Creep fatigue behaviour of Type 321 stainless

steel at 650uC
M. W. Spindler*1, G. Knowles2, S. Jacques2 and C. Austin2
Type 321 austenitic stainless steel has been used in the UK’s advanced gas cooled reactors for a
wide variety of thin section components which are within the concrete pressure vessel. These
components operate at typically 650uC and experience very low primary stresses. However,
temperature cycling can give rise to a creep fatigue loading and the life assessment of these
cycles is calculated using the R5 procedure. In order to provide materials property models and to
validate creep fatigue damage predictions, the available uniaxial creep, fatigue and creep fatigue
data for Type 321 have been collated and analysed. The analyses of these data have provided
evolutionary models for the cyclic stress strain and the stress relaxation behaviour of Type 321 at
650uC. In addition, different methods for predicting creep fatigue damage have been compared
and it has been found that the stress modified ductility exhaustion approach for calculating creep
damage gave the most reliable predictions of failure in the uniaxial creep fatigue tests. Following
this, validation of the new R5 methods for calculating creep and fatigue damage in weldments has
been provided using the results of reversed bend fatigue and creep fatigue tests on Type 321
welded plates at 650uC in conjunction with the materials properties that were determined from the
uniaxial test data.
Keywords: Creep fatigue, Creep damage, Type 321, Austenitic stainless

This paper is part of a special issue on Creep–Fatigue Crack Development

Nomenclature Nf Number of cycles failure as defined by

complete separation
a0 Crack depth used to define crack N9g Number of cycles to grow a fatigue
initiation in a specimen or structure crack from the nucleation size of
ai Crack depth of 0?02 mm which is used 0?02 mm to a0
to define crack nucleation in a speci- Ni Number of cycles to nucleate a fatigue
men or structure crack of size of 0?02 mm
A1, P1, m1, n1 Parameters in the stress modified duc- Npred Predicted number of cycles to the
tility exhaustion model, equations (14) initiation of a creep fatigue crack
and (15) Nobs Observed number of cycles to the
b, B0 Parameters in the stress relaxation chosen failure criterion
equation (2) Rp0?2% 0?2% proof stress
dc Creep damage per cycle Rp1% 1% proof stress
df Fatigue damage per cycle Rm Tensile strength
Dc Total creep damage per test t Time
Df Total fatigue damage per test tf Time at failure
E Young’s modulus th Dwell time
N Cycle number T Temperature
N5% Number of cycles to a 5% decrease in Z Elastic follow-up factor
maximum stress Ds Stress drop during the creep dwell
N25% Number of cycles to a 25% decrease in s02s
maximum stress ec Creep strain that accumulates during a
N0 Continuous cycling fatigue endurance creep dwell
to create a crack of depth a0 e_ c Instantaneous creep strain rate
ef Uniaxial creep ductility
DeT Total strain range
1
EDF Energy, Barnett Way, Barnwood, Gloucester GL4 3RS, UK eU Upper shelf total inelastic strain at
2
AMEC, Walton House, Risley, Warrington, Cheshire WA3 6GA, UK failure
*Corresponding author, email mike.spindler@edf-energy.com s Instantaneous stress

ß 2014 W. S. Maney & Son Ltd

Received 9 June 2014; accepted 4 August 2014
284 DOI 10.1179/0960340914Z.00000000051 Materials at High Temperatures 2014 VOL 31 NO 4
 Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

sMax Maximum stress in the cycle methods in R53 for calculating creep damage and for
s0 Initial stress during the creep dwell taking account of the weldments on creep fatigue crack
initiation have been improved by introducing new
methods, which are outlined in Refs. 6–8. In order to
Introduction validate these new methods, specifically for Type 321
The failure of components due to the effect of primary stainless steel at 650uC, the earlier and previously
stresses is a well recognised problem in power stations unpublished UK nuclear industry research programme
and due allowance is taken in pressure vessel design has been revisited and the results of this new investiga-
codes for pressure and other primary stresses. In tion into these old data are reported in this paper. The
addition, in specialised elevated temperature compo- approach that has been used here is to utilise the results
nents, such as those in the turbines of fossil fuelled and of uniaxial creep, continuous cycling fatigue and creep
in the structural components of nuclear power stations fatigue tests on Type 321 to determine the appropriate
the possibility of thermal cycling giving rise to a creep materials properties at 650uC. Following this, validation
fatigue failure mechanism is also taken into account. of the new R53 methods for calculating creep and fatigue
For example, creep and fatigue damages that may arise damage in weldments has been provided using the
due to thermal mechanical cycling are calculated in high results of reversed bend fatigue and creep fatigue tests
temperature design codes, such as ASME III1 and RCC- on Type 321 welded plates at 650uC in conjunction with
MRx2 and life assessment procedures such as R53 and the materials properties that were determined from the
CRIEPI.4,5 uniaxial tests. Nevertheless, it should be noted that all of
The UK’s advanced gas cooled reactors use CO2 as the data in this paper are from isothermal testing and
the primary coolant at temperatures of up to 650uC and additional data at lower temperatures would be required
this coolant is contained by a prestressed concrete to assess the effects of thermal cycling in real plant
pressure vessel, which surrounds the reactor, the boilers components.
and the gas circulators. Type 321 austenitic stainless
steel has been used for a wide variety of thin section Experimental and test results
components which are within the concrete pressure The material started as a 12?7 mm thick plate to
vessel. These components operate at about 650uC and BS1501:Pt3:1973 Grade 321S12, which had been hot
experience low primary stresses. However, during rolled from cast M2265, and softened at between 1000
startup, shut-down and refuelling operations, these and 1120uC as required by BS1501:Pt3:1973.9 The
components experience thermal cycling, which can result chemical composition of cast M2265 is given in Table 1.
in a significant mechanical strain range and a creep
dwell. In particular, the mechanical strain range can be Uniaxial testing
significant within structures containing complex geome- Elevated temperature uniaxial tensile, creep, low cycle
tries, changes in section thickness and welds. Within fatigue (LCF) and creep fatigue tests were conducted
EDF Energy, the life of these Type 321 components and on specimens in different heat treatment conditions,
weldments is calculated using the R53 assessment although not all conditions were examined by all test
procedure, and specifically Volume 2/3, which addresses types. These conditions are summarised in Table 2 and
the initiation of creep fatigue cracking in an initially are given a code for simplicity. Although the material
defect free structure. was originally supplied in the softened state after heat
During the late 1980s and early 1990s, when much of treatment at between 1000 and 1120uC, it was common
the development and validation of R5 was undertaken, to resolution heat treat at 1050uC (RST) after specimen
the UK’s nuclear industry conducted a research project manufacture to eliminate any machining residual
into the creep fatigue behaviour of Type 321 stainless stresses. The ageing of 750uC for 200 h was intended
steel and weldments at 650uC. Nevertheless, recently the to give rapid precipitation of TiC and M23C6 carbides

Table 1 Chemical composition in wt-% and mechanical properties of 12?7 mm plate from cast M2265

C Si Mn P S Cr Ni Ti Co N

0.04 0.47 1.7 0.03 0.002 17.12 9.04 0.29 0.21 0.015
Grain size Room temperature mechanical properties
6–7 ASTM Rp0?2%/MPa Rp1%/MPa Rm/MPa Elong./%
Test 1 312 368 621 52
Test 2 328 376 639 54

Table 2 Summary of different heat treatment conditions of test specimens

First heat treatment Second heat treatment

As-Rec None (material condition as received from the supplier)

RST 1050uC for 30 min in vacuum after specimen manufacture None
Age1 1050uC for 30 min in vacuum after specimen manufacture 750uC for 200 h in air (oxide removed
by longitudinal polishing)
Age2 750uC for 200 h in air, blank aged before specimen None
manufacture
SimHAZ 1300uC for 10 min WQ 750uC for 200 h in air , blank aged before
specimen manufacture

Materials at High Temperatures 2014 VOL 31 NO 4 285

Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

and therefore, simulate extended service at 650uC and Uniaxial push–pull low cycle fatigue and creep fatigue
any oxide was removed from the specimen surface by tests were conducted at two laboratories on material in
longitudinal polishing. Other heat treatments were two of the heat treated conditions, both laboratories
applied to blanks before final machining. The heat used the same threaded end specimen design, which had
treatment of 1300uC for 10 min was intended to simulate a 15 mm parallel length and 7 mm diameter. The tests
the temperatures experienced in the heat affected zone were fully reversed (R ratio of 21) under strain control
(HAZ) of a weldment. with a stain rate of 0?1% s21 using a side contacting high
Some elevated temperature tensile tests were con- temperature extensometer over a 12 mm gauge length.
ducted at 600, 625 and 650uC and the results are given in The specimens were heated to the test temperature of
Table 3. Most of these tests were conducted with the 650uC by three zone resistance furnace and held to
typical straining and loading rates allowed in testing within ¡2uC for the duration of the test. For the creep
standards, although one test was conducted the slower fatigue tests a strain controlled dwell was introduced at
strain rate of 0?238 h21. the maximum strain for times of 1, 5, 23?3 or 24 h. The
Constant load creep rupture tests were performed at evolution of the maximum stress, minimum stress and
three different laboratories at 550, 600, 625 and 650uC. stress at the end of the dwells were recorded for between
The creep deformation behaviour of these tests has been 8 and 20 of the cycles. In addition, the stress relaxation
examined in;10 nevertheless, the failure data are relevant behaviour of a mid-life dwell was recorded as stress
to the calculation of creep damage and are given in versus time. Initial cyclic hardening was observed in all
Table 4. tests, with the LCF tests and the creep fatigue tests on

Table 3 Results of elevated temperature tensile tests on cast M2265

Temp. Strain rate Rp0?2% Rp1% Rm Elong. RofA

Condition uC h21 MPa MPa MPa % %

Age2 600 Not known 128 165 320 31 65

SimHAZ 600 Not known 169 205 328 25 60
Age2 625 Not known 127 160 315 31 65
As-Rec 625 Not known 114 148 331 34 74
SimHAZ 625 Not known 167 203 326 28 65
RST 650 0.238 145 … 271.9 … 58.04
RST 650 3.6 105 … 331 33.1 69.5
RST 650 3.6 116 … 324 35.5 71.2
Age2 650 Not known 122 155 303 36 70
As-Rec 650 Not known 121 152 321 39 75
SimHAZ 650 Not known 163 193 308 29 60

Table 4 Results of creep to rupture tests on cast M2265

Loading strain

Temp. Eng stress Plastic Total Elong. RofA Time at failure

Condition uC MPa % % % % h

RST 650 180 0.491 0.61 44.4 57 461.2

RST 650 160 0.214 0.32 65.4 66 978
RST 650 140 0.287 0.38 52.2 71 2880
RST 650 110 0.157 0.23 47 64 9314
Age1 650 179.4 5.313 5.431 29.7 63.7 124
Age1 650 160 2.008 2.114 36.2 53.1 354
Age1 650 138.3 0.581 0.673 52.6 59.2 2913
Age1 650 113.3 0.107 0.182 38.1 69.9 6670
Age1 650 92.3 0 0.061 Unfailed .5690
As-Rec 650 125 0.29 0.373 47 70 5138
Age2 650 140 0.7 0.793 71 70 610
Age2 650 125 0.18 0.263 66 70 1708
Age2 650 110 0.06 0.133 49 70 4240
Age2 650 100 0.045 0.111 46 48 14215
Age2 650 100 0.045 0.111 Unfailed .3525
Age2 625 155 0.6 0.701 61 62 1072
Age2 600 170 1.36 1.470 64 64 1773
Age2 550 215 3.12 3.256 37 38 2858
SimHAZ 650 155 0.15 0.253 71 65 173
SimHAZ 650 140 0.04 0.133 55 59 816
SimHAZ 650 125 0.01 0.093 52 60 1481
SimHAZ 650 110 0.005 0.078 39 50 6397
SimHAZ 650 100 0 0.066 23 25 23082
SimHAZ 625 155 0.11 0.211 49 52 868
SimHAZ 625 140 0.024 0.116 41 60 2589
SimHAZ 600 185 0.51 0.630 51 55 501
SimHAZ 600 170 0.11 0.220 38 45 2997

286 Materials at High Temperatures 2014 VOL 31 NO 4

 Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

1 Overall dimensions (in mm) of reversed bend test specimen

aged material exhibiting a stable cyclic state, whereas the conducted. The 12?7 mm plate was cold rolled down to
creep fatigue tests on solution treated material showed 3 mm thick in five stages and after each stage the plate
cyclic softening after the peak hardening had been was heat treated in argon at 1050uC for 30 min and after
reached. Most LCF tests were taken to complete the final stage the 3 mm plate was heat treated in
separation (Nf) and the evolutionary data for maximum vacuum at 1050uC for 30 min. Bend specimens were
stress was used to define N5% and N25%. For tests 50 mm wide, 100 mm parallel gauge length and 3 mm
exhibiting a stable cyclic state the percentage decrease in thick. The ends were widened to 58 mm for 70 mm to
maximum stress was used to define N5% and N25%, provide gripping (Fig. 1). These grip ends were clamped
whereas for tests exhibiting cyclic softening N5% and to the centreline of rotating shafts; one shaft was free to
N25% were defined by the number of cycles to a move horizontally to prevent axial loads when the
percentage decrease in stress from the trend of decrease specimen is heated or deflected during testing. Pure
in maximum stress. Three LCF and one creep fatigue displacement controlled bending was applied to the
test failed outside of the gauge length and failure is taken centre 100mm by a pneumatic cylinder, which acts via
as being the same for N25% and Nf, although there was torque arms and connecting rods to both rotating shafts.
sufficient data to estimate N5%. Two tests were The displacement range was controlled by fixed stops
discontinued without failure. The summarised test that limit the travel of the pneumatic cylinder. Before
results are given in Table 5. each test, the displacement range was calibrated by using
a strain gauged specimen to determine the total outer
Reversed bend testing of plate and weldments fibre strain range that would result from the applied
In addition, to the uniaxial LCF and creep fatigue tests, cyclic displacement. A single creep dwell was introduced
reversed bend LCF and creep fatigue tests have been into the cycle by holding the displacement at one of the

Table 5 Results of uniaxial push–pull low cycle fatigue and creep fatigue tests on cast M2265

Test DeT Dwell time N5% N25% Nf

Condition I.D. % h cycles cycles cycles

RST T6/12 1.5 0 244 280 316

RST T6/28 1.51 0 314 352 440
RST T6/14 1 0 1514 1730 2005
RST T6/15 1 0 1976 2165 2220
RST T6/16 0.6 0 7071 7850 8850
RST T6/26 0.6 0 12 560 13 100 14 635
RST T6/17 0.39 0 85 080 85 080 85 080
RST T6/18 0.39 0 190 500 194 800 201 100
Age1 S12/12 0.35 0 234 525 236 100 236 850
Age1 S13/26 0.4 0 74 540 76 800 79 370
Age1 S13/25 0.6 0 8932 9260 9804
Age1 S13/22 0.62 0 10 336 11 371 11371*
Age1 S12/10 1 0 1983 2053 2053*
Age1 S12/11 1 0 1729 1842 1842*
RST T3/66 1.5 1 70 70 94
RST T2/66 1.5 23.3 115 118 153
RST T2/65 1 1 228 239 267
RST T3/55 1.02 5 199 217 257
RST S3/18 0.6 1 1221 1277 1311
RST S3/20 0.58 23.3 475 475 475
Age1 S2/131 0.4 1 3823 4271 4988
Age1 S12/14 0.6 1 1623 1720 2025
Age1 T6/29 1 1 554 564 564*
Age1 T3/69 0.6 24 .275 … Unfailed
Age1 T6/30 1 24 .125 … Unfailed
*Tests in which failure occurred outside of the extensometer gauge length.

Materials at High Temperatures 2014 VOL 31 NO 4 287

Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

2 Reversed bend weldment configurations (dimensions in mm)

limits. Clearly, this introduces a tensile dwell on one longitudinal welds used Type 347 filler and fillet welds
surface and a compressive dwell on the opposite surface. used Type 316L filler. All welds were made in two passes
The weldment configurations that were tested include and the welding process was tungsten inert gas weld
plain parent plate, butt welded plate, longitudinal roots and caps were left undressed, as with plant
welded plate and parent plates to which a fillet weld components, and in the butt and longitudinal welds
had been applied to one surface. The weldment the weld caps were proud of the parent plate. In
configurations are shown in Fig. 2 and are summarised common with the uniaxial testing, the parent specimens
in Table 6. Two types of weld filler metal were used were tested in either the solution treated condition or
for the butt welds, Type 347 and Type 316L, the aged at 750uC for 200 h. The welds made in solution

Table 6 Test matrix for reversed bend tests on cast M2265

Condition Weldment configuration Weld filler metal Weldment dwell

RST Parent … …
Age1 Parent … …
AW Butt weld Type 347 Weld cap in tension
Age1 Butt weld Type 347 Weld cap in tension
AW Butt weld Type 316L Weld cap in tension
Age1 Butt weld Type 316L Weld cap in tension
AW Longitudinal weld Type 347 Weld cap in tension
Age1 Fillet welds Type 316L Fillet on tension side
Age1 Fillet welds Type 316L Fillet on compression side

288 Materials at High Temperatures 2014 VOL 31 NO 4

 Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

3 Results of reversed bend continuous cycling and creep fatigue tests at 650uC. a results of continuous cycling tests; b
results of creep fatigue tests with 1 h dwells; c results of creep fatigue tests with 5 h dwells; d results of creep fatigue
tests with 10, 23?3 or 24 h dwells

treated plates and were tested in either the as welded models), creep failure and fatigue endurance data are
condition (AW) or after the ageing treatment of 750uC used to calculate creep and fatigue damage respectively.
for 200 hours. Reversed bend, R-B, (R ratio of 21) However, since the objective of this paper is simply to
continuous cycling fatigue and creep fatigue tests were validate the calculation of creep and fatigue damage in a
conducted on each weldment configuration at 650uC, range of isothermal tests only the properties at 650uC are
although the strain ranges used varied (between 2 and required.
0?25%) and dwell times were either 1, 5, 10 or 24 h long.
The dataset of reversed bend tests is particularly large Cyclic stress strain of Type 321 at 650uC
with 45 continuous cycling tests (173–2?76107 cycles) The main use for cyclic stress strain data in this paper is
and 50 creep fatigue tests (53–10 000 cycles) including 10 to estimate the start of dwell stress in the reversed bend
tests with endurances longer than 5000 h. Of particular tests. The uniaxial test data showed that the maximum
significance is that 85 of the reversed bend tests were on stress evolved with cycling, showing initially cyclic
weldments. The results of the reversed bend tests are hardening and then the continuous cycling tests and
summarised in Fig. 3. Most tests were taken to complete creep fatigue tests on aged material showed cyclic
separation and although load and displacement data stabilisation, whereas the creep fatigue tests on solution
were recorded during testing these data have not be used treated material softened after the initial hardening. The
in the current investigation. Therefore the only data that other significant observation was that the solution
will be used in this investigation from the reversed bend treated material started stronger and hardened more
tests are the test condition and the number of cycles to than the aged material. All of the available uniaxial
failure. Nevertheless, this is sufficient here as these data cyclic data from the creep fatigue tests have been fitted
will be used as validation of the new R53 methods for simultaneously using non-linear regression to determine
calculating creep and fatigue damage in weldments when a relationship for the evolution of the maximum stress as
used in conjunction with the materials properties that a function of cycle number, total strain range, dwell time
will be determined from the uniaxial tests.
and whether the initial heat treatment was solution
treated or aged. Note no attempt has been made to
Determination of deformation properties include the continuous cycling data in the fits. This is
from uniaxial data because the application of these data is to predict the
start of dwell stress for the reversed bend creep fatigue
In order to calculate creep and fatigue damage in a plant
tests. The equation is
component, it is necessary to estimate the total strain
  
range which arises from a thermal cycle; this requires sMax ~A exp(BNth DeT )zC0 DeC D1
T 1{exp {D0 De T N
1

cyclic stress strain data at a range of temperatures. In

(1)
addition, cyclic stress strain data at the dwell tempera-
ture are used to estimate the start of dwell stress. and the values of the materials parameters (A, B, C0, C1,
Furthermore, stress relaxation (or creep deformation D0 and D1) are given in Table 7.

Materials at High Temperatures 2014 VOL 31 NO 4 289

Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

4 Evolution of maximum stress showing test data and fit to equation (1) during cyclic testing of Type 321 at 650uC

Whenever possible, the parameters in the model fitting on the assumption that B0 evolved as the creep fatigue
were kept the same for both heat treatment conditions cycling progressed a model was constructed to predict
(A, B and D1) and were only allowed to differ if there stress relaxation behaviour for Type 321 at 650uC. In
was a significant improvement in the predictions. It can general, the stress drop increased as the initial stress
be seen from Fig. 4 that the model gives a reasonable increased. However, the stress drop continued to
prediction of the cyclic evolution in all tests. increase after stabilisation had occurred and sometimes
even continued to increase during cyclic softening. In
Stress relaxation of Type 321 at 650uC those examples where the stress drop decreased, this
Ideally stress relaxation data for a range of cycles would decrease was smaller than the decrease in the initial
be available; however, for these old tests, the only stress. These observations indicate that creep softening is
evolutionary data that were available was the initial occurring in subsequent dwells. The other important
stress (the same as the maximum stress used for the observation was that the value of b varied between tests
cyclic stress strain data) and stress at the end of the dwell and was correlated with the total strain range. Because
for between 8 and 20 of the creep fatigue cycles of each total strain range was one of the few known test
test. Nevertheless, the stress relaxation behaviour of a variables in the reversed bend tests, it was chosen as the
mid-life dwell had been fitted to the Feltham equation, explanatory variable for the variation of b. With regard
which is given by to the evolution of B0, there are a number of potential
explanatory variables; however, these are all correlated
s~s0 {s0 B00 ln(btz1) and for pure strain control with one other, such as cycle number, total time and
ec ~(s0 {s)=E (2) accumulated creep strain. It was decided that the
explanatory variable that would be most useful when
and the parameters B0 and b were available. It was used for a wide range of different creep-fatigue cycles
therefore decided to use the available parameters for the was the total accumulate creep strain from all preceding
mid-life cycle and to vary only B0 in the Feltham dwells (gec). However, a value of B0 is also required for
equation to give the measured stress at the end of the the first cycle where the creep strain starts at zero and it
dwell. This approach makes the assumption that the was decided to treat the first cycle value of B0 as a
slope of stress drop versus the logarithm of time evolves, function of the total strain range. Therefore, the
but that the overall shape of the stress relaxation following expressions for B0 and b were used when
behaviour remains similar. This gives a best estimate refitting the Feltham equation to the available stress
to the measured (albeit limited) relaxation behaviour for drop (s0–s) data from cycle 2 onwards.
individual uniaxial creep fatigue tests. However, it does
not give a general relaxation model that could be used to b~b0 DebT1 (3)
predict relaxation in the reversed bend tests. For this
reason, a second approach was also adopted. Building If cycle is 1 and ec ~0, B’’~b0 zb1 DeT (4a)

Table 7 Values for the parameters in equation (1) which describes cyclic evolution of maximum stress, where sMax is in
MPa, th is in hours and DeT is in %

Condition A B C0 C1 D0 D1

RST 107.5134 9.9707610 25

4.7869 0.728753 20.99916 0.74893
Age1 107.5134 0 4.3302 0.728753 21.423843 0.74893

290 Materials at High Temperatures 2014 VOL 31 NO 4

 Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

5 Comparison of original stress relaxation fits to mid-life dwell data with predictions made using equations (2)–(4)

6 Comparison of measured evolution in stress drop with fits made using equations (2), (3) and (4)

X 
For subsequent cycles, B’’~B0 zB1 log ec (4b) (Fig. 5). In addition, the measured stress drops from the
uniaxial creep-fatigue tests have been compared with
those fitted using equations (2)–(4) (Fig. 6).
The parameters in equations (3) and (4) are given in
Table 8. It can be noted that during fitting it was found
that the same set of parameters would fit the relaxation Determination of failure data from
data from both heat treatment conditions. uniaxial tests
Since the equations (2), (3) and (4b) were only fit to
The calculation of fatigue and creep damage requires
the stress drop at the end of the dwell, the goodness of fit
descriptions of fatigue endurance and either creep
of this equation set has been tested by comparing the
ductility or time to rupture depending on whether
relaxed stresses for the mid-life cycles that are calculated
ductility exhaustion or time fraction is used to calculate
from the original available values for B0 and b with
creep damage.
the relaxed stresses predicted using equations (2)–(4)
Fatigue endurance of Type 321 at 650uC
Table 8 Values for parameters used in equations (2), (3) For this paper, the fatigue endurance has been defined in
and (4) to predict stress relaxation in Type 321
two different ways. The first approach is specifically for
at 650uC, where stress is in MPa, DeT is in % and
time is in minutes calculating a cast specific best estimate fatigue damage
in the uniaxial creep fatigue tests at N5% and the second
b0 b1 b0 b1 B0 B1 approach is a general approach that can be used with the
reversed bend tests and to give a lower bound estimate
56.0513 2.76714 0.042723 20.013902 0.066776 0.010254
of fatigue endurance.

Materials at High Temperatures 2014 VOL 31 NO 4 291

Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

to the values for continuous cycling N5% and Nf which

are given in Table 5. The fitting used a log normal error
distribution and the resulting parameters are given in
Table 9 and a comparison between the data and the
models is given in Fig. 7.
Since equation (5) has only been fitted to data from a
single cast, it is not suitable for engineering assessments
of multiple cast of material using lower bound properties
to give a conservative estimate of fatigue damage. For
this reason, EDF Energy’s usual fatigue endurance
model for austenitic stainless steels, which gives Nf, has
also been used,11 to calculate general mean and lower
bound fatigue endurances. Nevertheless, it was found
that cast M2265 exhibited greater endurance than the
mean would have predicted and therefore the mean
fatigue endurance model has been adjusted by multi-
plying the total strain range by 1?17. An estimate of the
general lower bound fatigue endurance has been made
by factoring the adjusted mean by multiplying the total
strain range by 0?75. The general fatigue endurance
predicted by this model is shown in Fig. 7b, which also
shows the results of the continuous cycling reversed
bend tests. It can be seen from Fig. 7b that the
7 Continuous cycling fatigue endurance data and models continuous cycling fatigue endurances of the reversed
for Type 321 at 650uC. a uniaxial continuous cycling fati- bend tests are similar to those of uniaxial LCF tests.
gue N5% with fit to equation (5); b uniaxial and reversed
bend continuous cycling fatigue Nf with predictions from Creep rupture strength for time fraction
the general fatigue endurance model (note there has been The creep rupture strength of Type 321 stainless steels
no adjustment for specimen thickness by changing a0) which have been heat treated in the range 950–1070uC is
provided in PD6525.12 The data for cast M2265 and the
The cast specific best estimate fatigue endurance has mean line from PD6525 along with a lower bound given
been obtained by fitting the simple low cycle fatigue by 220% on stress are shown in Fig. 8. In addition,
equation Fig. 8 shows published13 creep rupture data for Type
 321, which have been heat treated in the range 1000–
ðDeT {a2 Þ 1=a1 1050uC. It can be seen from Fig. 8 that the creep rupture
NX% ~ (5) strength of cast M2265 at 650uC is comparable with that
a0
of other casts of Type 321 which have been heat treated
at similar conditions. In addition, it can be seen that the
published rupture strength model gives a realistic
Table 9 Values for parameters used in equation (5) to prediction of the failure times of cast M2265.
predict mean continuous cycling fatigue
endurance of Type 321 at 650uC: DeT is in %
Creep ductility for ductility exhaustion
a0 a1 a2 Within R53 creep ductility is defined as the creep strain
at failure (ef), which is calculated from the elongation
N5% 9.94083 20.349569 0.217904 minus the plastic loading strain as a function of the
Nf 11.83738 20.365023 0.221178
average creep strain rate (ef/tf). However, for simplicity

8 Creep rupture strength of cast M2265 Type 321 at 650uC

292 Materials at High Temperatures 2014 VOL 31 NO 4

 Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

9 Creep ductility of cast M2265 Type 321 at 650uC: note data from Ref. 13 are elongation values because plastic loading
strains are not reported

in assessments of plant components, this is often

simplified to a mean and lower bound ductility, which df ~1=N0 ðDeT ,T Þ (6)
is assumed to be a function of temperature only. where N0 is the continuous cycling fatigue endurance to
Therefore, for plant assessments, the creep ductility of create a crack of depth a0 at the appropriate total strain
Type 321 at 650uC is often assumed to be given by a range DeT and temperature T. The total fatigue damage
mean of 19% and a lower bound of 5?9%.11 It can be Df is then summed for the total number of cycles and
seen from Fig. 9 that cast M2265 exhibits high creep taking account of any cycles with different strain ranges
ductility when compared with this mean and lower that result in different damages per cycle. In the R5
bound and has high creep ductility compared with procedures, the fatigue strain range is enhanced for the
published data from.13 Nevertheless, it can also be seen presence of a creep dwell by adding the relaxation strain,
from Fig. 9 that the assumption of creep ductility being which for pure strain control is the same as the creep
independent of strain rate is reasonable and that the relaxation strain ec, and can be calculated from the stress
lower bound of 5?9% compares well with the published drop Ds, which was measured in these tests divided by
data from.13
E. This increases the total strain range used for the
fatigue calculations and is done to take account of the
Calculation of creep and fatigue damage effect of the dwell on increasing the inelastic strain range
uniaxial tests (which can be visualised as the width of the hysteresis
loop at zero stress). It should be noted that this
The models to predict creep fatigue failure that have
enhancement by the relaxation strain is not done in
been investigated in this paper calculate total creep
the design codes ASME III1 and RCC-MRx.2
damage Dc and total fatigue damage Df separately.
Nevertheless, it has been done here for all fatigue
Failure is conceded when the sum of the fatigue and
damage calculations so that the predictions of creep
creep damages reaches some failure criterion; for
damage can be compared fairly.
example, in R5,3 a linear damage summation is used
For the analysis of the uniaxial creep fatigue test data, it
and failure is conceded when DczDf§1. It is usual for
the failure criterion to vary depending on the method is important that the same failure criterion is used for both
used to calculate the creep damage and the interaction the continuous cycling fatigue and creep fatigue tests, as
diagram used in ASME III1 and RCC-MR,2 for this ensures that the crack sizes are approximately the
austenitic steels, is bilinear with a locus at (0?3,0?3). same. Therefore, with the cast M2265 Type 321 data, the
For the interpretation of uniaxial creep fatigue test best estimate N5% has been calculated from equation (5)
data, mean fatigue endurance and mean creep failure with the appropriate parameters from Table 9.
data for the same cast of steel have been used. Thus, best For plant analyses of thin section components a0 is
estimates of creep and fatigue damages were calculated. often set to 10% of the cross-section so that an uncra-
For plant assessments, lower bound fatigue endurance cked body stress analysis is appropriate. However, for
and creep failure data, which have been derived from multiheat datasets, only the number of cycles to the
multiheat datasets, would be used to give conservative complete separation of a test specimen Nf is commonly
upper bound estimates of the damages. These upper reported, which can be larger than N0. Thus, for plant
bound estimates of creep and fatigue damage have also analyses, an empirical model for fatigue crack nuclea-
been calculated for the uniaxial creep fatigue tests. The tion followed by a short fatigue crack growth law is
methods for calculating fatigue and creep damage are used to calculate N0 from lower bound values of Nf. This
described in the following sections. approach has also been used here in conjunction with
EDF Energy’s general mean and lower bound fatigue
Fatigue damage endurance Nf model for austenitic stainless steels,11
The approach for the calculation of fatigue damage per which has been adjusted by multiplying the total strain
cycle df in R53 is simply range by 1?17, which make the mean comparable with

Materials at High Temperatures 2014 VOL 31 NO 4 293

Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

the data for cast M2265 (Fig. 7b) and a factor of 0?75
for the lower bound. This fatigue endurance model is
used to calculate Ni the number of cycles to nucleate a
crack of size ai, the nucleation size of 0?02 mm, which is
calculated from the empirical equation
:
lnðNi Þ~lnðNf Þ{8:06Nf{0 28 (7)
where Nf is the number of cycles to failure, as defined by
complete separation. Equation (7) is a simplification of a
model due to Pineau.14 For a 7 mm diameter cylindrical
specimen, a single thumb nail crack of depth 0?852 mm
equates to a 5% reduction of cross-section area. The
number of cycles to grow a crack from 0?02 to
0?852 mm, Ng, is then calculated from an integration
of the two stage short fatigue crack growth law, which is
given by
da
for 0:02ƒaƒamin , !amin , where amin
dN
da (8)
~0:2 mm and for a§amin , !aQ
dN
where Q is a material parameter, which has been found
experimentally to vary between 1 and about 2. In this
case where Q is unknown a simplifying assumption has
been made to set Q51 which also means that the
coefficient of proportionality cancels out when inter-
polating between Ni and Nf (where a57 mm) to find N’g .
The endurance N0 which equates at a crack of size
a050?852 is therefore given by NizN’g .
10 Best estimate fatigue damage and time fraction creep
Time fraction approach damage. a interaction diagram; b Endurance Nobs/
The time fraction rule for the estimation of creep Npred versus start of dwell stress for mid-life cycle
damage is widely used in design codes such as ASME
III1 and RCC-MR.2 The time fraction rule for the creep
damage per cycle is where for pure strain control, the creep strain ec can be
ð calculated from the stress drop Ds, which was measured
th dt
dcTF ~ (9) in these tests divided by the Young’s modulus E.
0 tf (s,T)
where tf is the creep rupture time and is calculated as a Results creep and fatigue damage calculations
function of the stress and temperature, which prevail for for uniaxial creep fatigue tests
the increment of time dt during stress relaxation and th is The results of the best estimate creep and fatigue
the dwell time. For the analysis of the uniaxial creep damage calculations for time fraction and ductility
fatigue tests, the stress history during relaxation was exhaustion are shown graphically as damage interaction
described by the best estimate values for B0 and b which diagrams and total damage versus start of dwell stress in
were derived from the individual uniaxial creep fatigue Figs. 10 and 11. The corresponding figures for the upper
tests. The damage per cycle is obtained by summing the bound damages (using lower bound creep failure and
increment of damage calculated from 1000 steps spaced fatigue endurance data) are shown in Figs. 12 and 13.
in log time. It is useful to note that for the linear interaction Npred
equals 1/(dfzdc) and hence, the total damage (DfzDc)
R5 ductility exhaustion approach equals Nobs/Npred. Therefore, graphs of total damage or
The ductility exhaustion approach of R5 to calculate the Nobs/Npred simultaneously show how well both the
creep damage per cycle dcR5 is given by endurance and also the total damage are predicted and
ð th in particular because the fatigue damage is always the
e_ c same these graphs show how well the creep damage is
dcR5 ~ dt (10)
0 ef ðe_ c ,T Þ predicted as a function of the initial stress. With regard
where e_ c is the instantaneous creep strain rate and to the graphs for the time fraction, the calculation of
ef ðe_ c ,T Þ is the corresponding creep strain at failure at Npred is a little more complicated as the bilinear
the appropriate temperature T as a function of the creep interaction needs to be taken into account. For an
strain rate. Nevertheless, when the ductility is assumed interaction locus at (c,f), where c is the coordinate of the
to be independent of strain rate as in this case (Fig. 9), creep damage and f is the coordinate of the fatigue
equation (10) simplifies to damage; then for constant amplitude testing, this leads
to
ec
dcR5 ~ (11)
ef ðT Þ if df =dc §f =c then Npred ~fdf z½ð1{f Þ=cdc g{1 (12a)

294 Materials at High Temperatures 2014 VOL 31 NO 4

 Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

13 Interaction diagram showing upper bound fatigue

damage and upper bound simple ductility exhaustion
creep damage

(Fig. 11a) and became increasing conservative as the start

of dwell stress decreases (Fig. 11b), which also corre-
sponds to the longer observed endurances. The trends
that have been observed for the best estimate creep and
fatigue damages are also shown by the upper bound
estimates of damage (Figs. 12 and 13). The only
observation of concern is that the calculated upper bound
11 Best estimate fatigue damage and simple ductility damages are all very high, with all but one test being
exhaustion creep damage. a interaction diagram; b total greater than a factor of two above the appropriate
damage versus start of dwell stress for mid-life cycle interaction diagrams. In view of the observed over-
conservatism of the ductility exhaustion approach and
specifically because it was clear that there was an effect of
if df =dc ƒf =c then Npred ~f½ð1{cÞ=f df zdc g{1 (12b) stress on the calculated creep damage which was not
being taken into account (as shown in Figs. 11b and 13b),
If can be seen from Fig. 10 that the time fraction it was decided to develop a ‘stress modified ductility
approach gives a reasonable prediction of the best exhaustion’ model6–8 for Type 321 stainless steel.
estimate creep and fatigue damages, with an interaction
Stress modified ductility exhaustion model for
locus at (0?3,0?3) (Figs. 10) and therefore the predicted
endurances (Fig. 10b) are reasonable also. However, the
Type 321
ductility exhaustion approach was overly conservative The ‘stress modified ductility exhaustion’ (SMDE)
approach has recently been included into the R53
procedure and has previously been shown to give
improved estimate of creep damage for wrought Type
316H, cast Type 304L and Type 347 weld metal.6,7 In
particular, The ‘stress modified ductility exhaustion’
approach gives less pessimistic predictions of creep
damage than the simple ductility exhaustion approach,
for cycles with low stress creep dwells and therefore
should be applicable to Type 321 at 650uC.
The ductility exhaustion approach of R5 can be
simply modified to include the effect of stress on creep
damage by treating the ductility as a function of both
strain rate and stress. The resulting expression for the
creep damage per cycle calculated from a ductility
exhaustion approach, which includes the effect of stress
dcSM is given by
ðth
e_ c
dcSM ~ dt (13)
ef ðe_ c ,s,T Þ
0

12 Interaction diagram showing upper bound fatigue where th is the dwell time, ef ðe_ c ,s,T Þ is the inelastic
damage and upper bound time fraction creep damage strain at failure at the appropriate temperature as a

Materials at High Temperatures 2014 VOL 31 NO 4 295

Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

uniaxial creep fatigue test data. First, equation (14) was

simplified by removing the upper shelf ductility eU. This
is appropriate for creep fatigue tests because the upper
shelf ductility corresponds to necking followed by
ductile shear failures, which is a type of failure which
cannot physically occur during a strain controlled creep
fatigue cycle. Secondly, because creep fatigue data are
only available at 650uC the temperature parameter was
removed and therefore, equation (14) becomes
ef ~A1 e_ c n1 s{m1 (15)

The instantaneous values of stress and strain rate

during the dwell are calculated from equation (2) and
from the differentiated version of equation (2) respec-
tively, for 1000 steps spaced in log time. Determination
of the material parameters A1, n1 and m1 was carried out
using non-linear regression to minimise the sum of
squares of the logarithm of (DfzDc), where Df is the
best estimate cast specific fatigue endurance calculated
from N5%. It should be noted that this approach is not
being recommended here for general use to derive the
parameters for use with the SMDE approach. It has
only been used here due to the limitations in the
available cast specific creep data. The values that have
been derived for the material parameters A1, n1 and m1
are given in Table 10.
As expected and as shown in Fig. 14, the SMDE
calculation of creep damage using equation (15) with the
parameters from Table 10 gives a very accurate predic-
14 Best estimate fatigue damage and stress modified tion of the creep and fatigue damages for the uniaxial
ductility exhaustion creep damage. a interaction dia- tests on Type 321. However, since this model has been
gram; b total damage versus start of dwell stress for developed from a single cast, it is not appropriate to
a mid-life cycle calculate upper bound creep damages. Therefore,
equation (14) has been fitted to the published13 elonga-
function of both the creep strain rate and the stress. It tion at failure data for Type 321 heat treated in the range
has been shown experimentally6,7 that the uniaxial creep 1000–1050uC, albeit a temperature dependent upper
ductility of austenitic and ferritic steels can be repre- shelf has been used which is given by eU ~A2 ðP2 =T Þ.
sented by The reliability of fit of this model is shown in Fig. 15,

 which also shows that the upper and lower 95%
P1 n1 {m1 prediction intervals are given by the mean multiplied
ef ~Min eU , A1 exp e_ c s (14)
T by or divided by 2?142 respectively. It is now possible to
use this factor to calculate an upper bound creep
where A1, n1, m1 and P1 are material parameters. It damage, where the upper bound fatigue damage is as
should be noted that the upper shelf ductility eU is before, for the SMDE model and the creep fatigue
necessary for high ductility constant rate and creep tests damage predictions are shown in Fig. 16.
that fail by necking followed by ductile transgranular It is clear from comparing Fig. 14 with Figs. 10 and
failure, either by shear at 45u or by ‘cup and cone’. 11 and from comparing Fig. 16 with Figs. 12 and 13
Unfortunately, an examination of the creep ductility that the SMDE approach gives more realistic predic-
data for cast M2265 in Fig. 9 showed that the available tions of creep and fatigue failure for Type 321 at 650uC
creep data for cast M2265 would be unsuitable to fit than both the time fraction and the simple ductility
equation (14) to the available data. This is because the exhaustion approaches. In particular, the simple ductility
vast majority of the data for cast M2265 appear to fall exhaustion gives overly conservative predictions for both
on the upper shelf with high ductilities and there would best estimate and upper bound creep damages. With
be insufficient data to determine the remaining four regard to the time fraction approach, the best estimate
material parameters A1, n1, m1 and P1. It was therefore predictions are good; however, the upper bound predic-
decided to use a rather different approach and to derive tions are more pessimistic than the SMDE approach.
the materials parameters in equation (14) using the

Table 10 Values for parameters used in equation (15) to Calculation of creep and fatigue damage
predict mean creep damage using SMDE for
: reversed bend tests
Type 321 at 650uC: ec is in h21 and s is in MPa
Now that it has been established that the SMDE
A1 n1 2m1 approach is the preferred method to calculate creep
damage for Type 321 at 650uC, the method can be
loge (40.8485) 0.921167 27.142789
validated using the results of the reversed bend tests on

296 Materials at High Temperatures 2014 VOL 31 NO 4

 Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

15 Observed versus predicted elongation for published data on Type 321 fitted equation (14), with temperature depen-
dent upper shelf

parent and weldments. This is a particularly important and (8) where a0 has been taken to be 3 mm, which
activity because first, the SMDE model was actually corresponds to complete separation of the reversed bend
optimised to the uniaxial creep fatigue test data and tests. With regard to the weldment fatigue damage this
therefore, requires validation against some independent has used the latest recommendations in R5.3,8 First, the
test data and second, because the main application of weld endurance reduction (WER), which accounts for
the creep failure damage calculations for Type 321 at the fatigue endurance reduction due to the presence of
650uC will be to estimate the life of thin section small imperfections (e.g. inclusions, porosity, etc.) in the
weldments. Furthermore, this is a particularly valuable weldment constituent materials, is applied. This is
activity because there is a large dataset of tests on carried out by removing the number of cycles to nucleate
reversed bend specimens. The methods that have been a small crack of depth 0?02 mm, Ni, from the fatigue
used to calculate fatigue damage and creep damage in endurance and therefore, N0 is taken to be equal to N’g
reversed bend tests are now outlined: (rather than NizN’g ). Second, an appropriate weld
strain enhancement factor (WSEF), which accounts for
Fatigue damage strain enhancement due to the weldment geometry (if
The fatigue endurance has been calculated using the applicable) and the material mismatch between weld-
general mean and lower bound fatigue endurance Nf ment zones, is applied. For the butt and longitudinal
model for austenitic stainless steels,11 which has been welds in austenitic stainless steels, the recommended
adjusted by multiplying the total strain range by 1?17 for WSEF in R53 is 1?16 and for fully penetrated fillet welds,
Type 321 and a factor of 0?75 for the lower bound. it is 1?23. The WSEF is applied to the total strain range
However, the effect of the small thickness 3 mm of the at which the fatigue endurance is to be calculated. Note
reversed bend specimens compared with the 7 mm for the creep fatigue tests, the total strain range includes
uniaxial round bar specimens has been taken into the relaxation strain, which has been calculated as
account using the methods outlined for equations (7) outlined below in the section on creep damage. Further

16 Upper bound fatigue damage and upper bound stress modified ductility exhaustion creep damage. a interaction dia-
gram; b total damage versus start of dwell stress for mid-life cycle

Materials at High Temperatures 2014 VOL 31 NO 4 297

Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

17 Observed versus predicted endurance for reversed bend fatigue tests (adjusted by multiplying total strain range by
1?17 and factor of 0?75 for lower bound, with a053 mm for reversed bend and a057 mm for uniaxial tests). a predic-
tions using general mean fatigue endurance; b predictions using the general lower bound fatigue endurance model

details of the new R5 method to calculate fatigue Results creep and fatigue damage calculations
endurance in weldments are given in R53 and in Ref. 8. for reversed bend creep fatigue tests
In particular, it should be noted that this includes With regard to validating methods for calculating creep
methods for performing the simplified inelastic analysis and fatigue damage, it is important to tend to under-
to estimate the total strain range. However, since total estimate damage. This is because for test data, total
strain range was known for all of the reversed bend tests, damage equates to Nobs/Npred; therefore, if the total
these additional methods were not necessary here. It is damage is greater than the failure criterion given by the
therefore important to recognise that the validation interaction diagram, then this means that the observa-
contained in this paper relates to the materials data tion is larger than the prediction, which shows the
aspects of calculating weldment fatigue damage for Type prediction to be conservative. For plant assessments, the
321 at 650uC and do not validate the inelastic analysis total damage equates to the number of service cycles
methods in R5, for which the current validation is given (Nserv) divided by Npred. Therefore, in a plant assess-
within the R5 procedure.3 Predictions of fatigue ment, a total damage that is greater than the failure
endurance for the continuous cycling reversed bend criterion means that Nserv is greater than Npred and
tests on parent and weldments are shown in Fig. 17. It therefore failure is predicted. Conservatism is therefore
should be noted that for completeness, the uniaxial test assured by providing realistically bounding low values of
data have been included in Fig. 17 and the fatigue Npred, while avoiding over pessimism. Therefore, in this
endurance has been calculated using the same fatigue paper, the value of a0 has been set to the section
endurance model that has been used for the reversed thickness, which gives a larger value of Npred for a test.
bend tests, albeit with a057 mm. However, for a real plant assessment, a smaller value of
a0 would normally be chosen such as 10% of the
Creep damage thickness, which would be 0?3 mm for these reversed
For the reversed bend tests, there was no record of the bend tests. Other ways to maximise Npred for a test
maximum stress in the cycle, which could be used for the include ignoring the relaxation strain when calculating
start of dwell stress and there were no stress relaxation fatigue damage. This was not carried out in this paper
data. Therefore, cycle by cycle creep damages have been because of the need to fit the SMDE model to the
calculated using the start of dwell stresses that are uniaxial creep fatigue data directly. In this case, it was
predicted by equation (1), which evolve with every cycle, important not to minimise the fatigue damage as it
and the stress relaxation behaviour that is calculated would have meant that fitting would have overestimated
from equations (2)–(4), which also evolve with every the creep damage. After taking these considerations into
cycle. The cycle by cycle relaxation strain has also been account, it is possible to use the calculations of creep
used to enhance the total strain range for the fatigue and fatigue damage for the reversed bend tests to
damage calculations, as required by R5.3 Creep damage validate the models for Type 321 at 650uC.
has been calculated for the time fraction, simple ductility The results of the best estimate and upper bound
exhaustion and SMDE approaches. It is important to creep and fatigue damage calculations for time fraction,
note that no allowance has been made for the weldment simple ductility exhaustion and SMDE are shown grap-
in calculating creep damage and this analysis is therefore hically as damage interaction diagrams in Figs. 18–21,
testing the hypothesis that the creep damage in these where Figs. 20 and 21 also include the observed versus
thin section (3 mm thick) two pass weldments can be predicted endurances and total damage versus start of
estimated using the creep properties of the Type 321 dwell stress plots for the SMDE approach. It should be
parent material. This is the usual assumption for noted that for completeness, the uniaxial test data have
simplified R5 analyses of creep fatigue crack initiation. been included in Figs. 18–21 and the creep damage has

298 Materials at High Temperatures 2014 VOL 31 NO 4

 Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

18 Interaction diagrams showing fatigue damage and time fraction creep damage for reversed bend tests on weldments
(unfilled symbols are unfailed tests). a best estimate fatigue and time fraction creep damage; b upper bound fatigue
and time fraction creep damage

been calculated using the same cycle by cycle methods Discussion

that have been used for the reversed bend tests.
The results of the creep and fatigue damage calcula- With regard to the findings of this re-examination of the
tions for the reversed bend tests are very similar to the available uniaxial push–pull creep fatigue data on Type
findings of the uniaxial tests; the SMDE approach gives 321 at 650uC, these are broadly similar to the findings of
more realistic predictions of creep and fatigue failure for previous investigations into the creep fatigue behaviour
Type 321 at 650uC than both the time fraction and the of other austenitic stainless steels and ferritic steels.6–8
simple ductility exhaustion approaches. In particular, the The important features to note regarding the time
simple ductility exhaustion gives overly conservative fraction approach are that the best estimate damages
predictions for both best estimate and upper bound creep calculated with time fraction were realistic (Fig. 10),
damages; particularly, as the initial stress decreases. With although the upper bound estimates of time fraction
regard to the time fraction approach, the best estimate creep damage are overly pessimistic (Fig. 12). This
predictions are good; however, the upper bound predic- would be a problem when using time fraction in design
tions are more pessimistic than the SMDE approach. codes and life assessment, since lower bound time to

19 Interaction diagrams showing fatigue damage and simple ductility exhaustion creep damage for reversed bend tests
on weldments (unfilled symbols are unfailed tests). a best estimate fatigue and simple ductility exhaustion creep
damage; b upper bound fatigue and simple ductility exhaustion creep damage

Materials at High Temperatures 2014 VOL 31 NO 4 299

Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

20 Best estimate fatigue damage and stress modified ductility exhaustion creep damage for reversed bend tests on
weldments (unfilled symbols are unfailed tests). a interaction diagram; b observed versus predicted endurance; c
total damage versus start of dwell stress for a mid-life cycle

300 Materials at High Temperatures 2014 VOL 31 NO 4

 Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

21 Upper bound fatigue damage and stress modified ductility exhaustion creep damage for reversed bend tests on weld-
ments (unfilled symbols are unfailed tests). a interaction diagram; b observed versus predicted endurance; c total
damage versus start of dwell stress for a mid-life cycle

Materials at High Temperatures 2014 VOL 31 NO 4 301

Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

22 Creep ductility of cast M2265 Type 321 at 650uC: note data from Ref. 13 are elongation values because plastic loading
strains are not reported

failure is used to calculate the upper bound damage. Furthermore, the SMDE approach which has recently
Overly pessimistic design and life assessment calcula- been added to the R53 procedure gives the most realistic
tions would lead to costly decisions on plant manufac- predictions of best estimate (Fig. 14) and upper bound
turing operating and should be avoided. With regard to (Fig. 16) damages. Therefore, this SMDE approach is
the simple ductility exhaustion approach (equation (11)), recommended for use with Type 321 at 650uC.
this has been shown to be overly pessimistic in cases with It is useful to explore the relationship between the
low start of dwell stresses for both the best estimate simple ductility exhaustion, the SMDE approaches and
(Fig. 11) and upper bound (Fig. 13) estimates of damage. the creep ductility measured in constant load tests

23 Evolution of strain rate as function of stress for 23?3 h dwell uniaxial creep fatigue tests (as derived from mid-life
values for B 0 and b and observed stress drops for other dwells, with recalculated values of B 0). a RST S3/20,
DeT50?58%, 23?3 h; b RST T2/66, DeT51?5%, 23?3 h

302 Materials at High Temperatures 2014 VOL 31 NO 4

 Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

further. Figure 11b shows a clear relationship between fillet welds does show the general accuracy of the WER
the creep damage and the initial stress, which is not and WSEF. However, it can be noted that some of the
taken into account by the simple ductility exhaustion lower strain range endurance longitudinal weld tests
approach. This relationship suggests that the creep which were still conservatively predicted by the lower
ductility would increase as the stress decreases. bound failure endurance (Fig. 17b), were slightly less
However, Fig. 22 shows that the ductility does increase well predicted than the other tests. Post-test examination
a little between 215 and 160 MPa, although the creep showed that the longitudinal welds tended to initiate
ductility then falls with decreasing stress. It is therefore fatigue cracking in the Type 347 weld metal in both the
clear that simple models to describe creep ductility as a cap and root. This was influenced by the uneven ripples
function of either stress or strain rate alone would not on the surface of the weld cap and some defects
work to predict the creep damage in creep fatigue tests. particularly, in the weld root. These would have
The problem for constant load test data is that there is contributed to early initiation, which is taken into
the strong correlation between strain rate and stress, account by the WER. With regard to the continuous
which makes it very difficult to differentiate between the cycling tests on the butt and fillet welds, the R53
effects of strain rate and stress on the accumulation of treatment of weldments gives good predictions for all of
creep damage. Nevertheless, the evolutionally stress these weldments. The fillet welds tended to fail at the
relaxation behaviour from the creep fatigue tests on fusion boundary of either the toe or heal of the fillet
Type 321 at 650uC shows a wide range of behaviours in weld, and grow from the HAZ into the parent. The butt
strain rate and stress. This is illustrated for the two 23?3 welds failed in either the parent or the HAZ, with no
h dwell creep fatigue tests in Fig. 23. It can be seen from failures in the weld metal.
Fig. 23b that the strain rate and stress behaviour of the The observations that can be drawn from the reversed
higher strain range test (DeT51?5%) are similar to that bend creep fatigue tests are entirely consistent with those
of the constant load test data. This explains why the from the uniaxial push–pull tests. The SMDE approach
simple ductility exhaustion approach was realistic for giving the most realistic prediction of both the best
this test (Fig. 11b). However, the strain rates are much estimate (Fig. 20) and upper bound (Fig. 21) creep
higher for a given stress in the smaller strain range test damages at failure. With regard to the simple ductility
(DeT50?58%) (Fig. 23a). This is why the simple ductility exhaustion approach, this was overly pessimistic with
exhaustion approach was overly conservative for this low start of dwell stresses, which correspond to low
test. Nevertheless, the SMDE approach which takes strain ranges (Figs. 19) and the time fraction approach,
account of the effect of both strain rate and stress on while good with the best estimate predictions (Fig. 18a)
creep ductility gives realistic predictions of failure for all was overly conservative with the upper bound predic-
tests irrespective of the relaxation behaviour. The key tions (Fig. 18b). Nevertheless, there was a tendency for
feature of the SMDE model is that the exponent to the butt welds, longitudinal welds and the fillet welds
strain rate, n1, is positive but the exponent to stress, m1, with the fillet in compression during the dwell to be
is negative (Table 10). Therefore, the combination of conservatively predicted (Fig. 20). However, the best
high strain rate and low stress as exhibited by the smaller estimate predictions for the fillet welds with the fillet in
strain range tests (for example, Fig. 23a), gives higher tension during the dwell were a little optimistic (Fig. 20).
predicted creep ductility with SMDE, then the simple Notwithstanding this, all weldments were conservatively
ductility exhaustion and consequently more realistic predicted by the upper bound predictions using the
(less pessimistic) estimates of creep damage. Since low SMDE approach in conjunction with the new R53,8
strain ranges are typical in real plant assessments, it is treatment of weldments (Fig. 21). Post-test metallogra-
expected that the SMDE approach would give a phy shows that as for the continuous cycling tests, the
significant benefit over the simple ductility exhaustion fillet welds tended to fail at the fusion boundary of either
approach. the toe or heal of the fillet weld, and grow from the HAZ
The fatigue damage for the reversed bend tests was into the parent. The butt welds failed in either the parent
calculated using a different fatigue endurance model or the HAZ, with no failures in the weld metal. The
from the uniaxial push–pull tests. Between around 700 longitudinal welds initiated in the Type 347 weld metal
and 7000 cycles, EDF Energy’s usual fatigue endurance and grow into the parent plate. Cracking tended to be a
model is a little conservative compared with the data, mixture of intergranular creep and transgranular fatigue
as can be seen in Figs. 7b and 17a. Notwithstanding damage, which was dominated on the surface that
this observation, it is clear from Fig. 17a that the experienced a tensile dwell, which corresponds with the
continuous cycling reversed bend tests on parent plates roughly equal amounts of creep and fatigue damage
and weldments are reliably predicted by this fatigue predicted in Fig. 20a.
endurance model in conjunction with the latest recom- It was interesting to observe that although the cyclic
mendations in R53,8 for the treatment of weldments. The stress strain behaviour as measured in uniaxial creep
R53 treatment of weldments use the WER to account for fatigue tests on the resolution treated and aged speci-
the fatigue endurance reduction due to the presence of mens was different, which means different start of dwell
small imperfections and the WSEF to account for strain stresses, once this has been taken into account the stress
enhancement due to the weldment geometry and the relaxation and creep damage behaviours were closely
material mismatch between weldment zones. It should similar. Furthermore, there was very little difference
be noted that these tests are on a simple bend geometry between the butt welds made using either Type 347 or
with fully penetrated welds, so this does not validate the Type 316L weld metal.
more severe WSEF of 1?66, which is used for partially It is important to note that the predictions of fatigue
penetrated welds. The good agreement between the and creep damage that are made in this paper contain a
uniaxial parent, reversed bend parent, butt welds and series of simplifying assumptions. Nevertheless, these

Materials at High Temperatures 2014 VOL 31 NO 4 303

Spindler et al. Creep fatigue behaviour of Type 321 stainless steel

assumptions are commonly made when assessing the life data for this work. The clear tabulation of experimental
of real plant components. For example, that the results in these source reports was one of the key factors
behaviour of weldments can be predicted using the that made this work possible.
deformation properties of the parent used in conjunction
with factors to take account of the effect of the
weldment on fatigue endurance. The validation in this
References
paper shows that so long as realistic approaches are used 1. ‘Section 3 – Rules for construction of nuclear power plant
to predict creep and fatigue damage, such as using the components – Division 1 – Subsection NH – Class 1 components
in elevated temperature service’, ASME-BPVC-SEC 3 NH-2013,
‘stress modified ductility exhaustion’ approach, that ASME, New York, USA, 2013.
these simplifying assumptions give reliable predictions 2. ‘RCC-MRx, design and construction rules for mechanical compo-
of creep-fatigue failure in Type 321 weldments. nents of nuclear installations’, AFCEN, Paris, France, 2012.
Notwithstanding all of the complexity of analysing 3. ‘Assessment procedure for the high temperature response of
creep fatigue data, it is suggested that the main structures’, R5, Issue 3, Revision 002, EDF Energy, Gloucester,
UK, 2014.
contributions of this work have been the analysis of a 4. Central Research Institute of Electric Power Industry: ‘Guidelines
large body of LCF and creep-fatigue data, including six for high-temperature structural integrity assessment’ (Draft); 2001,
tests with 23?3 or 24 h dwells and many tests on thin Tokyo, Central Research Institute of Electric Power Industry.
section weldments that are relevant to real plant 5. Y. Takahashi: ‘Development of structural integrity assessment
assessments. In addition, it has been shown that the guideline for FBR components’, ASME PVP, 1998, 365, 207–
214.
cyclic and stress relaxation behaviour can be readily 6. M. W. Spindler: ‘An improved method for calculation of creep
accounted for using a cycle by cycle analysis. damage during creep-fatigue cycling’, Mater. Sci. Technol., 2007,
23, (12), 1461–1470.
Conclusions 7. M. W. Spindler and W. M. Payten: ‘Advanced ductility exhaustion
methods for the calculation of creep damage during creep-fatigue
The results of uniaxial push–pull continuous cycling and cycling’, J. ASTM Int., 2011, 8, (7). DOI: 10.1520/JAI103806.
creep fatigue tests on Type 321 stainless steel at 650uC 8. D. W. Dean, M. W. Spindler, M. Chevalier and N. G. Smith:
‘Recent developments in the R5 volume 2/3 procedures for assessing
have been used to derive evolutionary models for the
creep-fatigue initiation in defect-free components operating at high
maximum stress in the cycle, which for a peak dwell is temperatures’, Proc. ASME 2013 Pressure Vessels and Piping Conf.,
also the start of dwell stress, the evolutionary stress Paris, France, July 2013, ASME, Paper PVP2013-98134.
relaxation behaviour and a SMDE model. 9. ‘Steels for fired and unfired pressure vessels. Plates Part 3:
It has been shown that the SMDE model gives more Corrosion and heat resisting steels’, BS1501, British Standards
Institution, London, UK, 1973.
realistic estimates of creep damage than both the time
10. A. J. Moffat, J. P. Douglas, M. White, M. W. Spindler, C. Austin
fraction and simple ductility exhaustion approaches. and S. Jacques: ‘Development of the RCC-MR creep deformation
The latest R5 methods to calculate fatigue damage in model for the prediction of creep and stress relaxation in AISI Type
weldments in conjunction with the SMDE model for 321 stainless steel’, Proc. ASME 2014 Pressure Vessels and Piping
creep damage give realistic predictions for the results of Conf., Anaheim, CA, USA, July 2014, ASME, Paper PVP2014-
28917.
a large dataset of reversed bend fatigue and creep fatigue 11. ‘AGR materials data handbook’, R66, Revision 009, EDF Energy,
tests on three different configurations of weldments. Gloucester, UK, 2011.
12. ‘Elevated temperature properties for steels for pressure purposes.
Acknowledgements Part 1: Stress rupture properties’, PD6525, British Standards
Institution, London, UK, 1990.
This paper is published by permission of the EDF 13. ‘BSCC high temperature data’, The Iron and Steel Institute,
Energy Nuclear Generation Ltd and AMEC Technical London, UK, 1973.
14. A. Pineau: ‘High temperature fatigue behaviour of engineering
Services. A particular acknowledgment is given to Ian materials in relation to microstructure’, in ‘Fatigue at high
Bretherton who was the main author of many of the temperature’, (ed. R. P. Skelton), 305–364; 1983, London/New
internal company reports, which provided the source York, Applied Science Publishers.

304 Materials at High Temperatures 2014 VOL 31 NO 4

Copyright of Materials at High Temperatures is the property of Maney Publishing and its
content may not be copied or emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission. However, users may print, download, or email
articles for individual use.