This document summarizes a study that performed a life assessment of a fighter jet engine annular combustor liner using a combined fluid/structural approach. Computational fluid dynamics analyses were conducted to obtain the thermal loading on the combustor liner from the combustion process. Finite element analyses were then performed to calculate the resulting temperature and stress/strain distributions in the liner. A method was developed to analyze a complete measured flight profile with limited computational effort. The creep and fatigue life for a measured flight were calculated and compared to field experience data. While the exact number of cycles to crack initiation was difficult to predict, the locations and directions of cracking correlated well with field data.
1. See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/239399921
Gas Turbine Combustor Liner Life Assessment Using a Combined
Fluid/Structural Approach
Article in Journal of Engineering for Gas Turbines and Power · January 2007
DOI: 10.1115/1.2360603
CITATIONS
29
READS
1,561
4 authors:
Some of the authors of this publication are also working on these related projects:
MAINTENANCE DECISION SUPPORT AND SIMULATION (MADESI) View project
Development of an Inline Mains Inspection Technology View project
Tiedo Tinga
Netherlands Defence Academy and University of Twente
93 PUBLICATIONS 957 CITATIONS
SEE PROFILE
Jaap van Kampen
11 PUBLICATIONS 179 CITATIONS
SEE PROFILE
Bram de Jager
HOFOR
12 PUBLICATIONS 98 CITATIONS
SEE PROFILE
J. B. W. Kok
University of Twente
148 PUBLICATIONS 1,167 CITATIONS
SEE PROFILE
All content following this page was uploaded by Tiedo Tinga on 09 November 2016.
The user has requested enhancement of the downloaded file.
3. This is indicated by the dotted line in Fig. 2. Only the distribution
of the boroscope holes is not periodic as these holes are not
present in all sections ͑only four͒. Since the effect of the holes on
the flow behavior inside the liner was believed to be significant, a
boroscope hole was modeled in the 1/16th section ͑see Fig. 3͒.
2.1 Computational Fluid Dynamics Model. A fluid-
dynamical analysis of the combustor was performed using the
commercial CFD code ANSYS CFX 5.7. This code uses an un-
structured solver, which facilitates the meshing of the relatively
difficult geometry of the combustor ͑see Fig. 3͒. The geometry
was meshed with mainly triangular mesh elements. The total num-
ber of elements assembling the geometry is 1.4 million.
In this section, the models that were used in the CFD analysis
are described. Moreover, the translation step from the GSP data to
boundary conditions for the CFD model is treated.
2.1.1 Model Description. For all cases considered in this pa-
per, the flow in the combustor is turbulent. To predict the effects
of turbulence on flow, combustion and heat transfer, the standard
available k-epsilon turbulence model was used. At all inlet bound-
aries, a turbulence intensity of 10% of the mean inlet velocities
was assumed, corresponding to intense turbulence.
Combustion was accounted for by a single-step combustion
model which uses an effective Arrhenius law for the reaction rate.
The fuel was considered to be fully vaporized, i.e., no liquid phase
is included in the simulation. The aero gas turbine hydrocarbon
fuel JP10 was used, for which the Arrhenius constants were fitted
from a computational singular perturbation method based reduced
version of a detailed chemical reaction scheme ͑see De Jager et al.
͓13͔ and the Appendix͒. From laminar flame data the reaction rate
as a function of temperature is known. In order to establish the
activation energy constant in an Arrhenius-type relation for the
global conversion of JP10 into combustion products, a Gaussian
fit can be applied to this data. The Gaussian fit to the activation
energy and the pre-exponential factor is given by:
S = 261 0e−͑T−1938/358.2͒
͓kg/m3
/s͔ ͑1͒
These laminar fitted global rate constants are used in the turbu-
lent combustion calculation by use of a probability density func-
tion weighting over the mass fraction of fuel. This is represented
in Eq. ͑2͒:
S˜
= ͵SP͑yf͒dyf͓kg/m3
/s͔ ͑2͒
In this compact combustor, turbulence intensity is not too high
in the combustion zone. For that reason and for simplicity the pair
distribution function for the fuel mass fraction is assumed to be a
delta function. This way Eq. ͑2͒ can be used directly in a simple
finite rate combustion simulation. This approach is used in many
flamelet based combustion models ͓14͔.
The flame in the combustor is stabilized by a recirculation zone
of hot gases. This recirculation zone is the result of the swirling
inlet flow. The axial swirler vanes of the burner were not included
in the model. However, from the geometry of the actual axial
swirler vanes, a 45 deg angle between the tangential and axial
velocity was estimated. The swirlers of the burners in the annular
combustor are co-rotating, which justifies the use of periodic
boundary conditions at the radial planes bounding the model.
The influence of the combustor walls on the flow was modeled
with a wall function ͓14͔. The walls itself were modeled as non-
adiabatic walls. At the cooling air side of the walls ͑outside the
computational domain͒, a heat transfer coefficient was imposed.
This heat transfer coefficient was calculated from the Nusselt cor-
relation of Dittus-Boelter. The input for this correlation followed
from the geometry of the cooling passage and the total amount of
air flowing past the inner and outer liner ͑see Fig. 4͒. These mass
flows were determined by an analysis of the air distribution over
the combustor, which is treated in the next section. The different
mass flows yielded different heat transfer coefficients for inner
and outer liner, ranging from 140 to 1400 W/m2
K, depending on
the engine operating condition considered. To obtain the heat flux
to the cooling air, the temperature of the cooling air was also
required, which was assumed to be constant. Since the residence
time of the air in the cooling passage is in the order of a few
milliseconds, the cooling air has only a small temperature increase
Fig. 1 Cracks in inner and outer liner of combustor
Fig. 2 Combustor liner with 1/16th section that is modeled
indicated by dotted line
Fig. 3 Geometry of CFD „left… and FE models „right…
Fig. 4 Inlets and outlet in CFD geometry
70 / Vol. 129, JANUARY 2007 Transactions of the ASME
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 07/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
4. while flowing past the liner.
Besides convective heat transfer, radiation was accounted for as
well. The fluid was considered optically transparent, implying that
only radiation from the fluid to the walls was included in the
analysis. Additionally, the radiative heat loss from the liner to the
casing was modeled. For this purpose, a constant temperature was
assumed for the casing: 100 K below the cooling air temperature.
2.1.2 Boundary Conditions. The GSP generated input data had
to be transformed to mass flows at each inlet of the CFD model.
These inlets comprise: the annular air swirler inlet of the burner,
seven film cooling inlets in the inner liner, eight film cooling
inlets in the outer liner, three pairs of dilution holes in the inner
liner, three pairs of dilution holes in the outer liner, and the boro-
scope hole in the outer liner ͑see Fig. 4͒.
The pre-vaporized fuel was supplied at the center of the burner
mouth. The film cooling holes, with a 1.5 mm diameter, are lo-
cated in rows along the steps of the liner. For convenient model-
ing, the holes in each row were merged into a single slit-formed
inlet. The distribution of the air flow through the burner and
through the ensemble of the liner holes was determined by use of
Bernoulli’s law, assuming a global pressure difference over the
liner. Furthermore, the distribution of the air over each individual
air inlet hole was derived as proportional to the area of each inlet.
This approach was allowed as the flow is by approximation, prior
to combustion, isothermal, incompressible, and inviscous.
2.2 Finite Element Model. The finite element analysis was
performed using the commercial FE code MSC.Marc 2001. In the
geometry as shown in Fig. 3 a finite element mesh was created
with isoparametric eight-node hexagonal elements. Typically ten
elements were used in circumferential direction and two elements
across the liner wall thickness, resulting in a total number of
11,646 elements and 18,285 nodes.
The model was first used for a thermal analysis to calculate the
temperature distribution in the liner, using the CFD results as
input. Then a structural analysis was performed to calculate the
stress and strain distribution. The loads and boundary conditions
that were used in these analyses are described in the next
subsections.
2.2.1 Thermal Analysis. During the combustion process heat
is transferred from the flame to the liner walls by convection and
radiation. At the same time, cooling at the other side of the liner
causes heat transfer from the liner wall to the cooling gas. This
causes a temperature gradient across the wall thickness. These
processes have to be covered by the thermal analysis, using the
CFD results as input. This was done in the following manner ͑see
also Fig. 5͒:
• Prescribing the CFD calculated wall adjacent gas tempera-
ture and heat transfer coefficient at the inner wall of the liner
͑convective boundary condition, Qconv͒
Qconv = h͑Tgas − Twall͒ ͑3͒
• Prescribing a radiative heat input at the inner wall using the
CFD calculated maximum gas temperature ͑Tmax͒ inside the
liner. The radiative heat input is defined as
Qrad = SB*
͑Tmax
4
− Twall
4
͒ ͑4͒
where SB is the Stefan-Boltzmann constant and *
is the
emissivity of the wall. For a metal surface the value of *
is
about 0.75. However, a ceramic thermal barrier coating
͑TBC͒ was applied to the liner wall, reducing the emissivity
to 0.35 ͓15͔.
• Prescribing a wall adjacent gas temperature equal to the
combustor inlet temperature and a heat transfer coefficient,
obtained from the Nusselt correlation of Dittus-Boelter, at
the outer wall of the liner ͑convective boundary condition,
Qconv͒.
• Prescribing a radiative heat input at the outer wall using the
casing temperature as Tmax. It was assumed that the casing
temperature is 100 K below the cooling gas temperature due
to casing cooling from the outside.
Given these boundary conditions, the liner temperature distribu-
tion can be calculated.
Before the CFD calculated temperature field could be applied
as a boundary condition, it had to be transferred from the CFD
grid to the FE mesh. The pre-processing program MSC.Patran was
used to transform the spatial fields for the CFD calculated tem-
perature ͑T͒ and heat transfer coefficient ͑h͒ distribution to a con-
vective FE nodal boundary condition.
The temperature drop across the TBC in the current application
was approximated to be about 5 deg in a steady-state condition.
Since the effect is very limited, this aspect of the TBC was not
taken into account in the analysis. The effect of a reduced emis-
sivity has a larger influence on the thermal behavior and was
therefore included in the model as was described before.
2.2.2 Structural Analysis. Since only 1/16th part of the liner
was modeled, periodic boundary conditions were applied at the
edges of the model to account for the constraints applied by the
omitted part of the structure. Therefore, displacements in circum-
ferential direction were constrained at both end faces. Further-
more, the rigid body motion was constrained by fixing one point
of the model ͑on the lug near the attachment to the diffuser cas-
ing͒ in all three directions. Finally, at the downstream end, both
the inner and outer liner are fixed to the casing with metallic
guides. These guides are sliding in axial direction and also provide
some freedom to move in radial direction. Furthermore, the guides
are somewhat flexible, which means that after the offsets in radial
and axial direction have been reached due to deformation of the
liners, they are still not rigidly fixed in space. Since it was ex-
pected that the liner deformations would not exceed the offsets of
the guides, no boundary conditions were applied in this region.
The temperature distribution as calculated in the thermal analy-
sis was used as loading for the structural model. Due to differ-
ences in thermal expansion thermal stresses develop and the ma-
terial deforms elasto-plastically.
2.3 Life Assessment. Creep and low cycle fatigue ͑LCF͒ are
the two potential failure mechanisms for a combustor liner. For
both mechanisms a lifing method was defined, as will be de-
scribed in this section. Also, the interaction between the mecha-
nisms is discussed.
The creep strain accumulation was calculated during the FE
analysis. The creep strain rate depended on stress ͑in Pa͒ and
temperature T ͑in K͒ in the following way
˙cr = 3.48 · 10−20
e4.02.10−8
e0.0211T
͑5͒
FE analysis of one flight provides the creep strain accumulation
during that flight. However, analyzing the complete service life of
a combustor in this way is computationally not feasible. There-
fore, a method was defined to determine creep life based on a FE
Fig. 5 Schematic representation of heat flows in combustor
cross section
Journal of Engineering for Gas Turbines and Power JANUARY 2007, Vol. 129 / 71
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 07/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
5. calculation of limited duration ͑⌬t͒.
The most straightforward method is extrapolation of the dam-
age after one flight ͑using the average creep rate͒ towards the
critical amount of creep damage, also known as Robinson’s rule
͓16͔:
tcr =
cr,critical
cr,⌬t
⌬t =
cr,critical
˙cr,aver
͑6͒
where cr,critical is the creep strain that is associated with the end of
life.
However, it is well known that creep is not a linear process. The
first reason is the non-linear shape of the creep curve and the
second reason is the occurrence of stress relaxation. This makes
that the creep strain rate is initially very high but then decreases to
a constant value, as is illustrated in Fig. 6. If a FE creep analysis
is performed for one flight, only a small part of the creep curve is
covered and the constant creep rate is not reached. Extrapolation
with the average creep rate towards the critical creep strain then
yields a ͑very͒ conservative estimate of the creep life ͑see line 1 in
Fig. 6͒. Therefore, the tangent creep rate at the end of the time
period ͑˙cr,⌬t͒ was used for the extrapolation ͑line 2 in Fig. 6͒,
which yields a less conservative life assessment:
tcr =
cr,critical
˙cr,⌬t
͑7͒
The value of cr,critical is typically 2% creep strain. Note that this
value is lower than the creep strain at rupture. However, creep
rupture does not occur in this situation, since at creep strains of
2% other failure mechanisms, like coating spallation or grain
boundary oxidation, become active and are the life limiting
mechanisms.
For the low cycle fatigue ͑LCF͒ life analysis the cyclic loading
of the construction is required. The fatigue life of a component is
a combination of crack initiation and crack propagation. Normally
the initiation life is calculated with a Manson-Coffin-type model
and the crack propagation is predicted with a fracture mechanics-
type analysis. However, such a detailed crack analysis is beyond
the scope of this work and therefore a total life analysis is per-
formed, in which the LCF life is defined as the number of cycles
to a certain crack length. It is generally accepted that the initiation
of cracks is best correlated with the maximum shear strain,
whereas the propagation of cracks is governed by the maximum
principal strain component. Since the maximum principal strain
also provides information about the direction of crack propaga-
tion, this quantity was selected to calculate the LCF life distribu-
tion in the combustor. Further, for LCF conditions the load cycles
contain both elastic and plastic deformation, which means that the
LCF life is related to the total strain range ͑⌬tot=⌬el+⌬pl͒.
The relation between total strain range ͑⌬tot in %͒ and LCF life
͑NLCF͒ for the liner material Hastelloy-X was obtained from See-
ley et al. ͓17͔ and the functional relation used in the FE analysis is
given by
NLCF = ͩ ⌬tot
21.253
ͪ−2.36
͑8͒
which represents the 870°C curve.
Finally, it is possible to combine the creep and fatigue damage
to obtain the total creep and fatigue life of the combustor. How-
ever, it is very difficult to obtain the proper interaction between
the two failure mechanisms. The simplest method is to combine
the two mechanisms linearly, which is known as the Miner-
Robinson rule:
1
Ntot
=
1
NLCF
+
1
Ncr
͑9͒
3 Analysis of a Complete Flight
The integrated approach described in the previous section can
be used to analyze the liner loading and life consumption for a
single operating condition. However, a real flight is a sequence of
a large number of different operating conditions. Analyzing all
conditions separately is computationally not feasible. Therefore, a
method was developed to efficiently analyze a complete flight. In
this section the method is described and the accuracy is demon-
strated.
3.1 Method Description. The basic idea of the method is that
detailed CFD analyses were performed for a limited number of
characteristic conditions and that for any other condition thermal
input properties were obtained by interpolating between these
conditions. The time scale of the processes in the combustor fluid
domain ͑ϳ milliseconds͒ is very different from the time scale of
the overall engine transients ͑ϳ seconds͒. The transient behavior
of the engine ͑e.g., fast or slow accelerations͒ is captured by GSP
and translated into time sequences of the quantities that character-
ize the combustion process ͑fuel and airflow͒. Since the sampling
rate of these quantities is in the order of seconds, it is appropriate
to consider a transient mission analysis as a number of consecu-
tive quasi-steady-state CFD analyses. Therefore, there is no need
to do transient CFD analyses.
Appropriate quantities for interpolation, interpolation functions,
and characteristic conditions had to be defined. Therefore, a large
number of measured flights were analyzed with the gas turbine
simulation program GSP. Air inlet temperature ͑TT3͒ and fuel/air
ratio ͑Wf /Wair͒ proved to be the only two quantities required to
fully characterize the combustor operating condition. Also the op-
erating ranges of these parameters were established in this way,
resulting in the definition of five characteristic operating condi-
tions as indicated in Table 1. CFD analyses were performed for
these five conditions and the results were stored in a database.
Interpolation functions were selected for the quantities required
to define the boundary conditions in the FE analysis: wall adjacent
gas temperature, inside heat transfer coefficient, maximum gas
temperature, and outside heat transfer coefficient. GSP analyses
showed that the gas temperature rise in the combustor was almost
directly proportional to the fuel/air ratio, but the proportionality
Fig. 6 Creep strain extrapolation using the average strain rate
„1… and the tangent strain rate „2…
Table 1 Details of characteristic operating conditions
Condition 1 2 3 4 5
TT3͓K͔ 750 380 1000 750 750
Wf /Wair
0.0257 0.0257 0.0257 0.0033 0.0118
72 / Vol. 129, JANUARY 2007 Transactions of the ASME
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 07/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
6. constant depended on the initial gas temperature value. Therefore,
the following linear interpolation function was adopted for the
wall adjacent gas temperature:
Tgas = bTT3 + a
Wf
Wair
⇔
Tgas
TT3
= b + a
Wf
WairTT3
͑10͒
The heat transfer coefficient is largely determined by the
amount of turbulence in the flow, which is strongly related to the
mass flow. Therefore, the following linear interpolation function
was adopted for the heat transfer coefficient:
h = c + dWair ͑11͒
The interpolation functions for the maximum temperature in the
combustor ͑used for radiative heat input͒ and the outside heat
transfer coefficient were defined as
Tmax = 925 + TT3 + 22,946
Wf
Wair
͑12͒
houtside = 22.6 · Wair ͑13͒
The parameters a, b, c and d in Eqs. ͑10͒ and ͑11͒ were deter-
mined for every node on the liner inner surface by fitting the
functions to the five result cases that were available in the data-
base of CFD results. This was done by using a least squares fitting
method. The parameters were stored in a file, which was used
during the transient FE analyses. The functions in Eqs. ͑12͒ and
͑13͒ are valid for the entire combustor model and therefore do not
require fitting for all nodes.
The method to perform a transient analysis of a complete flight
is shown schematically in Fig. 7 and can be summarized as fol-
lows.
• A number of CFD analyses were performed for five charac-
teristic engine operating conditions that span the operating
range, see Table 1.
• The calculated wall adjacent temperature distribution and
heat transfer coefficient distribution for these conditions
were transformed to FE boundary conditions and stored in a
database.
• A recorded flight is analyzed with GSP as a transient analy-
sis. GSP calculates the time sequence of the input variables
TT3 and Wf /Wair.
• A transient thermal FE analysis is performed in which the
time-dependent boundary conditions are calculated by inter-
polating ͑for every node at the liner inner surface͒ between
the conditions in the CFD database, based on the given val-
ues of TT3 and fuel/air ratio.
• A transient structural FE analysis is performed in which the
calculated time-dependent temperature distribution is im-
posed as loading.
• A life assessment is performed based on the FE calculated
stress and temperature sequences.
3.2 Method Validation. In addition to the five CFD analyses
for the conditions given in Table 1, one extra CFD analysis was
performed to validate the interpolation method. The condition of
this analysis, referred to as condition 6, represents the engine de-
sign point characterized by TT3=817.6 K and Wf /Wair=0.0322.
This yields a fuel/air ratio of 0.0322 and a condition that is quite
different from any of the other conditions in Table 1.
The first validation was done by comparing the CFD calculated
Tgas and h with the interpolated Tgas and h. The result for every
node on the liner inner surface is shown in Fig. 8. This shows that
for the gas temperature the average of the complete set of points is
on the 1-to-1 line and the deviation from the calculated value is
typically less than 10%. For the heat transfer coefficient the cor-
relation is even better.
Another validation of the method is shown in Fig. 9, where the
calculated liner temperature distribution based on the CFD calcu-
lated and interpolated Tgas and h are compared. The distribution is
very similar and also the global temperature values are very com-
parable. Only at the hot spots the interpolated temperature was in
Fig. 7 Schematic representation of mission analysis method
Fig. 8 Comparison of CFD calculated and interpolated Tgas and h
Journal of Engineering for Gas Turbines and Power JANUARY 2007, Vol. 129 / 73
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 07/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
7. this case about 150° higher. This is a considerable deviation, but it
is only for a small number of local spots. That may imply that at
those spots the stresses relax quickly and that the effect on life
consumption is small. However, the exact influence of these tem-
perature deviations on life consumption was not determined. Fur-
thermore, the condition used for validation is a severe condition
which is quite different from the conditions used for interpolation.
Therefore, for other conditions that occur during a mission the
deviations may be smaller.
These two validation cases show that the method developed
reduced the computational effort for a mission analysis enor-
mously and still retained a very acceptable accuracy level. If a
higher accuracy is required, the number of CFD analyses can be
extended, which improves the interpolation method. However, the
consequence will be a more time-consuming and more costly
analysis.
4 Results
The integrated approach as described in Secs. 2 and 3 was used
to analyze a real flight and perform a combustor liner life assess-
ment. The variation of power lever angle ͑PLA͒ and flight altitude
for the selected flight are shown in Fig. 10. The analysis results
are shown in this section, starting with the flow properties pre-
dicted by the CFD analysis for a selected condition. Then the
predicted temperature and stress distribution are shown and the
results of the life assessment are presented. Finally, the results
were compared to the damage observed in practice to validate the
approach.
Fig. 9 Comparison of results based on real „left… and interpolated temperature distribution on outer liner
Fig. 10 Variation of power setting „PLA… and flight altitude dur-
ing analyzed mission
Fig. 11 Vector plot of the velocity field in a radial plane of the
combustor
74 / Vol. 129, JANUARY 2007 Transactions of the ASME
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 07/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
8. 4.1 Flow Properties. Figure 11 shows a vector plot of the
flow field in a radial cross section of the burner for the first of the
characteristic operating conditions ͑condition 1, see Table 1͒. As a
result of the swirling inlet flow, a recirculation zone is established,
providing flame stabilization. This recirculation zone is noticed by
reversing flow near the burner outlet. Additionally, since the cross
section is made in the plane of the boroscope hole, the effect of
the cooling air flowing through this hole is clearly seen.
A contour plot of the scaled temperature field at the same op-
erating condition is displayed in Fig. 12. The cooling air that is
introduced by the boroscope hole is clearly influencing the tem-
perature field. Moreover, due to the applied film cooling, the fluid
near the wall remains relatively cool. The maximum temperature
is found at approximately halfway the length of the combustor.
4.2 Thermal and Structural Results. The temperature distri-
bution in the liner walls at operating condition 6 is shown in Fig.
13͑a͒. This shows that the thermal loading of the liner is maxi-
mum at the downstream end of the inner liner. Figure 13͑b͒ shows
the tangential stress distribution in the liner walls. As with the
temperature, the extreme values were found at the downstream
ends of inner and outer liner.
The time variation of temperature and stress during the com-
plete flight for the four locations indicated in Fig. 14 is shown in
Fig. 15. This shows that significant variations in temperature and
stress occur during the flight. For locations on the inside of the
liners, stresses were mainly compressive, whereas locations on the
outside were in a tensile stress state. However, the situation can be
completely opposite at fast transients, as is observed just after
2000 s. Also, the average stress level decreased during the flight,
which is caused by stress relaxation due to creep.
4.3 Life Assessment. The temperature and stress history were
used to perform a life assessment on the combustor liner. Both the
creep life and low cycle fatigue life were calculated applying the
methods described in Sec. 2.3. The calculated distribution of creep
and LCF life are shown in Fig. 16, indicating the critical locations
on the liners. This plot shows that the creep life is minimal at the
final louvers of both inner and outer liner with a minimum value
of only 4 h. This calculated life is much too low and is probably
a result of an overestimated temperature in the liner. Since the
creep mechanism is extremely sensitive to the temperature, a
small deviation in temperature leads to significant changes in
creep life. This problem in predicting the creep life is a result of
the lack of accuracy in the prediction of absolute values that will
be addressed in the next subsection.
The minimum calculated fatigue life is around 1900 flights at
some very small areas, mainly at the final louvers of the inner and
outer liner. The majority of the structure has a fatigue life well
above 10,000 flights.
4.4 Model Validation. To gain insight into predictive capa-
bility, the model had to be validated, especially because a large
number of assumptions and approximations were made in the
CFD, FE, and lifing analyses. Ideally, the model should have been
validated and tuned to experimental data after each step in the
sequence. For example, detailed information about the tempera-
ture distribution in the liner at a specific operating condition
would give the possibility to tune the values of the outside heat
transfer coefficients to obtain a good correlation with the available
data. However, experimental data to do so was not available. Ac-
Fig. 12 Contour plot of the temperature field in a radial plane
of the combustor at condition 1
Fig. 13 „a… Temperature „in °C… and „b… tangential stress „in Pa… distribution at operating condition 6
Fig. 14 Location of nodes used for plotting the temperature
and stress history
Journal of Engineering for Gas Turbines and Power JANUARY 2007, Vol. 129 / 75
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 07/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
9. tually, the only way to validate the model was to compare the
calculated life to the values observed in practice. This means that
in the very complex chain of CFD—FE thermal—FE structural—
life assessment analyses no intermediate check of the model was
possible. It is quite likely that some assumption somewhere in one
of the analyses leads to a deviation in results.
The results of the life assessment were compared to the damage
observed in practice as shown in Fig. 1. The location of the cracks
is expected to coincide with the location of lowest calculated fa-
tigue life. As can be seen in Fig. 16, the lowest fatigue life was
calculated at the final three louvers of the inner and outer liner.
Furthermore, the creep damage on the final three louvers of the
inner liner is significantly higher than on those of the outer liner.
Since creep is believed to enhance crack initiation, it may be
concluded that cracks will start to propagate at the final three
louvers of the inner liner. This correlates well with the observation
in practice that cracks occur at the inner liner on the 4th to 9th
louvers.
The direction of crack growth is related to the orientation of the
stress tensor in the material. In other words, cracks propagate
mainly in a direction perpendicular to the maximum principal
stress. Comparison of the different components of the calculated
stress distribution in the louvers showed that the circumferential
stress was the largest stress component and also best resembled
the max principal stress distribution, as is shown in Fig. 17. This
means that cracks will propagate in axial direction, as was ob-
served in practice.
Finally, the number of cycles or flights to produce a fatigue
crack could be checked. This showed that the order of magnitude
of the calculated life corresponded to the real life, but the absolute
prediction is not very accurate.
The only proper way to improve the predictive capability of the
model is validation of intermediate results with experimental data.
Although the prediction of absolute values for combustor life at
certain operating conditions is hard, the current method is very
appropriate to perform comparative life assessments. In that case,
a wrong assumption somewhere in the model has a similar effect
for every condition that is analyzed, which reduces the inaccuracy
of the comparison considerably. Examples of such analyses are
comparison of different mission types in terms of life consump-
tion or the effect of increased firing temperature ͑due to compres-
sor or turbine deterioration͒ on life consumption.
5 Conclusions
The following conclusions can be drawn from this work:
• The results of a CFD analysis were successfully linked to a
FE model, which gave the possibility of calculating the ther-
mal loading of the combustor for an arbitrary operating con-
dition.
• An efficient method was developed to analyze a complete
measured flight. A compromise between accuracy and com-
putational effort must be found for each application.
• Although the actual number of cycles to crack formation
could not be predicted accurately, the location of the cracks
and the direction of crack growth correlated well to the ob-
servations in practice.
• The developed models and tools can be applied to perform
comparative life assessment, e.g., for different mission
types.
Fig. 15 Temperature and tangential stress variation on liner during a mission for the locations shown in Fig. 14
Fig. 16 Creep life „in hours… and fatigue life „in number of flights… distribution in the liner
76 / Vol. 129, JANUARY 2007 Transactions of the ASME
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 07/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
10. Appendix: JP10 Jet Fuel Chemistry
JP10 is a hydrocarbon fuel that is used in aero gas turbines. The
molecule consists of 10 C atoms and 16 H atoms and its system-
atical name is Tricyclo͓5.2.10͔decane. Stoichiometric oxidation is
given by the reaction
C10H16 + 14O2 → 8H2O + 10CO2
Williams et al. ͓18–20͔ provided a mechanism for the oxidation
of JP10 that consists of 42 species and 201 reactions. The basis of
this mechanism is the San Diego Combustion mechanism. Added
to this mechanism is a mechanism for the combustion of JP10.
This addition consists of the breakup of the JP10 molecule into
smaller hydrocarbons like butadiene, etc.
Laminar premixed 1D flame calculations at atmospheric and
Fig. 17 Distribution of „a… circumferential and „b… max. principal stress on the final louvers of the inner liner
Fig. 18 The temperature and species profiles for a premixed flame at a pressure of 1 and 18 atm
Journal of Engineering for Gas Turbines and Power JANUARY 2007, Vol. 129 / 77
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 07/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
11. elevated pressure yielded insight into the chemical reactions that
dominate the oxidation behavior of JP10. The laminar flame cal-
culations were analyzed using the computation singular perturba-
tion ͑CSP͒ ͓21͔ technique, in order to select dominant species that
govern the conversion of JP10 into products. This CSP analysis
can give a global reaction rate as a function of the species in the
mechanism or temperature.
Laminar Flames. Insight into global rates can be obtained by
simulating a laminar premixed flame. In this report the results of a
stoichiometric flame at 1 and 18 atm are shown. Figure 18 shows
that the mechanism is behaving differently at atmospheric and
high pressure. The temperature profile of the laminar flame at
atmospheric pressure has a less steep descent than the temperature
profile at 18 atm. The concentration of intermediate species and
combustion products is not changing very much with the different
pressures.
CSP Reduction. CSP analysis was carried out using the results
from the laminar flame. A theoretical background of this method
can be found in papers by Mastorakos et al. ͓21,22͔. This analysis
yields the species in Table 2 to be dominant in reaction rate de-
termination, if the mechanism is to be projected on one global
step. The global step that is constructed by CSP is given by the
following reaction:
CO2 → CO + O2
The results from CSP are similar to reduction of methane com-
bustion as is reported by Mastorakos et al. ͓22͔. An explanation
for this is that both methane and JP10 are hydrocarbons and glo-
Table 2 Non-steady-state species
CO CO2 O2 H2O N2
AR
Fig. 19 Composed species concentration and source term „reaction rate…
Fig. 20 CSP source term „reaction rate… fitted with Gaussian function
78 / Vol. 129, JANUARY 2007 Transactions of the ASME
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 07/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
12. bally the same type of CH bond needs to be broken. Other types
of CH bonds are also present in the JP10 mechanism, but they do
not seem to affect the global step. The global step is not a physical
step, it can rather be seen as an overall reaction that yields from
element and mass conservation, but with virtual species.
Rate Fitting and Turbulent Simulations. Using the CSP
analysis it is possible to construct a global reaction rate that rep-
resents the overall oxidation process composed of steady state
relations. This composed reaction rate is accompanied by a com-
posed species that represents the thermo-chemical state of the
mixture. In Fig. 19 this CSP analysis is projected on the laminar
flame results from the previous section. It is seen that for atmo-
spheric and high pressure there is a significant difference in the
concentration profile of . At high pressure the composed species
behaves as a fuel molecule: a high initial concentration and during
ignition a high consumption rate of the composed species. The
shape of this profile is similar to the JP10 concentration profiles in
Fig. 18. For atmospheric pressure, the composed species concen-
tration profile is not showing much resemblance with a fuel
molecule.
For turbulent combustion simulations the composed species and
global rate can be used as a single step mechanism. This gives the
advantage that only one extra transport equation is needed. From
the laminar flame data at 18 atm the reaction rate as a function of
temperature is known. In order to establish an Arrhenius-like re-
lation for the global conversion of JP10 into combustion product,
a Gaussian fit can be applied to this data ͑see Fig. 20͒. The Gauss-
ian function is given by
S = 2610e−͑T−1938/358.2͒
͓kg/m3
/s͔ ͑A1͒
This fitted rate can be used in a simple finite rate combustion
simulation.
References
͓1͔ Walls, D. P., and de Laneuville, R. E., 1997, “Damage Tolerance Based Life
Prediction in Gas Turbine Engine Blades Under Vibratory High Cycle Fa-
tigue,” ASME J. Eng. Gas Turbines Power, 119, pp. 143–146.
͓2͔ Harrison, G. F., and Tranter, P. H., 1995, “Modelling of Thermomechanical
Fatigue in Aero Engine Turbine Blades,” Proceedings of the 81st Meeting of
the AGARD Structures and Materials Panel on Thermal Mechanical Fatigue of
Aircraft Engine Materials, Oct. 2–4, Banff, Canada, pp. 1–12.
͓3͔ Tinga, T., Visser, W. P. J., de Wolf, W. B., and Broomhead, M. B. J., 2000,
“Integrated Life Analysis Tool for Gas Turbine Components,” ASME Paper
No. 2000-GT-646, NLR-TP-2000-049.
͓4͔ Scott Crocker, D., Nickolaus, D., and Smith, C. E., 1998, “CFD Modeling of
a Gas Turbine Combustor From Compressor Exit to Turbine Inlet,” ASME
Paper No. 98-GT-184.
͓5͔ Sivaramakrishna, G., Muthuveerappan, N., Venkataraman, S., and Sampathku-
maran, T. K., 2001, “CFD Modeling of the Aero Gas Turbine Combustor,”
ASME Paper No. 2001-GT-0063.
͓6͔ Malecki, R. E., and Rhie, C. M., 2001, “Application of an Advanced CFD-
Based Analysis System to the PW6000 Combustor to Optimize Exit Tempera-
ture Distribution—Part I: Description and Validation of the Analysis Tool,”
ASME Paper No. 2001-GT-0062.
͓7͔ McQuirk, J. J., and Spencer, A., 2000, “Coupled and Uncoupled CFD Predic-
tion of the Characteristics of Jets From Combustor Air Admission Ports,”
ASME Paper No. 2000-GT-125.
͓8͔ Snyder, T. S., Stewart, J. F., Stoner, M. D., and McKinney, R. G., 2001,
“Application of an Advanced CFD-Based Analysis System to the PW6000
Combustor to Optimize Exit Temperature Distribution—Part II: Comparison
of Predictions to Full Annular Rig Test Data,” ASME Paper No. 2001-GT-
0064.
͓9͔ Fuller, E. J., and Smith, C. E., 1993, “Integrated CFD Modeling of Gas Tur-
bine Combustor,” AIAA Paper No. 93–2196.
͓10͔ Lai, M. K., 1997, “CFD Analysis of Liquid Spray Combustion in a Gas Tur-
bine Combustor,” ASME Paper No. 97-GT-309.
͓11͔ Kiewel, H., Aktaa, J., and Munz, D., 2002, “Advances in the Inelastic Failure
Analysis of Combustor Structures,” in: High Intensity Combustors—Steady
Isobaric Combustion, Final Report of the Collaborative Research Centre, 167,
S. Wittig and O. Vohringer, eds., Wiley-VCH, Weinheim.
͓12͔ Visser, W. P. J., and Broomhead, M. J., 2000, “GSP—A Generic Object-
Oriented Gas Turbine Simulation Environment,” ASME Paper No. 2000-GT-
002, NLR-TP-2000–267, www.gspteam.com.
͓13͔ De Jager, B., Kok, J. B. W., and Van der Meer, Th. H., 2004, “Development of
Reduced Chemistry With CSP for Application in Turbulent n-Heptane flames,”
Proceedings of the European Congress on Computational Methods in Applied
Sciences and Engineering (ECCOMAS), P. Neittaanmäki, T. Rossi, K. Majava,
and O. Pironneau eds., W. Rodi, and P. Le Quéré assoc. eds., Jyväskylä, Fin-
land, July 24–28, pp. 1–13.
͓14͔ Peters, N., 1998, “Laminar Flamelet Concepts in Turbulent Combustion,” Pro-
ceedings of the International Symposium on Combustion, Munich, pp. 1231–
1250.
͓15͔ Wittig, S., and Vohringer, O., 2002, “High Intensity Combustors—Steady Iso-
baric Combustion,” Final Report of the Collaborative Research Centre, 167, S.
Wittig and O. Vohringer, eds., Wiley-VCH, Weinheim, pp. 1–463.
͓16͔ Robinson, E. L., 1952, “Effect of Temperature Variation on the Long-Time
Rupture Strength of Steels,” Trans. ASME, 74͑5͒, pp. 777–781.
͓17͔ Seeley, R. R., and Ishwar, V. R., 1999, “Fatigue in Modern Nickel-base Alloys
for Gas Turbine Applications,” Proc. Conf. Life Assessment of Hot Section Gas
Turbine Components, Oct. 5–7, Edinburgh, UK, pp. 61–82.
͓18͔ Williams, F. A., 2003, Technical Report, http://maemail.ucsd.edu/combustion/
cermech/.
͓19͔ Rupley, F. M., Kee, R. J., and Miller, J. A., 1989, “Chemkin II: A FORTRAN
Chemical Kinetics Package for the Analysis of Gas Phase Chemical Kinetics,”
Sandia Report No. SAND 89-8009.
͓20͔ Varatharajan, B., Li, S. C., and Williams, F. A., 2001, “The Chemistry of JP10
Ignition,” AIAA J., 39͑12͒, pp. 2351–2356.
͓21͔ Goussis, D. A., and Lam, S. H., 1994, “The CSP Method for Simplifying
Kinetics,” Int. J. Chem. Kinet., 26, pp. 461–486.
͓22͔ Mastorakos, E., Massias, A., Diamantis, D., and Goussis, D. A., 1999, “An
Algorithm for the Construction of Global Reduced Mechanisms With CSP
Data,” Combust. Flame, 117, pp. 685–708.
Journal of Engineering for Gas Turbines and Power JANUARY 2007, Vol. 129 / 79
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 07/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-useView publication statsView publication stats