Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Abstract
Air worthiness requirements of the aircraft engine emission bring new challenges to the combustor research and design. With
the motivation to design high performance and clean combustor, computational fluid dynamics (CFD) is utilized as the powerful
design approach. In this paper, Reynolds averaged Navier-Stokes (RANS) equations of reactive two-phase flow in an experi-
mental low emission combustor is performed. The numerical approach uses an implicit compressible gas solver together with a
Lagrangian liquid-phase tracking method and the extended coherent flamelet model for turbulence-combustion interaction. The
NOx formation is modeled by the concept of post-processing, which resolves the NOx transport equation with the assumption of
frozen temperature distribution. Both turbulence-combustion interaction model and NOx formation model are firstly evaluated by
the comparison of experimental data published in open literature of a lean direct injection (LDI) combustor. The test rig studied
in this paper is called low emission stirred swirl (LESS) combustor, which is a two-stage model combustor, fueled with liquid
kerosene (RP-3) and designed by Beihang University (BUAA). The main stage of LESS combustor employs the principle of lean
prevaporized and premixed (LPP) concept to reduce pollutant, and the pilot stage depends on a diffusion flame for flame stabili-
zation. Detailed numerical results including species distribution, turbulence performance and burning performance are qualita-
tively and quantitatively evaluated. Numerical prediction of NOx emission shows a good agreement with test data at both idle
condition and full power condition of LESS combustor. Preliminary results of the flame structure are shown in this paper. The
flame stabilization mechanism and NOx reduction effort are also discussed with in-depth analysis.
Keywords: air worthiness; computational fluid dynamics; low emission combustor; lean prevaporized and premixed; NOx reduc-
tion
zation in pilot stage. Based on this test rig, Fu, et al. is still believed that CFD could provide in-depth
firstly conducted experimental study of the ignition analysis where experimental study cannot go at current
performance of this chamber while assisted by some stage.
numerical simulation analysis [4]. Their work shows This paper presents a joint effort between ACAE
that the ignition position along the main stream di- and BUAA to study the combustion characteristics
rection has little influence on the lean ignition and NOx formation mechanism of the LESS com-
boundary of the whole chamber. This encouraging bustor. To achieve better understanding on the in-
result indicates that LESS combustor has pretty good teraction of LPP combustion and pilot flame under
stabilization performance working under the lean the lean burning conditions, Reynolds averaged Na-
burn condition. vier-Stokes (RANS) equations of reactive two-phase
Although some similar LPP concept combustors flow are performed. Results are compared against
have been developed for many years, such as twin experimental data with good agreements. Prelimi-
annular premixed swirler (TAPS) family of low NOx nary results of the flame structure are shown in this
combustor developed by General Electric Aircraft paper, which indicates that LESS combustor has
Engines (GEAE) [6-8], there is still a lack of funda- large potential to reduce NOx emission. The flame
mental understanding of LPP combustion mechanism. stabilization mechanism and NOx reduction efforts
The latest studies focusing on the combustion are also discussed with in-depth analysis.
mechanism of LPP reported is conducted by Dha-
nuka, et al. [9], who use planar laser-induced fluores-
2. Numerical Approach
cence (PLIF) to investigate unsteady flow character-
istics with the motivation to reveal the inside physics
of flame-flame interaction and to provide some sup- Numerical simulation is carried out in ACAE. The
ported data for further computational fluid dynamics Reynolds averaged governing equations are discretized
(CFD). Their work shows that the combustion inter- numerically by a finite volume approach. High solu-
action of the main stage and pilot stage would be the tion differencing scheme which is a hybrid of 1st and
source of combustion instability, and generates con- 2nd order is employed. Kerosene is modeled by the
siderable unsteady turbulent phenomena. However, mixture of 40% of n-C10H22 (decanenormal) and
the influence of the flame-flame interaction in LPP 60% of TMB-C9H12 (tetramethyl benzidine), thus the
scheme on NOx formation is still absent at current averaged molar mass is about 128.8 g/mol. In this
stage. work, Sauter mean diameter (SMD) distribution close
Experimental investigations of the complicated to downstream of the fuel nozzle obtained by fuel
flow physics of the premixed combustion and the droplets mass weighted rate is described instead of
pilot flame are made more difficult due to the highly direct modeling of the primary breakup processes of
transient nature of the processes and their interac- droplets. However, the spray fuel nozzle inlet condi-
tions. Thus the interaction of the pilot flame with tions are also given by the half cone angle (45°), the
premixed combustion is still an outstanding issue injection mass, injection direction, initial temperature
and needs further investigation as it has strong im- and the initial velocity. Atomization processes induced
plication on the combustor performance and NOx by aerodynamics inside the swirler cup are modeled by
reduction. enhanced Taylor analogy breakup (ETAB) breakup
Developments of CFD techniques in the last 30 model [14]. Detailed discussion with the application of
years have enabled more complicated engineering the atomization model could be referred to the work of
cases to be simulated. The numerical approaches to Raju [15] and Zhang, et al. [16-18].
purely turbulent diffusion flame and perfectly pre- Subsonic condition is imposed at the inflow by
mixed flame are rapidly developing in the context of specifying the flow mass rate and temperature. While
the studies of industrial combustion [10]. However, only for the main stage inflow boundary, the mass
there is still a threshold lying on the development road fraction ratio of fuel is also added with the assumption
of numerical simulation of LPP combustor. Because of of perfectly premixing. The pressure information is
premixed flame, partially premixed flame and non- given on the outflow boundary. The entire wall in the
premixed flame coexisting simultaneously in one combustion chamber is treated as impermeable and
unique chamber, a unified flame model which could no-slip. The droplets and wall interaction are neglected
handle these entire different combustion regimes in this paper due to lack of trustable model [19].
seems much more necessary [11].
Recent progress developed by Veynante and Legier, 2.1. Turbulence-combustion interaction model
et al. [12-13] shows that their extended flame model with
the hypothesis that turbulence brings effects of thicken The extended coherent flame model (ECFM) [20]
flame surface, could treat both partially premixed and used in this paper for turbulence and combustion in-
purely premixed flames with some reasonable results. teraction is a hybrid model which combines classic
Its application to turbulent spray combustion for air- flame surface density model (FSDM) with laminar
craft combustors has seen some successes [13]. Thus, it flamelet model. The coupling of ECFM with the
· 856 · ZHANG Man et al. / Chinese Journal of Aeronautics 25(2012) 854-863 No.6
flamelet model for effects of non-premixedness ex- duction and consumption of NO and the related spe-
tends the range of applicability of the model to deal cies.
with different combustion regimes inside the gas tur- As thermal NOx formation plays a dominate role in
bine combustor. the whole different mechanisms, only thermal NOx is
Laminar burning velocity of kerosene which is considered in this paper. The mechanism of thermal
modeled by SL in ECFM is obtained from experi- NOx formation is based on the extended Zel’dovich
mental test data given by Eberius, et al. [21]. However, mechanism, which consists of the following equa-
because the original test data only concentrate within tions as described by Bowman [25] and Hanson and
the range of equivalence ratio of 0.9 to 1.4, SL in the Salimian [26]:
entire flammability range and under any operation
conditions is extended by a fifth order polynomial ⎧ N 2 + O ⇔ NO + N
correlation and an experiential correlation proposed by ⎪
⎨ N + O 2 ⇔ NO + O (5)
Metghalchi and Keck [22], which can be written as ⎪
⎩ N + OH ⇔ NO + H
α β
⎛T ⎞ ⎛ p ⎞
S L = S L,ref ⎜ ⎟ ⎜ ⎟ (1)
⎝ Tref ⎠ ⎝ pref ⎠ 3. Validation
⎛ u′ l ⎞ ε
Kt = Γ K ⎜ , t ⎟ (4)
⎝ SL δ L ⎠ k Fig. 1 Geometry of LDI swirler.
where Kt is the turbulent time scale, u ′ the mean tur-
bulent velocity fluctuation, lt the integral length scale
of turbulence, k the turbulent kinetic energy, ε the tur-
bulent dissipation rate, and Γ K an efficiency function
taking into account the ability of all vortices to wrinkle
the flame.
before the calculation of LESS test rig (see Fig.6 and Table 1 Predicted NOx compared with experimental
Table 1). Comparisons of resolution of different grids data for grid independence study
show that all three mesh size cases have satisfactory 0.5 1.0 2.5 Experimental
Grid size
consistency with the experimental data. Considering million million million data
the accuracy and time cost of both results, it is be- NOx/10−6 300 319 320 325
lieved that the gird size of 1.0 million is accurate CPU number 24 48 120 N/A
enough for RANS, thus this mesh size has been taken Total CPU
6 10 16 N/A
as the grid for all subsequent calculations. time/h
of the chamber; Fig. 8, streamline distribution while crease of pilot stage fuel/air ratio, which can be proved
Fig. 9, the shear strain rate distribution. It should be by temperature contour in Fig. 7. Large area of high
noted that each figure contains two parts separated by temperature is shown in the center of the chamber,
the center dotted horizontal line of the chamber. The where the recirculation vortex core coincides with it.
upper pattern in each figure is for Case 2 and the lower Due to the nature of diffusion combustion of the pilot
one is the description of Case 5. stage, the maximum temperature value is achieved
along the line of equivalence ratio of 1.0.
strain rate distribution. layer between the main stream and the pilot stream,
and the other is the main stream mixed with the corner
vortex. Furthermore, Figs. 11-12 show that the pilot
stage fuel reacts with air along the shear layer as the
reason for scalar transport within it.
Figure 11 shows that the area of equivalence ratio Fig. 13 Shear strain rate distribution under the conditions
between the values of 1.0 to 0.72 has been extended in of Case 3 and Case 4.
both Case 3 and Case 4 than the condition of only pilot
flame on (refer to Case 2 in Fig. 7). This phenomenon
must be influenced by the main stage. Meanwhile, 4.6. NOx formation mechanism
within this region, the temperature gradient dissipates
along the streamline, which is also the exclusive fea- Figures 14-17 present further information about the
ture belonging to the diffusion flame. However, the reacted species distribution inside LESS chamber to
exact location of the interaction of the main flame and study the NOx formation mechanism. Figure 14 is the
pilot flame is difficult to determine in the current CFD O radical mass fraction distribution contour, Fig. 15
technique and Dhanuka’s work because of the lack of the thermal NO reaction rate distribution contour, Fig.
useful approach to separate the two kinds of combus- 16 the resident time of the fluid motion, and Fig. 17
tion. The only judgment can be made is that inside the the NO mass fraction contour.
round shape adjacent to the swirl cup (equivalence
ratio is 1.0), pilot flame holds the dominate position,
and within the region of equivalence between the value
of 0.72 and 0.68, pilot flame and main stage flame
interaction take place.
value of 0.68 due to main decomposition process of O2 vortex core. It should be noted that both these regions
accompanied by heat absorption. Thus, the NO gen- are based on the zone of Ux equal to 0.
eration process is followed by the accumulation of O
radical, where high intensity of thermal NO reaction
rate taking place coincides with the maximum amount
of O mass fraction.
As we know, the net generation amount of NO is not
only controlled by the chemical reaction rate, but also
by the resident time in a flowing fluid. However, at
current stage, the resident time scale of a moving fluid
is hard to measure, especially in the case with complex
geometry. In order to measure the resident time scale at
any location inside the chamber, a novel approach has
been developed in this work, as described below. Fig. 17 NO mass fraction distribution in Case 3 and Case 4.
In CFD work, the Courant number (or called CFL
number) is defined as 4.7. Analysis of reaction principle of LESS com-
U x Δt bustion
CCFL = (6)
Δx
In terms of the above analysis including the pilot
where Ux is the local velocity, ∆x the local girds space, stage burning characteristics and pilot-main stage in-
∆t the time step defined artificially, and CCFL the Cou- teraction feature, the mechanism of the pilot and main
rant number which could be easily obtained from any stage interaction could be summarized (see Fig. 18).
CFD codes, so the resident time τ could be defined by Figure 19 gives the NOx formation mechanism in these
Δx Δt fluid structures.
τ= = (7)
U x CCFL
According to the definition of resident time in this
work, the entire fluid resident time scale based on each
calculated node can be illustrated in Fig. 16.
quantitative measurement of level set of the interaction lean combustion low emission combustor scheme.
between pilot flame and the main stage in this work. Journal of Aerospace Power 2003; 18(4): 492-497. [in
As the NOx formation mechanism illustrated in Fig. Chinese]
[3] Lin Y Z, Lin Y, Zhang C, et al. Discussion on combus-
19, the overlapping region of the higher value of tem-
tion airflow distribution of advanced staged combustor.
perature and the resident time will be the place where Journal of Aerospace Power 2010; 25(9): 1923-1930.
NO mass fraction achieves its maximum. The qualita- [in Chinese]
tive analysis of this work indicates that the maximum [4] Fu Z B, Lin Y Z, Li J B. Experimental investigation on
value of NOx formation appears in the region adjacent ignition performance of LESS combustor. Proceedings
to the zero value axial velocity iso-surface. From this of ASME Turbo Expo 2011, 2011; 1-8.
point of view, with the motivation to design the low [5] Owen B, Lee D S, Lim L. Flying into the future: avia-
NOx emission combustor, the basic design principle tion emissions scenarios to 2050. Environment Science
lies in the organization of this large-scale recirculation and Technology 2010; 44(7): 2255-2260.
[6] Mongia H C. TAPS-A 4th generation propulsion com-
fluid motion. bustor technology for low emissions. AIAA-2003-2657,
2003.
5. Conclusions [7] Mongia H C. Recent progress in comprehensive mod-
eling of gas turbine combustion. AIAA-2008-1445,
Some findings of this work could be summarized as 2008.
follows: [8] Dhanuka S K, Driscoll J F, Mongia H C. Instantaneous
1) NOx predicted by CFD under the condition of en- flow structures in a reacting gas turbine combustor.
AIAA-2008-4683, 2008.
gine takeoff indicates that LESS combustor has large
[9] Dhanuka S K, Temme J E, Driscoll J F, et al. Unsteady
potential to reduce NOx emission. aspects of lean premixed-prevaporized (LPP) gas tur-
2) The flow pattern under the conditions of pilot bine combustor: flame-flame interactions. AIAA-2010-
stage and main stage interaction has been clearly re- 1148, 2010.
solved by the current CFD approach, which is verified [10] Westbrook C K, Mizobuchi Y, Poinsot T J, et al. Com-
by PLIF results in open literature. putational combustion. Proceedings of the Combustion
3) The principles of the fluid motion and NOx for- Institute 2005; 30(1): 125-157.
mation of LESS combustor have been illustrated in [11] Bilger R W, Pope S B, Bray K N C, et al. Paradigms in
turbulent combustion research. Proceedings of the
this work. Reaction zones could be divided into three
Combustion Institute 2005; 30(1): 21-42.
parts inside of the LESS chamber under the condition [12] Veynante D, Vervisch L. Turbulent combustion model-
of the pilot stage and main stage co-working. ing. Progress in Energy and Combustion Science 2002;
4) The overlapping region of the higher value of 28(3): 193-266.
temperature and the resident time is the place where [13] Legier J P, Poinsot T, Veynante D. Dynamically thick-
NO mass fraction achieves its maximum. The qualita- ened flame LES model for premixed and non-premixed
tive analysis of this work indicates that the maximum turbulent combustion. Proceedings of the 2000 Study-
value of NOx formation appears in the region adjacent ing Turbulence Using Numerical Simulation Databases,
to the zero value axial velocity iso-surface. 2000; 157-168.
[14] Tanner F X. Liquid jet atomization and droplet breakup
5) With the motivation to design the low NOx emis-
modeling of non-evaporating diesel fuel sprays. SAE
sion combustor, the basic design principle lies in the Transactions 1997; 106(3): 127-140.
organization of this large-scale recirculation fluid mo- [15] Raju M S. Numerical investigation of various atomiza-
tion. tion models in the modeling of a spray flame.
6) It is still hard to give a quantitative measurement NASA/CR-2005-214033, 2005.
of level set of the interaction between pilot flame and [16] Zhang M. Research on turbulent spray combustion and
the main stage in this work, which is critical for the flame stabilizing mechanism in RBCC chamber. PhD
design of LPP low emission chamber. Further efforts thesis, Northwestern Polytechnical University, 2010.
[in Chinese]
will be made in the coming research.
[17] Zhang M, Hu Z W, He G Q, et al. Large-eddy simula-
tion of kerosene spray combustion in a model scramjet
Acknowledgements chamber. Proceedings of the Institution of Mechanical
Engineers, Part G: Journal of Aerospace Engineering
The author appreciates the technical guidance from 2010; 224(9): 949-960.
senior engineer, Mr. HU Zhengyi of BUAA. The dis- [18] Zhang M, Hu Z W, He G Q, et al. LES of kerosene
spray combustion with pilot flame in a model dual
cussion with Dr. WANG Xiaofeng and LI Lin of
mode ramjet chamber. AIAA-2009-5385, 2009.
BUAA is also greatly acknowledged. [19] Gepperth S, Guildenbecher D, Koch R, et al. Pre-
filming primary atomization: experiments and model-
References ing. 23rd Annual Conference on Liquid Atomization
and Spray Systems, 2010; 1-9.
[1] http://www.acae.com.cn/. [20] ANSYS CFX-solver theory guide. ANSYS Inc., 2010.
[2] Lin Y Z, Peng Y H, Liu G E. A preliminary study of [21] Eberius H, Frank P, Kick T, et al. EU project computa-
NOx emission of staging/premixed and prevaporized tional fluid dynamics for combustion. No. GRD1-
No.6 ZHANG Man et al. / Chinese Journal of Aeronautics 25(2012) 854-863 · 863 ·