You are on page 1of 18

Nuclear Engineering and Design 232 (2004) 219236

Creep-fatigue behaviour of an AISI stainless steel at 550 C


M. Sauzay , M. Mottot, L. Allais, M. Noblecourt, I. Monnet, J. Perinet
Commissariat a` lEnergie Atomique, DEN-DMN-SRMA, Batiment 455, 91191 Gif-sur-Yvette Cedex, France
Received 19 November 2003; received in revised form 18 May 2004; accepted 20 May 2004

Abstract
In the past, a lot of experimental studies have been devoted to creep-fatigue interactions in austenitic stainless steels. Tests have
been carried mainly at temperatures of at least 600 C and at high applied strains, which are supposed to be the most damaging.
The present work is dedicated to mechanical tests, TEM observations and lifetime predictions at 550 C which corresponds to
the real industrial temperature in Liquid Metal Fast Breeder Reactors. It is shown that if pure fatigue test results are close to
those performed at 600 C, some of the creep-fatigue results are different, particularly for small applied strains which correspond
once more to the industrial conditions. In the 0.250.3% strain amplitude range, the stress is larger with hold time than without
whatever is the hold time up to 5 h. The numbers of cycles to failure are greatly reduced and no saturation with the hold time is
observed, contrary to higher temperature results. The stressstrain behaviour is discussed considering several high temperature
mechanisms such as ageing, recovery and viscoplasticity and using TEM observations and stress partitioning into kinematic,
isotropic and thermal stresses. Finally, a simple linear damage accumulation model is applied to the 550 C and 600 C tests,
using the measured stresses. The stress dependence on hold time can partly explain the observed failure results on fatigue life.
2004 Elsevier B.V. All rights reserved.

1. Historical background
For the development of structural material for Liquid Metal Fast Breeder Reactors (LMFBR), the creep
fatigue interaction is a very important topic to consider.
Indeed, many structural components are subjected to
repeated thermal stresses as a result of temperature gradients which occur on heating and cooling during startup and shut down cycles or during thermal transient.
For those components working at high temperature in
Corresponding author. Tel.: +33 1 69 08 35 67;
fax: +33 1 69 08 71 67.
E-mail address: sauzay@cea.fr (M. Sauzay).

0029-5493/$ see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.nucengdes.2004.05.005

the creep range, creep fatigue interaction induces a life


reduction which has to be evaluated. From the experimental viewpoint, as most of the studies devoted to
that topic, the life evaluation relies on low-cycle fatigue (LCF) tests with hold times (Wood et al., 1980;
Mottot et al., 1982; Ermi and Moteff, 1982; Cailletaud
et al., 1984; Levaillant et al., 1988).
Among the austenitic stainless steels of AISI 300 series, unstabilized 316L(N) has been used in France for
Superphenix plant. Then this material was selected for
the European LMFBR project EFR (European Fast Reactor). Although considerable LCF and creep-fatigue
interaction data have been obtained on this steel, these

220

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

Table 1
Chemical composition of the 316LN sheets (wt.%)

C
S
P
Si
Mn
Ni
Cr
Mo
N
B
Co
Cu

SP

SQ

0.021
0.007
0.03
0.4
1.74
12.3
17.2
2.4
0.08
0.0032
0.21
0.15

0.028
0.001
0.023
0.4
1.85
12.31
17.4
2.47
0.074
0.001
0.14
0.17

results come from mechanical tests conducted at higher


strains or temperature than the actual service ones.
The aim of the present work on type 316L(N) is to
investigate a field as close as possible to the actual service one (i.e. working at the service temperature 550 C
and considering very small strain amplitude (range under 0.5%) with hold times within 101440 min). Results of many test-machine years are reported in terms
of stressstrain creep-fatigue behaviour and LCF life
reduction with hold time. The evolution of the cyclic
behaviour with hold time is discussed thanks to a micromechanical model taking into account dislocation
mechanisms. In order to validate this model some TEM
observations are performed. The second part of the discussion is focused on the life prediction considering
usual creep-fatigue damage accumulation models.
2. Experimental procedures
2.1. Stainless steel
In 316L(N) steel, for inter crystalline corrosion matter, carbon content has been decreased. Then to compensate the effects of this low carbon content on the
mechanical properties, a controlled amount of nitrogen
was specified. These reasons have led to the 316L(N)
specifications reported in Table 1. Two plates of 30 mm
thickness of type 316L(N) steels were used. These
plates referred as SP and SQ have been considered in
many orther studies (Mottot et al., 1982; Ermi and Moteff, 1982; Cailletaud et al., 1984; Levaillant et al., 1988;
Argence and Pineau, 1995). The chemical composition
of the material is reported in Table 1. Both steels were

tested in the as-received condition corresponding to an


annealing treatment at 1050 C. Microstructural investigations can be found elsewhere (Tavassoli, 1986).
2.2. Creep-fatigue tests
LCF tests were conducted in air on MAYES
ESM100 servo-mechanical machines with resistance
furnace heating. Temperature along the gauge length of
the specimen was controlled to better than 2 C. The
cylindrical specimens of 16 mm gauge length and of
8 mm diameter with a shoulder radius of 16 mm were
machined in the longitudinal direction. They were
finished by fine turning to a roughness average value of
0.8 m. The axial strain was measured by a capacitive
extensometer directly fixed on the calibrated part. The
calibrated extensometer gauge length is of 10 mm.
The accuracy of this device is less than 0.5 m that
allows to conduct LCF tests with a strain amplitude
as small as 0.1%. The LCF tests were controlled on
the total strain measured with this extensometer. A
symmetrical triangular wave form with an axial strain
rate of about 3 103 s1 was used for the cyclic part
of the tests. For the hold time tests, the strain was held
constant at tensile peak strain. Thanks to a numerical
system, the tensile and compressive stress peaks were
recorded continuously. Stressstrain hysteresis loops
and stress relaxation curves during hold time were
recorded for distributed cycles.
2.3. TEM
After mechanical thinning down to 100 m, thin
foils with a 3mm diameter are prepared by electropolishing in a twin jet TENUPOLE device with the 721
electrolyte (70% ethanol, 20% butoxyethanol and 10%
perchloric acid) at 13 C with a current of 200 mA.
TEM observations are performed using a Philips EM
430 microscope operating at 300 kV.
3. Experimental results
3.1. Creep-fatigue behaviour
Pure fatigue tests at 550 C confirm the results obtained before at 600 C (Mottot et al., 1982; Cailletaud
et al., 1984; Levaillant et al., 1988; Argence and Pineau,
1995). During the first cycles, there is a hardening of

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

221

Fig. 1. Evolution of the stress amplitude with respect to the fraction of fatigue life. Pure fatigue, AISI stainless steel 316LN, 550 C.

the annealed alloy cycle by cycle (Fig. 1). After saturation the stressstrain cycles become quite stable, before
a drop of the maximal stress per cycle due to damage
at the end of the lifetime is observed. The maximal
stress continues to increase cycle by cycle until saturation, except for very small applied strains (range under
0.5%) for which no saturation is observed. Fig. 2 compares monotonic and cyclic stress strain (CSS) curves
and shows that the cyclic hardening is large. There is no
significant difference between the CSS curves obtained

at 550 C and 600 C. All these results are similar to


those usually observed in 316LN stainless steels.
The creep-fatigue tests are now considered. The
results obtained at 600 C had already been published
(Mottot et al., 1982; Cailletaud et al., 1984; Levaillant et al., 1988) and are now reminded. Only the saturated behaviour is described, but it should be noticed
that the saturation of the hardening requires less cycles
than without hold time. The comparison of the maximum tensile stress obtained for pure fatigue ( tmax0 )

Fig. 2. Cyclic and monotonic stressstrain curves. Pure fatigue, AISI stainless steel 316LN, 550 C.

222

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

Fig. 3. Variations of the tensile stress ratio with respect to the hold time. The maximum tensile stress measured for a given th hold time is divided
by the pure fatigue one (no hold time). Fatigue-relaxation, 600 C.

and for fatigue-relaxation ( tmaxth ) showed that the effect of the hold time on the maximum stress amplitude
is quite large (Fig. 3). For very small hold times (less
than 10 min), there is a stronger cyclic hardening than
without hold time (+7%), but for larger hold times,
there is a reduced hardening. The transition takes
place for hold times of about 10 min. For hold times
greater than 300 min, the reduction of the variation and
maximum tensile stresses is about 1020% of the pure

fatigue stress ones for the same applied strain range.


The larger is the applied strain, the larger is the relative
maximum stress decrease (with respect to the corresponding pure fatigue tests saturated stress levels).
During the hold times there is a stress relaxation due
to viscoplasticity. At the end of the hold time, a tensile
stress, tmin , is measured. Because of the viscoplastic
flow, the larger is the hold time (or/and larger is the
applied strain), the greater is the relaxed stress (Fig. 4).

Fig. 4. Variations of the relaxation stress ratio with respect to the hold time. The relaxed stress measured for a given th hold time is divided by
the maximum stress (measured for the same hold time). Fatigue-relaxation, 600 C.

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

The tests at 550 C give quite different results. The


tests performed with large and small applied strains
should be distinguished concerning the hold time effects. First, for large strains (variation larger than
1.2%), the tests lead to similar results to those performed at 600 C. For small hold times the hardening
is larger than without hold time (Fig. 5). The increase of
the maximal tensile stress is about 15%. For large hold
times, the hardening is smaller than at 600 C (about
12%). The hold time transition is larger than at 600 C
(about 1000 min at 550 C) and the stress reduction is
smaller for the same hold time and the same strain amplitude. Thermal activation of the viscoplasticity and
recovery could explain this shift. Concerning relaxation
during hold times, about 30% of the maximum stress
is relaxed during the larger dwell time (4000 min, see
Fig. 6). Second, in the small strains case, (range smaller
than 1%) and for hold times smaller than 300 min (experimental investigated field), there is only an increase
of the stress amplitude. No reduction of the saturated
stress amplitude is observed for small applied strains
(Figs. 5 and 7). The stress amplitude is increased by
15% in most of the corresponding tests. During the
dwell time, there is a quite instantaneous relaxation of
1015 MPa (Fig. 6). But, during the used hold times
(smaller than 300 min) no more relaxation is observed.
In fact, within the 4300 min dwell time range, the only
observed effect of the hold time is the 15% maximum

223

tensile stress increase. At 600 C, in terms of the mechanical behaviour, there is a small difference between
small and large applied strains. But this difference is
much more significant at 550 C.
The complex stressstrain behaviour, which depends on the temperature, on the dwell time and on the
applied strain could influence the damage mechanisms
and the lifetime. The experimental results concerning
lifetime are now described.
3.2. Creep-fatigue damage
The pure fatigue tests at 600 C and 550 C give
similar results (Fig. 8). In this range of temperature,
the MansonCoffin curves are little influenced by the
temperature. It is the same concerning the CSS curves.
The experimental dispersion on the number of cycles to
failure depends on the strain level. For a strain equal to
0.2%, the dispersion factor is about 6. The fatigue life
reduction factors are given in Figs. 9 and 10 (600 C)
and 11 (550 C). They are defined as the ratios between
the fatigue life measured for a given th hold time and
the fatigue life without hold time.
The results concerning the creep-fatigue tests carried at 600 C have already been published (Mottot
et al., 1982; Cailletaud et al., 1984; Levaillant et al.,
1988). The larger is the hold time, the smaller is the
number of cycles to failure. For large applied strains

Fig. 5. Variations of the tensile stress ratio with respect to the hold time. The maximum tensile stress measured for a given th hold time is divided
by the pure fatigue one (no hold time). Fatigue-relaxation, 550 C (compare with Fig. 3).

224

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

Fig. 6. Variations of the relaxation stress ratio with respect to the hold time. The relaxed stress measured for a given th hold time is divided by
the maximum stress (measured for the same hold time). Fatigue-relaxation, 550 C (compare with Fig. 4).

(/2 > 0.3%), these numbers are divided by a factor


of about 3 for dwell times larger than 30 min (Fig. 9).
There is a saturation of the hold time effect (except for
the larger applied strain amplitude 0.8%). This saturation could be due to the decrease of the stress amplitude.
But, it can be noticed that the life time increases with
the hold time. These two observations could partially be
explained by the decrease of the stress amplitude with
the dwell time and by the relaxation of the stress during

hold times (perhaps the stress becomes small enough


to create few damage during a large part of long time
relaxation tests). It should be noticed that hold times
associated with small strains are more damaging. The
life time reduction factors measured with intermediate
strains (/2 = 0.25%, 0.28%, 0.3%) are about 3 but
no saturation is observed even for a 700 min hold time
(Fig. 10). But, the damaging effect of the dwell time
is maximal with the smallest strains (/2 = 0.2%)

Fig. 7. Evolution of the stress amplitude for different hold times, with respect to the fraction of fatigue life. Pure fatigue, AISI stainless steel
316LN, 550 C, strain amplitude t = 0.3%.

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

(Fig. 10). The reduction factor is about 6 for the Krupp


sheet (an additional 316L(N) reference) and about 15
for the SQ one. For small strains, no clear saturation is
observed.
The experimental results concerning tests performed at 550 C are now described (Fig. 11). Dwell
times associated to small strains are still the most damaging. A 300 min hold time induces a fatigue life reduction factor of 10 for a 0.250.3% strain amplitude. This
decrease in fatigue life with the hold time is therefore
larger than that observed at 600 C. It confirms another
test result which has been performed in Springfield
(UK) on a Krupp sheet specimen with a strain amplitude of 0.2% and a 30 min hold time (Fig. 11). A reduction factor of about 50 has been measured. These results
confirm the large damaging effect of small strains. As
concerning the stressstrain behaviour, the results at
550 C have to be partially distinguished from those at
600 C, particularly for small applied strains. The hold
time effect is larger than at 600 C for the same test
conditions. And, even for large strains, no saturation
is observed. For large strains, the fatigue life reduction
factor is equal to 5 for a 4000 min hold time. In part 4,
a very simple linear accumulation rule is applied, but
using the real measured stressstrain evolution during
cycling including hold times. It will be shown that the
stressstrain dependence on the hold time partially explains the lifetime results. Finally, it should be noticed
that the most damaging test conditions correspond to

225

the real industrial ones in power plants (very long hold


times and small strains). These conditions need so long
test times that they are difficult to be performed in laboratories, therefore reliable extrapolations, based on the
involved physical mechanisms, should be made.
3.3. TEM observations of the dislocation
microstructures with and without hold time
The dislocation mechanisms of cyclic plasticity of
316L steels have been largely studied. The TEM observations show the formation of veins or walls structures
(PSBs, labyrinths . . .) at 20 C (Gerland et al., 1989;
Feaugas and Gaudin, 2001) and 600 C (Gerland and
Violan, 1986; Boulanger et al., 1985) for an applied
plastic strain amplitude over the 0.10.5% range. These
plastic strains correspond to the measured ones of our
applied total strain tests. The dislocation microstructures consist of soft zones (with low dislocation densities) and hard zones (with large dislocation densities). The volume fraction of hard phase is generally
510 times smaller than the soft phase one. Generally,
the walls are considered to be composed of dipoles
(Essmann and Mughrabi, 1979; Feaugas, 1999).
Our TEM observations on thin foils from a specimen tested at 550 C with a 0.3% applied strain amplitude confirm that the dislocation microstructure depends on the applied strain and on the observed grain
(Figs. 1214 ) (Feaugas and Gaudin, 2001). There are

Fig. 8. Failure curves. Pure fatigue, 550 C and 600 C.

226

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

Fig. 9. Fatigue life reduction factor with respect to the hold time. Large strains, 600 C. The fatigue life reduction factors are defined as the
ratios between the fatigue life measured for a given th hold time and the fatigue life without hold time.

several activated slip systems in a grain as observed in


Fig. 12, but no real two phases microstructure. In this
kind of grain, the dislocation density is about 1.3
1014 m2 . In the grain shown on Fig. 13 (see details in
Fig. 14), a labyrinth microstructure is observed. In the
channels, the dislocations are straight and there are a
lot of defects which could be created by the reactions
between the gliding dislocations during the repeated
cycles. In the soft phase (channels), the dislocation den-

sity is low (3 1013 m2 ), but the density of defects


is quite large (3 1021 m3 ). As detailed below, these
two-phase microstructures partially explain the cyclic
hardening of annealed metals which is larger than the
monotonic one (Fig. 2).
If the dwell time is large enough, recovery mechanisms induce the vanishing of the hard phase (dipoles).
This has been observed for a 316LN steel at 600 C by
Tavassoli (1986) and Ermi and Moteff (1982), for a 304

Fig. 10. Fatigue life reduction factor with respect to the hold time. Small strains (t < 1%), 600 C. The fatigue life reduction factors are
defined as the ratios between the fatigue life measured for a given th hold time and the fatigue life without hold time.

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

227

Fig. 11. Fatigue life reduction factor with respect to the hold time. Large and small strains, 550 C. The fatigue life reduction factors are defined
as the ratios between the fatigue life measured for a given th hold time and the fatigue life without hold time.

Fig. 12. Multiple planar slip in a grain (TEM). Pure fatigue. AISI stainless steel 316LN, t = 0.6%, 550 C.

228

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

to the stress level, the classical stress partition is now


used:
= R + X +

Fig. 13. Three grains with different dislocation microstructures


(TEM). AISI stainless steel 316LN, t = 0.6%, 550 C.

steel at the same temperature. Our TEM observations


confirm the vanishing of the walls for a specimen tested
at 550 C with a 300 min hold time and a 0.3% strain
amplitude (Figs. 15 and 16). The measured dislocation
densities are about of the same order of magnitude as
in the pure fatigue ones in homogeneous grains (about
2 1014 m2 ). There are less defects. These observations help to understand the stress dependence on the
hold time.

4. Discussion
4.1. Origin of the hold time inuence on the
mechanical behaviour
4.1.1. Stress partition
Several mechanisms could be involved in the evolution of the stressstrain behaviour with respect to the
hold time. In order to link the physical mechanisms

(1)

Three stresses are involved in this partition: the


isotropic stress (or friction stress), R, the kinematic
stress (or backstress), X, and the thermal stress,
(Lemaitre and Chaboche, 1987). Using the elastic part
of the unloading curve it is possible to measure the kinematic stress (that is the center of the elastic domain) and
the isotropic stress (elastic domain amplitude). From
a microscopic point of view, the three stress components of Eq. (1) correspond respectively to an athermal short-range stress, an athermal long-range stress
and a thermal stress. The total stress level depends on
these three stresses. Concerning the backstress, it is
usually considered to be due to plastic strain incompatibilities inside the material. Whatever their scale,
plastic strain incompatibilities seem to induce kinematic hardening but no isotropic hardening (because
the forward and the reverse gliding directions should
be distinguished). The incompatibilities could be intragranular or intergranular. The TEM observations have
shown classical two-phase dislocation microstructures
in many grains. These microstructures involve intragranular kinematic hardening because the soft phase
is plastically deformed and the hard phase is less deformed, which induces intragranular plastic incompatibilities. Such backstress, Xintra , has been modelled for
instance by Mughrabi (1983), or Feaugas and Gaudin
(2001). This hardening effect has to be added to the intergranular kinematic hardening which is due to plastic
incompatibilities between grains. The following backstress partition can be written (Feaugas and Gaudin,
2001):
X = Xintra + Xinter

(2)

The isotropic stress can be linked for example


to dislocationdislocation interactions or dislocation
locking by solute atoms (Haasen, 1984). The thermal
stress is due to the viscosity of the material and depends on the plastic strain rate (friction stress due to
solute atoms acting on gliding dislocations) (Haasen,
1984). The three stress components have been measured for pure fatigue tests (during unloading) and
for fatigue-relaxation tests (during unloading, after relaxation). These measurements could help to under-

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

229

Fig. 14. Details of the larger grain shown in Fig. 13 (TEM). AISI stainless steel 316LN, t = 0.6%, 550 C.

stand the evolution of the total stress because these


components could often be linked to physical mechanisms.
4.1.2. Solutiondislocation interactions and
isotropic stress
In 316L steels, solutedislocation interactions have
an important effect on the mechanical behaviour at
high temperature, particularly if hold times are used.
In Blanc (1986), these interactions are detailed. Two
types of interaction have to be considered at 550 C
or 600 C. First, if the stress is large enough, there is
a jerky flow or PortevinLe Chatellier effect, which
is observed during our pure fatigue tests. During this
kind of interaction, the solute atoms are dragged by
the mobile dislocations. Second, if the dislocations do
not move sufficiently rapidly, they can be locked by
the solute atoms. In this case, solute atoms can diffuse
towards the dislocation cores which can attract them.
Then, the atoms are captured by dislocations. An additive stress has to be applied in order to help these dislo-

cations to escape. If the stress is not large enough, dislocations cannot glide freely. This locking interaction
induces an increase of the isotropic hardening. During
the long hold times, the plastic strain rate decreases
greatly and so does the dislocation gliding speed, which
can induce dislocation locking by solute atoms. These
mechanisms cannot be easily observed but the isotropic
stress increase has been precisely measured using the
stressstrain curves obtained for the fatigue-relaxation
tests with the smallest hold times. Just after the relaxation, the isotropic stress is larger than the pure fatigue
one. But, the isotropic stress measured after the following minimal strain peak is smaller than the pure fatigue
one. This proves that the mean isotropic stress increase
is due to the relaxation time. Sequential tests have been
performed at 550 C and 600 C in order to have a better understanding of what happens for very large hold
times (about one month). First, a cyclic strain is applied
during 300 cycles in order to achieve the saturation of
the behaviour of pure fatigue test. Second, a relaxation
is applied during 15 days. Third, 300 pure fatigue cycles

230

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

Fig. 15. Two grains in a fatigue-relaxation specimen (TEM). t =


0.6%, th = 300min, 550 C. No real dislocation microstructure can
be found.

are applied once more with the same strain amplitude.


At these two temperatures, after relaxation, an increase
of the isotropic stress, R, is measured, the jump is about
90 MPa. Such sequential test could permit to evaluate
approximately the quantitative effect of hardening (or
softening) mechanisms and are helpful because such
15 days hold time is too large to be applied during a
real fatigue-relaxation test. The increase of the maximal tensile stress observed during fatigue-relaxation
tests for small dwell times is due to an increase of the
isotropic stress and can be partially due to this dislocation locking mechanism (by solute atoms) effect.
4.1.3. Dislocation microstructures and kinematic
hardening
For larger hold times, the maximum tensile stress
decreases. It is equal to the pure fatigue one at the transition time of about 10 min at 600 C and about 1000 min
at 550 C. After the transition times, the stress continues to decrease. The drop can be equal to 20% with
respect to the pure fatigue stress. Note that the case of
small strain amplitude at 550 C is an exception with
no decrease of the maximal tensile stress and only a
positive jump.

During dwell times, recovery mechanisms take


place too (Delobelle and Oytana, 1986). For example, edge opposite dislocations forming a dipole can
annihilate because of climb (Prinz et al., 1982). This
mechanism is thermally activated. It is due to vacancies
migration and the activation energy is the self-diffusion
one (about 280 kJ/mol in austenitic steels). If the dwell
time is large enough, recovery mechanisms induce the
vanishing of the hard phase (dipoles). The critical hold
time for homogeneous dislocation microstructure is
about 300 min for a specimen tested at 550 C and a
0.3% strain amplitude (Figs. 15 and 16). This would
avoid intragranular two-phase dislocation microstructures and reduce intragranular plastic strain incompatibilities, the kinematic hardening and a decrease of the
stress amplitude. The decrease is observed for large
applied strains at the two temperatures and for small
strains at 600 C. But, it is not observed for small applied strains at 550 C even if our TEM observations
show that the dislocation microstructure is quite homogeneous for a 300 min hold time. Following TEM
observations of Tavassoli (1986), at 600 C, for 30 min
hold time, all created dipoles are annihilated during
the relaxation times. At 550 C, our TEM observations
show that the dislocation microstructure becomes homogeneous for a 300 min hold time (Figs. 15 and 16).
Therefore, for the greater applied hold time (300 min),
the Xintra stress component could be already lost, which
corresponds qualitatively to a decrease that we have
actually measured. Feaugas and Gaudin evaluate the
value of Xintra at room temperature. The computed
value has to be multiplied by a factor 0.75 (which is the
ratio of the Young modulus at 600 C and at room temperature). Their model is applied to large strains. For an
applied plastic strain amplitude of 0.6%, Feaugas obtained: Xintra 90 MPa. At 600 C, this gives: Xintra
65 MPa. The decrease of the intragranular backstress
explains partially the decrease of the maximum tensile stress and stress amplitude observed at 550 C and
600 C. The stress reduction is not so efficient for small
strains for which the intragranular back stress is probably smaller than for large strains because of smaller
plastic strain incompatibilities.
4.1.4. Viscoplasticity and kinematic hardening
At 600 C, for larger hold times (within the 30 min
to 1440 min range), the maximal tensile stress continues to decrease. This is the same for large strains and

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

231

Fig. 16. Details of a grain (TEM). Fatigue-relaxation, t = 0.6%, th = 300 min, 550 C. There are less defects than in a pure fatigue specimen
(compare with Fig. 14).

hold times greater than 300 min at 550 C. During our


sequential tests (pure fatigue, then 15 days relaxation
and finally pure fatigue), backstress decreases of about
80 MPa are measured at 550 C and 600 C (these tests
concern small strains:  = 0.60.7%). The observed
drop cannot only be due to the vanishing of the intragranular back stress. This shows that another mechanism is involved. During the relaxation time, viscoplastic deformation takes place in the grains. This could
decrease the incompatibilities between the surrounding
grains and the intergranular back stress if the relaxation
time is long enough. This could explain why no stress
drop is observed for small strains and T = 550 C. In
these tests, dislocations seem to be locked by interstitials. In addition, the steels are less viscoplastic than at
higher temperature (thermoactivated slip mechanism).
Therefore, no relaxation and finally no kinematic hardening decrease take place. In the other tests, tempera-

ture and/or stresses are large enough to obtain efficient


relaxation and back stress drop. This mechanism has to
be modelled properly. In Sauzay et al. (2002), an evaluation of the required hold time for obtaining the end of
the relaxation of the stress is proposed. After that critical time, the residual thermal stress is negligible. If no
other mechanism is involved, and if the locking, dislocation microstructure homogeneization and relaxation
critical times are reached, the three stress components
and the fatigue-relaxation stressstrain behaviour can
be supposed to be saturated with respect to the hold
time.
Other mechanisms have to be studied. At 550 C
or 600 C, for large hold times (which are repeated),
intragranular precipitation can occurs which could also
modify the stressstrain behaviour.
The influence of the stress evolution with hold time
on the lifetime will now be studied.

232

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

Table 2
Number of cycles to failure for pure fatigue tests at 600 C (average
of several tests for each applied strain)

Table 4
Coefficients of the relation between the applied stress and the time
to failure (creep tests at 600 C)

t (%)

N25 ( 103 )

Coefficients

SP

SQ

Krupp

0.4
0.56
0.6
0.7
1.2
1.6

300
12
10
4
0.8
0.4

r
H (min)

9.64
1.63 1028

9.15
6.54 1026

11.21
1.36 1031

Creep stress unit: MPa.


pc

creep test, tF (), with an applied stress given by:


H
(4)
r
The coefficients r and H are given in Tables 4 and 5
(600 C and 550 C). Once more, the time is computed
using a linear accumulation rule:
 th
1
tFrelax = NF
dt
(5)
pc
t
0 F (t)
pc

tF () =

4.2. Application of the linear creep-fatigue


damage accumulation model
Several models propose to predict the number of
cycles to failure in fatigue-relaxation tests. The most
simple is based on the linear accumulation of creep
and fatigue damages. It corresponds to the following
equation:
NF
pf

NF

tF
tFrelax

=1

(3)

with NF the number of cycles to failure for a th hold


pf
time (and NF the number of cycles to failure without
hold time, i.e. for a pure fatigue test). The cumulated
hold time before failure is denoted as tF = NF th . If only
pure creep damage is considered, the failure would
take place after a cumulated time tFrelax which corresponds to a larger number of cycles than NF because
of additional fatigue damage. In this part, the hardening cycles are neglected and only saturated cycles are
considered. The linear rule can be applied using the
MansonCoffin curve (pure fatigue) and the results of
creep tests. The MansonCoffin curve gives the numpf
ber of cycles to failure for pure fatigue, NF , at the
same applied strain. The pure fatigue data are given in
Tables 2 and 3 (600 C and 550 C). Concerning the
creep damage, the time to failure corresponds to a pure
Table 3
Number of cycles to failure for pure fatigue tests at 550 C (average
of several tests for each applied strain)
t (%)

N25 ( 103 )

0.4
0.5
0.6
1.2
1.6

860
20
12
1.5
0.45

The comparison between the computed and observed times to failure are given in Tables 6 and 7 for
the two temperatures.
At 600 C, the predictions are quite reliable (Table 6
and Fig. 17). The ratio between predicted and measured values are included in the [1/3,3] range. For large
strains, fatigue and creep damages have the same order
of magnitude. Predictions seem to underestimate the
numbers of cycles to failure. For smaller strains, predictions overestimate them. In this case, the creep damage part becomes larger and the fatigue damage part
smaller. For the smallest applied strain (t = 0.4%),
the fatigue damage is very small (less than 0.03). The
creep damage is larger but remains small too (less than
0.3). The number of cycles to failure is overestimated
by a factor 3 for a 0.4% applied strain variation. It is
difficult to extrapolate the evolution of the lifetime with
hold time increasing.
At 550 C, the predictions are less reliable and depend on the strain level. The agreement between the
model and the experiment is quite good except for the
0.4% applied strain tests and the 1.2% test with the
Table 5
Coefficients of the relation between the applied stress and the time
to failure (creep tests at 550 C)
Coefficients

SP

SQ

r
H (min)

12
4.48 1035

12.6
1.07 1036

Creep stress unit: MPa.

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

233

Table 6
Comparison between the predictions of the linear and the measured numbers of cycles to failure (fatigue-relaxation tests, 600 C)
tmax (MPa)

R (MPa)

NF,comp

NF,exp

t = 1.6% (SP)
th = 90 min
th = 300 min

333
335

88
110

263
194

730
357

t = 1.2%
th = 300 min (SP)
th = 300 min (SQ)
th = 1,440 min (SP)
th = 1,440 min (SQ)

298
290
274
272

82
90
109
93

384
386
440
250

>314 (stopped)
248
398
212

t = 0.7%
th = 300 min (SP)
th = 300 min (SP)
th = 1,440 min (SP)

244
218
240

42
40
66

1,380
1,240
920

1,650
1,133
>100 (stopped)

t = 0.60.56%
th = 720min Krupp
th = 720 min (SQ)

212
198

72
41

2,900
3,100

1,290
>1,550 (stopped)

t = 0.4%
th = 90 min (SQ)
th = 30 min Krupp

187
175

16
11

20,300
36,000

6,500
10,900

The computation concerning the Krupp sheet use the results from its pure fatigue tests specifically. The maximum stress and the relaxed stress
are denoted as tmax , and R , respectively. They are measured at NF /2. The computed and experimental numbers of cycles to failure are denoted
as NF,comp and NF,exp , respectively.

Fig. 17. Experimental and computed numbers of cycles to failure (600 C).

234

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

Table 7
Comparison between the predictions of the linear and the measured numbers of cycles to failure (fatigue-relaxation tests, 550 C)
tmax (MPa)

R (MPa)

NF,comp

NF,exp

t = 1.2% (SP)
th = 600 min
th = 1,200 min
th = 4,000 min

344
306
310

54
66
95

700
1,140
1,000

360
312
217

t = 0.6% (SQ)
th = 90 min
th = 300 min
th = 300 min

281
270
277

17
15.5
15

2,500
1,300
970

2,250
>1,342 (stopped)
>248 (stopped)

t = 0.5% (SQ)
th = 90 min
th = 300 min

247
256

13.5
13

8,500
2,400

5,000
>317 (stopped)

t = 0.4% (SQ)
th = 30 min
th = 300 min

206
229

10
9

277,000
9,900

13,000
>512 (stopped)

The maximum stress and the relaxed stress are denoted as tmax , and R , respectively. They are measured at NF /2. The computed and experimental
numbers of cycles to failure are denoted as NF,comp and NF,exp , respectively.

Fig. 18. Experimental and computed numbers of cycles to failure (550 C).

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

largest hold time (4000 min) (Table 7 and Fig. 18). For
these particular tests, the stress level is small during
the largest part of the relaxation time ( 200 MPa,
Table 7). The used creep failure data (Table 5) correspond to larger stress levels. Specific data concerning
small stress levels are needed in order to check the reliability of the linear accumulation rule. The prediction
for 0.4% strain variation is very largely overestimated
(by a factor 20). It could be due to a lack of creep data
or to particular creep-fatigue interaction mechanisms
for small applied strains. Therefore, creep tests with
small applied stresses are needed leading to very long
tests. SEM observations could be useful in order to understand the specific creep-fatigue interaction for small
applied strains. A linear interaction rule and non-linear
pure fatigue (pure creep) damage models could be used
too (Cailletaud et al., 1984). For example the Kachanov
non-linear creep damage law can be used in addition
to a non-linear fatigue damage model (Lemaitre and
Chaboche, 1987). It could permit us to improve predictions at small strain and 550 C. A unified damage
law is used by Sermage et al. (2000). It is simple and
needs only a few material adjustable parameters. But it
seems not sure that the mechanisms of interaction between a fatigue crack and creep damaged grain boundaries near the crack tip are really taken into account
by those phenomenological models (Hales, 1980). It
should be noticed that several phenomenological models have been compared by Cailletaud et al. (1984).
The corresponding tentative extrapolations toward low
strain and long dwell time are scattered depending on
the chosen creep-fatigue damage model.

5. Conclusions
Many results have been obtained from fatiguerelaxation tests carried at 550 C on a 316LN steel. The
effect of small applied strains and long hold times has
been studied. All these test conditions are close to the
real power plant ones and lead to experimental results
which should be partially distinguished from the more
classical laboratory ones (higher temperature, larger
strains, smaller hold times).
The behaviour is affected by the hold time leading for long dwell times to a supplementary hardening
observed at small applied strains and little relaxation
is observed contrary to the results obtained for higher

235

temperatures. For large strains and long dwell times,


the saturated maximal stress is smaller with hold time
than without hold time.
The numbers of cycles to failure decrease when
the hold time is increased. This effect is larger than
for classical tests carried at 600 C and no saturation with respect to the dwell time is observed. The
damage effect could partially be explained by effect of dwell times on stressstrain behaviour. The
experimental behaviour results are discussed taking
into account several involved mechanisms (ageing,
recovery of the fatigue dislocation microstructures,
viscoplasticity . . .) and using TEM observations and
stress decomposition in kinematic, isotropic and thermal stresses. Concerning the fatigue-relaxation damage, a classical linear interaction model is applied and
discussed.

Acknowledgments
Pineau and X. Feaugas are acknowledged for stimulating discussions. This work is supported by the CEA
Program syst`emes du futur for a new generation of
nuclear power plants.

References
Argence, D., Pineau, A., 1995. Predictive metallurgy applied to
creep-fatigue damage of austenitic stainless steels. In: Proceedings of the Donald McLean Symposium, Structural Materials.
The Institute of Materials, pp. 229257.
Blanc, D., 1986. Effet de solute et deformation plastique dun acier
inoxydable austenitique. Th`ese de lEcole Superieure des Mines
de Paris, Paris, France.
Boulanger, L., Bisson, A., Tavassoli, A.A., 1985. Labyrinth structure
and persistent slip bands in fatigued 316 stainless steel. Phil. Mag.
A 51, L5L11.
Cailletaud, G., Nouailhas, D., Grattier, J., Levaillant, C., Mottot, M.,
Tortel, J., Escaravage, C., Heliot, J., Kang, S., 1984. A review of
creep-fatigue life prediction methods: identification and extrapolation to long term and low strain cyclic loading. Nucl. Eng. Des.
83, 267278.
Haasen, P., 1984. Mechanical properties of solid solutions and intermetallic compounds, third ed. In: Cahn, R.W., Haasen, P.
(Eds.), Physical Metallurgy, vol. II. North-Holland, Amsterdam,
pp. 13411409.
Delobelle, P., Oytana, C., 1986. Etude du comportement a`
haute temperature en plasticite-fluage dun acier inoxydable
austenitique (17-12SPH). J. Nucl. Mater. 139, 204227.

236

M. Sauzay et al. / Nuclear Engineering and Design 232 (2004) 219236

Ermi, A.M., Moteff, J., 1982. Correlation of substructure with timedependent fatigue properties of AISI 304 stainless steel. Metall.
Trans. A 13A, 15771588.
Essmann, U., Mughrabi, H., 1979. Annihilation of dislocations during tensile and cyclic deformation and limits of dislocation densities. Phil. Mag. A 40 (6), 731756.
Feaugas, X., 1999. On the origin of the tensile flow stress in the stainless steel AISI 316L at 300 K: back stress and effective stress.
Acta Mater. 47 (13), 36173632.
Feaugas, X., Gaudin, C., 2001. Different levels of plastic strain
incompatibility during cyclic loading: in terms of dislocation
density and distribution. Mater. Sci. Eng. A309/310, 382
385.
Gerland, M., Violan, P., 1986. Secondary cyclic hardening and dislocation structures in type 316 stainless steel at 600 C. Mater.
Sci. Eng. 84, 2333.
Gerland, M., Mendez, J., Violan, P., Ait Saadi, B., 1989. Evolution
of dislocation structures and cyclic behaviour of a 316L-type
austenitic stainless steel cycled in vacuo at room temperature.
Mater. Sci. Eng. A118, 8395.
Hales, R., 1980. A quantitative metallographic assessment of structural degradation of type 316 stainless steel during creep-fatigue.
Fat. Eng. Mat. Struct. 3 (4), 339356.
Lemaitre, J., Chaboche, J.-L., 1987. Mechanics of Solid Materials.
Springer-Verlag, Berlin.
Levaillant, C., Grattier, J., Mottot, M., Pineau, A., 1988. Creep and
creep-fatigue intergranular damage in austenitic stainless steels:
discussion of the creep-dominated regime. In: Solomon, H.D.,

Halford, G.R., Kaisand, L.R., Leis, B.N. (Eds.), Low Cycle Fatigue, ASTM STP 942. American Society for Testing and Materials, Philadelphia, pp. 414437.
Mottot, M., Petrequin, P., Amzallag, C., Rabbe, P., Grattier, J., Masson, S., 1982. Behavior in fatigue-relaxation of a high-creep resistant type 316L stainless steel. In: Amzallag, C., Leis, B.N.,
Rabbe, P. (Eds.), Low-Cycle Fatigue and Life Prediction, ASTM
STP 770. American Society for Testing and Materials, Philadelphia, pp. 152168.
Mughrabi, H., 1983. Dislocation wall and cell structures and longrange internal stresses in deformed metal crystals. Acta Metall.
31 (9), 13671379.
Prinz, F., Argon, A.S., Moffat, W.C., 1982. Recovery of dislocation
structures in plastically deformed copper and nickel single crystals. Acta Metall. 30, 821830.
Sauzay, M., Mottot, M., Allais, L., Noblecourt, M., 2002. A first step
in the modelization of the creep-fatigue behaviour based on dislocation microstructures. In: Blom, A.F. (Ed.), Proceedings of the
Eighth International Fatigue Congress (Fatigue 2002), EMAS.
Sermage, J.-P., Lemaitre, J., Desmorat, R., 2000. Multiaxial creepfatigue under anisothermal conditions. Fatigue Fract. Engng
Mater. Struct. 23, 241252.
Tavassoli, A.A., 1986. Dislocation concepts applied to fatigue properties of austenitic stainless steels including time-dependent
modes. Phil. Mag. A 54 (4), 521538.
Wood, D.S., Wynn, J., Baldwin, A.B., ORiordan, P., 1980. Some
creep-fatigue properties of type 316 steel at 625 C. Fat. Eng.
Mat. Struct. 3 (1), 3957.

You might also like