Fatigue Crack Growth in P91 Plates OMMI (Vol.

3, Issue 2) August 2004

www.ommi.co.uk


FATIGUE CRACK GROWTH IN LARGE CRACKED PLATES OF MARTENSITIC
P91 STEEL AT 550C

M Reytier, CEA Saclay, DEN/DM2S/SEMT/LISN, 91191 Gif-sur-Yvette, France



Abstract

In order to improve the high temperature defect assessment methods on the martensitic P91
steel (Z10 CDVNb 9.1), a new experimental programme on large scale components is
conducted at CEA Saclay. This study consists of fatigue tests at 550C on large plates
containing a semi elliptical surface crack subjected to cyclic bending, and also of standard
fatigue crack growth tests on CT specimens. By using either different beach marking and PD
measurements, or finite element calculations of the compliance, the crack propagation is
followed through the plate during the whole test. Furthermore, the stress intensity factors are
calculated for the deepest point and the surface point of the crack front according to the
French annex A16 handbook. This provides values that are compared with the application of
the Paris law determined from the CT specimen tests. The use of this Paris law for a semi
elliptical defect in P91 steel at 550C is then discussed.


Introduction

High temperature defect assessment procedures under fatigue, creep and creep-fatigue
loadings have been developed for the fast reactors and have been mainly validated on
austenitic stainless steels like 316L(N) [A16, Drubay 2003]. Concerning the GCR,
complementary studies are conducted in order to validate the existing methods and to get new
experimental data on 9%Cr martensitic steels. Moreover, if the geometry and the loading of a
standard CT specimen allow to propose a 2D analysis, the case of industrial loadings appears
much more complicated, notably because of surface defects which propagate and present
shapes that can be considered as half ellipse. Therefore, in the frame of the defect assessment
methods validation, the SEMT/LISN performs both standard tests on CT specimens to
determine the propagation laws and bending tests on large plates under high temperature
fatigue, creep or creep-fatigue loadings. These components present an initial semi elliptical
surface notch normal to the loading direction, and its initiation and propagation are studied.
Here, only the tests under fatigue are presented.


Experimental Procedure

Material

The material of the study is Z10CDVNb 9.1 martensitic steel whose chemical composition is
shown in Table 1. The sheet has been heat treated (1050C, 30min and then air cooled) and
then tempered (780C for one hour). Its thickness is 30 mm and each sample has been cut in
order to be loaded in the transverse direction and to make the crack propagate in the rolling
direction.
Fatigue Crack Growth in P91 Plates OMMI (Vol.3, Issue 2) August 2004

2

Elements C Mn P S Si Cu Ni Cr Mo Al
% wt 0.086 0.363 0.017 0.0010 0.324 0.068 0.149 8.910 0.917 0.018

Elements Nb V N
2
Sn Zr Ti As W B Sb O
2

% wt 0.080 0.198 0.0411 0.005 0.001 0.002 0.0122 0.010 0.0010 0.0006 0.0018
Table 1: Chemical composition of the steel sheet


Fatigue on CT specimens

The 25mm thick specimens are first fatigue pre-cracked at room temperature with a
decreasing K to a crack length of 27.5 mm (a/w=0.55). Then, they are side-grooved and
their net thickness is 20mm. Fatigue tests on these standard CT specimens are performed
according to ASTM E647 at 550C, with a frequency of 0.1Hz and a load ratio R=0.1 in order
to determine the Paris law. The set of tests is summarized in Table 2.

Specimens F (kN) a
precrack
(mm) a
final
(mm)
GBU4G 10 28.1 35.4
GBU4J 10 27.7 35.6
GBU4D 8 28.0 35.7
GBU4M 6 28.1 35.4
Table 2: Summary of the fatigue tests on CT specimens (R=0.1 and f=0.1Hz)

The crack length is followed by the compliance method, which is detailed here after.

The opening displacement at the load-line can be expressed [Tada 2000] by equation (1) in
which P is the load for 1mm thickness and E* = E / (1-
2
) with E = Young modulus and is
the Poisson coefficient.

) ) / .( 9314 . 9 ) / .( 609 . 20 ) / .( 9925 . 0 ) / .( 065 . 20 / . 219 . 12 1630 . 2 [
)
/ 1
/ 1
( ) / (
) 1 ( ) / (
5 4 3 2
2
1
1
*
w a w a w a w a w a
w a
w a
w a f
Eq w a f
E
P
+ +

+
=
=

Considering that the specimens present a thickness of (B.B
n
)

, it follows that:
n
B B E
w a f
F P
.
) / (
*
1
exp
exp
= =

Eq(2)

Then, for each recorded cycle, the compliance (
exp
/ F
exp
) is calculated, and a/w is determined
by solving equation (2). The obtained evolution of a/w for one specific specimen
(no.GBU4G) is presented in Figure 1 and compared with the measurements on the fracture
surface. The crack evolution is fitted by a parabola.




Fatigue Crack Growth in P91 Plates OMMI (Vol.3, Issue 2) August 2004

3
y = 1.26E-08x
2
+ 1.72E-04x + 2.86E+01
R
2
= 9.98E-01
25
27
29
31
33
35
37
0 5000 10000 15000 20000
number of cycles N
c
r
a
c
k

l
e
n
g
t
h

a

(
m
m
)
precrack measurement
end of test measurement


Figure 1: Example of crack length determination for specimen GBU4G and comparison
with measurements

Then the crack propagation rate is determined either by the tangent to the parabola or by
calculating the mean growth rate during the i
th
recorded cycle according to equation (3). It can
be noted in Figure 2 that both methods give comparable results.

1 1
1 1
) (
+
+

=
i i
i i
i
N N
a a
dN
da
Eq (3)

Then, the stress intensity factor is determined according to the ASTM E399-90 with equations
(4) and (5).
) / ( .
. .
2
W a f
B B W
F
K
n

= Eq (4)
2 / 3
4 3 2
2
) / 1 (
) ) / ( 6 . 5 ) / ( 72 . 14 ) / ( 32 . 13 ) / ( 64 . 4 886 . 0 )( / 2 (
) / (
w a
w a w a w a w a w a
w a f

+ + +
= Eq(5)



1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
0 5000 10000 15000 20000
number of cycles N
p
r
o
p
a
g
a
t
i
o
n

s
p
e
e
d

d
a
/
d
N
mean
tangent


Figure 2: Example of crack growth rate determination by the two methods for
specimen GBU4G

Fatigue Crack Growth in P91 Plates OMMI (Vol.3, Issue 2) August 2004

4
In order to ensure that the ligament length is sufficient to stay in a quasi-elastic domain all
through the test, a few points at the end of the test are rejected. The valid points must verify
equation (6) according to the ASTM E647-99. Finally, the Paris law obtained from four
different fatigue tests at 550C and K of 14 to 34 MPa.m
1/2
is given in equation 7 and
Figure 3.

2
2 . 0
max
) (
4
) (
Rp
K
a W

Eq (6)

) . ( ) / (
10 . 1 , 7
85 . 1 7
m MPa cy mm
K
dN
da
=

Eq (7)


y = 7.10E-07x
1. 85E+00
R
2
= 8.68E-01
1.00E-05
1.00E-04
1.00E-03
1.00E-02
1 10 100
delta K (MPa.m0.5)
d
a
/
d
N

(
m
m
/
c
y
)

Figure 3: Paris law for the set of tests


Fatigue tests on P91 large plates

These tests on structures deal with large plates (590 mm 440 mm 30 mm) whose central
width is reduced to 350 mm. A semi elliptical notch is machined in the plate centre. The notch
dimensions (depth (a) = 5 mm and width (2c) = 20 mm) have been chosen in order to be
isoK: this means that, for our bending loading, the notch presents the same stress intensity
factor K at depth and near the surface. First this plate is fatigue pre-cracked at room
temperature with constant F (Fmax=30 kN and R=0.1) and a frequency of 0.5Hz. A crack
propagation of around 1mm all along the notch is obtained.

Then, the plate is submitted to cyclic bending at 550C - 60000 cycles are performed. The
bending loading is obtained using two bending arms of 352.5 mm length (Figure 4). The
heating is provided by a radiant furnace fixed on the specimen in order to follow its
displacements during the test and to obtain a very homogeneous temperature (3C) on the
central zone of the plate. The test is performed by imposing a constant F (Fmax=47.1 kN
and R=0.1).

Fatigue Crack Growth in P91 Plates OMMI (Vol.3, Issue 2) August 2004

5

Deflection and CMOA diagram

f f f f

Figure 4: Principle of the fatigue test on large plates

Periodically beach-markings are realised under fatigue with a different R ratio in order to
temporarily modify the propagation conditions. These are then visible to the naked eye on the
rupture surface and constitute reference marks through the test to determine the crack
propagation. Finally fatigue at room temperature is imposed to finish the crack propagation.

Instrumentation is made up of 17 thermocouples (type K) to ensure a homogeneous
temperature distribution. Moreover, the displacement and the load produced by the press are
recorded. The arms rotation and the plate deflection are followed to correct the moment
effectively applied. Two laser micrometers are used to measure the crack opening
displacement (CMOD) and the crack opening angle (=CMOA). Finally, the crack
propagation is also monitored using the potential drop method (PD) or the compliance
method. More details on the instrumentation can be found in [Curtit 1999].


Numerical Calculations


Y
Z
GIBIFECIT

Figure 5: Example of mesh used to interpret the tests on cracked plates
Fatigue Crack Growth in P91 Plates OMMI (Vol.3, Issue 2) August 2004

6

In order to interpret these tests on structures, finite element calculations are necessary. They
allow to determine the crack propagation by using the compliance, but also to calculate the
global parameters of fracture mechanics K, J and C*. In this case of fatigue test, the aim of
this calculation is more particularly to build the compliance function linking, for a bending
moment, the opening angle (CMOA) to the propagation which is described by the couple
(a, c) or (a, c/a). 21 finite element calculations are performed by considering 7 crack depths
from 6.5 to 18 mm and 3 ratios c/a equal to 1.8 or 2 or 2.2. These numerical analyses are
based on 3D models, representing a quarter of the plate in the zone where the temperature is
homogeneous. The semi elliptical front is meshed (Figure 5) in order to simulate accurately
the experimental defects and the limit conditions allow to impose a pure bending moment. All
of the analyses are realised with massive quadratic elements by using the code CAST3M
[CAST3M]. The mechanical behaviour is supposed to be elastic with a Youngs modulus of
160GPa and a Poisson coefficient of 0.3.


Principle of Analysis and Results

The main difficulty of the analysis is to follow the evolution of the crack dimensions
throughout the test. The first part of the analysis consists of observing the fracture surface
(Figure 6). The initial semi elliptical crack keeps its shape during the test, and its length/depth
ratio (c/a) is between 1.8 and 2.2.












Figure 6: Fracture surface

Then, for the known points, the evolution of the defect dimensions can be built as a function
of potential drop or of the experimental CMOA, or of the numerical compliance. These
characteristic points are the end of pre-cracking in green in Figure 6, the 3 beach-markings in
blue (realised after 15000 cycles, 25000 cycles and 43000 cycles) and the final defect in red.
Either a linear correlation (PD1) or quadratic or cubic polynomial correlations (PD2 and PD3)
can be used with the potential drop (Figure 7) and a linear correlation is chosen to describe the
link between the propagation and the experimental cmoa.

The second technique used to estimate the crack dimensions from the cmoa measurement is to
build numerically the function linking the opening angle cmoa to the couple (a,c) or more
classically to the couple (a, c/a). The results are presented in Figure 8. It can be considered
that the opening angle varies linearly with the crack depth (a) and that the result is not very
dependent on c/a ratio between 1.8 and 2.2. Thanks to these calculations, for an experimental
measurement of cmoa, the crack depth (a) and the crack length (2c) can be determined. The

precrack
End of the test
3 Beach markings
Fatigue Crack Growth in P91 Plates OMMI (Vol.3, Issue 2) August 2004

7
results of these five analyses are presented in Figure 9. All of these methods give similar
results in very good agreement with the experimental data. This point constitutes an essential
validation of all the involved measurements and of the many successive analyses.


linear correlation
0
5
10
15
20
0 500 1000
DDP (mV)
a

(
m
m
)
0
10
20
30
40
50
c

(
m
m
)
beach marking
pre crack


quadratic correlation
0
5
10
15
20
0 500 1000
DDP (mV)
a

(
m
m
)
0
10
20
30
40
50
c

(
m
m
)
beach marking
pre crack


cubic correlation
0
5
10
15
20
0 500 1000
DDP (mV)
a

(
m
m
)
0
10
20
30
40
50
c

(
m
m
)
beach marking
precrack


linear correlation
0
5
10
15
20
0 0.2 0.4 0.6
delta cmoa
a

(
m
m
)
0
10
20
30
40
50
60
c

(
m
m
)
beach marking
precrack

Figure 7: Experimental correlations to estimate the crack dimensions

Then, the second step of the analysis is to determine the propagation speed for the crack depth
(a) and for the crack length at the surface (2c). As for the CT specimens, the mean crack
growth rate during the i
th
cycle of this plate test is calculated according to equation 3.

The third part of the analysis deals with the determination of the global parameter K. This has
been done by using a specific K
I
handbook [Chapuliot 2000].

Finally, a diagram showing the crack growth rate as a function of K is presented in
Figure 10. By using the cmoa measurements, the results are very scattered, whereas by using
the potential drop method, it is shown that the propagation speed is two times faster at the
surface than in depth. This could question the application of the Paris law, because for the
same K the crack growth rates are not the same. Moreover, using high exponents (2 or 3) for
the potential drop correlations leads to a slowdown in Figure 10, which is not seen by the
cmoa methods. This could be due to the effects of an inflexion point of the chosen functions,
Fatigue Crack Growth in P91 Plates OMMI (Vol.3, Issue 2) August 2004

8
to a relative incertitude of the beach-marking positions or to a real slowdown. On the other
hand, one other reason for the gap between the Paris law and the results in depth could be a
wrong estimation of K in depth because no account has been taken of any closure effect.
This would reduce the K and move the experimental points closer to the Paris law.
Moreover, the Paris law can be applied correctly at the surface, where no closure effect can
occur and where the bending stress is maximum and always positive. This emphasizes the
idea that if the local loading is correctly estimated, the Paris law is able to describe the crack
propagation in the plate.

0.15
0.2
0.25
0.3
0.35
0.4
0.45
5 10 15 20
a
n
u
m
e
r
i
c
a
l

C
M
O
A
c/a=2.2
c/a=2
c/a=1.8

Figure 8: Building of the compliance function for this geometry





0
5
10
15
20
25
30
35
40
45
0 50000 100000
Number of cycles N
c
r
a
c
k

d
i
m
e
n
s
i
o
n
s

a

e
t

c

(
m
m
)
PD3 PD2
PD1 cmoa
numerical CMOA (c/a = 2) f rac. surf . measurements
surf ace (c)
depth (a)

Figure 9: Determination of the propagation by the five correlation types


Fatigue Crack Growth in P91 Plates OMMI (Vol.3, Issue 2) August 2004

9

from linear correlation of PD(PD1)
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
10 100
DELTA K (MPa.m0.5)
d
a
/
d
N

e
t

d
c
/
d
N

(
m
m
/
c
y
)
depth
surface
CT paris law

f rom quadrat i c correl at i on of PD ( PD2)
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
10 100
DELTA K (MPa.m0.5)
d
a
/
d
N

e
t

d
c
/
d
N

(
m
m
/
c
y
)
depth
surface
CT Paris law

cubi c co rrel at i on of PD ( PD3 )
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
10 100
DELTA K (MPa.m0.5)
d
a
/
d
N

e
t

d
c
/
d
N

(
m
m
/
c
y
)
depth
surface
CT Paris law

experi ment al cmoa co rrel at i on
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
10 100
delta K (MPa.m0.5)
d
a
/
d
N

e
t

d
c
/
d
N

(
m
m
/
c
y
)
depth
surface
CT paris law

from FEcompliance
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
10 100
delta K (MPa.m0.5)
d
a
/
d
N

e
t

d
c
/
d
N

(
m
m
/
c
y
)
depth c/a=2
surface c/a=2
CT paris law


Figure 10: Crack propagation rate as a function of K for the five techniques of analysis


Conclusions

In order to validate high temperature defect assessment procedures for martensitic steels, the
Paris law of Z10CDVNb9.1 steel has been determined at 550C on CT specimens. At the
same time, tests on cracked structures are performed at the CEA/SEMT/LISN laboratory. In
this study, a fatigue test at 550C on a semi elliptical notch is detailed. The crack propagation
is followed by potential drop correlation, opening angle correlation and the compliance
method. Five techniques give similar results which constitutes a strong validation of these
different successive analyses. Moreover, K is calculated using a specific handbook. Finally,
the crack growth rate determination based on the cmoa measurements is scattered which is not
the case using the potential drop method. Using this last technique, it is shown that the crack
propagation at the plate surface is correctly estimated by the Paris law because no closure
effect can be involved there. On the other hand, the phenomenon of closure is probably more
pronounced for the deepest point, which may lead to an overestimation of the local loading
and therefore it would be wrong to call into question the applicability of the Paris law.


References

[A16 2002] Annex A16 of the RCC-MR code 2002.

[CAST3M] http://www-cast3m.cea.fr/cast3m/index.jsp

Fatigue Crack Growth in P91 Plates OMMI (Vol.3, Issue 2) August 2004

10
[Chapuliot 2000] S. Chapuliot Formulaire de KI pour les tubes comportant un dfaut de
surface semi elliptique longitudinal ou circonfrentiel interne ou externe, Rapport CEA-R-
5900, 2000.

[Curtit 1999] F. Curtit, Propagation de fissures semi-elliptiques en fatigue - Fluage 650C
dans des plaques dacier 316L(N) avec ou sans joints soudes, Thse en sciences et gnie des
matriaux lEcole Nationale Suprieure des Mines de Paris, 1999.

[Drubay 2003] B. Drubay, S. Marie, S. Chapuliot, M.H. Lacire, B. Michel, H. Deschanels,
A16: guide for defect assessment at elevated temperature, International Journal of Pressure
Vessels and Piping, 80, 2003, pp.499-516.

[Tada 2000] H. Tada, P.C. Paris, G. R. Irwin, The stress analysis of cracks handbook -
Third Edition ASME Press, 2000.