Sei sulla pagina 1di 12

Comprehensive Gas Turbine Combustion

Modeling Methodology


Hukam Mongia
GE Aviation, Cincinnati, OH, U.S.A.
Sundar Krishnaswami and PSVS Sreedhar
EACoE, GE Aviation, Bangalore, India






ABSTRACT: A comprehensive gas turbine combustion modeling
methodology must account for complex geometry, physics, and chemistry.
The methodology adopted here exploits the continuously evolving suite of
combustion models of FLUENT

, GE Aviations User-Defined Functions


(UDFs), and High-Performance Computing (HPC). As such, it provides a
framework to readily incorporate advances in simulation technology. Here we
document the systematic validation of submodels for turbulence, fuel injection,
droplet formation, droplet breakup, and turbulence-chemistry interactions.


Keywords: Gas Turbine Combustion, Modeling, Design Methodology.


INTRODUCTION

In the past, CFD has been mainly used as a qualitative guiding tool rather
than for the actual design of gas turbine combustors. This is primarily because
of the inadequacy of the available analytical and numerical models as well as
limited computing resources to predict the different performance
characteristics of the combustor. Now, with advances in modeling capabilities
aided by several fold increase in computing power, it is possible to work
towards realizing the goal of a CFD-based analytical combustor design
approach. Traditionally, a typical combustor design cycle involves conceptual,
preliminary and a detailed design complimented by several sub-scale
component tests along with a couple of full-scale annular rig tests and one or
more full engine test(s).
The goal set for a CFD-based design cycle is to reduce dependence on
the iterative analyze-build-test-refine design process with attendant
reduction in the cycle time of conceptual, preliminary and detailed design
stages and the number of rig tests. The readiness of CFD to be used for
design is assessed by its prediction capability of certain critical performance
characteristics like emissions, exit temperature profiles, wall temperatures,
International Aerospace CFD Conference
(Paris, June 18 19 2007)

lean blowout (LBO) fuel-air ratio etc, within certain specifications as stated in
Mongia, (2002).
For reliable aircraft gas turbine combustion calculations using CFD, the
following four steps need to taken:

1. Systematic assessment of the available turbulence models
against standard benchmark cases. This exercise is required to
establish the prediction capability of the solver and for appropriate
selection of the turbulence model.
2. Accurate prediction of the cold flow field in and around the
combustor (air flow through multiple inlet ports of the combustion
chamber, pressure, velocity & turbulence distribution), as
impacted by the compressor exit profiles, the diffusion system
and the combustion system details.
3. State-of-the-art modeling of the fuel atomization process,
resulting in an accurate estimation of the spray quality.
4. Finally an appropriate choice of the chemistry model, a turbulence
chemistry interaction model and rigorously established process
for exercising them, in order to predict the critical characteristics
of chemically reacting flow within the combustor.

CALCULATION PROCESS

In all the investigations carried out, the commercial software FLUENT

has
been used. Details of the four steps outlined in the introduction are described
below.

TURBULENCE MODELS ASSESSMENT

It is well known that the choice of turbulence model is dependent upon the
class of problems to be investigated. For typical combustor application, the
class of problems of interest can be grouped among the un-separated flows
like, the flat plate, channel, pipe and annulus and separated flows like
backward facing step, sudden expansion (with and without swirl). In addition,
several cooling arrangements in combustor require accurate prediction of the
heat transfer characteristics, and these arrangements can be simplified as
flow problems involving wall jets and impinging (single and multiple) jets.
Singh and Mongia (2007), have described in detail, in their paper, the
benchmark cases investigated, the approach followed and the comparisons
made. Specifically, their effort focused on four different turbulence models
Standard k- (SKE) (Launder and Spalding, 1974), Realizable k- (RKE) (Shih
et.al., 1995), Reynolds Stress Model (RSM) (Launder et.al., 1975) and Shear
Stress Transport (SST) model (Menter, 1994), and two different near wall
treatment, namely, standard wall function (swf) (Launder and Spalding, 1974)
and enhanced wall treatment (ewt) (Chen and Patel, 1988). Based on their
investigations, they have recommended the use of RKE-ewt model for all

combustor applications. All further investigations reported in this paper will be
using the aforementioned model, unless otherwise stated.

COMBUSTION SYSTEM ANALYSIS (CSA)

For reliable aircraft gas turbine combustion calculations, the foremost
requirement is the cold flow field in and around the combustor (air flow
through multiple inlet ports of the combustion chamber, pressure, velocity &
turbulence distribution), supplied by the diffusion system. A typical combustor-
diffusion system consists of an annular pre-diffuser, a dump section, fuel
nozzles, casing, cowlings, dome, liners, annular passages and support
structure. The objectives of a combustor-diffuser system are to supply the
combustion & cooling air to the combustor in prescribed amounts through
various inlets of the combustor, ensure optimum static-pressure recovery (Cp)
& total-pressure losses (), ensure uniformity of flow entering the annular
passages around the combustor.
Combustion System Analysis (CSA) entails the CFD based prediction of
the airflow distribution and pressure drop within the combustor-diffuser
system. During the years 2001 through 2004, considerable attention has
been devoted to the key aspects that would ensure the success of the CSA
approach. In addition to the appropriate choice of the turbulence models,
these include the investigation of the inlet conditions and the importance of
the grid quality. The summary of the earlier investigation has been published
in a series of six papers (Mongia et. al., 2004a-f). The approach has since
then been further improved and applied to 11 different combustors and the
results are summarized below.
Fig. 1 shows how well the approach works even for the combustor cooling
ports which inject a very small fraction of combustor air through many discrete
holes. Predicted airflow distribution through each axial row of the cooling
holes and nuggets against measured data for 11 different combustors are
presented in this figure. It can be seen that the agreement with data is
excellent, the prediction capability is within 0.16%, with a sigma of around 1%.
Fig. 2 shows the comparison of coefficient of pressure (defined as static
pressure recovery normalized by the inlet dynamic head) in the inner and
outer passages for 11 different annular combustors. The prediction of Cp is
within 5%. Clearly, from the results, one can infer that the CFD prediction with
respect to pressure drop and air flow distribution is as good as the
experimental results.


Fig. 1: Comparison of cooling air flow distribution for 11 different combustors

Fig. 2: Coefficient of Pressure for different combustors.

Cp at Outer and Inner passage inlets, = 0.09
y = 1.046x
R
2
= 0.6031
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Measur ed
P
r
e
d
i
c
t
e
d
Combustor1
Combustor2
Combustor3
Combustor4
Combustor5
Combustor6
Combustor7
Combustor8
Combustor9
Combustor10
Combustor11
Ideal Fit
Outer Passage: Red, Inner Passage: Magenta
Air Flow Distribution through all the holes, = 1.0%
y = 1.0016x
R
2
= 0.994
0%
10%
20%
30%
40%
50%
60%
70%
0% 10% 20% 30% 40% 50% 60% 70%
Measured
P
r
e
d
i
c
t
e
d
Combustor1
Combustor2
Combustor3
Combustor4
Combustor5
Combustor6
Combustor7
Combustor8
Combustor9
Combustor10
Combustor11
Ideal Fit

COMPREHENSIVE INJECTOR MODEL (CIM)

Having satisfactorily addressed the airflow and pressure drop, the next major
challenge is to establish and assess a robust fuel atomization and spray
distribution model. Development of a predictive model for fuel atomization
and subsequent spray distribution requires a comprehensive approach for
addressing various sub processes like the proper development of the flow (in
a pressure swirl atomizer), the formation of a sheet and its breakup using a
model for sheet instability, followed by the influence of the aerodynamic forces
to form droplets from the broken sheets, and the subsequent distribution of
drops.
There are two widely used approaches for numerically predicting the sheet
formation, and they are the Arbitrary Lagrangian-Eulerian (ALE) approach
advocated by J og et. al., (2000), and the Volume of Fluid (VOF) approach
developed by Hirt et. al. (1981). In the current investigation, a two phase VOF
technique available in FLUENT

is used to simulate the flow through a fuel


nozzle. The primary focus of this analysis is on the flow characteristics of the
atomizer, the formation of the liquid film as it exits from the orifice, and the
resulting cone angle. The turbulence model exercised is the RSM model
(Launder et. al., 1975) with the coupled solver and PRESTO (PREssure
STaggering Option, Patankar, 1980) scheme for pressure interpolation and
modified HRIC (Muzaferija, S. et. al., 1988) for volume fraction equation.
Investigations were carried out for the range of flow numbers normally
experienced by fuel atomizers in modern aviation engines. Fig. 3, below,
illustrates the air-core obtained for two different flow numbers through the
atomizer. One can observe that the liquid-air interface is sharply captured.
The region in red is representative of the fuel (J et-A) and the region in blue
is representative of air.











Fig. 3: Air-core and liquid film for two different Flow Numbers.

Using the liquid sheet and flow field parameters obtained from VOF as
inputs, a linear stability analysis has been carried out to predict the primary
droplet diameter and the breakup length. This involves conducting a temporal
linear stability analysis to study the primary atomization of annular liquid
sheets. The approach captures the instability of a swirling annular liquid
sheet subject to axi-symmetric disturbances and swirling inner and outer air
FN=2.7
FN=10.6

streams and is based on the paper by Ibrahim and J og, 2006. The VOF
simulation and linear stability analysis provide a comprehensive approach to
modeling atomization from pressure-swirl and airblast atomizers.
Having obtained the mean diameter through the stability analysis, the
droplet distribution is modeled using a two parameter Rosin Rammler
distribution, wherein the most probable droplet size is obtained from
processing the results of stability analysis, and the spread parameter is
assumed to be 2.5. Key physical properties of the liquid droplet, such as
surface tension, vapor pressure and thermal conductivity are obtained from
piecewise linear fits. Conical injection is assumed, and the parameters
relevant for setting up the injection are obtained from VOF.
The secondary breakup model used in the current investigation is the
Taylor Analogy Breakup (TAB) model (Taylor, G.I, 1963). This model was
chosen because the Weber number experienced for the range of conditions
investigated was less than 100. Coupling of the discrete phase with the
continuous phase was invoked through the discrete phase model available in
FLUENT

, which included accounting for the effect of instantaneous turbulent


velocity fluctuations on particle trajectories through the stochastic discrete
random walk model and invoking two-way turbulence coupling to account for
the effect of particles on the turbulence quantities in the continuous phase.
Using the above approach, the CIM model was validated against
atmospheric rig test data, and the comparisons obtained for the secondary
droplet size are shown in Fig. 4. The results show very good agreement with
data, thereby validating the process for comprehensive injector modeling.


















Fig. 4: SMD comparison for atmospheric Rig Test.



Secondary SMD
0
50
100
150
200
0 50 100 150 200
Rig data (microns)
P
r
e
d
i
c
t
e
d

(
m
i
c
r
o
n
s
)

REACTING FLOW MODEL

With an established combustion system analysis procedure and a
comprehensive injector modeling approach, the final step to successful
prediction of key characteristics of a combustor is a robust and reliable
reacting flow model. In most of the lean dome combustors, the fuel stream
undergoes some degree of partial premixing before it is subject to chemical
reaction. Therefore, the partially premixed combustion approach is most
appropriate for modeling the combustion process. Because of the requirement
of quick turnaround in calculations, the commonly used combustion models
are relatively simple, like for example, the assumed-shape PDF (fast
chemistry) and eddy-breakup. However, these models have limitations,
particularly when calculating the emissions from lean dome combustors.
In contrast, the laminar flamelet approach (Peters, 1986) offers significant
advantages, without adding additional calculation speed penalties. Briefly, a
flamelet model assumes that the effect of turbulence on a flame can, under
certain conditions, be modeled as a flame area increase. The main criterion
for making this assumption is that the flame length scale is much smaller than
the Kolmogorov turbulence length scale. For the combustor modeled here,
that assumption is not strictly correct, and the true combustion regime is
intermediate between flamelet combustion and distributed reaction. As an
engineering model, however, the mean properties of the flow field can often
be well-reproduced by a flamelet model even when it is not strictly valid.
In employing a single flamelet model for investigation, the key challenge is
a-priori estimate of the appropriate stoichiometric scalar dissipation. In the
current investigation, an iterative approach recommended by Shui-Chi Li
(2003), as described below, is adopted to determine the scalar dissipation.
To start with, a nominal value of scalar dissipation is assumed and the
corresponding flamelet and the PDF table is generated. The reacting flow is
run to convergence. From the converged CFD solution, the stoichiometric
scalar dissipation is estimated from a weighted average of the local scalar
dissipation on iso-mean mixture-fraction surfaces that envelope the thin
flame zone. A single flamelet is re-generated and the CFD solution is run
again, until the scalar dissipation used in flamelet generation and the CFD
based computed stoichiometric scalar dissipation are within 0.5% of one
another.
For the range of pressures, temperatures and fuel air ratios for dome
combustors investigated, a maximum of 3 iterations between the CFD solution
and flamelet generation was required.


Fig. 5: Comparison of NOx Emissions.

Fig. 5, above, shows the comparison of prediction of NOx emissions
against data for two different lean dome combustors, and for a range of
pressures from 29 psi to 200 psi, and temperatures in the range from 330F
710F. It can be seen clearly that the predictions agree with data within 1EI.
Shown below in Fig. 6 is the comparison of prediction of CO emissions.
Again, one can observe that the agreement is very good and is within 5EI.
Deviations at the highest values of CO EI can be explained due to the lack of
a good chemistry model to predict unburned hydrocarbons and inadequacy of
the flamelet approach for thick flame brushes, and it points towards adopting
the approach to solve transport equations for CO and unburned
hydrocarbons.
EINOx
y = 1.0038x
R
2
= 0.9204
3
5
7
9
11
13
15
3 5 7 9 11 13 15
Data
C
F
D
EINOx
Ideal Fi t
Ideal +1EI


Fig. 6: Comparison of CO emissions.

Fig. 7, below shows the comparison of exit temperature quality. One can
observe very good agreement between data and CFD.
Fig. 7: Comparison of exit temperature quality.
EICO
y = 0.9999x
R
2
= 0.9602
0
10
20
30
40
50
60
70
80
0 10 20 30 40 50 60 70 80
Data
C
F
D
EICO
Ideal Fi t
Ideal +5EI
RDG825-Pr of il e Factor
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
NDT AVG
H
e
i
g
h
t

F
r
a
c
t
i
o
n
RDG825-DATA
RDG825-CFD
RDG825-Pr of il e Factor
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
NDT AVG
H
e
i
g
h
t

F
r
a
c
t
i
o
n
RDG825-DATA
RDG825-CFD

SUMMARY

A systematic investigation for the prediction of the key characteristics of
combustors using CFD has been carried out. This entailed a rigorous
assessment of the turbulence models for combustor applications, establishing
and validating a combustion system analysis methodology for 11 different
combustors, a comprehensive spray injection modeling approach, and a
systematic way of exercising a single flamelet approach for reacting flow.
Good agreement with data, for the key performance parameters has been
demonstrated.
The authors would like to acknowledge S C Li, GE Aviation, Springdale for
sharing his approach of CFD-based estimation of scalar dissipation, and the
team in GE Aviation, Bangalore, for their efforts in the validation process.

REFERENCES

Chen, H. C., and Patel, V. C. (1988) Near wall turbulence models for complex
flows including separation, AIAA J ournal, 26, 641-648.

Ghanshyam Singh and Mongia, H. C. (2007) Comparison of turbulence
models for gas turbine applications, Workshop on Diagnostics and control of
Pollutant Formation and Emission in Combustion, J adavpur University, India.

Hirt, C. W., and Nichols, B. D., (1981) Volume of Fluid (VOF) method for the
dynamics of free boundaries, J ournal of computational physics, 39, 201-225.

Ibrahim, A. A., and J og, M. A. (2006), Effect of liquid and air swirl strength and
relative rotational direction on the instability of an annular liquid sheet, Acta
Mechanica, 186, 113-133.

J og, M. A., Sakman, A. T., J eng, S. M., and Benjamin, M. A. (2000),
Parametric study of simplex fuel nozzle internal flow and performance, AIAA
J ournal, 38, 1214-1218.

J ones, W. P., and Whitelaw, J . H. (1982) Calculation methods for reacting
turbulent flows A review, Combustion and Flame 48, 1-48.

Launder, B. E., and Spalding, D. B (1974) The numerical computation of
turbulent flows, Computer Methods in Applied Mechanics and Engineering, 3,
269-289.

Launder, B. E., Reece, G. J ., and Rodi, W. (1975) Progress in the
development of a Reynolds stress turbulence closure, J ournal of Fluid
Mechanics, 68, 537-566.

Menter, F. R., (1994) Two equation eddy viscosity turbulence models for
engineering applications, AIAA J ournal, 32, 1598-1605.


Muzaferija, S., Peric, M., Sames, P., and Schellin, T., (1998), A two fluid
Navier Stokes solver to simulate water entry, Proceedings of 22
nd
Symposium
on Naval Hydrodynamics, 277-289.

Mongia, H. C., (2002) Development of combustion chamber design tools, TRF
workshop, IIT Madras.

Mongia, H.C., Hsiao, G., Burrus, D., Sreedhar, PSVS, Rao, A. and Naik, P,
(2004a), Combustor Diffuser Modeling Part I: Inlet Profiles & 2-D Calculations,
AIAA Paper 2004-4168. 40th AIAA/ASME/SAE/ASEE J oint Propulsion
Conference and Exhibit, Fort Lauderdale, Florida, J uly 11-14.

Mongia, H.C., Hsiao, G., Burrus, D., Sreedhar, PSVS, Rao, A. and Ismail, S.
(2004b), Combustor Diffuser Modeling Part II: Inlet Profiles and 3-D
Calculations, AIAA Paper 2004-4169. 40th AIAA/ASME/SAE/ASEE J oint
Propulsion Conference and Exhibit, Fort Lauderdale, Florida, J uly 11-14.

Mongia, H.C., Hsiao, G., Burrus, D., Sreedhar, PSVS, (2004c), Combustor
Diffuser Modeling Part III: Validation w/ Typical Separating Single Passage
Diffusers Combustor Diffuser Modeling, AIAA Paper 2004-4170. 40th
AIAA/ASME/SAE/ASEE J oint Propulsion Conference and Exhibit, Fort
Lauderdale, Florida, J uly 11-14.

Mongia, H.C., Hsiao, G and Burrus, D., Sreedhar, PSVS, Rao A and Ismail,
S, (2004d), Combustor Diffuser Modeling Part IV: Effect of Cowling Geometry,
Mixer Size and Nozzle Blockage, AIAA-2004-4171, 40th
AIAA/ASME/SAE/ASEE J oint Propulsion Conference and Exhibit, Fort
Lauderdale, Florida, J uly 11-14.

Mongia, H.C., Hsiao, G., Sreedhar, PSVS, Rao, A. and Ismail, S. (2004e),
Combustor Diffuser Modeling Part V: Validation with a Three Passage
Diffuser Rig Data, AIAA Paper 2004-4172. 40th AIAA/ASME/SAE/ASEE J oint
Propulsion Conference and Exhibit, Fort Lauderdale, Florida, J uly 11-14.

Mongia, H.C., Mueller, M., Dai, Z., Sreedhar, PSVS, Rao, A. and Ismail, S.
(2004f), Combustor Diffuser Modeling Part VI: Validation with a Four Passage
Diffuser Rig Data, AIAA Paper 2004-4173. 40th AIAA/ASME/SAE/ASEE J oint
Propulsion Conference and Exhibit, Fort Lauderdale, Florida, J uly 11-14.

Patankar, S. V., 1980, Numerical Heat Transfer, Hemisphere Publication.

Peters, N (1986), Laminar flamelet concepts in turbulent combustion, 21
st

Symposium (International) of Combustion, 1231-1250.



Peters, N., Riesmeier, E. and Honnet, S.. (2004), Flamelet modeling of
pollutant formation in a gas turbine combustion chamber using detailed
chemistry for a kerosene model fuel, J ournal of engineering for gas turbines
and power, 126, 899-905.

Shih, T. H., Liou, W. W., Shabbir, A., Yang, Z., and Zhu, J . (1995) A new k-e
eddy viscosity model for high Reynolds number turbulent flows Model
development and validation, Computers Fluids 24, 227-238.

Taylor, G. I., (1963), The shape and acceleration of a drop in a high speed air
stream, Technical Report in the scientific papers of G I Taylor ed. G.K.,
Batchelor.

Zimont, V. L., (2000) Gas premixed combustion at high turbulence. Turbulent
flame closure model combustion model, Experimental thermal and fluid
science, 21, 179-186.

Potrebbero piacerti anche