See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/347311177

Evaluation of Mean Stress Correction on Fatigue Curves of Grade 91 and Alloy 617
in ASME Section III Division 5

Conference Paper · August 2020

DOI: 10.1115/PVP2020-21572

CITATIONS READS

0 5

4 authors, including:

Yanli Wang Sam Sham

Oak Ridge National Laboratory Idaho National Laboratory
61 PUBLICATIONS 673 CITATIONS 110 PUBLICATIONS 685 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Creep-Life Prediction for Ferritic-Martensitic Steels Using Crystal Plasticity Modeling View project

All content following this page was uploaded by Sam Sham on 07 June 2021.

The user has requested enhancement of the downloaded file.

 Proceedings of the ASME 2020
Pressure Vessels & Piping Conference
PVP2020
July 19-24, 2020, Minneapolis, Minnesota, USA

Draft-PVP2020- 21572

EVALUATION OF MEAN STRESS CORRECTION ON FATIGUE CURVES OF GRADE 91

AND ALLOY 617 IN ASME SECTION III DIVISION 5

Y. Wang 1, M. D. McMurtrey2, R. I. Jetter3, T.-L. Sham4

1Oak Ridge National Laboratory, Oak Ridge, TN

2Idaho National Laboratory, Idaho Falls, ID
3
RI Jetter Consulting, Pebble Beach, CA
4Argonne National Laboratory, Lemont, IL

ABSTRACT
The current ASME Boiler and Pressure Vessel (B&PV) Code
Section III, Division 5, Subsection HB, Subpart B has only one
design fatigue curve for grade 91 steel (Gr. 91) at 540 oC (or 1000
o
F). The ASME Section III Working Group on Creep-Fatigue
and Negligible Creep (WG-CFNC) has taken an action to
incorporate the temperature-dependent design fatigue curves for
Gr. 91 developed by Japan Society of Mechanical Engineers
(JSME) into ASME Section III Division 5. During the process,
issues regarding the effect of mean stress on fatigue analysis, and
how to consider the mean stress effect for elevated-temperature
design, were brought up. To evaluate whether the design fatigue
curves of Gr. 91 needed adjustment to account for mean stress,
critical tests were designed and performed at 371 oC (700 oF) and
540 oC (1000 oF). This study is similar to the work performed on
Alloy 617 when its fatigue design curves were established for
temperature range of 538-704°C (1000-1300°F) as part of the
Code Case package for Alloy 617 to be used as Class A
construction material in Division 5. The effects of mean stress
on Alloy 617 were evaluated at 550°C (1022°F). The results
showed that the mean stresses introduced by the non-zero mean
strain could not be maintained under strain-controlled fatigue
and resulted in negligible effect on the fatigue life. Mean stress
correction was not recommended for Alloy 617 fatigue design
curves in Division 5. This study shows the same conclusion for
Gr. 91.
Keywords: mean stress correction, fatigue curves, elevated
temperature
1. BACKGROUND
High-temperate components under cyclic loading may
experience mean stress effects. Mean stress can be categorized
into either load-controlled or strain-controlled quantities. For
load-controlled mean stress, such as that introduced by internal
pressure, current Creep-Fatigue design procedures (elastic or
inelastic analysis route) in ASME B&PV Code Section III
Division 5 account for creep damage introduced by sustained
loading. Thus, the mean stress effect due to sustained primary
loads shows up in the calculation of the life fraction used to enter
the damage interaction diagram (Figure HBB-T-1420-2, “Creep-
Fatigue Damage Envelope”. On the other hand, the effect of
mean stress under strain-controlled cycling is introduced by a
non-zero mean strain. Mean stress under strain-controlled cyclic
loading corresponds to cases such as welding residual stress or
residual stress by thermal transients. Since the fatigue curves are
based on fully reversed strain-controlled cycling, there is a
negligible amount of mean stress during testing. Evaluation of
the mean stress effect under strain-controlled fatigue is deemed
to be necessary when fatigue design curves are developed.
The current ASME B&PV Code Section III Division 5
(2019 edition), Subsection HB, Subpart B has only one design
fatigue curve for Gr. 91 at 540°C (or 1000°F). The ASME
Section III Working Group on Creep-Fatigue and Negligible
Creep (WG-CFNC) has taken action to incorporate the Japan
Society of Mechanical Engineers (JSME) temperature-
dependent fatigue curves for Gr. 91 in Division 5. The mean
stress evaluation on Gr.91 in this study is to address the issue
regarding the effect of mean stresses on fatigue analysis raised
by WG-CFNC. Similar evaluation on the mean stress effect was
also performed on Alloy 617 when its fatigue design curves were
established as part of the Code Cases N-872 (low temperature)

1
Contact author: wangy3@ornl.gov

1 Copyright © 2020 by ASME

and N-898 (high temperature) data package for Alloy 617 to be
used as Class A construction material in Division 5. Based on
recommendations from the Section III Working Group on
Fatigue Strength, the Code Case N-872 used the existing fatigue
design curves in I-9,5. I-9.5 uses a mean stress correction branch
for cycles to failure above 230,000 cycles. The mean stress
correction for the 800 °F fatigue design curve is based on
analysis using a Goodman diagram approach rather than actual
experimental data [1]. The mean stress evaluation on Alloy 617
in this study is to determine whether mean stress correction is
appropriate for the fatigue curves for temperatures above 800 °F.
The primary temperature range of concern for the Alloy 617 high
temperature code case is the 1,000-1,300 °F, and a test
temperature of 550 oC in the low end of the range to ensure that
it was below the creep range.
2. MATERIALS AND METHODS
2.1 Materials
The Gr. 91 plate with Heat No. 30176 was manufactured by
Carpenter Technology Corporation in the early 1980s, and it has
been preserved at the Oak Ridge National Laboratory (ORNL)
as an archival material [2]. The chemical compositions of this
plate are listed in Table 1 below. This plate was characterized for
its mechanical properties and the data were used as reference
data for the Gr. 91 development program. The as-received Gr. 91
plate was hot forged followed by hot rolling. To support this
study, the plate was normalized at 1050°C for 1-hr and then
tempered at 760°C for 2-hr, followed by air cooling. The heat-
treated plate has typical tempered martensitic microstructure
with average Vickers hardness value of 227 kg/mm2.
Alloy 617 plate with Heat No. 314626 was manufactured by
ThyssenKrupp VDM USA, Inc with a nominal thickness of 35
mm. The as-received plate had been hot-rolled and annealed. The
chemical compositions of the plate are listed in Table 2. This
Alloy 617 plate was used to generate data in support of Alloy
617 Code Cases N-872 and N-898 that were fully approved in
2019.
levels were determined to be the strain to reach the maximum
mean stress under tensile, (𝜎𝜎𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 )𝑚𝑚𝑚𝑚𝑚𝑚 , per the Recommended
Test Procedure as
𝛥𝛥𝛥𝛥
(𝜎𝜎𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 )𝑚𝑚𝑚𝑚𝑚𝑚 = 𝑆𝑆𝑢𝑢 –
2
where Su is the ultimate tensile strength and Δσ is the stress range
taken from a zero mean strain baseline fatigue test for the same
strain range. The pre-straining procedure introduces an initial
stress in the test specimen, prior to the cyclic loading. The strain
ranges were selected to result in the high-cycle-fatigue region
where the original fatigue curve was significantly different from
the mean stress correction. The details of the testing parameters
are summarized in Table 3 for Gr.91 and Alloy 617. The Alloy
617 tests were performed with a mean stress slightly below the
recommended value due to limitations from the extensometers
(which had a total range of ±2.2%). The strain rate for testing on
Alloy 617 was the standard rate of 1E-3/s. The Gr.91 was
initially tested at a strain rate of 5E-3/s but was later increased to
1E-2/s for the test at 371oC. The increase in the strain rate was to
shorten the duration of the tests, and it was considered to have
insignificant effect on the fatigue life based on the JSME fatigue
curves data base [4].

2.2 Experimental Method

The mean stress evaluation experiments followed the
recommended test procedure, “Recommended Test Procedure
For Fatigue Testing Of Metallic Materials Below The Creep
Range,” by ASME Section III, Working Group-Fatigue Strength
[3]. Tests with and without mean strain were performed under
strain-controlled cyclic loading at 371°C (700°F) for Gr.91 and
550°C for Alloy 617 below the creep region. For comparison FIGURE 1. (A) STRAIN-CONTROLLED LOADING PROFILE
purpose, tests were also performed at higher temperature of WITH ZERO INITIAL MEAN STRESS (I.E., BASELINE TEST,
540°C (1000°F) for Gr. 91 within the creep range. FULLY REVERSED CONTINUOUS FATIGUE WITH R=-1 ON
The evaluation is to compare the fatigue life under strain- STRAIN), AND (B) STRAIN-CONTROLLED CONTINUOUS
controlled testing with and without pre-straining effect, and the CYCLING LOADING PROFILE WITH A NON-ZERO PRE-
fatigue failure cycles against the best fit fatigue curve. The STRAIN, Ɛ0, FOR THE MEAN STRESS TESTING.
loading profiles of the fatigue with and without pre-straining are
schematically shown in FIGURE 1. The pre-straining strain

2 Copyright © 2020 by ASME

 Table 1. Chemical compositions of Gr. 91 plate with heat number 30176 (weight %).
C P Si* Ni Mn N Ti Sn V Fe As
0.081 0.010 0.11 0.09 0.37 0.055 0.004 <0.001 0.209 balance 0.001
Zr S Cr Co Mo Al W Cu Nb B
<0.001 0.003 8.61 0.010 0.89 0.007 <0.01 0.04 0.072 <0.09
*Note: heat 30176 is lower in Si content than the SA-387 specification of 0.2–0.5%.

Table 2. Chemical compositions of Alloy 617 plate with heat number 314626 (weight %).
C S Cr Ni Mn Si Mo Ti Cu Fe Al Co B
0.05 <0.002 22.2 R54.1 0.1 0.1 8.6 0.4 0.04 1.6 1.1 11.6 <0.001

Table 3. Testing parameters for strain-controlled mean stress evaluation on fatigue life of
Gr. 91 (Heat No. 30176) and Alloy 617 (Heat No. 314626).
Materials Test temperature Strain range, Δε1 Strain rate Pre-straining level, Stress level after
𝜺𝜺𝟎𝟎 pre-straining
Gr.91 371oC 0.36% 5E-3/s, 1E-2/s 0.15% 252.4 MPa
540oC 0.33% 5E-3/s 0.18% 306.3 MPa
Alloy 617 550oC 0.36% 1E-3/s 3.36% 278.2 MPa
550oC 0.48% 1E-3/s 3.21% 252.0 MPa

3. RESULTS AND DISCUSSIONS cycles had a remarkable resemblance to that at 540°C. It is

3.1 Mean Stress Test Results on Gr. 91 known that Gr. 91 is a cyclic softening material, and this cyclic
The strain-controlled fatigue tests on Gr. 91 with and softening effect at the strain range levels in this study may have
without pre-straining were carried out until the failure criteria played a role in relaxing the mean stresses under the strain-
were met. Table 4. summarizes the failure criteria and the fatigue controlled condition.
failure cycles of the four tests at 371oC and 540oC. The
representative hysteresis loops of the strain-controlled mean
stress fatigue tests are compared with the standard fatigue tests
in FIGURE 2 and FIGURE 3. For both temperatures, the first
cycle of the non-zero mean strain fatigue test showed plastic
deformation but decreased to near elastic from the second cycle.
The initial non-zero mean stresses (shown as the ‘+’ sign on the
plots) decreased when the specimens were loaded with
additional cycles.
The measured mean stresses as a function of loading cycles
are compared in FIGURE 4. for tests at 540°C and FIGURE 5.
for 371°C. At both test temperatures, the initial non-zero mean (A)
stress introduced by pre-straining quickly decreased as more
strain-controlled cycles were applied. The decrease in the mean
stresses as a function of applied cycles is more evidently shown
in the linear plot scale. The initially applied maximum mean
stress could not be sustained under stain-controlled cyclic
loading. At 540°C the mean stress dropped from the initial 196
MPa to about 50 MPa at 5,000 cycles. Similarly, at 371°C, the
mean stress dropped from the initial 143.4 MPa to about 50 MPa
at 10,000 cycles. That is, more than 96% of the applied cycles
had mean stress levels of 50 MPa or less for the two strain-
controlled fatigue tests with non-zero initial mean stress. Note
also that the mean stresses for the standard strain-controlled (B)
fatigue tests with R = −1 may deviate from zero for most of the FIGURE 2. REPRESENTATIVE HYSTERESIS LOOPS OF
cycles, but the mean stress level remained small for the duration STRAIN-CONTROLLED FATIGUE TESTS AT 540°C FOR GR. 91
of the tests. (HEAT 30176). (A) WITH NON-ZERO INITIAL MEAN STRESS,
It is interesting to note that despite the lower testing AND (B) WITH ZERO INITIAL MEAN STRESS.
temperature of 371°C with negligible creep deformation, the
decaying trend of the initial mean stress as a function of applied

3 Copyright © 2020 by ASME

 (A)
(A)

(B)
FIGURE 3. REPRESENTATIVE HYSTERESIS LOOPS OF
STRAIN-CONTROLLED FATIGUE TESTS AT 371°C FOR GR. 91
(HEAT 30176). (A) WITH NON-ZERO INITIAL MEAN STRESS
AND (B) WITH ZERO INITIAL MEAN STRESS.
(B)
FIGURE 5. MEASURED MEAN STRESSES AS A FUNCTION OF
LOADING CYCLES FOR MEAN STRAIN EFFECT TESTING AT
371°C WITH THE APPLIED CYCLES IN LOGARITHMIC SCALE
(A) AND LINEAR SCALE (B).

The cycles to failure data points are plotted against the

JSME best fit curve in FIGURE 6. At both temperatures, this
heat of Gr. 91 showed higher fatigue resistance than the average
JSME best fit curves. At 540°C, the test with pre-straining and
non-zero initial mean stress lasted longer than the test with
(A) standard fatigue with zero initial mean stress, indicating mean
stress correction is not needed at this temperature. At 371°C, the
non-zero mean stress fatigue showed slightly shorter life cycles
than with standard fatigue, but the difference was small and well
within the data scatter for fatigue testing; therefore, it was
concluded that the non-zero mean stress has a negligible effect
on strain-controlled fatigue life at this temperature.

(B)
FIGURE 4. MEASURED MEAN STRESSES AS A FUNCTION
OF LOADING CYCLES FOR MEAN STRAIN EFFECT TESTING
AT 540°C WITH APPLIED CYCLES IN LOGARITHMIC SCALE
(A) AND LINEAR SCALE (B).

4 Copyright © 2020 by ASME

 similar. As a result, the test was running with very little mean
stress, although it had a mean strain of 3.36%. The mean stress
had a maximum value of about 37 MPa in the first cycle, and the
dropped to about 20 to 30MPa with additional cycles.

FIGURE 6. COMPARISON OF THE STRAIN-CONTROLLED

MEAN STRAIN FATIGUE TEST RESULTS WITH THE JSME BEST
FIT CURVES FOR GR. 91 AT 540°C AND 371°C.

3.2 Mean Stress Test Results on Alloy 617

Two sets of strain-controlled fatigue tests with and without (A)
pre-straining were performed at 550 °C (1022 °F) to gauge the
mean stress effect of Alloy 617. The tests are to determine
whether mean stress correction to the fatigue curve at 1000-
1300°F is appropriate. It is noted that the fatigue design curves
in current Section III Div. 5 have the 2 and 20 “knock-down”
factors built into the rules [5], unnecessary mean stress
correction would result in excessive conservativism.
The hysteresis loops for strain controlled fatigue tests with
and without pre-straining for 0.36% strain range are shown in
FIGURE 7 and the maximum and minimum stresses are
compared in FIGURE 8 . Both tests started out with significant
inelastic strain, indicated by the wide loops, and reached a fully
elastic state for both tests by 50,000 cycles. Cyclic hardening is (B)
evident by the increasing stress with cycle for the baseline test FIGURE 7. REPRESENTATIVE HYSTERESIS LOOPS OF
with zero mean strain. For the test with pre-straining, the first STRAIN-CONTROLLED FATIGUE TESTS AT 550°C FOR ALLOY
loop started at about 278 MPa and stayed at about that level until 617 (HEAT 314626) AT STRAIN RANGE 0.36%. (A) WITH ZERO
the maximum strain was reached. With the exception of the first INITIAL MEAN STRESS, (B) WITH NON-ZERO INITIAL MEAN
cycle, the hysteresis loops for the mean stress test were narrower STRESS
than those of the baseline test. The maximum tensile stress of the
mean stress test started higher, but the minimum stress was very

Table 4. Summary of the test results of strain-controlled mean stress evaluation on Gr. 91 (heat 30176) at 540°C and 371°C.
Fatigue with zero mean strain Fatigue with non- zero mean strain
Test
Cycles to Cycles to
temperature Failure criteria Failure criteria
failure failure
Failure by fracture into two
371°C 6.7E6 17% load drop runout 5.07E6
parts with 14% load drop

Failure by fracture into two

540°C 4.59E5 1.3E6 17% load drop runout
parts with 17% load drop

5 Copyright © 2020 by ASME

FIGURE 8. MAXIMUM AND MINIMUM STRESSES FOR EACH
CYCLE OF THE BASELINE AND MEAN STRESS TESTS AT
550°C FOR 0.36% STRAIN RANGE
The hysteresis loops for the 0.48% strain range tests with
and without pre-straining are compared in FIGURE 9 and the
maximum and minimum stresses are compared in FIGURE 10.
the hysteresis loops for the mean stress test were narrower than
those of the baseline test. The maximum stress of the baseline
test rapidly approaches that of the mean stress test. During the
first 10,000 cycles, the peak stress is higher for the test with pre-
straining, however, beyond 10,000 cycles, the stresses reach a
similar level for both the mean stress and baseline test
(A)
(B)
FIGURE 9. REPRESENTATIVE HYSTERESIS LOOPS OF
STRAIN-CONTROLLED FATIGUE TESTS AT 550°C FOR ALLOY
617 (HEAT 314626) AT STRAIN RANGE 0.48%. (A) WITH ZERO
INITIAL MEAN STRESS, AND (B) WITH NON-ZERO INITIAL
MEAN STRESS
FIGURE 10. MAXIMUM AND MINIMUM STRESSES FOR
EACH CYCLE OF THE BASELINE AND MEAN STRESS TESTS
AT 550°C FOR 0.48% STRAIN RANGE.
Both the mean stress and baseline tests with the 0.36% strain
range extended well beyond the expected life from the best fit
fatigue curve and were interrupted prior to failure at 132,000
cycles and 452,000 cycles for the mean stress and baseline tests,
respectively. The failure cycles for the 0.48% strain range tests
were 29,000 cycles for the mean stress test and 39,000 cycles for
the baseline test, both were slightly higher than the average
cycles of 25,000 from the best fit fatigue curve but within the
data scatter. When the mean stress and baseline tests are plotted
on the fatigue curve, as in FIGURE 11, all points are to the right
of the original fatigue curve and significantly to the right of the
mean stress correction curve. This implies that the mean stress
correction would be overly conservative and that the original
fatigue curve accurately modeled fatigue life for baseline and
strain controlled mean stress tests.
FIGURE 11. FATIGUE TESTS AT 550°C (1022°F) ARE
PLOTTED WITH THE BEST FIT FATIGUE CURVES (I.E.,
FATIGUE CUVRVES WITHOUT 2 AND 20 “KNOCK-DOWN”
FACTORS).
6 Copyright © 2020 by ASME
4. CONCLUSION
In summary, current test data on Gr. 91 and Alloy 617
demonstrate a negligible effect of applied mean strain on strain-
controlled fatigue life; i.e., strain-controlled cycling with non-
zero mean strain does not reduce the cycle life when compared
with cycling test with zero mean strain. Based on the above
results and discussions, mean stress correction is not
recommended for the fatigue design curves in Division 5 for Gr.
91 nor for Alloy 617. To include the mean stress correction
would result in an overly conservative model.
COPYRIGHT NOTICE
This manuscript has been co-authored by UChicago
Argonne LLC, under Contract No. DE-AC02-06CH11357, by
Battelle Energy Alliance, LLC, under Contract No. DE-AC07-
05ID14517 and by UT-Battelle LLC, under Contract No. DE-
AC0500OR22725, with the U.S. Department of Energy. The
United States Government retains and the publisher, by
accepting the article for publication, acknowledges that the
United States Government retains a nonexclusive, paid-up,
irrevocable, world-wide license to publish or reproduce the
published form of this manuscript, or allow others to do so, for
United States Government purposes.
ACKNOWLEDGEMENTS
The research was sponsored by the U.S. Department of
Energy, Office of Nuclear Energy, under Contract No. DE-
AC02-06CH11357 with Argonne National Laboratory, managed
View publication stats
and operated by UChicago Argonne LLC, under contract No.
DE-AC05-00OR22725 with Oak Ridge National Laboratory
(ORNL), managed and operated by UT-Battelle, LLC, and under
contract No. DE-AC0500OR22725 with Idaho National
Laboratory (INL), managed and operated by Battelle Energy
Alliance, LLC. Programmatic direction was provided by the
Office of Nuclear Reactor Deployment of the Office of Nuclear
Energy.
REFERENCES
[1] Strizak, J. P., Brinkman, C. R., Booker, M. K., and
Rittenhouse, P. L., The Influence of Temperature, Environment
and Thermal Aging on the Continuous Cycle Fatigue Behavior
of Hastelloy X and Inconel 617, ORNL/TM-8130, 1982, Oak
Ridge National Laboratory, Oak Ridge, Tennessee.
[2] DiStefano, J. R. et al., Summary of Modified 9Cr-1Mo
Steel Development Program: 1975-1985, ORNL-6303, 1985,
Oak Ridge National Laboratory, Oak Ridge, Tennessee.
[3] Recommended test procedure for fatigue testing of
metallic materials below the creep range REVISION 4, Prepared
by Subgroup on Fatigue Strength, ASME Boiler and Pressure
Vessel Committee, April 1994, The American Society of
Mechanical Engineers.
[4] JSME Codes for Nuclear Power Generation Facilities-
Rules on Design and Construction for Nuclear Power Plants,
Part II Fast Reactor Code, JSME S NC2-2016, 2016.
[5] Wright, R. N., Updated Draft ASME Boiler and Pressure
Vessel Code Case for Use of Alloy 617 for Construction of
Nuclear Components for Section II, Division 5, INL/EXT-17-
42999, 2018, Idaho National Laboratory, Idaho Falls, Idaho.
7 Copyright © 2020 by ASME