9

Proceedings of ASME Turbo Expo 2018

Turbomachinery Technical Conference and Exposition
GT2018
June 11-15, 2018, Oslo, Norway

GT2018-76578

MULTI-PHYSICS SIMULATIONS WITH OPENFOAM IN THE RE-DESIGN OF A

COMMERCIAL COMBUSTOR

Marco Konle Ludovic de Guillebon Christopher Beebe

MTU Aero Engines AG MTU Aero Engines AG Vericor Power Systems
Munich, Germany Munich, Germany Alpharetta, US

ABSTRACT k Turbulent kinetic energy [m2.s-2]

The numerical description of combustion provides a good p Pressure [Pa]
understanding of the thermal boundary conditions of a Pr Laminar Prandtl number -
combustor already in the early design phase. With an accurate Prt Turbulent Prandtl number -
simulation, the thermal load on the combustor wall as well as si Direction vector -
the turbine inlet profile can be estimated and potentially Sc Laminar Schmidt number -
tailored to the structural requirements. Sct Turbulent Schmidt number -
At MTU Aero Engines AG, the authors set up a multi- t Time [s]
physics solver in OpenFOAM to support the development T Temperature [K]
process for combustors. This article shows the activities for a Tij Reynold stress tensor [N.m-2]
redesign of a commercial engine combustor and the successful ui Velocity [m.s-1]
numerical description of all related physical aspects. The Yk Mass fraction of species k -
comparison of combustor liner temperatures predicted via
Conjugate Heat Transfer (CHT) with thermal paint results show α Absorption coefficient for radiation -
excellent agreement. The numerical results allow the evaluation ε Turbulent dissipation rate [m2.s-3]
of different redesign options with respect to maximum liner κ Reactive volume fraction -
temperatures and impact on turbine inlet temperature profiles. µ Molecular viscosity [kg.m-1.s-1]
The application of a multi-physics solver to real engine µt Turbulent viscosity [kg.m-1.s-1]
combustor designs is limited to the availability of ρ Density [kg.m-3]
computational power and acceptable running time. The σ Stefan-Boltzmann constant [W.m-2.K-4]
applicability of the presented approach to complex engine τc Chemical timescale [s]
designs is not only shown by the successful simulation results. τij Viscous stress tensor [N.m-2]
The choice of an open source code allows the usage of massive τmix Turbulent timescale [s]
parallelization. A parallelization up to 1024 CPUs was tested to τT Taylor timescale [s]
evaluate the potential of speeding up the simulations. The τη Kolmogorov timescale [s]
results are encouraging and show that complex meshes with ω Turbulent eddy frequency [s-1]
~15 million cells can be simulated within acceptable ωk Mass reaction rate of specie k [kg.m-3.s-1]
calculation time. ωrad Enthalpy source term from radiation [W.m-3]
ωT Heat release due to combustion [W.m-3]
Ω Solid angle [sr]

NOMENCLATURE ā Reynolds averaging mean part for scalar a

E Emission contribution [W.m-2] ã Favre averaging mean part for scalar a
h Sensible enthalpy per unit mass [[Link]-1]
I Radiation intensity [W.m-2]

1 Copyright © 2018 ASME

Downloaded From: [Link] on 09/12/2018 Terms of Use: [Link]

 ACRONYMS character of the reacting flow field of the combustor
CHT Conjugate Heat Transfer investigated in this paper requires a temporal resolution and,
fvDOM finite volume Discrete Ordinates Method thus, the URANS approach (k--SST) is chosen to describe the
ID Inner Diameter flow field and turbulence. The solver includes additionally
OD Outer Diameter gaseous radiation as well as the radiative heat load contributed
OTDF Overall Temperature Distribution Factor by soot formed in the combustion process. Finally, the
PaSR Partially Stirred Reactor conjugate heat transfer provides the information required to
RTDF Radial Temperature Distribution Factor evaluate the combustor design with respect to thermal loads
SST Shear-Stress-Transport and structural life. Briefly summarized the models combined:
URANS Unsteady Reynolds Average Navier-Stokes
 The mass, momentum, species, and enthalpy conservation
equations below are solved to predict the flow behavior
INTRODUCTION and are solved within the Unsteady RANS turbulence
To predict accurately the thermal load on combustor walls, approach according Menter [6], the k--SST model.
the numerical modeling of such components has to cover a lot
 
of physical aspects. Besides the precise prediction of the   u~i   0 . (1)
t xi
reacting flow field, the gas radiation capturing also the soot
contribution as well as the heat transfer onto the combustor

 u~i     u~i u~ j    p    ij  Tij 
walls via conjugate heat transfer (CHT) have to be taken into . (2)
account. Due to the high demand of computational power for t xi x j xi
the described modeling depth, MTU used the open source CFD
~
     
     Yk
software OpenFOAM and developed a multi-physics solver to  ~  ~
 Yk   Yk u~i    t    k . (3)
capture all these aspects. This modeling approach was t xi xi   Sc Sct  xi


successfully validated with different academic and semi-
technical test cases. Main vehicle for this validation was the so- ~
     t 
called FIRST combustor designed and investigated by the DLR 
t
 
~
h 

xi
~

 h u~i    
xi   Pr Prt
 h
   T  rad . (4)

Stuttgart [1]. The geometry as well as the test data were  xi 
provided as test case for the international sooting flame (ISF)
workshop 2016. Additionally, the combustor designs of the where the subscripts i and j represent the xi and xj direction
European combustion research projects TIMECOP-AE [2], components, k the species and t the turbulent part [7]. ui is
[3]and TECC-AE [4] were used to validate the solver with the velocity, ρ the density, p the pressure, Yk the mass
respect to different combustor designs and operation ranges. fraction of species k, and h the enthalpy. µ and µt are the
Results were compared to the test data provided by the DLR molecular and turbulent viscosity, Sc and Sct the laminar
combustion group as well as to numerical results published in and turbulent Schmidt number, Pr and Prt the laminar and
the community (e.g. [5]). The numerical approach to cover all turbulent Prandtl number. τij and Tij symbolize the viscous
physical aspects of interest as well as its validation was and the Reynold stresses. Moreover, ωk, ωT, and ωrad are
published in more detail in [4]. The validated solver was source terms for the equations. The production and
recently used to support the re-design process of a real engine consumption of reactive species due to combustion is
combustor. Main aspects for this application were: described by ωk and heat release due to combustion by ωT.
 numerical description of flow field around and inside the ωrad represents the radiation effects. The solver used by the
combustor authors is intended to calculate flow fields with Mach
 evaluation of flow splits (primary combustion air, cooling numbers below 0.3. However, the density changes caused
air flows) by the combustion required to implement all further
 evaluation of local equivalence ratio distribution models into the OpenFOAM solver rhoPimpleFoam. To
 description of local reaction zone and related radiation reduce the computational time, the simulations were
 metal temperatures of combustor liners and turbine inlet started from RANS results. As mentioned above, the
temperature profile transient character of the problem, however, required the
temporal modelling approach of URANS. For average
In the following section, details of the modelling approaches field information, transient phenomena like e.g. a
and evaluation steps are presented, before the numerical results precessing vortex core had to be averaged out. The
will be compared to experimental data. comparison of these averaged field information with
RANS solutions confirmed the necessity of the transient
simulation.
NUMERICAL MODELING
The MTU internally developed multi-physics solver is
based on the open source tool of OpenFOAM. The transient

2 Copyright © 2018 ASME

Downloaded From: [Link] on 09/12/2018 Terms of Use: [Link]

  The combustion is solved using the Partially Stirred  For the soot formation in the combustor, the two-equation
Reactor model (PaSR) and, with it, the empirical Arrhenius model of Magnussen/Tesner [11] is implemented. This
law (Kärrholm, [8]). The PaSR model is a finite rate semi-empirical soot model calculates in a first step the
chemistry method which assumes that the real flame is specific concentration of radical nuclei and in a second
much thinner than any computational cell. Thus, each cell step the formation of soot out of these nuclei. The
is divided into a reacting part, in which all present species oxidation of nuclei and soot particles (i.e. destruction of
are homogeneously mixed and react together, and a non- soot) is modelled by scaling the reaction rate of the
reacting part. The reactive volume fraction κ, which is a combustion model to the soot concentration. The impact of
multiplying factor for ωk and ωT in Eq. (3) and (4), results turbulence on the mean reaction rate according the Eddy
of the turbulence-chemistry interactions: Dissipation Concept (Magnussen [11]) limits the soot
formation. Besides that, the limitation to a maximum soot
c
 (5) level is implemented (Kleiveland, [12]). Further details,
 c   mix especially about the validation of this model, can be found
in [4].
  t (6)
 mix  Cmix   T  Cmix 
 The calculation of the thermal load on the combustor walls
is performed via a Conjugate Heat Transfer (CHT)
where τc represents the chemical timescale and τmix the solution. Again starting with the CHT solver implemented
turbulent timescale. τη and τT are the Kolmogorov and in OpenFOAM, the authors implemented the models
Taylor timescale, respectively, and Cmix a model constant. described above into this solver.
This constant as well as the turbulent Schmidt number Sct There are several approaches used for Conjugate Heat
for the thermal diffusion were calibrated via measured Transfer simulations. The fully-coupled approach solves
temperature profiles at the exit of the combustors used for the solid response simultaneously to all other equations
the validation [2-4]. Since the commercial combustor solved for the fluid domain. The implementation is,
investigated here was operated with natural gas only, any therefore, straight forward. However, due to the different
modelling effort with respect to liquid fuel was cancelled. time scales of the solid response and inside the flow field,
The chemical reaction was described by the Arrhenius the fully-coupled approach is very time and resources
inputs for one step chemistry of methane. consuming. Thus, another possibility used e.g. in the
commercial solver Ansys CFX is to use different time
 High temperatures in combustors lead to significant scales for the solid and fluid domains. The basic
radiative heat transfer. To cover the contribution of the hot assumption is that the solid response is calculated with a
walls as well as the gaseous radiation to the thermal loads quasi steady-state fluid field and, thus, the solid result is
on the combustor, this phenomenon has to be taken into only weakly influencing the liquid result. This approach
account. To predict radiation, two aspects has to be provides acceptable results as long as the reacting flow
considered: the spatial and the spectral radiation. In field has only a limited transient character close to the
OpenFOAM, the fvDOM (finite volume Discrete Ordinate walls. Highly fluctuating flow fields due to transient flow
Method) solves a simplified form (Eq. 7) of the radiative structures require a careful data post-processing to evaluate
transfer equation (Viskanta and Mengüç, [9]) for a finite potential temperature fluctuations inside the solids. For this
number of discrete solid angles (16 rays in the present reason, and additionally, to increase the performance of the
study) and returns the ωrad source term. solver and to reduce the computational time significantly,
the solver was adapted to allow the calculation of the wall

I si     I     T 4  E  . (7) heat flux in the solid via a frozen-fluid-option. Figure 1
xi  4 shows a scheme of this approach: Starting from an
with I the radiation intensity, si the xi-component of the established flow solution, the fully-coupled model
direction vector, Ω the solid angle considered, T the including fluid domain and solid domains is simulated with
temperature, σ the Stefan-Boltzmann constant. E is the a reasonable time step to achieve a Courant number ~0.4 in
emission contribution, an offset which is set to 0W.m-2 for the fluid domain. Since the solid response is significantly
this study. The absorption coefficient of gas, α, is slower (in the order ~ 1s), a pure CHT simulation with
calculated with the spectral model. In this study, the grey time steps in the order ~1e-07s is not achievable in
gas model is used and takes into account the acceptable time. Thus, the CHT solver iterates for a real
absorption/emission of the two species CO2 and H2O [10]. time of e.g. 0.001s before the fluid domain gets frozen and
Due to the significance of soot radiation, the contribution only the conduction inside the solid parts is solved with
of soot to the gaseous radiation were added to the standard time steps 1000 times larger. This way, the quasi steady
implementation. state solution for 1s stabilization can be achieved. By
iterating such loops and monitoring characteristic locations

3 Copyright © 2018 ASME

Downloaded From: [Link] on 09/12/2018 Terms of Use: [Link]

 of the liners, transient temperature fluctuations inside the overall pressure ratio of 10:1. More details about the engine
solids can be detected. can be found on [Link] including data sheets for the
different applications. The computational domain covers the
diffusor casing at the compressor exit, the combustor (casing,
liners, burner head, injectors) and the entry into the high
pressure turbine. Figure 4 shows a snapshot of an axial cut
through the domain.

Fig. 1: Solver scheme to achieve quasi steady state

temperatures for the solid domains.

Figure 2 displays two monitoring points set on both sides

of the fluid-solid interface, spatially separated by about 0.5mm.
The figure visualizes also the loop strategy: While the solid is
solved to every iteration, the fluid domain is periodically
frozen. As consequence, the curve for the fluid monitoring
point shows a stepwise behavior, but the transient fluctuation is
obvious. Furthermore, the trend of the monitoring points for
both, fluid and solid, is converging which indicates the
achievement of a stationary temperature solution. For the
Fig. 3: Picture of the Vericor TF50F engine.
combustor design investigated in this paper, the fluctuations
inside the solids are very limited (fig. 2 red curve). Thus, the
The numerical study conducted by MTU had the objective
numerical approach provides very good results for metal
to analyze geometric variations in the combustor liners design
temperatures to be expected during stationary operation of the
with respect to flow split, reaction zone location, and resulting
engine. The local distribution of maximum temperatures can be
liner temperatures. To achieve a reliable flow split result, the
predicted to allow reliable analysis of combustor liner life
model requires a sufficient resolution of the cooling features.
times.
The high density of nodes in these areas is visible in fig. 4. The
number of cells for this configuration is in the order of ~15
million cells.

Fig. 2: Monitoring points in fluid and solid domain for a quasi-

steady-state of CHT.

APPLICATION TO A COMMERCIAL COMBUSTOR

The design investigated in this paper is based on the TF50
Fig. 4: Numerical domain of investigated combustor TF50F
combustor of Vericor Power Systems. This engine is an
aeroderivative gas turbine, which originates from the Lycoming
Figure 5 shows a plot of the isothermal velocity field on
T55 military helicopter engine and the ALF502/LF507
the central plane for the base configuration operating in the
commercial turbofan engines, and has been adapted for use in
field. The hole pattern provides dilution air through the first
mechanical drive applications (TF50F) but also used in marine
and second combustor liner segments (location #1 and #2,
applications (TF50B) i.e. ships and hovercrafts. Figure 3 shows
respectively in fig. 5) of the combustor and additional cooling
a picture of the engine. The TF50 is rated at a maximum
air before the 180° turn via so-called scrubber tubes (loc. #3) to
continuous power (MCP) output of 3800 kW (5100 hp), at an
cool down the outer transition liner. Despite of a low swirl level

4 Copyright © 2018 ASME

Downloaded From: [Link] on 09/12/2018 Terms of Use: [Link]

 of the combustor head the cooling air of the first hole row (#1) field. Additionally, the reaction zone is visualized by the black
gets sucked to the burner exit by the recirculation zone. lines showing 5% of maximum reaction rate and, thus,
envelopes the reaction zone. The flame is not compact but very
elongated over a wide area. The resulting thermal load is spread
over the whole axial length of the combustor liners. And the
difference in the air jet penetration leads, as expected, to a
significant shift of the flame to the ID side. The conventional
combustor system shows high temperatures on the burner axis
with maximum temperatures above 2100K. The air flow
through the scrubber tubes (fig. 5 loc. #3) does not lead to a
significant mixing but just to a tailoring of the temperature
profile at the combustor exit.

Fig. 5: Isothermal velocity field on the middle plane of the

TF50F

To evaluate the reliability of the numerical simulations, the

flow split of the base configuration was analyzed and compared
to existing information based on calibrated 1D tools and test
data. Figure 6 shows the values of percentage cooling air over
the primary and secondary hole rows on the outer and inner
combustor liner segments. The significant pressure drop which
occurs by turning from the OD cavity to the ID cavity leads to
significantly smaller values on the inner side. As a result, the
penetration of the ID flows is significantly lower. This is also Fig. 7: Temperature field and reaction rate isoline (black line
visible in the velocity magnitude plot shown in fig. 5. The envelopes the reaction zone) for the TF50F.
simulation results could be confirmed by the mentioned 1D
tools and, thus, the modelling approach was verified for further CFD ANALYSIS OF A RE-DESIGN OPTION
use. The studies of potential redesign options were focused on
variations of the hole pattern on the combustor liner sections
close to the primary reaction zone. Figure 8 shows one variant
of modified combustor liner hole pattern, here called
configuration A. The first and second combustor liner segments
got additional holes to provide a significant amount of air close
to the combustor head. Since the available amount of cooling
air was a fixed boundary condition, the flow of the scrubber
tubes (fig. 5 loc. #3) was re-distributed to the front.
Additionally, the swirl of the combustor head was significantly
increased to achieve a compact reaction zone. The isothermal
flow field indicates already a significant shift of the flame to
the ID side due to even weaker penetration on the ID side.
Figure 9 shows the resulting temperature field including the
black lines enveloping the reaction zone. The expected shift is
Fig. 6: Flow split of cooling air over the first and second very strong and, thus, leads to a significant increase in the
combustor liner segments for the TF50F. thermal load on the ID side as well as to very high flow
temperatures in this area.
As consequence of the non-symmetric flow penetration
through the OD and ID holes, respectively, shown in fig. 6,
the reaction zone is expected to be unsymmetrically shifted to
the ID. Figure 7 shows the contour plot of the temperature

5 Copyright © 2018 ASME

Downloaded From: [Link] on 09/12/2018 Terms of Use: [Link]

 The resulting configuration was then manufactured and
tested, while the analytical analysis was extended to the CHT
simulation for this configuration. In the following figures, the
experimental and the numerical results for this configuration
will be presented and discussed.
Figure 11 shows thermal paint results for the combustor
testing. The different colors indicate zones of different
temperatures. Since the color changes are linked to specific
temperature values, the borders between different colors
indicate discrete temperatures. The areas of same colors just
give a qualitative information. The legend included in the
figure is the color palette of the thermal paint used for the
testing. Main aspects of this kind of measurement technique is
to identify local hot spots and temperature maxima as well as a
spatial information.

Fig. 8: Velocity field on the middle plane of the modified

combustor.

Fig. 11: Thermal paint results of steady state operation at max.

continuous power for combustor configuration B.

Areas of special interest are the combustor dome, the

second combustor liner segments with the dilution holes, and
the outer transition liner (fig. 11 loc. A, B, C, respectively). At
Fig. 9: Temperature field and reaction rate isoline (black line the combustor dome (A), the thermal paint colors indicate high
envelopes the reaction zone) for the modified combustor temperatures between the different sectors with the coloring
(configuration A). yellow to brown. The combustor liners show low temperatures
close to the dome achieved via the film cooling, but at the end
Based on the learning so far, the re-design option was further of the first segment length, some hot spots are visible (yellow
modified and the holes in the first segments of the combustor color). The film cooling on the second segment works
liner were closed. Figure 10 overlays the hole pattern of the efficiently up to the dilution holes row. On its backside (B), the
liner re-design (configuration A, blue contours) studied so far color change orange to yellow indicates a significant thermal
and the finally tested configuration B (green contours). load on this section. The numerical analysis confirms this by
revealing a strong heat release of the reacting flow close to the
OD at this location. On the outer transition liner (C), the last
hole row leads again to a cooling of the solid parts and, thus, to
a reduced temperature level (orange with some yellow hot
spots).
Figure 12 shows the numerical results of the tested
hardware with respect to liner temperatures. The simulation
results of one sector model covering one burner head was
Fig. 10: Hole pattern of configuration A (blue) and B (green). multiplied to visualize the temperature distribution. For

6 Copyright © 2018 ASME

Downloaded From: [Link] on 09/12/2018 Terms of Use: [Link]

 comparison reason, the temperature field is visualized with the
color range of the thermal paint. With respect to the different
areas of interest discussed for the thermal paint results, one can
see: The temperature distribution on the combustor dome (A)
also shows high temperatures between the well cooled burner
heads, the color range goes from orange around the heads to
yellow and even brown hot spot areas in the edges of the
combustor segments. Around the dilution air holes (B), a very
similar color distribution is achieved by the numerical
simulation. The maximum temperature on the backside is Fig. 13: Radial profiles of turbine inlet temperatures for
caused by heat release close to the walls in the wake of the averaged (RTDF, l.h.s.) and maximum values (OTDF, r.h.s.).
penetrating jets. Finally, the outer transition liner (C) shows
acceptable cooling efficiency by the orange color. EFFICIENCY EVALUATION OF CFD APPROACH
The high accuracy of the numerical predictions is shown in
the comparison of predicted wall temperatures and
experimental data. But the potential of such a numerical setup
is only given for complex combustor designs if the numerical
calculation can be achieved within an acceptable calculation
time. To fulfill this requirement, the massive parallelization is
an option. Since the usage of an open source CFD code without
any license costs allows this step, the question to be answered
is the scalability of the multi-physics solver with increasing
CPU power. Figure 14 shows the result of a study parallelizing
the numerical calculation from 32 CPUs up to 1024 CPUs. The
evaluation reveals that doubled numbers of cores led to a
significant reduction of calculation time by almost 50% up to
1024 cores. Increasing the number of CPU led to a significant
reduction of nodes which had to get handled per CPU. The
Fig. 12: CHT results for comparison to the thermal paint results final simulation with 1024 CPU had in the order of ~15,000
of fig. 11 (OD liners and combustor head). nodes per CPU. For the re-design work presented in this paper,
the massive parallelization allowed the authors to evaluate
Overall, the agreement of the thermal paint results and the design changes in less than a week. Encouraged by these
numerically analyzed temperature distribution is good and results, the code will be further developed and used for
acceptable for the objective of the investigation. Back-to-back numerical studies at MTU.
comparisons of CFD results allow the evaluation of design
changes with respect to combustor liner temperatures expected
at steady state operation. The indirect validation of the
numerical approach via the combustor wall temperatures in
combination with the former validation with test data including
exit temperature profiles (see [4]) allow the assumption that the
combustor exit profiles are reliable as well. For this reason, the
final step to evaluate the re-design option is to compare the
turbine inlet profiles of the base combustor and the modified
one. Figure 13 shows the averaged radial temperature profile
(RTDF) of both configurations as well as the radial maximum
temperature profile (OTDF) caused by local hot spots. The
main finding is the significant reduction of the maximum
temperature around 50% span. Introducing a stronger swirl
level and closing the first dilution holes led to a compact
reaction zone and to a significant improvement of the mixing Fig. 14: Evaluation of speed up over increased CPU power
behavior. Local hot spots and hot streaks on the axis could be shows almost linear scaling.
avoided or at least locally reduced. Further testing will evaluate
the potential of the presented design with respect to durability
and emissions.

7 Copyright © 2018 ASME

Downloaded From: [Link] on 09/12/2018 Terms of Use: [Link]

 CONCLUSIONS [11] Magnussen B.F., “Modeling of NOx and Soot Formation
The paper presented results of combustor simulations with by the Eddy Dissipation Concept”, International Flame
a multi-physics solver based on the open source CFD package Research Foundation First Topic Oriented Technical Meeting,
OpenFOAM. The in-house solver was applied to a commercial Amsterdam, 1989.
engine combustor and used for evaluation of different re-design [12] Kleiveland R.N., “Modeling of soot formation and
options. The successful comparison of thermal paint results oxidation in turbulent diffusion flames”, PhD Thesis,
with predicted combustor wall temperatures revealed the high Norwegian University of Science and Technology, Trondheim,
potential of the modelling approach. Especially the successful 2005.
test of massive parallelization presented in the paper
encourages to further improve the solver. The implementation
of more high-sophisticated models in the future will help to
come to a more quantitative tool. Nevertheless already the
current solver with its possibility to evaluate accurately designs
via back-to-back comparisons allows studies of complex
geometries as commercial combustor designs already short
term.

REFERENCES
[1] Geigle K. P., Hadef R., and Meier W., “Soot formation and
flame characterization of an aero-engine model combustor
burning ethylene at elevated pressure”, Journal of Engineering
for Gas Turbines and Power, Vol.136, 2014, pp. 021505-1-
021505-7.
[2] Freitag S., Meier U., Heinze J., Behrendt T., Hassa, C.,
“Measurement of initial conditions of a kerosene spray from a
generic aero engine injector at elevated pressure”, ILASS –
Europe 2010, 23rd Annual Conference on Liquid Atomization
and Spray Systems.
[3] Meier U., Heinze J., Freitag S., Hassa C. „Spray and flame
structure of a generic injector at aero engines conditions“,
ASME Journal Gas Turb. Power, 134, 2012.
[4] Konle M., de Guillebon L., Cottier F., “Multi-Physics
Simulations of an Aero Engine Combustor with OpenFoam”,
Proceedings of 1st Global Power and Propulsion Forum, GPPF-
2017-0045, Zurich, 2017.
[5] Puggelli S., Bertini D., Mazzei L., Andreini A.,
“Assessment of scale-resolved computational fluid dynamics
methods fort he investigation of lean burn spray flames“,
ASME Journal Gas Turb. Power, 139, 2017.
[6] Menter F.R., “Two-equation eddy-viscosity turbulence
models for engineering applications”, AIAA-Journal, 32(8),
1994, pp.269-289.
[7] Poinsot T., Veynante D., “Theoretical and numerical
combustion”, Edwards, second edition, 2005.
[8] Kärrholm F., “Numerical modeling of diesel spray
injection, turbulence interaction and combustion”, PhD Thesis,
Chalmers University of Technology, Göteborg, 2008.
[9] Viskanta R., Mengüç M.P., “Radiation heat transfer in
combustion systems”, Progress in Energy and Combustion
Science, 13(2):97-160, 1987.
[10] Barlow R.S., Karpetis A. N., Frank J. H., Chen J.-Y.,
“Scalar profiles and NO formation in laminar opposed-flow
partially premixed methane/air flames”, Combustion and
Flame, 127:2102–2118, 2001.

8 Copyright © 2018 ASME

Downloaded From: [Link] on 09/12/2018 Terms of Use: [Link]