You are on page 1of 12

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

net/publication/282733129

Numerical simulation of turbulent liquid jet breakup using a sub-grid criterion


with industrial application

Conference Paper · September 2014

CITATIONS READS

4 1,129

4 authors, including:

Mahdi Saeedipour Simon Schneiderbauer


Johannes Kepler University Linz Johannes Kepler University Linz
27 PUBLICATIONS   46 CITATIONS    90 PUBLICATIONS   598 CITATIONS   

SEE PROFILE SEE PROFILE

Stefan Pirker
Johannes Kepler University Linz
161 PUBLICATIONS   1,523 CITATIONS   

SEE PROFILE

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

Multiscale modelling of the primary breakup of liquid jets View project

Multi-scale Modelling of Multiphase Processes View project

All content following this page was uploaded by Mahdi Saeedipour on 19 October 2015.

The user has requested enhancement of the downloaded file.


ILASS – Europe 2014, 26th Annual Conference on Liquid Atomization and Spray Systems, 8-10 Sep. 2014, Bremen, Germany

Numerical simulation of turbulent liquid jet breakup using a sub-grid


criterion with industrial application

Mahdi Saeedipour*1, Simon Schneiderbauer1, Stefan Pirker1, Salar Bozorgi2


1
Department of Particulate flow modeling, Johannes Kepler University, Austria
2
LKR Leichtmetallkompetenzzentrum Ranshofen GmbH, Austrian Institute of Technology,
Ranshofen, Austria
*Corresponding author: mahdi.saeedipour@jku.at

Abstract
Disintegration and breakup of turbulent liquid jets with industrial application are numerically investigated using a sub-
grid breakup criterion and an Eulerian-Lagrangian coupling. An analytical model for flow instability leading to turbulent
breakup is considered to realize the droplet formation near the liquid jet interface, and then numerical simulations
using volume of fluid approach (VOF) are carried out to model the global spreading and primary breakup of the
turbulent liquid jet. We further present a novel idea to capture the droplet formation for such industrial application by
using an Eulerian-Lagrangian hybrid concept. This approach helps to reduce computational costs by using the sub-
grid model instead of resolving small liquid structures in finer grid resolutions. A finite volume-based solver is
developed within the framework of OpenFOAM to solve 3-D Navier-Stokes equations coupled with VOF equation and
presented Eulerian-Lagrangian coupling method. This methodology is tested for simulation of different liquid jets
validated against experimental data and showed good agreement. The generated droplets can be used for further
analysis of sub-processes like droplet-wall interactions and solidification.

Introduction
Fluid instabilities near liquid-gas interface might result in disintegration and breakup and influence the dynamic
behavior of the flow in most industrial flows [1]. High pressure die casting (HPDC) is one of the important but less
known manufacturing methods and is widely used for mass production of components based on aluminum,
magnesium or zinc alloys. In this process liquid metal is injected into the die at high speed (30-100 m/s) and under
high pressure through complex gate and runner systems [2]. The liquid jet is issuing near interface instability and
turbulence breakup during injection and filling. On the one hand, unstable wavy jet and disintegration might result in
cold run defects in the final product. On the other hand droplet formation and atomization may strongly increase the
porosity defect in the final casting. An in-depth fundamental understanding of the fluid instabilities and turbulent
breakup helps to predict defects and enables further considerations in filling pattern and solidification process in such
turbulent two-phase flows. Several factors such as liquid-gas turbulence, nozzle cavitation, velocity profile relaxation
mechanisms and liquid-gas properties have been identified to play role in the primary breakup process [3]. Over the
years, a number of efforts have been done to simulate liquid jet breakup with different methods mostly based on
solving Navier-Stokes equations coupled with a proper interface capturing method like Volume of fluid (VOF) [4],
Level set method [5] or both [6-7]. Regarding advantages and disadvantages of each different method, the
appropriate choice of modelling strategy may still vary from one problem to the other. (i) The dependency of droplet
formation to turbulence modeling and (ii) severe grid resolution requirement make the numerical simulation of primary
breakup still a computationally challenging problem. In recent years some hybrid frameworks have been developed
based on multiscale simulation. They involve mesoscopic particles or droplets suspended in macroscopic flow fields
[8]. In these frameworks numerical simulation is performed by either Eulerian-Eulerian or Eulerian-Lagrangian
approaches. The former resolves both scales in Eulerian formulation, while the latter includes solving the continuous
phase in an Eulerian scheme and model the discrete phase using Lagrangian formulation. Herrmann [9] developed a
coupling method for primary breakup of liquid jet in which all interface dynamics and physical processes occurring on
scales larger than the available grid resolution are fully resolved and all dynamics and processes occurring on sub-
grid scales are modeled in a Lagrangian framework by using a level set/vortex sheet method. Ham et al. [10]
developed a hybrid Eulerian-Lagrangian technique in which they resolve primary breakup using LES and surface
ILASS – Europe 2014, 8-10 Sep. 2014, Bremen, Germany

capturing method and leave the secondary breakup to the Lagrangian framework. Tomar et al. [8] performed a
multiscale simulation of primary jet breakup using an Eulerian–Lagrangian two-way coupling. Inspired from some
works above, Vallier [11] used similar methodology for multiscale simulation of cavitation sheets breakup. Grosshons
et al. [12] combined VOF-LES with LPT (Lagrangian particle tracking) to simulate jet atomization in a hybrid
framework.
In the present work an analytical model for flow instability leading to turbulent breakup is considered to realize a sub-
grid criterion for droplet formation, and then numerical simulations using volume of fluid approach (VOF) are carried
out to model the global spreading and primary breakup of the turbulent liquid jet. We further present a novel idea to
capture the droplet formation with less computational costs for such industrial application by using an Eulerian-
Lagrangian hybrid concept.

Physical model
Primary breakup of liquid jet and its secondary breakup called atomization occur due to instability. In other words, the
growth of disturbances along with the liquid-gas interface would cause instability and breakup. According to some
theoretical and experimental works carried out in 1990s [13-16], primary breakup of turbulent liquid-gas flow with a
density ratio greater than 500 is largely controlled by liquid turbulence properties with negligible aerodynamic effects
at these conditions. Droplet formation due to turbulent primary breakup occurs as a consequence of unbalanced
situation between disruptive and consolidating energy effects. If the kinetic energy of an eddy of characteristic size
relative to its surroundings plus the added mechanical energy due to the pressure drop caused by acceleration of the
surrounding gas over the tip of the perturbance could overcome the surface tension energy near the interface, one
droplet can be released [13] as illustrated in Figure 1. This sub-grid criterion reads:
̅ (1)

where (vli) is the velocity fluctuation perpendicular to main jet direction, (Csa) and (Csi) are two empirical constants
[12]. The term (li) denotes the turbulent length scale and can be estimated by the Taylor micro-scale.

Figure 1. The physical description of turbulent breakup near the liquid-gas interface (Redrawn from [13]).

It is declared in [13], li must be smaller than largest present eddies in the flow. As it is indicated in Figure 2, Taylor
micro-scale (λ) is the largest length scale located in inertial sub-range of turbulence energy cascade. It is obviously
smaller than the largest eddies present in flow and could be a good candidate for this physical model. Therefore li is
assumed equal to λ and is modeled by turbulent kinetic energy (k), viscosity (ν) and turbulence dissipation rate (ɛ) as
follows [17]:
ILASS – Europe 2014, 8-10 Sep. 2014, Bremen, Germany

√ (2)

Figure 2. Different length scales and ranges in turbulence energy cascade (Redrawn from [17]).

Numerical modelling description


The two-phase flow behavior of a liquid jet is governed by the incompressible Navier-Stokes equations with two-
phase mixture assumption which are coupled with volume of fluid equation for appropriate interface capturing (as
formulated in equations (3) to (5)). Surface tension ( ) is treated as a source term in momentum equation by CSF
method [18] acting only on the interfacial areas.

( ⃗) (3)

( ⃗) (4)


( ⃗ ⃗) ⃗ (5)

As stated in the introduction, by using a very high grid resolution, applying an appropriate turbulence modelling and
correct computational schemes, this set of equations could resolve all the liquid small structures after breakup (e.g.
ligaments, droplets and etc.), but needs huge computational expenses. Due to the limited computational resources
for industrial applications, a new sub-grid model for droplet breakup and a relatively accurate turbulence model
Reynolds stress model (RSM) are used to model the droplets smaller than grid size.
The sub-grid breakup criterion is applied based on an energy balance between disruptive and consolidating forces
near the interface as discussed in previous section. If this criterion is satisfied in any cell containing the interface, a
specific number of droplets are released at that position of the computational domain relative to the local properties of
the cell. The released droplets are then considered as Lagrangian droplets as illustrated in Figure 3.
In case of using a relatively coarse grid (in the same order of nozzle diameter), droplets are formed at least in one
order of magnitude smaller than the grid spacing in computational domain. The number of droplets in each
computational cell is proportional to the turbulence frequency (fr), turbulence length-scale (li), grid spacing (Δx) and
time step (dt) as follows:

( ) . (6)
ILASS – Europe 2014, 8-10 Sep. 2014, Bremen, Germany

Figure 3. Schematic of sub-grid breakup model and Eulerian - Lagrangian coupling.

Figure 4. An arbitrary computational cell containing liquid-gas interface and its length scales.

Figure 4 illustrates these length scales in an arbitrary cell containing the liquid-gas interface. While droplets with the
same properties of corresponding eddies are being released as Lagrangian representative particles, conservation
equations are influenced by different source terms as presented in equations (7) to (9).

( ⃗) (7)

( ⃗) ⃗
(8)


( ⃗ ⃗) ⃗ (9)

A numerical local coupling is established between continuous phase and Lagrangian droplets over the cells subjected
to breakup to keep the conservation equations still valid. In the equations above source terms could be modelled
based on the source term of volume of fluid equation. Indeed, by considering the difference between liquid volume
fraction before and after droplet release in each cell, the sink term for alpha equation can be formulated in equation
(10). Then the other source terms are modelled based on equation (10) only by adjusting physical dimensions.

(10)
ILASS – Europe 2014, 8-10 Sep. 2014, Bremen, Germany

Considering momentum equation, there are two additional source terms: Smom and Sdroplets. Both have negative
effects on the momentum equation which means they are acting like sink term for the momentum transport. The term
(Smom) is the momentum of released droplets in each cell containing droplet release. In other words, as droplets are
released in a specific computational cell, their momentum should be extracted locally at that time step. It could be
modelled based on number of droplets, their volume and their velocity after formation. The term (Sp) is the influence
of the droplets on the flow and is in interaction with the Lagrangian part of the solution. In the Lagrangian part, the
dynamics of particles are governed by Newton’s second law. Therefore position vector (xp) and velocity vector (Up) of
each particle are calculated from equations (11) and (12).

⃗ (11)


∑ (12)

In equations above, (mp) denotes the droplet mass calculated from the droplet volume and density at each time step.
The vector (F) is the representative of all forces acting on a droplet such as drag force, buoyancy force, pressure
gradient force, lift force and etc. In this work for small droplets, the first two forces are dominant and defined in
equations (13) and (14) [11].
⃗ ⃗
(13)

( ) (14)

where (g) is the gravitational acceleration and ( ) is the relaxation time of the droplets and reads [11]:

(15)
| |

The drag coefficient (CD) for a spherical droplet comes from experimental studies. There are different correlations for
this coefficient in the literature. Here the correlation from [19] is used as below:

[ ] (16)

where (Re) is the droplet Reynolds number defined as follows:


| |
(17)

The force exerted by a particle on a unit volume of fluid is proportional to the difference in particle momentum
between the instant it enters (tin) and leaves (tout) a specific cell.

[( ) ( ) ] (18)

Therefore the contribution of all the particles which occupied the cell k of volume (Vk) during one time step is written
as:

∑ [( ) ( ) ] (19)

A finite volume-based solver is developed within the framework of OpenFOAM [20] to solve incompressible 3-D
Navier-Stokes equations coupled with VOF equation with implementation of the above coupling method. This solution
is tested for simulation of different liquid jets and the results are presented in following section. This approach helps
to reduce computational costs by using the sub-grid model instead of resolving small liquid structures in finer grid
resolutions. The generated droplets can be used for further analysis of sub-processes like droplet-wall interactions
and solidification.
ILASS – Europe 2014, 8-10 Sep. 2014, Bremen, Germany

Result and discussion


In order to validate this coupling methodology, three turbulent liquid jets with different materials and injection
conditions are considered (see Table 1). The simulation results for the first case are compared to experimental study
by Mayer and Branam [21], while the next two cases are compared to data from in-house experiments. Three
structured computational grid networks are generated with the grid spacing in the order of magnitude of half
millimeter. For turbulence modelling, Reynolds stress model (RSM) is implemented to capture the velocity
fluctuations in different directions more precisely in comparison with other RANS-based models. Different inlet
boundary conditions are considered for each case.

Table 1. Three different turbulent liquid jets as the simulation test cases.

case U(m/s) Re We Oh Experiment


Ethanol 20 31000 31000 0.005 Mayer [21]
Hot Water 17 170000 16000 0.021 In house
Cold water 23 69000 22000 0.021 In house

For each case, droplet size distribution of generated droplets is plotted. The Sauter mean diameter (SMD) is
calculated by considering Normal distribution. The value is then compared to empirical correlation presented in [16]
reading:

( ) (20)

where (x) is streamwise distance from the jet exit, (cx) is an empirical parameter based on measurement and (ʌ) is
the radial integral length scale at the jet exit. In addition, the liquid volume flow rate during the simulation is compared
to theoretical liquid flow rate showing the mass conservation. The SMD values for each test case are presented in
Table 2.
Figure 5 shows a qualitative validation for this coupling method. A turbulent Ethanol jet with the velocity of 20 (m/s)
simulated is compared to experiment. The figure reveals that the instability and droplet formation around the core of
the jet are qualitatively matched with the experiment. In Figure 6 (left) the volume fraction of the liquid in the whole
domain is plotted over the simulation run time. The linear volume flow rate from the ingate is also plotted for
corresponding injection conditions. At the beginning of the simulation, values of both theory and simulation are the
same, while at the onset of droplet formation (t=1 ms), the simulation values start to deviate from the theoretical
volume flow rate. This could be justified by droplet extraction from the surface of the jet. If one add the volume of
droplet formed at each time step to the volume fraction coming from the Eulerian part of the simulation at that time
step, the volume of fluid in the system is conserved. This proves that in this methodology mass conservation is
satisfied. In Figure 6 (right) also shows the droplet size distribution for this jet at x/d=20. By considering normal
distribution over this histogram, SMD is about 130 micron. This is about 7% different from the value calculated by
empirical correlation in equation (20) at the same streamwise distance from the jet exit. This small validation case
shows the feasibility of using this method for further investigations.
Then two other different liquid jets are studies both numerically and experimentally to validate this sub-grid breakup
concept. In Figure 7 the simulation results for the hot water at different time steps are presented against the
experiments. These sets of experiments were performed in our laboratory using high speed camera. Figure 8
illustrates conservation of mass for whole simulation time and droplet size distribution for hot water jet at x/d=20. By
considering normal distribution over this histogram, SMD is 82 micron and is about 7% deviated from the value
predicted from the empirical correlation (20). The last case is the study of whole injection process for cold water with
the injection velocity of 23 (m/s) both experimentally and numerically. As can be seen in Figure 9, the experimental
snapshots and predicted form of the free jet by simulation look qualitatively similar. Also instabilities and droplet
formation over the jet surface are well captured for this test case. Similar two previous cases, Figure 10 shows mass
conservation for this case and droplet size distribution at x/d=20. The SMD calculated from normal disturbing of this
histogram is 79 micron which is only 3% deviated from the empirical correlation.
ILASS – Europe 2014, 8-10 Sep. 2014, Bremen, Germany

Table 2. Sauter Mean Diameter (SMD) values from simulation and empirical correlation at x/d=20 (in micron).

case SMD (simulation) SMD (empirical) Deviation


Ethanol 130 140 7%
Hot water 82 90 7%
Cold water 79 82 3%

Figure 5. Simulation results compared with experiments for Ethanol jet at t=2 ms.

Figure 6. Total liquid volume fraction rates for Ethanol jet (left), Droplet size distribution of ethanol jet at x/d=20 (right).
ILASS – Europe 2014, 8-10 Sep. 2014, Bremen, Germany

Figure 7. Simulation results compared with experiments for hot Water jet: at t=3.5 ms (left), at t=5 ms (right).

Figure 8. Total liquid volume fraction rates for hot Water jet (left), Droplet size distribution of hot Water jet at x/d=20 (right).
ILASS – Europe 2014, 8-10 Sep. 2014, Bremen, Germany

Figure 9. Simulation results compared with experiments for cold Water jet: t=1 (left), t=2 (middle), t=3 (right).

Figure 10. Total liquid volume fraction rates for cold Water jet (left), Droplet size distribution of cold Water jet at x/d=20 (right).

Conclusions
In this research, disintegration and breakup of turbulent liquid jets with industrial application are numerically
investigated using a sub-grid breakup criterion and an Eulerian-Lagrangian coupling. This work is basically motivated
by industry to study the fluid dynamics of liquid jets and sheets in High pressure die casting (HPDC) process. An
analytical model is considered for droplet formation criterion. A numerical hybrid framework is established to predict
the general behavior of liquid jet by solving Navier-Stokes equations coupled with VOF method. Droplets smaller than
the grid size are generated in the domain and their dynamics are modeled in a Lagrangian approach. Although this
novel idea of coupling influences the conservation equations by exerting different source terms, it helps to reduce
computational costs for simulation of jet breakup for many industrial applications. Three different turbulent liquid jets
are simulated using this methodology. The results are compared with the experimental data and show agreements in
modeling the global spreading of the jets. By considering droplet size distribution from simulation, Sauter mean
diameter (SMD) shows only 3-7 % deviation from the SMD result using the empirical correlation in the literature. This
method could be extended by more detailed experimental verification as well as further studies on sub-processes
involved with droplets like droplet-wall interactions and solidification.
ILASS – Europe 2014, 8-10 Sep. 2014, Bremen, Germany

Acknowledgments
This work was funded by the Christian-Doppler Research Association, the Austrian Federal Ministry of Economy,
Family and Youth, and the Austrian National Foundation for Research, Technology and Development. The authors
also would like to thank for supporting the project MEEEPRO-CAST in the framework of the EU-programme Regio 13
sponsored by the European Regional Development Fund (EFRE) and by the State of Upper Austria.

Nomenclature
α Volume of fluid [-]
2 3
ɛ Turbulence dissipation rate [m /s ]
k Turbulence kinetic energy [j]
2
ν Kinematic viscosity [m /s]
d Jet diameter [m]
D Droplet diameter [m]
m Mass [kg]
fr Frequency [Hz]
2
g Gravitational acceleration [m/s ]
F Force [N]
t Time [s]
λ Taylor micro scale [m]
li Eddy size [m]
3
ρ Density [kg/m ]
x Streamwise distance from the jet exit [m]
xp Droplet position vector [m]
U Velocity [m/s]
3
V Volume [m ]
CD Frag coefficient [-]
Re Reynolds number [-]
We Weber number [-]
Oh Ohnesorge number [-]
Cx Empirical parameter in SMD correlation [-]
Csa Empirical parameter in breakup model [-]
Csi Empirical parameter in breakup model [-]
SMD Sauter mean diameter [m]

References
[1] Schneiderbauer, S. et al., IOP Conference series, Material Science Technology 27:012067 (2012).
rd
[2] Saeedipour, M. et al., 143 TMS annual meeting, San Diego, USA, 16 - 20 February 2014.
[3] Srinivasan, V. et al., Atomization and Sprays 18:571-617 (2008).
[4] Hirt, C.W. and Nichols, B.D., Journal of Computational Physics 39:201-225 (1981).
[5] Osher, S. and Sethian, J.A., Journal of Computational Physics 79:12-49 (1988).
[6] Sussman, M. and Puckett, E.G., Journal of Computational Physics 162:301-337 (2000).
[7] Menard, T. et al., International Journal of Multiphase Flow 33:510-524 (2007).
[8] Tomar, G. et al., Computers & Fluids 39:1864-1874 (2010).
[9] Herrmann, M., CTR Annual Research Briefs: 185–196 (2003).
[10] Ham, F. et al., Computational methods in Multiphase Flow II, Advances in Fluid Mechanics, WIT Press
37:313-322 (2003).
[11] Vallier, A., 2010, "Eulerian and Lagrangian cavitation related simulations using OpenFOAM."
th
[12] Grosshans, H. et al., 7 International Symposium on Turbulence and Shear Flow Phenomena TSFP7,
Ottawa, Canada, 28 - 31 July 2011.
rd
[13] Wu, P.K. and Faeth, G.M., 31 Aerospace Science Meeting & Exhibit, Reno, USA, 11 - 14 January 1993.
[14] Wu, P.K. and Faeth, G.M., Physics of Fluids 7 (11):2915-2917 (1995).
[15] Sallam, K.A. et al., International Journal of Multiphase Flow 25:1161-1180 (1999).
ILASS – Europe 2014, 8-10 Sep. 2014, Bremen, Germany

[16] Sallam, K.A. et al., International Journal of Multiphase Flow 28:427-449 (2002).
[17] Pope, S.B., 2000, "Turbulent Flows."
[18] Brackbill, J.U. et al., Journal of Computational Physics 100:335-354 (1992).
[19] Ashgriz, N., 2011, "Handbook of Atomization and Sprays: theory and applications."
[20] OpenFOAM, The open source CFD toolbox, http://www.OpenFOAM.org .
[21] Mayer, W.O.H. and Branam, R., Experiments in Fluids 36:528–539 (2004).

View publication stats

You might also like