate and droplet diameter correlations, and the break fraction f can
be integrated along the location of the gas/liquid interface. Local perturbations to the flow field, such as those due to
walls, internal structures, or adjacent injectors are accounted for through the gas/liquid solution, as well as changes
in temperature, which can affect the fluid properties, such as surface tension, viscosity, and density, all of which
affect the stripping rate and droplet size. During a timestep Δ t , where breakup occurs, a mass m×Δ  t of liquid is
m b
removed from the mixture, along with its momentum and enthalpy, and ρ + fm less liquid mass is available for
md
breakup. Mass stripped from the liquid jet is transferred to an Eulerian droplet group, via the particle source term
Wp4 discussed in the next section. The primary droplets are subject to secondary breakup and vaporization through
the use of dispersed phase models discussed below.

B. Eulerian Droplet Transport

The current droplet transport model employs a continuum type of approach to modeling a large number of
droplets, by tracking a mass fraction of droplets and average particle surface area (Sp) in a given control volume.
This along with the droplet density can be used to calculate the Sauter Mean Diameter (SMD), i.e. the diameter of a
droplet whose volume (Vp) to surface area ratio is the same as that for the entire continuum cloud, as
Vp
d32 = 6 (11)
Sp

which experiments have shown is an appropriate particle size metric to base vaporization and/or combustion rate
calculations14, where the droplet size changes throughout the domain. This approach is similar to that used by
Kelleners15 to avoid tracking a prohibitively large number of droplets, or using binning to cover a range of particle
radii.
The liquid droplets are tracked within the computational domain using an Eulerian particle approach with
second-order upwind flux numerics. Conservation equations for the Eulerian continuum cloud are solved for the
mass, momentum, enthalpy, and droplet surface area within a computational cell. In the present formulation, the
droplets are assumed volumetrically dilute relative to the gas for computational simplicity, and use of a droplet
volume fraction is not needed. A methodology for relaxation of the dilute assumption to account for a dense

6
American Institute of Aeronautics and Astronautics
dispersed phase of droplets has been implemented in an earlier version of the CRAFT CFD® code16, and future
efforts are planned to investigate the effect of a dense dispersed phase on liquid jet breakup and spray distribution.
The Eulerian continuum-cloud particulate equations are written in generalized coordinates as follows:
∂ Qp ∂ Ep ∂ Fp ∂ Gp
+ + + = W p1 + W p 2 + W p 3 + W p 4 (12)
∂t ∂ξ ∂η ∂ζ
where:

⎡ ρp ⎤ ⎡ ρp ⎤
⎢ ⎥ ⎢ ⎥
⎢ ρ pu p ⎥ ⎢ ρ p u pU p ⎥
⎢ρ v ⎥ ⎢ρ v U ⎥
Qp = ⎢
p p ⎥
Ep = ⎢
p p p ⎥
; (13)
⎢ ρ p wp ⎥ ⎢ ρ p w pU p ⎥
⎢ ⎥ ⎢ ⎥
⎢ ρ p hp ⎥ ⎢ ρ p h pU p ⎥
⎢ ⎥ ⎢ ⎥
⎣⎢ S p ⎦⎥ ⎣⎢ S pU p ⎦⎥
In this set of equations, ρp represents the particle cloud density, which can be expressed as follows:
ρ p = Nm p (14)

where N is the local number density of the droplets, and mp is the averaged mass of the droplets. Assuming that the
droplets are spherical in shape,
π
mp = d 3p ρ s (15)
6
where dp is taken as the Sauter-Mean Diameter and ρs is the droplet material density.
The remaining variables in the above equation denote the particle velocity (up,vp,wp), enthalpy (hp), surface area
per unit volume (Sp) and the cell contravariant velocity, Up. The fluxes Fp and Gp are defined similarly for the other
computational directions. The source vector Wp1 includes inter-phase drag and heat transfer terms, Wp2 denotes
source terms related to the condensation and evaporation of the droplets with the surrounding gas phase (discussed
below), Wp3 contains the gravitational source term, and Wp4 is related to primary breakup of the liquid phase. Details
on the inter-phase drag and heat transfer terms, and gravitational source terms can be found in Ref. [17]. Mass
stripped from the liquid jet, as described in the previous section, is transferred to an Eulerian droplet group, via the
particle source term Wp4, where the subscript L denotes a property of the mixture-liquid phase, m is the rate of mass
of leaving the mixture liquid phase, and u is the gas/liquid mixture velocity vector.

⎧ m ⎫
⎪ mu  ⎪
⎪ ⎪
1 ⎪⎪ mv  ⎪⎪
Wp4 = ⎨ ⎬ (16)
Vcell ⎪ mw 
⎪
⎪ mh  L ⎪
⎪ ⎪
⎪⎩4m ρ L d p ⎪⎭
Equal quantities of mass, momentum, and energy are removed from the mixture liquid phase.

C. Droplet Secondary Breakup

The methodology to analyze how larger droplets (such as those produced in the primary breakup stage) breakup
into smaller droplets, then vaporize and burn follows directly from that used in modeling spray combustion
problems. Secondary breakup, at a basic level, is directly related to the Weber number,
2
We = 2 ρ g r U g − U p / σ (17)

and can be classified in terms of three basic mechanisms 18, namely:

7
American Institute of Aeronautics and Astronautics
 Table II. Droplet breakup mechanisms
Weber Number Breakup Mechanism Description
6<We≤80 Bag Breakup Thin bag forms behind droplet rim
80<We≤350 Stripping Breakup Shear forces strip droplets from liquid ligaments
We>350 Catastrophic Breakup Droplet immediately disintegrates

Our most computationally efficient droplet breakup/vaporization methodology makes use of an Eulerian
framework, which tracks the droplet SMD in each computational cell. The secondary break-up methodology,
applied at an engineering level, determines how many child droplets a single parent droplet breaks into, based on the
droplet Weber number. A simplified secondary breakup model, based on the characteristics of the regimes in Table
II, has been incorporated into the CRAFT CFD® code. Incorporation of a more rigorous secondary breakup model
into the CRAFT Tech CFD codes is currently being pursued, based on the Taylor Analogy Breakup Model (TAB) of
O’Rourke 19. Refinements to the TAB method for spray atomization and droplet breakup, such as those discussed for
example in Ref. [18] are being investigated.

D. Droplet Vaporization
The secondary breakup methodology is coupled to a droplet vaporization model based on a modified form of the
Hertz-Knudsen equation, which gives a mass transfer rate as the difference between incoming fluxes from the gas
phase, and evaporative fluxes from the droplet of radius r.

⎡ Ppartial Psat ⎤
m p = N 4π r 2 ⎢ − eff
⎥ (18)
⎣⎢ 2π RT 2π RTd ⎥⎦

Ppartial is the partial pressure of the droplet substance vapor in the gas phase, T is the gas phase temperature, Td is the
droplet temperature, and R is the gas constant for the material changing phase. An effective saturation pressure,
Psateff, given by the Kelvin-Helmholtz equation, which accounts for surface tension, σ, is used.
⎡ 2σ ⎤
Psat = Psat exp ⎢ ⎥ (19)
⎣ ρ d RTd r ⎦
eff

Here Psat is the flat film saturation pressure at the droplet temperature. Surface tension is also expressed as a function
of droplet temperature.
The particle source term Wp2 is constructed to transfer mass between the gas and droplet phases. For the disperse
phase, the source term becomes

⎧ m p ⎫
⎪ ⎪
⎪ m p u p ⎪
⎪ ⎪
1 ⎪ m p v p ⎪
Wp2 = ⎨ ⎬ (20)
Vcell ⎪ m p w p ⎪
⎪ m h* ⎪
⎪ p ⎪
⎪m S ρ ⎪
⎩ p p p⎭
The value used for h* in the above equation is dependent on whether condensation or evaporation is occurring
within the given cell. For condensation (i.e., when m p > 0 ), the enthalpy of the condensable gas species within the
mixture is utilized; for evaporation, the enthalpy of the droplet plus its heat of vaporization is used. A similar source
term expression is used in the gas phase equations but with opposite sign.

E. Material Data Base

A number of material properties are necessary to apply the droplet models presented above. In order to
generalize the code application to allow consideration of alternate liquid materials, a material data base has been
developed which includes relevant material properties used by both the primary breakup model and the dispersed

8
American Institute of Aeronautics and Astronautics
 Table III. Material data base
phase models. Included in the data base are temperature dependent
Liquids
properties for liquid surface tension, viscosity, density, and saturation
Water
pressure. For computational efficiency and accuracy, the properties are
Methane
implemented as polynomial fits to temperature, allowing the physical Hexane
mechanisms to respond to changes in local droplet temperature at each Cyclohexane
time step. The data base has been populated with a number of materials Heptane
shown in Table III, including a variety of pure hydrocarbon liquids using Octane
the NIST WebBook tables which were generated from various Nonane
experimental sources 20. The material data base is housed outside of the Decane
CRAFT CFD® code, such that new materials can be added easily, Benzene
without a code modification. The code logic accesses the data base for Toluene
Nitrogen
the required material as necessary during a given simulation. Ammonia

IV. Computational Methodology

Vapor generated by droplet vaporization is transferred to a gas phase as a source term in the species conservation
equation where it is treated with the CRAFT CFD® structured, finite-volume code or the CRUNCH CFD®
unstructured, finite-volume code that solve the compressible Navier-Stokes equations. The solver is fully implicit
and uses Roe/TVD numerics for the inviscid fluxes and second-order central differencing for viscous and diffusive
terms. The codes are sufficiently general to simulate finite-rate chemistry and multi-phase flows and includes
standard polynomial curve fits for the thermodynamic properties. A variety of advanced turbulence model
formulations are applied, including a compressibility corrected form of the k-ε model, which is used here. Excellent
computational efficiency is achieved through a fully parallelized implementation of the flow solver using a
combined shared and distributed memory (using MPI) data parallel model.

V. Illustrative Capabilities

A. Liquid Co-Flow Jet Breakup

The simulation of the breakup of a round water jet by a higher speed annular air jet was performed and compared
to the experimental observations of Lasheras et al. 21. Figure 3 shows a schematic of the problem, corresponding to
one set of experimental conditions. The inner jet is a 100% liquid core with a nozzle mean exit velocity of 0.52 m/s.
The liquid core is surrounded by an outer air jet with a mean exit velocity of 85.4 m/s. The simulation was run at
room temperature and pressure (300 K and 1 atm). Comparisons to experimental observations are made qualitatively
between time-average RANS predictions of liquid volume fraction and instantaneous liquid jet photographs.

Gas

Liquid 3.8 mm 5.6 mm

Gas

Momentum Ratio Gas Velocity Liquid Velocity Weber #

(M) (m/s) (m/s)
31.7 85.4 0.52 447

Figure 3. Schematic of co-annular gas/liquid jet problem.

The nature of surface breakup varies significantly with injection conditions, and various water-to-air momentum
ratios were tested by Lasheras et al. 21, who define the momentum ratio as:
ρ gU g2
M= (21)
ρlU l2

9
American Institute of Aeronautics and Astronautics
The observed structure of the jet core indicates that at increased momentum ratio (M) the liquid jet will break up
more readily, as expected. Here we consider a higher-momentum ratio case, where liquid jet breakup was more
prominent. A Weber number of 447, defined as:
2
We = ρ g DL U g − U p / σ (22)

suggests that this jet will be subject to more catastrophic breakup, with finer ligaments21.
Figure 4a presents a plot of the liquid volume fraction (jet and droplets) with the primary breakup model
activated, showing the liquid jet core being disintegrated, as droplets are shed off of the gas/liquid interface. Figure
4b shows the volume fraction of droplets alone. The time-averaged image of the liquid jet breakup observed in
Figure 4 compares well to the liquid jet breakup captured in the instantaneous photographs provided in Figure 2h of
Lasheras et al. 21. Here, the focus is on the primary breakup mechanism in the near field, and the simulation was not
extended to capture dispersed phase spray phenomena, such as droplet secondary breakup, evaporation, and
turbulence dispersion, which will tend to disperse the droplets more, as they extend to the far field. While there
appears to remain a relatively large liquid volume fraction in the jet core region in Figure 4a, the experimental
observations show this was not a finely atomized jet, and large parcels of liquid appeared to be shed from the jet
core. Therefore, a fully atomized spray cone was not expected for this simulation. Figure 5 shows the SMD is in the
range of 350-450 microns in a majority of the spray cone, indicating relatively large primary droplets.
A rough estimate of the liquid breakup length is provided in Lasheras et al.21 as:
xb 6
≈ 1/ 2 (23)
DL M

where the liquid breakup length xb is the liquid length before droplet stripping occurs. For the current case, this
relation yields an estimated liquid breakup length of xb = 1.1 DL, where DL is the liquid nozzle exit diameter.
Inspection of the predicted jet breakup reveals the formation of droplets beginning at approximately 1.4 DL
downstream of the nozzle exit, which is in reasonable agreement with the estimate from Eq. (23).

(a)

(b)
Figure 4. Co-flow jet with primary breakup – Liquid volume fraction;
(a) total liquid/droplets, (b) droplets.

10
American Institute of Aeronautics and Astronautics
 Figure 5. Co-flow jet with primary breakup - Droplet diameter.

B. Transverse Liquid Jet Breakup

The gas/liquid-primary breakup model is next applied to a transverse round liquid jet discharging into a
crossflow air stream. Here the complexities of a three-dimensional flow field are encountered, providing a more
challenging test of the computational framework. The experimental investigation of a water jet into a subsonic
crossflow of Wu et al. 22 is used as a basis for this problem. The evolution of the jet is reminiscent of the process
shown schematically in Figure 1, where a round liquid jet is discharged vertically into a horizontally flowing
crosswind. Wu et al. 22 consider varying injector diameters, liquids, and gas/liquid velocities, and use instantaneous
photographs to illustrate the effect of changes in these parameters on the jet breakup. Significantly, Wu et al. 22
focus on the near-field primary breakup of the liquid column and provide correlations for liquid column trajectory
and fracture point. This type of experimental reporting provides a valuable data base for calibration and validation
of the breakup models. Here, we consider a water jet discharging from a 0.5mm diameter orifice into a 70 m/s air
crosswind, at a liquid-to-air momentum ratio q=9.9, defined by Wu et al. 22 as
ρ LU L2
q= (24)
ρGU G2

This momentum ratio provides a liquid velocity of 9.3 m/s. The liquid temperature is 300K, and the gas temperature
and pressure are 306K and 140000Pa, respectively.
Wu et al. 22 provide instantaneous photographs of the liquid jet discharge and breakup, which show that at q=9.9,
the column holds together for approximately five jet discharge diameters, before breaking up. Iso-surfaces of liquid
volume fraction of the predicted liquid jet column and primary droplets are shown in Figure 6, without secondary
breakup or droplet vaporization. Qualitatively, this figure shows the liquid column holding together until the
instability has developed sufficiently to generate primary droplets. These droplets are stripped from the liquid at the
rate predicted using the relation in Eq. (7). The primary droplet SMD ranges from 300 – 350 microns, as shown in
Figure 7, indicating that they are relatively large parcels, as expected, and a secondary breakup mechanism would be
needed to disintegrate them into smaller droplets as they are subjected to the higher velocity freestream.
The primary droplet Weber number, shown in Figure 8, is in the 30-35 range as the droplets move out into the
higher speed stream, away from the shielding effect of the liquid jet. Based on the criteria in Table II, the primary
droplets would be subject to a bag-type breakup, in which our simplest secondary breakup model decomposes a
parent droplet into two child droplets for 6 < We ≤ 80. The child droplets can also experience secondary breakup,
and this cascading effect will continue until a droplet is small enough to be carried by the gas stream such that its
Weber number drops below the critical value of six. The distribution of the secondary and primary droplets shortly
after activation of the secondary breakup model is shown in Figure 9, and reveals the breakup of the primary
droplets as suggested by the Weber number distribution in Figure 8. The corresponding droplet SMD distribution in
Figure 10 shows a dramatic reduction in droplet size down to the 100 micron range, as the droplets continue to
breakup due to interaction with the higher velocity freestream gas flow. As the simulation is run further out in time,
the secondary droplets will tend to disperse more due to their smaller size.
Here, we have demonstrated the functionality of the proposed liquid breakup model to predict liquid column
breakup in a gaseous crossflow. Further calibration of the primary breakup model is needed to adjust the constant CA
in Eq. (7) to consistently predict the observed column breakup length, over the range of momentum ratios
considered by Wu et al. 22 Observations of alternate liquids are also reported by Wu et al. 22. It is planned to use this
data set, along with others available in the literature, to calibrate the primary breakup model for transverse liquid jet
breakup. One of the challenges is to obtain relevant quality data for droplet size in this near-field region, which is
scarce, due to the difficulty in making these measurements in regions of high droplet density.

11
American Institute of Aeronautics and Astronautics
Figure 6. Transverse liquid jet with primary breakup – Figure 7. Transverse liquid jet with primary breakup –
Iso-surface of liquid/droplet volume fraction. Primary droplet Sauter Mean Diameter

Figure 8. Transverse liquid jet with primary breakup – Primary droplet Weber number.

Figure 9. Transverse liquid jet – Primary & secondary Figure 10. Transverse liquid jet – Primary & secondary
breakup; liquid/droplet volume fraction. breakup; droplet Sauter Mean Diameter.

12
American Institute of Aeronautics and Astronautics
C. Unified Secondary Breakup and Droplet Vaporization in a Transverse Jet
The previous examples of a liquid jet
discharging into a high-speed gas flow show the
formation of relatively large primary droplets. As
seen in the transverse liquid jet example, droplets
in a high shear environment are subject to Freestream
secondary breakup which can result in further M=2.0

disintegration of the droplet, based on the droplet

Weber number.
Considered here is the response of hypothetical Jet w/ Toluene Droplets
M=1.0
pre-existing droplets to the secondary breakup Droplet radius = 50 micron
mechanism coupled to droplet vaporization. The
Figure 11. Transverse jet with droplet injection schematic.
simulation compares the penetration of vented 50
micron radius droplets into a high speed stream
with (1) no secondary breakup, and (2) high We
number, catastrophic type breakup (where each
droplet breaks into numerous droplets, each No Breakup
having a resulting radius that is approximately 2% Mean Radius =
of the original droplet size). The problem is 47 microns
shown schematically in Figure 11, with toluene
droplets injected via a slot jet into a Mach 2 air
freestream. Calculated Weber numbers are well
above 350 over most of the droplet field,
Secondary Breakup
indicating that they would undergo catastrophic Mean Radius =
breakup per the criteria of Table II. Figure 12 1 micron
presents the droplet radius behavior (note the
difference in scales) for each case, and shows the
prominent effect of secondary breakup on droplet
Figure 12. Transverse jet with droplet injection -
size. Without secondary breakup, vaporization
Droplet radius.
reduces the droplet radius by less than 10%, while
the combined effect of breakup and vaporization results in more than an order-of-magnitude reduction in droplet
size. The left-hand side of Figure 13 shows the droplet cloud penetration, while the right-hand-side shows the
gaseous toluene distribution due to droplet vaporization. The impact of secondary breakup on droplet vaporization
and transport is clearly evident in this unit problem, as the smaller droplets provide a significantly larger surface-
area/mass ratio, affecting drag, heat transfer, and vaporization, and consequently the distribution of gaseous fuel.

No Breakup

Secondary
Breakup

Particle Mass Loading C7H8 Species Mass Fraction

Figure 13. Transverse jet with droplet injection - Droplet distribution (left) and vapor formation (right).

13
American Institute of Aeronautics and Astronautics
D. Surface Tracking with Grid Adaptation
While the proposed formulation for predicting liquid jet
primary breakup eliminates the need to spatially resolve a sharp
unsteady gas/liquid interface within a single computational
volume, sufficient spatial resolution is required to accurately
predict local gas/liquid differential velocity for the breakup
models and avoid excessive numerical diffusion. Here, we
demonstrate CRAFT Tech’s CRISP CFD® grid adaptation
methodology 25,26 to locally refine the solution around the
gas/liquid interface, while allowing the grid to remain more
coarse in regions away from the interface. A sample
application of the generalized VOF model to track a high-speed
gas/liquid interface, with grid adaptation, considers a liquid
blob flying through air at Mach 2. The liquid was a cylindrical
blob 50 cm long and 50 cm in diameter. The liquid was
impulsively started at a velocity of 700 m/s. A snapshot of the
results is shown in Figure 14, showing the important physical
characteristics captured in the simulation. Pressure contours Figure 14. Bulk liquid flyout without breakup.
show the bow shock that formed ahead of the blob. Also
evident is the gas/liquid interface highlighted by the dark black
lines. The simulation was run without the primary breakup model, to more readily illustrate the grid adaptation to a
fixed interface.
The liquid blob case was started with a coarse grid and refined using grid adaptation. The current capabilities of
the CRISP CFD® grid adaptation package allow the grid to be refined dynamically as the solution progresses
without the need for any user involvement. Upon starting the problem within the unstructured Navier-Stokes solver
CRUNCH CFD®, an “error estimate” is projected ahead of the solution at each time step to determine the fitness of
the current grid. Once the fitness falls below pre-defined levels, the grid is automatically adapted based on user
defined refinement factors. The grid is adapted based only on the original grid, not prior adaptations, in order to
apply the refinement only where needed, and avoid excessive computational expense.
Figure 15 shows the grid adaptation for a sequence of times for the liquid blob simulation, with contours of
liquid mass fraction overlaid. Figure 16 shows the adaptation with pressure contours overlaid. As the blob moves
through the domain, the grid is adapted to cluster points along the gas/liquid interface, and also along the shock
boundary, in response to specification of the adaptation parameters of pressure and liquid mass fraction.
The CRISP CFD® capability will allow a computational mesh to adjust to the changing interface location as
primary breakup of a liquid jet occurs. Significantly, the user can start the calculation with a relatively coarse grid
without prior knowledge of where the interface location will be after breakup, and CRISP CFD® will refine the grid
dynamically, to provide specified resolution at the interface.

Figure 15. Grid adaptation with liquid mass fraction.

14
American Institute of Aeronautics and Astronautics
 Figure 16. Grid adaptation with pressure.

VI. Conclusion
A robust CFD methodology for simulating liquid jet discharge and breakup in a compressible gas flow is
outlined for use in practical engineering applications. The proposed approach is cast within a RANS framework and
utilizes a volume-of-fluid type methodology to efficiently capture the gas/liquid interface location. This approach is
not intended to resolve the actual liquid surface within a computational cell, but is ideally suited for application of
the proposed surface breakup model based on the free surface wave concept of Mayer10. Comparison of predictions
to experimental observations without surface breakup shows good agreement in capturing the gas/liquid interface
location. Use of matrix preconditioning techniques allows for more robust numerics in these gas/liquid problems
with widely disparate material densities and sound speeds. The RANS-based approach to predicting the liquid jet
breakup and droplet evolution, along with the utilization of grid adaptation methodology, allows the capturing of
local flow-field effects on the liquid atomization, while making multi-scale system level simulations practical.
Several illustrative problems have been presented showing the capability of the coupled gas/liquid-primary
breakup methodology to predict liquid jet disintegration and droplet/ligament formation for water jets in high-speed
air flows. The primary droplets stripped from a liquid jet are transferred to dispersed-phase Eulerian droplet groups,
which are subject to secondary breakup and vaporization. It has been demonstrated how each mechanism in the
sequence from liquid jet discharge through droplet vaporization must be accounted for, in order to accurately predict
the distribution of vapor resulting from liquid jet atomization.
Future modeling work is planned in the area of droplet coalescence, droplet turbulent dispersion and eddy-
tossing, and dense dispersed phase effects to further generalize the methodology to capture the relevant phenomena
which may be encountered throughout a flow field. Further validation/calibration of the primary breakup model for
liquids of interest is needed. Validation of each dispersed phase model is being pursued. Once the individual models
are refined, validation of the integrated model sequence will be carried out for various aero-propulsive applications.

Acknowledgments
This work was supported in large part by CRAFT Tech Internal R&D funding, with portions of the technology
described based on earlier efforts supported by NASA/DoD.

References
1
Hirt, C.W., and Nicholls, B.D., “Volume of fluid (VOF) method for dynamics of free boundaries,” J. Comput. Phys., 39, 201-
221, 1981.
2
Kawamura, T., and Miyata, H., “Simulation of nonlinear ship flows by density-function method,” J. Soc. Naval Architects,
176, 1-10, 1994.
3
Osher, S. and Sethian, J. A., “Fronts Propagating with Curvature-Dependent Speed: Algorithms Based on Hamilton-Jacobi
Formulations,” Journal of Computational Physics, Vol. 79, pp. 12–49, 1988.
4
Hosangadi, A., and Ahuja, V., “A New Unsteady Model for Dense, Cloud Cavitation in Cryogenic Fluids,” Paper No. AIAA-
2005-5347, 17th Computational Fluids Dynamic Conference, Toronto, Ontario, CA, Jun. 6-9, 2005.
5
Feldman, G., Hosangadi, A., Brinckman, K.W., and Dash, S.M., “Application of Improved Gas/Liquid Methodology to High
Speed Venting and Bulk-Flyout Problems,” AIAA Paper 2007-0922, 45th Aerospace Sciences Meeting and Exhibit, Reno,
NV, Jan. 8-11, 2007.
6
Ahuja, V., Hosangadi, A. and Arunajatesan, S., “Simulations of Cavitating Flows Using Hybrid Unstructured Meshes,”
Journal of Fluids Engineering, May/June, 2001.

15
American Institute of Aeronautics and Astronautics
 7
Hosangadi, A. and Ahuja, V., “Numerical Study Of Cavitation In Cryogenic Fluids,” Journal of Fluids Engineering, Vol.
127, pp. 267-281, March 2005.
8
Fisher, M.F, Schmidt, G.R., and Martin, J.J., “Analysis of Cryogenic Propellant Behaviour in Microgravity and Low Thrust
Environments,” Space Cryogenics Workshop, 10th, Cleveland, OH, June 18-20, 1991.
9
Tonello, N.A., Dash, S.M., and Perrell, E.R., “Advanced Computational Framework For Dynamic Bulk-Liquid Gas
Interactions,” AIAA Paper-99-2205, 35th AIAA/ASME/ SAE/ASEE Joint Propulsion Conference and Exhibit, Los Angeles,
CA, June 20-24, 1999
10
Mayer, E., “Theory of Liquid Atomization in High Velocity Gas Streams,” ARS Journal, Vol. 31, No. 12, December 1961, pp
1783-1785.
11
Sutton, R.D., Schuman, M.D., and Chadwick, W.D., "Operating Manual For Coaxial Combustion Model," NASA CR-
129031, April 1974.
12
Tonello, N.A., An Experimental and Computational Study of The Mechanisms of Detonation Transmission in Layered H2-O2
Mixtures, Ph.D. Thesis, The University of Michigan, Ann Arbor, MI, 1996.
13
Oran, E.S. and Boris, J.P., Numerical Simulations of Reactive Flows, Elsevier, New York, 1987.
14
Kuo, K.K., “Principles of Combustion,” John Wiley and Sons, New York, 1986
15
Kelleners, P., 2003, “Simulation of inviscid compressible multi-phase flow with condensation,” Annual Research Briefs,
Center for Turbulence Research, Stanford University. 49-67.
16
Hosangadi, A., Sinha, N., and Dash, S.M., "A Unified Hyperbolic Interface Capturing Scheme for Gas/Liquid Flows," AIAA-
97-2081, 13th AIAA CFD Conferences, Snowmass, CO, June 29-July 2, 1997.
17
Kenzakowski, D.C. and Brinckman, K.W., “CFD Simulations of NASA B-2 Spray Chamber
During Rocket Fire,” AIAA Paper 2008-0790, 46th Aerospace Sciences Meeting and Exhibit, Reno, NV, Jan. 7-10, 2008.
18
Tanner, F.X., “Development and Validation of a Cascade Atomization and Drop Breakup Model for High-Velocity Dense
Sprays,” Atomization & Sprays, Vol. 14, Issue 3, 2004.
19
O’Rourke, P.J. and Amsden, A.A., “The TAB Method for Numerical Calculation of Spray Droplet Breakup,” SAE Paper
872089, 1987.
20
Lemmon, E.W., McLinden, M.O., and Friend, D.G., "Thermophysical Properties of Fluid Systems" In NIST Chemistry
WebBook, NIST Standard Reference Database Number 69, Eds. P.J. Linstrom and W.G. Mallard, June 2005, National
Institute of Standards and Technology, Gaithersburg MD, 20899 (http://webbook.nist.gov).
21
Lasheras, J.C., Villermaux, E., and Hopfinger, E.J., “Break-up and atomization of a round water jet by a high-speed annular
air jet,” J. Fluid Mech, Vol. 357, pp. 351-379, 1998.
22
Wu, P.-K., Kirkendall, K.A., Fuller, R.P., and Nejad, A.S., “Breakup Processes of Liquid Jets in Subsonic Crossflows,”
Journal of Propulsion and Power, Vol. 13, No. 1, 1997, pp 64-73.
23
Lorenzetto, G.E. and Lefebvre, A.H., “Measurements of Drop Size on Plain-Jet Airblast Atomizer,” AIAA Journal, Vol. 15,
No. 7, 1977.
24
Rizk, N.K. and Lefebvre, A.H., “Influence of Atomizer Design Features on Mean Drop Size,” AIAA Journal, Vol. 21, No.8,
pp 1139-1142, 1983.
25
Cavallo, P.A., and Grismer, M.J., "A Parallel Adaptation Package for Three-Dimensional Mixed-Element Unstructured
Meshes," Journal of Aerospace Computing, Information, and Communications, Vol. 2, No. 11, 2005, pp. 433-451.
26
Cavallo, P.A., Sinha, N., and Feldman, G.M., "Parallel Mesh Adaptation Method for Moving Body Applications," AIAA
Journal, Vol. 43, No. 9, 2005, pp. 1937-1945.

16
American Institute of Aeronautics and Astronautics

View publication stats