ABS TECHNICAL PAPERS 2003

nd

22

Proceedings of OMAE 2003

International Conference on Offshore Mechanics and Arctic Engineering
June 8-13, 2003, Cancun, Mexico

OMAE 2003-37238
SENSITIVITY OF FATIGUE ASSESSMENT TO THE USE OF
DIFFERENT REFERENCE S-N CURVES
Xiaozhi Wang
American Bureau of Shipping
16855 Northchase Drive
Houston, TX 77060, USA
cwang@eagle.org

Zhan Cheng
American Bureau of Shipping
16855 Northchase Drive
Houston, TX 77060, USA
zcheng@eagle.org

ABSTRACT

INTRODUCTION

The S-N curve based fatigue assessment approach is

the most widely used one in both ship and offshore
industry, in contrast with a fracture mechanics approach.
The S-N approach, implemented by either simplified or
spectral method, has to apply S-N curves to calculate
fatigue strength. The S-N curve, which represents the
number of cycles (N) of a constant amplitude stress range
(S) that will cause a fatigue failure, is normally developed
based on experimental data. Which S-N curve should be
applied to a particular detail depends very much on the
geometry of the detail, welding information as well as
loading condition.

The S-N curve method is the most widely used

procedure to establish the fatigue strength (life) of
structural details in both offshore and ship engineering
practices. The application of the S-N curve method is
governed by a variety of engineering standards that
typically share some features. These common features
include:
The format and implicit confidence bounds of the
design S-N curves (which are established from
experimental data from specimens that are
considered representative of the structural detail
being assessed);
The categorization of generic structural details into
particular S-N curve classifications (which are
based on the geometry and loading of the detail;
and also occasionally specifics of the welding procedure, or
the extensiveness of the non-destructive
examination (NDE) to be provided during
structural fabrication,
or both);
Adjustments to the design S-N curves to account
for other considerations such as structural
element thickness and the relative corrosiveness
of the environment that the detail will
experience.

There are various S-N curves published by different

institutions, e.g., S-N curves published by UK HSE, IIW,
AWS etc. The newly developed ABS Guidance on
Fatigue Assessment of Offshore Structures, [1], proposes
the ABS S-N curves, in which two categories of joints,
tubular and non-tubular, are included, and both size and
environment effects are taken into account. However, the
application in Gulf of Mexico is also influenced by API
recommendations. In API RP 2A, AWS S-N curves are
referred, which in US practice is accepted for fixed
(buoyant and non-buoyant) platform deck structures.
The objective of this paper is to address the
difference between different S-N curves and to present the
detailed results of fatigue assessment by using different SN curves for non-tubular joints. Conclusions made based
on the study provide more background on the S-N curve
application in fatigue assessment.

While sharing these general features, the specifics of

each of these features in the various reference standards
can be quite different. This can result in disparities of
fatigue life predictions.

Sensitivity of Fatigue Assessment to the Use of Different Reference S-N Curves

141

ABS TECHNICAL PAPERS 2003

Internationally, relevant sources of specifications
governing the application of the S-N curve approach to
offshore structures are the U.K. Welding Institute (TWI)
and the International Institute of Welding (IIW). The
former are well known to the offshore industry as they
were published in various editions by the U.K. regulatory
bodies, the Department of Energy (DEn) and the Health
and Safety Executive (HSE). In the USA and in other
countries where the prevalent offshore structural
engineering standards have relied on the APIs
recommended practice-RP 2A, the API and by reference,
the American Welding Society (AWS) S-N curves are
used to establish fatigue strength.
ABS has studied the various S-N curves in use for
offshore structures and has developed its Guidance on
the Fatigue Assessment of Offshore Structures, [1]. The
Guidance provides proposals that reflect the features of
the international standards. This has resulted in the
development of new sets of S-N curves that are referred
to in the Guidance as the ABS S-N curves. But, the
Guidance also recognizes the application of API and
AWS S-N curves for major areas of fixed (buoyant and
non-buoyant) platform structures that are sited on the
U.S. Outer Continental Shelf.
This paper compares the ABS and AWS S-N curves
for non-tubular joints in offshore structures and presents
the results of fatigue assessments that have been
performed using these data. This exercise will illustrate
the background and application of these standards, and it
will also demonstrate the sensitivity of the fatigue life
predictions to the reference standards employed.

Offshore Structures [1] employs a combination of DEn

(1990) curves [2] and HSE (1995) curves [3]. The basic
S-N curves in-air are the same as defined by DEn (1990)
and adjustments for structures under corrosive
environments such as cathodic protection and free
corrosion in seawater are the same as defined by [3]. The
bases for this choice are: 1) the history of successful
practice, 2) worldwide acceptance, and 3) conservative
performance in the high cycle range.
In general, ABS in-air S-N curves, for both tubular
and non-tubular joints, can be represented by a twosegment S-N curve (Figure 1). When the number of
cycles, N, is less than NQ,, which is taken as 10-7 here,
the relationship between N and stress range (S) is:
N = ASm
(1)
where A and m are the fatigue strength coefficient
and exponent respectively, as determined from fatigue
tests.
When N is greater than NQ cycles,
N = CSr

(2)
where C and r are strength parameters determined
from fatigue tests.
The parameters defining S-N curves are listed in
Table 1, in which the first eight curves are for nontubular joints of various classifications and the last one
for tubular joints.
Figure 1: Two-segment S-N curve

NS m = A

number of cycles to failure in S-N curve

representation
NQ Number of cycles when S-N curve has
slope change
S stress range
A,m fatigue strength coefficient and exponent
determined from fatigue tests
C,r fatigue strength parameter determined from
fatigue tests
t
plate thickness of the member under assessment
tB basic (reference) thickness
SB stress range at basic (reference) thickness
k thickness exponent factor

ABS S-N CURVES FOR NON-TUBULAR JOINTS

IN OFFSHORE STRUCTURES
In order to establish S-N curves for offshore
structures, ABS has performed a comprehensive review of
fatigue test results and fatigue strength models for welded
joints. The ABS Guidance on the Fatigue Assessment of

142

Log(S)

NOMENCLATURE

NS r = C

m
SQ

1
r

Log (N)

Curve
Class

B
C

NQ

1.011015

4.0

1.021019

6.0

13

3.5

2.591017

5.5

1.5210

12

3.0

15

4.3310

5.0

1.041012

3.0

2.301015

5.0

6.3010

11

3.0

14

9.9710

5.0

F2

4.301011

3.0

5.281014

5.0

2.5010

11

3.0

14

2.1410

5.0

1.601011

3.0

1.021014

5.0

3.0

15

5.0

4.2310

12

1.4610

4.0510

Table 1 Parameters for ABS in-air S-N curves

Sensitivity of Fatigue Assessment to the Use of Different Reference S-N Curves

ABS TECHNICAL PAPERS 2003

When a joint is exposed to seawater, the above S-N
curves should be modified to reflect the impairing effects
of the corrosive environment on the fatigue strength of
the joints. For non-tubular joints in seawater with
cathodic protection or no protection, penalties are applied
to the in-air S-N curves to reflect the reduced fatigue
strength.
SIZE EFFECTS OF NON-TUBULAR JOINTS
When thickness correction is taken into account, the
S-N curve shown by Equation (1) can be expressed as
S
Log10 (N ) = Log10 ( A) mLog10
(3)
k
(t B / t )
where N is the number of cycles to failure, S is stress
range, and
Log 10 ( A) = Log10 ( A1 ) 2 Log10 ( N )

(4)

In which Log10(A1) is the mean value obtained by the

least squares regression analysis and Log10(N) is the
standard deviation of Log10(N). The design curve is
defined as two standard deviations of Log10(N) below the
mean S-N curve, which corresponds to a 2.5% probability
of failure.

from 56 MPa to 341 MPa and only 4 specimens have a

fatigue life exceeding 107 cycles.
The parameters of F-curves used in the two codes are
shown in Table 3. The basic F-curves in [2] and [3] are
identical but with different thickness correction formulae.

N<107

Codes

N>107

Log10(A)
m
Log10(A)
DEn (1990)
11.801
15.001
3
& HSE (1995)
Table 3 Parameters of F-curves (in air)

m
5

The design F-curves with thickness correction are

plotted against the test data, with one thickness in each
figure (for illustration, only three figures, as shown in
Figures 2 5, with thickness of 25 mm, 50 mm and 75
mm, are presented in this paper). F-curve of 16 mm
thickness (i.e. without thickness correction) in HSE
(1995) [3] is also plotted in the figures where it is
appropriate for reference. These series of figures
demonstrate the general detrimental effect of increasing
plate thickness. These figures also illustrate large safety
margins between the test data and design curves, with
HSE (1995) curve at the most.
1000

ABS [1]
HSE [3]

Stress Range (MPa)

Thickness correction to the stress range is also

included in Equation (3), where t is plate thickness of the
member under assessment, tB is the basic thickness (i.e.
the minimum thickness to which the thickness correction
should be applied) and k is the thickness exponent. In
Table 2, tB and k used in the two fatigue codes, i.e. DEn
(1990) [2] and HSE (1995) [3], are shown.

Test Data
HSE [3]-16mm
100

10

One of the objectives of this study is to compare the

above mentioned two thickness correction formulas with
the test data that were used in reviewing thickness effect
by [3].

DEn (1990)
0.25

HSE (1995)
0.30

tB

22 mm

16 mm

Table 2 Parameters of plate thickness correction for

non-tubular joints
An analysis is undertaken on data from tests on aswelded T-butt and cruciform joints that belong to the Fcurve joint classification, as presented in the HSE (1992)
background document [4]. The specimens vary in
thickness from 16 mm to 200 mm. There are a total of
146 specimens, of which 125 specimens have equal main
plate and attachment thicknesses. The stress ranges vary

1.00E+04

1.00E+05

1.00E+06

1.00E+07

1.00E+08

1.00E+

Figure 2: F-curves with thickness correction and

test data (plate thickness 25 mm).
1000

ABS [1]
HSE [3]
Stress Range (MPa)

Parameters
k

1.00E+03

Test Data [4]

HSE [3] 16mm
100

10
1.00E+03

1.00E+04

1.00E+05

1.00E+06

1.00E+07

1.00E+08

1.00E+

Figure 3: F-curves with thickness correction

and test data (plate thickness 50 mm)

Sensitivity of Fatigue Assessment to the Use of Different Reference S-N Curves

143

ABS TECHNICAL PAPERS 2003

1000

ABS [1]

Stress Range (MPa)

HSE [3]
Test Data
HSE [3] 16mm
100

10
1.00E+03

1.00E+04

1.00E+05

1.00E+06

1.00E+07

1.00E+08

1.00E+0

Figure 4: F-curves with thickness correction and

test data (plate thickness 75 mm)
The following formula can be employed to correct
the stress range data measured at various plate
thicknesses to the one at the basic thickness:

S B = S /(t B / t ) k

(5)

This formula is identical to that in Equation (3). The

parameters used in the corrections are listed in Table 2.
For a different viewpoint, the correction of Equation
(5) is applied to the data and then compared to the basic
curves (without the thickness correction).
In this analysis, only data for specimens with equal
main plate and attachment thicknesses were included.
The data with fatigue life longer than 107 cycles were
also excluded due to the small number of data which is
not enough to regress the curve segment for N > 107.
With the corrected data, quasi-design S-N curves can be
produced. These curves are constructed by taking the
least square line and shifting it two standard deviations
(on a log basis) to the left. The quasi-design S-N curves
and the basic F-Curves, without thickness corrections, are
plotted in Figure 5 for comparison. Figure 5 shows that
there are relatively high safety margins between the
regressed S-N curves and design curves, with the HSE
(1995) curve [3] having the largest margin.
1000

ABS [1] F-Curve without Thickness Correction

Stress Range (MPa)

HSE [3] F-Curve without Thickness Correction

Regressed S-N Curve with HSE Thickness
Correction
Regressed S-N Curve with ABS Thickness
Correction

100

In reviewing the background document [4] that

supports HSE Fatigue Guidance [3], it is found that with
the thickness correction of [3], all test data locate above P
curve (i.e. D-curve in ABS [1] and DEn [2]), while the
test specimens are as-welded T-butt and cruciform joints
which belong to the F-curve joint classification. This gap
indicates that thickness correction formula in [3] is very
conservative. Therefore, in recognition of possible
excessive conservatism for particular joints, a clause is
included in [3] so that alternative corrections may be
used if they are supported by results from experiments or
from fracture mechanics analyses. Based on Figure 5, it is
found that the use of the F curve for this detail with
reference thickness 16mm is conservative. It is therefore
proposed by ABS [1] that the reference thickness of 22
mm is applied, together with the exponent of k (= 0.25).
FATIGUE ASSESSMENT - ABS S-N CURVES VS.
AWS S-N CURVES
For non-tubular connections, API RP 2A [5] cites the
ANSI/AWS D1.1-92 [6] S-N curves. The S-N curves in
the latest AWS (2002) [7] document are essentially the
same as those in [6]. The AWS and ABS in-air curves are
compared in this section, as shown in Figure 6. However,
the comparison is not exact. Observations that contrast
the two are:
Although both codes have eight classes or
categories of plate joint types, there are
differences in the definition of the detail
category.
ABS specifies a thickness correction and there is
no thickness correction in the AWS
requirements.
Overall there is no direct correspondence of
categories. Therefore, Figure 6 does not portray
a one to one match between S-N curves defined
in ABS [1] and AWS [7].
The fatigue life for a variety of details calculated by
the simplified method, is presented next. The basic
assumptions employed in the so called simplified fatigue
analysis procedure are: a) a linear cumulative damage
model, (i.e., Palmgren-Miners Rule) is used in
connection with the S-N curve, b) long-term stress ranges
on a detail can be characterized by a Weibull probability
distribution parameter, h.

10
1.00E+03

1.00E+04

1.00E+05

1.00E+06

1.00E+07

1.00E+08

1.00E+0

Figure 5: Regressed S-N curves and design F-curves

144

Sensitivity of Fatigue Assessment to the Use of Different Reference S-N Curves

ABS TECHNICAL PAPERS 2003

1000

1000

B (ABS [1])
C (ABS)
D (ABS)

C (ABS [1])

Fatigue Life (year)

F2 ABS)
G (ABS)
W (ABS)

Stress Range (MPa)

Stress Range (MPa)

E (ABS)
F (ABS)

100

A (AWS [7])
B (AWS)

100

h=0.75
_____________________
t =22mm t=30mm
ABS C
20
16
32
AWS B 32
10

B' (AWS)
C (AWS)
D (AWS)
E (AWS)
E' (AWS)
10
1.00E+04

F (AWS)
1.00E+05

1.00E+06

1.00E+07

h=1.0
_____________________
t =22mm t=30mm
ABS C
20
16
AWS B 35
35

1.00E+04

1.00E+0

1.00E+05

B (AWS [7])

1.00E+06

In this comparison study, two Weibull shape

parameters, 0.75 and 1.0, and two plate thicknesses, 22
mm and 30 mm, are considered. For example, in Figure
7, the structural detail is classified as C curve according
to ABS [1] and B curve according to AWS [7]. Figure 7
also shows these two curves. For the case of Weibull
parameter being 0.75, assume that the fatigue stress
range at this detail with a plate thickness of 22 mm
produces a fatigue life of 20 years based on the C curve
from ABS [1], the estimated fatigue life is 32 years when
using B curve from AWS [7]. If, however, the plate
thickness increases to 30 mm, by assuming the same
fatigue stress range at this detail, the predicted fatigue
life will be 16 and 32 years according to ABS [1] and
AWS [7], respectively. The thickness correction
introduced in ABS [1] reduces the fatigue life for thicker
plate. But in AWS [7], as mentioned above, the thickness
effect is ignored. It should be noted that all of these
comparisons do not include additional factor of safety.

1.00E+09

1000

B (ABS [1])

Fatigue Life (year)

100

Stress Range (MPa)

Figures 7 16 present comparisons for 10 selected

structural details. There details are respectively:
continuous welds essentially parallel to the direction of
stress-butt or fillet welds with no start/stop; continuous
welds essentially parallel to the direction of stress-full
penetration butt welds; continuous welds essentially
parallel to the direction of stress-butt or fillet welds with
start/stop; as welded transverse butt welds in plates full
penetration; transverse butt welds in plates full
penetration with backing strip (tack welds) outside
groove); transverse butt welds in plates full penetration
with backing strip (tack welds) inside groove); transverse
full penetration butt welds in plates (perpendicular to the
direction of stress); welded attachments on the surface or
edge of a stressed member bevel butt or fillet welded
(attachment length > 150 mm); welded attachments on
the surface or edge of a stressed member bevel butt or
fillet welded (attachment length < 150 mm); loading
carrying transverse fillet and T butt welds (toe cracking).

1.00E+08

Figure 7: Continuous welds essentially parallel to the

direction of stress-butt or fillet welds with no start/stop

h=0.75
_____________________
t=22mm t=30mm
ABS B
20
16
AWS B 18
18
10

h=1.0
_____________________
t=22mm t=30mm
ABS B
20
16
AWS B 18
18

1.00E+04

1.00E+05

B (AWS [7])

1.00E+06

1.00E+07

1.00E+08

1.00E+09

Figure 8: Continuous welds essentially parallel to the

direction of stress-full penetration butt welds

1000

D (ABS [1])
E (AWS [7])

Stress Range (MPa)

Figure 6: S-N curves for non-tubular joints in air

1.00E+07

100

Fatigue Life (year)

h=0.75
___________________
t =22mm t=30mm
ABS D
20
16
AWS E
5
5

10

h=1.0
_____________________
t =22mm t=30mm
ABS D
20
16
AWS E
6
6
1
1.00E+04

1.00E+05

1.00E+06

1.00E+07

1.00E+08

1.00E+

Figure 9: Continuous welds essentially parallel to the

direction of stress-butt or fillet welds with start/stop

Sensitivity of Fatigue Assessment to the Use of Different Reference S-N Curves

145

ABS TECHNICAL PAPERS 2003

1000

1000

D (AWS [7], 50mm < R < 150mm)

C (ABS [1])

E (AWS [7], R < 50mm)

Stress Range (MPa)

Stress Range (MPa)

h=0.75
_____________________
t =22mm t=30mm
ABS C
20
16
AWS C 10
10
10

h=1.0
_____________________
t =22mm t=30mm
ABS C
20
16
AWS C 11
11

1.00E+04

F2 (ABS [1], R > 1.25t)

Fatigue Life (year)

Fatigue Life
100

1.00E+05

100

10

C (AWS [7])
1
1.00E+06

1.00E+07

1.00E+08

1.00E+09

h=0.75
_____________________
t =22mm t=30mm
ABS F2
20
16
44
44
AWS D
AWS E
19
19
h=1.0
_____________________
t =22mm t=30mm
ABS F2
20
16
AWS D
49
49
AWS E
20
20

1.00E+04

1.00E+05

Ground smooth
1.00E+06

1.00E+07

Figure 10: As welded transverse butt welds

in plates full penetration

1000

E (AWS [7])

F2 (ABS [1], Attachment length > 150mm, d > 10mm))

E (AWS [7])

Fatigue Life (year)

Stress Range (MPa)

Stress Range (MPa)

G (ABS [1])

10

1.00E+09

Figure 13: Transverse full penetration butt welds

in plates (perpendicular to the direction of stress)

1000

100

1.00E+08

h=0.75
_____________________
t =22mm t=30mm
ABS G
20
16
AWS E
33
33
h=1.0
_____________________
t =22mm t=30mm
ABS G
20
16
AWS E
35
35

1.00E+04

1.00E+05

100

10

1
1.00E+06

1.00E+07

1.00E+08

Fatigue Life (year)

h=0.75
_____________________
t =22mm t=30mm
ABS F2
20
16
19
19
AWS E
h=1.0
_____________________
t =22mm t=30mm
ABS F2
20
16
20
20
AWS E

1.00E+04

1.00E+09

1.00E+05

1.00E+06

1.00E+07

1.00E+08

1.00E+0

Figure 11: Transverse butt welds in plates

full penetration with backing strip (tack welds)
outside groove)

Figure 14: Welded attachments on the surface

or edge of a stressed member bevel butt or
fillet welded (attachment length > 150 mm)

1000

1000

D (AWS [7])
F (ABS [1], Attachment length < 150mm, d > 10mm))

100

Fatigue Life (year)

Fatigue Life (year)

Stress Range (MPa)

Stress Range (MPa)

F (ABS [1])

h=0.75
_____________________
t =22mm t=30mm
ABS F
20
16
30
AWS D 30
10

h=1.0
_____________________
t =22mm t=30mm
ABS F
20
16
AWS D 34
34
1.00E+05

1.00E+06

1.00E+07

1.00E+08

1.00E+09

10

h=1.0
_____________________
t =22mm t=30mm
ABS F
20
16
AWS C 78
78

C (AWS [7])

1.00E+04

1.00E+05

1.00E+06

1.00E+07

1.00E+08

1.00E+09

Figure 12: Transverse butt welds in plates full

penetration with backing strip (tack welds) inside groove)

146

h=0.75
_____________________
t =22mm t=30mm
20
16
ABS F
AWS C 65
65

1
1.00E+04

100

Figure 15: Welded attachments on the surface or

edge of a stressed member bevel butt or
fillet welded (attachment length < 150 mm)

Sensitivity of Fatigue Assessment to the Use of Different Reference S-N Curves

ABS TECHNICAL PAPERS 2003

1000

F (ABS [1], full penetration)

F2 (ABS [1], patial penetration/fillet)
Stress Range (MPa)

Fatigue Life (year)

100

10

h=0.75
_____________________
t =22mm t=30mm
20
16
ABS F2
DEn F
29
23
AWS C
96
96
h=1.0
_____________________
t =22mm t=30mm
ABS F2
20
16
29
23
DEn F
AWS C 114
114

1.00E+04

1.00E+05

C (AWS [7])
1.00E+06

1.00E+07

1.00E+08

1.00E+

When the cognizant governmental authority

mandates the use of such standards, these will be
considered for use in ABSs classification of structures.
However, the designer is naturally cautioned that use
must be made of the entire fatigue assessment procedure
including: how the stress range producing loads are to be
obtained, requirement for increased inspection & NDT
during fabrication, added factors of safety, etc. It is
inappropriate to unjustifiably mix elements of different
fatigue assessment procedures.

Figure 16: Loading carrying transverse fillet

and T butt welds (toe cracking)
By reviewing Figures 7 16, it is difficult to
conclude which set of S-N curves is more conservative.
Actually, the relative conservatism of the two sets of
curves varies with detail class or category. For some
detail classes, e.g. for full penetration butt welds of
continuous welds essentially parallel to the direction of
stress, the predicted fatigue lives are very close to each
other. But for some other detail classes, e.g. loading
carrying transverse fillet and T butt welds, the fatigue
lives predicted with the AWS curves can reach more than
seven times those based on ABS curve, when the input
stress range is the same corresponding to 20-year fatigue
life by ABS curve.

ACKNOWLEDGMENTS
Authors would like to acknowledge the valuable
comments from John F. Conlon, ABS Consultant,
Professor Paul H. Wirsching, University of Arizona, and
Dr. Richard Yee from Nuclear Safety Solutions Ltd.
REFERENCES
1.
2.
3.

CONCLUSIONS
This paper illustrates some of the differences in
fatigue assessment results which can occur when using SN curves recommended by different sources. The recent
ABS S-N curves were produced based on reanalysis of
data used to establish the curves issued in the UK. An
adjustment to the UK S-N data related to thickness has
been modified in the ABS proposal, and this has been
mentioned in this paper.
The UK based curves are wildly employed and may
eventually be the basis of the future ISO criteria
applicable to fixed offshore structures. On the other hand,
the use of API (and AWS) fatigue criteria is expected for
offshore hydrocarbon production structure sited in the US
continental shelf.

4.
5.
6.
7.

American Bureau of Shipping (to be published),

Guidance on the Fatigue Assessment of Offshore
Structures.
United Kingdom Department of Energy (1990),
Offshore Installations: Guidance on Design and
Construction, Fourth Edition, HMSO
Health & Safety Executive (1995), Offshore
Installations: Guidance on Design, Construction and
Certification, Third Amendment to Fourth Edition,
London
Health & Safety Executive (1992), Fatigue
Background Guidance Document, Report OTH 92
390, London
The American Petroleum Institute (2001), API
Recommended Practice 2A-WSD (latest edition)
American Welding Society (1992), Structural
Welding Code Steel, American Welding Society
specification ANSI/AWS D1.1
American Welding Society (2002), Structural
Welding CodeSteel, American Welding Society
specification ANSI/AWS D1.1M, 18th Edition.

Sensitivity of Fatigue Assessment to the Use of Different Reference S-N Curves

147

ABS TECHNICAL PAPERS 2003

148

Sensitivity of Fatigue Assessment to the Use of Different Reference S-N Curves