Professional Documents
Culture Documents
Fatigue and Creep Fatigue Crack Growth in 316 Stainless Steel Cracked Plates at 6508C
Fatigue and Creep Fatigue Crack Growth in 316 Stainless Steel Cracked Plates at 6508C
www.elsevier.com/locate/ijpvp
Abstract
This paper presents experimental results on fatigue and creep±fatigue experiments consisting of a wide cracked plate subjected to cyclic
bending loads at 6508C. Six tests have been realised with different loading histories. The specimen is a large 316SS plate containing a wide
semielliptical surface notch precracked by fatigue.
Fatigue and creep crack growth laws are determined, using an adapted stress intensity factors compendium calculated by F.E. elastic
analysis, and compared to tests results conducted on CT specimen. A good correlation of the experimental results for fatigue tests is obtained.
For creep crack growth law, the CT specimen gives an intermediate result between plate results at the deepest point and at the surface point.
A16 guideline assessment procedures are used; results are found to be in good agreement with experimental crack growth. q 2002
Elsevier Science Ltd. All rights reserved.
Keywords: Crack growth; Fatigue; Creep; Creep±fatigue; Defect assessment; 316 Stainless steel; Plates
1. Introduction machined defect was 2.5 mm deep and had a total length
of 80 or 85 mm depending on the case.
This article presents the results of the FIPAC and HIDA Two bending arms were secured to the plate, and a
experiments carried out on plates of 316L(N) austenitic steel tension/compression force was applied to them as shown
subjected to bending at 6508C, focusing on: in Fig. 1. The force was applied by an electromechanical
actuator with a capacity of 200 kN.
² the effect of loading conditions on propagation rate, For heating, a furnace was suspended on the set-up by
² the effect of closure, means of a series of struts enabling it to follow the move-
² the creep propagation law. ments of the bending arms and the plate. Seventeen thermo-
couples allow the measurement of the temperature all over
The results are used to validate the methodology the plate and then to insure the homogeneity of the tempera-
proposed in document A16 [1] for analysing the propagation ture distribution.
of defects at high temperatures.
2.2. Instrumentation
2. Description of the HIDA and FIPAC experiments
During all the experiments, the force, the inclination of
2.1. Principle
the levers and the de¯ection were recorded, and propagation
was monitored using the potential drop method.
The experiments [2] were carried out on plates of
For the last two experiments (FIPAC4 and HIDA2), a
316L(N) steel containing a semielliptic crack subjected to
system for measuring crack mouth opening displacement
cyclic bending loads with and without holding at a tempera-
by means of 2 laser interferometers was used.
ture of 6508C [3].
Fig. 1 shows the plates used. These were 24.5 mm thick
and 350 mm wide at the cracked section. The initial
2.3. Tests matrix
* Corresponding author. Tel.: 133-1-6908-9257; fax: 133-1-6908-8784.
E-mail address: smarie@cea.fr (S. Marie). Table 1 gives all the tests performed.
0308-0161/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0308-016 1(01)00099-0
848 S. Marie, C. Delaval / International Journal of Pressure Vessels and Piping 78 (2001) 847±857
Fig. 2. Crack growth rates at depth and surface versus imposed DK.
(the instant when the creep component is the smallest) is coef®cient of closure U, corresponding to the ratio
relatively close to this Paris law. DKeff =DK; was determined for different instants of
Conversely, the effect of hold time is far more apparent in precracking.
the case of the FIPAC4 experiment (in which Tm 3 h), Fig. 3 shows the force versus crack opening displacement
where the propagation rates are greater than with the law for three cycles of the fatigue pre-cracking period in
obtained on compact specimens. experiment FIPAC4.
It can be seen that, at the beginning of the cycle (with a
3.2. Analysis of closure effect positive force), the compliance is identical to that at the ®rst
instants of loading. Then, when loading of the crack
3.2.1. The case of fatigue loading becomes suf®ciently great, the defect opens. The effective
An analysis of the closure effect during fatigue pre- stress intensity factor DKeff corresponds to this instant.
cracking was carried out on the results of the FIPAC4 and Fig. 4 shows the variation of experimental closure coef®-
HIDA2 experiments (with a negative load ratio R). The cient U as a function of applied load ratio R deduced from
Table 2
Material properties
Tensile properties
Cyclic curve
2
1 1 n Ds 1=m
De 100 Ds 1 ; K 647; m 0:296; De
%; Ds
MPa
3E K
Paris law
da n
C f DKeff ; Cf 4:662 £ 1027 ; n 2:339; da=dN
mm=cycle; DKeff
MPa m1=2
dN
² Primary±secondary transient time: tffp
100C=C1 C2 s n2n 1 1=
C2 21 ; C1 5:863 £ 10211 ; C2 0:565; n1 4:233; C 1:018 £ 10225 ; n 9:407;
ec
%; s
MPa; t ffp
h
² Secondary creep
t $ tffp : ec C1 tC2 s n 1 1 100C s n
t 2 tffp ; C1 1:37 £ 10210 ; C2 0:2953; n1 4:233; C 1:018 £ 10225 ; n 9:407; ec
%;
s
MPa
pre-cracking in the HIDA2 and FIPAC4 experiments. This disappeared after the tenth cycle, despite a load ratio
®gure also shows the laws proposed: of R 21:
in document A16 [1]: In fact, this phenomenon is the result of competition
1 2 0:5R 1 between the effect of closing and opening of the defect
U pour R , 0 U pour R . 0 during the hold time. The reduction in closure effect is there-
12R 1 2 0:5R
fore dependent on hold time Tm and load ratio R: the drop
by Polvora [4]: being more rapid when Tm is long and R is large.
U 0:2R2 1 0:5R 1 0:7 pour 2 1 , R , 0:5 3.3. Analysis of the creep part of the FIPAC4 and HIDA2
experiments
U 0:0143R 1 0:4261 pour R , 21
Measurement of crack opening displacement and the
All the experimental results lie between the two curves. crack opening angle during the HIDA2 and FIPAC4
As regards simpli®ed methods, the relationship given in experiments made it possible to use the experimental
document A16 would appear to be the best suited, at least method of calculation of C p proposed in Ref. [3], and
for negative load ratios, as it constitutes an upper-bound for establish the da=dt 2 Cp law for plate tests.
the experimental results [5].
3.3.1. Experimental method of calculating C p
The method is based on the crack opening angle rate [6],
3.2.2. Case of creep fatigue loading and coef®cients determined by ®nite element calculations
For a creep fatigue cycle, the closure effect situation is (b and kb):
different. Fig. 5 shows three cycles obtained during the
FIPAC4 experiment (carried out with a hold time of 3 h). p kb_a
V0 ; V1 ; V2 2 s b pa dCMOA
Cs_a b
n1
It can be seen that the effect of closure has totally kb
V0 ; V1 ; V2 dt
852 S. Marie, C. Delaval / International Journal of Pressure Vessels and Piping 78 (2001) 847±857
In order to validate the calculated values of C p, two ®nite 3.3.2. Calculation of crack growth associated with creep in
element calculations were carried out for experiments the FIPAC4 experiment
FIPAC4 and HIDA2. This consisted of ®rst modelling the In the FIPAC4 experiment case, it is not possible to
load build-up by an elastoplastic calculation using a cyclic distinguish crack growth associated with creep from that
law for the material. The strain and stress ®elds obtained associated with fatigue without making assumptions on
were then used as initial conditions for the creep calculation. the manner in which the two types of damage combine.
This modelled the behaviour of the material using the To evaluate the da=dt 2 C p law of the FIPAC4 experiment,
primary and secondary laws for the material (see Table 2). total growth during a cycle was assumed to be the sum of the
The calculations were carried out with the CEA code fatigue increment associated with the stress intensity factor
CASTEM2000 [7]. and the crack increment associated with creep. This hypoth-
The laws in question were not those obtained during esis corresponds to the simpli®ed method in document A16.
characterisation tests. Indeed, Curtit has demonstrated a There is therefore a crack increment associated with the
non-negligible effect of work hardening of the material creep part of loading by the following relationship:
during fatigue pre-cracking with the creep laws [3]. For
daflu datot 2 dafat :
creep and creep±fatigue tests, the reference stress Ds ref is
140 MPa (calculated from a limit load relationship) and then where datot is the total increment, effectively experimentally
is higher than the yield stress (125 MPa, see Table 2). For measured, and dafat the crack increment associated with the
plates, Curtit has numerically reset, by ®nite element calcu- fatigue part of loading by integrating the Paris law.
lation, a primary law making it possible to replicate the
experimental behaviour. It is this law that was used (see 3.3.3. Determination of the da=dt 2 C p law
Table 2): the loading history between FIPAC and HIDA Fig. 7 show variation of the crack opening angle of the
experiments and the experiments of Curtit were basically simulated crack (theoretically involved in calculation of C p)
similar, both it concerns procedure and loading level, with the experimental curves for the FIPAC4 and HIDA2
enabling us to apply the same law. experiments.
Fig. 6 shows the mesh used for the FIPAC4 test. In view As concerns the FIPAC4 experiment, it was found that
of the symmetry, only a quarter of the plate was modelled. the initial values of the angle of opening are correctly
The crack modelled was the ®nal defect observed at the predicted, but thereafter the variation is greatly underesti-
fracture surface. For the FIPAC4 experiment, the cyclic mated by the calculation: this is due to the fact that the
part of the test was disregarded, being equated to a pure simulation considers a stationary crack, which is not the
creep test in the simulation. case during the tests where substantial propagation is
S. Marie, C. Delaval / International Journal of Pressure Vessels and Piping 78 (2001) 847±857 853
Fig. 7. Crack opening angle F.E. simulation for HIDA2 e and FIPAC4 tests.
observed. The consequences for the estimate of C p are As concerns the depth points, it was found that the agree-
shown in Fig. 9. Although the order of magnitude of the ment was relatively good with the law obtained with
two changes is identical, the calculation indicates a decreas- compact specimens (which however indicates rates that
ing curve whereas the experimental results show an are slightly greater than was found with the plates).
increase, associated with propagation. Finally, the propagation rates, at equal loading (in term of
For the HIDA2 experiment, it was found that the agree- C p), are faster at the surface than at a depth. The experi-
ment between variation of the crack opening angle as indi- mental results on plates thus bracket the creep propagation
cated by the experiment and determined by calculation is law obtained from compact specimens (which represents an
relatively good, which con®rms the validity of the results `average' law).
given by simulation. Furthermore, the values of the C p Two possibilities allow the explanation of the difference
obtained experimentally (Fig. 9) can be considered to be of crack growth rate between surface and depth:
satisfactory in the light of ®nite element calculation in
view of the accuracy of the experimental data. Residual stresses built up after each unloading could
Fig. 8 shows the points obtained with the FIPAC4 and justify this phenomenon: but Curtit [3] evaluated these
HIDA2 experiments in the da=dt 2 Cp diagram. residual stresses in the case of very similar test (same
For the HIDA2 experiment, only the depth law is deter- geometry, comparable load level) and show that they
mined, as the propagation obtained at the surface is are not suf®cient to explain the crack growth rate
extremely slight. Similarly, propagation measured at a difference, because the stress intensity factor variation
depth is very small: the rates of propagation calculated are due to these stresses is not higher than 1, 2 MPa m 1/2.
to be considered to represent a scale of magnitude rather The work hardening±creep interaction, already
than reliable values. mentioned, could explain this result: It is show that this
Fig. 8. Creep crack growth rate versus C p for HIDA2 and FIPAC4 tests.
854 S. Marie, C. Delaval / International Journal of Pressure Vessels and Piping 78 (2001) 847±857
interaction reduce the kinematics of creep, and hence 4.1.1. Calculation of reference stress
stress relaxation. But, as the plate is submitted to a bend- For determining fracture parameters, document A16 [1]
ing, this phenomenon is non homogeneous through the uses the reference stress s ref concept. To estimate this stress,
thickness: It is more important where the stress is maxi- it is necessary to calculate two equivalent stresses: an
mum, that is at the surface. Then, the reducing of creep equivalent elastic stress and an equivalent plastic stress
kinematics is highest at the surface: the stress relaxation s eq_el and s eq_pl.
is also minimum at this area. The crack tip loading is then To do so, the linearised elastic stresses for the defect
more severe at the surface point, which explain that for a section are ®rst determined. These are then combined in
given value of C p, crack growth rate is higher at the Illiouchine's [8] two equivalent stress formulas to obtain
surface point than at the deepest point. s eq_el and s eq_pl.
The reference strain e ref is then determined:
s eq_el
e ref 1 epl
s eq_pl
4. Comparison of A16 methods with experimental results E
where e pl(s ) is the plastic strain corresponding to stress s
4.1. The methodologies proposed in document A16 with the stress±strain curve used.
The reference stress is then deduced from this reference
Document A16 [1] proposes different methodologies for strain by means of the behaviour law.
estimating the propagation of a crack and its effect on the
behaviour of a structure, particularly for fatigue, creep and 4.1.2. Fatigue problem analysis
fatigue±creep loading. For a fatigue loading, the method proposed in document
Fig. 10. A16 method for fatigue crack growth: case of FIPAC2 test.
S. Marie, C. Delaval / International Journal of Pressure Vessels and Piping 78 (2001) 847±857 855
A16 involves estimating the crack increment for a certain defect is supposed to be semi-elliptical. This could explain
number of cycles by integrating the Paris law: To do so, it is differences with previous results, obtained considering the
necessary to calculate stress intensity factor DK, obtained real shape of the defects.
from DJ:
p 1 2 0:5R 4.2.1. The case of pure fatigue: FIPAC1 and FIPAC2
DK U E p DJs U for R , 0 experiments
12R
The FIPAC1 and FIPAC2 experiments were carried out
1 for fatigue, under the same loading conditions: R 0:1 and
U for R . 0
1 2 0:5R Fmax 19 kN: As the FIPAC2 experiment is better docu-
where U is the closure coef®cient, E p is equal to E in plan mented and covers the greater propagation, we compared
stress and E=
1 2 n2 in plan strain. For a loading with an the results supplied by the A16 method to those obtained
imposed force: with this test.
To be able to apply the method in document A16, the
Eeref defect is assumed to be semi-elliptical in shape. Further-
DJs DJel
s ref more, the loading considered is pure bending loading
where DJel is the elastic value of J with a semi-elliptical (membrane stress is disregarded). Finally, the Paris law
defect. used is that proposed by Curtit [3], adapted to our material.
The crack implement is obtained by integrating the Paris Fig. 10 shows the results obtained. The agreement is
law: excellent with the experimental points, if allowance is
made for the uncertainty as to the number of cycles at
da C DK m experimental initiation (which implies horizontal offset of
where m and C are coef®cients of the Paris law. the experimental points), particularly for the surface point.
The result was however predictable as, in the previous
4.1.3. The case of creep propagation section, there was found to be good agreement between the
The crack increment obtained during a time dt is obtained Paris law obtained with compact specimens and the points
by integrating the da=dt 2 Cp law. To do so, it is necessary derived from tests on plates.
to ®rst estimate the parameter Csp : For loading with an
imposed force, this is expressed as follows: 4.2.2. The case of fatigue±creep propagation: FIPAC3,
FIPAC4 and HIDA1 experiments
Ee_ ref As before, only bending loading was considered. Fig. 11
Csp Jel
s ref shows the results obtained, using the primary creep law for
where e_ ref is the strain rate associated with the reference work-hardened material [3].
stress by means of the creep laws. In this case: For the FIPAC4 and HIDA1 experiments, the results are
excellent. For the FIPAC3 experiment, the A16 method is in
Z dt q
da A Csp dt: close agreement with the test results up to approximately
0 750 cycles. Thereafter, the experimental propagation
accelerates, which is associated with substantial crack
4.1.4. The case of fatigue±creep propagation propagation on the side opposite to that of the initial defect,
For fatigue±creep propagation, document A16 proposes which the simpli®ed method cannot take into account.
summing the fatigue and creep propagation without any
interaction between the two phenomena: 4.2.3. The case of creep propagation: HIDA2 experiment
For the HIDA2 experiment, carried out for pure
datot dafat 1 daflu creep, the creep law used for the simulations is that
To estimate the crack growth associated with creep, for for work-hardened material. As shown in Fig. 12, the
cycle N, document A16 proposes integration of the da=dt 2 A16 method substantially overestimates the propagation
C p law for the holding time, as per the Polvora [4] obtained. However, the result is broadly satisfactory, as
methodology: the creep time is not reset each time the the propagation considered are very small (less than
holding period Tm begins: 0.3 mm). Furthermore, the values of C p calculated for
the end of the test are in close agreement with the
Z
N 1 1Tm q
da A Csp dt experimental values (Fig. 12)
NTm In fact, the high values of C p calculated for the beginning
of the crack growth (t , 400 h) result, by integration of the
4.2. Analysis of the results obtained da=dt 2 Cp law, in initial crack increments that are too
large.
The material data used in the experiment are given in It is therefore dif®cult to draw conclusions as to the
Table 2. For all applications considered in this part, the quality of prediction of the method in document A16 for a
856 S. Marie, C. Delaval / International Journal of Pressure Vessels and Piping 78 (2001) 847±857
Fig. 11. A16 method for creep±fatigue crack growth: (a) case of FIPAC3 test; (b) case of FIPAC4 test; (c) case of HIDA1 test.
creep experiment on the basis of the present experiment, for studied using a stress intensity factor formula adapted to the
which the propagation was too small. shape of the defects observed:
Fig. 12. A16 method for creep crack growth: case of HIDA2 test.
Concerning the closure effect: use of this work hardening law makes it possible to replicate
the behaviour of a plate in bending by ®nite element calcu-
² the results obtained with the FIPAC4 and HIDA2 experi- lation, as shown by Curtit.
ments show that, for fatigue, the closure coef®cients are However, this creep law for work-hardened material
bracketed by the relationships proposed in document A16 certainly depends on the conditions of loading and the
and by Polvora. In the case of a design analysis, the A16 geometry in question, as would appear to indicate the
relationships is better suited as it represents an upper- difference in the rate of creep propagation between that at
bound for this coef®cient, the surface and at a depth. Understanding this interaction
² as concerns fatigue±creep loading, blunting of the crack would appear to be the key to determining the appropriate
tip during the hold time tends to eliminate this closure law for the con®guration in the context of a simpli®ed
effect. method.