Failure Analysis On HPT Blade of CFM56-7B: Gao Yu, Chen Xin-Feng and Li Nan

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

2nd International Conference on Electronic & Mechanical Engineering and Information Technology (EMEIT-2012)

Failure Analysis on HPT Blade of CFM56-7B

Gao Yu1, a, Chen Xin-feng2,b and Li Nan3,c


1
M&E Division, China Academy of Civil Aviation Science and Technology, Beijing, 100028, China.
2
M&E Division, China Academy of Civil Aviation Science and Technology, Beijing, 100028, China.
3
R&D Division, China Academy of Civil Aviation Science and Technology, Beijing, 100028, China.
a
[email protected], [email protected], [email protected]

Keywords: Transition zone. Edge of contact. Electronic speculum. Weibull distribution. Parameter
estimation. Finite element method. Johnson median rank time method. Bayes estimation.

Abstract. The high pressure turbine blade of CFM56-7B engine is risky to crack, it’s failure modes
include transition zone crack, shank rupture, min-neck damage, edge of contact and split-shelf clog,
which could be modeled and simulated by finite element method. The root cause of crack is found by
aforementioned simulation and electronic speculum to modify the blade design. And the reliability of
blades could be calculated by Weibull distribution to get soft-time limitation in replace of related
blades, in which the Weibull parameter would be estimated with Johnson median rank time method
and Bayes estimation method for the censored data.

Introduction
Recent years, CFM56 engine has a great market share in China, it reach 56% in domestic engine
market till 2011. It is necessary to study the reliability of engines. CFM56-7B is used widely among
CFM56 series, statistical study on CFM56-7B operational data shows that HPT blade is damageable.

Analysis algorithm
The operational data which come from airlines would be used to calculate the reliability degree
with Weibull distribution method. The estimation method such as Johnson median rank time method
and Bayes estimation [1] should be applied to estimate parameters for the insufficient censored data.
In the risk assessment, severity degree and relatively failure rate would be statistical analyzed to
calculate risk matrix, the result as Table 1, the most risky failure is crack of HPT blade.
Table1 Risk management of CFM56-7B
failure mode measure level
3# bearing failure Improve production process medium
Fuel filter bypass light on Periodic clean the fuel tank medium
HPT blade crack Replace blades by SB 72-696 high
HMU failure Improve design and real-time monitoring high
Weibull distribution could be used to discribe the reliability module of parts which is conform to
bath curve. For Weibull distribution function[2],
t
F (t ) = 1 − exp[ −( ) β ]
a (1)
For the insufficient censored data, Weibull parameter such as shape parameter and scale parameter
could be estimated with Johnson median rank time method and Bayes estimation method.
Johnson median rank time method. Assuming the failure time sequence changes for the dropout
parts, the median sequence number could be calculated for the failure parts, the median sequence
number of No. r is J r = J r −1 + I r , I i = (n + 1 − J r −1 ) / (n + 2 − i ) .Assuming J0=0, the sample failure
probability of failure parts by the median rank time method is,
Fn ( xr ) = ( J r − 0.3) / (n + 0.4) . (2)

Published by Atlantis Press, Paris, France.


© the authors
2116
2nd International Conference on Electronic & Mechanical Engineering and Information Technology (EMEIT-2012)

The sample distribution function could be fitted by nonlinear least squares curve to estimate the
parameter P=[shape parameter, scale parameter].
Bayes estimation[3].For bi-parameter index Weibull distribution EW (a,θ ) ,
a = a0 , x a0 = t
,
Weibull distribution function is,
 F (t ) = (1 − e ) , t > 0, θ > 0
−t θ

 − t θ −1
. (3)
 f (t ) = θ e (1 − e )
−t

Assuming there is r parts in n (1≤r≤n),the joint density function for t(1)≤t(2)≤ ≤t(r) is,
n! r
p (t θ ) = ∏ θ e−ti (1 −e−ti )θ −1.[1 − (1 − e−tr )θ ]n−r
(n − r )! i =1 . (4)
Bayes estimation for θ with scale square loss function is,
(a + r − k + 1) j =0 C j n − r (−1) j (b − T − jR) a + r − k
n−r


n−r
C j n − r (−1) j (b − T − jR) a + r − k +1
j =0
. (5)
Then estimate the parameter P=[shape parameter, scale parameter].

Failure modes of blade


The blade of CFM56-7B has been modified several times, it may failure in four modes such as
transition zone crack, shank liberation, min-neck wear and edge of contact(EOC).
Analysis on transition zone crack.
Crack mostly happen in the transition zone of HPT blade under high temperature and pressure. For
low ductile-brittle casting design and thick inner coating, the local stress and heat stress is high
around the cooling passage of blade. The OEM such as PCC and Howmet differ much from each
other on cooling passage design and positioning core ball chute of tolerance design.
Make Weibull distribution, it’s found that more transition zone crack appeared on PCC blade than
Howmet. So the thin coating to the inner wall would benefit. Calculating with insufficient censored
data and found P=[5, 15794] by Johnson median rank time method, P=[ 5.44, 15033] by Bayes
estimation. After getting the failure distribution function with Eq. 1, the reliability degree function
β
could be calculated by exp[−(t / a) ] .The results of Table 2 show that Howmet blade is reliable at
11000 cycles. In the same way, PCC blade is reliable at 14000 cycles.
Table 2 Reliability of Howmet blade
Cycle since new 10000 11000 12000 13000 14000
Johnson median rank time method 0.903 0.849 0.776 0.685 0.579
Bayes estimation method 0.897 0.833 0.746 0.635 0.507
Shank rupture
If crack of transition zone is deep enough, shank is easily to rupture ,which would lead to IFSD.
It’s tailored that blade between 11000 and 17800 cycle is easily to rupture due to tensile overload. It’s
metallographic analyzed that shank is sensitive to rupture if crack beside the cooling passage have
propagated into matrix intergranularly. Surface corrosion is a key point to make crack.
In one case of electronic speculum on blade as Fig.1,2, it’s found that fatigue crack appear at No.3
cooling passage which is regarded as the origin of crack. The arrow from the figure show the crystal
route on crack propagation before overload[4]. Sulfide is found on the route by eddy current method
which could lead to corrosion on the passage wall and make pit under concentrated stress.

Fig.1 electronic speculum on the shank Fig.2 electronic speculum on TZ crack

Published by Atlantis Press, Paris, France.


© the authors
2117
2nd International Conference on Electronic & Mechanical Engineering and Information Technology (EMEIT-2012)

Blade crack is often found in most engine OEMs, sulfide would make crack to some extent[5],
transition zone and inner cooling passage is most susceptible to corrosion. It’s not suggest to repair
when blade corroded, refer to anti-corrosion technology from other OEM, one method is platinum
plating at surface , aluminum coating into the transition zone, then platinum plating to the dovetail,.
Another way is to change the casting, examine if the coating abrade the groove in assembly process.
Calculate the reliability degree by the parameter estimation method, it’s found that P=[8,15830] by
the Johnson method, P=[5.37,16999] by Bayes method, as Table 3 said, the blade is reliable before
13000 cycles.
Table 3 Reliability of blade
Cycle 12000 13000 14000 15000 16000
Johnson method 0.897 0.813 0.688 0.522 0.337
Bayes estimation 0.857 0.789 0.703 0.600 0.486
Contact damage
Blade is easily to crack for high contact stress such as min-neck damage and EOC, which usually
lead to dovetail fracture and could not be detected by borescope. It’s analyzed by finite element
method that the contact stress is high at min-neck when dovetail contacting to the groove ,it usually
be fatigue damaged at the influence of heat stress and centrifugal effect.
Min-neck damage. Modeling to blade and make stress analysis by finite element method, which
found that stress at min-neck is high enough to crack, as Fig.3. Actually it usually take place below
the min-neck of No.1 cooling passage at cycle between 11000 and 17000, that verify the method.
EOC.EOC usually caused by high shear stress of dovetail, or little gap between blade and disk . If
the hours circulation ratio is low, throttle is frequently activated, EOC usually lead to low cycle
fatigue.SB is issued by OEM to increase the eddy current test for blade. Modeling to blade and make
contact stress analysis, as Fig.4.It found that the stress at contact area is high enough to crack[6].
Actually, EOC usually take place at upper pressure area of No.2 cooling passage at cycle between
11700 and 17300, that verify the method.

Fig.3 Modeling anaylsis on min-neck Fig.4 Modeling analysis on EOC


For the insufficient data, Johnson method is used to make Weibull distribution, which found
P=[6,15671]. So it suggest that blade should be detected by eddy current method at 12000 cycle.
Split-shelf clog
Some blades have split-shelf at cooling passage of trailing edge, it work as controlling the cooling
air in the inner blade. It does works but dust is easily clogged blow the split-shelf which cause
overheat locally. Finally, heat stress make crack and blade scrap. The location of crack usually take
place at 25% or 50% of airfoil. Modeling to cooling passage and make heat stress analysis, as Fig.5,6.
It’s found that the maximum heat stress is located at 25% of trailing edge of airfoil. If no split-shelf
exit, the heat stress is distributed equably in the passage. So no split-shelf is suggested.

Fig.5 Pre-mod of split-shelf Fig.6 Post-mod of split-shelf

Published by Atlantis Press, Paris, France.


© the authors
2118
2nd International Conference on Electronic & Mechanical Engineering and Information Technology (EMEIT-2012)

Research to reliability of blade


Air pollution is a key factor to the blade damage from the study above, make Weibull distribution
by Johnson method based on the data from all related damage modes, as Fig.7. It found that
P=[7,16346] and the blades beyond 20000 cycle should not be used anymore, so soft-time
replacement is suggested to control the blade configuration.

Fig.7 Weibull distribution on blade corrosion

Precaution
The HPT blade of CFM56-7B work under high pressure and temperature, it easily crack for high
or low cycle fatigue. The failure mode include transition zone crack, min-neck contact, EOC and
shank liberation. Although crack upon platform could be detected by borescope regularly, the crack
below the platform could not be detected by NDT method, the soft-time control is the only effective
way.OEM suggest the soft-time based on world fleet data, but the domestic data is more accurate for
Chinese fleet. It’s suggested that blade beyond 14000 cycle is not worth repairing. For blade from
PCC,16000 cycle is a threshold to change the blade, but domestic status is lower than that, about
14000 cycle. For Howmet blade, 20000 cycle is a threshold to change the blade, but domestic
threshold is 11000 cycle.
The design of blade is still under development, it’s suggested that new thin coating should be used
in transition zone to avoid oxidation corrosion. New casting should be applied to strengthen cooling
effect. The design of min-neck should be improved to increase contact stress, such as film coating in
dovetail. For airfoil upon platform, over-grinding is inhibit. Split-shelf should be improved to reduce
the dust accumulation inner the cooling passage. It should strengthen the configuration management
and performance monitoring of blades.

Reference
[1] FU Hui-min , YUE Xiao-rui: Multiple regression analysis method for mixed data. JOURNAL OF
AEROSPACE POWER, 26 (2011).
[2] Xiong Kai-ping: Parameter stimation of Weibull distribution for incomplete samples. Huazhong
University of Science and Technology ,(2006).
[3] Li Feng, Shi Yi-min,Jing Yuan: Bayesian estimation for environmental factor of two parameter of
Weibull distribution. Systems Engineering and Electronics, 30(2008).
[4] Tang Hai-jun: Investigation and Analysis on an In-Flight Shutdown of a CFM56-7B Engine.
Aviation Maintenance & Engineering, 6(2011).
[5] Yan Xiao-jun,Sun Rui-jie,Deng Ying: Experimental study on fatigue curve law of turbine blade
under combined high and low cycle loading. Journal of Aerospace Power, 26(2011).
[6] Peng Mao-lin,Yang Zi-chun: Analysis on the influences of the contact gap on the contact response
of turbine disc-blade structures. Journal of Aerospace Power, 26(2011).

Published by Atlantis Press, Paris, France.


© the authors
2119

You might also like