Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 80 (2014) 282 292

3rd International Symposium on Aircraft Airworthiness, ISAA 2013

Effect of cavity structure on the risk of engine life limited

parts
Zhang Gonga*, Zhou YanPeia
Aeroengine Airworthiness Certification Center Preparatory Office, China Academy of Civil Aviation Science and Technology,
Beijing 100028, P.R.China

Abstract
An approach integrating one way fluid structure interaction and probabilistic risk assessment is developed to evaluate
the effect of cavity on the risk of engine life limited parts, by which the effect of the variation of structure parameters
in the cavity on the safety of gas turbine disks is investigated. The results indicate that the structure parameters affect
the failure risk of the disk with initial anomalies since they changes the distribution of the stress and temperature of the
disk. It demonstrate that engine is a system which functional connection with other subsystems, especially life limited
parts. In addition, a sensitivity analysis is conducted to determine the sensitivities of the safety with respect to structure
parameters.
2014 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
2013 Published by Elsevier Ltd. Selection and peer-review under responsibility of ENAC
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Selection and peer-review under responsibility of Airworthiness Technologies Research Center, Beihang University/NLAA.
Keywords: Airworhiness; Engine Life Limited Parts; Cavity; Probabilistic Risk Assessment

1. Introduction
The rotor-stator cavity, commonly used in the aero-engine secondary flow cooling system, has
significant impact on the integrity of turbine disks which are engine life limited parts (ELLP) since the
fracture of a gas turbine disk could directly induce non-containment of high energy debris, which is one of
the hazardous events according to airworthiness regulations [1,2]. Therefore it is necessary to consider the
influence of the cavity during the design and certification procedures, otherwise it could cause hazardous
effect and result in catastrophic events like loss of aircraft and human life.

* Corresponding author. Tel.: +86-10-64481253; fax: +86-10-64481206.

E-mail address: zhangg@mail.castc.org.cn.

1877-7058 2014 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Selection and peer-review under responsibility of Airworthiness Technologies Research Center, Beihang University/NLAA.
doi:10.1016/j.proeng.2014.09.087

Zhang Gong and Zhou YanPei / Procedia Engineering 80 (2014) 282 292

The rotor-stator cavity is such a system in which a disk rotates close to a casing where the bleeding
airflow from the compressor in the space between the casing and disk is used for cooling and seal. Fig. 1
shows a typical cavity structure and its numerical model. The traditional design process of the cavity is
rather complex, involving considerable iterations and usually leads to several independent studies such as
fluid flow and heat transfer, structure analysis and safety assessment since the relationship between
subsystems of the cavity is neglected during the traditional process. As a result, there are extensive
investigations concerning fluid flow and heat transfer of rotor-stator systems [3-6]. Based on the previous
studies, Owen et al. pointed out that two important ways to improve turbine efficiency are enhancing the
cooling effectiveness and reducing the mass flow of cooling air [7]. These efforts played an important role
in the design of the cavity. However, most of them evaluated the effect of the cavity on cooling effectiveness
or flow resistance rather than failure risk of the disk. That is to say, to date most investigations have been
concerned with the engine performance, despite the safety of ELLP.

Sc

rin

R0

Cooling
air
S

b r0

Fig. 1: (a) typical rotor-stator cavity structure; (b) axisymmetric model of a rotor-stator cavity

There are also many investigations regarding the safety issue, but none of them focuses on other
subsystems around the ELLP like the cavity by far. Miles et al discussed the effect of elements on the risk
prediction based on a probabilistic approach which is established by using finite element analysis and linear
elastic fracture mechanism and used to study ELLP life and reliability. The Design Assessment of
Reliability with Inspection (DARWIN) is introduced as a numerical simulation to predict fracture
probability of disk with initial anomalies [8]. Leverant et al [9,10] then discussed some enhancement of
DARWIN to assess the risk of rotating system. In addition, Enright et al [11-13] evaluated the effect of
various distributions of anomalies and random variables on the risk prediction and enhanced DARWINs
capability. It is validated that DARWIN with sufficient database could be an acceptable method of
compliance with the requirement of FAR 33.70 and CS-E 515 by the study on DARWINs process and data
availability[14]. These researches are all beneficial to improve the safety of aero-engine. However, these
attempts have been made with respect to ELLP itself solely rather than deep consideration with its crucial
working environment together.
Although many studies have been conducted to address performance of the cavity and the safety of the
disk separately, the effect of the cavity structure on the risk of ELLP is rarely studied. This paper mainly
focuses on the effect of structure parameters in a rotor-stator cavity on fracture probability of the disk with
initial anomalies. A numerical approach integrating fluid-structure-interaction (FSI) and probabilistic risk
assessment (PRA) is established to investigate the variation of failure probability of the disk with respect to
structure parameters in the cavity. A sensitivity analysis is also performed to identify the dominant variable
in this field.

283

284

Zhang Gong and Zhou YanPei / Procedia Engineering 80 (2014) 282 292

2. Methodology
Fig. 2 shows the framework we established to perform the investigation. It contains three steps: the loads
of the disk according to the variation of structure parameters of the cavity is obtained by using one-way
FSI; then the loads, such as temperature and stress distribution, are used as the input data of PRA to predict
the probability of fracture of the disk (Pf); the sensitivity of Pf to the structure parameters is evaluated finally.
Boundary
Conditions of
Rotating Cavity

One way
FSI

Probabilistic Risk
Assessment

Sensitive
Analysis

Stress and
Temperature
Distribution

Failure Probability
of Disk

Significant
Variable of
Rotating Cavity

Fig. 2 the flowchart of methodology

The variation of boundary conditions (BCs) of the cavity can be denoted by structure parameters in
numerical simulations. Three structure parameters Rc, G and Gc are investigated in this study. They are
formulated by
Rc=rin/R0

(1)

G=S/R0

(2)

Gc=Sb/R0

(3)

where, rin is the inlet position of the cooling air; R0 is the radius of the disk, which is considered as the
reference length of the system; S is distance between the disk and cavity; Sb is the length of the outlet of the
cavity, respectively. These structure parameters governed the performance characteristics of the cavity with
the other parameters such as flow parameters fixed.
The structure parameters is transferred to the loads of the disk by FSI which is an effective tool for the
problems in the cavity where the interaction between fluid analysis and structural analysis is taken into
account. Then the fracture probability of the disk is predicted with the loads of the disk obtained from FSI
simulation as the input data. In order to calculate Pf, a numerical approach, following the flowchart of PRA
as shown in Fig. 3, is established. The approach integrates the kernel failure function analysis with stress
intensity factor as the fracture driven force, Monte Carlo (MC) method with the initial crack length
distribution as the only probabilistic driver. The kernel failure function of PRA is
(4)
g =Kc - K
where Kc, K and g are fracture toughness of the material, stress intensity factor of the crack and fatigue limit
state function respectively.
Pf is defined as the probability of violating the limit state function, which means that the failure occurs
when K equals or exceeds Kc. Kc is one of the material properties and naturally a function of temperature.
It can be obtained by interpolation of the available material property data when the temperature distribution
of the disk is gotten. K is a function according to crack length, crack geometry and loads and can be calculate
by various process, including numerical and theoretical. It is derived from Newmans method [15] in this
study. The crack propagation is assumed to be controlled by Paris Law:

285

Zhang Gong and Zhou YanPei / Procedia Engineering 80 (2014) 282 292

da
dN

C ('K )n

(5)

where, a is the crack length and N the number of load cycles. C and n are constant and index of the Paris
Law. Variable K denotes the difference between the maximal and minimal values of K and it can be
calculated by current crack length and the load of disk obtained from FSI simulation. Eq. (5) can be solved
by using the Runge Kutta method.
Stress
Distribution

Temperature
Distribution

Stress
Intensity
Factor

Crack
Propagation
Rate

Fracture
Toughness

Failure
Analysis

P  K c  K i  d 0

pi

pF

1

1  p
m

is

f i 1

 f (s)ds

Probabilistic
Driver

Initial Crack Length

Distribution

Fig. 3 the flowchart of PRA

The initiation flaw size is chosen as the only one probabilistic driver of MC method to calculate Pf as a
function of flight cycles. When the fatigue limit stat function is established base on previous effort including
FSI simulation, SIF calculation and Kc interpolation, Pf is calculated. This is a probable approach to predict
Pf, on which a computer code is also developed. The approach and computer code can be verified and
validated by the calibration test data which is provided in Advisory Circular (AC) 33.14-1 [16] and AC
33.70-1 [17] as an acceptable means of compliance.
The sensitive analysis used to identify the significance of elements in system is very important in safety
analysis of the cavity since it can identify the dominant variable. The sensitivity of failure probability is
defined as the partial derivative of Pf with respect to the structure parameters. The non-dimensional
sensitivity is also assessed since the magnitude of structure parameters is not same which may result in
confused direct sensitivity. The non-dimensional sensitivity is a sampling based sensitive estimate of Pf
with finite difference method. They are obtained by perturbing the interested parameters and can be
approximated by

SeV

wPf
wV / V

| 0.01u

Pf (V  'l u V )  Pf (V  'l u V )
2'l

(6)

286

Zhang Gong and Zhou YanPei / Procedia Engineering 80 (2014) 282 292

where Se represents the non-dimensional sensitivity of Pf towards the interest parameter; represents one
interest structure parameter; 'l represents the perturbation size which is chosen as 0.5, 1 and 2.5 percentage
of nominal condition. A separate analysis must be performed for the sensitive estimate of each parameter
at the same perturbation size to determine the sensitivity of Pf.
3. Numerical analysis
Present parametric study is carried out to investigate the effect of Rc, G and Gc on the fracture probability
of the disk. The essential details for the numerical simulation are introduced and discussed in this section.
3.1. Model and geometry
The domain of the cavity consists of a stationary fluid part and a rotary solid part. The numerical model
of FSI problems is axisymmetric and only half of the meridian plan since the cavity system is assumed
axisymmetric, as shown in Fig. 1. The radii of the outer and the inner rims of the disk are 320mm and
40mm, respectively. In order to evaluate the effect of structure parameter on the safety of ELLP, the radius
of inflow position of coolant varies from 40 mm to 292 mm with an increment of 28 mm while the distance
between the cavity and the disk varies from 23 mm to 73 mm with an increment of 5 mm and the length of
cavity outflow varies from 0.5 mm to 2.5 mm with an increment of 0.2 mm, respectively. The other details
of the cavity geometry are listed in table 1. The outflow section is extended 40mm from the outer rim due
to the simulation convergence. The number of the mesh is about 20000 since the FSI simulation model of
the cavity changes. The mesh has been validated that the effect of meshing density on the results are
negligible.
3.2. Boundary conditions and material properties
The numerical method used in FSI simulation employs a k- turbulence model which has been validated
successfully to compute the fluid flow and heat transfer in a rotor-stator system. It is a pressure-based,
incompressible and steady solver, while the thermal and stress analysis employs thermo-elastic model to
calculate temperature and stress distribution. The material used in the simulation are assumed isotropic and
homogenous. The material properties used in fluid domain and solid domain are listed in table 2. The
material properties used to calculate SIF and evaluate fracture, including crack propagation rate and fracture
toughness variation subjected to different temperature, are listed in table 3.
BCs of this simulation include the mass flow of the cooling air, the rotation speed of the disk and the
heat flux on the outer rim of the disk, etc., which are listed in table 4. The external load to simulate blade
loading for structural analysis is about 50 MPa, which is applied to the outer rim. Hoop stress is the
maximum principle in this study. It is assumed that the crack grows under constant amplitude cyclic
loadings, also known as simple loading cycle, until it failed. More than six million samples are used in
Monte Carlo simulation to get a highly accurate result. Fig. 5 shows the exceedance curve of the initial
crack length distribution as the random variable of PRA, which is obtained from industry defined data [16].
In addition, the aspect ratio of the crack is assumed to be 1.
Table 1 Geometry parameter of simulation model
R0/mm

r0/mm

b/mm

/mm

320

40

30

Depend on rin

287

Zhang Gong and Zhou YanPei / Procedia Engineering 80 (2014) 282 292
Table 2 Material properties
Material properties
Density /kg m-3

Fluid Domain

1.173

Thermal conductivity f/W m-1 K-1

-1

0.05589

-1

Specific heat Cp/J kg K

1095

Dynamic viscosity /10-5 kg m-1 s-1

3.73

-3

Solid Domain

Density /kg m

8240

Thermal conductivity s/W m-1 K-1

-1

21

-1

Specific heat Cp/J kg K

580

Youngs modulus E/109 Pa

161

Poissons ratio

0.3

Coefficient of thermal expansion /10-6 K-1

14.6

Table 3 Crack propagation rate and fracture toughness

Paris index n

Fracture toughness Kc/106 /Pa m

5.19

103.5

89

500

7.0410

-12

4.86

87

600

2.1510-10

3.98

83

650

69.5

Temperature T /

Paris constant C

20

9.0810

300

-13

Table 4 Boundary conditions for the simulation

Boundary type

BCs details

Inlet flow velocity /m s

-1

70

Temperature of inlet coolant /K

780

Pre-swirl angle of inlet coolant /

Rotation speed /rad s-1

1047.2

Outlet pressure /Pa

202625

Heat flux on the rim/J m-2 s-1

20000

The wall of stationary casing

No slip, adiabatic

The leeside of the disk

No axis displacement, adiabatic

Exceedance probability

0.1

0.01

1000

10000

Defect area/Mils2

Fig. 5 Initial crack size distribution

288

Zhang Gong and Zhou YanPei / Procedia Engineering 80 (2014) 282 292

4. Results and discussion

4.1. Effect of structural parameter
The numerical results of simulation are shown in Figs. 6-9. Fig. 6 illustrates the typical failure probability
of the disk as a function of load cycles. Pf of the disk subjected to nominal condition at 20,000 cycles is
3.2110-11 which is much less than 10-8, therefore it meets the safety requirement of the airworthiness
regulations since the load of the disk at the simulation conditions is less severe than real operating
conditions of aero-engines. It is implied Pf will violate the safety requirement if the load cycles continue to
increase since it increase monotonically with the increasing of load cycles.
Pf at 20,000 cycles decreases from 3.2310-11 to 3.21110-11 with Rc rising from 0.125 to .09125, as
illustrated in Fig. 7. It is observed that the trend of Pf with Rc is different from the trend with G and Gc. Pf
increases from 3.2010-11 to 3.2210-11 then decreases with G increasing from 0.072 to 0.22, and increases
from 3.2210-11 to 3.2310-11 with Gc increasing from 1.5610-3 to 7.8110-3, as shown in Fig. 8 and Fig.
9, respectively. It is illustrated that structure parameters have certain but not significant impact on the safety
of ELLP.
-11

2.0x10

-11

1.0x10

-11

Pf

Pf

3.0x10

3.23x10

-11

3.22x10

-11

3.21x10

-11

0.0
0

5000

10000

15000

0.0

20000

0.3

0.6

Fig. 6. Failure probability of the disk

3.22x10

-11

3.21x10

-11

3.20x10

-11

Fig. 7. Failure Probability with increasing Rc

3.23x10

-11

3.22x10

-11

Pf

-11

Pf

3.23x10

0.9

Rc

Cycles/N

0.07

0.14

0.21

Fig. 8. Failure probability with increasing G

0.002

0.004

0.006

0.008

Gc

Fig. 9. Failure probability with increasing Gc

The results of FSI simulation are shown in the Fig. 10. It is indicated that the flow structure changes
greatly with Rc increasing while Gc and G has not so significant effect on the flow structure but they all

289

Zhang Gong and Zhou YanPei / Procedia Engineering 80 (2014) 282 292

influence the distribution and numerical value of temperature and stress. As a result of FSI simulation, the
max hoop stress of the disk and the temperature at the location with maximum stress related to different
structure parameters are derived and illustrated in Fig. 11 and 12 respectively. It should be pointed out that
the structure parameters have no effect on the centrifugal stress since it only depend on the rotation speed
of the disk. The structure parameters have certain impact on safety of ELLP since the variation of structure
parameters changes the flow structure in the cavity and as a result, the temperature and thermal stress
distribution of the disk. The hoop stress, the principle stress of the disk in this study, is affected by the
thermal stress, which result in the variation of stress intensity factor. The crack propagation rate and fracture
toughness of the material are affected by the elevated temperature as well. The different flow structure of
the cavity induced by variation of structural parameters result in vary stress and temperature distributions
on the disk as well as the max hoop stress and elevated temperature at the location with max stress as
illustrated in Fig.11 and 12. Therefore, Pf changes as mentioned above.

Rc

Gc
Fig. 10 Temperature and stress distribution variation with structure parameters

Gc

Temperature at max stress/K

809.6

Gc

809.4
6

Rc

809.8

Max hoop stress/10 Pa

Rc

782.0

810.0

809.2
809.0
808.8
808.6
808.4
808.2

781.5

781.0

780.5

780.0

808.0
0

10

Conditions

12

10

12

Conditions

Fig. 11. Max hoop stress of the disk

Fig. 12. Temperature at the location with max stress

290

Zhang Gong and Zhou YanPei / Procedia Engineering 80 (2014) 282 292

4.2. Sensitive analysis

Table 5 shows the sensitivity of Pf towards the structure parameters obtained from the results of
numerical simulation. The different signs depict the different characteristics of the system responding to
the structure parameters of the cavity. A positive sign indicates that Pf increases with the increasing of the
parameter, and conversely, a negative sign indicates that Pf decreases with the increasing of the parameter.
Actually, sensitive analysis is more concerned with the absolute value rather than the sign. SRc is -5.01014
when Rc is 0.125, which means a unit augment of Rc results in 5.010-14 decrease of Pf, and drops to 3.210-13 when Rc is 0.9125. Gc seems to be the dominant variable among the structure parameters in the
cavity since the sensitivity absolute value of Pf with respect to Gc is about one order of magnitude more
than Rc and G, as shown in table 5. The results is confusing since the magnitude of Gc is much less than Rc
and G. Hence it is necessary to identify the dominant variable by non-dimensional sensitivity estimates.
Table 5 The results of sensitivity analysis
Structure parameter

Value

Sensitivity of Pf

Rc

0.125

-5.010-14

0.5625

-2.310-13

0.9125

-3.210-13

0.072

2.910-12

0.15

1.610-12

0.23

-6.410-13

Gc

1.56

1.710-11

10-3

4.69

9.910-12

7.81

1.810-11

The non-dimensional sensitivity of Pf with regard to the perturbed structure parameters at the nominal
condition is shown in Fig.13. The perturbation size is chosen as 0.5, 1, 2.5 percentage of the nominal
condition respectively. The results indicate that the perturbation size has a significant impact on the nondimensional sensitivity estimates since the cavity system, especially flow structure, responds differently to
the structure parameter when they changes. The dominant variable in the cavity is Rc when l is 0.5% and
1%, followed by G and Gc. When the perturbation size increases to 2.5%, G becomes the dominant variable
in the cavity. The results also reveal that engine is one system which functional connection with other
subsystems and the effect of the structure parameters in the cavity is quite significant, which needs careful
consideration in the design and certification process of ELLP life prediction.
1.0x10

-14

Rc
0.0

Se

G
-1.0x10

-14

-2.0x10

-14

-3.0x10

-14

Fig. 13. Non-dimensional sensitivity of structure parameters

Gc

'l=0.5%
'l=1%
'l=2.5%

Zhang Gong and Zhou YanPei / Procedia Engineering 80 (2014) 282 292

5. Conclusions
A numerical simulation was performed for evaluating the safety of a cavity. The fracture probability of
the disk with respect to the structure parameters in the cavity are investigated. The dominant variable of the
system is also discussed from the numerical results obtained by a non-dimensional sensitivity estimates.
An approach for the safety assessment of the cavity is established by integrating one way FSI and PRA
method. Based on the approach, a computer code integrating SIF calculator, Runge-Kutta method and
Monte Carlo simulation, is also developed to predict the fracture probability of the disk with initial
anomalies.
The numerical results indicate that the structure parameters in the cavity have certain but not significant
impact on Pf. The effect of structure parameters are different since they have different impacts on the flow
structure and as a result, the stress and the temperature distribution of the disk. The results of the sensitivity
estimates reveal that it depends on the perturbation size, thus the dominant variable changes with the
variation of the perturbation size. Rc is the dominant variable to estimate the risk of the cavity when the
perturbation size is 0.5% and 1%. The results also demonstrate that the engine is a system which functional
connection with other subsystems and suggest that the cavity needs careful consideration in the design and
certification process of ELLP life prediction.
References
[1] U.S. Department of Transportation FAA. e-CFR 14 part 33 Airworthiness Standards: Aircraft engines. Washington D.C.: FAA,
2009.
[2] European Aviation Safety Agency. Certification Specifications for Engines. Cologne: EASA, 2010.
[3] Farthing P.R., Long C.A., Owen J.M., et tl. Rotating Cavity with Axial Throughflow of Cooling Air: Heat Transfer. Journal
of Turbomachinery. 1992,114(1): 229-236.
[4] Farthing P.R., Long C.A., Owen J.M. et al. Rotating Cavity with Axial Throughflow of Cooling Air: Flow Structure. Journal
of Turbomachinery. 1992, 114(1): 237-246.
[5] Karabay H., Pilbrow R., Wilson M., et at. Performance of Pre-Swirl Rotating-Disc Systems. Journal of Engineering for Gas
Turbines and Power, 2000, 122(3): 442-450.
[6]Karabay H., Wilson M., Owen J. M. Predictions of Effect of Swirl on Flow and Heat Transfer in a Rotating Cavity. International
Journal of Heat and Fluid Flow, 2001, 22(2): 143- 155.
[7] Owen J M. Modelling internal air systems in gas turbine engines. Journal of Aero-space Power, 2007, 22(4):505-520.
[8] Melis M.E., Zaretsky E.V., August R. Probabilistic analysis of aircraft gas turbine disk life and reliability. Journal of Propulsion
and Power, 1999, 15(5): 658-666.
[9] Leverant G.R., Littlefield D.L., McClung R.C., et al. A Probabilistic Approach to Aircraft Turbine Rotor Material Design.
ASME International Gas Turbine and Aero-engine Congress, Paper 97-GT-22, 1997.
[10] Leverant G.R., Millwater H.R., McClung R.C., et al. A new tool for design and certification of aircraft turbine rotors. Journal
of Engineering for Gas Turbines and Power, 2004, 126: 155-159.
[11] Enright M.P., McClung R.C., Huyse L. A probabilistic framework for risk prediction of gas turbine engine components with
inherent or induced material anomalies. ASME paper GT2005-68982, 2005.
[12] Enright M.P., Huyse L., McClung R.C. Fracture mechanics- based probabilistic life prediction of components with large
numbers of inherent material anomalies. Proceedings 9th international conference on structural safety and reliability (ICOSSAR),
2005: 601-607.
[13] Enright M.P., Huyse L. Methodology for probabilistic life prediction of multiple-anomaly materials. AIAA Journal, 2006,
44(4): 787-793.

291

292

Zhang Gong and Zhou YanPei / Procedia Engineering 80 (2014) 282 292

[14] Millwater H.R, Enright M.P, Fitch S.H.K. A convergent probabilistic technique for risk assessment of gas turbine disks
subject to metallurgical defects. AIAA-2002-1382, 2002.
[15] Newman J.C., Raju I.S., Stress-intensity factor for cracks in three-dimensional finite bodies. NASA TM 83200, 1981
[16] U.S. Department of Transportation FAA. Advisory circular 33.14-1: Damage tolerance for high energy turbine engine rotors.
Washington D.C.: FAA, AC33.14-1, 2001.
[17] U.S. Department of Transportation FAA. Advisory circular 33.70-1: Guidance material for aircraft engine-life-limited parts
requirements. Washington D.C.: FAA, AC33.70-1, 2009.