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

net/publication/37180149

Cold flow analysis of an aero-engine gas turbine combustor configuration

Article · January 2001

Source: OAI

CITATIONS READS
4 655

6 authors, including:

Muralidhara Hs
National Aerospace Laboratories
5 PUBLICATIONS 6 CITATIONS

SEE PROFILE

All content following this page was uploaded by Muralidhara Hs on 05 December 2019.

The user has requested enhancement of the downloaded file.

ISOABE, Sept. 2-7, 2001, Bangalore, India.

Cold Flow Analysis of an Aero-Engine Gas Turbine

Combustor Configuration
H.S.Muralidhara+, T.R.Shembharkar+, B.R.Pai+
P.Ramanujam*. G.N.Jayaprakash*, P.N.Pandey*

+ Scientists, Propulsion Division, National Aerospace Laboratories, Bangalore.

* Scientists, Combustion Group, Gas Turbine Research Establishment, Bangalore.

Abstract
The present work is related to the computation of cold flow in a typical aero-engine gas
turbine annular combustor and validation of the predictions with the experimental results. The
primary motivation of the study was to simulate the complex 3-D turbulent flow in the complete
combustor system, starting from the compressor exit to the turbine inlet, in a coupled manner. A
general-purpose CFD code PHOENICS was employed for this purpose. Structured BFC grid
was generated for the combustor configuration. Three-dimensional Reynolds averaged Navier-
Stokes equations were solved with the standard k- turbulence model. Computations were
carried out for three flow conditions for the combustor and the predictions were compared with
experimental results like mass flow distribution and total pressure loss. Around 2.5% deviation
from experimental values was observed for mass flow splits along outer annulus, core and inner
annulus. Deviation in total pressure loss was found to be less that 1%.

Nomenclature
f Friction factor
k Specific turbulent kinetic energy
lm Characteristic length
m Mass flow rate
S Sources/sinks of 
Sf Source term due to friction
T Temperature (k)
V Velocity vector (m/s)
Vaxial Axial component of velocity (m/s)
 Conserved property
 Density (Kg/m3)
 Exchange coefficient for 
 Shear stress
 Vane angle
 Dissipation rate of turbulent kinetic energy
 Prandtl number

1.0 Introduction

1
ISOABE, Sept. 2-7, 2001, Bangalore, India.

The flow in a modern gas turbine combustor like the one, which is under consideration, is
indeed very complex. Computational Fluid Dynamics (CFD) can play an important role in the
design and development of such complex equipment. However, the complex flow in a complex
geometry does not render itself easily and precisely to the CFD simulation in straightforward
manner. In many situations, intelligent approximations regarding geometrical details and flow
models have to be made to carry out a sensible CFD simulation. These aspects of flow
simulation should be validated before they are confidently used for analysing new flow
situations.
CFD simulations are frequently included in the design process. Some of the earlier
simulations for combustor flow have been surveyed by Sturgess [1], Correa and Shyy [2], and
Rizk and Mongia [3]. Most of the simulations include only flow inside the combustor liner
either with assumed profiles and flow splits at the various liner inlets or these inlet conditions are
separately calculated and prescribed. One-dimensional annulus codes have traditionally been
utilised to provide a reasonable prediction of flow splits, but they cannot capture important two
and three-dimensional flow features that can significantly affect total pressure losses. Non-
uniform profiles of flow conditions, including velocities, jet angles and turbulent properties, at
inlet boundary conditions can also have a significant influence on the flow field in a combustor.
Turbulence properties at inlets are particularly difficult to estimate. McGuirk and Spencer [4]
discussed the sensitivity of discharge coefficient and jet trajectory of dilution holes to small
changes in the dilution hole geometry. They emphasised the need to provide realistic profile at
dilution hole inlet. Lawson [5] used a one-dimensional code to predict flow splits and two-
dimensional CFD model to predict the flow profile at the exit plane of the swirl cup. This profile
was then applied as a boundary condition in the three-dimensional model. Martino et. al. [6] also
used a one-dimensional code to determine air flow distribution through the rows of holes across
the liner as well as pressure drop and injection angles. Fuller and Smith [7] were able to predict
exit temperature profiles of an annular through-flow combustor that were fairly in good
agreement with measurements. They modelled the air swirler and combustor flow field in a
decoupled manner and used a two-dimensional model to provide boundary conditions at the exit
plane of swirl nozzle. They also demonstrated the sensitivity of the overall solution to the details
of swirler flow. Crocker et. al. [8] discussed the importance of detailed prescription of fuel
nozzle/dome swirler boundary conditions and described an approach for accurate prediction of
flow splits and effective area. Shembharkar et. al. [9] in a study of a reverse flow combustor,
computed the flow in two parts separately; the cold flow in the outer annulus and the reactive
flow inside the liner. These two solutions were then matched iteratively at the common
boundaries namely the various holes. Lai [10] included the swirler passages in his model, which
is an important step in reducing the uncertainty of prescribed boundary conditions.
In a recent study, Crocker, et. al. [11] took a logical next step of modelling the entire flow
field from compressor diffuser to the turbine inlet with the flow inside and outside the combustor
liner in a coupled manner. Their model includes an air blast fuel nozzle, dome, and liner wall
with dilution holes and cooling louvers. In their analysis, the flow outside the dome feeding the
fuel nozzle and the flow through the fuel nozzle were decoupled. Effusion cooling through the
dome face was also modelled in a decoupled manner. The flow rates feeding the air swirler and
effusion cooled dome walls were modelled as exit boundaries and were matched to the inlet flow
rates on the downstream side of the dome. The swirler inlet boundaries were modelled according
to the procedure of Crocker et al. [8]. The liner walls with film cooling and dilution holes were

2
ISOABE, Sept. 2-7, 2001, Bangalore, India.

modelled in a fully coupled manner. The holes feeding the film cooling slots were modelled by
concentrated resistances, as it was not practical to model the individual holes.
The present work is related to the computation of flow in a typical aero-engine gas turbine
combustor. In these computations, effort has been made to fully couple the outside and inside of
liner flow fields, by modelling the swirler flow through extra source terms in flow equations and
by making use of poracity factors, for the hole dimensions smaller than grid sizes.
Computations were carried out for given three flow conditions using a general-purpose
CFD code PHOENICS version 3.1. Some of the important aspects of the flow simulations were:
1. The complex geometry was simulated by appropriate BFC grid and the volume porosity
concept.
2. The analysis considered the diffuser, the dump diffuser, the outer and inner annuli and the
liner simultaneously.
3. The various air admission holes were simulated as accurately as feasible by the grid
topology. The poracity concept was used to simulate the holes smaller than the local grid
size.
4. The flow through the swirler was modelled by introducing extra source terms in the flow
equations. These terms accounted for the flow turning and pressure drop due to the swirler.
5. The turbulent flow was modelled by 3-D Reynolds averaged Navier-Stokes equations
employing the k- turbulence model.
6. The predictions from CFD simulation were validated with experimental results.

2.0 Theoretical Formulation

2.1 Mathematical model

The flow is governed by the Reynolds averaged Navier-Stokes equations which have the
general form:

(  )
 Div( V   grad )  S (1)
t
For =1, V, H, k and , the equations differ in their effective exchange coefficient  and
the source term. These quantities are listed in Table-1. The code PHOENICS solves the set of
above equations iteratively using a pressure based algorithm called SIMPLEST; a modified
version of SIMPLE.

Table -1
Source term and effective exchange coefficients for the solved variables.

Property   S
solved
Mass 1 0 0
Momentum V  Div -Div( grad V)
(Vector – grad p
form)
Enthalpy H /h 0

3
ISOABE, Sept. 2-7, 2001, Bangalore, India.

Turbulent k /k Gk-

kinetic
energy
Dissipation  / C1Gk/k –C22/k
rate of
turbulent
kinetic
energy

2.2 Swirler Modelling

Ideally, the flow through swirler should be computed by properly gridding the blade
passages in the swirler. But, this would amount to dealing with a very large number of grid
points with its associated problems like excessive computational time and difficulty in obtaining
convergence of the solution. As discussed above in introduction, most of the investigators have
used simpler models for the flow through the swirler. The present work also models the swirler
flow. The swirler vanes primarily do two things; one, they turn the flow and secondly they cause
pressure drop. These two effects can be simulated by introducing extra source terms in the
proper momentum equations written for the vane-less control volumes inside the swirler. These
source terms are as follows:

2.2.1 Source term for pressure drop

Sf = f (½)V2 (2)

This term introduces a frictional effect in the flow field, which accounts for frictional loss
due to swirler vanes, loss due to flow turning and diffusion, and sudden expansion loss near the
swirler exit. The combined effect of these losses is the pressure drop across the swirler.

2.2.2 Source term for flow turning

Sf = m Vaxial Tan() (3)

This term introduces an additional tangential force, and is responsible in turning the flow
along the vane.

2.3 Boundary Conditions

2.3.1 Input Conditions

The required inlet conditions for the CFD calculations were derived from the
experimental conditions. Table-2 shows these inlet conditions.

Table 2
Boundary conditions for a typical combustor
Run No. Case-2 Case-1 Case-3
Run:209 Run:203 Run:202

4
ISOABE, Sept. 2-7, 2001, Bangalore, India.

Mach Number 0.3053 0.3451 0.4088

Velocity (m/s) 111.7 120.5 146.6
3
Density (kg/m ) 2.57 2.96 3.07
Total temp (k) 344.2 316.2 338.7
K 15.0 18.0 26.0
 88.4 116.2 201.7
Ref. Pressure 2.60640 2.51777 2.85429
N/m2 (x105)

2.3.2 Turbulence Model

The standard k- turbulence model was used to account for turbulence in the flow. The
inlet values for k and  was computed by assuming 5% turbulence level at the inlet.

V1 = 0.05Vinlet (4)

k = (V1)2/2 (5)

 = 0.1643 k1.5/lm (6)

Where lm= 0.108 m (length of combustor)

2.3.3 Exit boundary condition

At exit, uniform static pressure as measured in experiment, was specified.

2.3.4 Cyclic boundary condition

As there are 18 injectors in the combustor, only 20 degree sector was considered for flow
computation and cyclic boundary conditions were applied at the end planes in circumferential x-
direction (ix=1 and ix=17).

3.0 Results and Discussion

Computations have been carried out for the three cases mentioned above. This section
presents and discusses these results. The predictions for the combustor are compared with
experimental results like mass flow split and total pressure loss. More details about the flow
features obtained from the numerical predictions are presented in the form of velocity vectors,
temperature, pressure and turbulent viscosity contour plots.

3.1 Grid

5
ISOABE, Sept. 2-7, 2001, Bangalore, India.

Structured Grid generation in a complex geometry like the combustor under consideration
is always a tedious exercise. The grid should align with the complicated boundaries, should be
locally fine enough to geometrically simulate small size openings and should have high density
in critical regions. Besides, the grid should also lead to a satisfactorily converged solution in a
reasonable time. It is not always easy to satisfy all these requirements simultaneously.
Moreover, the conventional method of successive refinement of grid to arrive at suitable grid is
also not practical in such complex situation. Hence, based on geometrical consideration, a grid
consisting of 17 cells in X direction, 50 cells in Y direction and 110 cells in the Z direction was
chosen in a 20 deg. sector of the combustor which is the computational domain (total 93,500
cells). Care was taken to have sufficient number of grid points in the region of dome, flare and
curved casing. Primary holes, secondary holes, dilution holes and exit bleed holes were
represented by equivalent cross sectional area of rectangular shape. Swirler was represented by a
rectangular annulus of equivalent flow area and grided. As the injector hole, ring-cooling holes,
film cooling holes and flare holes were very much smaller than the local grid size, these areas
were represented by area poracity concept.
Fig.1a shows the isometric views of grids in the 20 degree sectors of a combustors. Fig.1b
shows the grid on a X-plane (ix=1); we can observe that sufficiently fine grid was taken in
various holes and along liner wall. Fig.1c shows the grid on a Y-plane (iy=39) while Fig.1d
shows the grid on a Z-plane (iz=42), which also passes through the injector exit. In Fig.1c and
1d, we can see that the circular holes were approximated by rectangular holes of equivalent
areas.

3.2 Flow fields

The predicted flow fields are discussed below with the help of velocity vector plots and
contour plots of pressure, temperature and turbulent viscosity at various sectional planes of
combustors.
Fig. 2a shows velocity vector plot on a X-plane (ix=9) passing through the middle of the 20
degree sector i.e. in-line with injector plane, primary holes and dilution holes. The smooth flow
deceleration in the pre-diffuser, two strong recirculating flow regions in the dump area, excellent
penetration from primary hole and dilution hole inside the liner, two strong recirculating flow
regions in primary zone and flow acceleration near the exit are visible in the plots. The primary
hole air jets and dilution hole air jets from inner and outer annuli intersect nearly in the middle
region of the liner. Fig. 2b shows similar vector plot on another X-plane (ix=5), 5 degree away
from mid-plane, which is in-line with secondary holes. Here, the secondary hole air jet
penetration can be seen; the penetration is not as deep as seen for other holes in Fig. 2a. The
recirculating flow zones resemble those seen in Fig. 2a but the velocity profiles in the annuli
seem to be more uniform on this plane.
Fig. 2c and Fig. 2d show velocity vector plots on two Y-planes, one at the mid-height of
the sector (iy=25) and second near the outer liner casing (iy=40), respectively. Fig. 2c shows the
effect of swirler blockage on the flow in its neighbourhood while Fig. 2d shows the effect of
injector stem.
As noted earlier, the swirler was modelled numerically rather than geometrically by adding
extra source terms in momentum equations. The effect of this modelling can be seen in Fig. 2e
which shows the velocity vector plot on a Z-plane passing through the injector exit (iz=42). The
swirling flow emanating from the swirler can be clearly seen in the figure.

6
ISOABE, Sept. 2-7, 2001, Bangalore, India.

Fig. 3a shows the static pressure contours on the central X-plane (ix=9). Here, the pressure
values are relative to the reference pressure indicated in the figure. The plot shows the smooth
pressure rise in the pre-diffuser and local high pressure zones near dome, inner annulus and outer
annulus. There is relatively lower pressure inside the liner which helps smooth air admission
from outer and inner annuli into the liner through various holes on liner walls. The recirculating
zones are relatively constant pressure zones as seen in the plot. Fig. 3b shows the static pressure
plot on the exit plane of the combustor (iz=110). This figure is included here to indicate the
uniform static pressure condition for the openings (core and bleed holes) at the exit plane that has
been used in the computations.
Fig. 4 shows the turbulent viscosity contours on the central X-plane (ix=9). The plot clearly
shows that the turbulent viscosity levels are quite high inside the liner as the toroidal swirling
flow from swirler and impinging air jets from primary and dilution holes cause extremely high
turbulence in the region.

3.3 Mass Flow Distribution

In the experimental investigation, results were obtained for the mass flow split along outer
annulus, core and inner annulus, and total pressure loss across dump diffuser and combustor.
The same quantities were also estimated from the numerical solutions and compared with the
experimental results.
The flow split is plotted in Fig. 5a, 5b and 5c. These figures show the mass flow
distribution in terms of percentage of total mass flow (mass flow split) for the three major flow
paths, outer annulus, core and inner annulus. The figures contain the predicted values and the
measured values for the combustor. The degree of agreement varies from case to case. The
predictions consistently indicate slightly higher mass flow in the outer annulus which is
accompanied by correspondingly smaller mass flow in inner annulus. It can be observed in all
the three cases that the outer annulus is receiving about 4% more mass flow than the inner
annulus. This trend is in qualitative agreement with experiments for two cases but differs in one
case. There is no appreciable change in core mass flow in the predictions for the combustors
though these are somewhat higher than measured values. It is interesting to note that this trend is
in conformity with flow area available in the two major flow paths. The outer annulus flow area
is about 3.13% higher than the inner annulus flow area. The experiment shows similar trend for
the case 2 and 3 but opposite one for case 1. The agreement is reasonably good for the case 2
and 3 though the deviation is more pronounced for the case 1. In the two cases, an average
deviation of about 2.5% in mass flow split is observed from the experimental results. These
results also highlight an interesting feature about the flow split. As far as the predictions are
concerned, the flow split is relatively insensitive to the inlet Mach number. The experimental
results confirm this trend for outer annulus flow but seem to readjust the core and inner annulus
flow with different Mach number; the readjustment being much higher for case 1.

3.4 Total Pressure Loss

It is quite instructive to know the total pressure loss caused by different portions of the
combustor. From the predicted flow fields, the losses were estimated at important locations like
pre-diffuser exit (iz=18), outer dump diffuser (iz=55), inner dump diffuser (iz=55) and the
combustor exit (iz=110). These locations were selected in order to compare the predicted losses

7
ISOABE, Sept. 2-7, 2001, Bangalore, India.

with the measured values in the experimental investigation. These total pressure loss data are
plotted in Fig. 6a and 6b both of which show total loss and pre-diffuser loss along with either
outer dump diffuser loss (Fig 6a) or inner dump diffuser loss (Fig. 6b). The predictions indicate
very small total pressure loss for the pre-diffuser in all cases and it forms only a small fraction
of the total loss. These values could not be compared as there were no measurements for pre-
diffuser loss. The predicted losses for the inner and outer dump diffuser agree reasonably well
with experiments, as observed in these figures. These figures indicate that both predictions and
experiments confirm to an obvious fact that the total pressure loss increases with inlet Mach
number. There is very good agreement between the predictions and the experiments with regard
to the total pressure loss. Generally, the deviation is less than about 1% with measured values.

5.0 Conclusions
The code PHOENICS was used to simulate complex 3-dimensional turbulent flow in a
typical aero-engine gas turbine combustor configuration. The simulation took into account the
complete combustor, right from the pre-diffuser inlet to the combustor exit, simultaneously.
Care was taken to simulate all the geometrical complications like swirler, flare, injector air
passage, various air admission holes, film cooling holes, bleed holes etc. with structured BFC
grid. The effect of the swirler was simulated by introducing extra source terms in momentum
equations. Computations were carried out for three flow conditions and the predictions were
compared with experimental results like mass flow split and total pressure loss. The difference
between predicted and measured mass flow split varies from case to case, but on an overall basis,
it may be said that around 2.5% deviation from experimental values is observed for mass flow
split along outer annulus, core and inner annulus. There is very good agreement for total pressure
loss. The deviation in total pressure loss is generally found to be less that 1%. The predictions
indicate that the flow split does not vary significantly with inlet Mach number unlike total
pressure loss, which increases with increase in Mach number.

6.0 Acknowledgements
This work has been carried out in a project sponsored by the Gas Turbine Research
Establishment(GTRE), Ministry of Defence, Bangalore. The authors gratefully acknowledge the
sponsorship. They also thank the Centre for Mathematical Modelling and Computer Simulation
(C-MMACS), Bangalore for providing access to their computing facilities.

7.0 References
1. Sturgess G.J., “Calculation of aerospace propulsion combustors: a view from industry”,
AMD Volume 81 (Editors So. R.M.C., Whitelaw J.H. and Mongia H.C.)
2. Correa and Shyy W., “Computational models and methods for continuous gaseous turbulent
combustion”, Prog. Energy Combust. Sci. 13, 249-292, 1987.
3. Rizk N.K. and Mongia H.C., “Three-dimensional analysis of gas turbine combustors”,
Journal of Propulsion, 7(3), 445-451, 1991.
4. McGuirk.J.J., and Spencer.A., “Computational methods for modelling port flows in Gas
Turbine combustors,” ASME Paper 95-GT-414.

8
ISOABE, Sept. 2-7, 2001, Bangalore, India.

5. Lawson.R.J., “Computational modelling of an aircraft engine combustor to achieve target

exit temperature profiles,” ASME Paper 93-GT-164.
6. Martino P.Di., Cinque G., Paduano C. and Cirillo L., “CFD modelling of a practical annular
reverse flow combustor”, Proc. of 12th international Symposium on Air breathing Engine
(ISABE-95), 529-539, 1995.
7. Fuller.E.J., and Smith.C.E., “Integrated CFD modelling of gas turbine combustors,” AIAA
Paper 93-2196.
8. Crocker.D.S., Fuller.E.J., and Smith.C.E., “Fuel nozzle aerodynamic design using CFD
analysis,” ASME Paper 96-GT-127.
9. Shembharkar.T.R, Muralidhara.H.S, Chidananda.M.S, and Pai.B.R. “Computation of flow
in the LCGT combustor and the scroll,” NAL-PD-PR-9808, 1998.
10. Lai.M.K., “CFD Analysis of liquid spray combustion in a gas turbine combustor,” ASME
Paper 97-GT-309.
11. Crocker.D.S., Nickolaus.D., and Smith.C.E., “CFD Modelling of a gasturbine combustor
from compressor exit to turbine inlet,” J. of Engg. for gas turbine and power, 121, 89-95,
1999.

9
Fig.1a Isometric view of combustor grid Fig.1b Grid in X-Plane (ix=1)

Fig.1c Grid in Y-Plane (iy=39) Fig.1d Grid in Z-Plane at injector exit (iz=42)
Fig.2a Velocity vectors at injector plane (ix=9) Fig.2b Velocity vectors at the plane of .
secondary holes (ix=5)

Fig.2c Velocity vectors at the middle

Y-plane(iy=25)

Fig.2e Velocity vector in z-plane, at

injector exit(iz=42)
Fig.2d velocity vector near liner in
Y-plane (iy=39)
Fig.3a Temperature Contours at injector plane (ix=9)

Ref. Pr. =2.6064E+5

Fig.3b Pressure contours at exit (iz=110)

Fig.4 Turbulent viscosity at injector plane (ix=9)

 Mass Flow Distribution (%) Total Pressure Loss
Outer
12.00
Pre dif.-CFD
50
10.00

Tot. Pr. Loss (% )

Tot. pr. Expt
48 8.00
Mass Flow (%)

6.00 Tot. pr.-CFD

46 Expt.
CFD
4.00 Outer dump diff. Expt.
44
2.00
Outer dump diff.-CFD
42
0.00
40 0.25 0.3 0.35 0.4 0.45

0.25 0.3 0.35 0.4 0.45 Inlet Mach Number

Inlet Mach number

Fig.5a Percentage mass flow through outer annulus Fig.6a Total pressure losses of pre-diffuse,
outer dump diffuser and combustor.

Mass Flow Distribution (%) Total Pressure Loss

Core
12.00
Pre dif.-CFD
20
10.00
18
Tot. Pr. Loss (% ) Tot. pr. Expt.
8.00
Mass Flow (% )

16
Expt. Tot. pr.-CFD
14 6.00
CFD
12
4.00 Inner dump diff. Expt.
10
2.00
8 Inner dump diff.-CFD

6 0.00
0.25 0.3 0.35 0.4 0.45 0.25 0.3 0.35 0.4 0.45
Inlet Mach number Inlet Mach Number

Fig.5b Percentage mass flow through core region Fig.6b Total pressure losses of pre-diffuse, inner
dump diffuser and combustor.

Mass Flow Distribution (%)

Inner

50

48
Mass Flow (% )

46 Expt.
CFD
44

42

40
0.25 0.3 0.35 0.4 0.45
Inlet Mach number

Fig.5c Percentage mass flow through inner annulus

View publication stats