WRC 432-1998@ (Asmesecviii)

BULLETIN 432 JUNE 1998
IJIIIIIIIIIIIIII
Welding Research Council
VVRC
bulletin
FATIGUE STRENGTH REDUCTION AND STRESS
CONCENTRATION FACTORS FOR WELDS IN
PRESSURE VESSELS AND PIPING
REPORT NO . 1 : Interpretive Review of Weld

Fatigue-Strength-Reduction and
Stress-Concentration Factors
C . E. Jaske
REPORT No . 2 : Fatigue-Strength-Reduction Factors

for Welds Based on NDE
J . L . Hechmer and E. J . Kuhn, III
These Bulletins contain final Reports from projects sponsored

by the Welding Research Council, important papers presented
before engineering societies and other reports of current
ISSN 0043-2326 interest.
WELDING RESEARCH COUNCIL, INC.
3 PARK AVE . (27th Floor),
NEW YORK, NY 10016-5902
INTENTIONALLY LEFT BLANK
WRC - The Welding Research Council brings together science and engineering specialists in
developing the solutions to problems in welding and pressure vessel technology. They exchange
knowledge, share perspectives, and execute R and D activities. As needed, the Council organizes
and manages cooperative programs.
MPC – A Council of the WRC, the Materials Properties Council is dedicated to providing
industry with the best technology and the best data that can be obtained on the properties of
materials to help meet today’s most advanced concepts in design and service, life assessment,
fitness-for-service, and reliability and safety.
PVRC – A Council of the WRC, the goal of the Pressure Vessel Research Council is to
encourage, promote and conduct research in the field of pressure vessels and related pressure
equipment technologies, including evaluation of materials, design, fabrication, inspection and
testing.
For more information, see www.forengineers.org
WRC Bulletins contain final reports from projects sponsored by the Welding Research Council, important
papers presented before engineering societies and other reports of current interest.
No warranty of any kind expressed or implied, respecting of data, analyses, graphs or any other
information provided in this publication is made by the Welding Research Council, and the use of any
such information is at the user’s sole risk.
All rights are reserved and no part of this publication may be reproduced, downloaded, disseminated, or
otherwise transferred in any form or by any means, including photocopying, without the express written
consent of WRC.
Copyright © 1998 The Welding Research Council.

All rights, including translations, are reserved by WRC.
Printed in the United States of America.
ISSN 0043-2326
Library of Congress Catalog Number: 85-647116
Welding Research Council

20600 Chagrin Blvd.
Suite 1200
Shaker Heights, OH 44122
www.forengineers.org
FATIGUE STRENGTH REDUCTION
AND STRESS CONCENTRATION
FACTORS FOR WELDS IN PRESSURE
VESSELS AND PIPING
REPORT NO. 1
Interpretive Review of Weld Fatigue-Strength-Reduction
and Stress-Concentration Factors
C. E. Jaske
REPORT NO. 2
Fatigue-Strength-Reduction Factors for
Welds Based on NDE
J. L. Hechmer
E. J . Kuhn, III
WRC Bulletin 432—June 1998
Publication of this report was sponsored by the Pressure Vessel Research Council
of the Welding Research Council, Inc.
WELDING RESEARCH COUNCIL, INC.

3 Park Ave (27th Floor)
New York, New York 10016-5902
ISBN No . 1-58145-439-2
Library of Congress
Catalog Card Number : 99-60962
Copyright © 1998 by
Welding Research Council, Inc.
All Rights Reserved
Printed in U .S.A.
ii
CONTENTS
Report No . 1 : Interpretive Review of Weld Fatigue-Strength-Reduction

and Stress-Concentration Factors 1
Objectives 1
Introduction 1
Fatigue-Strength-Reduction Factor 1
Stress Co ncentration Factor 2
Stress Indices 2
Type of Weld 3
Welds in Pressure Vessel and Piping Design 3
Parameters that Affect FSRFs 6
Joint Configuration and Weld Detail 6
Welding Process and Practice 8
Cyclic Plasticity 11
Dissimilar Metal Welds 12
Residual Stress 12
Post-Weld Heat Treatment 13
Type and Degree of NDE Performed 13
Metallurgical Notch Effect 14
Status of Weld FSRFs 14
Published Data on Weld FSRFs 14
Generic Procedure for Determining FSRFs 24
Experimental Determination of FSRFs 24
Analytical Determination of FSRFs 25
Report No. 2: FSRF for Welds Based on NDE 27
Executive Summary 27
Nomenclature28
Abbreviations 28
Definitions 28
Introduction 28
Goal 29
Basis for the Approach 29
Position 29
Foundation 29
Post-Weld Heat Treatment 30
Background for Initiating This Project 30
FSRF Definition 30
Defining FSRF Based on NDE 30
The NDE Process 31
Weld Types 31
NDE Quality Levels 31
Basis For Establishing FSRF Values 31
Discussion of NDE Tech niques 32
RTm and UT m 33
Ground versus as-welded 33
Surface versus volumetric examination 33
Visual inspection 34
B ackside of partial and fillet welds 34
C onditions Peripheral to Quality Levels 35
ACKNOWLEDGEMENTS 35
REFERENCES 36
APPENDICES 36
I. Data From Other Fatigue Curves Used to Validate Weld FSRF 36
II. Comparisons to ASME Code S-N Fatigue Curve 43
III. FSRF Definition 49
1V Other Fatigue Issues 51
V. Level 7, Weld Toe Notches, and Non-Inspectable Weld Roots 52
FOREWORD
This Bulletin contains two reports prepared under the auspices of

the Steering Committee on Cyclic Life and Environmental Effects in
Nuclear Applications and the Committee on Piping, Nozzles and
Vessels of the Design Division in the Pressure Vessel Research Council
(PVRC) of the Welding Research Council (WRC) . Together, the two
reports present complementary technical descriptions and discussions
of the methodology and analytic procedures utilized in the ASME
Boiler and Pressure Vessel Codes to ensure that the effects of welds on
the fatigue performance of pressure equipment are adequately in-
cluded in the design process.
Report No . 1 presents an in-depth review of the background and
basis of the Fatigue Strength Reductions that are in the Code . Report
No. 2 covers a more specific aspect of the relation between fatigue
performance of welds and the type and amount of non-destructive
examination (NDE) performed on the weld . Both reports include
suggestions for increasing the understanding of the various factors
that influence the fatigue performance of welds and suggestions for
Code modifications that can be considered . It is hoped that these
reports will be utilized by the developers and users of the ASME Code
and will contribute to modifications and improvements in the Code.
S. Yukawa, Chr., Committee on Cyclic Life and Environmental
Effects in Nuclear Applications
G. L. Hollinger, Chr., Design Division
Report No.1 : Interpretative Review
of Weld Fatigue-Strength-Reduction
and Stress-Concentration Factors
C . E . Jaske
Objectives local concentrations of stress . These irregularities

may be design details of the joint configuration,
The objectives of this report are to (1) clarify the
normal variations in the profile of the weld, or weld
current procedures for determining values of fatigue-
flaws . Fatigue strength is the value of applied cyclic
strength-reduction factors (FSRFs), (2) collect rel-
stress corresponding to a certain value of cyclic life.
evant published data on weld-joint FSRFs, (3) inter-
Also, the weld metal and the heat-affected zone
pret existing data on weld joint FSRFs, (4) facilitate
(HAZ) of base metal adjacent to the weld may have
the development of a future database of FSRFs for
either lower or higher fatigue strengths than the
weld joints, and (5) facilitate the development of a
normal base metal . Such differences in fatigue strength
standard procedure for determining the values of
are a result of local variations in material composi-
FSRFs for weld joints . The main focus of this report
tion and microstructure . All of the above factors can
is on weld joints in Class 1 nuclear pressure vessels
reduce the local fatigue strength of a welded joint
and piping . However, relevant fatigue data on simi-
compared with that of the base metal in a pressure
lar weld joints for other applications, such as bridges
vessel or pipe . To account for reductions in local
and offshore structures, also are reviewed and inter-
fatigue strength, FSRFs, and/or stress-concentration
preted.
factors (SCFs) are employed in the fatigue design
of Class 1 pressure vessels, while stress indices are
Introduction
employed in the fatigue design of Class 1 piping.
Cracks can initiate and grow in pressure vessels
and piping subjected to cyclic loading if the ampli- Fatigue-Strength-Reduction Factor
tude and number of stress fluctuations are suffi- Article NB-3200 in Section III of the ASME Code
ciently high . This cracking process is defined as provides the following definition : "Fatigue strength
fatigue . The likelihood of fatigue damage in pressure reduction factor is a stress intensification factor
vessel and piping materials increases as the magni- which accounts for the effect of a local structural
tude of the local stress increases and as the inherent discontinuity (stress concentration) on the fatigue
fatigue resistance of the material decreases . Design strength . " Extending this definition to include weld
rules for Class 1 nuclear pressure vessels and piping joints, an FSRF is the ratio of the fatigue strength
require that the possibility of fatigue damage caused (S f) of a component without a discontinuity or weld
by cyclic loading be taken into account . Section III of joint to the fatigue strength of that same component
the ASME Boiler and Pressure Vessel Code (ASME with a discontinuity or weld joint:
Code) provides design fatigue curves for use in Sf (component without
design analysis . These curves relate allowable stress- discontinuity or weld joint)
intensity amplitude (S a ) to the number of cycles (N) FSRF = (1)
Sf(component with
for materials permitted by the Code.
discontinuity or weld joint)
Welded joints are the most likely sites for fatigue
damage in pressure vessels and piping because they Weld FSRFs can be determined experimentally by
often contain geometric irregularities that produce fatigue testing both unwelded and welded specimens
WRC Bulletin 432 1

or components of the same configuration . The welded The SCF is determined by elastic analysis . Hand-
specimens that are used for such tests contain joints books contain SCFs that have been determined for
typical of those used in actual components and are typical discontinuities encountered in mechanical
not simply standard fatigue specimens machined engineering equipment and structures . Many of
from welded joints. these were determined experimentally by means of
As is illustrated in Fig . 1, the relative fatigue photoelasticity. With the advent of computer technol-
strength of a welded joint vs base metal is usually a ogy and general-purpose finite element analysis
function of fatigue life . For this reason, the weld- software, SCFs can be readily calculated by means of
joint FSRF is usually a function of fatigue life . Only finite element stress analysis.
at long lives, where the fatigue curves approach a When an SCF is used to estimate a weld FSRF, it is
fatigue limit, does the FSRF approach a constant implicitly assumed that the reduced fatigue strength
value . When the FSRF increases as a function of of the weld joint is caused by the increased stress at a
fatigue life, as is shown in Fig . 1 and as is usually the local discontinuity associated with the weld . The
case for pressure vessel and piping materials, the SCF usually overestimates the geometric FSRF or
FSRF at the fatigue limit can be used as a conserva- what is commonly referred to as the effect of notches
tive estimate of the FSRF for shorter fatigue lives . In on fatigue strength, especially in the low-cycle fa-
general, FSRFs should be determined as a function tigue (LCF) regime where there is significant plastic-
of fatigue life, or the variation of FSRFs as a function ity at the notch . In the high-cycle fatigue (HCF)
of fatigue life should be accounted for analytically. regime, where there is only a very small amount of
The ASME Code provides procedures for experi- local plasticity at the notch, the SCF is still larger
mentally determining FSRFs . However, the cost and than the FSRF but closer to it than in the LCF
time required to perform fatigue tests on weld joint regime.
specimens and components make it desirable to use In fatigue design, the SCF is often adjusted to
analytical estimates of FSRFs for typical fatigue estimate the FSRF using an empirical relationship
design applications. that takes the notch sensitivity of the material into
account . However, both local properties of the weld
Stress-Concentration Factor and HAZ metal and residual stresses produced by
The stress-concentration factor (SCF) is commonly the welding process may also affect the fatigue
used to estimate the FSRF because fatigue strength strength of the weld joint . Additional adjustment
is related to stress amplitude . The SCF is the ratio of factors must then be applied to account for local
the peak local stress (S r ) at a geometric irregularity material properties and residual stresses.
to the nominal stress:
Stress Indices
Spat notch) The design of Class 1 piping is covered by Article
SCF = S (2)
P(away from notch) NB-3600 in Section III of the ASME Code . Instead of
log (Fatigue Life, cycles)

Fig . 1—Fatigue curves for a welded joint and base metal.
2 WRC Bulletin 432

using the FSRFs applicable to pressure vessels, the specimens used to develop these fatigue data do
these piping design rules employ stress indices . The not strictly contain ASME Code welds, the welds
local value of Sp is computed using equations that used in the fatigue test specimens are often of
apply stress indices to the internal pressure (K1 and comparable quality to ASME Code welds . For this
C 1 ), moment (K2 and C2 ), and thermal (K3 and C 3) reason, the information obtained from fatigue stud-
loadings . C l , C 2 , and C 3 are the secondary stress ies of other types of welds can provide valuable
indices, and K1 , K2 , and K3 are the local stress guidance on methods for establishing FSRFs.
indices . For example, for moment loading, the nomi-
nal bending stress (S b ) is used to compute Sp , as Welds in Pressure Vessel and Piping Design
follows : Welds are a key factor in the design of pressure
vessels and piping . Most joints in vessels and pipes
(3)
Sp = K2 C 2 S b are made by welding to avoid the potential problems
equal wall thickness, C 2 is equal to 1 .0 . If this joint is that are associated with mechanically fastened joints.
flush, K2 = 1 .1 . If this joint is as-welded, K2 = 1 .8. To provide background on welds used in nuclear
Thus, this K2 index for piping is the same as an components, this section briefly reviews information
FSRF for moment loading of a vessel girth butt weld. from Section III of the ASME Code.
The ASME Code divides welded joints into four
Type of Weld categories, A through D . Figure 2 illustrates typical
The type of weld can significantly affect the fatigue locations for joints of these categories . Category A
strength of a welded joint . The main focus of this includes longitudinal joints, all joints within spheres,
report is on the fatigue performance of welded joints and circumferential joints that connect hemispheri-
in Class 1 components that conform to the require- cal heads to main shells . Category B includes circum-
ments of the ASME Code . Section III, Subsection NB ferential joints within the main shell, communicat-
of the Code addresses nuclear power plant compo- ing chambers, or nozzles and circumferential joints
nents, while Section VIII, Division 2 of the Code that connect non-hemispherical formed heads to
addresses non-nuclear components . Any welds al- main shells . Category C includes joints that connect
lowed by these parts of the Code are covered by the flanges, Van Stone laps, tubesheets, or flat heads to
discussions in this report . Article NB-3200 provides the main shell, communicating chambers, or nozzles.
the requirements for design by analysis and, as Category D includes joints that connect communicat-
pointed out previously, defines fatigue strength reduc- ing chambers or nozzles to the main shell, spheres,
tion factor . Excluding bolting, this article indicates or heads ; it also includes joints that connect nozzles
that FSRFs greater than five do not need to be used, to communicating chambers . Each category of welded
except for crack-like defects and certain specific joint has different fabrication and examination re-
piping geometries . Thus, for typical welded joints in quirements.
Class 1 components, their effect on fatigue strength The most common type of welded joint used in
is taken into account by employing FSRFs up to five. pressure vessels and piping design is a butt joint.
The fatigue performance of welds other than those Butt joints utilize full penetration welds between
used in Class 1 components also is covered because a pieces that are in approximately the same plane.
large amount of fatigue data have been developed for Figure 3 illustrates typical butt joints for each
weldments used in other applications . These applica- category of welded joint . All Category A and most
tions include offshore structures, highway bridges, Category B and C joints are full penetration butt
pipelines, and other steel structures . Even though welds . The exceptions are that pipes (Category B and
Fig . 2—Typical welded joint locations and categories (from Section III of the ASME Code).
Interpretative Review 3
(When a does not exceed 30 deg ., joint
Category B Butt Joint meets requirements for butt joints)
Category B Angle Joint
Category C Butt Joint

Category D Butt Joint
Fig . 3—Typical butt joints (from Section III of the ASME Code).
C) and nozzles (Category C only) of 2-inch or less used . Fillet welds may be used to transitions
nominal size may be socket welded . Category B between joined parts or to provide a seal . If
includes angle joints with offset angles not greater fillet welds are used, they must be ground to
than 30 degrees . For Category B joints with integral provide a smooth surface with transition radii
backing strips or backing strips that are not re- at their intersections with the joined parts.
moved, the ASME Code requires that an FSRF of at 4. Partial penetration welded nozzles and branch
least two be used for fatigue design analysis . Cat- piping connections. Partial penetration welds
egory C joints must be either a butt joint or a full may be used to attach nozzles and make branch
penetration corner joint . Figure 4 shows acceptable piping connections, provided that they meet the
details for Category C full penetration corner joints. ASME Code design requirements . Further-
Category D joints are either full or partial penetra- more, specific details for attaching nozzles by
tion welds made using one of the five following means of partial penetration welds are pre-
details: scribed by the ASME Code . If a fatigue analysis
is required, an FSRF of at least four must be
1. Butt welded nozzles and branch piping connec-
employed for partial penetration welded nozzles,
tions . These connections are made by means of
while branch piping connections may be ana-
full penetration butt welds, and backing strips
lyzed in accordance with Article NB-3600.
must be removed, if they are used.
5. Oblique full penetration nozzles . These connec-
2. Corner welded nozzles and branch piping con-
tions are made by means of full penetration
nections . These connections are made by means
welds, and backing strips must be removed, if
of full penetration welds, and backing strips
they are used.
must be removed, if they are used.
3. Deposited weld metal of openings for nozzles Category A, B, and C joints in vessels and similar
and branch piping connections . Full penetra- full penetration welded joints in other components
tion welds are used to make connections to must be examined by radiography and by liquid
builtup weld-metal deposits on components, penetrant or magnetic particle inspection . Ultra-
and backing strips must be removed, if they are sonic examination also will be required in the near
4 WRC Bulletin 432

(a) (b)
Type 1 Corner Welds
0.25tn , but not

less than 1/4 in .,
the minimum for
either leg
Id) le)
Type 2 Corner Welds
Fig. 4—Acceptable weld details for Category C full penetration corner joints (from Section III of the ASME Code).
future . Fillet and partial penetration welded joints, corner welded nozzles also must be examined
socket welds, and structural attachment welds must ultrasonically to ensure that the fusion zone
be examined by liquid penetrant or magnetic par- and base metal under the attachment surface
ticle inspection . Special radiographic techniques that are free of lack of fusion and laminar defects.
may include multiple exposures are required for 2. Corner welded branch and piping connections
Types 1 and 2 Category C full penetration corner that do not exceed a nominal size of 4 inches
welds (see Fig . 4) . Also, ultrasonic examination must need to be examined only with liquid penetrant
be performed to ensure that the fusion zone and base or magnetic particle inspection.
metal under the attachment surface of the Type 2 3. Weld buildup deposits, weld fusion zones associ-
Category C welds are free of lack of fusion and ated with the deposits, and base metal under
laminar defects. These examination requirements the deposits must be examined ultrasonically
help ensure that the welded joints have the desired to ensure that they are free of lack of fusion and
integrity. laminar defects.
Radiography and liquid penetrant or magnetic 4. Partial penetration welds must be examined by
particle inspection also are required for Category D liquid penetrant or magnetic particle inspec-
joints in vessels and similar joints in other compo- tion.
nents, except as indicated below: 5. Full penetration welds in oblique vessel nozzles
1 . Just as with Type 2 Category C corner welds, need not be examined radiographically if ultra-
sonic and liquid penetrant or magnetic particle (NDE), and local variation of material properties . In
inspections are performed in a specified man- addition to the direct effect of each parameter, inter-
ner. actions among the parameters may affect the FSRF
6 . Oblique full penetration welds in branch and of a welded joint. The following subsections discuss
piping connections, pumps, and valves are to be the major effects of each of the parameters sepa-
examined in the same manner as corner welded rately. Important parameter interactions also are
branch and piping connections. discussed, as is appropriate.
Again, these examination requirements are pre-
scribed to ensure that the welded joints have the Joint Configuration and Weld Detail
desired integrity. Joint configuration and weld detail normally have
This brief review shows that welds are important the greatest influence on weld joint FSRFs because
in pressure vessel and piping design. The ASME they are the primary factors that determine the
Code provides both fabrication and examination amount of local stress in a welded joint . Increasing
requirements for welded joints to ensure that they local stress at a constant nominal stress decreases
have sufficient structural integrity and strength, fatigue strength and, thus, increases the FSRF . In
including fatigue strength . Thus, except for crack- other words, factors that increase the concentration
like defects and piping configurations for which of local stress also tend to increase the FSRF of a
values are given in Article NB-3680, FSRFs greater welded joint.
than five need not be used in establishing Level A Butt welds in pressure vessels are often associated
Service Limits (Article NB-3222) for typical welded with joints between sections of unequal thickness.
joints in Class 1 components . When integral backing The ASME Code requires a uniformly tapered transi-
is used or backing strips are not removed, an FSRF tion region between such sections if the thickness of
of at least two must be employed . An FSRF of at least the thicker section differs from that of the thinner
four must be applied for partial penetration welded section by more than 25% of the thinner section's
nozzles, and branch piping connections may be ana- thickness . Harvey (1991) points out that the stress
lyzed in accordance with Article NB-3600 . Thus, it is distribution in a uniformly tapered transition region
recognized that pressure vessel and piping design can be approximated by that in symmetrical wedge
has a direct effect on the fatigue strength of welded loaded as illustrated in Figure 5 . The theoretical
joints. stress concentration factor, K t , (ratio of the maxi-
mum stress to the average stress) for this wedge is
Parameters That Affect FSRFs given by the following relation:
A number of important parameters may influence Kt = 2 tan (a)/[a + 0 .5 sin (2a)] (4)
the FSRFs for welded joints in pressure vessels and
piping . These parameters include joint configuration Figure 6 shows a plot of K t vs a, where the indicated
and weld detail, welding process and practice, amount section taper is equal to the cotangent of a in
of cyclic plasticity, dissimilarity of metals joined, radians . When the section taper is 3 to 1 or more, the
residual stress, post-weld heat treatment (PWHT), maximum stress is less than 10% greater than the
type and degree of non-destructive examination average stress . For this reason, uniform transition
ttftttttttftt ftttttftttftt
Rtttttfttitt!
1111111 1111111Q
(o) ca) cc)
Fig . 5—Approximation of a tapered transition section (Harvey, 1991).
6 WRC Bulletin 432

1 1 1 I I 1 I I 1 1 1 I 1 1 1 I I 1 I i 1 1 1~
Taper = cot (a);

0 a is in radians
11 I 1 I I I 1 1111111 1 1 1 1 1 I I 1
0 2 4 6 8 10 12
Wedge Transition Taper

Fig . 6—Stress concentration factor as a function of wedge taper.
tapers of at least 3 to 1 are usually employed in weld profile . Also, the local details at the junction
pressure vessel design. between the taper and the uniform-thickness section
The tapered region at a welded butt joint is often will affect the stress concentration . Whereas the
offset as illustrated in Figure 7 . When such a joint is effect of offset can be taken into account by stress
axially loaded, the local stress near the weld joint is analysis, the effects of local details and profile are
increased . The effect of joint offset in pressure ves- usually accounted for by using an FSRF.
sels can be taken into account by means of stress Partial penetration welds have significantly lower
analysis . For piping, Article NB-3600 of the ASME fatigue strength than full penetration welds . As
Code provides the following equations for calculating pointed out by Harvey (1991), the SCF for a lack-of-
the secondary stress indices at welded transitions: penetration defect in an axially loaded butt-welded
0 .33(D °/t)0 .3 + 1 .5(8/t) but not >1 .8 (5a) plate is approximately 4 to 5 for defect lengths less
C1 =
than 20% of the wall thickness and is increasingly
C 2 = 1 .7 + 3 .0(8/t) but not >2 .1 (5b) greater than 4 to 5 as the defect length increases to
C 3 = 1 .0 + 0 .03(D ° /t) but not >2 .0 (5c) values greater than 20% of the wall thickness . With
such large SCFs, the corresponding FSRFs are ex-
where Do is the outside diameter, t is the wall pected to be large . In fact, as mentioned previously,
thickness, and 8 is amount of offset . The require- the ASME Code requires use of an FSRF of at least 4
ments for using Eqs . 5a, 5b, and 5c are given in for partial penetration welds because of the large
Article NB-3600. SCFs associated with such welds.
If an offset weld joint is not machined flat as shown As shown in Fig . 8, the amount of weld reinforce-
in Fig. 7, the fatigue strength is reduced additionally ment can significantly affect the fatigue strength of
by the stress concentration associated with the local welded joints (Harvey, 1991) . Weld reinforcement or
overfill refers to the extra thickness of the weld bead
compared with the thickness of the base metal . The
FSRF increases as the angle of reinforcement de-
creases or the amount of reinforcement increases.
The maximum fatigue strength (FSRF = 1 .0) is ob-
tained when the weld is ground flush, and the angle
reinforcement is 180 degrees or there is no reinforce-
ment.
When a single-V configuration is used for a butt
thickness (t) offset (8) weld, the joint will tend to warp in the direction of
the widest part of the V. As reported by Iida and Iino
Fig . 7—Offset weld joint . (1977), this type of warpage reduces fatigue strength.
CC
130 140 150 160 170 180
Reinforcement Angle (0), degrees

Fig . 8—Effect of buff-weld reinforcement on fatigue strength reduction factor (Harvey, 1991).
For this reason, a double-V configuration is typically 10 and 11 . The names indicated in these figures are
recommended for a welded joint that is subject to the standard terminology for weld flaws and defects
cyclic loading. (AWS, 1989) . Each of these flaws or cracks will
Fillet welds and welded socket joints can be suscep- reduce the fatigue strength of a welded joint . Since
tible to fatigue damage . Figure 9 shows the details fatigue cracking typically starts at a surface, surface-
and dimensions for Section III, ASME Code fillet connected flaws and cracks are usually more detri-
welds and welded socket joints . When fillet welds are mental to fatigue strength than imbedded flaws or
used for transitions between parts or for seals, they cracks . The welding process must be controlled to
must be ground to provide a smooth surface with avoid the formation of such flaws and cracks . Addi-
transition radii between the weld and base metal. tionally, the weld must be examined nondestruc-
This type of preparation enhances weld joint fatigue tively to ensure that significant flaws and cracks are
strength . Fillet welds (see Fig . 9) also can be used to not present or that locations that require repairs are
make structural attachments to pressure vessels, properly identified . Producing sound, fatigue-resis-
provided that an FSRF of at least four is used if the tant welded joints necessitates the use of proper
weld is subject to cyclic loading. welding processes and practices.
Girth fillet welds are used to make welded socket Some important factors that affect the fatigue
joints, as shown in Fig . 9 . If the minimum values of strength of welded joints are incorporated in Ameri-
the secondary stress indices for a welded socket joint can Welding Society (AWS) design fatigue curves
equal 1 .09 tn , where to is the wall thickness of the (AWS, 1994) . Although these fatigue curves, which
nozzle or branch connection reinforcement, then the are shown in Fig . 12, are not applicable to pressure
product C 2 K2 equals 4 .2 . The value of C 2 K2 can be
vessels and piping, they illustrate the effects of
reduced when the secondary stress indices exceed
important parameters on the fatigue strength of
1 .09 tn . Socket joints should be avoided for material welded joints as determined by fatigue testing such
and environmental conditions that can cause crevice joints . Each curve is based on a statistical analysis of
corrosion . The crevices associated with socket joints
fatigue data from tests of components or actual
are likely to increase the potential of corrosion
welded joints and represents 95% survival with 95%
fatigue at such joints . Thus, it is generally desirable
confidence for the indicated category.
to avoid fillet and welded socket joints in locations
Curve A is for plain unwelded carbon steel pipe,
subjected to significant cyclic loading.
whereas the other curves apply to various types of
Welding Process and Practice weld joints in circular sections of similar material . To
Welding process and practice can greatly affect the produce weld joints with the highest fatigue strength,
fatigue strength of welded joints . Improper welding the final pass is removed by grinding to produce a
can produce the flaws and cracks illustrated in Figs . flush joint . For example, Curve B in Fig . 12 shows
8 WRC Bulletin 432

GENERAL NOTE:
The size of an equal leg fillet weld is the leg length of the largest inscribed right isosceles triangle.
Theoretical throat = 0 .7 x size of weld.
(a) Equal Leg Fillet Weld
Theoretical throat
Surface of vertical member
Convex fillet weld
GENERAL NOTE:
The size of an unequal leg fillet weld is the shorter leg length of the largest right triangle which can pe inscribed
within the fillet weld cross section.
(b) Unequal Leg Fillet Weld
x min. to nominal pipe
7 x min .
to
wall thickness
1/16 in . approx.
before welding
1/16 in . approx.
ttt~~~ before welding
x min . 1 .4tn or the thickness of the hub, whichever

is smaller, but not less than 1 /8 in ., where
Cx min . = 1 .09tn
to = nominal pipe wall thickness
lc) Minimum Welding Dimensions for Id) Minimum Welding Dimensions for
Socket Welding Flanges Socket Welding Fittings
Fig . 9—Details and dimensions of fillet welds and welded socket joints (from Section III of the ASME Code).
the allowable cyclic stress range as a function of weld joints covered by the AWS Code, but not rel-
number of load cycles for flush full-penetration butt evant to this report.
welds . This curve is below that for plain pipe (Curve Metallurgical factors often influence weld joint
A) and above that for as-welded full-penetration butt fatigue strength . Hydrogen-induced cracks or non-
welds (Curve C 1 ) . Of course, one should be aware metallic inclusions in the HAZ or weld metal can act
that, in some cases, grinding may produce other as crack-initiation sites . Hydrogen-induced cracking
undesirable results . For example, grinding austen- is caused by improper welding procedure . Inclusions
itic stainless steel welds can lead to intergranular in the HAZ of the base metal are a result of the
stress-corrosion cracking (IGSCC) problems in sus- steel-making process, while inclusions in the weld
ceptible combinations of material and environment. metal are a result of the welding process . If corrosion
Curve D covers members with welded attachments attack is possible, inclusions can act as sites for the
and full-penetration cruciform and T joints . The initiation and growth of pits that subsequently be-
remaining curves in Fig . 12 are for other types of come locations for fatigue crack initiation . Some
(A)
UNDERBEAD CRACKS
(B)
(C)
UNDERFILL
UNDERFILL
(D)
(E)
(F)
Fig . 10—Flaws encountered in welded joints (AWS, 1989).
weld processes can cause elements to segregate joint detail, fatigue cracking often initiates at the
within the weld joint . This segregation can in turn weld toe and grows into the HAZ and base metal.
lead to local regions of material that are more Also, variations of material strength across the weld
susceptible to fatigue cracking or IGSCC than the joint can produce strain concentrations when the
overall weld joint. stress amplitude is large enough to cause cyclic
Local weld details play an important role in the plastic flow. These strain concentrations then act as
fatigue strength of welded joints . Because of weld- crack-initiation sites .
10 WRC Bulletin 432

LEGEND
1 CRATER CRACK
2 FACE CRACK
3 HEAT-AFFECTED ZONE CRACK
4 LAMELLAR TEAR
5 LONGITUDINAL CRACK
6 ROOT CRACK
7 ROOT SURFACE CRACK
8 THROAT CRACK
9 TOE CRACK
10 TRANSVERSE CRACK
11 UNDERBEAD CRACK
12 WELD INTERFACE CRACK
13 WELD METAL CRACK
IBl
Fig . 11—Cracks encountered in welded joints (AWS, 1989).
Cyclic Plasticity dence of yielding. In the HCF regimen (typically

Cyclic plasticity plays a major role in the fatigue >50,000 cycles to failure), the cyclic plastic deforma-
damage process . Some cyclic plastic deformation is tion is quite small and highly localized, and the
necessary for fatigue damage to occur . In the LCF material's macroscopic stress-strain response is elas-
regimen (typically <50,000 cycles to failure), signifi- tic.
cant cyclic plasticity occurs, and the material 's mac- Ductile steels and alloys, such as those used in
roscopic stress-strain response shows definite evi- pressure vessels and piping, are more tolerant of
100
(690)
50 2000
(345)
' CATEGORY A
nnai
20
(138)
CI AND X I
10 -''.11111110111rcr.
(69) F C2 AND X2
5
(34 .5) ;~\U•I~~Z!!C::~~---mil
nni.!Ij~UIll DT E K X11
\II
2
(13 .8) 1 50
2
2 10 4 4 6 8 10 5 2 4 6 8 10 6 2 4 6 8 10 7 4 6 8 10 8
CYCLES OF LOAD N
Fig . 12—AWS design fatigue curves for welded joints in circular sections (AWS, 1994).
defects and notches in the LCF regime than in the Residual Stress
HCF regime because local plastic flow tends to Residual stress has a major effect on the HCF
reduce the stress concentrating effects of defects and strength of welded joints . It acts like a mean stress
notches in the LCF regime . However, these stress that increases both the maximum and minimum
concentrating effects are not reduced in the HCF values of cyclic stress . In the LCF regime, residual
regime . For this reason, FSRFs tend to be relatively stress becomes decreasingly important as the num-
small at short fatigue lives and increase to larger ber of cycles to failure decreases because of the
values at long fatigue lives . This effect of cyclic corresponding increase in the amount of cyclic plas-
plasticity is evident in Fig. 12, where the Curves A ticity. Plastic deformation mechanically relieves the
through E tend to converge at short lives and diverge residual stress, so when the cyclic stress amplitude
at long lives . Thus, the FSRF should be a function of is equal to or greater than the cyclic yield strength,
fatigue life or the analytical procedures should ac- there is no significant effect of residual stress on
count for the effect of cyclic plasticity. fatigue strength.
Dissimilar Metal Welds The ASME Code accounts for residual stress by
Welds made between ferritic and austenitic alloys applying a mean-stress reduction factor to the de-
are called dissimilar metal welds . The fatigue sign fatigue curve (Jaske and O'Donnell, 1977) . The
strength of these welds is affected by the same design fatigue curve is reduced in the HCF regime to
factors that affect the fatigue strength of other account for the maximum effect of mean stress using
welds . Additionally, the unique nature of these welds the Goodman relation or some other appropriate
makes them particularly sensitive to thermal fa- parameter that approximates the experimentally
tigue cracking . Because ferritic and austenitic alloys observed fatigue behavior. The maximum mean stress
have significantly different thermal expansion and that a material can sustain at a given cyclic life is
thermal conductivity properties, dissimilar metal estimated by subtracting the stress amplitude at
welds experience higher local strain than similar that life from the material's cyclic yield strength . As
metal welds when they are subjected to thermal long as this difference is greater than zero, a mean-
cycling . Fatigue cracks develop in the HAZ of the stress effect is predicted, and the design fatigue
ferritic base metal of dissimilar metal welds . Ther- curve is reduced . This approach assumes that the
mal diffusion of carbon at high temperature can residual stress will always be as high as the cyclic
reduce the fatigue strength of the ferritic material. yield strength.
The complexity of this problem makes thermal fa- In some cases, the residual stress can be reduced
tigue susceptibility difficult to characterize by an by heat treatment or special fabrication procedures.
FSRF. Procedures have been developed for predict- Warm prestressing may redistribute residual stresses
ing the creep-fatigue life of dissimilar metal welds and reduce their severity but also may produce
(Roberts et al ., 1985) . Also, the thermal fatigue residual stresses in some locations . If the residual
strength of dissimilar metal welds can be improved stress can be accurately predicted or measured, the
by using nickel-based filler metals rather than iron- design fatigue curve can be adjusted to take into
chromium-nickel austenitic alloy filler metals . account the effect of the actual level of mean stress.
12 WRC Bulletin 432

Using actual rather worst-case mean stress can nealed at temperatures in the range of 1,750 to
greatly reduce fatigue design conservatism in the 2,050°F and then rapidly cooled to temperatures
HCF regime but will have little effect in the LCF below 900°F. However, this type of rapid cooling can
regime. reintroduce significant residual stresses into thick-
section components . For this reason, welded austen-
Post-Weld Heat Treatment itic stainless steel components are often given no
Post-weld heat treatment (PWHT) is employed to PWHT, and the existence of residual stresses equal
relieve residual stresses and to modify the material's to the yield strength is assumed for fatigue analysis.
microstructure . Relief of residual stresses is impor-
tant for welded joints that are subjected to HCF Type and Degree of NDE Performed
loading because it can significantly improve their The NDE flaw acceptance criteria affect the fa-
fatigue strength . Stress relief is less important for tigue resistance of welded joints . The nature and
LCF loading than for HCF loading because the cyclic severity of indications that are allowed to be left
plasticity associated with LCF tends to relieve re- without repair after NDE can affect weld joint fa-
sidual stresses, as discussed previously. Beneficial tigue strength . The type and degree of NDE per-
microstructural changes associated with PWHT are formed influence the probability of finding rejectable
usually associated with modifications of carbides or indications . Both of these factors should be taken
grain size in the HAZ. into account in establishing an FSRF for a welded
The ASME Code requires that all ferritic steel joint.
welds receive PWHT unless specified rules for exemp- As discussed previously, the ASME Code requires
tion from PWHT are satisfied . Base metals that have both radiographic and liquid penetrant or magnetic
similar welding characteristics are grouped by P particle method examinations for welded joints ex-
numbers to reduce the number of required welding cept for certain Category D joints where radio-
procedure qualifications . The temperature range and graphic examination is not possible or where ultra-
duration of PWHT are based upon P-number groups. sonic examination can be used in its place . Also, only
For example, P-No . 1 metals include low carbon liquid penetrant or magnetic particle method exami-
steels and have a holding temperature range of 1,100 nations are required for fillet, partial penetration,
to 1,250°F for PWHT. The minimum holding time at socket, and structural attachment welds . Radio-
temperature varies from half an hour to several graphic and ultrasonic examinations are volumetric
hours and depends upon section thickness . The techniques that can detect internal defects, whereas
P-No. 3 group of metals, which includes SA-508-2 the liquid penetrant and magnetic particle methods
and SA-533B-1 steels, have PWHT requirements can detect surface-breaking defects . Because fatigue
that are similar to those of the P-No . 1 group. cracking often initiates at the surface, it is important
Specific exemptions to mandatory PWHT are pro- to perform surface NDE on welded joints that are
vided by the ASME Code . Typically, these exemp- subjected to cyclic loading . In some cases, fatigue
tions apply to sections that are thinner than a cracking can initiate below the surface . For example,
specified nominal value, which depends on metal P the fatigue cracking of socket joints often initiates at
number, alloy carbon content, and the minimum defects in the root pass weld, and ultrasonic examina-
preheat for welding . They also are restricted to tion should be used to find such defects.
certain types of welds, such as circumferential butt The inspectability of a welded joint can influence
and socket welds that connect pipes and tubes to the probability of defect detection . For example,
nozzles and fillet welds . Thus, the exemptions are for corner joints (see Fig . 4) are generally more difficult
cases where welding residual stresses are expected to examine than butt joints (see Fig . 3) . In such
to be reasonably low. cases, the examination procedures should be de-
The commonly used austenitic stainless steels, signed to take into account weld-joint complexity to
such as Types 304 and 316, are in the P-No . 8 group ensure that the probability of detection (POD) is not
of metals . The ASME Code neither requires nor significantly reduced . If the POD is reduced, the
prohibits the PWHT of these metals because some effect of this reduction should be quantified, so the
heat treatments can impair their corrosion resis-
weld-joint FSRF can be appropriately modified.
tance . Welded austenitic stainless steel components
The defect acceptance criteria can have a direct
are sometimes stress relieved at temperatures in the effect on fatigue strength . As the acceptable indica-
range of 400 to 750°F. This type of PWHT does not tion size increases, the fatigue strength is expected
degrade their corrosion resistance, but it provides
to decrease. For radiographic examination, the follow-
only a moderate amount of stress relief . Heating
ing maximum values of elongated indications are
austenitic stainless steels to temperatures in the
permitted by the ASME Code:
range of 900 to 1,500°F can sensitize them, making
them susceptible to IGSCC . Stabilized or low-carbon 1. 1/4 inch for section thickness up to 3/4 inch,
grades are used to minimize this problem . To obtain 2. 1/3 of section thickness for section thickness
significant stress relief and avoid sensitization prob- from 3/4 to 2- 1/4 inch, and
lems, austenitic stainless steels are solution an- 3. 3/4 inch for section thickness over 2-% inch.
For surface examination, rounded indications greater Published values of Kf exhibit considerable vari-
than 3/16 inch are not allowed . Also, the type of ability and should be used only for the conditions and
indication can affect fatigue strength . For example, materials for which they were developed . The type of
the ASME Code does not allow any types of cracks. loading and the definition of failure can affect the
Cracks are much more likely to be fatigue-crack values of K f determined by fatigue testing . Small
initiation sites than rounded defects. specimens tested under load control typically fail
Metallurgical Notch Effect once a significant crack initiates . However, fatigue-
Local variations in material properties can cause a crack propagation can significantly contribute to
corresponding localization of stress or strain in a total fatigue life under load-controlled cycling of
loaded component even when no notch or geometric large specimens or displacement-controlled cycling.
discontinuity is present . This type of stress or strain Much of the variability in Kf values can be elimi-
concentration is referred to as a metallurgical notch. nated if fatigue life is defined as the number of cycles
Metallurgical notches can develop at a welded joint to crack initiation . To the extent possible this ap-
because there is often a significant variation in local proach was used in the current study . However,
materials properties at welds . A welded joint is because the definition of crack initiation is not the
usually designed so that the weld metal is slightly same for all fatigue studies, the type of loading and
stronger than the base metal to avoid failure in the failure criteria may still have affected the FSRFs
weld metal . If the strength mismatch between the that were developed in the current work.
weld and base metal is too large, the base metal can Because different investigators use different defini-
yield and concentrate strain well before the weld tions for fatigue failure, it was necessary to make
metal yields. judgments as to which data were reasonable to
In ferritic steels, the fine-grained HAZ is normally compare . For example, in the long-life regimen near
stronger than the coarse-grained HAZ . Under cyclic or more than 10 5 cycles to failure, the fracture of
loading, the weaker area may be subject to cyclic small 6-mm diameter specimens was assumed to
plasticity even though the adjoining material re- give similar results to through-wall leakage in pipe
mains elastic . If the local area undergoes cyclic specimens with similar wall thicknesses . Also, as
softening, there can be a corresponding local accumu- pointed out later in the discussion of a generic
lation of strain . However, if the local area undergoes procedure for determining FSRFs, different defini-
cyclic hardening, there can be a corresponding in- tions of fatigue failure based on a consistent crack-
crease in the local stress . This type of behavior is
initiation criterion are expected to give similar re-
complex and difficult to model analytically, so these
sults . Although this approach is subject to some
effects are usually quantified by testing welded
small errors, it is reasonable to use considering the
joints rather than by their analysis . The test results
inherent variability in fatigue data and the varia-
then reflect both metallurgical and geometric notch
effects. tions in test procedures used by various investiga-
tors.
Status of Weld FSRFs The determination of Kf values for a variety of
materials and notches requires a large amount of
The section reviews the current status of FSRFs fatigue testing. To reduce the need for fatigue test-
for welded joints in pressure vessels and piping. ing, values of Kf have been correlated with values of
First, published data on weld FSRFs are reviewed.
Kt , which can be determined from elastic stress
Then, the current procedures for establishing them
analysis . A notch sensitivity factor, q, is used to
are presented and discussed . This information pro-
characterize a material's notch sensitivity in fatigue,
vides the background for the generic procedure for
and q is defined by the following expression:
determining FSRFs that is presented in the next
section of this report. q = (Kf – 1)/(Kt – 1) (6)
Published Data on Weld FSRFs The value of q can vary from 0 to 1 . When there is no
It is a well-established engineering practice to use effect of a notch on fatigue strength, Kf = 1 and q = O.
FSRFs to account for the effect of notches on fatigue The maximum notch effect occurs when K f = Kt and
strength (Dieter, 1976) . As was illustrated for welds q = 1 . Figure 13 shows a plot of q vs notch root
in Fig . 1, the fatigue curves for notched and un- radius, r, that is widely used for fatigue design
notched specimens are compared on the basis of purposes (Peterson, 1974) . Since q varies as a func-
net-section stress, and the FSRF or fatigue-notch tion of r, it is not truly a material constant and is
factor, Kf, at the fatigue limit is the ratio of the based on the results of fatigue testing.
fatigue limit for unnotched specimens to the fatigue The following expression was used to estimate the
limit for notched specimens . When a material does values of q that are shown in Fig . 13 (Peterson,
not exhibit a fatigue limit, the value of Kf is deter- 1974) :
mined at a specific fatigue life, such as 10 8 cycles to
failure . q=1/(l+air) (7)
14 WRC Bulletin 432

Notch Radius, r, millimeters
0 1 3 4 5 ; 9 10
1.0
0.9
0.8
Q 0.1 1
0.6
14
4A.
6' n Quenched and Tempered Steel
J
(I Annealed or Normalized Steel
0.5 Average - Aluminum Alloy (bars and sheets)
.c 0 .4 F~
Note J
N 0 .3
Approximate Values
0 .2 (note shaded band) —
Not verified for very
0 .1 ' deep notches t Ir>4 -
0 .04 0.08 0.12 0 .16 0.20 0 .24 0 .28 0.32 0 .36 0 .40
Notch Radius, r, inches
Fig. 13—Fatigue notch sensitivity factor as a function of notch root radius (Peterson, 1974).
where a is Neuber's constant . The following values of less than Kt, an alternative approach is required . For
Neuber's constant were used to prepare the three this reason, Topper et al . (1969) applied Neuber ' s
curves shown in Fig. 13 : a = 0 .0025 for quenched and rule to LCF of notched specimens . Neuber's rule
tempered steel (most likely martensitic microstruc- states that Kt is equal to the geometric mean of the
ture), a = 0 .01 for annealed and normalized steel actual stress concentration factor, Kr , and the actual
(most likely ferritic or bainitic microstructures), and strain concentration factor, KE:
a = 0 .02 for aluminum alloy. Also, values of Neuber's Kt = (7,c u)v2 (8)
constant for low-alloy steel as a function of ulti-
mate tensile strength, S,, (Peterson, 1974), are pre- Values of K, and KE are not constant when yielding
sented in Fig . 14 . As shown, the value of a is an occurs . To apply Neuber's rule to fatigue, Topper et
inverse exponential function of ultimate tensile al . (1969) used Kf in place of Kt and defined KQ and K.
strength. in terms of stress and strain ranges . They found that
Since the above approach uses Kf values based on the parameter Kf(ASzeE) 112 = (OQOEE) w could be
the fatigue limit or long-life fatigue strength, it used to correlate the fatigue strength of notched
applies only to HCF. In the LCF regime, where specimens with the fatigue strength of unnotched
stresses exceed the yield strength and K f is much specimens, where AS is nominal stress range, De is
0 .02
c
0.015
5-
o 0 .005
tv
z
0
0 50 100 150 200 250 300

Ultimate Tensile Strength (S u ), ksi
Fig . 14—Neuber constant for fatigue of low-alloy steel (Peterson, 1974).
nominal strain range, E is elastic modulus, ,‘S,Q is fatigue behavior. Barsom and Vecchio (1997) and
local stress range, and De is local strain range . When Lundin (1988) have reviewed much of the available
the nominal stress-strain behavior is essentially fatigue data for weldments . Representative ex-
elastic, KAS=(OQOeE) 1 " 2 . To apply this parameter, amples of weldment fatigue testing are the studies of
Topper et al . (1969) estimated values of Kf from Kt Andrews and Jones (1991), Baudry, et al . (1986),
using Eqs . 6 and 7 . This approach provides a con- Booth (1980, 1981), Cole and Vittori (1991), Higuchi
stant FSRF that can be applied to LCF as well as et al. (1996a,b), Lawrence (1990), Nishijima et al.
HCF because it accounts for the effect of cyclic (1983, 1984a,b, 1985a,b, 1986a,b, 1987, 1990), Rad-
plasticity. ziminski et al . (1973), Vaessen et al . (1982), Vosi-
The parameter Kf(OSzeE) 112 can be applied in kovsky and Bell (1991), and Yagi et al . (1991a,b).
design using fatigue data for unnotched specimens, a This type of weldment fatigue data has been used to
Kf value calculated from Eqs . 6 and 7, and calculated develop design fatigue curves for highway bridges
values of AS and De . The data for unnotched speci- (Barsom and Vecchio, 1997), offshore structures
mens are plotted in terms of (OQO€ )112 vs fatigue life. (AWS, 1994), and Class 2 and 3 nuclear piping
This approach is restricted to predicting crack initia- (Rodabaugh, 1983) . This direct use of weldment
tion or failure where there is an insignificant amount fatigue data eliminates the need for employing
of crack propagation, such as is typically the case for FSRFs in design calculations.
the small laboratory specimens used to develop data The ASME Code design fatigue curves for Class 1
for ASME design fatigue curves . A fracture-mechan- nuclear pressure vessels and piping are based on
ics approach should be used to predict the portion of tests of polished unnotched specimens (Anon ., 1969;
fatigue life associated with crack propagation . Thus, Jaske and O'Donnell, 1977 ; English et al ., 1978), so
the parameter K f(ASOeE) 112 provides a straightfor- FSRFs should be employed when they are applied to
ward basis for defining FSRFs for fatigue-crack the design of weld joints . To evaluate fatigue design
initiation life. methods, Rodabaugh and Moore (1971), Rodabaugh
The Kf( . SAeE) 1i2 parameter also was used to (1983), Koves (1988), and Scott and Wilkowski (1995)
successfully predict the fatigue life of both notched have compared fatigue data for weldments and
weld-metal specimens and T-weldment specimens welded components with the ASME Code design
made from a carbon-manganese structural steel (van fatigue curves . Since the ASME curves incorporate
der Zanden et al ., 1972) . To account for the effect of safety factors, these comparisons do not yield direct
weld residual stress producing a cyclic mean stress measures of FSRFs.
under fatigue loading, they modified the parameter, Weldment fatigue-crack initiation data should be
based on the work of Smith et al . (1970), by incorpo- compared with the average fatigue curves from
rating maximum local stress, Amax, in place of &r. which the ASME design fatigue curves were devel-
Monotonic and cyclic stress-strain curves for the oped to obtain direct measures of FSRFs . In practice,
steel were used to compute the values of (Uma AE) 112 it often is difficult to obtain fatigue data for weld
for their fatigue life predictions . Thus, inelastic joints where crack-initiation life is well defined and
analysis is generally required to apply this approach there is a large amount of scatter in the fatigue data
to weldments . However, the analysis need not be used to define the average curves . Thus, when
complex and typically can be performed using simpli- fatigue lives for weld joints are compared with
fied rules and a hand-held calculator or personal fatigue lives for unnotched polished specimens, it
computer. should be kept in mind that the former fatigue lives
As shown by the above example, the determina- include some indeterminant amounts of crack growth
tion of FSRFs is generally much more complex for that are expected to vary as a function of size and
weldments than for notches . In addition to the section thickness, while the latter fatigue lives in-
influence of stress concentrations at local discontinui- clude significant data scatter.
ties, the fatigue behavior of weldments is influenced Once a significant fatigue crack initiates, remain-
by weldment imperfections and residual stresses ing fatigue life should be based on crack-growth
(Leshnoff and Riccardella, 1992) . Typical weldment analysis . Maddox (1986) showed that fracture-
imperfections, such as slag inclusions and porosity, mechanics methods and fatigue-crack propagation
are not crack-like defects, so they cannot be evalu- data can be used to develop fatigue curves for
ated using fracture mechanics and fatigue-crack weldments with lack of penetration defects . He used
propagation data . Thus, these imperfections should an initial crack depth of 0 .01 inch and a final crack
be considered as potential sites for fatigue-crack depth equal to 80% of the plate thickness in his
initiation and accounted for in FSRFs that are calculations . Thus, the fracture-mechanics approach
developed from fatigue tests of weld joint specimens can be used to predict the fatigue life of welded joints
or components. that contain significant crack-like fabrication flaws
The development of design information on weld- or fatigue cracks.
ments, including FSRFs, has concentrated on fa- The average fatigue curves that were used to
tigue testing of weldments because of the inherent develop the ASME design fatigue curves for carbon
complexities involved in analytically predicting their and low-alloy steels (Anon ., 1969) and austenitic
16 WRC Bulletin 432

steels (Jaske and O'Donnell, 1977) and (English et As reviewed by Rodabaugh (1983), the ASME Code
al ., 1978) are shown in Fig . 15 . The curves for carbon fatigue evaluation methods for Class 2 and Class 3
and low-alloy steels extend to only 10 6 cycles to piping are based on the Markl's (1952) fatigue test-
failure, whereas the curves for austenitic steels ing of SA106 Grade B steel pipe with girth butt
extend to 10 11 cycles to failure . All of the fatigue welds . For bending fatigue with failure defined as
curves fall fairly close together in the LCF region of the development of a through-wall crack in the pipe,
less than 10 4 cycles to failure . Beyond that region, he found that the mean trend of the data was
the curves for carbon and low-alloy steels fall well represented by the following expression:
below those for austenitic steels.
The curves for carbon and low-alloy steels were AS = 490,000Nf-o .2 (9)
lowered in the region of 105 to 10 6 cycles to failure, where AS is stress range in psi and N f is in cycles . In
N f, to include the maximum effect of mean stress. a similar fashion, Lawrence (1990) found that the
Beyond 10 6 cycles to failure, there are three curves mean trend of a large collection of fatigue data for
for austenitic steels . Curve A is an extension of the AISC Category C weldments (see Barsom and Vec-
curve for less than 106 cycles to failure and applies to
chio, 1997) was represented by the following expres-
base metal where the major loading is strain con-
sion :
trolled and the stress intensity range is no greater
than 27 .2 ksi . Curve B applies (1) to base metal
(10) AS = 394,050N f -° .196
where the major loading is load controlled and the
stress intensity range is corrected for mean stress Equation 10 has essentially the same slope as Eq . 9,
and is no greater than 27 .2 ksi or (2) to welded joints but it has a smaller coefficient, indicating that the
where the major loading is strain controlled and the Category C weldments have lower fatigue strength
stress intensity range is no greater than 27 .2 ksi. than the butt-weld pipes.
Curve C applies (1) to base metal where the major To account for weldments having different fatigue
loading is load controlled and the stress intensity strengths when their fatigue curves had the same
range is not corrected for mean stress and is not slope, Markl introduced the stress intensification
greater than 27 .2 ksi or (2) to welded joints where factor, i . Using i, the more general form of Eq . 9 is
the major loading is strain or load controlled and the
(11) i0S = 490,00N f - ° .2
stress intensity range is greater than 27 .2 ksi . Thus,
only curves B and C apply to austenitic steel welded Then, the fatigue curve for Category C weldments
joints . could be approximated by letting i = 490,000/
10 7 I 1119 111I119 11 11119 11 11119 11 1111111 1111111y 111111y 11 1 nu' 1 111119 1 I I I lI
Austenitic Steels, Curve A

Austenitic Steels, Curve B
Austenitic Steels, Curve C
- Carbon Steels
'y 10 6
0. --- Low-Alloy Steels
I
104 I I11111J 1 1IIIIIi
111111111 I IIlllld III01J 1 1111111 I 1 11111I 1 m id 1 1 I IIIni 1 t 111n~
10
10
10 10 1 3 5 107 9 1011
N f, cycles
Fig . 15—Average fatigue curves used to develop ASME design fatigue curves (Anon ., 1969 ; Jaske and O' Donnell, 1977;
English et al ., 1978) .
394,050 = 1 .24 . Of course, i = 1 .0 for girth butt welds 1—n /Sn
in carbon steel pipe. Ke=1 .0+ —1
n(m—1),3Sm
Markl's fatigue curve for butt-welded carbon steel
pipe (Eq . 7) and dotted line) is compared with the for 1 .0 < Sn/(3Sm) < m (12b)
average ASME Code fatigue curve for Class 1 carbon Ke = 1 .0/n, for Sn/(3S m ) m (12c)
steel components (solid line) in Fig . 16 . Markl's curve
for welded pipe falls well below the ASME curve for where S,, is the range of primary plus secondary
unnotched base-metal specimens, and the two curves stress intensity, Sm is the allowable stress inten-
diverge in the LCF region . This divergence implies sity, and m and n are material parameters . The
that the FSRF varies as a function of fatigue life . The value of m is 3 .0 for carbon steel, 2 .0 for low-alloy
problem with this comparison is that the stress steel, and 1 .7 for austenitic stainless steel . The
amplitudes for the tests of unnotched base-metal value of n is 0 .2 for carbon and low-alloy steel and 0 .3
specimens were computed from strain amplitudes, for austenitic stainless steel . The variation of Ke
whereas the stress amplitudes for Markl's work were as a function of S n/(3S m ) is illustrated in Fig . 17 . In
computed from the elastic stiffness for tests con- the elastic region, Ke = 1 .0 . In the plastic region, it
ducted in displacement control. increases to a limiting value of 5 .0 for carbon and
FSRFs determined by direct comparison of results low-alloy steel and 3 .33 for austenitic stainless
in the LCF range often depend on fatigue life be- steel.
cause they include effects of plasticity . As pointed out Using S m = 20 ksi, m = 3 .0, and n = 0 .2, Ke values
previously, an alternate approach is to determine a were calculated and multiplied times stress ampli-
single FSRF for HCF and then account for the effects tudes calculated from Markl's equation to construct
of plasticity by analysis . For example, elastic-plastic the short-dashed curve shown in Fig. 16 . Also,
analysis can be used to compute strain-based stress Markl's equation was compared with the ASME
amplitudes for Markl's test results, so they can be average curve at 10 5 cycles, and an FSRF of approxi-
compared with the ASME Code fatigue curve. mately 2 .0 was found . Applying the FSRF of 2 .0 to
When shakedown limits are exceeded, the ASME the ASME curve gave the long-dashed curve shown
Code provides a simplified method for elastic-plastic in Fig. 16 . These two curves are in good agree-
analysis . An elastic-plastic factor, Ke , is computed ment, indicating that the ASME Code evaluation
and multiplied times the stress amplitude used to methods with an FSRF of 2 .0 can be used to predict
enter the design fatigue curve:
the fatigue behavior of butt-welded carbon steel
Ke = 1 .0, for Sn/(3S m ) 1 .0 (12a) pipe .
I 107 1 1 1 111111 1 1 1 111111 1 1 1 1111 1 l I 1 11111 1 1 1 1 1 1 1±
Carbon Steels
Markl's Equation
Markl's Equation * Ke
V1 106 - - - (Carbon Steels)/2 .0
a-
W
w
Cl) 105
I 11111111 I t I l 11111 I 1 1 111111 I 11111111 1 1 1 1 1 11
10 2 10 3 10 4 105 10 6
N f , cycles
Fig . 16—Comparison of Markl's equation for girth butt-welded carbon steel pipe with the average ASME code fatigue
curve for carbon steel .
18 WRC Bulletin 432

I I_1111 ~ I I 1 I ~ 1111 ~ I I I 1 I1I
Carbon Steel
Low-Alloy Steel
Austenitic Steel
r t 1 + i t r r r t l l l i 1 + t l t l i + + I 11
0 1 2 3 4 5 6
Ratio of Primary Plus Secondary Stress
Intensity Range to Three Times the
Allowable Stress Intensity [S n/(3S m )]
Fig . 17—Elastic-plastic factor as a function of the ratio of primary plus secondary stress
intensity range to three times the allowable stress intensity.
In design applications, Ke values are applied to fatigue data for butt-welded plates of a carbon steel,
calculated values of stress amplitude so that the a low-alloy steel, and a stainless steel, respectively.
allowable number of fatigue cycles can be obtained They tested axially loaded specimens fabricated
from the design fatigue curve . In the current work, from 20-mm-thick plate and evaluated the effect of
K e values were applied to fatigue data only so that an stress ratio, R (ratio of minimum to maximum cyclic
FSRF could be estimated. The estimated FSRF of 2 .0 stress) on fatigue strength . Their data for tests at
can be used to evaluate FSRFs for cases where stress R = -1 (fully reversed loading) are compared with
intensification factors have been developed for other the ASME average curves in Figs . 19 and 20 . Analy-
carbon steel weldments, as follows: sis of their data showed that FSRF values of 2 .00 for
the carbon steel and 1 .25 for the low-alloy steel
FSRF = 2 .0i for carbon steel (12)
provided reasonable correlations with the ASME
For example, an FSRF for the AISC Category C curves for carbon and low-alloy steels, respectively
weldments (see Eq . 10) is estimated to be 2 .0 x (Fig . 19) . An FSRF value of 2 .50 provided a reason-
1 .24 = 2 .48. able correlation for their data on Type 304-HP
Rodabaugh (1983) compared the results of fatigue stainless steel (Fig . 20) . The FSRF value of 2 .00 for
tests of 6-inch diameter piping components of SA106B carbon steel is the same as that for Markl's data,
carbon steel and Type 304 stainless steel with Markl's while the FSRF value of 2 .50 for Type 304-HP
equation . These tests were performed by the General stainless steel is the same as that deduced from the
Electric Company (Heald and Kiss, 1974) and speci- data for butt-welded Type 304 stainless steel pipe
men failure was defined as the development of a (Fig . 18) . Thus, as noted by Rodabaugh (1983), butt
through-wall crack . Rodabaugh (1983) found that welds in stainless steel have larger FSRFs than
the results for butt-welded carbon steel pipe were comparable butt welds in carbon steel.
close to or slightly above Markl's mean curve . In Figure 21 presents a modified Goodman diagram
contrast, most of the results for butt-welded stain- of fatigue strength (stress amplitude vs mean stress)
less steel pipe were slightly below Markl 's mean at 10 7 cycles to failure for the data of Nishijima et al.
curve, as shown by the open circles in Fig . 18 . After (1983, 1984a, 1986a) . The symbols at stress ampli-
applying Ke values to the data for stainless steel pipe tudes greater than zero indicate the fatigue data,
(solid circles in Fig . 18), an FSRF of 2 .50 was applied while those at stress amplitudes equal to zero indi-
to Curve A for austenitic steels to obtain a reasonable cate the ultimate strength of the steels . The dotted
correlation with the test data . This FSRF value of lines are best fits to the data for each steel, while the
2 .50 for butt-welded stainless steel pipe is greater solid lines are Goodman relations for each steel . The
than the value of 2 .00 previously developed for Goodman relation is a straight line between the
butt-welded carbon steel pipe. fatigue strength at R = -1 (zero mean stress) and
Nishijima et al . (1983, 1984a, 1986a) developed the ultimate strength and is used to adjust the
I 107 1 1 1 111111 I 1 I i 11111 I 1 1 11111 1 11 111111 1 i I I I I l4
Curve A
(Curve A)/2 .50
Markl's Equation
o Type 304SS Butt Welds
106 • Type 304SS Butt Welds *Ke
105
' ---. :
I 11111111I 11111111I 11111/11I 11111111 I I 111111
104
10102 3 105 6
Nf, cycles
Fig . 18—Comparison of fatigue data for girth butt-welded type 304 stainless steel pipe with ASME Code fatigue curve and
Markl's equation.
1 I 1107 111111
I I 1 1 11111 I I 1111111 1 1 1 1 11111 I 11IIII11 1 1111119 1 1 111111
Carbon Steels
(Carbon Steels)/2 .00
Low-Alloy Steels
- - - (Low-Alloy Steels)/1 .25
106 o SB42 Butt-Welded Joints
• SPV50 Butt-Welded Joints
H
a
--> No Failure
Cl)
105
1 11111111
1 11111111 l 11111111 111111111 111111111 11111111 1 111111
104
10
102 101 3 104 5 6 107 8
N, cycles
Fig . 19—Fatigue of butt-welded steel plate (Nishijima et al ., 1983, 1984a).
20 WRC Bulletin 432

I
I
I
1
I 1
II 10 7 1111111
111111)
1111111
1
11111111
I 1 1111111
I111if
I 1 1111111 I 1 1111111
Curve A
Curve B
Curve C
- - - (Curve A-C)/2 .50
106 • 304-HP Butt-Welded Joints
.N
a
N
105
t.
I I 1111111I I 1111111 11 1111111I I 11111111 I 11111111 I 1111111- T"1 rrti,tl— 1-1-111+1

104
103
10 10 102 4 5 6 7 8 9
N, cycles
Fig . 20—Fatigue of butt-welded stainless steel plate (Nishijima et al ., 1986a).
---~--- Carbon Steel

---L--- Low-Alloy Steel
-4c — 304-HP Stainless Steel
20,000 1 1 1111111111 1
Dotted lines are best fits to data, -

solid lines are the Goodman relation, and -
dashed lines are maximum stress at yield -
a 15,000
0
0 20,000 40,000 60,000 80,000 100,000
Mean Stress, psi

Fig. 21—Effect of mean stress on fatigue strength of butt-welded plates at 10 7 cycles to failure (Nishijima et al ., 1983,
1984a, 1986a).
ASME fatigue curves for the maximum effect of the development of "standard" fatigue curves for
mean stress . The Goodman relation is a conservative 50-mm diameter joints and evaluations of the effects
representation of the data for the carbon steel, but a of diameter, weld bead shape, wall thickness, slip-on
non-conservative representation of the data for the gap, and root defects on fatigue strength. The rotat-
low-alloy steel and the stainless steel. ing bending fatigue testing included the develop-
The maximum effect of mean stress is assumed to ment of standard fatigue curves for 20-mm diameter
occur when the stress amplitude plus mean stress joints and evaluations of the effects of PWHT, wall
equals yield strength for carbon and low-alloy steels thickness, and root defects on fatigue strength . The
or cyclic yield strength for stainless steel . These 20-mm diameter joints had significantly lower fa-
stress amplitudes are indicated by the horizontal tigue strength in rotating bending than in four-point
dashed lines in Fig. 21 . The data for the carbon and bending . The stainless steel joints has better fatigue
low-alloy steel are conservatively bounded by the strengths than the carbon steel ones . At 10 7 cycles to
predicted maximum effect of mean stress, but the failure, FSRFs for the 50-mm diameter joints were 4
data for the stainless steel at R = 0 .5 fall below the for the stainless steel and 5 for the carbon steel.
predicted maximum effect of mean stress, indicating Fatigue strength was increased by increasing wall
that this approximation may be slightly non-conser- thickness, placing a final refinement welding pass on
vative for stainless steel . Additional studies of the the weld toe, eliminating the slip gap, and implement-
effect of mean stress or stress ratio of the fatigue ing PWHT. Formulas were developed for relating the
strength of stainless steel weldments are needed to size of weld root defects to the FSRF.
provide more data for this type of evaluation. The "standard" fatigue curves for socket weld
Recently, Higuchi et al . (1996a,b) evaluated the joints developed by Higuchi et al . (1996a,b) are
fatigue strength of small-diameter (20 to 50 mm), compared with the ASME average fatigue curves in
socket-welded pipe joints and developed FSRFs for Figs . 22 and 23 . An FSRF of 2 .50 was needed to
those joints based on the results of their testing. correlate the results for carbon steel joints with the
They performed rotating bending fatigue tests on ASME average curve, as shown in Fig . 22 . In con-
carbon steel and Type 316L stainless steel joints to trast, an FSRF of 4 .00 was needed for the stainless
develop fatigue curves up to 108 cycles to failure steel joints, as shown in Fig . 23 . The value for carbon
(Higuchi et al ., 1996a) . Also, they performed four- steel joints is different from that of Higuchi et al.
point bending fatigue tests on joints of these same (1996b) because they based their value of 5 .00 on
steels to develop fatigue curves up to 10 7 cycles to their own fatigue data instead of on the ASME
failure (Higuchi et al ., 1996b). average fatigue curve . These FSRFs for socket weld
The four-point bending fatigue testing included joints are larger than those for butt welds.
11 7 10 I 1 1111111 111111111 11111111` 11111111` 111111 111111111 1111111/
Carbon Steels
106
(Carbon Steels)/2 .50
Std Sockets, 4 Pt Bending
- - - Std Sockets, Rotating Bending t
104
1 1 1111111 1 1 1111111 111111111 II I 1 11111 II I I t 1111111 11111111

103 111111
102
10 101 3 4 5 6 7 8
N, cycles
Fig . 22—Fatigue of carbon steel socket weld joints (Higuchi et at ., 1996a,b).
22 WRC Bulletin 432

10 7 I I I 111111 I I1111111 1111111 111111111
- Curve A
- Curve B
Curve C
--(Curve A-C)/4
106 Std 316L Sockets,
4 Pt Bending
- Std 316L Sockets,
Rotating Bending
105
N
--
104
1 11111111 111111111 111111111 I II 111111 I II111,11 1 11111111 111111111 I ItMI

10 3
10
10
10 10 1 2 103 104 05 6 107 8 109
N, cycles
Fig . 23—Fatigue of stainless steel socket weld joints (Higuchi et al ., 1996a,b).
Scott and Wilkowski (1995) showed that fatigue Fig. 24 . The results are in the region of 10 5 cycles to
data for full-scale carbon steel butt-welded pipe can failure and fall well below the ASME curve and
fall well below the average ASME fatigue curve for Markl equation . An FSRF of 4 .00 is required to
carbon steel . The data that they reported are corn- correlate these data with the average ASME curve.
pared with the ASME curve and Markl's equation in Scott and Wilkowski (1995) attributed this low fa-
I I I I III1 1 I I 111111 I I I 1 IWI I I I I 1111' I I I 11112
Carbon Steels
Markl's Equation
- (Carbon Steels)/4 .00
• Full-Scale Component Welds
a
N
W
w
U)
N f, cycles
Fig . 24—Full-scale butt-welded pipe fatigue data (Scott and Wilkowski, 1995).
tigue strength to ASME Code acceptable imperfec- following subsections outline a generic approach
tions in the weld joints that provide sites for crack that can be used to develop FSRFs for welded joints
initiation under cyclic loading. If these fatigue lives used in pressure vessels and piping.
are predominantly crack propagation, the fracture-
mechanics approach used by Maddox (1986) may be Experimental Determination of FSRFs
applicable. Fatigue testing of welded joints is required to
develop FSRFs experimentally. The joints should be
produced using materials and procedures that are
Generic Procedure for Determining FSRFs
representative of their intended use . They should be
Values of FSRFs can be determined experimen- inspected to ensure that they contain no indications
tally or analytically. However, even the analytical that are unacceptable to the ASME Code . The weld-
approaches for determining FSRFs usually require joint test specimens should be subjected to cyclic
an experimentally based material parameter, such bending, tension-compression, or pressure loading
as the Neuber's constant defined by Eq . 7 . Experimen- using acceptable testing equipment . Figure 25 shows
tal approaches are based on fatigue testing . The typical weld joints and types of loading that are used
experimental procedures discussed in this section for fatigue testing. Straight butt joints or tee joints
are directed toward Class 1 components and may be are usually used for piping specimens . Vessels or
somewhat different from existing test procedures for pipes with end caps are used for pressure cycling.
Class 2 and 3 components, especially piping prod- The type of test and loading should be selected to
ucts . Based on this report, those existing procedures reflect those expected in the components for which
should be reviewed and modified if necessary. The the FSRF is being developed.
1 . Full-Penetration T Joint — 5. Fillet Joint — Bending Load

Bending Load
1-
6. Fillet Joint — Axial Load

2. Cruciform T Joint —
Bending Load
L
3. Butt Joint — Bending Load 7. Full-Penetration K Joint —
Bending Load
4. Cruciform T Joint — Axial Load

8. Fillet T Joint — Bending Load
Fig . 25—Typical weld joints used for fatigue testing .
24 WRC Bulletin 432

Failure of the specimens should be defined by Kharshafdjian . This database would incorporate the
crack initiation using a nondestructive technique fatigue data for weld joints cited in this report.
such as ultrasonic testing or measurement of electric Additionally, papers published by Commission XIII
potential drop, a decrease in load under displace- and Iida and co-workers should be obtained and
ment controlled cycling, an increase in displacement reviewed for relevant data . Those data should then
under load controlled cycling, or a leak in a pressur- be added to the database . Also, the fatigue data
ized specimen . The technique should be calibrated so developed for PVRC by Scavuzzo and Markl's fatigue
that crack initiation corresponds to a crack that is data on socket joints should be included in the
approximately 0 .050-inch deep. database . Finally, numerical examples should be
Specimens should be tested at several different developed to illustrate the analytical development of
stress amplitudes using constant amplitude cycling FSRFs.
to failure, so a plot of stress amplitude vs number of
cycles to failure can be developed . Data should be References
developed over a range of fatigue lives that vary by a Andrews, R . M . and Jones, R. L ., "The Fatigue Performance of Single
factor of at least 10 . The applied stress range calcu- Sided Closure Welds in Offshore Structures," Proceedings of the 10th
International Conference on Offshore Mechanics and Arctic Engineering,
lated using elastic analysis should be no more than Vol . III-B, American Society of Mechanical Engineers, New York, 1991, pp.
377-383.
2S y to avoid the need for elastic-plastic analysis in Anon., Criteria of the ASME Boiler and Pressure Vessel Code for Design by
computing the nominal stress amplitude . The fa- Analysis in Sections III and VIII, Division 2, American Society of Mechani-
cal Engineers, New York, 1969.
tigue data should be plotted on a graph of stress AWS, Standard Welding Terms and Definitions, ANSI/AWS A3.0-89,
American Welding Society, Miami, 1989.
amplitude vs number of cycles to failure and com- AWS, Structural Welding Code—Steel, ANSI/AWS D1 .1-94, American
Welding Society, Miami, 1994.
pared with the ASME average fatigue curve, cor- Barsom, J . M . and Vecchio, R. S ., "Fatigue Behavior of Weldments," WRC
rected for the effect of mean stress, for the same Bulletin, 422, June 1997.
Baudry, G ., Amzallag, C . and Bernard, J . L., "Fatigue Behavior of
material as the weld joint . A curve that is shifted Heterogeneous Weldments Between Stainless and Low Alloy Steels,"
Fatigue Life : Analysis and Prediction, ASM International, Materials Park,
below the average ASME curve by a constant factor OH, 1986, pp . 381-392.
should be developed . This curve should fit the experi- Booth, G . S ., "Constant Amplitude Fatigue Tests on Welded Steel Joints
Performed in Air," European Offshore Steels Research Seminar, Welding
mental data in a reasonable and conservative fash- Institute, Cambridge, UK, 1980, pp . III/4-1 to III/4-15.
Booth, G. S ., "Constant Amplitude Corrosion Fatigue Strength of Welded
ion . An extrapolation of the data to longer cyclic lives Joints," Fatigue in Offshore Structural Steels, Paper 2, Thomas Telford Ltd .,
London, 1981, pp . 5-16.
should not fall below this curve . The factor by which Cole, I . and Vittori, O ., "TMCP Steels and the Thickness Effect,"
this curve is shifted below the average ASME curve Proceedings of the 10th International Conference on Offshore Mechanics
and Arctic Engineering, Vol . III-B, American Society of Mechanical Engi-
is then the experimentally determined FSRF . This neers, New York, 1991, pp. 331-337.
Dieter, G . E ., Mechanical Metallurgy, Second Edition, McGraw-Hill, Inc .,
FSRF value should be compared with other pub- New York, 1976.
lished data on weld-joint FSRFs to ensure that it is English, W. F., Greene, R. L., Hughes, D . A. and Post, R . I ., "Discussion of
Fatigue Design Criteria for Pressure Vessel Alloys," Journal of Pressure
reasonable for the type of joint that was tested. Vessel Technology, 100, 1978, pp . 236-238.
Harvey, J . F., Theory and Design of Pressure Vessels, Second Edition, Van
Nostrand Reinhold, New York, 1991.
Analytical Determination of FSRFs Heald, J . D . and Kiss, E ., "Low Cycle Fatigue of Nuclear Pipe Compo-
nents," Journal of Pressure Vessel Technology, 96, 1974, pp . 171-176.
Analytical determination of FSRFs is typically Higuchi, M ., Nakagawa, A ., Chujo, N., Iida, K., Matsuda, F. and Sato, M.,
based on calculating the stress concentrating effects "Effects of Weld Defects at Root on Rotating Bending Fatigue Strength of
Small Diameter Socket Welded Pipe Joints," Pressure Vessels and Piping
of weld details and profiles . There are no widely Codes and Standards, PVP-Vol . 338, Vol. 1, American Society of Mechanical
Engineers, New York, 1996a, pp . 3-10.
accepted models for determining FSRFs from local Higuchi, M ., Nakagawa, A ., Hayashi, M ., Yamauchi, T., Saito, M., Iida,
K., Matsuda, F. and Sato, M ., "A Study on Fatigue Strength Reduction
variations in material structure and properties . Elas- Factor for Small Diameter Socket Welded Pipe Joints,"Pressure Vessels and
tic stress analysis should be performed to determine Piping Codes and Standards, PVP-Vol. 338, Vol . 1, American Society of
Mechanical Engineers, New York, 1996b, pp. 11-19.
the stress concentration factor for the welded joint. Iida, K. and Iino, N ., "Effect of Angular Distortion on Fatigue Strength of
Transverse Butt Welds in High Strength Steels," International Institute of
Since it is normally difficult to model all of the local Welding Paper XIII, 1977, pp . 827-877.
variations at a weld joint, a conservative approxima- Jaske, C . E . and O'Donnell, W. J ., "Fatigue Design Criteria for Pressure
Vessel Alloys, " Journal of Pressure Vessel Technology, 99, 1977, pp . 584-592.
tion of the nominal weld profile should be employed. Koves, W. J., "Comparison Pressure Vessel and Piping Criteria," Paper
No. 88-PVP-5, American Society of Mechanical Engineers, New York, 1988.
The calculated value of Kt then should be used Lawrence, F. V., "Coping With Weldment Geometry in Fatigue Design:
Problems and Options," Sixth Annual North American Welding Research
with Eqs . 6 and 7 and estimates of the parameters Conference, Columbus, OH, 1990.
"a" and "r" to estimate a value of K f. As reviewed Leshnoff, S . and Riccardella, P. C ., Metal Fatigue in Operating Nuclear
Power Plants, EPRI TR-100252, Electric Power Research Institute, Palo
earlier, correlations of "a" with tensile strength have Alto, CA, 1992.
Lundin, C . D ., "Bibliography on Fatigue of Weldments and Literature
been developed for carbon and low-alloy steels . Simi- Review on Fatigue Crack Initiation from Weld Discontinuities," WRC
lar correlations need to be developed for stainless Bulletin, 333, May 1988.
Maddox, S . J ., "Fatigue Analysis of Welded Joints Using Fracture
steels . If such correlations are not satisfactory, an- Mechanics, " Fatigue Life: Analysis and Prediction, ASM International,
Materials Park, OH, 1986, pp . 155-166.
other procedure for estimating Kf values for stainless Markl, A. R . C ., "Fatigue Tests of Piping Components, " Transactions
ASME, 74, 1952, pp . 287-303.
steels would have to be developed . The value of "r" Nishijima, S ., et al ., Data Sheets on Fatigue Properties for Butt Welded
used in these correlations should correspond to the Joints of SB42 Carbon Steel Plate for Boilers and Other Pressure Vessels—
Effect of Stress Ratio, NRIM Fatigue Data Sheet No . 34, National Research
expected size of local defects at the weld joint. Institute for Metals, Tokyo, 1983.
Nishijima, S ., et al ., Data Sheets on Fatigue Properties for Butt Welded
Joints of SPV50 Steel Plate for Pressure Vessels—Effect of Stress Ratio,
Future Work NRIM Fatigue Data Sheet No. 40, National Research Institute for Metals,
Tokyo, 1984a.
For future work, an FSRF database should be Nishijima, S ., et al ., Data Sheets on Fatigue Crack Propagation Properties
established . This database should be patterned after for Butt Welded Joints of SB42 Carbon Steel Plate for Boilers and Other
Pressure Vessels—Effect of Stress Ratio, NRIM Fatigue Data Sheet No . 41,
those currently being developed for PVRC by Gary National Research Institute for Metals, Tokyo, 1984b.
Nishijima, S ., et al ., Data Sheets on Fatigue Crack Propagation Properties Rodabaugh, E . C . and Moore, S . E ., Phase Report No. 115-10 on
for Butt Welded Joints of SPV50 (Si-Mn, 500N/mm2 YS) Steel Plate for Comparisons of Test Data With Code Methods for Fatigue Eval-
Pressure Vessels—Effect of Stress Ratio, NRIM Fatigue Data Sheet No . 46, uation, ORNL-TM-3520, Oak Ridge National Laboratory, Oak Ridge, TN,
National Research Institute for Metals, Tokyo, 1985a. 1971.
Nishijima, S ., et al ., Data Sheets on Fatigue Properties for Weld and HAZ Scott, P. M. and Wilkowski, G. M ., "A Comparison of Recent Full-Scale
Materials of SPV50 (Si-Mn, 500N/mm2 YS) Steel Plate for Pressure Vessels, Component Fatigue Data With the ASME Section III Fatigue Design
NRIM Fatigue Data Sheet No . 47, National Research Institute for Metals, Curves," Fatigue and Crack Growth : Environmental Effects, Modeling
Tokyo. 1985b. Studies, and Design Considerations, PVP-Vol . 306, American Society of
Nishijima, S ., et al., Data Sheets on Fatigue Properties for Butt Welded Mechanical Engineers, New York, 1995, pp . 129-138.
Joints of SUS304-HP (18Cr-8Ni) Hot Rolled Stainless Steel Plate—Effect of Smith, K. N ., Watson, P. and Topper, T. H ., "A Stress-Strain Parameter for
Stress Ratio, NRIM Fatigue Data Sheet No. 53, National Research Institute the Fatigue of Metals," Journal of Materials, 5, 1970, pp. 767-778.
for Metals, Tokyo, 1986a. Topper, T. H ., Wetzel, R . M. and Morrow, J. D., "Neuber's Rule Applied to
Nishijima, S ., et al ., Data Sheets on Fatigue Crack Propagation Properties Fatigue of Notched Specimens, " Journal of Materials, 4, 1969, pp . 200-209.
for Butt Welded Joints of SUS304-HP (18Cr-8Ni) Hot Rolled Stainless Steel van der Zanden, A. M., Robins, D. B . and Topper, T. H ., "Fatigue Life
Plate—Effect of Stress Ratio, NRIM Fatigue Data Sheet No . 54, National Prediction for Weldments With Internal Cavities," Testing for Prediction of
Research Institute for Metals, Tokyo, 19866. Material Performance in Structures and Components, ASTM STP 515,
Nishijima, S ., et al ., Data Sheets on Fatigue Properties for Weld and HAZ
Materials of SB42 (C-Si, 420N/mm2 TS) Carbon Steel Plate for Boilers and American Society for Testing and Materials, Philadelphia, 1972, pp.
Other Pressure Vessels, NRIM Fatigue Data Sheet No . 57, National 268-284.
Research Institute for Metals, Tokyo, 1987. Vaessen, G . H . G ., de Back, J. and van Leeuwen, J . L., "Fatigue Behavior
Nishijima, S ., et al ., Data Sheets on Fatigue Properties for Weld and Base of Welded Steel Joints in Air and Seawater,"Journal of Petroleum Technol-
Metals of SUS304-HP (18Cr-8Ni) Hot Rolled Stainless Steel Plate, NRIM ogy, 34, 1982, pp . 440-446.
Fatigue Data Sheet No . 65, National Research Institute for Metals, Tokyo, Vosikovsky, O . and Bell, R ., "Attachment Thickness and Weld-Profile
1990. Effects on Fatigue Life of Welded Joints," Proceedings of the 10th Interna-
Peterson, R . E ., Stress Concentration Factors, John Wiley & Sons, New tional Conference on Offshore Mechanics and Arctic Engineering, Vol . III-B,
York, 1974. American Society of Mechanical Engineers, New York, 1991, pp . 339-362.
Radziminski, J . B ., Srinivasan, R., Moore, D., Thrasher, C . and Munse, Yagi, J ., Machida, S ., Tomita, Y., Matoba, M . and Soya, I ., "Influencing
W. H ., Fatigue Data Bank and Data Analysis Investigation, UILU-ENG-73- Factors on Thickness Effect of Fatigue Strength in As-Welded Joints for
2025, University of Illinois, Urbana, IL, 1973. Steel Structures," Proceedings of the 10th International Conference on
Roberts, D. I., Ryder, R . H . and Viswanathan, R ., "Performance of Offshore Mechanics and Arctic Engineering, Vol . III-B, American Society of
Dissimilar Welds in Service," Residual-Life Assessment, Nondestructive Mechanical Engineers, New York, 1991a, pp. 305-313.
Examination, and Nuclear Heat Exchanger Materials, PVP-Vol . 98-1, Yagi, J ., Machida, S ., Tomita, Y., Matoba, M . and Soya, I., "Thickness
American Society of Mechanical Engineers, New York, 1985, pp . 55-63. Effect Criterion for Fatigue Strength Evaluation of Welded Steel Struc-
Rodabaugh, E. C ., Comparisons of ASME Code Fatigue Evaluation tures,"Proceedings of the 10th International Conference on Offshore Mechan-
Methods for Nuclear Class 1 Piping With Class 2 or 3 Piping, NUREG/CR- ics and Arctic Engineering, Vol . III-B, American Society of Mechanical
3243, E . C . Rodabaugh Associates, Hilliard, OH, 1983 . Engineers, New York, 1991b, pp . 315-322.
26 WRC Bulletin 432

Report No. 2 : Fatigue-Strength-Reduction Factors
Based on NDE
J. L. Hechmer and E . J . Kuhn, III
Executive Summary welds for analytical use of FSRF. The quality is in

the material, not in the analysis.
This report was developed in a project that was
Fatigue damage originates at notches and flaws
initiated within the Babcock & Wilcox Co ., Nuclear
and NDE is intended to detect flaws, allowing for
Equipment Division . Its development is co-funded
their removal . The different NDE techniques have
with the PVRC (Grant 97-22) . It addresses applying different detection capabilities, which lead to levels
a fatigue-strength-reduction factor (FSRF) to a weld
of quality depending on which techniques are ap-
surface, based on the non-destructive examination plied . The base condition is the current ASME Code
(NDE) that is performed . The development is fo- (1997) requirements for Category A welds, i .e ., full
cused on Class 1, pressure vessels of the ASME volumetric and full surface examination . The ASME
Boiler and Pressure Vessel Code, Section III, NB and Code approach and requirements indicate that an
NC-3200 and Section VIII Division 2 (ASME Code, FSRF = 1 .0 is adequate for Category A welds . At the
1997). other extreme, the root of a fillet weld receives no
Welds are suspect as having a lower fatigue life NDE ; its appropriate FSRF may be as high as 4.0 . All
than base metal . This is based on the assumption recommended FSRF are between 1 and 4.
that welds are more prone to flaws and notches than A total of seven quality levels are defined in terms
base metal . This may be true, but the ASME Code of the required NDE . FSRF are applied to these
(1997) requires NDE to ensure that the weld metal levels for three weld types : full penetration welds,
has high quality relative to flaws and notches . It is partial penetration welds, and fillet welds . With a
the position of this report that the fatigue life is a brief title, the seven quality levels are defined as
function of the quality of the material and the NDE requiring the following NDE:
gives an assessment of this quality . This leads to the
conclusion that weld metal will have a fatigue life 1. RT or UT and MT or PT with visual (full
consistent with the prediction of the ASME Code S-N volumetric and surface examination)
2. RT m or UTm and MT or PT with visual (RTm and
curves and equivalent to that of base metal provided
a full NDE is applied . With reduced NDE, the UT m are partial volumetric)
application of an FSRF in the analysis maintains 3. MT or PT with visual (no volumetric)
consistency. 4. RT m or UTm with visual
The ASME Code (1997) has additional require- 5. Visual only
ments that are coupled with the NDE requirements 6. RT, RT m , UT, UTm (no surface examination)
to produce quality welds . High quality welds start 7. None
with the qualified welder, demonstrated weld proce- The appropriate FSRF is to be applied to a weld
dures, and testing requirements . Also, the ASME surface depending on the quality level of that sur-
Code has rules for weld profiles and post-weld heat face . The root surface and the final surface can have
treatment (PWHT) requirements . In the global view, different quality levels.
the ASME Code is consistent in addressing the Except for the root of partial penetration and fillet
factors related to obtaining quality welds . It is not welds, each weld type addresses machined (ground)
the intent of the project to exchange high quality and as-welded surfaces . It is presumed that as-
WRC Bulletin 432 27

welded surfaces are adequate for meaningful RT, UT, FSRF = fatigue-strength-reduction factor(s)
MT, or PT, when required . The FSRF are presented NDE = non-destructive examination(s)
for each category in matrix form in Table 1 of the MT = magnetic particle testing (NDE)
report and are also presented in Table 2, using an PT = dye (liquid) penetrant testing (NDE)
increasing FSRF format that is related to quality RT = radiographic (x-ray) testing (NDE)
level. UT = ultrasonic testing
The report develops and defines the basis for each VT = visual examination testing
FSRF. For example, it explains why one NDE tech- NED = Babcock and Wilcox Co . Nuclear Equip-
nique has a greater impact on fatigue life than ment Division
another technique, i .e ., its omission requires a higher PVRC = Pressure Vessel Research Council
FSRF. The report also gives a more definitive discus- PWHT = post-weld heat treatment
sion of the weld quality levels . Peripheral issues PV = pressure vessel
such as cladding and root pass MT are discussed but SCF = stress-concentration factor(s)
are not quantified for FSRF. Thus, the report ad- S-N = stress vs cycles (N)
dresses the peripheral issues that have been raised. SWC = Structural Welding Code
The proposed FSRFs are presented without a TWI = The Welding Institute
variation based on stress level . This implies that the WRC = Welding Research Council
FSRF is the same regardless of stress level, i .e ., it
has the same value at very high applied stress (e .g ., Definitions
twice yield) as at the endurance limit . This is a
conservative assumption . However, the attenuation Design factor (2 & 20)—A factor applied to the best
of FSRF as a function of applied stress level is well fit curve of test data to obtain an S-N fatigue design
beyond the scope of this project . The topic is briefly curve . The ASME Code (1997) design factor is ob-
addressed in Appendix III, FSRF Definition. tained by applying the more conservative of 2 on
The relationship between FSRF and quality level stress or 20 on cycles, applied to the best fit of the
is evaluated based on published test data and on test data.
fatigue curves used outside of the ASME Code (1997). Equivalent fatigue quality or equivalent fatigue
Four published, non-ASME Code fatigue curves are life—The ASME Code (1997) S-N fatigue design
evaluated. Each has multiple fatigue curves for curve is appropriate for fatigue analysis of the weld
various weld types . In general, the multiple curves metal with the appropriate FSRF, at the same level
are parallel, indicating that they can be collapsed of confidence as the base metal.
into a single curve with an FSRF applied to the weld Fatigue quality level—Seven levels are defined;
type . FSRF are determined, based on the weld types. each is based on the extent of the NDE that is
Those that can be correlated with the weld types and performed . "Fatigue quality" relates to the actual
NDE requirements are compared with the quality cyclic life of the part for consistent loading condi-
levels . This approach shows that the proposed FSRF tions . The Code required NDE for Category A welds
are conservative. is expected to give the weld the same cyclic life as the
Since the ASME Code (1997) fatigue design curves, base metal under the same loading condition.
as well as other fatigue curves, are developed from Full volumetric—The RT or UT gives 100% cover-
design factors applied to test-data-failure curves, a age of the weldment with high quality definition of
comparison of design curves is made (Appendix II) . It indications.
shows that the two approaches are equivalent for Weld fatigue quality—The expected fatigue life
quality level 1 . Also, recent PVRC fatigue test data relative to the fatigue life of "fatigue quality level
for pipe welds are assessed . It shows that the #1".
proposed FSRF are conservative.
It is proposed that the ASME Code (1997) adopt Introduction
this approach for both pressure boundary and non- Over the last few years, there have been many
pressure boundary welds as a non-mandatory Appen- discussions at meetings of the ASME Code commit-
dix . The proposed values are probably more conserva- tees, the Pressure Vessel Research Council (PVRC),
tive than is necessary and may be overly conservative. and Pressure Vessel & Piping (PVP) Conference on
However, the guidelines are suitable as a non- the fatigue quality of welds . Some argue that welds
mandatory appendix and would be of great benefit are suspect as having a lower level of fatigue quality
for designers of pressure vessel equipment. than base metal . The common expectation of design-
ers is that welds have the same fatigue life as the
Nomenclature base metal ; thus, the fatigue quality of the weld and
base metal are the same . The suspicion of lower weld
Abbreviations quality is caused by the perception of welds being
AISC = American Institute of Steel Construc- more prone to flaws and notches . This can be shown
tion by failures occurring in weldments (including the
FEA = finite element analysis heat affected zone and its base metal interface) for a
28 WRC Bulletin 432

disproportionate amount of cases . A literature search deterministic basis to have a lower fatigue quality
performed by the ASME Code's Subgroup Fatigue than base metal . The equalizer is expected to be
Strength showed that the vast majority of fatigue NDE ; i .e ., the defects will be found by NDE and
failures occur at fillet welds, not full penetration removed.
welds . As a result, the majority of fatigue test
Goal
programs are directed at fillet welds . These factors
This report develops and defines the basis for
may lead to the conclusion that welds have lower
establishing FSRF for welds based on various combi-
fatigue quality than base metal, but this is not
nations of the common NDE procedures . The goal is
consistent with experience from vessels built to the
to have the ASME Code (1997) adopt this approach
requirements of the ASME Code (1997).
as non-mandatory guidelines for the designer.
Many articles have been written discussing weld
processes and procedures that are detrimental to Basis for the Approach
fatigue quality (Barsom and Vecchio, 1997 ; Jaske, Fatigue damage originates at notches and flaws
1998) . Also, Lundin (1976, 1981, 1988) developed an and NDE is intended to detect flaws and allow for
excellent initiation into fatigue of welds including a their removal . The different NDE techniques have
large bibliography. The practices that cause welds to different detection capabilities for different types
have defects are addressed by the ASME Code (1997) and locations of flaws . This leads to levels of quality
through weld processes and NDE . This report ad- as a function of which NDE techniques are applied.
dresses these issues by proposing conditions where Thus, all NDE techniques are useful, but some are
an FSRF should be applied for welds. more important than others in detecting the flaws
Fatigue damage is a function of stress at a point, that are significant to fatigue life.
cycles applied, and material properties . The stress is
Position
a function of the applied load and the local geometry.
With full volumetric and surface NDE, weld metal
When the stress analysis is performed, the model
has fatigue life consistent with the ASME Code,
normally will account for gross discontinuities and
Section III, NB (and NC-3200) and Section VIII,
blend radii, but not account for local notches and
Division 2 S-N fatigue curves . With appropriate
flaws . In the analysis, local notches and flaws are
FSRF for a reduced NDE, weld metal will still have a
addressed by the superposition of stress concentra-
fatigue life consistent with the S-N curves . In the
tion factors (SCF) or FSRF. However, identifying the
following discussions, this is stated as weld metal
existence of local notches and flaws and characteriz-
having fatigue life equivalent to that of base metal.
ing them is not practical . Only through a global
The development of a relationship between NDE
approach can their impact be quantified.
and FSRF for welds is intended to be a positive basis
It is generally expected that fatigue failures will
for applying the appropriate FSRF . It is not a
originate at notches and flaws . NDE can identify
recommendation to eliminate NDE in low stress
significant flaws, thus allowing them to be removed.
areas . However, it may be appropriate as a basis to
This is the role of NDE in the construction process.
only allow partial penetration and fillet welds to be
Specifically for welds, the role of NDE is to confirm used in low stress areas.
that the fabrication process has been correctly per-
formed and significant surface (planar) flaws have Foundation
been removed . This greatly reduces the probability Although it seems obvious, it is prudent to remind
of significant flaws. the reader that the guidelines being presented are
Considering the processes that produce plates, based on the vessel being built to ASME Code (1997)
forgings, and welds, it is reasonable to expect welds requirements . It may be applicable to other pressure
to have more flaws and notches than base metal. vessels and to internals, but its development was
Also, welds are more prone to having flaws and based on the high quality, Class 1 vessels where
notches, because the NDE applied to welds is less fatigue is a failure mode that must be analytically
consistent than the NDE applied to plates and included in the design process.
forgings . Welds may receive different NDE pro- In developing the guidelines, the focus was on low
cesses, depending on their location and its geometry. alloy steels and carbon steels . Again, this does not
For example, Category A, pressure boundary welds preclude applying the guidelines to other materials.
(ASME Code, 1997) receive full volumetric and sur- More ductile materials may have lower FSRF for the
face NDE, whereas welds for lugs may only receive a same NDE quality level. However, the accept-reject
surface NDE . The conclusion is that the potential criteria may also be different . The material and
magnitude and quantity of flaws and notches in criteria differences may cause changes in the specif-
welds are directly related to the NDE applied and ics of the guidelines, but the overall view of the
the accept-reject criteria . Thus, SCF and FSRF are guidelines should be appropriate for all ductile mate-
important factors in fatigue analyses of welds. rials.
Since welds, prior to NDE, are likely to have more It is inherent in the proposal that the ASME Code
flaws (per cubic inch) than base metal, welds are (1997) NDE acceptance criteria are used . The accep-
more likely on a probabilistic basis rather than tance criteria for NDE techniques can vary for
Fatigue Strength 29
different products . For example, all design codes and all significant flaws are expected to be detected.
may not use the same reject criteria for RT. These This defines a notch condition that is equivalent to
differences establish the overall quality of the prod- base metal with normal surface finish . From this it is
uct, i .e ., establish the probability of failure . This concluded that Category A welds, with their full
report does not address the accept-reject criteria ; it volumetric examination, should be considered as
is assumed that the ASME Code (1997) require- having the same fatigue quality as base metal.
ments are applied . In addition to material type and However, deleting some of the NDE required for
the accept/reject criteria, the foundations include: Category A welds may result in a significant flaw
remaining. The guidelines in this report were devel-
• High quality welds are developed ; this starts oped to address the condition of reduced NDE.
with the qualified welders, demonstrated weld
procedures, and testing requirements. FSRF Definition
• ASME Code rules for weld profiles. The following gives a brief definition for FSRF as
used in this report and a brief discussion of the
Post-Weld Heat Treatment (PWHT) relationship between FSRF and SCF. A more de-
The procedure is consistent with or without a tailed discussion is given in Appendix III . When a
PWHT. However, without a PWHT, the material may condition exists that is not included in the stress
not be consistent with the S-N curve test material analysis but will reduce the fatigue life, an FSRF can
due to a difference in ductility . The analyst must be used to determine the effective stress that is
consider this, but it is out of the scope of this report. consistent with the actual fatigue life . In practice,
The issue of higher residual stress is not relevant the FSRF is applied to the calculated alternating
because the ASME Code (1997) assumes the maxi- stress at a point prior to entering the S-N curve to
mum effect of mean stress. obtain the allowable cycles . Thus, the FSRF ac-
counts for a condition that is not included in the
Background for Initiating This Project
stress analysis model.
Fatigue quality, especially of welds, has been An SCF is defined as the total stress at a point
discussed extensively in meetings of the ASME (e .g ., notch location) divided by the nominal (mem-
Boiler and Pressure Vessel Code, the PVRC, and the brane-plus-bending) stress at the cross section . At
PVP conferences . A large percentage of attendees the endurance limit, the FSRF and the SCF may
argue that welds should have an FSRF applied in the have close to the same value . The closeness is a
fatigue analysis . However, this position has invari- function of the severity of the notch relative to the
ably been challenged by requesting example cases applied stress.
where failure resulted from an inadequacy in the FSRF accounts for the effect(s) on the fatigue life
ASME Code (1997) fatigue analysis procedure . Only due to local structural discontinuities (notches) that
Section III NB and NC-3200 and Section VIII Divi- are not included in the fatigue model . The FSRF is
sion 2 have specific fatigue procedures, based on an based on the change in cyclic life, but for design use,
analysis that predicts the stress at a point (not based the FSRF is restructured as a stress ratio . There is a
on nominal applied load) . No fatigue failures have stress level that will cause fatigue failure, at every
been identified and documented as being caused by specific cyclic life, for a component without any
an ASME Code inadequacy. Thus, the argument on notches . If a notch is present, a lower stress level will
which the conclusion is based is invalid. cause fatigue failure at the same cyclic life . The ratio
A counterargument is that designers may be apply- of these two stresses gives the FSRF. This definition
ing FSRF values even though the ASME Code (1997) is discussed in more detail in Appendix III.
does not specifically require it . Although this may The NB-3200 rules for fatigue analysis are applied
occur, it is not an across-the-board practice . As a to a point ; thus, a weld would be analyzed at both the
minimum, the Code is open to criticism for a lack of root surface and the final surface . Each of these two
guidelines for defining and applying FSRF. locations might receive different NDE . For example,
The relationship between fatigue quality and NDE if there is a non-removed backing strip at the root
is further demonstrated by the ASME Code (1997) pass, it could not receive a surface examination (only
design rules coupled to NDE and FSRF . An example a volumetric) . As a result, the root pass location
is the comparison of the ASME Code Category A and might require an FSRF = 2 .5 . On the other hand, the
B welds and their use of backing strips . Both require final pass might receive a full surface examination
a full radiographic test (RT) ; Category A excludes (e .g ., MT and VT) and require an FSRF of 1 .6 or less
non-removed backing strips ; Category B allows non- (the next section discusses FSRF values in detail).
removed backing strips but requires an FSRF to Thus, the FSRF is not applied to a weld; it is applied
account for the backing strip effect . Therefore, if the to a weld surface.
NDE is not capable of detecting the existence of a
significant notch or flaw, an FSRF should be in-
Defining FSRF Based on NDE
cluded in the fatigue analysis.
The Category A weld receives a full volumetric The development of FSRF for welds is a multi-step
examination and a surface examination (PT or MT), process . The NDE processes are defined and the weld
30 WRC Bulletin 432

types are categorized . Weld fatigue quality levels are NDE Quality Levels
established based on the applied NDE . This pro- The seven NDE quality levels, as presented in
duces a matrix of weld categories vs weld fatigue Table 1, are based on what the NDE is expected to
quality levels . An FSRF is then established for each detect and its relative impact on fatigue life . Earlier,
combination within the matrix . The process for these were referred to as "quality levels" and not as
establishing values is discussed . Validation of the categories . Quality level 1 includes full volumetric
weld-fatigue relationship is by comparison to avail- examination (RT or UT), and full surface examina-
able test data. tion including magnetic particle (MT) or liquid pen-
etrant (PT), and visual examination (VT with reject
The NDE Process criteria) (see "Definitions" for "full") . This is the
The common NDE processes that are considered general requirement for all pressure boundary
are radiography (RT), ultrasonic (UT), magnetic [ASME Code (1997) Category A] welds and merits an
particle (MT), dye penetrant (PT), and visual (VT). FSRF of 1 .0 . The maximum FSRF that needs to be
RT and UT are volumetric examinations ; MT, PT, applied is 4 ; this is for quality level 7, which has no
and VT are surface examinations . Although RT and NDE, e .g ., root of a fillet weld . Quality level 7
UT differ in their capabilities to detect various flaw requires an FSRF value of 3 to 4 ; this is based on
types (i .e ., slag, porosity, planar), their primary goal experience and a conservative envelope of test data
is to establish an overall fatigue quality of the weld. as discussed later. Tables 1 and 2 present FSRF in
Good weld processes, properly applied, will result in two formats : Table 1 is in a matrix format and Table
very few flaws . Thus, RT and UT ensure that the 2 is in a sequence format that may be easier to follow.
weld process is sound and is correctly applied.
On the other hand, the surface NDE focuses on Basis For Establishing FSRF Values
eliminating surface flaws that are equivalent to The following gives the fundamental logic in estab-
severe notches . The acceptance criteria must pre- lishing the FSRF values in Tables 1 and 2 . The bases
clude any notch that would have a significant impact are discussed in more detail in subsequent sections.
on the fatigue life ; i .e ., reduction in actual fatigue life Appendix I presents data from major test-program
relative to the analytically predicted fatigue life . The fatigue curves interpreted in terms of FSRF values
surface examinations are considered as more signifi- for various conditions . The appendix summarizes
cant than volumetric because fatigue cracks nor- the data and relates the data to the proposed FSRF
mally initiate at the surface (or near surface) rather values.
than at embedded flaws . (More discussion on this • A machined or ground, full penetration weld
item is given later.) that receives full volumetric examination (UT
or RT) and surface (MT or PT and VT) NDE plus
Weld Types appropriate PWHT should have equivalent fa-
Three basic weld types are considered : full penetra- tigue quality to base metal . If the base metal is
tion, partial penetration, and fillet. Each of the three left in an unmachined condition, the weld may
geometries has different limitations in the quality of have better fatigue properties than the base
NDE that can be performed . For example: metal . Therefore, these conditions preclude the
• Partial penetration welds present significant need for applying an FSRF. This quality level is
difficulty in reaching a high level of volumetric required for ASME Code (1997) Category A
NDE. welds.
• Fillet welds preclude a quality volumetric exami- • A weld that receives no NDE may have defects
nation. of any size . Pressure vessel experience indicates
• Both partial penetration and fillet welds have a that a FSRF = 4 is sufficient . Also, test data
(Appendix I .C) indicate that FSRF = 4 may be
root that precludes any significant NDE . Thus,
too conservative . The judgment is that the appro-
the three general weld categories have distinct
priate FSRF for non-inspected welds will be
differences in their relation to NDE.
between 3 and 4 ; this is discussed later in more
The weld types are evaluated for machined and detail (see Appendix V). Examples of locations
as-welded surfaces . The as-welded surface condition that cannot be inspected include the backside
must be suitable for the NDE that is being applied; (root) of partial penetration and fillet welds.
i .e ., it must be suitable for meaningful RT, UT, MT, or • Fatigue quality is more dependent on surface
PT, consistent with the planned quality level, even if NDE than on volumetric NDE, because fatigue
this requires some surface preparation. cracks will normally originate at the surface,
The partial penetration and fillet welds can be particularly at the weld-toe.
examined only on the final surface ; thus, these are • Reduced RT or UT occurs when a full volumetric
further divided as final or toe surface and root examination is not obtained ; full is defined as
locations . In the application of the FSRF, different 100% coverage with normal butt-weld quality.
values may be appropriate to the two locations . Reduced RT or UT is designated as quality level
Fatigue Strength 31
Table 1—Weld Surface Fatigue-Strength-Reduction Factors
Weld Condition Quality Levels*

(surface condition) 1 2 3 4 5 6 7
Full penetration
Machined 1 .0 1 .2–1 .5 1 .5 2 .0 2–2 .5 2 .5–3 .0 3 .0–4 .0
As-welded 1 .2 1 .3–1 .6 1 .7 2 .0 2–2 .5 2 .5–3 .0 3 .0–4 .0
Partial penetration
Final surface
Machined n .a . 1 .2–1 .5 1 .5 2 .0 2–2 .5 2 .5–3 .0 3 .0–4 .0
As-welded n .a . 1 .3–1 .6 1 .7 2 .0 2–2 .5 2 .5–3 .0 3 .0–4 .0
Root n .a . n .a . n .a . n .a. n .a. n .a . 3 .0–4 .0
Fillet
Toe–machined n .a . n .a . 1 .5 n .a. 2–2 .5 2 .5–3 .0 3 .0–4 .0
Toe–as-welded n .a. n .a . 1 .7 n .a. 2–2 .5 2 .5–3 .0 3 .0–4 .0
Root n .a. n .a . n .a . n .a. n.a . n.a . 3 .0–4 .0
n .a . = not applicable (inspection cannot be performed).

*Quality levels (see Table 2 for definition):
1, RT or UT (volumetric), MT or PT (surface), and VT (surface);
2, RTm or UT m and MT/PT plus VT, where RTm and UT m are partial volumetric;
3, MT/PT plus VT with or without progressive MT/PT;
4, RTm or UT m and VT with or without MT-root pass;
5, VT (visual) only ; 6, RT, UT, RTm , or UTm only; 7, none.
Subscript m, modified, i .e ., volumetric examination is not 100%.
Acceptance criteria for visual are addressed later.
Table 2—Criteria for Weld FSRF impact on fatigue quality is greater than that of
Quality the RT or UT.
FSRF Level Definition • As-welded surfaces can vary in quality from
excellent to poor . Therefore, it is appropriate to
1 .0 1 Machined or ground weld that receives a
full volumetric, and a surface that apply an FSRF to as-welded surfaces . The as-
receives MT/PT and a VT (visual). welded factor of 1 .2 is based on test data dis-
1 .0–1 .2 1 As-welded weld that receives a full volu- cussed in Appendix I.
metric, and a surface that receives
MT/PT and VT. • At the quality level where MT/PT is not per-
1 .2–1 .5 2 Machined or ground weld that receives a formed, surface flaws comparable to as-welded
partial volumetric, and a surface that conditions can exist . Thus, the machined and
receives MT/PT and VT.
1 .3–1 .6 2 As-welded weld that receives a partial as-welded conditions become equivalent (see
volumetric, and a surface that receives Table 1).
MT/PT and VT. • Volumetric examination is relevant to fatigue
1 .5 3 Machined or ground weld surface that
receives MT/PT and VT, but the weld quality even in the absence of surface examina-
receives no volumetric. tion . Its impact on fatigue life plays the same
1 .7 3 As-welded weld surface that receives role as the volumetric examination applied to
MT/PT and VT, but the weld receives no
volumetric. base metal, i .e., it demonstrates that the fabrica-
2 .0 4 Weld has received a partial or full volu- tion/weld process has been controlled . The FSRF
metric and the surface has received VT, should be less than that of no inspection but
but no MT/PT.
2 .0–2 .5 5 VT only of the surface ; no volumetric nor greater than VT-only.
MT/PT examination.
2 .5–3 .0 6 Volumetric only. The definition of quality levels presented in Table
3 .0–4 .0 7 Weld backsides that are non-definable 2 establishes the FSRF based on these relationships.
and/or receive no inspection .
Discussion of NDE Techniques
2 and establishes the concept of and UT,, Table 1 gives brief definitions to describe the
(more discussion later). quality levels ; these levels are given in an expanded
• All surfaces that receive MT or PT are expected definition in Table 2 along with the appropriate
to receive a visual examination (VT) ; the reverse FSRF for the level . In addition, some quality levels
is not necessarily true . Except for roots of par- have potential variations in the FSRF, and it can be
tial penetration and fillet welds, all weld sur- difficult to choose the appropriate FSRF value . For
faces are expected to have a VT. the final selection, the design engineer should con-
• A VT precludes severe surface notches, but it firm the desired quality level and review this with
does not preclude the existence of tight planar NDE and welding engineers . It is also appropriate to
indications . This is considered to be the level of discuss the NDE techniques when the quality level
examination used in the Harrison-Maddox test has a range for the FSRF. The following discusses
data evaluation (Appendix I .C), which shows the NDE techniques, their relationship to choosing a
that the data can be encompassed by an FSRF of quality level, and the reason for variations where
2 .0 to 2 .5 . Since VT is a surface examination, its they exist.
32 WRC Bulletin 432

RT11 and UT, The ASME Code (1997) depends surface still may not have the quality of machined
on RT as the volumetric method for Categories A and and ground surfaces due to roughness . Therefore, it
B, but other weld geometries preclude a full RT. is appropriate to apply an FSRF to as-welded sur-
Since the ASME Code depends primarily on RT, the faces even if some grinding is performed.
initial assessment is to determine whether the RT Quantifying the FSRF for as-welded surfaces de-
gives a full volumetric examination for the specific pends on understanding the relationship between
weld . If not, the value of a superposed UT can be as-welded and base metal surface finish as related to
determined . The result can be that neither RT or UT the S-N curve design factor (2 & 20) . The test data on
nor a combined RT/UT will give a full volumetric which the 1 .2 factor is based do not address this but
examination . If this occurs, the RT/UT should be rather compare a ground surface to the as-welded
applied to the appropriate extent along with the surface . In general, as-produced, base metal surface
appropriate FSRF in the analysis. finish does not have the quality of a ground surface.
Quality level 2 uses the term RT m or UTm ; Table 2 The ASME Code (1997) allows for this roughness in
indicates that these are partial volumetric examina- the design factor ; thus, it is already built into the
tions . These generally occur when welds are placed S-N design curve . Therefore, the FSRF = 1 .2 is
at discontinuities, what the ASME Code (1997) calls expected to be conservative and could be appropriate
Category C and D welds . They may also apply to a for some non-pressure boundary cases where signifi-
double wall RT, e .g ., piping welds . The pressure cant roughness is accepted.
boundary shells' long seam and circle seam welds Surface vs Volumetric Examination . Most field
(without backing strips) will normally receive a full failures, as reported in the literature, identify a
volumetric examination . On the other hand, a nozzle surface notch (flaw) as the initiation site . The cyclic
insert weld is often at a discontinuity and may loading sharpens the notch into a fatigue crack and
require special techniques to get a full volumetric this can usually be identified and reported . The
NDE . The designer should review the RT/UT plan defect then undergoes crack growth, prior to leakage
with NDE personnel to determine if the modified or fracture ; this can also be identified . This scenario
inspection technique will produce a full volumetric is typical and the initial step in failure assessment.
examination . The correct selection of quality level Since surface examinations directly address surface
can then be made. flaws, fatigue quality is more dependent on surface
Quality level 2 shows a span on the FSRF (1 .2 to NDE than volumetric NDE.
1 .5 for machined/ground condition) . Based on the Two additional factors support the expectation
assessment with the welding and NDE engineers, and fact that most failures originate on the surface.
the designer can determine the surface condition They are as follows : (1) the membrane-plus-bending
and establish an FSRF for use in the fatigue analysis stresses and the thermal stresses are normally higher
of the weld . The RT/UT procedure on which the on the surface than at embedded locations and (2)
FSRF is based becomes locked in ; i .e ., it cannot be the free surface has less constraint than an embed-
changed without the knowledge and concurrence of ded location . Low constraint allows the shear plane
design engineering. to be worked, distorted, and separations to be formed.
Ground vs As-Welded. The purpose of grinding If embedded flaws have sufficient size, they can
the surface is to remove notches, including weld-toe respond to the load with a constraint similar to the
notches, but not necessarily to remove the reinforce- free surface . Therefore, fatigue initiation is more
ment . Some grinding of the as-welded surface may likely to occur on the surface, but it can occur from
be necessary to obtain meaningful NDE . Just as large size embedded flaws.
important as obtaining smoothness on the surface is RT and UT can detect many surface flaws and
blending the weld into the base metal . This is a key notches. However, surface NDE techniques (MT, PT,
issue when the weld is part of a discontinuity blend and VT) are better suited to finding surface flaws
radius (see Appendix V for more details). Ground than the ASME Code (1997) RT or UT procedures.
surfaces are expected to be as smooth as (or smoother Also, MT/PT normally have more restrictive reject
than) machined plates or forgings and to be flush criteria, especially for planar flaws, than RT/UT.
with the adjacent surface. As-welded conditions can Most flaws are embedded, so volumetric NDE is
be rough, but the roughness is a function of the important in demonstrating the general quality of
applied NDE . When the required NDE reaches qual- the weld, in the same manner that volumetric exam-
ity level 4, the potential for surface notches replaces inations are used to demonstrate the quality of
the FSRF for (as-welded) roughness. forgings . Therefore, volumetric examinations estab-
Depending on the weld process and the welder's lish material quality, whereas surface examinations
skill, an as-welded surface can vary in quality from detect the truly significant flaws . Flaws that are
excellent to poor. When the as-welded surface is near-surface (distance is a function of grain size)
adequate for RT, UT, or MT/PT, it has a higher relate more to surface notches than to embedded
fatigue quality than the normal as-welded condition. flaws, relative to quality level . When using the MT
This is because significant notches that cause a procedure, near-surface flaws will be detected ; the
reject condition are removed . However, an as-welded depth is a function of the MT procedure . PT only
Fatigue Strength 33
addresses the surface . In summary, surface flaws mock-up for all welds . However, these are part of the
and notches are the most significant cause for fa- justification for the VT acceptance criteria . The
tigue damage and direct surface examinations mock-up need not be on the actual component being
(MT/PT and VT) are the best method for detection. considered, but it must be consistent with it . The
However, volumetric examinations are necessary for mock-ups can be evaluated by finite element analy-
assuring quality material, including quality weld ses (FEA) to define SCF, recognizing that SCFs are
deposit. conservative relative to FSRFs (Snow et al ., 1970).
Visual Inspection. A visual inspection (VT) The mock-up should be consistent with an SCF less
should be performed to the maximum extent pos- than 4 determined by FEA; if not, the surface may
sible . The FSRF for the "visual (VT) only" quality not meet a reasonable workmanship requirement.
level is 2 .0 to 2 .5 with the appropriate value depen- Therefore, the mock-ups and FEA are appropriate
dent on inspection quality and the required accep- for new weld conditions, and past data can be
tance criteria . Guidelines for acceptance criteria maintained for future assessments.
should be considered for ASME Code (1997) incorpo- The ASME Code NB-4424(c) (1997) permits 1/32"
ration rather than having it as the responsibility of undercuts (30 mils/0 .75 mm), but does not allow
the designer. The focus of this inspection and its cracks in either base metal or weld metal . Thus, this
acceptance criteria are to preclude severe notches, undercut is covered by the S-N curve design factor (2
which are the primary consideration in establishing & 20) . Also, it is a reject limit for VT and should be
FSRF. included in the defined acceptance criteria.
With today's tools, virtually every weld can receive In summary, a VT precludes severe surface notches,
a visual inspection except the backside of some but it does not preclude the existence of a crack-like
partial penetration welds, fillet welds, and where indication . Mock-ups are a good tool in defining
permanent backing strips are used . The quality of workmanship and in assuring that the necessary
the VT can vary substantially. For example, the final quality will be met . Engineering must design for or
surface of a typical pressure boundary weld (MT/PT concur with manufacturing on the workmanship
applied) is easy to (VT) inspect and to ensure that quality of the weld surface for consistency with the
the acceptance criteria have been met. At the other FSRF used in analysis . Together, design engineering
extreme, the backside of a thermal sleeve to nozzle and manufacturing must agree on acceptance crite-
weld may require a video camera to extend a few ria that is consistent with the design analysis and
inches into a 3/16" annulus to evaluate the acceptance doable.
criteria . However, for the vast majority of conditions, Backside of Partial and Fillet Welds. The
the VT should detect significant notches. backside (root) of partial penetration and fillet welds
The FSRF = 2 .5 relates to a significant notch as cannot be adequately inspected ; therefore, the root
the acceptance criteria. For difficult inspection condi- requires an FSRF of 3 to 4 . The final surface of these
tions, the procedure requires a workmanship sam- welds can receive a surface examination ; thus, the
ple . The design analysis alternatives are as follows: FSRF for the final surface should be less than 3 . The
(1) to assume no NDE (VT or RT/UT) and use an following discussion is a summary of this topic ; more
FSRF = 3 to 4 or (2) to justify a value between 2 .0 details are given in Appendix V.
and 2 .5 . A high quality workmanship sample and The root of a partial penetration J-groove (insert)
acceptance criteria based on this sample permit an welds is not expected to fuse through-thickness,
FSRF = 2 .0 provided the inspectors have adequate producing a notch at the root of unknown severity.
tools for making judgments, e .g ., clear access and When fillet welds are applied in the same manner,
5X VT. they would have the same type of notch at the root.
In addition to assessing notches, the VT should Since none of the standard NDE procedures can
address the blend (contour) from the weld to the base detect these notches, an FSRF = 4 is reasonable . To
metal . The key geometries are when the weld in- justify a reduced FSRF would require a weld-process
cludes a blend radius or there is a weld reinforce- development program. An FEA will often model this
ment . This always applies to fillet welds . The smooth- condition as a singularity that produces a large SCF
ness of the transition is the consideration. directly related to the FE grid refinement . The weld
Weld mock-ups can be used to define a workman- FSRF is applied to the local membrane-plus-bending
ship sample for a backside without direct access . The stress.
mock-up must be consistent with the geometric The root of a partial penetration attachment weld
condition, including the worst case fit-up that will be (e .g., attachment of a bracket to a shell) produces a
allowed in production . Thus, the workmanship sam- small flaw that is probably rounded and is parallel
ple ensures that the as-built backside should meet with the stress flow. Even though it receives no NDE,
the FSRF requirements of the analysis . Review of an FSRF = 3 is probably sufficient . Use of fillet welds
the backside of the mock-up gives the inspector an for the same application produces a much larger flaw
indication of what to expect ; anything less should be than the partial penetration weld . This large flaw
rejected. can be the full width of the bracket (being attached)
It is not the intent of this process to require a and may concentrate the parallel shell stresses . An
34 WRC Bulletin 432

FSRF of 4 for this condition (weld notch and geomet- do not include weld buildups in establishing
ric SCF) is reasonable . For bracket loads, the flaws FSRF per quality level.
at the root will concentrate the stress, especially for • The scope of the project does not include clad-
membrane loads . However, the flaws are well re- ding, but it is reasonable that cladding does not
moved from the surface . Thus, the concentration require an FSRF. Cladding may receive a UT for
may be significant, but the resulting FSRF should be (base metal) bond and/or defect and a final
less than 4. surface PT and VT. Recognizing that the ASME
In addition to weld-process development, a test Code (1997) does not require UT for bond, the
program may show that the FSRF = 3 or 4 is too designer is responsible for considering this in
conservative . S-N data from testing the full weld- establishing the fatigue analysis procedure.
ment would give a more accurate FSRF than the • The recommended weld FSRF does not include
proposed values . The testing approach is appropri- stress concentration due to design notches, e .g .,
ate but can become expensive . If the partial penetra- discontinuity blend radii which include the weld.
tion and fillet welds are used only in low stress areas, Thus, the notches caused by design discontinui-
the FSRF of 3 or 4 should not predict a fatigue ties produce an SCF that is applied prior to the
problem. weld FSRF. A combined SCF and FSRF = 4
seems reasonable as an upper limit . (Additional
Conditions Peripheral to Quality Levels discussion is given in Appendices IV and V.)
• The valley between weld passes is expected to
The development of the proposed guidelines in-
have a contour appropriate for RT/UT and/or
cluded major input from welding and NDE experts.
MT/PT even if none is performed . Otherwise,
Within these discussions, items arose that did not
the FSRF (quality levels 4, 5, and 6) may not be
neatly fit into the guidelines . However, some of these
applicable.
items are significant and may be important to the
• Applying the FSRF to through-thickness ther-
designer. The following are the items that were
mal peak stresses is expected to be overly
judged to require attention.
conservative . This is discussed in Appendix IV
• Although backing strips produce a back surface but is not addressed in the proposal.
that would be in quality level 6 (RTm/UTm only), • In a number of places, the report mentions
Code experience indicates that the FSRF be- "weld process development ." This is an impor-
tween 2 .0 and 2 .5 is conservative . Thus, backing tant concept for the designer to understand and
strip conditions may use an FSRF of 2 .5 . The address . The weld process must receive a proce-
backing strip condition is assumed to have an dure qualification (PQ) . In general, the PQ
RT m . Thus, level 6 value of 2 .5–3 .0 is consistent assures weld strength, ductility, and overall
with the 2 .5 for backing strips . This is not quality. However, the PQ is not directly related
intended to preclude the use of FSRF = 2 if it to fatigue quality. For high quality welds, mock-
can be justified . If the RTm is not performed, the ups and destructive examination are needed to
FSRF of 2 .5 may not be appropriate and FSRF = demonstrate that the process can consistently
3 .0 should be considered. produce defect-free welds . The mock-ups should
• Progressive MT may have benefit for fatigue include worst case conditions, such as maxi-
quality when no volumetric examination has mum worst-case fit-up . The weld quality im-
been performed . However, no benefit has been pacts the type, quantity, and distribution of
assumed because no data were available to flaws . Detailed discussion of the weld process
quantify the benefit. development is outside the scope of this project.
• Root-pass MT has benefit for fatigue quality • In general, the NDE for establishing the pro-
when surface MT is not performed . The benefit posed FSRF are the standard approaches re-
exists even when a full volumetric examination quired by the ASME Code (1997) . There are a
is performed . However, its benefit cannot be number of advanced techniques that have been
quantified at this time, so no benefit has been well developed, even some that are not specifi-
defined. cally addressed in Section V of the ASME Code.
• Buttering of a weld prep is expected to receive a Advanced techniques are not considered in this
preliminary NDE as well as the NDE applied to project, but their use should allow the design
the final weld . The quality of the weld prep engineer to reduce the FSRF in the proposed
should be defined by the NDE applied to the guidelines.
final weld.
• Full volumetric examination of a weld buildup Acknowledgements
must be defined independent of full volumetric Development Team
for through-thickness welds . When a weld The authors thank the PVRC for their support of
buildup receives the NDE of RTm or UTm with a this project . Also, a special thanks to Harold Graber,
surface MT/PT and VT, an FSRF = 1 .0 is Jeff Kikel, Mike Snyder, Bob Yeager, and Eric Weber
reasonable . However, the proposed guidelines who were on the team that developed the initial
Fatigue Strength 35
quality levels and established the FSRF. The proce- addition, the sources are referenced and available
dure has been reviewed by many people within the for more detailed study as needed for any external
Babcock and Wilcox Co . and also by members of the acceptance of the proposed procedure . Thus, the
ASME, Boiler and Pressure Vessel Code, Subgroup appendix consists of the introduction with its summa-
on Design Analysis . These efforts were beneficial and ries and the detailed reviews of the four sources.
are appreciated. The following are the sources that were used to
assess the proposed procedure for setting FSRF
PVRC Reviewers values for welds:
A large number of people from PVRC have re-
viewed and commented on the report . The major A. WRC Bulletin 398 : Reduction of S-N Curves for
effort was by the special review team, consisting of Ship Structural Details, 1995 Based on data
Tom Ahl (CB&I), Rich Basile (Exxon), Jerry Bitner from University of Illinois (Stambaugh)
(Consultant), Woody Caldwell (Eastman Chemical), B. The (British) Welding Institute : Research Bul-
Owen Hedden (Consultant), and Robert Vecchio (Lu- letin Vol 17, No 5, May 1976 Fatigue Design
cius Pitkin) . This team made many suggestions that Rules for Welded Steel Joints (Gurney)
greatly raised the quality of the report. C. The (British) Welding Institute : Derivation of
We extend a special thanks to Sumio Yukawa Design Rules for Pressure Vessels (Harrison
(Consultant) for the final review and critique and for and Maddox, circa 1980)
upgrading the PVRC report into a WRC Bulletin. D. Structural Welding Code ANSI/AWS D1 .1-92,
1992
References In general, each of the sources includes fatigue
ASME, Criteria of the ASME Boiler and Pressure Vessel Code for Design
by Analysis in Sections III and VIII, Division 2, American Society of curves for welds that are based on full scale test data
Mechanical Engineers, 1969. using various weld-geometries . The fatigue curves
ASME, Boiler and Pressure Vessel Code, Section III, Rules for Construc-
tion of Nuclear Power Plant Components, NB, Class 1 Components and consist of a set of parallel fatigue curves that relate
Section VIII, Rules for Construction of Pressure Vessels, Division 2–Alter-
nate Rules, American Society of Mechanical Engineers, 1997. to various weld configurations (called categories).
Barsom, J. M. and Vecchio, R ., "Fatigue of Welded Components," WRC
Bulletin, 422, June 1997. Their parallel-curve format allows FSRF to be readily
Gurney, T. R ., "Fatigue Design Rules for Welded Steel Joints," The calculated and based on geometry and loading . They
(British) Welding Institute, Research Bulletin Vol 17, Number 5, May 1976.
Hardraft, H . F. and Kuhn, P., An Engineering Method for Estimating all have at least one curve for base metal or ma-
Notch-Size Effect in Fatigue Tests on Steel, NACA (now NASA), Tech Note
2805, 1952. chined butt-welds . In the detailed review, FSRF
Harrison, J. D . and Maddox, S . J ., "Derivation of Design Rules for values are developed by normalizing the individual
Pressure Vessels," The (British) Welding Institute, circa 1980.
Heald, J . D . and Kiss, E ., "Low Cycle Fatigue of Nuclear Pipe Compo- curves by the base metal curve or by the full penetra-
nents," Journal of Pressure Vessel Technology, 96, August 1974, pp . 171–
178. tion, machined butt-weld curve . Therefore, the infor-
Jaske, C ., "Interpretative Review of Weld Fatigue-Strength-Reduction
and Stress Concentration Factors," WRC Bulletin, 432, 1998. mation from these four sources is the FSRF that the
Lundin, C . D ., "The Significance of Weld Discontinuities—A Review of sources are applying to various weld configurations.
Current Literature," WRC Bulletin, 222, December 1976.
Lundin, C . D ., "Review of Worldwide Weld Discontinuity Acceptance These FSRF are then related to the proposed FSRF-
Criteria," WRC Bulletin, 268, June 1981.
Lundin, C . D ., "Bibliography on Fatigue of Weldments and Literature for-welds procedure.
Review on Fatigue Crack Initiation From Weld Discontinuities," WRC
Bulletin, 333, May 1988. Summary of Each of the Four Sources
Neuber, H ., Theoretical Determination of Fatigue Strength at Stress
Concentration, Report AFML-TR-68-20, Air Force Materials Lab, Wright- A. WRC Bulletin 398. NDE for these welds is
Patterson Air Force Base, Dayton, Ohio, 1968.
O' Donnell, W. J . and Purdy, C . M ., "The Fatigue Strength of Members presented in a ship-building code . The maximum
Containing Cracks," Trans. ASME, 86, May 1964. FSRF is 3 .0 ; thus, the worst fillet weld category has
Petershagen, H ., "The Influence of Undercut on the Fatigue Strength of
Welds—A Literature Survey," Welding in the World, Vol . 28, 1990. an FSRF of 3 .0 . For configurations more relevant to
Peterson, R. E ., Stress Concentration Factors, John Wiley & Sons, New
York, 1974. Code pressure vessels, the maximum fillet weld
Stambaugh, K. A ., Lesson, D . H., Lawrence, F. V., Hou, C . Y., and Banas, FSRF is 2 .2 which is less than the recommended
G ., "Reduction of S-N Curves for Ship Structural Details," WRC Bulletin,
398, January 1995. value of 3 to 4 from Table 1 . Two of the categories
Scavuzzo, R. J ., Srivatsan, T. S . and Lam, P. C ., " Fatigue of Butt-Welded
Pipe," PVRC Grants #91-11, 92-12, 93-21, and 94-14, Welding Research have the same weld except one is ground and one is
Council, New York, February 1997.
Snow, A. L., Langer, B . F. and Gibbons, W. G ., " Low-Cycle Fatigue as-welded . The FSRF is 1 .2 for as-welded ; thus, it is
Strength of Notched Ni-Cr-Fe Alloy 600 and Low Alloy Steel," Journal of the same as recommended in Table 1.
Materials, 5, 1970, pp . 719–737.
SWC, "ANSI/AWS D1.1-92," Structural Welding Code, American Welding B. The (British) Welding Institute–Gurney.
Society, Miami, 1992.
Gurney 's work (1976) addresses backing strips . The
backing strip configuration, compared with the full
Appendix I butt-weld with transverse load, gives an FSRF = 1 .8.
Data From Other Fatigue Curves
Therefore, the ASME Code (1997) experience-based
Used to Validate Weld FSRF values of FSRF = 2 .0/2 .5 are conservative . The
Introduction proposed FSRF of 2 .5 (see Conditions Peripheral to
Four sources of information and data are dis- Quality Levels) is consistent with level 6 (RT/UT
cussed in this appendix . This introduction includes a only) Table 1 values of FSRF = 2 .5–3 .0 . Considering
summary of each of these sources, followed by a the Gurney work and the ASME Code experience, an
summary of the four sources relative to the FSRF-for- FSRF = 2 .0 is probably a more appropriate value.
welds procedure . The appendix then gives a more Gurney's maximum FSRF based on his as-rolled
detailed review of each of the four sources . In plate data is less than 3.
36 WRC Bulletin 432

C. The (British) Welding Institute-Harrison shows that the high values (FSRF from 2 to 4) are
and Maddox. Harrison and Maddox presented test justified by the data . Also, FSRF = 1 for full penetra-
data (1980) for full size pressure vessels components. tion and ground welds is justified along with the 1 .2
There is no indication of what NDE was applied to value for as-welded . The cases with FSRF between 1
the welds . In general, most of the construction does and 2 have not been verified, but they certainly
not lend itself to quality NDE . When the data are would be expected to have an FSRF between 1 and 2.
compared with the ASME Code (1997) design curve The FSRFs from the four sets of fatigue curves are
using an FSRF = 2, all of the test data points are obtained from the stresses at 1E6 cycles . Except for
above the adjusted ASME Code design curve . Thus, the TWI (The Welding Institute) curves (B), all the
one can assume that the maximum FSRF that is curves are parallel . Thus, the FSRF could be ob-
necessary regardless of NDE is 2 . Recognizing that tained at any cycle level . The TWI curve reduces the
the test specimens and procedure may not fully FSRF at cycles less than 1E6 ; this matches testing
represent production conditions, one might opt for for FSRF data . Since the sources do not interpret
an upper bound FSRF = 2.5 to 3 .0. their data for reduced FSRF below 1E6 cycles, other
Although these data are the thrust of the British data are needed for this adjustment . The parallel
report, they are not the basis for the Harrison and curves are based on two standard deviations below
Maddox (1980) recommended S-N curve . They pro- the mean.
posed three classifications ; their FSRF values are
almost negligible compared with the test data ap-
Detailed Review of the Four Sources
proach and they show base metal and flush welds
with MT/PT as equivalent. A. WRC Bulletin 398—Reduction of S-N Curves for
Ship Structural Details
D. Structural Welding Code (SWC). In this Based on Data from University of Illinois-W. H.
approach, butt-welds require volumetric NDE . The Munse
FSRF for the best weld is 1 .3, which receives a The bulletin categorizes welds, A through G, based
volumetric but no MT/PT or visual . It compares with on quality obtained from test data . Categories A
the proposed quality level of 6 with an FSRF of 2 .5 to through E include the conditions described below ; F
3 .0 . Thus, the recommended approach is conserva- and G are not considered relevant to pressure ves-
tive relative to the SWC. The full butt-weld has an sels .
FSRF of 1 .4 for the non-ground vs ground, compared
with the recommended Table 1 value of 1 .2 . Recom- A. Base metal
mending the lower value is consistent with Gurney's B. Longitudinal and transverse butt joints, weld
recommendation, based on available data . The over- ground, axial load
all FSRF for fillet welds of 2 .3 to 3 indicates that the Transverse butt, as-welded, bending
proposed 3 to 4 is conservative. Lateral attachment, fillet, axial load
C. Welded I-beam, fillet or bevel with fillet, bend-
Summary of Information From the Four Sources ing
For general fillet welds that might be used in Partial penetration (double bevel) butt-weld,
pressure vessels, the four references give FSRF ground, axial load
values from 2 to 3 . Thus, the recommended 3 to 4 is D. Lateral attachment to plate, fillet, resisting
conservative . It follows from this that quality level 6, bending
with a recommended FSRF of 2 .5 to 3 is also Cruciform joint, double bevel with fillet, axial
conservative . This along with other findings from the load
references are presented in Table I-1, which repeats Cruciform joint, fillet, bending
Table 1 but highlights (bolds) the verified areas and Transverse butt-welded I-beam, double bevel,
underlines the areas considered justified. The table as-welded, bending
Table I-1—FSRF Conservative by Comparison to Other Sources
Weld Category Quality Levels

(surface condition) 1 2 3 4 5 6 7
Full penetration
Machined 1 .0 1 .2-1 .5 1 .5 2 .0 2-2 .5 2 .5-3 .0 3 .0-4.0
As-welded 1 .2 1 .3-1 .6 1.7 2 .0 2-2 .5 2.5-3 .0 3.0-4.0
Partial penetration
Final surface
Machined n .a. 1 .2-1 .5 1 .5 2 .0 2-2 .5 2 .5-3 .0 3 .0-4.0
As-welded n .a . 1 .3-1 .6 1 .7 2 .0 2-2 .5 2 .5-3 .0 3 .0-4.0
Root n .a . n .a . n.a . n .a . n .a . n .a . 3 .0-4.0
Fillet
Toe n .a . n .a . 1 .7 n .a . 2-2 .5 2 .5-3 .0 3 .0-4.0
Root n .a. n .a . n.a . n .a . n.a. n .a . 3 .0-4 .0
n .a . = not applicable .
Fatigue Strength 37
Welded beam with intermittent welds, bending Table A .1-FSRF
Partial penetration (double bevel) butt-weld, FSRF to:
WRC 1E6 Q
as-welded, axial load Category Description (ksi) Plate Weld
E . Attachments—fillet welds
A Plate 24
NDE for these welds is presented in a ship-building B Full butt 19 1 .26
C Partial 16 1 .5 1 .2
code. D As-welded 13 1 .8 1 .4
The S-N curves are plotted in Fig . A .1 ; it shows E Fillets 11 2 .2 1 .7
parallel lines for each WRC category. How the cat- F 9 .5 2 .5 2 .0
G 8 3 .0 2 .4
egory A curve compares with the ASME Code is not
considered important because the material and the
design factors are different . The benefit is to define etries, the fillet weld FSRF is 2 .2 . This implies that,
FSRF based on the welds . Since the S-N curves are for pressure vessels, FSRF = 2 .2 is appropriate
parallel, FSRF can be readily calculated. provided NDE is applied.
Table A .1 shows the FSRF calculations based on Two of the categories have the same weld except
category A (base metal) and on Category B (full one is ground and one is as-welded:
penetration ground butt-weld) . First, the stress range
C. Partial penetration (double bevel) butt-weld,
at 1E6 cycles is obtained ; these are provided in the
ground, axial load
first column (1E6 Q) . Next, the stresses are normal-
D. Partial penetration (double bevel) butt-weld,
ized, first to the Category A stress range (column 2,
as-welded, axial load
FSRF to Plate), then to the category B stress range
(column 3, FSRF to Weld) . The normalized values The FSRF are 1 .5—1 .8 and 1 .2—1 .4 depending on
are the FSRF. whether they are normalized by WRC Category A or
The maximum FSRF in Table A.1 is 3 .0 . This B, respectively. They both give a factor of 1 .2 in going
indicates that the worst fillet weld has an FSRF of from a ground weld to an as-welded condition . This is
3 .0 based on base metal . For more relevant geom- relevant to pressure vessel welds.
Number of Stress Cycles N

Fig . A .1-S-N curves from WRC Bulletin 398.
38 WRC Bulletin 432

B . The (British) Welding Institute (TWI)-Gurney, TWI Category B (as-rolled plate) and on TWI Cat-
Research Bulletin Volume 17, Number 5, May 1976: egory C (full butt-weld, transverse load) . First, the
Fatigue Design Rules For Welded Steel Joints
stress range at 1E6 cycles is obtained ; these are
The paper generates S-N design curves based on
provided in the first column (1E6 cr or r at 1E6
test data, primarily generated at TWI . Gurney (TWI)
cycles) . Next, the stresses are normalized, first to the
categorizes the curves A through G plus F 2 and W,
TWI Category B stress range (column 2, FSRF to
based on the test data . Table B .1 gives the major
considerations per category. In the overview, Gurney Plate), then to the TWI Category C stress range
first establishes six types of weld conditions ; mul- (column 3, FSRF to Weld) . The normalized values
tiple TWI categories exist in each type . The types of are the FSRF.
weld conditions are as follows: The key categories are C, F, and G . Category B
would be significant if a forging OD was equivalent
1. Material free from welding to an as-rolled plate . In going from as-rolled plate to
2. Continuous welds with applied stress parallel a full butt-weld, the FSRF is 1 .2 compared with the
to weld proposal of the weld being as good as the base metal
3. Transverse butt-welds if adequate NDE is performed.
4. Welded attachments The backing strip (TWI category F) compared with
5. Load-carrying fillet and T butt-welds the full butt-weld with transverse load (TWI Cat-
6. Details in welded girders egory C) gives FSRF = 1 .8 ; therefore, the 2 .5 pro-
The S-N curves are plotted in Fig . B .1 (the curves posed value is conservative.
represent mean minus 2 standard deviation values); As obtained through WRC Bulletin 398, all FSRF
it shows parallel lines except for TWI Categories B are less than 3 .0.
and C . Thus, the benefit of D through F is to define
FSRF based on the welds . Table B .2 shows the FSRF C . The (British) Welding Institute-Harrison
and Maddox: Derivation of Design Rules
calculations based on category B (full penetration for Pressure Vessels
ground butt-weld), using the stress range at 1E6. As stated in the title, Harrison and Maddox (1980)
Table B .2 shows the FSRF calculations based on
addressed pressure vessels, but followed the same
type of categories as Gurney . They first assembled a
data base of full size tests, primarily from European
Table B .1-Major Considerations per TWI Category sources . These can be assembled in five types of weld
Category Consideration conditions:
A Base metal, machined and polished 1. Base metal

B Plate in the as-rolled condition 2. Main seam
Full penetration butt welds, dressed flush with sur- 3. Toe cracks
face, finish machined, NDE
C Plate in rolled condition but flame cut edges by con- 4. Root cracks
trolled procedure, free of cracks 5. One side
Butt or fillet welds with welds made by an auto-
matic submerged or open arc process and with no There is no indication of what NDE was applied to
stop-start positions within the length, load par-
allel with weld the welds . Lack of NDE could be the cause of the
Butt welds ground flush, NDE, load perpendicular scatter ; other conditions could contribute . With the
to weld total lack of any indication that NDE was performed,
D Butt or fillet welds with welds made by an auto- it is reasonable to assess the results on a non-NDE
matic submerged or open arc process, allows stop-
start positions within the length, load parallel basis.
with weld Weld conditions 3 and 4 cover a broad spectrum as
Full penetration butt welds, flush, made in fiat posi-
tion, either manual or automatic process other shown in Fig. C .1 . Only sketch 5 can be considered
than submerged arc for Class 1 construction . Not even sketch 5 can be full
E Full penetration butt weld, flush using process not volumetric inspected . Weld condition 5 (one side) has
defined in C or D
F Full penetration butt welds with backing strip an as-welded backside with no surface or visual
Attachment fillet welds perpendicular to load, inspection . Harrison and Maddox consider this to be
length <6" (150 mm) and edge distance >0 .4" (10 the worst condition.
mm)
Cruciform or T -joint full penetration welds with Harrison and Maddox plot these data against the
corner undercutting ground out ASME Code (1997) design curve to show that the
F2 Full penetration butt welds between plates of Code is non-conservative . However, this study gives
unequal widths
Attachment fillet welds perpendicular to load no basis for accepting the test specimen and results
length >6" (150 mm) and edge distance >0 .4" (10 as consistent with ASME Code, Class 1 vessel design
mm) and quality. Therefore, the comparison of the test
Cruciform or T-joint fillet or partial penetration
welds with corner undercutting ground out data to the ASME Code S-N curve is not directly
G Attachment fillet welds perpendicular to load, edge applicable but is useful for assessing FSRF guide-
distance <0 .4" (10 mm) lines . The controlling data are plotted (Fig . C .2) with
Fatigue Strength 39
300 rr a. { .1
100
so..
''k%\\\%'%.
.!'%\%Ik
''%
a
20
10s
10' I0g
Endurance, N Cycles
Fig . B .1-S-N curves from Gurney (1976).
Table B .2-FSRF the ASME Code design curve and a FSRF = 2

TWI 1E6 a' FSRF to: adjusted ASME Code design curve superposed on the
Category Description (N/mm2) Plate Weld data . The controlling data point is a main seam weld.
In comparison with the constructions shown in Fig.
B Full butt, parallel load 178
as-rolled plate C .1, a main seam-weld with NDE should not have
C Full butt, transverse load 151 1 .2 the lowest fatigue quality . This indicates a high
D Process defined 115 1 .5 1 .3 probability that the weld received no NDE . It is
E Process defined 101 1 .8 1 .5
F Backing strip, some fillet 86 2 .1 1 .8 important that such low quality welds (Fig . C .1) can
welds be covered by an FSRF = 2 ; this indicates that
F2 Minor changes from F 75 2 .4 2 .0 themaximum FSRF for welds that meet Code quality
G General fillet 63 2 .8 2 .4 could be 2 .0 . However, the approach used to plot the
40 WRC Bulletin 432

r
Fig . C .1-Geometries from TWI report.
1 1 I 11111 I I IIIIJ 1 I I I lilt

1
6
ASME DESIGN CURVE • Table 1 Base Metal 3
WITH FSRF = 2.0 n Table 2 Toe Cracks 2
♦ Table 3 Root Cracks
O Table 4 Main Seam

6
q Table 6 One Side
3
2
3
2
1 1 1 111111 1 1 1 111111 1 1 1 111111 1 1 1 11111

2 3 a 1E+4 2 3 5 1E+5 2 3 5 1E+6 2 3 5 1E+7
Number of Cycles, N
Fig . C .2-Data vs FSRF = 2 .0 .
Fatigue Strength 41
data is difficult to interpret and may not be consis- team does not understand the basis and believes
tent with how ASME Code analyses are done . Thus, that NDE will override the weld process . Classifica-
the FSRF = 2 may not be conservative . However, the tion III lumps all the conditions that are in the
FSRF = 2 certainly gives an indication that quality FSRF = 2 to FSRF = 4 weld quality level.
welds do not require FSRF > 2. Table C .2 shows the FSRF calculations based on
Harrison and Maddox develop three categories for Harrison and Maddox (1980) classifications . First,
use in the British design code ; these are shown in the stress range at 1E6 cycles is obtained ; these are
Table C .1 and plotted in Fig . C .3 . The first classifica- provided in the first column (1E6 o) . Next, the
tion indicates that base metal and flush welds with stresses are normalized, first to the Class I stress
MT/PT are equivalent. Classification II makes a range [column 2, FSRF to Plate (and ground weld)],
distinction by weld process, indicating that shielded then to the Class II stress range (column 3, FSRF to
metal arc (SMA) is a higher quality than submerged Weld) . The normalized values are the FSRF.
arc . The proposal does not address this ; the B & W Table C .2 FSRF are almost negligible compared
Table C .1-Harrison and Maddox : Weld Classification Table C .2-Harrison and Maddox : FSRF
Classification Description 1E6 a FSRF to:

Classification Description (N/mm?) Plate Weld
I Base metal, including repairs, flush with MT/PT
Main seam welds, machined flush, MT/PT I Base metal 99
Main seam weld, SMA, overfill left on MT/PT welds
II Main seam weld made by submerged arc with II Submerged arc 81 1 .2
overfill left on III Backing strip 72 1 .4 1 .2
III Main seam with backing strip Nozzle weld
Nozzle to shell welds Fillet welds
Reinforcing weld to shell pad
Fillet welded attachments Note : Geometric SCF must be superposed . Most of the data
come from Maddox tests . Curve based on 99.5 (with a three sigma
Note: No mention is made of MT/PT in II or III . deviation) survival probability.
1000
900
soo
700
600
300
400
300
N
A ~•
•
1~
100
90
SO
70
60
so
40
3 s a s
1o3 104 10 5 10 6 10'
Endurance, cycles
Fig . C .3-Proposal for British design curve.
42 WRC Bulletin 432

with the proposal . However, the low FSRF values are quality level of 7 . The fillet weld values (SWC
consistent with the data that Harrison and Maddox Categories D and E) of 1 .8 and 2 .3 relative to SWC
discuss ; i.e ., the data have so much scatter that the Category B are reasonable if Category B relates to
base S-N curve is reduced, but various effects have quality level 4.
minor impact . The B & W team considers this
unrealistic; it probably results from inadequate un- Appendix II
derstanding or interpretation of the data. Comparisons to ASME Code
D . Structural Welding Code (SWC) S-N Fatigue Curve
The Structural Welding Code (SWC) is very de- The quantifying of FSRF values for welds is in-
tailed and covers a large quantity of conditions and tended for application in ASME Code (1997) design
variations . Table D .1 shows the major conditions (A analyses . Other Codes, as shown in Appendix I,
through E) that are considered relevant to pressure address weld quality on the basis of their geometry
vessels . Butt-welds require volumetric NDE ; the and weld process . Appendix II compares the Code
acceptance criteria for this NDE have not been with FSRF for two data sets as an additional validity
determined . Figure D .1 shows the fatigue curve for check on the applicability of the recommended FSRF
redundant structures . The non-redundant curve is values ; these are as follows:
similar but reduced below 1E6 cycles and slightly
A. ASME Code S-N vs the MSC Fatigue Curve
non-parallel. Table D .2 gives the FSRF based on
B. ASME Code S-N vs PVRC 1991/1994 Fatigue
SWC Categories A and B.
of Butt Welded Pipe
Table D .2 shows the FSRF calculations based on
SWC Category A (base metal) and on SWC Category The first item (see Section A) shows that the ASME
B (full penetration ground butt-weld) . First, the Code (1997) is conservative relative to AISC, except
stress range at 1E6 cycles is obtained ; these are by a small amount for a small percent of the cycle
provided in the first column (1E6 u) . Next, the range . Also, the AISC FSRF values are consistent
stresses are normalized, first to the SWC Category A with the SWC ; thus, the recommended FSRF for
stress range (column 2, FSRF to Plate), then to the welds based on NDE is conservative relative to
Category B stress range (column 3, FSRF to Weld). AISC.
The normalized values are the FSRF. The second item (Part B of this appendix) presents
The FSRF for the best weld is 1 .3 . This weld data from recent pipe weld tests and compares the
receives a volumetric, but no MT/PT and no indica- data to both the ASME Code (1997) failure curve and
tion of a visual . It compares with the proposed the design curve . All the data points are above the
quality level of 6 with an FSRF of 2 .5 to 3 . Thus, the design curve, and the design factors of 2 & 20 cover
proposal is conservative relative to the SWC. the FSRF. Thus, these data confirm the conserva-
The full butt-weld, ground vs non-ground has an tism of the ASME Code S-N curve for this type of
FSRF of 1 .4 . This is probably related to the weld-base- construction and the conservatism of the recom-
metal juncture that results from the reinforcement, mended FSRF values.
rather than the ground vs as-welded condition . This
should be reasonably well covered by the NDE for A. ASME Code S-N vs the AISC Fatigue Curve
quality levels up through 5.
The overall FSRF for fillet welds of 2 .3 to 3 .0 Appendix K, Strength Design Considerations, of
indicates that the proposed values are conservative. the AISC Code addresses the fatigue failure mode . It
This assumes that the SWC requires no NDE for states, "Members and connections subject to fatigue
fillet welds, thus requiring an FSRF = 3 to 4 from the loading shall be proportioned in accordance with the
provisions of this appendix . " The AISC provides
Tables A-K4 .1 and A-K4.3 to numerically define the
Table D.1-SWC Categories S-N curve . Table A-K4.3 presents the stress range
levels for multiple categories of conditions ; Table
Category Description
A-K4 .2 defines the categories . Category A is "base
A Base metal, rolled surfaces, smoothness <1,000 metal with rolled or cleaned surface ; flame-cut edges
B Full penetration butt-welds, ground, volumetric with ANSI smoothness of 1000 or less ." Category A is
NDE
Fillet welds with parallel loading closest in definition to the ASME S-N curve . Table
C Full penetration butt-welds, reinforcement not II .A-1 combines the AISC requirements of Tables
removed, volumetric NDE A-K4 .1 and A-K4 .3 for Category A.
Fillet welds, lateral attachments with length limita-
tions As shown in Table II .A-1, the AISC starts fatigue
D Fillet welds, lateral attachments with length limita- considerations at 20,000 cycles compared with ASME
tions Code (1997) 100 cycles . Somewhere between 2E4 and
E Fillet welds, lateral attachments without length
limitations (L > 4") 5E4 cycles, the ASME Code design factor of 2 applied
Intermittent fillet welds attaching longitudinal to the smooth specimen failure curve controls the
stiffeners design curve rather than the 20 on cycles . This
Fatigue Strength 43
60 414.0
5o 345.0
40 275.0
30 CATEGORY A 207.0
CATEGORY B —' 138.0

`~ CATEGORY C FOR STIFFENERS' 103.5
e nILnn maws
Tr t- 69.0
CATEGORY F — 1 =aim
_NMI 62:1
"~~ —
ri~~tnn~~~
55.2
48.3
~ CATEGORY 0 41A
! 1 ( I
f l l 34.5
CATEGORY C FOR OTHER ATTACHMENTS J CATEGORY E 27.6
l 20 .7
2
!III 13 .6
1 ~
6.9
2 3436789 6 2 3 4 5 6789 7 2
10 10 4x10 7
CYCLE UFE
• TRANSVERSE STIFFENER WELDS ON GIRDER WEBS OR FLANGES
Design Stress Range Curves for Categories A to F

Redundant Structures (see 9 .4, Structural Welding Code)
Fig . D .1–Structural Welding Code S-N curves.
Table D.2—FSRF Failure Locations and Crack Origins are as Follows

FSRF to: Test Location Crack Origin
SWC 1E6 0u Plate Weld
Category Description (ksi) A B 1 HAZ at base metal Inside surface
2 Toe of weld Outside surface
A Base metal (plate) 30 3 Toe of weld Outside surface
B Full butt, ground, volumetric NDE 23 1 .3 4 Toe of weld Outside surface
C Full butt, volumetric NDE 16 1 .9 1 .4
D Fillet welds, limited length 13 2 .3 1 .8
E Fillet welds 10 3 .0 2 .3
Test Code
Table 11 .A-1—AISC Fatigue Data for Category A Point Strain Cycles Strain FSRF
Loading Stress Range 1 3,890 825 8,972 2 .3

Condition Cycle Range (ksi) 2 1,964 4,621 4,373 2 .2
3 708 1 .3E5 1,318 1 .9
1 2E4–1E5 63 4 10,680 202 17,400 1 .6
2 1E5–5E5 37 5 5,800 791 9,138 1 .6
3 5E5–2E6 24 A 9,250 523 10,943 1 .18
4 >2E6 24 B 13,400 285 14,574 1 .09
44 WRC Bulletin 432

indicates the condition is load-control rather than B . PVRC Grants (91-11, 92-12, 93-21, 94-14):
deformation-control . Figure II .A-1 compares ASME Fatigue of Butt Welded Pipe
with AISC . The "ASME DESIGN" curve is from the
A PVRC sponsored fatigue test of pipe welds was
book ; the "ASME MEAN" is the failure curve recalcu-
conducted under the direction of Dr . Scavuzzo (1997)
lated using a factor of 2 on stress . Figure II .A-1
at Akron University. This testing is an extension of
shows:
the Markl data (many years of testing and publica-
1. The ASME curve is very conservative relative tion) . The basic goal is to extend the Markl work to
to AISC from 2E4 to 1E5. strain levels that produce fatigue failure in less than
2. Between 1E5 and about 1 .3E5, ASME is non- 2,000 cycles . The specimen was filled with water and
conservative relative to AISC. failure is defined as the initiation of a leak . Four
3. From 1 .3E5 to 4E5, ASME is again conserva- specimen types were tested ; these are as follows:
tive relative to AISC.
1. Carbon Steel, 1 .5" Schedule 80, A-53 Type F
4. From 4E5 to 1E6 (ASME max cycles), ASME is
(t = 0 .200", D/t = 7 .5)
non-conservative relative to AISC.
2. Carbon Steel, 4" Schedule 40, A-106 (t = 0 .237",
Overall, Fig . II .A-1 shows the ASME curve is D/t = 17)
conservative relative to AISC . For the small zones 3. Stainless Steel, 1 .5" Schedule 80, Type 304
where ASME may not be conservative relative to (t = 0 .200")
AISC, the difference is small and considered to be 4. Stainless Steel, 4" Schedule 40, Type 304
outside the accuracy of the process . At 5E5 cycles, (t = 0 .237")
the factor between AISC and ASME is only 1 .125
compared with the 2 on stress and something greater Test Summary
than 20 on cycles . Therefore, the ASME Code fatigue The test results are presented in Fig . II .B-1 and
curve is considered as consistent or conservative Table II .B-1 . Control specimens (base metal) were
relative to AISC. run for the 1 .5" pipe weld ; two for the carbon steel
The AISC categories are quite similar to those in (CS) pipe and one for the stainless steel (SS) pipe.
the SWC even though the fatigue curves are differ- The 1 .5" specimens were radiograph (RT) cleared;
ent . Both of these codes had access to the same data the 4" specimens had no NDE . The 1 .5" specimens
base . The AISC and SWC FSRF (per category) are were tested at strains greater than 5,000 p in/in
also about the same . A simple comparison between (N < 1,300 cycles), whereas the strains for the 4"
Fig. II .A-3 and the SWC FSRF values as presented specimen ranged from 8,000 to 700 p in/in (N from
in Appendix I, Table D .2, confirms this . Therefore, 600 to 2E5 cycles).
the proposed FSRF values based on NDE are consis- Figure II .B-la (CS, 1 .5" pipe) shows that the test
tent with the AISC, as well as the basic fatigue specimens have only a mild reduction in life relative
curve . to the control specimens . For example, if a best-fit
ASME SEC III (1992), FIG 1-9 .1 FOR C AND LOW ALLOY STEEL w/UTS<80 KSI
100 AISC, 9th EDIT, TABLE A-K4.3, CATEGORY A
♦ ASME DESIGN,(ie 2 on Stress or 20 on Cycles)

e AISC, CAT A
- ASME MEAN (w F=2 on Stress removed)
100000 1000000 10000000

CYCLES
Fig . II .A-1-ASME vs AISC fatigue curves for un-notched specimens (i .e ., FSRF = 1 .0).
Fatigue Strength 45
Note that Snominal,alt is computed without SCFs or FSRFs
100 i.e . the allowable for each category has the FSRF "built-in"
v)
n
• -1t CAT A
I--
•
-a-
C) -4 CAT B
10 -N=.n--
-A CAT B'
•CATC
€I CAT D
-e- CATE
A CATE
1
10000 100000 1000000 10000000
CYCLES
Fig . II .A-2-AISC (9th ed .) fatigue allowables.
10
8 n CAT A
+ CAT B
6 * CAT B'
• CAT C
4 $CATD
m .= +CATE
2 *CATE'
0—
10000 100000 1000000 10000000
CYCLES
Fig . 11 .A-3-AISC (9th ed .) equivalent fatigue FSRFs . Note that the FSRFs inferred from the AISC fatigue allowables may not be comparable to
FSRFs used with ASME curves.
46 WRC Bulletin 432

a. CS 1 .5" Pipe b. CS 4" Pipe
0 0 Control Tests ° Weld Tests
0 Weld Tests
1 .0E+5 1E+4 —
1.0
5— 5
3
2— 2
1 .0E+4 1E+3
5
3— 3
2— 2
1 .0E+3 1 1 iIr1111 1E+2 r u1UUl u 11109 1 11111
3 2 5 2 3 52 3 52 3
1 .0E+2 1 .0E+3 1E+2 1E+3 1E+4 5 1E+5
Cycles Cycles
c. SS 1 .5" Pipe d . SS 4" Pipe

1E+4
•Control Test * * Weld Tests
5
+ Weld Tests 3
2— 2 *
+ + *
5— 5
3
2— 2 ma.
1E+3 i u u milt r r r i rrnf 1E+2 1119 1 x119 r umWulf

2 3 5 23 E+32 3 52 3 52 3
1 E+2 1E+3 5 1E+4 1 1E+4 1E+5 5 1E+6
Cycles Cycles
Fig. II .B-1—Scavuzzo et al . (1997) test data.
Fatigue Strength 47
Table II .B-1-Fatigue Tests of 1 1/2" and 4" Butt Welded Pipes*
1 .5" Diameter 4" Diameter
Specimen Displ . Strain Alt. Specimen Displ . Strain Alt.
No. Amp. Microin . Cycles No . Amp . Microin. Cycles
1-CS-C 1 .70 9,250 523 1-CS-W 1 .5 3,107 2,382

2-CS-C 2 .50 13,400 285
3-CS-W 1 .70 10,270 249 3-CS-W 1 .35 2,975 2,834
4-CS-W 1 .50 10,100 304 4-CS-W 0 .8 1,649 21,008
5-CS-W 1 .20 8,300 615 5-CS-W 0 .4 708 1E5
6-CS-W 2 .30 14,720 155 6-CS-W 2 .2 3,890 825
7-CS-W 1 .05 6,530 836 7-CS-W 2 .4 7,310 1,058
8-CS-W 1 .30 8,480 461 8-CS-W 0 .65 1,150 52,263
9-CS-W 1 .10 7,800 786 9-CS-W 0.73 1,287 90,773
10-CS-W 1 .40 9,170 386 10-CS-W 1 .25 3,278 3,925
11-CS-W 2 .50 15,640 183 11-CS-W 2 .6 5,753 674
12-CS-W 1 .90 10,680 202 12-CS-W 1 1,964 4,621
13-CS-W 1 .00 6,230 708 13-CS-W 2 .8 6,013 635
14-CS-W 0 .90 5,800 791 14-CS-W 0 .85 1,502 38,980
15-CS-W 1 .1 2,063 16,327
1-SS-C 1 .75 9,300 751 1-SS-W 1 .2 2,042 14,107
2-SS-W 1 .75 9,535 596 2-SS-W 2 .4 6,221 1,587
3-SS-W 2 .50 13,200 195 3-SS-W 0 .6 993 2E5
4-SS-W 1 .40 7,830 1,224
5-SS-W 2 .00 10,730 478
6-SS-W 2 .25 12,560 311
7-SS-W 1 .60 8,510 673
CS = carbon steel ; SS = stainless steel ; C = control; W = welded ; Disp . Amp . = Displacement Amplitude ; Strain Microin . = Strain (p
in/in) ; Alt . = Alternating.
*From Scavuzzo et al . (1997).
curve of the test data is drawn parallel to the control data, the designated stresses for the curves are
data-line, the reduction in cyclic life is small. divided by the modulus, E = 30E6 psi, to obtain the
Figure II .B-lb (CS, 4" pipe) shows a reasonable strains (E = Q/E).
level of scatter, over a relatively wide range of strain. Figure II .B-4 shows that there are three data
Figure II .B-lc (SS, 1 .5 " pipe) has only one control points well below the ASME Code "best fit" curve;
specimen . Its data point fits within the test specimen these are designated 1, 2, and 3 . All three are from 4"
scatter band . Figure II .B-ld (SS, 4" pipe) shows only pipe specimens . For the 1 .5" pipe specimens, the 2
3 data points ; it shows a reasonable distribution. lowest data points are designated as 4 and 5 . The
Figure II .B-2 (Fig . 19 from the PVRC report) following table shows the development of FSRF for
compares the best-fit for the 1 .5" pipe data to the these 5 data points. For this table, the FSRF are
Markl best-fit results . All of the data except one point related to the ASME Code failure curve . Also pre-
fall above the Markl curve . If a best-fit curve, sented, in the following table, is the FSRF for the
parallel to the Markl curve, were drawn starting at control data points, which are designated A and B for
the 1,500 cycle data point (4" pipe data), the best-fit the 1 .5" CS pipe ; the 4 " pipe has no control speci-
1 .5" data curve would still indicate a change in slope mens . The appropriate FSRF is based on the control
for the higher strains . Thus, the new welds may have specimen.
better fatigue quality than the Markl welds, but the The FSRF for the five test-data points represent
S-N curve is not a linear continuation below 2,000 the full specimen vs the smooth database (Code)
cycles . Figure II .B-2 is a stress/cycle plot . The report
specimen . The maximum value is 2 .3 . For this FSRF
indicates that the calculation of stress from the
definition, the FSRF for non-inspected welds is 2 .0 to
measured strain is consistent with the Markl ap-
2 .5 ; for RTm (modified radiograph) welds, it is less
proach . This figure also includes Heald and Kiss
than 2 .0 and perhaps as low as 1 .6.
(1974) data for comparison.
The second FSRF definition (the test-data point vs
Data Assessment the control-data points) is more appropriate because
Figure II .B-3 combines the 1 .5" and 4" pipe data the FSRF for the two control-data points represent
grouped by test material : CS(a) and SS(b) . It indi- the design factor (2 & 20) that is put on the smooth
cates that there is no substantial difference between test specimens . For the second FSRF definition,
the 1 .5" and the 4" pipe data ; a reasonable S-N curve FSRF = 2 .0 is probably sufficient for the non-
can be fit to the data. inspected welds and 1 .5 is sufficient for the RTm
Figure II .B-4 superimposes the ASME (1969) S-N condition.
"best fit" and the ASME Code (1997) S-N design The low FSRF for the control specimens is not
curves onto the CS data . To obtain strain values for surprising. The specimen thickness is comparable to
these curves for comparison with the piping test that of the smooth test specimens . Great care ap-
48 WRC Bulletin 432

STRESS AMPLITUDE, psi
10000001
: . .. . : .ti,,,$ :: .. .
1 I ♦ . ♦ •
1 • • • f , t . .... .... . • . 1 • . . 1 t , 1 . • • 1 , t 1 -
. • .' . .. ' ' • . . . .. ♦•♦•

. •
100000 ~~
..
. s~
J . . .
!
! 1 , 1 1 •
t
. • • 1 . • t 1 „ 1
- -
a'ss•
. C$ CQt1TRQ ; t •▪ . . i • • . t . . i . . . . •
S. ` `• •~ .1~ t t . ..
1WSS WELDED : :
10000 : : : ; : . . : : :
. . . .
: ; ; ; ;
-: LEAST.▪ SQUARE .FI't.
HEALD'S DATA :- CS(b)
▪ . . . . . .
X HEALQ'S bit'TA : S(5)' At ▪• - • t
- MARKL'S EQUATION
1000 )ig MARKL & GEORGE :

100 1000 10000 100000
CYCLES TO FAILURE
Fig . ll .B-2-Comparison of Scavuzzo et al . (1997) and Heald and Kiss (1974) data . Heald and Kiss nominal stresses calculated using measured
stiffness.
pears to have been given to making the welds since allowable cycles . Thus, the FSRF accounts for a condi-
they appear to be a higher quality than the original tion that is not included in the stress analysis model.
Markl specimens . However, this does show the poten- For an applied membrane or bending load, the
tial conservatism in the ASME Code 2 & 20 design stress distribution is linear through the thickness . A
factor. To put it into perspective, Fig . II .B-4 also notch or flaw concentrates the stress, causing a
compares the specimen results to the ASME Code maximum stress (amax) to occur . The SCF is defined
design curve . All data points are above the design as the maximum stress divided by the nominal
curve . If these could be considered typical welds, no stress, i.e ., SCF = amax/anom where anom is either the
FSRF is necessary for predicting the fatigue life. applied membrane-plus-bending stress or the local
membrane-plus-bending stress (M + B) . Tradition-
Appendix III ally, the SCF is applied to the component or principal
FSRF Definition M + B stress prior to calculating the alternating
stress . Frequently, SCF predicts an overly conserva-
Many authors use the terms SCF and FSRF as if
tive cyclic life.
they were synonymous or as if the reader under-
The actual reduction in cyclic life, caused by a
stands their differences . However, the definitions for
notch or flaw, relates to test data . At an applied
FSRF are not consistent among authors . The defini-
stress level (S a '), testing the FSRF condition will
tion used in this report is as follows:
produce a reduced cyclic life (N i ') relative to the
When a condition exists that is not included in the stress no-notch condition . For the same cyclic life (Ni '), the
analysis, but will reduce the fatigue life, an FSRF is applied stress (Sa ) for the non-FSRF condition di-
used to obtain the effective stress that is consistent with vided by the applied stress for the FSRF condition
the actual fatigue life . Specifically, the potential for
flaws and notches is not included in FEA models . In (S a ') gives the FSRF. Thus, FSRF = S a/Sa '.
practice, the FSRF is applied to the calculated alternat- There are two tests that are used to determine the
ing stress prior to entering the S-N curve to obtain the FSRF. One determines the stress level for infinite life
Fatigue Strength 49
a. CS Butt Weld Tests
1E+5
5 q q 1 .5" Control Tests

O 1 .5" Weld Tests
3
A 4" Weld Tests
2
1E+4
3 AA
2
1E+3
3
2
1 I 1 111111 1 1 1 1 1 1111 I 1 1 1 1 1111

1E+2
1E+2 2 3 5 1E+3 2 3 5 1E+4 2 3 5 1E+5
Cycles
1E+5
b. SS Butt Weld Tests
5
• 1 .5" dia. Control Test
3
+ 1 .5" dia . Weld Tests
C 2
0 4" dia. Weld Tests
C
p 1E+4
+ +
+1 +0
C.)
5
E
3
C
2
Cu
L
1E+3 O
Q 5
3
2
I 1 11 1 1111 1 1 1 111111 1 1 1 111111 1 1 1 I

1E+2
2 3 2 3 2 3 2 3 5
1E+2 5 1E+3 5 1E+4 5 1E+5
Fig . II .B-3-Scavuzzo et al . (1997) data grouped by test material.
(nominally > 1E6 cycles) for a notched or FSRF The second test approach determines the cyclic life
condition specimen (Sa ') relative to an un-notched (Ni ') of the notched or FSRF condition specimen at
(S a ) specimen . The FSRF is the ratio of S a/S a '. any specific stress level above the endurance limit.
However, this FSRF is accurate only at the endur- Using the cyclic life approach, FSRF' = N ;fN i ' where:
ance limit. Ni is the number of cycles to failure for the un-
At the endurance limit, the FSRF will be equal to notched geometry at the defined stress and N i ' is the
or lower than the SCF. If the applied stress level number of cycles to failure for the actual (notch)
were increased, the FSRF would decrease, but the geometry at the same stress level . The factor is
SCF does not change . Thus, the FSRF is a function of designated FSRF' because it is a cycle ratio rather
stress level as well as notch/flaw geometry. than a stress ratio . If multiple stress levels are
50 WRC Bulletin 432

5
3 Test Data
Q—0 CS 1 .5" Control Tests
2 0 CS 1.5" Weld Tests
A CS 4" Weld Tests
1E+4 ASME "Best fit" for Carbon Steel
A
AA
1
2
e
1E+3
ASME Design for Carbon Steel,

5 UTS < 80 ksi
I I I l l 1111 I 1111111 I I 111111

A1
1E+2
1E+2 2 3 5 1E+3 2 3 5 1E+4 2 3 5 1E+5
CYCLES
Fig . II .B-4-Scavuzzo et al . (1997) data vs ASME Code "best fit" and design curves .
tested both with and without the notch or FSRF and Kuhn (1952) gave a similar interpretation of the
condition, two fatigue curves can be generated . The Neuber theory. O'Donnell and Purdy (1964) address
FSRF can then be defined by a stress ratio at any the issue ; their approach relates to what is some-
specified cycle level. times called the DEL approach. Snow et al. (1970)
In the "FSRF for welds based on NDE " procedure, generated some interesting data, which give insight
the FSRF relates to the endurance limit value ; i .e ., to the potential impact of using an FSRF rather than
its value is constant for all stress levels . This is a an SCF. He showed that the FSRF can be a factor of 2
conservative approach but not as conservative as or even 5 lower than the SCF. The topic deserves a
using an SCF. In reality, the FSRF will decrease as full review, but this is considered beyond the scope of
the applied stress increases . The FSRF will go the current effort.
asymptotically to 1 .0 at the stress level that pro-
duces a crack in a few cycles.
Appendix IV
There are techniques for defining the FSRF based
Other Fatigue Issues
on a geometric SCF (at a notch) . They are based on
the Neuber theory (1968), which essentially says During the early reviews of this report, a number
that the controlling stress is the average stress over of questions were raised which have an impact on
some volume of material . The theory is that the the accuracy of the fatigue analysis . The questions
amount of material relates to the grain size, thus that do not directly relate to the FSRF for welds
making the correlation a material property . Peterson issue are presented in the following discussions.
(1974) presented his interpretation of the Neuber These are considered to be important issues that
theory but modified Neuber 's equations . Hardraft may not have consistent interpretations in the analy-
Fatigue Strength 51
sis community. In that sense, these related issues the proposed procedure for welds addresses undetec-
should be addressed separately, just as the "FSRF for ted flaws/notches, metallurgical notches are not in-
Welds" topic is being addressed separately, even cluded.
though an interaction exists between the topics . The
questions are presented close to how they were Applying an FSRF
stated . The discussion is presented as a singular The traditional approach is to apply the FSRF to
view without defense . The pressure vessel industry S a before entering the fatigue curve . This is the
would benefit from developing consistent guidelines. approach in the report . However, this is a complex
question and there are definable conditions where it
Should the Maximum Concentration may be overly conservative . For example, should
for a Weld in a Blend Radius Be 4 .0? FSRF be applied to peak thermal stress or to the
FEA can be used to calculate total component FEA stress at a singularity? Additional work is
stresses at a blend radius (design notch) . As an needed to address these issues.
alternate, an SCF (e .g ., from Peterson) could be Another example uses a fillet weld with the stress
applied to the membrane and bending stresses to direction parallel to the bead . The test data on fillet
obtain the total stress . If the blend radius included a welds show that welding direction vs load direction
weld, superposing a weld FSRF should be consid- has an impact on fatigue life . This can be interpreted
ered . The superposition of a design-notch SCF and as allowing the FSRF to be applied to the component
weld FSRF is currently considered an open question. stress . However, this is not necessarily true, espe-
The conservative approach is to do just that, super- cially across the board . It has thus not been incorpo-
pose them . For low total stress conditions, this may rated in the current work . Welding (direction) pat-
give reasonably accurate results . For high total terns are more prone to flaws oriented in a specific
stress conditions, this is probably too conservative. direction than random directions . Thus, applied
SCF are accurate only for elastic conditions ; when S }, stresses that are normal to the flaw will have less life
is exceeded, they rapidly lose accuracy. For high than those where the stresses are parallel to the
stress conditions, especially those caused by FSRF, flaw. This forces the mean curve to shift, but may not
most of the fatigue life is crack growth, not crack change the scatter band . Also, using this presumes
initiation . Once a crack is initiated, the impact of the that the welding patterns are consistent (especially
SCF and the FSRF decreases . This leads to the weld repairs) and known . Also, starts and stops must
conclusion that there is a maximum FSRF level that be controlled . The test results must address this or
makes sense, perhaps a value of 4 . Obviously, the they are self-predicting prophesies ; i .e ., all the test
applied stress level plays a role in this decision . In specimens were welded in one pattern and there is
summary, one might superpose the SCF and FSRF; if no other way to do it or the rules require this . Much
the combined value exceeds some value (e .g ., 4 or 5) more discussion is appropriate.
and Sa exceeds S3„ a reduction is appropriate. When the FSRF is developed from a geometric
SCF (at a notch), there is a basis for direction
Some Visual Inspections and Some Failure
Assessments Make It Appear orientation. It is consistent to apply the FSRF to the
That FSRF = 4 Is Inadequate principal stress (consistent with SCF direction) rather
For the proposed procedure, there is no worse than to Sa .
condition than no NDE . The available test data that
are relevant to pressure vessels indicate that FSRF -- Appendix V
4 should be acceptable as a maximum value . How- Level 7, Weld-Toe Notches,
ever, it may be possible to make welds that need an and Non-Inspectable Weld Roots
FSRF greater than 4 . If it can be shown that there
It is probably part of human nature to produce
are conditions where FSRF > 4 are appropriate, the
what is expected . If the design engineer says that
first choice should be redesign to allow for a higher
NDE is not required for a specific weld, the welder is
quality weld . Otherwise, the designer must assume
likely to give the weld less attention than he wbuld
an FSRF > 4 . The comment implies a visual (VT)
give to a weld that will receive a full NDE . The
should always be applied . As a minimum, non-
welding engineer is likely to use the least expensive
inspectable welds should not be used in areas of
process that will do the job but may be more prone to
significant stress . The ASME Code (1997) does not
flaws . We can also view the "human nature" from the
allow non-inspectable welds in high stress locations
positive side where full NDE is required . If the
for Class 1 pressure vessels . However, the issue of
inspection people detect a lot of rejectable flaws, the
" how high a stress is acceptable" is open.
grinders get a lot of work, the welders get more work,
The Procedure Does Not Address and even the inspectors get more work . However,
Metallurgical Notches management is likely to ask, " Why are you destroy-
The issue with metallurgical notches is calculating ing our profit margin?" They are liable to find a new
an accurate stress rather than addressing flaws. welding engineer and to develop a new welder quali-
Doing so requires a knowledge of the cyclic stress- fication program . The bottom line is that the knowl-
strain curves for both base and weld materials . Since edge of required NDE improves the weld quality.
52 WRC Bulletin 432

Level 7 (1990) discussed weld-toe notches (undercuts) in
Quality level 7 has FSRF = 3 to 4 . The quality of terms of three types, namely, (1) wide and curved, (2)
the weld process is expected to be the major determi- narrow (crack-like), and (3) shallow and narrow (Fig.
nant in whether the appropriate FSRF is closer to 3 V-1).
or 4 . For example, a weld process that has been used Type 1, wide and curved undercuts, can be de-
successfully on many welds where full NDE is ap- tected visually. Thus, with the correct accept-reject
plied should merit an FSRF = 3 even if it receives no criteria, Type 1 can be blended (e .g ., 3 :1 taper and
NDE . However, if the weld process has no demon- bottom radius equal to depth) and a FSRF would not
strated quality in either a weld development pro- be required . Petershagen (1990) states:
gram or in production with suitable NDE, an FSRF =
4 may be appropriate . This applies regardless of the No indication has been found that test results have been
weld type, but it is more likely that a full penetration used in establishing permissible figures for undercut . It
can therefore be concluded that permissible values for
weld process (especially if it is automated) would undercut in existing codes are based on engineering
have a higher quality than a fillet weld process. assessment and are directed towards good workman-
Therefore, it may appear that the full penetration ship rather than fitness for purpose.
weld is always superior to the partial penetration
The "good workmanship" should be covered by a
weld or the fillet weld, but the reality is that the
quality is a function of the weld process and other generous radius at the bottom coupled with a gener-
parameters such as general geometries and applied ous taper. Allowable surface roughness can have
stress levels. more severe notches than a well-blended notch.
The general geometries relate to both the disconti- Therefore, considering the factors that establish the
nuity and a local notch . For example, the full penetra- ASME Code (1997) design curve (2 & 20 on the test
tion weld is less likely to be at a discontinuity than data failure curve), the blended Type 1 undercut
the partial penetration or fillet weld . Thus, the latter should not require an FSRF to have "equivalent
two types have the weld notch superposed on the fatigue life" of base metal.
discontinuity notch, but the full penetration weld is Type 2, narrow (crack-like) is described by Peter-
more likely to only have the weld notch . The applied shagen (1990) as follows:
stress level is a misinterpreted parameter . The pres-
With an undercut of Type 2, depth measurement be-
ence of a discontinuity produces additional stresses comes difficult, speculative, or even impossible . Dye
to the applied stresses, just as additional or more penetrant or magnetic particle testing can be used to
severe notches increase the total stress . The percep- detect a crack-like undercut of this type.
tion of full penetration welds having a higher quality
than partial penetration or fillet welds may be due to Once detected, the Type 2 can be repair welded or
the total stress level vs the nominal applied stress at blended as discussed for Type 1 . Weld repair may
the crack initiation location . Thus, in the global view, require PWHT or the use of special weld procedures
it is reasonable to expect butt-welds to have a lower such as temper bead. It also requires re-inspection . If
FSRF than partial penetration and fillet welds when: the Type 2 is shallow relative to the through-
thickness dimension, blending may be a better choice
• The butt-welds have a well-developed process. than weld repair. The conclusion is that the Type 2
• The butt-welds are removed from discontinui- undercut is expected to be detected by the full
ties. surface examination and removed.
• The butt-welds are not inspected due to an Regarding Type 3, Petershagen (1990) states:
established low stress condition.
Microflaws of Type 3 are generally not detectable visu-
Another aspect is the perception that locations ally or by non-destructive testing. They are found by
with low stress levels do not require much NDE . This careful sectioning and magnification of the order of
is easily translated to "thick (large) parts have high 250 :1.
stresses and thin (small) parts have low stress
levels ." The translation is not valid especially for This type of crack should not exist whether the
small parts . Applying FSRF = 4 to locations with low material is base metal or weld metal . For example,
(insignificant) stress levels would probably produce austenitics are prone to cracking if the micro-
acceptable usage factor, whereas an FSRF = 3 may structure or heat treatment is not adequately con-
cause a usage factor greater than 1 when the location trolled . The same applies to welds due to such
has significant stress levels . As a result, an FSRF = conditions as hydrogen cracking or ductility dip
4 may produce no usage factor for a low stressed, full cracking . The development of a weld process should
penetration weld but show usage factors well above always include a destructive examination for cracks.
1 .0 for a highly stressed fillet weld. This somewhat understates the problem since Peter-
shagen described Type 3 as having depths of 0.25
Weld-Toe Notches mm (0 .010 ") . If these micro-cracks exist in the weld
The weld-toe surface notch is a common condition (or base metal), they will not be detected by current
especially when no NDE is required . Petershagen NDE procedures . Certainly, there is a need for the
Fatigue Strength 53
potential impact of micro-cracks on fatigue life. Non-Inspectable Weld Roots
Assembling available data and recommending test- Both partial penetration and fillet welds are used
ing to determine the impact are outside the scope of for attachments ; in addition, partial penetration
the current project. welds are used for small nozzle penetrations . For
In addition to the weld-toe notch, the blend of the attachments, the FSRF can be significantly different
weld to the base metal is significant to fatigue life. In for the partial penetration vs fillet welds . Figure
addition to presenting the three types of weld-toe V-2a shows the beveled weld-prep for the attachment
notches, Fig . V-1 shows an extreme, no blend radius partial penetration weld . Full penetration is not
condition . When the weld intersects the base metal expected ; thus, there is likely to be intermittent
at a right angle, there is a severe geometric notch
rounded flaws at the nose of the weld-prep (root).
regardless of undercut . This geometry precludes
These are fully embedded ; thus, their impact on
detecting a Type 2 weld-toe notch by MT/PT . It is
fatigue life should be minor. If there occurred a lack
appropriate to treat this location as non-inspectable
with an FSRF = 4. of bond between the bevel and the weld, one could
If the weld to base metal juncture has a defined project a significant concentration with an SCF > 3.
blend radius (Fig . V-1), the severity of the geometric However, the flaw is embedded and the stresses to be
notch is greatly reduced. With a sufficient blend concentrated would be low. Relative to the surface
radius, MT/PT or VT can be performed and the location, regardless of whether the stresses are
weld-toe notches can be detected . Thus, the blend imposed by the shell (shown in the figure) or imposed
radius may result in an FSRF of 2 .0 rather than by a load on the attachment, the FSRF would be
FSRF = 4 for no blend radius. significantly less than the SCF. An FSRF = 3 is
Type 1 Type 2
Wide and Curved Narrow and Crack Like
Lack of Blend Radius
Figure V-1—Weld toe notches.
54 WRC Bulletin 432

a. Partial Penetration Weld b. Fillet Weki
Figure V-2—Non-inspectable weld roots.
expected to be appropriate and should not present a notch at the root of unknown geometry or severity.
fatigue problem. Since none of the standard NDE can detect these
Figure V-2b shows the same attachment with a notches, an FSRF = 4 is reasonable. Unlike the
fillet weld . It is difficult to project whether the attachment weld, the root location may see signifi-
attachment (plate) will carry much load, regardless cant stress levels . Also, depending on the quality of
of whether the load is imposed by the shell (shown in the weld process, there may be a lack of bond
the figure) or imposed by a load on the attachment. between the weld and cylinder at the root pass . This
However, the large lack of bond between the attach- flaw may be very tight with a sharp root superposed
ment plate and the shell may respond to loads as a on the rounded root notch . It is more likely (with a
free surface (low constraint) . Since the ASME Code well-developed weld process) that the root of the flaw
(1997) required NDE will not detect these flaws and will have only a radius . Depending on the depth and
they could be severe, an FSRF = 4 could be realistic root radius, the FSRF will vary from 3 to 4.
unless the weld process is very well developed . As a An FEA will often model the root as a sharp corner
minimum, the fillet weld root should have a higher with no blend radius . This is a singularity that
FSRF than the partial penetration root for attach- produces a stress concentration with an SCF directly
ment geometries. related to the FE grid refinement . To reasonably
The ASME Code, Section III, Fig . NB-4244(d)-1 assess the stresses for a fatigue analysis, the FSRF
shows geometries for small nozzle and branch piping is applied to the membrane-plus-bending stress at
connections . These could be welded with either par- the location . Thus, the FSRF is caused only by the
tial penetration or fillet welds . The weld produces a welding notch.
Fatigue Strength 55

WRC 432-1998@ (Asmesecviii)

Uploaded by

Copyright:

Available Formats

WRC 432-1998@ (Asmesecviii)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

WRC 432-1998@ (Asmesecviii)

Uploaded by

Copyright:

Available Formats

BULLETIN 432 JUNE 1998

REPORT NO . 1 : Interpretive Review of Weld

REPORT No . 2 : Fatigue-Strength-Reduction Factors

These Bulletins contain final Reports from projects sponsored

For more information, see www.forengineers.org

Copyright © 1998 The Welding Research Council.

Welding Research Council

WRC Bulletin 432—June 1998

WELDING RESEARCH COUNCIL, INC.

Report No . 1 : Interpretive Review of Weld Fatigue-Strength-Reduction

Report No. 2: FSRF for Welds Based on NDE 27

This Bulletin contains two reports prepared under the auspices of

Objectives local concentrations of stress . These irregularities

WRC Bulletin 432 1

log (Fatigue Life, cycles)

2 WRC Bulletin 432

Category B Angle Joint

Category C Butt Joint

4 WRC Bulletin 432

Type 1 Corner Welds

0.25tn , but not

Type 2 Corner Welds

6 WRC Bulletin 432

Taper = cot (a);

Wedge Transition Taper

130 140 150 160 170 180

Reinforcement Angle (0), degrees

8 WRC Bulletin 432

(a) Equal Leg Fillet Weld

Surface of vertical member

Convex fillet weld

(b) Unequal Leg Fillet Weld

x min. to nominal pipe

x min . 1 .4tn or the thickness of the hub, whichever

Fig . 10—Flaws encountered in welded joints (AWS, 1989).

10 WRC Bulletin 432

Fig . 11—Cracks encountered in welded joints (AWS, 1989).

Cyclic Plasticity dence of yielding. In the HCF regimen (typically

12 WRC Bulletin 432

14 WRC Bulletin 432

0 50 100 150 200 250 300

16 WRC Bulletin 432

10 7 I 1119 111I119 11 11119 11 11119 11 1111111 1111111y 111111y 11 1 nu' 1 111119 1 I I I lI

Austenitic Steels, Curve A

I 107 1 1 1 111111 1 1 1 111111 1 1 1 1111 1 l I 1 11111 1 1 1 1 1 1 1±

I 11111111 I t I l 11111 I 1 1 111111 I 11111111 1 1 1 1 1 11

18 WRC Bulletin 432

I 11111111I 11111111I 11111/11I 11111111 I I 111111

20 WRC Bulletin 432

I I 1111111I I 1111111 11 1111111I I 11111111 I 11111111 I 1111111- T"1 rrti,tl— 1-1-111+1

---~--- Carbon Steel

Dotted lines are best fits to data, -

Mean Stress, psi

11 7 10 I 1 1111111 111111111 11111111` 11111111` 111111 111111111 1111111/

1 1 1111111 1 1 1111111 111111111 II I 1 11111 II I I t 1111111 11111111

22 WRC Bulletin 432

1 11111111 111111111 111111111 I II 111111 I II111,11 1 11111111 111111111 I ItMI

Fig . 23—Fatigue of stainless steel socket weld joints (Higuchi et al ., 1996a,b).

I I I I III1 1 I I 111111 I I I 1 IWI I I I I 1111' I I I 11112