Professional Documents
Culture Documents
Edel Bauer 2016
Edel Bauer 2016
Edel Bauer 2016
PII: S1359-4311(16)30274-5
DOI: http://dx.doi.org/doi: 10.1016/j.applthermaleng.2016.02.131
Reference: ATE 7860
Please cite this article as: W. Edelbauer, D. Zhang, R. Kopun, B. Stauder, Numerical and
experimental investigation of the spray quenching process with an euler-eulerian multi-fluid
model, Applied Thermal Engineering (2016), http://dx.doi.org/doi:
10.1016/j.applthermaleng.2016.02.131.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will
undergo copyediting, typesetting, and review of the resulting proof before it is published in its
final form. Please note that during the production process errors may be discovered which could
affect the content, and all legal disclaimers that apply to the journal pertain.
Numerical and Experimental Investigation of the Spray Quenching Process with an
e-mail: dongsheng.zhang@avl.com
3
AVL-AST d.o.o., Trg Leona Stuklja 5, SI-2000 Maribor, phone +386 31 707 796,
e-mail: rok.kopun@avl.com
4
Nemak Linz AG, Zeppelinstrae 24, AT-4030 Linz, phone +43 664 / 516 2052,
e-mail: bernhard.Stauder@nemak.com
Highlights:
New powerful automatic time step sub-cycling procedure for the evaporation model
Simulation time speed-up due to new workflow for periodically frozen fluid flow
Good agreement between measured and simulated cooling curves of the test specimen
Abstract
An Eulerian spray model for simulations of spray quenching and cooling in water mist chambers has been
developed. Spray quenching allows targeted cooling of heated mechanical components to obtain increased material
strength properties by modifying the micro-structure evolution. For the numerical multi-domain simulation it was
necessary to implement a heat transfer model for the fluid-solid interface taking into account the increased heat
transfer due to phase change of impinging droplets. The droplet size distribution of the spray is represented by
discrete size classes for which separate transport equations are solved. The main challenge of the simulation was the
long process time of several minutes. To reduce simulation time, a new time step sub-cycling method for the
evaporation model and a method for periodically frozen fluid flow have been developed. The aim of this paper is to
present an enhanced Euler-Eulerian spray model capable for the simulation of the spray quenching process. For
1
Page 1 of 35
model validation, measurements of the temperature history at certain locations within a quenched test geometry have
been performed and compared with the numerically predicted results from the CFD simulations. The paper discusses
the numerical method, the experimental set-up, and demonstrates the feasibility of the workflow for industrial
applications.
Nomenclature
Variables:
A Area (m2)
d Diameter (m)
m Mass (kg)
2
Page 2 of 35
nph Number of phases (-)
p Pressure (Pa)
t Time (s)
T Temperature (K)
V Volume (m3)
3
Page 3 of 35
Kinematic viscosity (m/s)
Density (kg/m3)
Ratio of wall normal velocity scale and turbulence kinetic energy (-)
Subscripts:
C Collisions
E Evaporation
eff Effective
i, j Index notation
int Interface
k, l Phase indices
M Mass
P Primary break-up
Q Heat
s Species index
S Secondary break-up
sat Saturation
vap Vapor
w Wall
Superscripts:
t Turbulence
Abbreviations:
4
Page 4 of 35
DDM Discrete droplet method
1. Introduction
The ongoing demand on reduced weight and increased mechanical and thermal load of IC engine components leads
to sophisticated manufacturing processes of the components. Quenching is a well-established process for casted
aluminum alloy components like cylinder heads and engine blocks to increase the material strength properties by
modifying the micro-structure evolution. The residual stresses and the mechanical properties strongly depend on the
local time-temperature history obtained during the quenching process. One can distinguish between immersion
quenching and spray quenching. For the latter process galleries of nozzles perform the atomization of the cooling
fluid. The heat transfer from the solid to the fluid is determined by the impinging spray droplets and their
evaporation. In praxis, the optimization of these processes is expensive and time consuming due to many trial and
error experiments. The application of advanced numerical methods still plays a minor role in the development of the
production process.
The aim of this paper is to present a simulation approach for industrial spray quenching processes by applying an
Euler-Eulerian multi-fluid approach. Beside of modeling evaporation/condensation and droplet-wall heat transfer,
the main difficulty is to treat the time-scales of the involved processes. Typical spray quenching processes take place
within the time frame between 60 and 1200 seconds. The time required for single spray droplet to be heated,
evaporated and impinged is several orders of magnitude faster. Thus, simulation of the complete spray quenching
process would require unfeasible high number of time steps. By certain techniques presented in this paper, like sub-
cycling and temporarily frozen fluid flow field, the total number of time steps could be reduced and a significant
One key feature of the spray quenching simulation is the heat transfer model between impinging droplets and the
solid surface. This process is considerably different to that of immersion quenching which was discussed in previous
works by Srinivasan et al. [1] and Kopun et al. [2]. Since the late 1960s, spray cooling was investigated extensively
by Eugene and Mizikar [3], Gaugler [4] and others. Pioneering works mainly have been conducted for individual
droplet cooling and/or very low liquid droplet mass flux to obtain a basic understanding. Table 4 gives an overview
5
Page 5 of 35
about several experimental spray cooling and quenching studies, and it summarizes the range of their characteristic
parameters. The following four major heat transfer mechanism occurs during spray cooling and have been reported
by several authors, i.e. Pais et al. [5], Yang et al. [6], Rini et al. [7], Pautsch and Shedd [8], Shedd and Pautsch [9]
and others:
Mudawar and Estes [10], Visaria and Mudawar [11], and Lin and Ponnappan [12] proposed empirical models for the
critical heat flux of spray cooling. The applicability of these models is limited to their validated conditions. All three
boiling regimes appear in the spray quenching process, and consequently, although spray quenching is widely used
in industry, in the open literatures only limited investigations on industrial scale spray quenching process could be
found, as there are Wendelstorf et al. [13], Al-Ahmad and Yao [14], and Mascarenhas and Mudawar [15]. The
common conclusion from these publications is that the measured heat transfer rate strongly depends on the
temperature difference between wall and fluid, the local droplet mass flux, the droplet velocity and their size. Yao
and his co-operators (Cox [16], Choi and Yao [17] and Deb and Yao [18]) investigated laboratory-produced, mono-
dispersed spray cooling and obtained correlations for the heat transfer for a wide range of droplet mass fluxes from
0.016 to 2.05 kg/m2s. Mudawar and his team (Klinzing et al. [19], Bernardin and Mudawar [20], Mascarenhas and
Mudawar [15]) published a series of papers on the spray cooling, but the application range is narrow and strictly
defined. Based on extensive experimental data, Yao and Cox [16] provided two correlations for the spray heat
transfer. The heat transfer rate is either a function of the spray Reynolds or the Weber number. Wendelstorf et al.
[13] performed a series of measurements using full cone nozzles, where the droplet mass flow rates range from 3 to
30 kg/m2s and the temperature difference between hot surface and spray droplets varies from 150 to 1100 K. The
authors present a reasonable boiling curve correlation as function of the full temperature difference and the droplet
mass flux. Due to its wide application range, this correlation is adopted in this paper.
The modeling of droplet evaporation and condensation is a further important part of the spray quenching model. It is
based on single droplet models. Especially for the Euler-Lagrangian Discrete Droplet Method (DDM), as suggested
by Dukowicz [21], several specific evaporation models can be found in the literature, such as Amsden et al. [22],
6
Page 6 of 35
Varnavas and Assanis [23], Abramzon and Sirignano [24], [25], and several others. The latter model, which is
developed for non-unity Lewis numbers, has been adapted for Euler-Eulerian multi-fluid approach and is applied in
The paper is organized as follows. After the introduction in the first section, the second section discusses the
modeling approach, containing a summary of the entire Euler-Eulerian multi-fluid model, the turbulence,
evaporation and the heat transfer model for the droplet wall interaction. Then the sub-cycling methodology of the
evaporation model and the applied approach of the frozen fluid flow field are presented. The third section discusses
the performed experiment, the simulation set-up and the comparison between measured and numerically predicted
results. The presented models of this work have been implemented by the authors into the CFD code AVL-FIRE.
The standard model for the simulation of gas flows moderately laden with droplets is the Euler-Lagrangian Discrete
Droplet Method (DDM). Beside of many benefits, one drawback of the DDM is that an adequate number of spray
parcels in every computational cell is required to provide a statistically reliable description of the droplet size
spectrum in the entire computational domain. This can lead to high computational effort due to the large number of
spray parcels required for covering the entire computational domain. Furthermore, MPI parallelization efficiency of
the DDM suffers from parcels crossing the domain interfaces. In order to avoid these difficulties, the Euler-Eulerian
multi-fluid approach has been applied for the spray simulations in the present study. The solved transport equations
Based on the work of Drew and Passman [26], the complete set of conservation equations is solved for an arbitrary
number of interpenetrating phases. Each phase is characterized by its volume fraction . For the Euler-Eulerian
spray approach, one gaseous phase and an arbitrary number of phases for the droplet size classes are defined. The
mass, momentum and energy conservation equations for phase k with volume fraction k, are described by the
equations
(1)
(2)
7
Page 7 of 35
(3)
.
The terms kl, Mkl,i and Hkl represent the interfacial mass, momentum and heat transfer terms between phase k and
another phase l respectively. Uniform pressure, pk = p for all phases is assumed. The energy equation is solved for
the relative total enthalpy hk, which includes the thermal and the kinetic energies as described by
from hk with the integral specific heat capacity cP,k. The term qk is the heat flux vector consisting of a conductive and
The gaseous phase consists of an arbitrary defined number of gas species. For the spray quenching simulations the
gaseous phase is defined as composition of oxygen O2, nitrogen N2 and water vapor H2O. For each species s of the
gaseous phase k=g the transport equation for the vapor mass fraction Ys , described by
(4)
is solved. DY,eff is the effective diffusion coefficient. SY,s is the volumetric mass transfer rate of species s from phase
In general the nature of the flow in the spray quenching process is turbulent. The applied turbulence model is based
on the eddy-viscosity concept. Different to the isotropic standard k- model, as proposed by Launder and Spalding
[27], the applied four equation k---f model, proposed by Hanjalic et al. [28] and enhanced by Basara et al. [29],
[30] is not isotropic. It damps the turbulence close to the wall depending on the wall normal velocity scale . This
The k---f model is particularly suitable for predicting near wall turbulence. The authors propose a version of the
eddy-viscosity model based on Durbin's elliptic relaxation concept [31]. The aim is to improve numerical stability of
the original - model by solving a transport equation for the velocity scale ratio instead of the wall
8
Page 8 of 35
normal velocity scale . This model has been adopted for multi-fluid applications, and it is fully available in the
applied CFD code. For every phase the k---f model solves transport equations for the turbulence kinetic energy k,
the turbulence dissipation rate and the velocity scale . Additionally, elliptic relaxation equations for f, which is a
parameter closely related to the pressure-strain redistribution term, are solved for each phase. According to the
closure assumptions for multiphase turbulence of Lee et al. [32], it is assumed that the turbulence level of the
dispersed phase is equal to the continuous phase, and the interaction between these phases is neglected. A summary
of the k-f model equations can be found in the publication of Greif et al. [33].
The Eulerian spray model describes the interfacial exchange term of the Euler-Eulerian multi-fluid model with
respect to the droplet physics. The frame of the model is briefly described in this section. Further details can be
found in the publications of von Berg et al. [34], [35], and Edelbauer et al. [36]. Model enhancements and
validation, preferably on high pressure Diesel injections, are described by Vujanovi et al. [37] and Petranovi et al.
[38], [39]. The Eulerian spray model solves transport equations for the gaseous phase and an arbitrary number of
liquid droplet phases, where the transport equations are discussed in section 2.1. Each liquid phase represents a
droplet size-class. The phase specification of the Eulerian spray model is shown in Table 5. The first phase is
gaseous and the others are liquid representing the droplet size distribution.
In the Eulerian spray model the interfacial mass exchange term kl in equation (1), gets its contribution from droplet
evaporation E, primary break-up P, secondary break-up S, and collisions C as described by the equation
(5)
The mass exchange between two phases for a process acts in both directions, so that one can write Process,kl =
-Process,lk. This means that a mass gain for phase k is a mass loss for phase l and vice versa, so that mass
conservation over all phases is ensured. For the intended spray quenching simulations, the droplet size distribution at
the inlet is pre-defined based on the nozzle specification sheet. Consequently, the spray inlet boundary condition is
not located at the nozzle orifice, it is located a certain distance downstream, where break-up and collision processes
are in most instances finalized, and the droplet size distribution can be assumed to be constant. Thus, collision,
primary and secondary break-up models are not enabled in the performed simulations. Evaporation takes place
between the liquid phases from 2 to nph and the gaseous phase 1. The interfacial mass exchange rate due to
evaporation between the droplet phase k and the gaseous phase 1 is determined by the equation
9
Page 9 of 35
(6)
where Nk denotes the number of droplets per unit volume, and determines the evaporation mass exchange rate
of a single droplet. Assuming spherical droplets of diameter dk, the liquid volume fraction can be expressed by
(7)
where for the volume fraction k and the droplet number density Nk, separate transport equations are solved. This
means that within each size class the droplet diameter varies in space and time. Additionally the size class borders
are continuously adapted. The transport equation for the droplet number density Nk yields
(8)
The source term on the right-hand side, kl, gets its contributions from all droplet producing or annihilating
interfacial mass transfer processes. Evaporation and thermal expansion Th due to temperature change lead to
changes of the droplet diameter, but do not influence the number of droplets. Thus, E,kl and Th,kl are zero. The
child droplets from primary and secondary break-up or the coalescence from collision would create the sources P,kl,
S,kl, C,kl. Since these processes are not taken into account, the right-hand side of equation (8) is zero.
The momentum exchange between the gaseous phase 1 and the droplet phases k from 2 to nph, Mk1, is determined by
drag forces, MD,k1, and turbulent dispersion forces, MT,k1, as described by the equation
(9)
The drag coefficient cD is based on a correlation from ORourke [41] and depends on the droplet Reynolds number
(10)
The turbulent dispersion force is modeled following the approach of Lopez de Bertodano [42], with a constant or
This section briefly describes the applied evaporation/condensation model for the Eulerian spray simulation, and
discusses the effects of the simulation time steps. The applied model of Abramzon and Sirignano [24], [25] is a
10
Page 10 of 35
single droplet model for non-unitary Lewis number. It has been adapted for the Euler-Eulerian framework to provide
the mass transfer term in equation (6). In this section index k=d denotes the droplet phase and index g the
gaseous phase. The single droplet mass transfer rate is described by the equation
(11)
(12)
The overbar in equation (11) indicates mixture properties at a certain reference condition in the gas layer around the
droplet. is the mixture gas density, is the diffusion coefficient, and is the modified Sherwood number, as
(13)
Yamb is the vapor mass fraction of the ambient gas, and Yd is the vapor mass fraction on the droplet surface, which
depends on the ambient pressure p, the saturation pressure psat and the fluid properties, as described by the equation
(14)
M is the molecular weight and x is the mole fraction of the gas species. The heat transferred into the droplet interior
(15)
where is the specific heat capacity of the vapor at reference conditions, and L is the latent heat of the fluid. BT
is the Spaldings heat transfer number. From equation (15) one can see that the heat transfer is directly proportional
to the mass transfer. By multiplying with the droplet number density Nk, one obtains the interfacial heat
transfer term Hkl of the enthalpy transport equation (3). Furthermore, infinite conductivity inside the droplet is
assumed leading to homogenous temperature distribution within the droplet. This model has been implemented by
the author into the CFD code FIRE, and it was applied and validated on high pressure Diesel injections, as shown
in previous publications (Edelbauer [43]; Petranovi et al. [38], Vujanovi et al. [37], and Edelbauer et al.[36]). The
application of this evaporation model for spray quenching is considerably different. In typical Diesel sprays, the
11
Page 11 of 35
droplet size varies between approximately zero and fifty micrometers during evaporation, and the injection period is
in the order of few milliseconds. Consequently, typical simulation time steps are in the order of few microseconds,
or smaller depending on the resolution of the computational mesh. These time steps are usually small enough to
cover the physical processes for evaporation/condensation as well as for droplet break-up. For typical spray
quenching processes the situation is different. There, the process usually takes place over several minutes. Injection
pressure is usually much lower and the droplet size varies between zero and a few hundred micrometer. While the
size of the droplets is in comparable order, the total process time is much longer. To achieve reasonable
computational times, simulation time steps between 10-3 and 10-2 seconds are desired. These large time steps may
lead to problems which are illustrated on two examples in Appendix A to get an understanding about the relevant
The conclusion from these tests is that the evaporation effects under typical quenching conditions, are much faster
than the intended time step sizes for simulation of 10-3 to 10-2 seconds. The fast changes in temperature and diameter
cannot be captured by those time steps. Especially the sharp initial gradient of the liquid temperature can lead to
numerical problems and inaccurate solution, if the time step is not small enough within this period. To overcome
this problem, a time step sub-cycling procedure has been developed and implemented into the numerical solution
The idea of the time step sub-cycling procedure is not new. Amsden et al. [22] describes a methodology for the
DDM approach. The idea is to calculate the evaporation and condensation of every spray parcel with smaller time
steps, and to integrate the obtained mass and heat transfer rates between the liquid and the gaseous phase. The
choice of the sub time step is based on the assumption that during one CFD solver time step t the heat/mass
transfer to/from a computational parcel should not exceed some fraction of the energy/mass available for transfer.
This idea has been picked up for the Eulerian spray model to obtain integrated transfer rates which are then divided
by the CFD solver time step to provide the final sources for the volume fraction and enthalpy transport equations.
Amsden et al. (1989) derived the maximum time step for the Lagrangian spray parcels from considerations about the
maximum diffusive mass transfer by assuming unitary Schmidt number, as described by the equation
(16)
12
Page 12 of 35
The equation here already takes into account the gas volume fraction g which reduces the effective gas volume in
the computational cell. Factor fmax determines the maximum change of the gas mass in the cell, the chosen default
value of fmax is 0.5. Vcell is the cell volume, g is the gas density, Sh is the Sherwood number, g is the molecular
viscosity of the gas, and nd is the number of parcels, here droplets, in the computational cell. A similar criterion from
Amsden et al. [22] based on the maximum change of the gas energy in the cell leads to the maximum sub-cycling
(17)
g is the thermal conductivity and cp,g is the specific heat capacity of the gas. Nu is the Nusselt number. For adapting
these correlations in the Euler-Eulerian multi-fluid approach, the number of droplets nd per computational cell has to
(18)
Substituted it into the equations (16) and (17), the sub-cycling time steps yield
(19)
(20)
Nusselt and Sherwood number are calculated according to the well-known correlation of Ranz and Marshal [44]
number of sub-cycling time steps nsub for the computational cell Vcell and the droplet phase d is then simply
(21)
The test cases in Appendix A have shown that the initial change of the liquid temperature may become very large.
Thus, it is reasonable to use small sub time steps at the beginning and to increase the sub time steps later. Therefore,
the time steps are calculated from a geometric series as described by the equation
(22)
13
Page 13 of 35
For this approach the first sub-cycling time step tsub,1 and coefficient q have to be determined. It has been decided
that the last sub time step tsub,n is nsub times larger (but maximum ten times) than the first sub time step tsub,1, as
described by
(23)
(24)
and the first sub time step tsub,1 can be determined by the equation
(25)
All further sub time steps are then calculated by the equation
(26)
During the sub-cycling procedure the droplet diameter and temperature are continuously updated. The new droplet
diameter and temperature for the next sub time step i are obtained from simple mass and heat balance and calculated
by the equations
(27)
where md,i is the single droplet mass. The final exchange terms from evaporation for the volume fraction and
enthalpy transport equations between gaseous phase g and droplet phases d are then described by the equations
(28)
Optionally, it is possible to make estimations about the temperature and the vapor mass fraction in gaseous phase for
each sub-time step, as shown next. It can be shown from equation (15) that the rate of heat transferred into the gas
equals . Consequently, the gas temperature at the next sub-time step is described by the
equation
14
Page 14 of 35
(29)
The vapor mass fraction of the gas is directly influenced by the evaporated mass of all droplet phases in the
(30)
The updates of Tg,i+1 and Yamb,i+1 are only reasonable for CFL numbers less than unity. For CFL>1 the change of
temperature and vapor content in the computational cell is mainly influenced by the convective mass transport, and
consequently by the condition of the upwind cell. Thus, the gas property update should not be performed there.
The evaporation model is implemented as depicted in Figure 5, where the flow chart describes the steps performed
to determine the interfacial exchange terms of the volume and enthalpy transport equation due to evaporation. For
every cell a loop over all gas droplet interfaces, index iint, is performed. nsub according to equation (21) is calculated,
and the sub-cycling time step loop, index isub, is performed. After calculation of the time step size tsub, the mass and
heat transfer rates for the single droplet, and , are calculated. Then the intermediate values for droplet
diameter and temperature are calculated by equation (27), and optionally also the intermediate temperature and
vapor mass fraction of the gaseous phase in the computational cell is executed by equations (29) and (30). Here it
must be mentioned that, this update is a rough estimation, since convective effects are not taken into account.
Furthermore, the estimated values of Tg,i+1 and Yg,i+1 depend on the interface index iint and fits best at the last
interface and the last sub time step, since then the accumulation of transferred heat and mass in the current
During spray quenching, the prevailing heat transfer proceeds between the hot solid surface and the sub-cooled
droplets. It is reported by many researchers that the heat transfer rate between the hot surface and the impinging
spray droplets strongly depends on the temperature difference T=Tw-Tdrop and on the local droplets mass flux Vs.
Based on extensive measurements with water as coolant, Wendelstorf et al. [13] proposes the following correlation
(31)
This equation results from experiments with full cone nozzles (VKE 6/60, 6/90 and 8/60), where the nozzle diameter
15
Page 15 of 35
was 350 m, T was varied between 150 and 1100 K, and Vs was in the range between 3 and 30 kg/m2s. Figure 6
left shows the curve for three different mass fluxes of Vs1=3, Vs2=15 and Vs3=30 kg/m2s which cover the
experimental range and consequently the validity of this correlation. The corresponding heat flux is shown in
Figure 6 right. One can observe the critical heat flux occurs at T between 200 and 250 K, and the Leidenfrost
temperature limit is located at T greater than 500 K. Other researchers report that the heat transfer coefficient
additionally depends on the size of the impinging droplets. In the measurements performed by Wendelstorf et al.
[13] this effect was, owing to the high droplet mass fluxes, negligible.
It has to be noted that equation (31) is an integral heat transfer correlation. The measurements were performed on
laboratory samples (quenched surface is a circular disc with a diameter of 70 mm), and the temperature difference
T is defined as the difference between the hot wall and the coolant temperature in the nozzle. In the presented CFD
simulations, this correlation is applied locally for every discretized surface element. Droplet mass flux Vs and
temperature difference T are determined locally from the cells adjacent to the surface element. To overcome these
uncertainties between integral and local heat transfer correlation and the effect of different nozzle types, a correction
coefficient CTC, which is now a model parameter, has been introduced. Similar corrections are reported by Pola,
On the hot wall, evaporation mass exchange E,lg occurs between the droplets or the liquid film and the surrounding
gas, represented by phases l and g. The energy for this additional phase change results directly from the solid wall
(32)
where Tsat is the saturation temperature, Tw is the heated wall temperature and Aw is the surface area on the wall
boundary.
Due to multi-fluid flow and long process time the numerical simulation of spray quenching/cooling is
computationally expensive. In a primary survey, it is found that the effort for solving the transport equations of the
fluid, i.e. volume fraction, momentum, turbulence as well as pressure correction equation, dominates compared to
the effort for solving solid and fluid energy equations. Observing the flow field shows that after the spray droplets
arrive at the hot surface, the flow field varies very slowly with time proceeding. This behavior is also observed in
16
Page 16 of 35
the experiment. Consequently, it is practically meaningful and reasonable to keep velocity, pressure, volume fraction
and turbulence fields periodically frozen. This means that after the droplets impinged on the hot surface, the
transport equations for momentum, volume fraction, turbulence and the pressure correction equation are solved
periodically while enthalpy equations for fluid and solid domains are solved continuously. After a short period of a
few time steps where all transport equations are solved, there follows a longer period where only enthalpy equations
3. Computational Results
For the current study, experimental investigations of the spray quenching process have been performed. A simple
cuboid alloy made of AlSi7MgCu0 was quenched as displayed in Figure 7. Inside the work piece eight
thermocouples of type K (Ni-CrNi) with a temperature range of T=1200 K were located. Figure 7b shows the
dimensions of the cuboid sample and the positions of the eight thermocouples, illustrated by red dots from T1 to T8.
The holes for the thermocouples are positioned in the middle of the solid piece in a depth of 15 mm and a height of
30 mm. An air forced oven was used to heat up the work piece to homogeneous temperature of 780 K. Then the
sample was carried to the test station and placed on a fire refractory plate with the thickness of 40 mm. The
temperature profiles during quenching were recorded with the multi-meter DEWE-PORT 2000 at the frequency of
2 Hz.
The experimental set-up of the spray quenching process of the rectangular block specimen is summarized in Table 6.
A twin-fluid atomization nozzle operated with water and air was applied for the current study. The reproducibility of
the spatial position was secured by fixing screws, as shown in Figure 7a.
Figure 8 demonstrates the physical domain and the boundary conditions applied in the CFD simulations, and in
Table 7 the simulation set-up is summarized. The present computational domain consists of 800 000 cells in total,
where the solid domain has approximately 50 000 cells. The mean cell size of the solid domain is approximately
2 mm. The mesh dependency test, presented in Appendix B, has shown that this is an adequate cell size. The
computational domain has rather large dimensions in order to cover the whole working piece and to ensure enough
distance to the numerical boundary conditions. Atmospheric pressure is set at the static pressure boundaries on the
top, bottom and sides as depicted by the blue surfaces in Figure 8a. At the back side of the domain, a no-slip wall
17
Page 17 of 35
boundary condition with constant environment temperature of 20 C is applied (yellow surface). The fire refracting
plate, illustrated by the black surface, was not explicitly modelled as solid body, but an adiabatic wall boundary
condition with zero heat flux is applied. Since the experimental configuration is symmetric, only half of the
geometry is modeled in order to reduce the computational effort. Therefore symmetry boundary conditions,
illustrated by the green surface, are applied. For the spray, an inlet boundary condition with prescribed velocity and
volume fraction is applied. The normal distance between injector orifice and solid is 75 mm. Instead of resolving the
fine nozzle orifices of the six-hole atomizer, a ball-shaped inlet boundary condition is located 11.5 mm from the
nozzle orifice and 63.5 mm from the solid. The domain interface between fluid and solid domains, illustrated by the
red surface, is determined by a no-slip wall boundary condition. There, the wall temperature is calculated in order to
obtain balanced heat flux on both sides of the interface. The applied correction coefficient CTC is a model parameter
which reflects the differences between the empirical heat transfer correlation in equation (31) and the present
numerical model, as discussed in section 2.6. The maximum time step in the simulation is 0.01 seconds.
In the current study, two different quenching simulations have been investigated and compared with the
measurement data. The first simulation uses the time step sub-cycling methodology for the evaporation model as
described in section 2.5 but with the non-frozen fluid flow, whereas the second simulation uses both, sub-cycling
and temporary frozen fluid flow, as described in section 2.7. The initial droplet size diameters and the inlet volume
fractions are specified according to the data of the injection specification in order to approximate the droplet size
distribution from the manufacturer. The three droplet phases together with the gaseous phase lead to four phases in
total. The gaseous phase consists of three species, oxygen O2, nitrogen N2 and water vapor H2O for which separate
transport equations are solved. The three thermocouples, T1, T5, and T7 in Figure 8b, are used for comparison
between measured and numerically predicted temperatures results. Both simulations, non-frozen flow (NFF) and
The comparison of measured temperatures and numerically predicted results is presented next. It focuses on the
temperature and liquid phase distributions. The latter is represented by the so called total volume fraction which is
the sum of all liquid volume fraction values in the computational cell.
18
Page 18 of 35
94 hours. Figure 9 displays the comparison between measured and numerically predicted temperature histories at the
three monitoring locations, T1, T5, and T7. In general a good agreement was achieved, although the numerical
simulations slightly overestimate the cooling rates. It can be seen that at the monitoring points, which are closer to
the prolongation of the spray axis, positions T1 and T5, faster cooling than at the monitoring point T7 occurs. This is
due to the fact that at T1 and T5 more droplets are impinging on the hot surface and consequently, the amount of heat
consumed for evaporation is higher. It can be seen in Figure 6 right that with the initial solid temperature of 500 C
the cooling process starts below the Leidenfrost point. Consequently, the cooling curve passes the transition and
nucleate boiling regime. At the end time of 200 s, the maximum deviation between simulated and measurement
The total liquid volume fraction distribution at different time instants is presented in Figure 10, where red color
represents water droplets and blue is gas. It can be seen that at 0.01 seconds the spray has not reached the solid
surface. This takes place at around 0.03 seconds. During this free flow period only droplet evaporation takes place.
When the droplets get in contact with the hot surface, the evaporation rate increases dramatically due to the heat
transfer from the solid to the liquid. One can observe that the spray shape is relatively similar over the whole
simulation time. But at later time, when the work piece is already cooled down, one can observe higher droplet
concentration. Between 150 and 200 seconds, the quenched surface at the spray axis prolongation has almost
reached environmental temperature, and therefore the evaporation rate is very small.
Figure 11 displays the surface temperature distribution and the total liquid volume fraction at different time instants
during the entire quenching cooling process. At 0.01 seconds the solid has its initial temperature of 773.15 K, while
after 10 seconds the cooling is clearly visible. At this time monitoring point T 1 shows a temperature of
approximately 700 K while T7 has around 750 K. It can be seen that between 50 and 150 seconds the temperature
field is very heterogeneous and the temperature difference within the work piece reaches T=150 K. The numerical
results of all monitoring point locations show that the surface temperature has dropped substantially at the areas
closer to the prolongation of the spray axis, whereas slower cooling is observed at the outer monitoring points where
less droplets impact the hot surface, i.e. T7 and T8. As expected, the local droplet mass flux has dominant influence
The heat transfer coefficients and heat flux at three different time instants for one droplet phase are illustrated in
Figure 12. There, solid surface temperature, heat transfer coefficient and the heat flux are shown for the second
19
Page 19 of 35
droplet phase with 37 m class diameter (Table 7). It is important to mention that the total heat flux is the sum over
all droplet phases and the gas phase, and that is consequently considerably higher. The change of the heat transfer
owing to the transition between different heat transfer regimes, as discussed in Figure 6, is visible. For better
contrast in the figures, the color-bar has been adapted. While at the beginning at 0.04 s, where the spray has just
impinged the hot surface, the wall heat flux shows a minimum value, a few seconds later at 5 s the heat flux is
significantly higher, and at 100 s heat flux and heat transfer coefficient are reduced again. It can be concluded that
for this droplet phase the critical heat flux in the impinged area is passed at around 5 s. It can be also observed that
the change in the heat transfer coefficient at the three time instants is not as significant as in the evaluation in
Figure 6. This difference results from the fact that an integral heat transfer correlation is applied locally for every
the fluid flow was periodically frozen for every 3 seconds as discussed in section 2.7. During the initial period of 0.5
seconds all equations were solved followed by a frozen period where only enthalpy equation was solved up to 3
seconds. Then all equations were solved for further 0.5 s followed by a further frozen period of 2.5 s. This procedure
was repeated up to the final time of 200 seconds. This measure leads to a total simulation time of 37 hours compared
to 94 hours for the previous case. A speed-up of 2.5 was achieved. The temperature histories of the simulations with
non-frozen and alternating frozen fluid flow of the three monitoring points, T1, T5, and T7, are shown in Figure 13.
The difference between the two simulations is very small demonstrating the feasibility of this methodology. The
frozen fluid flow simulation shows a slight delay in the cooling curve.
This work presents an Euler-Eulerian spray model for spray quenching applications. The model was originally
developed for high-injection Diesel sprays, and has been enhanced for spray quenching simulations. Therefore, a
spray droplet-solid heat transfer model, taking into account the phase change of droplets impinging on a hot surface,
has been implemented into a CFD code. Two domains, fluid and solid, are coupled through a heat exchanging
interface. While in the fluid domain the complete set of transport equations is solved for one gaseous and an
arbitrary number of liquid droplet phases, in the solid domain only the enthalpy transport equation is solved. The
droplet size distribution is represented by the diameter classes of the liquid droplet phases. Numerical investigations
20
Page 20 of 35
of the evaporation and condensation on single droplets have been performed to demonstrate that droplet temperature
and diameter changes take place on short time scales. These time scales are considerably smaller than the desired
CFD time steps required for covering the quenching process. In order to reduce the number of time steps, a sub-
cycling procedure has been developed for the evaporation and condensation model. Additionally, a methodology for
periodically frozen fluid flow field for further reduction of the computational effort is presented. Experiments of the
spray quenching process have been performed on a prismatic specimen. The solid temperature histories have been
measured on certain locations within the work piece. There is a good overall agreement between the measured and
numerically predicted temperature curves. Theses curves are the base for stress simulations which will be performed
in the next step, in the same way as it was performed by Kopun et al. [45] and Li et al. [46]. Nevertheless, the
achieved accuracy of the cooling curves allows qualitative prediction of the stresses, for quantitative prediction
The current work has shown that the Euler-Eulerian spray approach is suitable for industrial spray quenching
applications. Through measures like time step sub-cycling and periodically frozen fluid flow the computational
effort could be considerably reduced. In the current work a speed-up of 2.5 between simulations without and with
periodically frozen fluid flow was achieved. As next step, further simulations with increased number of droplet size
classes, in order to have a better representation of the droplet size distribution, will be performed. The application of
this method on more complex work pieces, e.g. cylinder heads, is part of further investigation.
Acknowledgements
Parts of this research work were performed within the FFG funded project QUENCH-IT, project number 828697.
Other parts are funded by the European Union, European Social Fund and SPIRIT Slovenia, Slovenian Public
References
[1] Srinivasan, V., Moon, K. M., Greif, D., Wang. D.M., Kim M., 2010. Numerical simulation of immersion
quench cooling process using an Eulerian multi-fluid approach. Appl. Therm. Eng. 30 (5), 499-509.
http://dx.doi.org/10.1016/j.applthermaleng.2009.10.012.
[2] Kopun, R., Skerget, L., Hribersek, M., Zhang, D., Stauder, B., Greif, D., 2014, Numerical simulation of
immersion quenching process for cast aluminium part at different pool temperatures. Appl. Therm. Eng. 65
21
Page 21 of 35
[3] Eugene, A., Mizikar, A., 1970. Spray cooling investigation for continuous casting of billets and blooms.
[4] Gaugler, R. E., 1966. An experimental study of spray cooling of high temperature surfaces. PhD thesis,
[5] Pais M. R., Chow L. C. and Mahefkey E. T., 1992. Surface roughness and its effects on the heat transfer
[6] Yang J., Chow L. C. and Pais M. R., 1996. Nucleate Boiling Heat Transfer in Spray Cooling. J. Heat
[7] Rini D. P., Chen R. H. and Chow L. C., 2002. Bubble Behavior and Nucleate Boiling Heat Transfer in
[8] Pautsch A. G. and Shedd T. A., 2005. Spray impingement cooling with single- and multiple-nozzle arrays.
Part I: Heat transfer data using FC-72. Int. J. Heat Mass Transfer 48, 31673175.
[9] Shedd T. A. and Pautsch A. G., 2005. Spray impingement cooling with single- and multiple-nozzle arrays.
Part II: Visualization and empirical models. Int. J. Heat Mass Transfer 48, 31763184.
[10] Mudawar I. and Estes K. A., 1996. Optimizing and Predicting CHF in Spray Cooling of a Square Surface.
[11] Visaria M. and Mudawar I., 2008. Theoretical and experimental study of the effects of spray inclination on
two-phase spray cooling and critical heat flux. Int. J. Heat Mass Transfer 51, 23982410.
[12] Lin L. C. and Ponnappan R., 2004. Critical Heat Flux of Multi-Nozzle Spray Cooling. J. Heat Transfer 126,
482-485.
[13] Wendelstorf, J., Spitzer, K. H., Wendelstorf R., 2008. Spray water cooling heat transfer at high temperature
and liquid mass fluxes. Int. J. Heat Mass Transfer 51, 4902-4910.
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2008.01.032 .
[14] Al-Ahmadi, H. M., Yao, S. C., 2008. Spray cooling of high temperature metals using high mass flux
[15] Mascarenhas, N., Mudawar, I., 2012. Methodology for predicting spray quenching of thick-walled metal
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2012.02.018.
22
Page 22 of 35
[16] Yao, S. C., Cox, T. L., 2002. A general heat transfer correlation for impacting water sprays on high-
[17] Choi, K. J., Yao, S. C., 1987. Heat transfer mechanisms of horizontally impacting sprays. Int. J. Heat Mass
[18] Deb, S., Yao, S. C., 1989. Analysis on film boiling heat transfer of impacting sprays. Int. J. Heat Mass
[19] Klinzing, W. P., Rozzi, J. C., Mudawar, I., 1992. Film and transition boiling correlations for quenching of
[20] Bernardin, J. D., Mudawar, I., 1997. Film boiling heat transfer of droplet streams and sprays. Int. J. Heat
[21] Dukowicz, J. K., 1980. A particle fluid numerical model for liquid sprays. J. Comp. Phys, 35, 229-253.
[22] Amsden, A. A., ORourke, P. J., Butler, T. D., 1989. KIVA-II: A Computer Program for Chemically
[23] Varnavas, C., Assanis, D. N., 1996. A high temperature and high pressure evaporation model for the
[24] Abramzon, B., Sirignano, W. A., 1989. Droplet vaporization model for spray combustion calculations. Int.
[25] Abramzon, B., Sirignano, W. A., 1987. Approximate Theory of a Single Droplet Vaporization in a
Convective Field: Effects of Variable Properties, Stefan Flow and Transient Liquid Heating. Proc. in the 2nd
[26] Drew, D. A., Passman, S. L., 1998. Theory of Multicomponent Fluids, Springer, New York.
[27] Launder, B. E., Spalding, D. B., 1974. The numerical computation of turbulent flows. Comput. Meth. Appl.
[28] Hanjalic, K., Popovac, M., Hadziabodic, M., 2004. A robust near-wall elliptic-relaxation eddy-viscosity
turbulence model for CFD. Int. J. Heat Fluid Flow 25, 1047-1051.
http://dx.doi.org/10.1016/j.ijheatfluidflow.2004.07.005.
[29] Basara, B., 2006. An Eddy Viscosity Transport Model Based on Elliptic Relaxation Approach. AIAA
23
Page 23 of 35
[30] Basara, B., Krajnovic, S., Girimaji, S., Pavlovic Z., 2011. Near-Wall Formulation of the Partially Averaged
[31] Durbin, P. A., 1991. Near-wall turbulence closure modelling without damping functions. Theor. Comput.
[32] Lee, S. J., Lahey Jr., R. T., Jones Jr., O. C., 1989. The Prediction of Two-Phase Turbulence and Phase
Distribution Phenomena Using a k-e Model. Japanese J Multiphase Flow 3 (4), 335-368.
[33] Greif, D., Sampl, P., Edelbauer, W., 2014. Cavitating injector flow simulations considering longitudinal
[34] Von Berg, E., Edelbauer, W., Alajbegovic, A., Tatschl, R., Volmajer, M., Kegl, B., Ganippa, L. C., 2005.
Coupled Simulations of Nozzle Flow, Primary Fuel Jet Breakup, and Spray Formation. J. Eng. Gas
[35] Von Berg, E., Alajbegovic, A., Tatschl, R., Krueger, C., Michels, U., 2001. Multiphase Modelling of Diesel
Sprays with the Eulerian/Eulerian Approach. Proc. in 17th European Conference on Liquid Atomisation and
[36] Edelbauer, W., Suzzi, D., Sampl, P., Tatschl, R., Krueger, C., Weigand, B., 2006. New Concept for On-
Line Coupling of 3D Eulerian and Lagrangian Spray Approaches in Engine Simulations. Proc. in 10th Int.
Conf. on Liquid Atomisation and Spray Systems, ICLASS06, Kyoto, Japan, September.
[37] Vujanovi, M., Petranovi, Z., Edelbauer, W., Baleta, J., Dui, N., 2015. Numerical modelling of diesel
spray using the Eulerian multiphase approach, Energy Conversion and Management 104, 160-169.
http://dx.doi.org/10.1016/j.enconman.2015.03.040.
[38] Petranovi, Z., Edelbauer, W., Vujanovi, M. , Dui, N., 2013. Validation of Eulerian-Eulerian Approach
for Diesel Sprays, and 3D Coupling with Lagrangian Spray Approach. Proc. in 25th European Conference
[39] Petranovi, Z., Edelbauer, W, Vujanovi, M., Dui, N., 2014. The ORourke droplet collision model for the
Euler-Eulerian framework, Proc. in 26th European Conference on Liquid Atomization and Spray Systems,
[40] Pola A., Gelfi M. and Vecchia G. M La (2013). Simulation and validation of spray quenching applied to
24
Page 24 of 35
[41] ORourke, P. J., 1981. Collective Dop Effects on Vaporizing Liquid Sprays. Ph. D. Thesis, Los Alamos
[42] Lopez de Bertodano, M.A., 1998. Two fluid model for two-phase turbulent jets, Nucl. Eng. Des. 179,
65-74.
[43] Edelbauer, W., 2014. Coupling of 3D Eulerian and Lagrangian Spray Approaches in Industrial Combustion
[44] Ranz, W. E., Marshall Jr., W. R., 1952. Evaporation from drops Part II. Chem. Eng. Prog. 48 (4),
172-180.
[45] Kopun, R., Greif, D., Kovai, Z., Tatsch, R., 2013, Numerical Investigation of Immersion Quenching
Process for Cast Aluminum Parts Using an Eulerian Multi-Fluid Approach, Proc. in 27th ASM Heat
[46] Li, Z., Lynn Ferguson, B., Greif, D., Kovacic, Z., Urbars, S., Kopun, R., 2015. Coupling CFD and oil
quench hardening analysis of gear components, Proc. in 28th ASM Heat Treating Society Conference,
25
Page 25 of 35
Appendix A: Evaporation on Single Droplet under Typical Conditions
The aim of the two test cases presented in this section is to demonstrate the magnitude of the relevant time scales of
the evaporation/condensation process at typical spray quenching conditions. Therefore, the evaporation model
described in section 2.4 has been implemented as single droplet model into an explicit ordinary differential equation
(ODE) solver. There the ambient gas conditions, Yamb and Tg, are kept constant, and the evolution of droplet
The conditions of test case 1 are summarized in Table A.1. The initial droplet diameter is 10 m and the initial
temperature is 20 C. The ambient gas temperature is 300 C and there is no ambient vapor. These conditions are
typical for droplets entering the hot gas layer next to heated solid surface during the spray quenching process.
Figure A.1 shows the time evolutions of droplet temperature and diameter, and the Spaldings heat and mass transfer
numbers, BT and BM. Positive values of BT and BM indicate evaporation, negative condensation. One can observe that
there is a very fast change of the droplet temperature from 293.15 to approximately 326 K. Then the liquid
temperature as well as BT and BM stay constant when the so-called wet bulb temperature is reached. In this period
heat and mass transfer are in balance, meaning that the energy transferred from the gas to the droplet is fully
consumed by the mass transfer and the latent heat. Consequently the first and second term on the right-hand side of
equation (15) are in balance, and is zero. During this time the d 2-law, meaning that the time for evaporation is
proportional to the square of the droplet diameter, is valid. After 0.004 s the droplet is completely evaporated
indicating that the time for evaporation is in the same order as the desired time step size for simulation.
The initial conditions of the second test case are summarized in Table A.2 and the results are shown in Figure A.2.
There the initial droplet diameter is again 10 m, initial droplet temperature is 45 C, and the ambient gas
temperature has only 80 C but there is a considerable amount of vapor in the ambient gas. With these conditions
condensation takes place at the initial phase until the droplet temperature reaches equilibrium at approximately
1.510-4 s. In this initial phase BT and BM are negative. Then again balance between heat and mass transfer at the wet
bulb temperature is reached, and the droplet starts to evaporate without further temperature change indicated by
constant positive Spaldings heat and mass transfer numbers. Since the temperature difference between gas and
droplet of test case 2 is much smaller than that of case 1, the evaporation is slower.
26
Page 26 of 35
Initial droplet diameter dd 10 m
Initial droplet temperature Td 45 C (293.15 K)
Ambient gas temperature Tg 300 C (573.15 K)
Ambient vapor mass fraction Yamb 0.0
Table A.1: Single droplet evaporation for test case 1
Figure A.1: Droplet radius and temperature evolution (left) and Spalding heat and mass transfer number (right) of a
single droplet with initial temperature of 20 C and constant ambient conditions of 300 C and Yamb=0.0.
27
Page 27 of 35
Figure A.2: Droplet radius and temperature evolution (left) and Spalding heat and mass transfer number (right) of a
single droplet with initial temperature of 45 C and constant ambient conditions of 80 C and Yamb=0.137.
A grid size dependency test on a quasi-two-dimensional sector mesh has been performed to check the required
resolution. The simulation set-up is similar to that described in section 3.2. Three different meshes, the coarse
Mesh 1, the finer Mesh 2 and the finest Mesh 3, as depicted in Figure B.3, are applied for this test. The solid part,
marked with red color in Figure B.3, is cooled by the spray injected from the right hand side. The mesh resolution of
the solid part and the mean cell size, determined by the square root of the element surface area in the solid cross-
section, are listed in Table B.3. The height of the boundary layer on the fluid side is adopted according to the mesh
28
Page 28 of 35
Figure B.3: Computational meshes for the grid size dependency test: coarse Mesh 1 (left), fine Mesh 2 (middle),
finest Mesh 3 (right).
The cooling curves of the three different mesh configurations are shown in Figure B.4. The monitoring point for this
comparison is located in the center of the solid part. While the case with Mesh 1 shows faster cooling, the cases with
Mesh 2 and Mesh 3 show very similar cooling behavior. This means that the resolution of these meshes is sufficient.
For the simulations presented in section 3, a cell size of 2 mm has been chosen.
Figure B.4: Comparison of the cooling curves at a certain monitoring point for three meshes with different
resolution.
29
Page 29 of 35
Figure 5: Flowchart of the evaporation procedure.
Figure 6: Heat transfer coefficient (left) and heat flux (right) for three droplet mass fluxes of V s1=3, Vs2=15 and
Vs3=30 kg/m2s.
30
Page 30 of 35
Figure 7: Rectangular solid used in the experiment with the position of the thermocouples (a) and the test station for
spray quenching measurements (b).
Figure 8: Computational domain set up with boundary conditions (a) and positions of the monitoring points (b).
31
Page 31 of 35
Figure 9: Comparison of experimental (Exp) and numerically predicted solid part temperature for non-frozen fluid
flow (CFD_nFF) at monitoring points (a) T1, (b) T5, and (c) T7.
Figure 10: Total liquid volume fraction distribution at different time steps: (a) 0.1 s, (b) 10 s, (c) 50 s, (d) 100 s, (e)
150 s, (f) 200 s in cuts along the spray axis.
32
Page 32 of 35
Figure 11: Solid temperature and total liquid volume fraction distributions at different time steps: (a) 0.1 s, (b) 10 s,
(c) 50 s, (d) 100 s, (e) 150 s, (f) 200 s.
33
Page 33 of 35
Figure 12: Temperature, heat transfer coefficient and heat flux at different time steps for the second droplet phase
with 37 m class diameter: (a) 0.04 s, (b) 5 s, (c) 100 s.
Figure 13: Comparison of part temperature histories between non-frozen (CFD_nFF) and frozen fluid flow
(CFD_FF) at monitoring points (a) T1, (b) T5, and (c) T7.
34
Page 34 of 35
Table 4: Overview about some experimental studies on spray cooling and their characteristic parameters.
**Mudawar published a series of papers on spray cooling and or quenching, cooling area varied in each paper.
***Yao and Cox (2002) provided the correlation by compiling the available experimental data.
Phase 1 2,..,nph
Content Gas Droplets
Table 6: Experimental set-up
35
Page 35 of 35