Fatigue Assessment of Offshore Structures Commentary On The Guide For The

Commentary on the Guide for the Fatigue Assessment of Offshore Structures (April 2003)
COMMENTARY ON THE GUIDE FOR THE
FATIGUE ASSESSMENT OF OFFSHORE STRUCTURES

(APRIL 2003)
JANUARY 2004 (Updated July 2014 see next page)
American Bureau of Shipping

Incorporated by Act of Legislature of
the State of New York 1862
Copyright 2004
American Bureau of Shipping
ABS Plaza
16855 Northchase Drive
Houston, TX 77060 USA
Updates
July 2014 consolidation includes:
February 2013 version plus Corrigenda/Editorials
February 2013 consolidation includes:
January 2004 version plus Notice No. 1
April 2010 consolidation includes:
June 2007 version plus Corrigenda/Editorials
June 2007 consolidation includes:
June 2007 Corrigenda/Editorials
Foreword
Foreword
This Commentary provides background, including source and additional technical details, for the ABS
Guide for the Fatigue Assessment of Offshore Structures, April 2003, which is referred to herein as the
Guide. The criteria contained in the Guide are necessarily brief in order to give clear descriptions of the
fatigue assessment process. This Commentary allows the presentation of supplementary information to
better explain the basis and intent of the criteria that are used in the fatigue assessment process.
It should be understood that the Commentary is applicable only to the indicated version of the Guide. The
order of presentation of the material in this Commentary generally follows that of the Guide. The major
topics of the Sections in both the Guide and Commentary are the same, but the detailed contents of the
individual Subsections and Paragraphs will not typically correspond between the Guide and the Commentary.
In case of a conflict between anything presented herein and the ABS Rules or the Guide, precedence is
given to the Rules or the Guide. This Commentary shall not be considered as being more authoritative than
the Guide to which it refers.
ABS welcomes comments and suggestions for improvement of this Commentary. Comments or suggestions
can be sent electronically to [email protected].
ABS COMMENTARY ON THE GUIDE FOR THE FATIGUE ASSESSMENT OF OFFSHORE STRUCTURES . 2004
iii
Table of Contents
COMMENTARY ON THE GUIDE FOR THE
FATIGUE ASSESSMENT OF OFFSHORE STRUCTURES

(APRIL 2003)
CONTENTS
SECTION 1
SECTION 2
SECTION 3
iv
Introduction ............................................................................................ 1
1
General Comments ............................................................................. 1
2
Basic Terminology .............................................................................. 1
3
The Deterministic Method and the Palmgren-Miner Rule to Define
Fatigue Damage ................................................................................. 1
4
Application of the Palmgren-Miner (PM) Rule ....................................2
5
Safety Checking with Respect to Fatigue ........................................... 3
TABLE 1
Deterministic Stress Spectra ..................................................... 2
TABLE 2
Tubular Joints: Statistics on Damage at Failure,

( Lognormal Distribution Assumed) ......................................... 2
TABLE 3
Plated Joints: Statistics on Damage at Failure,

( Lognormal Distribution Assumed) ......................................... 3
Fatigue Strength Based on S-N Curves General Concepts ............. 4

1
Preliminary Comments........................................................................ 4
2
Statistical Analysis of S-N Data .......................................................... 5
3
The Design Curve ............................................................................... 5
4
The Endurance Range ........................................................................ 6
5
Stress Concentration Factors Tubular Intersections ....................... 7
TABLE 1
Details of the Basic In-Air S-N Curves....................................6
FIGURE 1
An Example of S-N Fatigue Data Showing the Least

Squares Line and the Design Line [HSE(1995)] ....................... 4
FIGURE 2
The Design S-N Curve for the ABS-(A) Class D Joint .............. 7
FIGURE 3
Weld Toe Extrapolation Points for a Tubular Joint ................... 8
S-N Curves .............................................................................................. 9

1
Introduction ......................................................................................... 9
2
A Digest of the S-N Curves Used for the Structural Details of
Offshore Structures ............................................................................. 9
3
General Comparison ......................................................................... 10
Tubular Intersection Connections ..................................................... 11

4.1
Without Weld Profile Control ......................................................... 11
4.2
With Weld Improvement ................................................................ 12
Plated Connections ........................................................................... 13
Discussion of the Thickness Effect ................................................... 14
6.1
Introduction.................................................................................... 14
6.2
Fatigue Test Data on Plated Joints ............................................... 15
6.3
Design F-Curves with Thickness Adjustment ................................ 15
6.4
Thickness Adjustments to Test Data and Their Regressed

S-N Curves .................................................................................... 15
6.5
Discussion ..................................................................................... 16
6.6
Postscript....................................................................................... 16
Effects of Corrosion on Fatigue Strength.......................................... 28

7.1
Preliminary Remarks ..................................................................... 28
7.2
A Summary of the Results ............................................................. 28
7.3
The Summaries ............................................................................. 28
TABLE 1
Coverage of the Two Main Sources of S-N Curves Used

for Offshore Structures ............................................................ 11
TABLE 2
AWS-HSE/DEn Curves for Similar Detail Classes ................. 13
TABLE 3
Parameters of Plate Thickness Adjustment for Plated

Joints ....................................................................................... 14
TABLE 4
Parameters of Plate Thickness Adjustment for Tubular

Joints ....................................................................................... 15
TABLE 5
Parameters of F-curves .......................................................... 15
TABLE 6
Details of Basic Design S-N Curves HSE(1995) .................... 29
TABLE 7
Life Reduction Factors to be Applied to the Lower Cycle

Segment of the Design S-N HSE Curves ............................... 29
TABLE 8
Life Reduction Factors to be Applied to the Lower Segment

of the Design S-N DNV Curves............................................... 30
FIGURE 1
API, DEn, and ABS S-N design Curves for Tubular Joints;
Effective Cathodic Protection; No Profile Control
Specified ................................................................................. 12
FIGURE 2
F-Curves with Thickness Adjustment and Test Data;

16 mm Plate ............................................................................ 17
FIGURE 3

20 mm Plate ............................................................................ 17
FIGURE 4

22 mm Plate ............................................................................ 18
FIGURE 5

25 mm Plate ............................................................................ 18
FIGURE 6

26 mm Plate ............................................................................ 19
FIGURE 7

38 mm Plate ............................................................................ 19
FIGURE 8

40 mm Plate ............................................................................ 20
FIGURE 9

50 mm Plate ............................................................................ 20
FIGURE 10

52 mm Plate ............................................................................ 21
FIGURE 11

70 mm Plate ............................................................................ 21
FIGURE 12

75 mm Plate ............................................................................ 22
FIGURE 13

78 mm Plate ............................................................................ 22
FIGURE 14

80 mm Plate ............................................................................ 23
FIGURE 15

100 mm Plate .......................................................................... 23
FIGURE 16

103 mm Plate .......................................................................... 24
FIGURE 17

150 mm Plate .......................................................................... 24
FIGURE 18

160 mm Plate .......................................................................... 25
FIGURE 19

200 mm Plate .......................................................................... 25
FIGURE 20
Test data with DEn(1990) Thickness Adjustment and their

Regressed S-N Curves (All Thicknesses) .............................. 26
FIGURE 21
Test Data with HSE(1995) Thickness Adjustment and their

FIGURE 22
Test Data with DNV(2000) Thickness Adjustment and their

FIGURE 23
Regressed S-N Curves and Design F-curves ......................... 27
SECTION 4
Fatigue Design Factors ........................................................................ 31

1
Preliminary Remarks ......................................................................... 31
2
The Safety Check Expression........................................................... 31
3
Summaries of FDFs Specified by Others ......................................... 32
SECTION 5
The Simplified Fatigue Assessment Method ..................................... 34

1
Introduction ....................................................................................... 34
2
The Weibull Distribution for Long Term Stress Ranges ................... 34
vi
2.1
Definition of the Weibull Distribution .............................................. 34
2.2
A Modified Form of the Weibull Distribution for Offshore

Structural Analysis ......................................................................... 35
Typical Values of the Weibull Shape Parameter for Stress ........... 35

3.1
Experience with Offshore Structures ............................................. 35
3.2
Experience with Ships ................................................................... 36
Fatigue Damage: General.................................................................36

4.1
4.2
General Expression for Fatigue Damage ....................................... 36
5.1
Expression for Damage at Life, NR ................................................ 37
5.2
Miners Stress ................................................................................ 38
5.3
The Damage Expression for Weibull Distribution of Stress

Ranges .......................................................................................... 38
Fatigue Damage for Bilinear S-N Curve ........................................... 38
Safety Check Using Allowable Stress Range ................................... 40
The Simplified Method for Which Stress is a Function of Wave

Height ................................................................................................ 40
SECTION 6
Fatigue Damage for Single Segment S-N Curve .............................. 37
8.1
The Weibull Model for Stress Range; Stress as a Function of

Wave Height .................................................................................. 40
8.2
The Weibull Model for Stress Range; Stress as a Function of

Wave Height; Considering Two Wave Climates ............................ 41
The Weibull Distribution; Statistical Considerations ......................... 42

9.1
9.2
Estimating the Parameters from Long-Term Data; Method of

Moment Estimators ....................................................................... 42
9.3
Estimating the Parameters from Long-Term Data; Probability

Plotting .......................................................................................... 42
9.4
Another Representation of the Weibull Distribution Function ........ 45
9.5
Fitting the Weibull to Deterministic Spectra ................................... 46
9.6
Fitting the Weibull Distribution to the Spectral Method .................. 47
TABLE 1
Data Analysis for Weibull Plot ................................................. 43
TABLE 2
Deterministic Spectra .............................................................. 46
FIGURE 1
A Short Term Realization of a Long-Term Stress Record ...... 34
FIGURE 2
Probability Density Function of s ............................................. 36
FIGURE 3
Characteristic S-N curve ......................................................... 37
FIGURE 4
Bilinear Characteristic S-N curve ............................................ 39
FIGURE 5
Weibull Probability Plot ........................................................... 44
FIGURE 6
Long Term Distribution of Fatigue Stress as a Function of

the Weibull Shape Parameter ................................................. 45
FIGURE 7
Long-Term Stress Range Distribution of Large Tankers,

Bulk Carriers, and Dry Cargo Vessels Compared with
the Weibull .............................................................................. 46
FIGURE 8
Probability Density Function of Stress Ranges of the i-th

Sea State ................................................................................ 47
The Spectral Based Fatigue Assessment Method ............................ 49

1
Preliminary Comments...................................................................... 49
2
Basic Assumptions............................................................................ 49
3
The Rayleigh Distribution for Short Term Stress Ranges................. 50
4
Spectral Analysis; More Detail .......................................................... 51
5
Wave Data ........................................................................................ 51
6
Additional Detail on Fatigue Stress Analysis; Global Performance
Analysis ............................................................................................. 52
vii
SECTION 7
SECTION 8
viii
7.1
General Considerations ................................................................. 53
7.2
The Stress Process in Each Cell ................................................... 53
Fatigue Damage Expression for Wide Band Stress ......................... 54

8.1
Preliminary Comments .................................................................. 54
8.2
Definitions ...................................................................................... 55
8.3
The Equivalent Narrow Band Process ........................................... 56
8.4
The Rainflow Method ..................................................................... 56
8.5
A Closed Form Expression for Wide Band Damage ...................... 57
The Damage Calculation for Single Segment S-N Curve ................. 58
10
The Damage Calculation for Bi-Linear S-N Curve ............................ 59
TABLE 1
A Sample Wave Scatter Diagram ........................................... 52
FIGURE 1
Fatigue Assessments by Spectral Analysis Method ............... 50
FIGURE 2
Realizations of a Narrow Band and Wide Band Process

(Both Have the Same RMS and Rate of Zero Crossings) ...... 55
FIGURE 3
Segment of Stress Process to Demonstrate Rainflow

Method .................................................................................... 56
Deterministic Method of Fatigue Assessment ................................... 61

1
General ............................................................................................. 61
2
Application to a Self-Elevating Unit ................................................... 61
TABLE 1
Deterministic Stress Spectra ................................................... 61
TABLE 2
Wave and Other Parameters to be Used in the Fatigue

Assessment............................................................................. 62
Fracture Mechanics Fatigue Model ..................................................... 63

1
Introduction ....................................................................................... 63
2
Crack Growth Model (Fatigue Strength) ........................................... 63
SECTION 9
The Safety Check Process ............................................................... 53
2.1
Stress Intensity Factor Range ....................................................... 63
2.2
The Paris Law ................................................................................ 63
2.3
Determination of the Paris Parameters, C and m ........................... 64
Life Prediction ................................................................................... 65

3.1
Relationship Between Cycles and Crack Depth ............................ 65
3.2
Determination of Initial Crack Size, ai ............................................ 65
3.3
Determination of the Failure (Critical) Crack Length, ac. ................ 66
TABLE 1
Paris Parameters for Structural Steel .....................................65
FIGURE 1
A Model of Crack Propagation Rate versus Stress Intensity

Factor Range .......................................................................... 64
References ............................................................................................ 67
Section 1: Introduction
SECTION
Introduction
General Comments
For over a half century, ABS has been involved in the development of fatigue technology, starting in 1946
with the formation of the Ship Structure Committee (SSC) for the specific goal of addressing avoidance of
serious fracture in ships. The SSC, with strong financial support from ABS, has executed several fatigue
research projects. Over the years, ABS has also provided support to numerous joint industry/agency
fatigue projects in addition to independent investigators for their own in-house projects.
The current state of the art in fatigue technology represents worldwide contributions of a large numbers of
investigators from government agencies, professional organizations, classification societies, universities
and private industry, most notably petroleum companies. ABS has synthesized this body of knowledge to
provide fatigue design criteria for marine structures. This document provides a review of the most relevant
literature, describes how ABS criteria were established and compares ABS criteria with those of other
organizations.
Because welded joints are subject to a variety of flaws, it is generally expected that fatigue cracks will start
first at the joints. Therefore, the focus of this document will be on the joints, but the general principles and
some of the fatigue strength data will apply to the base material.
Basic Terminology
NT (or T)
Design life; the intended service life of the structure in cycles (or time)
Nf (or Tf)
Calculated fatigue life; the computed life in cycles (or time) of the structure using the
design S-N curve
fatigue damage at the design life of the structure
maximum allowable fatigue damage at the design life of the structure
FDF
fatigue design factor; FDF 1.0
The FDF accounts for:
i)
Uncertainty in the fatigue life estimation process
ii)
Consequences of failure (i.e., criticality)
iii)
Difficulty of inspection
The Deterministic Method and the Palmgren-Miner Rule to Define

Fatigue Damage
Fatigue assessment in the Guide relies on the characteristic S-N curve to define fatigue strength under constant
amplitude stress and a linear damage accumulation rule (Palmgren-Miner) to define fatigue strength under
variable amplitude stress.
Fatigue stress is a random process. Stress ranges in the long-term process form a sequence of dependent
random variables, Si; i = 1, NT. For purposes of fatigue analysis and design, it is assumed that Si are
mutually independent. The set of Si can be decomposed and discretized into J blocks of constant amplitude
stress, as illustrated in Section 1, Table 1.
Section
Introduction
TABLE 1
Deterministic Stress Spectra
Stress Range
Si
Number of Cycles
ni
S1
n1
S2
n2
S3
.
.
n3
SJ-1
nJ-1
SJ
nJ
Applying the Palmgren-Miner linear cumulative damage hypothesis to the block loading of Section 1,
Table 1, cumulative fatigue damage, D, is defined as:
D=
ni
N
i =1
................................................................................................................................. (1.1)
where Ni is the number of cycles to failure at stress range Si, as determined by the appropriate S-N curve.
Failure is then said to occur if:
D > 1.0......................................................................................................................................... (1.2)
Application of the Palmgren-Miner (PM) Rule

The PM rule is a simple algorithm for predicting an extremely complex phenomenon (i.e., fatigue under
random stress processes). Results of tests, however, have suggested that the PM rule is a reasonable
engineering tool for predicting fatigue in welded joints subjected to random loading.
Statistical summaries of random fatigue tests have been reported by the UK Health and Safety Executive
[HSE(1995)]. Let be a random variable denoting damage at failure and let i denote damage at failure in
a test of the i-th specimen in a sample of size, n. i will depend on how the constant amplitude S-N curve
is defined (e.g., as a median (best fit) curve through the center of the data or a design curve on the safe side
(lower) of the data). The sample mean and standard deviation of can be computed from the random
sample (i ; i = 1, n). An empirical distribution can be fitted as well.
A limited number of tests on tubular joints is available. In HSE(1995), a lognormal distribution is assumed
for . Statistics computed from the data presented are summarized in Section 1, Table 2. It is noted that
the scatter is quite broad, and it is likely that the wide distribution is largely a result of the inherent scatter
in fatigue data and not the suitability of the PM algorithm. For reference purposes, the probability of
being less than the reference curve is also presented in Section 1, Table 2.
TABLE 2
Tubular Joints: Statistics on Damage at Failure,
( Lognormal Distribution Assumed)
Best fit curve
Design curve
~
Median,
COV, C
1.41
4.42
0.98
0.98
Percent less than S-N curve

34
3.5
Section
Introduction
For plated joints *, there is a relatively large database. Again, a lognormal distribution for is assumed,
and the statistics are presented in Section 1, Table 3.
TABLE 3
Plated Joints: Statistics on Damage at Failure,
( Lognormal Distribution Assumed)
Best fit curve
Design curve
~
Median,
COV, C
Percent less than S-N curve
1.38
4.44
0.70
0.70
33
1.5
Safety Checking with Respect to Fatigue

The safety check expression can be based on damage or life. While the damage approach is featured in the
Guide, either approach below can be used.
Damage
The design is considered to be safe if:
D .......................................................................................................................................... (1.3)
where
= 1.0/FDF ............................................................................................................................... (1.4)

Life
Nf NT FDF.............................................................................................................................. (1.5)
* Note:
In the Guide, to conform to practice, the two general categories of structural details are referred to as tubular (really
meaning tubular intersection) details and non-tubular details. In the context of the HSE (1995), the non-tubular details
are referred to as plate or plate type details. The plate terminology will be used in this Commentary.
Section 2: Fatigue Strength Based on S-N Curves General Concepts
SECTION
Fatigue Strength Based on S-N Curves General

Concepts
Preliminary Comments
This Section introduces general concepts related to the S-N curve-based method of fatigue assessment.
The next Section contains detailed information regarding S-N curves.
For the stress-based approach to fatigue, the S-N curve defines fatigue strength. An example of S-N data
and a design curve are shown in Section 2, Figure 1. Each point represents the cycles to failure N of a specimen
subjected to constant range stress S. Log(N) is plotted versus Log(S). Section 2, Figure 1 presents the results
of fatigue tests on tubular joints where failure is defined as first through wall cracking.
FIGURE 1
An Example of S-N Fatigue Data Showing the Least Squares Line
and the Design Line [HSE(1995)]
1000
Hot Spot Stress Range, S (MPa)
Least
Squares
Line
++ +
+
+ ++ +++ ++
+++ +++ ++
+
Design
+++++++ + + +
++++++ + +
Line
++
+
100
++++
+ +
+
+ ++ +
Best Fit S-N Line Through 16 mm Data

Design Line for 16 mm Data
Experimental Data for 16 mm Thick Tubular Joints
10
10 000
100 000
1 000 000
107
108
Fatigue Endurance, N (Cycles)
A design curve is defined on the safe (lower) side of the data. Note that an implicit fatigue design factor is
thereby introduced. For purposes of safety checking, the design S-N curve defines fatigue strength, but
one should keep in mind that there is a large statistical scatter in fatigue data (relative to other structural
design factors) with cycles-to-failure data often spanning more than two orders of magnitude.
Section
Fatigue Strength Based on S-N Curves General Concepts
Statistical Analysis of S-N Data

The design curve is established as follows: First, it is noted that when S-N data are plotted in a log-log
space, the data tend to plot as a straight line, as suggested in Section 2, Figure 1. A linear model can be
employed, the form of which is:
log(N) = log(A) m log(S)........................................................................................................... (2.1)
Base 10 logarithms are generally used. A and m are empirical constants to be determined from the data. A
is called the fatigue strength coefficient and m is called the fatigue strength exponent. The parameter m is
the negative reciprocal slope of the S-N curve, but for convenience, it is often referred to simply as the slope.
Another component of the model is the standard deviation of N given S, denoted as (N|S), or simply, .
This parameter describes the scatter in life.
To estimate A, m and , the least squares method can be employed, thus providing parameters (A and m) to
define the median S-N curve (i.e., a curve that passes through the center of the data). Note that S is the
independent variable and N is the dependent variable. It is assumed that log(N) has a normal distribution,
which means that N will have a lognormal distribution.
For many welded joint fatigue data, the parameter m is approximately equal to 3.0. Therefore, for convenience
and consistency, a fixed value of m = 3 is assumed and least squares analysis is then employed to estimate
A and . Let A and denote the estimates. For the sample data of Section 2, Figure 1:
m=3
log(A) = 12.942
= 0.233
The coefficient of variation (standard deviation/mean) of cycle life N is required for a reliability analysis.
The form for the COV is:
CN = 10 (
/ 0.434 )
1 ................................................................................................................ (2.2)
For the example:

CN = 0.58, or 58%
The Design Curve

The design S-N curve is defined as the median curve minus two standard deviations on a log basis.
Thus, the basic S-N curves are of the form:
log(N) = log(A) m log(S)
where
log(A) =
log(A1) 2
predicted number of cycles to failure under stress range S
A1
constant relating to the mean S-N curve
standard deviation of log N
inverse slope of the S-N curve
The relevant values of these terms are shown in the table below for the ABS In-Air S-N curves for platetype (non-tubular) details.
The in-air S-N curves have a change of inverse slope from m to m + 2 at N = 107 cycles.
Section
TABLE 1
Details of the Basic In-Air S-N Curves
A1
Standard Deviation
Class
B
A1
2.343 1015
log10
15.3697
loge
35.3900
m
4.0
log10
0.1821
loge
0.4194
A
1.01 1015
1.082 1014
14.0342
32.3153
3.5
0.2041
0.4700
4.23 1013
3.988 1012
12.6007
29.0144
3.0
0.2095
0.4824
1.52 1012
3.289 10
12
12.5169
28.8216
3.0
0.2509
0.5777
1.04 1012
1.726 1012
12.2370
28.1770
3.0
0.2183
0.5027
0.63 1012
F2
1.231 1012
12.0900
27.8387
3.0
0.2279
0.5248
0.43 1012
0.566 1012
11.7525
27.0614
3.0
0.1793
0.4129
0.25 1012
0.368 1012
11.5662
26.6324
3.0
0.1846
0.4251
0.16 1012
If cycles to failure were lognormally distributed, then a specimen selected at random would have a probability
of 2.3% of falling below the design curve.
There may be confusion over this probability compared to those mentioned previously in Section 1, Tables
2 and 3. Different random variables are being referred to. In Section 1, Tables 2 and 3, the random
variable is delta, the damage at failure. The statistics for delta are computed for both the best-fit curve and
the design curve. Note that the fatigue test results are based on random stresses. The title of the column in
the tables labeled, Percent less than S-N curve could have been alternatively labeled, Percent of
specimens that had lives below the S-N curve.
The basic S-N curves are established from constant amplitude tests. Assuming a lognormal distribution for
life, the design curve is that curve below which 2.3% of the specimens are expected to fall. So, random
fatigue test results are being compared to constant amplitude test results. It would not necessarily be
expected that the results would be the same, but it is gratifying to see that the results are so close.
The Endurance Range

Test data are much more limited in the range beyond 107 cycles. It appears that there may be an endurance
limit near this point (i.e., a stress below which fatigue life would be infinite). However, a more prudent
extrapolation of the S-N curve into the high cycle range involves a change in slope. For in-air structure,
the slope (actually the negative reciprocal slope) beyond 107 cycles is:
r = m + 2 ..................................................................................................................................... (2.3)
While defined by engineering judgment, this form seems to have performed well for an extended period of
time. This algorithm is used by DEn(1990) and others, but ISO(2000) specifies the knee of the curve at
108 cycles.
Section
As an example, consider the ABS-(A) class D curve.
FIGURE 2
The Design S-N Curve for the ABS-(A) Class D Joint
Stress Range, S (MPa)
1000
100
Cycles to Failure, N
10
4
10
10
10
10
10
For plate joints that are cathodically protected, HSE(1995) specifies the knee at 106 cycles. For joints
exposed to free corrosion, most organizations do not specify an endurance limit (i.e., the S-N curve is
extrapolated into the high cycle range without a change in slope).
Stress Concentration Factors Tubular Intersections

A major theme of the presentation in Section 2 of the Guide is that the fatigue assessment should employ
applicable stress concentration factors (SCFs) and the appropriate S-N curve. For a tubular joint, the S-N
curves recommended by DEn(1990)/HSE(1995) and API RP2A are meant to be used with SCFs obtained
for the hot-spot locations at the weld toe.
The SCF equations referenced in the Guides Appendix 2 are meant to have precedence. However, allowance
is made (Guide Paragraph 3/5.5) to also use, as appropriate, the parametric equations referenced in the API
RP2A when it is permitted to use the APIs tubular joint S-N curves (e.g., structure sited on the U.S. Outer
Continental Shelf, subject to US Minerals Management Service Regulation).
Where conditions are such that the recommended parametric SCF equations cannot be applied confidently,
then the SCFs can be obtained experimentally or numerically via finite element analysis. In either case, it
is necessary to have a stress extrapolation procedure to weld toe locations that is compatible with the S-N
curve. This is directly analogous to the extrapolation procedure for non-tubular details given in the Guide.
The DEn provided guidance, as shown in Section 2, Figure 3, on the specific locations where the stresses
should be obtained for extrapolation to the hot-spot locations at the weld toe.
Section
FIGURE 3
Weld Toe Extrapolation Points for a Tubular Joint
a = 0.2(rt)0.5, but not smaller than 4 mm.
r
t
Line 1.
Line 2.
Brace.
A1
0.65(rt)0.5
Line 3.
A3
A2
B1
a
B3
B2
0.65(rt)0.5
a
0.4(rtRT)0.25
5
a
B4
A4
Line 4.
Chord.
Section 3: S-N Curves
SECTION
S-N Curves
Introduction
In the offshore industry, fatigue assessment and design are based primarily on S-N curves to define strength.
These curves define the integrity of both plate-type details and tubular welded joints under oscillatory
loading. ABS has performed a comprehensive review of fatigue test results and fatigue strength models
employed for steel structural details for the purpose of defining the ABS requirements.
For the sources of design S-N curves, documents from three organizations, API (2000), AWS(2002), DEn
(1990)/HSE(1995), are commonly cited by designers and analysts in the offshore industry. Agencies and
organizations that provide structural design criteria for welded joints use these S-N curves and variations
thereof. In order to gain a perspective on current practice, a digest of the S-N curves cited in various design
criteria documents is provided in Subsection 3/2 below.
The approach used in the ABS Guide for the classification of details, the S-N curves and adjustments made
to the curves, may be referred to as a hybrid of the DEn(1990) and HSE(1995) criteria. The ABS Guide
criteria uses:
The classification of details and basic S-N curves from the DEn(1990), which is almost identical to that
found in HSE(1995) for plate-type details [a comparative description of DEn(1990) and HSE(1995) is
given below in Subsection 3/2ii)].
For plate-type details, the thickness adjustment applies when t > 22 mm using tref = 22 mm and exponent
of 0.25, and for tubular intersection details, the thickness adjustment applies when t > 22 mm using
tref = 32 mm and exponent of 0.25.
The HSE(1995) Environmental Reduction Factors (ERFs), which is akin to Corrosiveness in the
ABS Guide are for plate type details: 2.5 where effective Cathodic Protection (CP) is provided and 3.0
for Free Corrosion (FC) conditions, and for tubular intersection details, the ERFs are 2.0 for CP and
3.0 for FC conditions.
A Digest of the S-N Curves Used for the Structural Details of

Offshore Structures
i)
DEn (1990), Gurney (1979); A suite of eight curves for plated joints. Change in slope at 1E7
cycles, used successfully for many years by DEn and other criteria based on DEn
ii)
HSE(1995). Citations and comparisons to HSE and DEn criteria are difficult. The version of the
fatigue criteria contained in the DEn Guidance Notes that was issued in 1990 was labeled the 4th
Edition. It is referred to here as DEn(1990). Following DEn practice, changes to an edition were
issued as amendments to that edition. Revision of the fatigue criteria in the 4th Edition was planned
for publication in the 3rd amendment of the DEn Guidance Notes in 1995. At the same time, the
DEn was undergoing organizational change, and the HSE became its successor organization. The
document planned for release was relabeled, and is referred to here as HSE(1995). There were
changes in the details of the criteria presentation between what had been planned as the 3rd amendment
of the Guidance Notes, 4th Edition in 1995 and the superseding HSE(1995) document. However,
immediately after the HSE(1995) fatigue criteria were issued, it was withdrawn along with all of
the other DEn Guidance Notes.
Section
S-N Curves
For information, the essential features of the HSE(1995) fatigue criteria compared to DEn(1990)
are as follows.
The guidance provided on the classification of structural details and the assigned S-N curve to
each class remained the same (see Appendix 1 of the Guide showing the classifications using
the sketches of the various structural details and loading). Changes included in HSE(1995)
were added guidance related to tubular member details and a change in the W S-N curve.
Also, in the detail classification guidance (for plate type details), it was planned to replace mention
of the individual (8) S-N categories with one S-N curve, the P curve that was equivalent to
the D curve in DEn(1990). Then, the detail classes would be related to the P curve by a
classification factor.
The basic S-N curve for tubular intersection details was revised. In DEn(1990), the T curve is
close to the D. The revised HSE(1995) T curve (in air) is higher than the 1990 T curve.
However, the application of Environmental Reduction Factors (EFRs) and a revised thickness
adjustment might produce significant reductions from the basic case.
In the DEn(1990), no reduction to an (in air) S-N curve is called for when effective Cathodic
Protection is present. Based on additional testing, it was deemed necessary to include in
HSE(1995) penalties for the Cathodic Protection (CP) case and to increase the penalties for
the Free Corrosion case. For plate type details, the penalty factors are 2.5 and 3.0 for (CP)
and (FC), respectively. For tubular intersection details, the respective penalty factors were 2.0
and 3.0. (The specific details of how these are applied are discussed in Subsection 3/7.)
Another planned, significant change between HSE(1995) and DEn(1990) concerns the adjustment
to the S-N curves for thickness. The limiting thickness (above which adjustments are to be made),
and the exponent and reference thickness in the adjustment equation were all affected.
iii)
ABS (2001) Rules for Building and Classing Steel Vessels. Since the original introduction in
1994, the criteria for fatigue strength in these Rules employ the DEn (1990) curves.
iv)
Eurocode 3 (1992). Uses a suite of 14 curves, with initial segments having slopes of 3.0. Beyond
5E6 cycles, the slopes are 5.0 for the curves up to 1E8 cycles, beyond which the curves are flat
(endurance limit).
v)
IIW (1996). In general application, a suite of 14 S-N curves is presented. Each has an endurance
limit at 5E6 cycles, after which the curve is flat. For marine application to be used together with
Palmgren-Miner summation, another suite of 14 S-N curves that basically matches the Eurocode 3
curves is recommended: Beyond 5E6 cycles the curve has a slope of 5 and the curve has a cut-off
limit at 1E8. The concept of a FAT class defines the joint detail.
vi)
DNV (2000); RP-C203 for offshore structures. Uses a suite of 14 curves [as in iv) and v)] that also
incorporate the HSE(1995) curves. This reference also has S-N curves that reflect FC and CP
conditions. It also has a curve for tubular joints, in-air and for CP and FC conditions in seawater.
vii)
ISO/CD 19902 (2000). The ISO draft standard appears to be based on DEn(1990), but the basic
2-segment S-N curves have a change of slope at 1E8 cycles, which is not the same as DEn(1990).
S-N curves are also provided for tubular intersection details and cast steel tubular joints.
viii)
API (2001 a & b). RP2A (both WSD and LRFD) has S-N curves for tubular intersection joints.
Defines X and X curves for joints with and without weld profile control, respectively. Cites
ANSI/AWS D1.1- for plate joints.
ix)
API RP2T(1997). Cites RP2A for definition of S-N curves.
General Comparison
Section 3, Table 1 summarizes the characteristics of the S-N design curves of DEn(1990)/HSE(1995) and
API/AWS relative to environment, cathodic protection, and weld improvement.
10
Section
S-N Curves
TABLE 1
Coverage of the Two Main Sources of S-N Curves
Used for Offshore Structures
Detail Type
Tubular
Intersection
Non-Tubular
(Plate)
Notes:
1&2
Corrosion Condition
API (2000)
Notes 1&4
AWS D1.1
DEn(1990)
HSE(1995)
Notes 2&3
In-Air
Cathodic Protection
Free Corrosion in the Sea Water
In-Air
Cathodic Protection
Note 5
Free Corrosion in the Sea Water
Fatigue life enhancement via Weld Improvement techniques is explicitly permitted:

--in API RP2A by weld profiling
--in DEn/HSE by weld toe grinding
DEn/HSE is the basis of the ABS criteria
API RP 2A treats corrosion differently from the other codes. API RP 2A uses one curve with
different endurance limits to represent the three corrosion cases (in-air, in seawater with free
corrosion, and in seawater with cathodic protection). DEn/HSE use three curves to represent
the three cases.
While AWS does not address modification of S-N curves for CP, API RP2A specifies an
endurance limit at 2 108 cycles for plate type details.
Tubular Intersection Connections
4.1
Without Weld Profile Control

A summary of the API and HSE(1995) having no weld profile control is presented as follows.
API RP 2A(2000) uses the X curve for the following three corrosion cases with various endurance limits:
In the air, endurance limit = 2 107 cycles
Cathodic protection, endurance limit = 2 108 cycles
Free corrosion in sea water, no endurance limit
HSE(1995) defines a T curve and its derivatives for the three corrosion cases:
In-air,
Cathodic protection, (CP)
Free corrosion in sea water, (FC)
The ABS Guide specifies a T curve and recognizes three corrosion cases:
In-air, (A)
Cathodic protection, (CP)
Free corrosion in sea water, (FC)
Section 3, Figure 1 presents the S-N curves for the CP case for tubular joints for: HSE (1995) T with CP,
API RP2A the X curve, and the ABS T (CP) curve, as provided in the ABS Guide. The latter is based on
the use of the DEn(1990) T curve, which is adjusted as recommended in HSE (1995).
11
Section
S-N Curves
FIGURE 1
API, DEn, and ABS S-N design Curves for Tubular Joints;
Effective Cathodic Protection; No Profile Control Specified
4.2
With Weld Improvement (1 February 2013)

A summary of the API and HSE/DEn S-N curves for joints of tubular members having weld improvement
is presented in the following.
API RP 2A(2000) uses the X curve for the following three corrosion cases with various endurance limits:
In-air, endurance limit = 107 cycles
Cathodic protection, endurance limit = 2 108 cycles
Free corrosion in seawater, no endurance limit.
The crediting of weld profile control (i.e., concave weld profile) and other fatigue strength enhancements
are not mentioned in the Guide for use with the ABS S-N curves. The main reason for this is to discourage
(however, not ban) the use of such a credit in design. In this way, the credit will be available if needed in
the future [say, if design changes occur after structural fabrication begins and even later in the structures life
should reconditioning or reuse be considered]. Out of necessity and in a limited, particular circumstance,
the Guide (in its Appendix 3) allows the use of the API X curve, which requires weld profile control and NDE.
Grinding is preferably to be carried out by rotary burr and to extend below the plate surface in order to
remove toe defects and the ground area is to have effective corrosion protection. The treatment is to
produce a smooth concave profile at the weld toe with the depth of the depression penetrating into the plate
surface to at least 0.5 mm below the bottom of any visible undercut. The depth of groove produced is to be
kept to a minimum, and, in general, kept to a maximum of 1 mm. In no circumstances is the grinding depth
to exceed 2 mm or 7% of the plate gross thickness, whichever is smaller. Grinding has to extend to areas
well outside the highest stress region.
The finished shape of a weld surface treated by ultrasonic/hammer peening is to be smooth and all traces of
the weld toe are to be removed. Peening depth below the original surface is to be maintained at least 0.2
mm. Maximum depth is generally not to exceed 0.5 mm.
Provided these recommendations are followed, when using the ABS S-N curves, a credit of 2 on fatigue
life may be permitted when suitable toe grinding or ultrasonic/hammer peening are provided. Credit for an
alternative life enhancement measure may be granted based on the submission of a well-documented,
project-specific investigation that substantiates the claimed benefit of the technique to be used.
12
Section
S-N Curves
Plated Connections
For plated connections, API RP2A cites the ANSI/AWS D1.1-92 [AWS(1992)] S-N design curves. The S-N
curves of the newer AWS(2002) document are essentially the same as AWS(1992). The AWS and DEn
(1990) curves are compared below. Both references use sketches to help the designer in the selection of a
details classification.
The comparison is not exact. Observations that contrast the two main reference sources are:
i)
DEn has eight classes or categories of joint types. AWS has six.
ii)
DEn is more discriminating in the number of joint types or details.
iii)
There are differences in the definition of the detail category.
iv)
DEn employs a thickness adjustment (see Subsection 3/6). There is no thickness adjustment in the
AWS criteria.
v)
Except for free corrosion in seawater, AWS specifies a stress endurance limit in the high cycle range.
DEn changes to a shallower slope.
vi)
Overall, there is no direct correspondence of categories, but there are a few that are similar. These
are summarized in Section 3, Table 2.
TABLE 2
AWS-HSE/DEn Curves for Similar Detail Classes
Detail Class
Base or parent material
Full penetration butt welds, groove welds
ANSI/AWS(1992)
DEn(1990)
C (L < 50 mm)
D (50 < L < 100)
E (L > 100)
F (L < 150 mm)

F2 (L > 150 mm)
Parent material of cruciform T-joints
Load carrying fillet welds transverse to the direction of

stress (parent material)
F (d > 10 mm)
G (d < 10)
Load carrying fillet welds transverse to the direction of

stress (weld material)
Parent material at the end of butt welded attachments
For conditions of effective cathodic protection (CP):

i)
API specifies a stress endurance limit on the AWS curves at 2 108 cycles.
ii)
The DEn CP curves have a break at 106 cycles. The slope to the left is m; to the right, it is m + 2).
The DEn curves are lowered from the in-air curves by a factor of 2.5 on life, again maintaining the
break point at 106 cycles.
For conditions of free corrosion, both curves have no endurance limit or slope change in the high cycle
range (i.e., the low cycle curve with a slope of 3.0 is continued into the high cycle range). In addition, the
DEn curves are lowered by a factor of 3.0 on life.
13
Section
S-N Curves
Discussion of the Thickness Effect
6.1
Introduction
The ABS-recommended thickness adjustment (size effect) is based on studies of fatigue test data as well as
models used by others. A summary of this study is presented below.
The basic S-N design curve has the functional form:
log10 N = log10 A mlog10 S ......................................................................................................... (3.1)
where N is cycles to failure, S is stress range, and A and m are respectively, the fatigue strength coefficient
and exponent.
The size effect in fatigue in which larger sections tend to be weaker is manifest in welded joint fatigue by a
thickness adjustment. In API, HSE/DEn and other codes, the effect of plate thickness is addressed by a
similar adjustment formula:
t
Sf = S
tR
Sf = S
t > t0.................................................................................................................. (3.2)

t t0.................................................................................................................. (3.3)
where
Sf
allowable stress range,
allowable stress range from the nominal S-N design curve,
q, tR =
parameters (tR is the reference thickness),
t0
thickness above which adjustments should be made,
actual thickness.
A thickness adjusted S-N curve can be constructed when t > t0.

t
log10 (N) = log10 (A) m log S
tR
+q
.................................................................................... (3.4)
The parameters q and tR are determined empirically. For plated joints, Section 3, Table 3 summarizes these
parameter values from the references: DEn (1990), HSE (1995) and DNV (2000). (Size effect is not considered
in ANSI/AWS D1.1.)
TABLE 3
Parameters of Plate Thickness Adjustment for Plated Joints
Parameters
DEn (1990)
HSE (1995)
DNV (2000)
0.25
0.30
0.0 0.25
depending on detail
classification;
0.25 for F-curve
tR
22 mm
16 mm
25 mm
These values do not depend upon the environment (i.e., they are the same for the in-air, cathodic protection
and free corrosion curves).
The objective of this section is to compare the three parameter sets with the test data on plated joints that
were used in reviewing the thickness effect by HSE (1995) and to recommend the algorithm to be used by
ABS in the Guide.
14
Section
S-N Curves
For reference, the tubular joint parameters are also given in Section 3, Table 4.
TABLE 4
Parameters of Plate Thickness Adjustment for Tubular Joints
Parameters
API (2000, 1993)
HSE (1995)
DNV (2000)
0.25
0.30
0.25 for SCF < 10.0

0.30 for SCF > 10.0
tR
25 mm
16 mm
32 mm
6.2
Fatigue Test Data on Plated Joints

An analysis was undertaken of data from tests on as-welded T-butt and cruciform joints that belong to the
F classification [HSE(1995)]. The specimens varied in thickness from 16 mm to 200 mm. There are a total
of 146 specimens in which 125 specimens have equal main plate and attachment thickness. Stress ranges in
the tests varied from 56 MPa to 341 MPa and only four specimens had a fatigue life exceeding 107 cycles.
6.3
Design F-Curves with Thickness Adjustment

The parameters of the basic F-curves used in the three codes are shown in Section 3, Table 5. The F-curves
in DEn (1990) and HSE (1995) are identical, but with different thickness adjustment formulae. The DNV
(2000) F-curve is slightly less conservative than the other two.
TABLE 5
Parameters of F-curves
N < 107
N > 107
Codes
log10 (A)
log10 (A)
DEn (1990)
11.801
15.001
HSE (1995)
11.801
15.001
DNV (2000)
11.855
15.091
The design F-curves with thickness adjustment (Equation 3.4) are plotted in Section 3, Figures 2 through 19.
In ascending order, each curve has a different thickness. The test data for each thickness are plotted. The
HSE (1995) F-curve of 16 mm thickness (i.e., without thickness adjustment) is also plotted in figures where it
is appropriate for reference. These series of figures demonstrate the general detrimental effect of increasing
plate thickness. There exist relatively large safety margins between the test data and design curves, with
the HSE (1995) curve having the largest gap.
6.4
Thickness Adjustments to Test Data and Their Regressed S-N Curves

For a different viewpoint, the adjustment of Equation 3.2 is applied to the data and then compared to the
basic curves (without the thickness adjustment).
In this analysis, only data for specimens with equal main plate and attachment thicknesses were included
because HSE used the same strategy in their study on thickness effect. Data with fatigue lives longer than
107 cycles were also excluded due to the small sample size (i.e., insufficient data to regress the curve segment
for N > 107). With the adjusted data, quasi-design S-N curves were produced. These curves were constructed
by taking the least squares line and shifting it two standard deviations (on a log basis) to the left. The adjusted
data, (the quasi-design S-N curves,) and the basic F-curves, without thickness adjustments, are plotted
together for comparison. The results for DEn (1990), HSE (1995) and DNV (2000) are shown in Section 3,
Figures 20 through 22, respectively. The comparison across the codes is demonstrated in Section 3, Figure 23.
The conclusion stated previously is justified. There are relatively large safety margins between the regressed
S-N curves and design curves, with HSE (1995) curve having the largest margin.
15
Section
6.5
S-N Curves
Discussion
In reviewing the commentary document [HSE (1992)] that supports the HSE Fatigue Criteria [HSE (1995)],
it is found that with the thickness adjustment of HSE (1995), all test data locate above the P-curve [i.e.,
D-curve in DEn (1990)], while the test specimens were as-welded T-butt and cruciform joints that belong
to F-curve of joint classification. This gap indicates that HSE (1995) thickness adjustment formula is too
conservative. Perhaps, in recognition of the possible excessive conservatism for particular details, a clause
is included in HSE (1995) so that alternative adjustments may be used if they are supported by results from
experiments or from fracture mechanics analyses.
A statement that the basic 16 mm P-curve is equivalent to the 22 mm D-curve in DEn (1990) is found in a
commentary paper on the HSE (1995) [Stacey and Sharp (1995)]. Therefore, one may ask why it is necessary
to make a thickness adjustment to joints with a 22 mm thickness.
In a commentary paper of DNV RP-C203 [Lotsberg and Larsen (2001)], a similar study was conducted
and a conclusion is that use of the F-curve for this detail with reference thickness 16 mm is conservative.
6.6
16
Postscript
Due to the discrepancy between the thickness adjustment formulae, there is a question as to how the thickness
adjustment formula of HSE (1995) was derived. It is speculated by the authors of this Commentary that
the algorithm was obtained by borrowing the form for tubular joints, or by using a curve other than the
F-curve as the target curve for regression analysis, or perhaps using some other procedure. The origin of
the algorithm is not documented in HSE (1992). Thus, the procedure used to derive the thickness adjustment
formula of HSE (1995), particularly the choice of 16 mm as basic thickness, is not clear.
Section
S-N Curves
FIGURE 2
F-Curves with Thickness Adjustment and Test Data; 16 mm Plate
1000
DEn(1990)
HSE(1995)
DNV(2000)
Stress Range (MPa)
Test Data
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
FIGURE 3
1000
DEn(1990)
HSE(1995)
DNV(2000)
Stress Range (MPa)
Test Data
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
17
Section
S-N Curves
FIGURE 4
1000
DEn(1990)
HSE(1995)
DNV(2000)
Stress Range (MPa)
Test Data
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
FIGURE 5
1000
DEn(1990)
HSE(1995)
DNV(2000)
Test Data
Stress Range (MPa)
HSE(1995)-16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
18
Section
S-N Curves
FIGURE 6
1000
DEn(1990)
HSE(1995)
DNV(2000)
Test Data
Stress Range (MPa)
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
FIGURE 7
1000
DEn(1990)
HSE(1995)
DNV(2000)
Test Data
Stress Range (MPa)
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
19
Section
S-N Curves
FIGURE 8
1000
DEn(1990)
HSE(1995)
DNV(2000)
Test Data
Stress Range (MPa)
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
FIGURE 9
1000
DEn(1990)
HSE(1995)
DNV(2000)
Test Data
Stress Range (MPa)
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
20
Section
S-N Curves
FIGURE 10
1000
DEn(1990)
HSE(1995)
DNV(2000)
Test Data
Stress Range (MPa)
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
FIGURE 11
1000
DEn(1990)
HSE(1995)
DNV(2000)
Test Data
Stress Range (MPa)
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
21
Section
S-N Curves
FIGURE 12
1000
DEn(1990)
HSE(1995)
DNV(2000)
Test Data
Stress Range (MPa)
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
FIGURE 13
1000
DEn(1990)
HSE(1995)
DNV(2000)
Test Data
Stress Range (MPa)
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
22
Section
S-N Curves
FIGURE 14
1000
DEn(1990)
HSE(1995)
DNV(2000)
Test Data
Stress Range (MPa)
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
FIGURE 15
1000
DEn(1990)
HSE(1995)
DNV(2000)
Test Data
Stress Range (MPa)
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
23
Section
S-N Curves
FIGURE 16
1000
DEn(1990)
HSE(1995)
DNV(2000)
Stress Range (MPa)
Test Data
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
FIGURE 17
1000
DEn(1990)
HSE(1995)
DNV(2000)
Stress Range (MPa)
Test Data
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
24
Section
S-N Curves
FIGURE 18
1000
DEn(1990)
HSE(1995)
DNV(2000)
Stress Range (MPa)
Test Data
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
FIGURE 19
1000
DEn(1990)
HSE(1995)
DNV(2000)
Stress Range (MPa)
Test Data
HSE(1995) 16mm
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
25
Section
S-N Curves
FIGURE 20
Test data with DEn(1990) Thickness Adjustment
and their Regressed S-N Curves (All Thicknesses)
1000
DEn F-Curve without Thickness Correction

Test Data with Thickness Correction
Stress Range (MPa)
Regressed S-N Curve
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
FIGURE 21
Test Data with HSE(1995) Thickness Adjustment
1000
HSE F-Curve without Thickness Correction

Stress Range (MPa)
Regressed S-N curve
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
26
Section
S-N Curves
FIGURE 22
Test Data with DNV(2000) Thickness Adjustment
1000
DNV F-Curve without Thickness Correction

Stress Range (MPa)
Regressed S-N Curve
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
FIGURE 23
Regressed S-N Curves and Design F-curves
1000
Stress Range (MPa)
DEn F-Curve without Thickness Correction

HSE F-Curve without Thickness Correction
DNV F-Curve without Thickness Correction
Regressed S-N Curve with HSE Thickness Correction
Regressed S-N Curve with DEn Thickness Correction
Regressed S-N Curve with DNV Thickness Correction
100
10
1.00E+03
1.00E+04
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
27
Section
S-N Curves
Effects of Corrosion on Fatigue Strength
7.1
Preliminary Remarks
ABS recommendations for considering the effects of corrosion on fatigue strength are based on a review of
corrosion effects published in specifications, guidance and recommended practice documents relating to
marine structures. A digest of the corrosion requirements relative to fatigue is presented for each of several
documents in 3/7.3, below. There is no particular significance to the ordering of the documents presented.
7.2
A Summary of the Results

A review of the requirements suggests only that fatigue strength is reduced in the presence of free corrosion.
One approach is providing separate S-N curves for in-air and free corrosion conditions. Another is to
specify a reduction factor on in-air life when operating in a corrosive environment.
It is generally thought that effective cathodic protection restores fatigue strength to in-air values. However,
both HSE and DNV specify a reduction of the in-air curves for CP joints exposed to seawater. Moreover,
for DNV ship requirements, factors are provided for reduction of in-air S-N curves for those cases where
cathodic protection has become ineffective later in life.
Some documents provide no adjustments for corrosive environments.
ABS archives contain results of corrosion studies on marine structures. These results suggest: (1) it is very
difficult to characterize corrosion in a general, useful engineering context, and (2) there is enormous
statistical variability in corrosion rates.
7.3
The Summaries
API RP2T [API(1997)]
No specific reference to corrosion requirements.
API RP2A [API(2000, 1993)]
i)
For all non-tubular members, refer to ANSI/AWS D1.1-92 (Table 10.2, Figure 10.6). No endurance
limit should be considered for those members exposed to corrosion. For submerged members where
cathodic protection is present, the endurance limit is set at 2 108 cycles.
ii)
The S-N curves are the X and X curves. These curves assume effective cathodic protection. For
splash zone, free corrosion or excessive corrosion conditions, no endurance limit should be considered.
Fatigue Design of Welded Joints and Components [IIW (1996)]

The basic fatigue requirements presented assume corrosion protection. If there is unprotected exposure,
the fatigue class should be reduced. The fatigue limit may also be reduced considerably.
Offshore Installations: Guide on Design, Construction, and Certification, [HSE (1995)]
This document defines basic design curves for plates (P curve) and for tubular joints (T curve). A classification
factor is applied to the P curve to account for different joint types. There are three sets of the basic curves:
(1) in-air, (2) seawater with corrosion protection, and (3) free corrosion. (3) is lower than (2) and (2) is
lower than (1).
The S-N curves are defined in Section 3, Table 6.
28
Section
S-N Curves
TABLE 6
Details of Basic Design S-N Curves HSE(1995)
Class
P
P
Environment
Air
Log10A
12.182
15.637*
m
3
5**
SQ (N/mm2)
53
NQ (cycles)
107
P
P
P
Seawater (CP)
Seawater (CP)
Seawater (FC)
11.784
15.637*
11.705
3
5**
3
84
1.026 106
Air
12.476
67
107
95
1.745 106
16.127*
5**
Seawater (CP)
12.175
Seawater (CP)
16.127*
5**
Seawater (FC)
12.000
Fatigue strength coefficient (C; see Section 5, Figure 4) beyond NQ
**
Fatigue strength exponent (r; see Section 5, Figure 4) beyond NQ
The parameters of Section 3, Table 6 can be translated into reduction factors to be applied to life in the
lower life segment of the in-air S-N curves. These factors are defined in Section 3, Table 7.
TABLE 7
Life Reduction Factors to be Applied to the Lower Cycle Segment
of the Design S-N HSE Curves
Tubular
Joints
Plated
Joints
Cathodic Protected
2.0
2.5
Free Corrosion
3.0
3.0
ISO CD 19902, International Standards Organization [ISO/CD 19902 (2000)]

This is a draft document.
Basic in-air S-N curves are defined for tubular joints, cast joints and other joints.
Joints with cathodic protection. The basic in-air curves apply for N greater than 106 cycles. If significant
damage may occur with N less than 106 cycles, a factor of 2 reduction on life is recommended.
Free corrosion. A reduction factor of 3 on life is required. There is to be no slope change at 108 cycles.
Note:
The editing panel found these statements confusing, so they have requested a re-write.
RP-C203, Fatigue Strength Analysis of Offshore Structures, Det norske Veritas [DNV (2000)]
There are 14 S-N curves, each representing a joint classification. These S-N curves are specified separately
for: (1) in-air, (2) seawater with cathodic protection, and (3) seawater with free corrosion.
In-air. The S-N curves have a break at 107 cycles with a slope of m = 3 in the low cycle range and m = 5
in the high cycle range.
Cathodic protection. The S-N curves in the low cycle range are reduced by the factor of 2.5 on life for
both tubular and plated joints. The curves have a break at 106 cycles.
Free Corrosion. The S-N curves in the low cycle range are reduced by the factor of 3.0 on life for both
tubular and plated joints (see Section 3, Table 8). There is no break in the curves (i.e., m = 3) for all values
of S.
29
Section
S-N Curves
TABLE 8
Life Reduction Factors to be Applied to the Lower Segment
of the Design S-N DNV Curves
Tubular
Joints
Plated
Joints
Cathodic Protected
2.5
2.5
Free Corrosion
3.0
3.0
Eurocode 3 Design of Steel Structures, BSI Standards, 1992 [Eurocode 3, (1992)]

No specific reference to corrosion.
Fatigue Assessment of Ship Structures, Classification Notes No. 30.7, Det norske Veritas, [DNV
(1998)]
A factor is specified for reduction of in-air curves for those cases where cathodic protection is effective for
only a fraction of the life.
BS 7608 Fatigue Design and Assessment of Steel Structures, British Standards Institute [BS 7608
(1993)]
For unprotected joints exposed to seawater, a factor of safety on life of 2 is required. For steels having a yield
strength in excess of 400 MPa, this penalty may not be adequate.
ABS Design Curves; Guide on the Fatigue Assessment of Offshore Structures
The ABS in-air curves for both plated and tubular members are those given in DEn(1990). The basis for this
choice is: (1) the history of successful practice, (2) worldwide acceptance, and (3) relatively conservative
performance in the high cycle range.
The API (2000) curves are permitted as an alternative for application in the Gulf of Mexico based on the
history of successful practice and their mandated use by U.S. Regulatory Bodies.
Adjustment for thickness (see Equations 3.2 and 3.3)
For plated details: q = 0.25; tR = 22 mm
For tubular details: q = 0.25; tR = 32 mm; This applies for thicknesses greater than 22 mm.
The following adjustments to the in-air curves for corrosion were subsequently recommended by the
HSE(1995), these were adopted by ABS.
Tubular Details
With CP. A penalty factor of 2.0 on life applied to the low cycle segment of the in-air S-N curve and
no penalty on life applied to the high cycle segment of the in-air S-N curve.
Free corrosion. A penalty factor of 3.0 on life applied to the low cycle segment of the in-air S-N curve
and continuation of the obtained curve to the high cycle range.
Plated Details
With CP. A penalty factor of 2.5 on life applied to the low cycle segment of the in-air S-N curve and
no penalty on life applied to the high cycle segment of the in-air S-N curve.
Free corrosion. A penalty factor of 3.0 on life applied to the low cycle segment of the in-air S-N curve
and extrapolation of the obtained curve to the high cycle range.
The following adjustments to the in-air curves for corrosion are recommended for the API X and X curves.
Tubular joints
30
CP; endurance limit at 2 108 cycles.
FC; no endurance limit.

Section 4: Fatigue Design Factors
SECTION
Fatigue Design Factors
Preliminary Remarks
The purpose of a fatigue design factor is to account for uncertainties in the fatigue assessment and design
process. The process includes operations of estimating dynamic response and stresses under environmental
conditions. The uncertainties include the following:
Statistical models used to describe the sea states
Prediction of the wave-induced loads from sea state data
Computation of nominal element loads given the wave-induced loads
Computation of fatigue stresses at the hot spot from nominal member forces
Application of Miners rule
Fatigue strength as seen in the scatter in test data, where a typical coefficient of variation on life is
approximately 50-60%.
Environmental effects on fatigue strength (e.g., corrosion)
Size effects on fatigue strength
Manufacturing, assembly and installation operations
In addition to uncertainties, the fatigue design factor should also account for:
Ease of in-service inspection of a detail
Consequences of failure (criticality) of a detail
While reliability methods promise the most rational way of managing uncertainty, the concept of a factor
of safety on life [referred herein as a fatigue design factor (FDF)], maintains universal acceptance.
The Safety Check Expression

The safety check expression can be based on damage or life. While the damage approach is featured in the
Guide, either approach below can be used and are exactly equivalent.
Refer to Subsection 1/2 for terminology. Subsection 1/5 is repeated here for reference.
Damage.
D .......................................................................................................................................... (1.3)
where
1.0
................................................................................................................................... (1.4)
FDF
Life.
Nf (FDF) (NT) .......................................................................................................................... (1.5)
Fatigue design factors specified in relevant documents are summarized in this Section.
31
Section
Summaries of FDFs Specified by Others

The following is a summary of factors of safety on life that have been extracted from documents relevant
to marine structural fatigue. The safety factors by themselves do not tell the whole story and may not address
all of the issues raised above. However, it is instructive and helpful in the development of the Guide to review
those factors that have been published in relevant documents.
It should be noted that safety factors associated with free corrosion and CP are not included in these factors
and should be applied separately.
API RP2T [API (1997)]
General structure. In general, it is recommended that the design fatigue life of each structural element of
the platform be at least three times the intended service life of the platform.
Tendons. high uncertainties exist The component fatigue life factor of ten is considered a reasonable
blanket requirement.
API RP2A [API (2000, 1993)]
In general, the design fatigue life of each joint and member should be at least twice the intended service
life of the structure (i.e., FDF = 2.0).
Fatigue Design of Welded Joints and Components, [IIW (1996)]
For fatigue verification, it has to be shown that the total accumulated damage is less than 0.5 (i.e., FDF = 2.0).
ABS Rules for Building and Classing Steel Vessels, Part 5, The American Bureau of Shipping [ABS
(2001)]
No safety factor specified (i.e., an implied factor of safety on life of 1.0). However, since computed stress
is based on net scantlings, the nominal FDF is greater than 1.0.
Offshore Installations: Guidance on Design, Construction and Certification, UK Department of Energy
[DEn (1990)]
No specific value given. In defining the factor of safety on life, account should be taken of the accessibility
of the joint and the proposed degree of inspection as well as the consequences of failure.
ISO CD 19902, International Standards Organization [ISO CD 19902 (2000)]
In lieu of more detailed fatigue assessment, the FDF can be taken from the following table:
Failure Critical
Inspectable
Uninspectable
No
2.0
5.0
Yes
5.0
10.0
RP-C203 Fatigue Strength Analysis of Offshore Structures, Det norske Veritas [DNV (2000)]
Design fatigue factor from OS-C101, Section 6, Fatigue Limit States
Design Fatigue Factor (DFF) (Table A1 of DNV-OS-C101 Design of Offshore Steel Structures, General
(LRFD Method), Section 6)
The following DFFs are valid for units with low consequence of failure and where it can be demonstrated
that the structure satisfies the requirement for the damaged condition according to the Accidental Limit
State (ALS) with failure in the actual joint as the defined damage.
DFF
32
Structural element
Internal structure, accessible and not welded directly to the submerged part.
External structure, accessible for regular inspection and repair in dry and clean conditions.
Internal structure, accessible and welded directly to the submerged part.
External structure, not accessible for regular inspection and repair in dry and clean conditions.
Non-accessible areas, areas not planned to be accessible for inspection and repair during operation.
Section
Eurocode 3 Design of Steel Structures, BSI Standards [Eurocode 3, (1992)]

This document lists safety factors on stress. These are converted to FDF in the following table.
Fail Safe (a)
Components
Non Fail Safe (b)

Components
Periodic inspection and maintenance

(accessible joint)
1.00
1.95
Periodic inspection and maintenance

(poor accessibility)
1.52
2.46
Inspection and access
Notes:
(a)
local failure of one component does not result in failure of the structure
(b)
local failure of one component leads rapidly to failure of the structure
Fatigue Assessment of Ship Structures, Classification Notes No. 30.7, Det norske Veritas [DNV
(1998)]
Accepted usage factor is defined as 1.0 (FDF = 1.0)
BS 7608 Fatigue Design and Assessment of Steel Structures, British Standards Institute [BS 7608
(1993)]
The standard basic S-N curves are based on a mean minus two standard deviations.... Thus, an additional factor
on life (i.e., the use of S-N curves based on the mean minus more that two standard deviations) should be
considered for cases of inadequate structural redundancy.
33
Section 5: The Simplified Fatigue Assessment Method
SECTION
The Simplified Fatigue Assessment Method
Introduction
The simplified fatigue assessment method employs the Weibull distribution to model the long-term distribution
of sea states. In fact, other distributions could be used, but the Weibull is standard practice in the marine
industry. In this Section, the Weibull distribution is defined and described. Expressions for fatigue damage
at the design life NT of the structure are derived. Also, the allowable stress range approach to safety checking
is derived.
Statistical considerations associated with the Weibull distribution are provided in Subsection 5/9.
The Weibull Distribution for Long Term Stress Ranges
2.1
Definition of the Weibull Distribution

A segment of a long-term stress record at a fatigue sensitive point is shown in Section 5, Figure 1.
FIGURE 1
A Short Term Realization of a Long-Term Stress Record
S
Stress, S(t)
i+1
time, t
The stress range, Si, for the i-th trough and peak is defined. Stress ranges, Si, i = 1, n, form a sequence of n
dependent random variables. In the linear damage accumulation model, this dependency is ignored. Thus, it
will be assumed that Si, i = 1, n is a random sample of independent and identically distributed random variables.
Let S be a random variable denoting a single stress range in a long term stress history. Assume that S has a
two-parameter Weibull distribution. The distribution function is:
s r
Fs(s) = P(S s) = 1 exp

s > 0 ......................................................................... (5.1)
where and are the Weibull shape and scale parameters, respectively. The shape parameter is predetermined
from a detailed stress spectrum analysis or by using historical, empirical data (see Subsections 5/3 and 5/9).
34
Section
The parameters in terms of the mean and standard deviation of S are:
S
S
1.08
S
1
.................................................................................... (5.2)
where S and S are the mean and standard deviation of S respectively. The expression for is approximate,
but for engineering purposes, very close to the exact. (x) is the gamma function defined as:
(x) =
t x 1e t dt ..................................................................................................................... (5.3)
The gamma function is widely available in mathematical analysis programs (e.g., MatLab) and also in
some programmable calculators.
2.2
A Modified Form of the Weibull Distribution for Offshore Structural Analysis

The magnitude of stresses is defined by . However, for design and safety check purposes, it is convenient
to represent in terms of the long term stress spectra as described in the following.
Define a reference life, NR. This could be a time over which records are available (e.g., three years). It
could also be chosen as the design life NT.
Define a reference stress range SR which characterizes the largest stress anticipated during NR. The probability
statement defines SR:
P( S S R )
1
........................................................................................................................ (5.4)
NR
SR is the value that the fatigue stress range S exceeds on the average once every NR cycles.
From the definition of the distribution function, FS(SR) = P(S SR), it follows from Equations 5.1 and 5.4
that:
SR
ln N R 1 /
.......................................................................................................................... (5.5)
The parameter, , is a measure of the amplitude of S(t) and will be independent of the length of time NR
considered.
In the special case where NR is taken as the design life NT, the corresponding stress range ST is defined by
Equation 5.4. The fatigue stress range S will exceed ST on the average once every NT cycles. Thus, ST can
be interpreted as a maximum stress applied during the design life and may be used for static design. ST is
sometimes called the once-in-a-lifetime stress. Using the Weibull formulation, as described in Subsection 5/6,
the fatigue strength can be formulated in terms of maximum allowable stress and a safety check expression
based on stress can be derived.
Typical Values of the Weibull Shape Parameter for Stress
3.1
Experience with Offshore Structures

The Weibull shape parameter for stress, S typically varies between 0.7 to 1.4, depending on the dominant
period of the structural response and the considered wave environments. For example, Gulf of Mexico fixed
platforms experience = 0.7, whereas the same platforms in the North Sea would have > 1, maybe as
high as 1.4 if the platform is slender and experiences significant dynamics. API RP2A separates the longterm distribution into components, hurricane waves and operational sea state. = 1.0 is used for both.
In general, in the absence of data, one might choose a suitable value based on experience from the fatigue
analysis of similar structures.
35
Section
3.2
Experience with Ships

Ships operating mainly in the Northern or Southern oceans will generally have a Weibull shape parameter,
> 1, maybe as high as 1.3 or a little more. However, there are cases where < 1 (see Section 5, Figure 7).
Part 5C of the ABS Rules for Building and Classing Steel Vessels provides an empirical algorithm for the
shape parameter (long term stress distribution parameter) that depends upon the length of the ship and
also the location of the detail in the ship.
Fatigue Damage: General
4.1
Preliminary Remarks
This Subsection provides detail on the fatigue damage expressions that are used in the Guide. Considered
are both the single segment and the bi-linear S-N curves which are used to describe fatigue strength in design
criteria documents.
4.2
General Expression for Fatigue Damage

Assume that a statistical model (e.g., Weibull) has been fit to stress range data. Let S be a random variable
denoting stress range. The probability density function of S is shown in Section 5, Figure 2.
FIGURE 2
Probability Density Function of s
f (s)
S
f(s )
i
Stress, s
An expression for damage is derived in the following.

Consider a small increment of s of width s at stress level si, as shown in the figure. The number of cycles
to failure at stress level si as obtained from the S-N curve defining fatigue strength is:
Ni = N(si). ..................................................................................................................................... (5.6)
The number of cycles of applied stress at level Si is:
ni = NR [fs(si)s] ........................................................................................................................... (5.7)
36
Section
where NR is any reference life. The term in brackets is the fraction of the total cycles associated with si.
Substituting Equation 5.7 into Equation. 1.1 (the basic damage expression of the PM rule defined in
Subsection 1/3), it follows that the damage at reference life NR is:
DR =
N R f s ( s i )s
................................................................................................................ (5.8)
N (si )
where the summation is taken over the whole sample space of s. In the limit as s 0, the summation
becomes an integral.
DR = N R
f s ( s )ds
.................................................................................................................... (5.9)
N (s)
Equation 5.9 is the damage at life NR. This expression will be used in the following to derive closed form
expressions for damage.
Fatigue Damage for Single Segment S-N Curve
5.1
Expression for Damage at Life, NR

Assume that the single segment S-N curve, as shown in Section 5, Figure 3, defines the fatigue strength.
FIGURE 3
Characteristic S-N curve
Log(S)
NS
=A
1
m
Log(N)
The analytical form of the S-N curve is:

N(s) = As-m ................................................................................................................................ (5.10)
Damage becomes, from Equation 5.9:
DR =
NR
A
f ( s )ds .............................................................................................................. (5.11)
But the integral is by definition the expected value of Sm.
E ( S m ) = s m f ( s )ds ................................................................................................................ (5.12)

0
37
Section
Thus the expression for damage at any arbitrary life NR can be written:
D=
5.2
N R E (S m )
......................................................................................................................... (5.13)
A
Miners Stress
In the special case where s is constant amplitude:
E(Sm) = Sm ................................................................................................................................. (5.14)
Given E(Sm) from a random stress process, Equation 5.14 is used to define se as:
Se = [E(Sm)]1/m ........................................................................................................................... (5.15)
Se can be thought of as an equivalent constant amplitude stress which produces the same fatigue damage as
the random stress process. Se is called Miners stress.
Using Se, damage at any life NR for the single segment S-N curve can be written as:
DR =
5.3
N R S em
............................................................................................................................ (5.16)
A
The Damage Expression for Weibull Distribution of Stress Ranges

The expression for damage is derived from Equations 5.11 and 5.12. It is necessary to determine the probability
density function for the Weibull distribution. The probability density function is fs(s) = dFs/ds, where Fs is
defined in Equation 5.1.
s
f s ( s ) =

r 1
s r
exp

s > 0 ....................................................................... (5.17)
Upon substituting Equation 5.17 into Equation 5.12, it follows that:

m
E ( s m ) = m + 1 ............................................................................................................... (5.18)
where (.) is the gamma function defined in Equation 5.3.

The general expression for damage at life NR is obtained by substituting Equation 5.18 into Equation 5.13.
Damage at the design life NT is obtained by letting NR = NT.
D=
NT m m
+ 1 ............................................................................................................... (5.19)
A
The fatigue safety check is:
1.0
................................................................................................................................. (5.20)
FDF
Fatigue Damage for Bilinear S-N Curve

The bilinear S-N curve, as shown in Section 5, Figure 4, is specified in the Guide for in-air and cathodic
protection. Fatigue damage for the case of Weibull-distributed stress and the bilinear S-N curve is addressed
in this Subsection.
38
Section
FIGURE 4
Bilinear Characteristic S-N curve
Log(S)
NS
=A
NS = C
m
SQ
1
r
Log (N)
NQ
Fatigue strength is given as:

N(s) = Asm
for s > sQ
N(s) = Csr
for s < sQ .......................................................................................... (5.21)
The basic damage equation is given by Equation 5.9. For the bilinear case, Equation 5.9, damage at reference
life NR, becomes:
DR =
NR
C
sQ
s r f s ( s )ds +
NR
A
sQ
s m f s ( s )ds ........................................................................... (5.22)
Upon integration of Equation 5.22 and after some reduction, the expression for damage at the design life
NT, for the two-segment case of Section 5, Figure 4 is derived [Wirsching and Chen (1988)].
D=
NT m
A
m
N r r
+ 1, z + T
o + 1, z ........................................................................ (5.23)
(z,a) and o(z,a) are incomplete gamma functions (integrals z to and 0 to z, respectively). For reference,
these functions are available in MatLab.
( z , a ) =
o ( z , a) =
a=
b=
t a 1 e t dt ........................................................................................................... (5.24)
z
O
t a 1 e t dt .......................................................................................................... (5.25)
+ 1 ................................................................................................................................. (5.26)
+ 1 .................................................................................................................................. (5.27)
sQ
z =
................................................................................................................................ (5.28)
The safety check expression of Equation 5.20 is applied.
39
Section
Safety Check Using Allowable Stress Range

A safety check expression can also be developed in terms of the once-in-a-lifetime stress range, SR.
First, consider the single segment S-N curve. Let D = (1/FDF). From Equation 5.5, can be expressed in
terms of SR. Substituting D and into Equation 5.19 and solving for SR, it follows that:
m/
A(ln N R )
S R =
m
(FDF )N T + 1
1/ m
................................................................................................. (5.29)
This is actually the maximum allowable stress range that would be permitted once on the average every NR
cycles, and as such would be interpreted as fatigue strength. The prime is included to distinguish it from
stress. The fatigue stress is SR and is the computed fatigue stress that will be exceeded once on the average
every NR cycles.
The safety check expression is:
SR SR ...................................................................................................................................... (5.30)
In practice, it is most likely that NR will be taken as the design life NT so that stress and strength (SR and SR)
will refer to NT.
If SR has been computed, the strength (maximum allowable stress range) SS at any other life NS can be
computed using Equation 5.5.
ln N S
S S = S R
ln N R
1/
.................................................................................................................. (5.31)
Now consider the bilinear S-N curve (Section 5, Figure 4 and Equation 5.23). Let D = (1/FDF). From
Equation 5.5, can be expressed in terms of SR. Substituting D and into Equation 5.23 and solving for
SR, it follows that:
m/
(
ln N R )
S R =
(FDF )N m + 1, z / A + r m r + 1, z / C
T
o
1/ m
................................................ (5.32)
Again, the prime superscript is used to denote strength. The strength modification of Equation 5.31 will apply.
The safety check expression of Equation 5.30 would apply.
The Simplified Method for Which Stress is a Function of Wave Height
8.1
The Weibull Model for Stress Range; Stress as a Function of Wave Height
Another refinement of the simplified method is a model in which: (a) the long term distribution of wave
heights H is known (or assumed), (b) H is assumed to have a Weibull distribution, and (c) the fatigue stress
range S is related to H by the power law. Such a model is presented in API RP2A [API(2000, 1993)] and
described in the following. This model is not presented in the Guide. It appears here only for reference.
API RP2A [API(2000, 1993)] defines a damage model based on Weibull wave heights. This model was
based on an earlier work by Nolte and Hansford (1976). Consider the random instantaneous ocean wave
elevation process as a sequence of waves. Let H be a random variable denoting the height (i.e., the vertical
distance from the trough to the following peak) of a single wave.
40
Section
It is assumed that stress range S is proportional to the wave height H that produces that stress.
S = ChHg ................................................................................................................................... (5.33)
where Ch and g are empirical constants. If H is Weibull, then S will also be Weibull (from basic probability
theory). To compute Miners stress:
Sem = E(Sm) = E(Chm Hgm) = ChmE(Hgm) .................................................................................. (5.34)
From Equations 5.5 and 5.18, it follows that:
E H gm = H Rmg (ln N R )
gm /
gm
+ 1 ................................................................................. (5.35)

where is the Weibull shape parameter of H. HR is a wave height that is exceeded on the average once
every NR cycles (analogous to SR of Equation 5.5). Damage can be computed by substituting Equations 5.34
and 5.35 into Equation 5.16.
D=
8.2
NR m
C h E ( H gm ) ................................................................................................................ (5.36)
A
The Weibull Model for Stress Range; Stress as a Function of Wave Height;
Considering Two Wave Climates
Assume that there are two distinct wave climates in a region, one characterized as the usual wind-generated
seas and the other the exceptional cyclonic storm seas of a hurricane (e.g., Gulf of Mexico). Both are assumed
to have Weibull distributions. The parameters are:
Ho1
design wave height for the period of wave climate 1
Ho2
design wave height for the period of wave climate 2
N1
number of wave cycles for the period of wave climate 1
N2
number of wave cycles for the period of wave climate 2
Weibull shape parameter for wave heights for wave climate 1
Weibull shape parameter for wave heights for wave climate 2
as needed, a calibration factor for wave climate 1
as needed, a calibration factor for wave climate 2
Damage at life n is equal to the sum of the damage in each wave climate.
D=
NR
(Yo + Y1 ) ..................................................................................................................... (5.37)
A
where
NR = N1 + N2 ............................................................................................................................. (5.38)
Y1 =
N 2 mg
gm / 1 gm
+ 1 ..................................................................................... (5.39)
H o1 (ln N 1 )
NR
1
Y2 =
N 2 mg
gm / 2 gm
+ 1 ................................................................................... (5.40)
H o 2 (ln N 2 )
NR
2
41
Section
The Weibull Distribution; Statistical Considerations
9.1
Preliminary Remarks
The Weibull distribution is commonly used to describe long term wave height and long term stress range
processes. Statistical considerations of the Weibull distribution are presented below for reference. Specifically,
parameter estimation and probability plotting are described. Also, the operations of fitting the Weibull
distribution to deterministic and spectral models are presented.
9.2
Estimating the Parameters from Long-Term Data; Method of Moment Estimators

Given a long-term measured history, the stress ranges Si, i = 1, n are recorded. It is assumed that the
observations in this random sample are independent and identically distributed (i.e., all dependencies are
ignored). It must be emphasized that this history be representative of the entire service life.
The parameters can be estimated using the method of moments. Compute the sample mean and standard
deviation.
S=
1
n
S
i =1
1
sS =
n 1
.............................................................................................................................. (5.41)
n
(S
i =1
1/ 2
..................................................................................................... (5.42)
S and sS are estimates for s and S. From Equation 5.2, the method of moment estimators for and are:
sS
=
=
1.08
S
1
+ 1
............................................................................................................................ (5.43)
............................................................................................................................ (5.44)
While the more complicated maximum likelihood estimators are generally considered by mathematicians
to be of higher quality, extensive and unpublished Monte Carlo simulation by Wirsching has demonstrated
that the performance of the two estimator types to be essentially identical.
Example. An illustration of a Weibull statistical analysis for which the sample size, n = 25. The random
sample:
S = (12.5377 9.8050 8.2736 12.3759 11.0321 14.6142 21.6679 9.4326 17.2867

4.3187 24.5273 5.7944 5.5171 17.1173 13.2568 3.7245 0.8759 17.8508 4.7251
6.1859 11.7465 5.9809 8.6302 2.3726 26.4654)
Using Equations 5.41 and 5.42, the sample mean and standard deviation are, respectively, S = 11.04, sS = 6.83.
Substituting these values into Equations 5.43 and 5.44, the method of moment estimates for and are:
= 1.68
9.3
= 12.36
Estimating the Parameters from Long-Term Data; Probability Plotting

A probability plot involves a transformation that allows the distribution function to be plotted as a straight
line. Consider the distribution function of Equation 5.1. Taking the log of both sides twice results in:
ln[ln(1 FS)] = ln s ln ................................................................................................. (5.45)
Let
Y = ln[ln(1 FS)]
X = ln s ...................................................................................................................................... (5.46)
42
Section
The result is a linear relationship in which is the slope of the plotted line and ln is the y-intercept.
Given the random sample, the distribution function is estimated in the following manner:
i)
The sample is ordered, smallest to largest. S(i) denotes the i-th smallest value.
ii)
The estimate of the distribution function corresponding to S(i) is:
Fi =
i 0.5
.............................................................................................................................. (5.47)
n
Fi and Si are translated into Yi and Xi using Equation 5.46. The points are plotted on rectangular paper. If
the data follow a linear trend, it is suggested that the Weibull distribution is a reasonable model. This decision is
purely subjective.
A straight line can be fitted through the points. The least squares estimators of and are then:
~ = slope of Y-X line .............................................................................................................. (5.48)
~
= exp
y - intercept
......................................................................................................... (5.49)
~
Example: Consider the random sample as given above. The following table provides the details of the
preparation of the data for plotting. Xi and Yi are computed using Equation 5.46.
TABLE 1
Data Analysis for Weibull Plot
i
S(i)
Fi = (i 0.5)/n
Yi
Xi
0.88
0.02
3.90
0.13
2.37
0.06
2.78
0.86
3.72
0.10
2.25
1.31
4.32
0.14
1.89
1.46
4.73
0.18
1.62
1.55
5.52
0.22
1.39
1.71
5.79
0.26
1.20
1.76
5.98
0.30
1.03
1.79
6.19
0.34
0.88
1.82
10
8.27
0.38
0.74
2.11
11
8.63
0.42
0.61
2.16
12
9.43
0.46
0.48
2.24
13
9.81
0.50
0.37
2.28
14
11.03
0.54
0.25
2.40
15
11.75
0.58
0.14
2.46
16
12.38
0.62
0.03
2.52
17
12.54
0.66
0.08
2.53
18
13.26
0.70
0.19
2.58
19
14.61
0.74
0.30
2.68
20
17.12
0.78
0.41
2.84
21
17.29
0.82
0.54
2.85
22
17.85
0.86
0.68
2.88
23
21.67
0.90
0.83
3.08
24
24.53
0.94
1.03
3.20
25
26.47
0.98
1.36
3.28
43
Section
Y versus X is shown in Section 5, Figure 5.
FIGURE 5
Weibull Probability Plot
2
Y = ln(-ln(1-F ))
S
1
0
-1
-2
y = -4.01 + 1.59x
-3
X = ln(S)
-4
-0.5
0.5
1.5
2.5
3.5
As shown in the figure, the slope is 1.59 and the y-intercept is 4.01. Thus, the least squares estimators
from Equations 5.48 and 5.49 are:
~ = 1.59
4.01
= 12.45
1.59
= exp
Some observations:
44
i)
The data plot is nearly a straight line on Weibull paper. This suggests that the Weibull is a reasonable
model for the random variable, S. This does not mean that nature actually chose the Weibull.
ii)
The method of moment estimators and least squares estimators are in relatively good agreement.
The difference is due to the randomness in the process.
iii)
For large sample sizes, the maximum likelihood estimators are minimum variance unbiased estimators.
However, method of moment estimators have been found to be insignificantly different from
maximum likelihood.
iv)
When the data do not plot as a straight line, the Weibull model might be a poor choice and the analyst
must exercise engineering judgment regarding the possible choice of another distribution. Another
indication that the Weibull model might be inappropriate would be poor agreement between method
of moment and least squares estimators.
Section
9.4
Another Representation of the Weibull Distribution Function

For fatigue purposes, the long-term Weibull distribution of fatigue stresses is often described graphically
as shown in Section 5, Figure 6.
FIGURE 6
Long Term Distribution of Fatigue Stress as a Function
of the Weibull Shape Parameter
0.8
1.4
0.6
S/So
= 1.0
0.4
= 0.7
0.2
Number of Exceedances
0
1
10
100
1000
10
10
10
10
10
Section 5, Figure 6 is another graphical representation of the distribution function. Let NT be the total number
of stress cycles in the design life. Let n be the number of values of the stress range during the service life
that exceed the value, s (i.e., the number of exceedances of S by stress level, s). Then, for any value of S
selected at random, the probability of exceeding the value s is simply equal to n/NT.
P(S > s) =
number of values of S which exceed s

n
=
= 1 P(S s) = 1 FS(s) ....... (5.50)
total number of stress ranges in design life
NT
By combining Equations 5.1, 5.4 and 5.50, the following equation, a restatement of the distribution function,
is obtained.
S
exp
ST
n
ln N T
..................................................................................................... (5.51)
NT
where n is the number of exceedances of the fatigue stress of stress level S. Section 5, Figure 6 is a plot of
(S/ST) versus n for the case where NT = 108 and for three values of .
An example of the use of this plot for marine structures is shown in Section 5, Figure 7.
45
Section
FIGURE 7
Long-Term Stress Range Distribution of Large Tankers, Bulk Carriers,
and Dry Cargo Vessels Compared with the Weibull
Bending Stress - Range (S/So)
W.H. Munse
"Fatigue Characterization of
Fabricated Ship Details for Design".....
SSC-318 (1983)
12
102
102
103
10422
105
1066
Number of Exceedances, N Cycles
107
108
= NT
A crude method for estimating the Weibull shape parameter, , is to plot observed stress data on the figure
as shown and then choose that value of which provides the curve matching the experimental data. The
method of moments estimator for is recommended.
9.5
Fitting the Weibull to Deterministic Spectra

A deterministic spectra defined over the service life (Section 5, Table 2) provides the stress information for
computing fatigue damage and executing a safety check. However, a Weibull distribution can be fitted to
this deterministic spectra. One reason for doing this is to develop information on the Weibull shape parameter
working towards characteristic values for various structure types.
TABLE 2
Deterministic Spectra
Stress
Range
Number of
Cycles
s1
n1
si
ni
sk
Note that: (1) N S

S=
nk
, and (2) there are ni values of si. Thus, the random sample is generated as:
(n1 values of s1 plus n2 values of s2 plus ni values of si plus nk values of sk.) ......(5.52)
The method of moment estimators, as described above in 5/9.2, are used to estimate and .
46
Section
9.6
Fitting the Weibull Distribution to the Spectral Method

While the spectral method provides definition of fatigue stresses for analysis, the Weibull parameters can
be extracted from the spectra. As in the case of the deterministic method, a random sample is constructed.
In the spectral method, the long-term stress process is discretized into J short term stationary sea states.
Consider the i-th sea state. The stress ranges in this sea state will have a density function (e.g., as shown in
Section 5, Figure 8).
FIGURE 8
Probability Density Function of Stress Ranges of the i-th Sea State
f (s)
i
f i (S m )
Stress, s
If the stress process is assumed to be Gaussian and narrow band, then S will have a Rayleigh distribution.
However, this is not a necessary assumption to execute this process.
To construct the random sample of stress ranges, divide the sample space into M small increments. One such
increment of width, s, is shown in the figure. The continuous distribution of S is transformed to a discrete
distribution of M constant amplitude stresses. One of those stresses, sm, is shown at the center of the interval.
The following definitions lead to the number of cycles of Sm.
TS
service life
% of time in the i-th sea state
Ti
TS
time in the i-th sea state
center frequency of the stress process in the i-th sea state
Ni
Ti
number of cycles in the i-th sea state
nm
[fi(sm) s]
Ni
number of cycles of sm
47
Section
The contribution to the random sample is nm cycles of sm. Then,
48
Repeat the above for all of the intervals of the i-th sea state
Repeat for all sea states
Pool all of the values of S and use the method of moments
Section 6: The Spectral Based Fatigue Assessment Method
SECTION
The Spectral Based Fatigue Assessment Method
The spectral fatigue analysis is a systematic approach to develop a rational, first-principle approach whereby
the dynamic loads and the wave environments are explicitly considered. The essence of the spectral fatigue
analysis method is to model the long-term random sea state process as several short-term stationary Gaussian
processes, each defined by the wave spectral density function. The fraction of time that each short-term
process acts is specified. A frequency response function relating the wave spectral density and the spectral
density of fatigue critical components, and including the structural dynamics, is developed. The fatigue stress
spectral density is computed. If the process can be assumed to be narrow band (a conservative assumption),
the stress ranges will have a Rayleigh distribution and a closed form expression for fatigue damage contributed
by each of the short-term processes can be used.
A summary of the process of fatigue assessment by the spectral fatigue analysis method is provided in
Section 6, Figure 1.
Basic Assumptions
Basic assumptions in spectral fatigue analysis are:
i)
The global performance analysis and the associated structural analysis are assumed to be linear. Hence,
scaling and superposition of stress range transfer functions due to unit wave height are valid.
ii)
Non-linearities due to non-linear motions and wave loadings are treated by the equivalent linearization.
iii)
Structural dynamic amplification, transient loads and effects such as springing and ringing are
insignificant and, hence, use of quasi-static finite element analysis is valid.
iv)
Short-term stress processes are assumed to be stationary Gaussian.
v)
Short-term stress processes are also assumed to be narrow band so that the stress ranges have a
Rayleigh distribution. (If a decision is made that the stress process must be treated as wide band, a
correction to the narrow band assumption is provided.)
vi)
The linear damage accumulation rule (Palmgren-Miner) applies.
A fundamental limitation of the spectral method is the assumption of linearity (i.e., items i) and ii) above).
Thus, the spectral method may not be appropriate in those cases where non-linearities are considered to be
important (e.g., TLP tendons).
49
Section
The Spectral-based Fatigue Assessment Method
FIGURE 1
Fatigue Assessments by Spectral Analysis Method
DEFINE
OPERATIONAL
PROFILE
DEFINE THE LONGTERM DESIGN WAVE

ENVIRONMENT
DEFINE THE
FATIGUE CRITICAL
STRUCTURAL
DETAILS
PERFORM FREQUENCY DOMAIN ANALYSES OF

SHORT-TERM FATIGUE STRESS PROCESSES
direction seakeeping analysis for ships

global finite element analysis
local finite element analysis
frequency response functions for fatigue stresses
spectral density functions for fatigue stresses
DERIVE SHORT-TERM
DISTRIBUTIONS OF
STRESS RANGE
CALCULATE AND SUM

SHORT-TERM FATIGUE
DAMAGES
SAFETY CHECK
EXPRRESSION FOR
FATIGUE
The Rayleigh Distribution for Short Term Stress Ranges

The long term (non-stationary) sea state is modeled by several short-term stationary sea states. It is assumed
that the short term processes are Gaussian and narrow band. Thus, the distribution of the peaks is Rayleigh
[Wirsching, Paez, and Ortiz (1995)]. The Rayleigh distribution is a special form of the Weibull distribution
having parameters:
= 2.0
= 2 2 ....................................................................................................... (6.1)
Thus, it follows from Equation 5.19 that damage at the any reference life NR is:
DR =
NR
2 2
A
m
+ 1 ..................................................................................................... (6.2)
2
This expression forms the basis of the damage expression for the spectral method.
50
Section
Spectral Analysis; More Detail

Under the previously stated assumptions, a more detailed description of the spectral analysis method is provided
in the following:
i)
Characterization of Sea Environment. The sea environment is represented by the number of occurrences
of various sea states, each defined by a set of spectra. A two-parameter (significant wave height
HS, zero up-crossing period, TZ) wave-scatter diagram shall be used to characterize the sea states
(See Section 6, Table 1 for example). All sea state spectra are defined by the wave spectrum
relationship such as the Pierson-Moskowitz relationship. Wave direction probability is included in
the sea environmental characterization.
ii)
Global Performance Analysis. Waves of appropriate frequencies, heights and directions are selected.
The response and the loading of the structure are computed for each wave condition.
iii)
Structural Analysis. A global structural analysis is performed to determine the applied loading or
displacement for the critical structural details (load transfer function per unit wave amplitude as a
function of frequency). The local structural analysis is performed to determine the stress transfer
function (per unit wave amplitude as a function of wave frequency) at each location of interest in
the structural detail.
iv)
Stress Concentration Factor (SCF). The geometric SCF is accounted for in the fatigue assessment.
The SCF can be determined either by parametric equation or by fine mesh FEA.
v)
Hotspot Stress Transfer Function. As needed the stress transfer function is multiplied by the SCF
to determine the hotspot stress transfer function.
vi)
Long-term Stress Range. Based on the wave spectrum, wave scatter diagram and hotspot stress
response per unit wave amplitude, the long-term stress range is determined. This is done by
multiplying the ordinate of the wave amplitude spectrum for each sea state by the ordinate squared
of the hotspot stress transfer function to determine the stress spectrum. The stress range distribution
is assumed to follow a Rayleigh distribution. The long-term stress range is then defined through a
short-term Rayleigh distribution within each sea state for different wave directions.
vii)
S-N Classification. For each critical location considered in the analysis, S-N curves are assigned
based on the structural geometry, applied loading and, when applicable, welding procedure and quality.
viii)
Fatigue Life. Estimation of fatigue damage (or life) using the Palmgren-Miner Rule and subsequent
comparison with design requirements.
Wave Data
The basis of the spectral fatigue approach lies in the wave environment data that are contained within the
wave scatter diagram that is to be used. Clearly, physically measured wave data is the most reliable form.
In lieu of the physical measurement data, visually observed wave data and some well compiled authoritative
sources (e.g., ABSWAVE, BMT Global Wave Statistics, Walden Data, etc.) can be used.
The wave data should be available in scatter diagram form which consists of M number, of cells that
contain the probability of occurrence of specific sea states. Any cell will effectively contain three data
items, namely
i)
The significant wave height, Hs, (typically in meters),
ii)
The characteristic wave period, TZ, (in seconds) and
iii)
The fraction of the total time that the wave condition exists, pi.
Wave data are characterized as a wave scatter diagram, an example of which is shown in Section 6, Table 1.
51
Section
TABLE 1
A Sample Wave Scatter Diagram
Wave Period (sec)
Wave Height (m)
5s
7s
9s
11s
13s
15s
Sum Over
All
Periods
17s
0.0-0.75
20.91
11.79
4.57
2.24
0.47
0.06
0.00
0.60
40.64
0.75-1.75
72.78
131.08
63.08
17.26
2.39
0.33
0.11
0.77
287.80
1.75-2.75
21.24
126.41
118.31
30.24
3.68
0.47
0.09
0.56
301.00
2.75-3.75
3.28
49.60
92.69
32.99
5.46
0.68
0.12
0.27
185.09
3.75-4.75
0.53
16.19
44.36
22.28
4.79
1.14
0.08
0.29
89.66
4.75-5.75
0.12
4.34
17.30
12.89
3.13
0.56
0.13
0.04
38.51
5.75-6.75
0.07
2.90
9.90
8.86
3.03
0.59
0.08
0.03
25.46
6.75-7.75
0.03
1.39
4.47
5.22
1.93
0.38
0.04
0.04
13.50
1.09
2.55
3.92
1.98
0.50
0.03
0.02
10.09
0.54
1.36
2.26
1.54
0.68
0.20
0.04
6.62
0.01
0.10
0.11
0.10
0.05
0.02
0.00
10.75-11.75
0.00
0.03
0.08
0.17
0.06
0.00
11.75-12.75
0.05
0.00
0.14
0.22
0.06
0.01
12.75-13.75
0.02
0.07
0.09
0.03
0.02
0.06
0.02
0.00
0.01
0.11
0.01
0.01
0.02
0.01
0.01
0.08
138.59
29.05
5.63
0.92
2.69
1000.00
7.75-8.75
8.75-9.75
9.75-10.75
0.01
13.75-14.75
14.75-15.75
Sum Over All Heights
0.02
118.97
345.43
358.72
0.40
0.34
0.48
0.01
0.22
The long-term sea state is discretized as a set of M short-term sea states as shown. Each sea state, which is
assumed to be a stationary Gaussian process, contributes to the total fatigue damage.
The numbers in each cell (divided by 1000) define the joint probability of occurrence of Hs and Tz, denoted
as pi. The fraction of the total time each cell is acting is defined by pi, i = 1, M.
For a more refined analysis, wave directionality can be accounted for. Each of the M cells can be subdivided
into discretized wave direction. The probability distribution of wave direction needs to be defined.
Fatigue damage is computed for each of the M cells and summed to compute total damage.
Additional Detail on Fatigue Stress Analysis; Global Performance

Analysis
In order to translate wave data into fatigue stresses, it is necessary to construct a frequency response function,
and to do this, it is necessary to perform a motion response analysis. This analysis should be based on sixdegree-of-freedom global motion models with full simulation of the mass properties of the operating platform
or ship. For a TLP, the global performance model may be uncoupled from the tendons and risers, meaning
that the responses of these components are not included in the model, although the static stiffness must be
included.
The analysis should be conducted in the frequency domain based on panel model diffraction loads or strip
theory to determine the hydrodynamic loads, stiffness matrix at the quasi-static offset position in the
environment, and full spectral definition of the sea states. A suitable range of wave frequency, number of
frequency points and wave headings shall be used in the analysis. Ideally, for ship type structure, the following
are suggested:
52
i)
Frequency range:
0.20 to 1.80 rad/sec
ii)
Frequency Increment:
0.05 rad/sec
iii)
Wave headings:
0 to 360 degrees in 15 degree increments
Section
The global motion analyses will serve as the basis for load development in FEA, discussed later. The actual
interface from the global analysis to the structural analysis consists of six loading components for each analyzed
wave period and direction: the real and imaginary applied unit amplitude wave diffraction and radiation
loads, the associated inertial loads and the other cyclic loading (e.g., TLP tendon dynamic reactions). The
successful interface of these load components, as discussed above, is dependent on a consistent geometry and
mass model between the motion and structural analyses and consistent generation of the loading components
in the motion analysis. Consistent modeling can be obtained by interfacing the model geometry directly
from the motion analysis wherever possible. Consistent mass can be obtained by interfacing with the same
weight control database for both the motions and structural analyses, when available.
The Safety Check Process
7.1
General Considerations
As a general reference to random processes for the following discussions, see Wirsching, Paez, and Ortiz
(1995).
7.2
The Stress Process in Each Cell

Fatigue stresses are to be established by the following process:
i)
Determine the stress transfer function, H(|), for a particular location of interest in the structural
detail for a particular load condition. This can be done in a direct manner, whereby, the structural
analysis is carried out at every frequency and heading for which the spectral analysis is intended to
be carried out and the resulting stresses are used to generate the stress transfer function explicitly.
ii)
For each of the M cells or short-term sea states, generate a stress spectra, S(|Hs, Tz, ), by
linearly scaling the wave spectral density, S(|Hs, Tz), in the following manner:
S(|Hs, Tz, ) = | H(|)|2 S(|Hs, Tz) .................................................................................. (6.3)
iii)
For each of the M cells or short-term sea states, calculate the spectral moments. The n-th spectral
moment, n, is calculated as follows:
n = n S ( H S , TZ , )d ....................................................................................................... (6.4)
Most fatigue damage is associated with low or moderate seas. Therefore, confused short-crested sea conditions
should be considered.
Confused short-crested seas result in a kinetic energy spread which is modelled using the cosine-squared
approach, (2/)cos2. Generally, cosine-squared spreading is assumed from +90 to 90 degrees on either
side of the selected wave heading. Applying the wave spreading function, the spectral moment may be
modified as follows:
n =
' = + 90
2
2 ' n
cos S ( H S , TZ , )d ........................................................................... (6.5)
= 90
Often, the correction for short crested seas is ignored, leading to conservative damage estimates.
iv)
Using the spectral moments, compute: (1) the RMS i, (equal to the standard deviation for a zero
mean process), (2) rate of zero crossings (equivalent frequency) foi, and, if the decision is made to
include a wide band correction, (3) rate of peaks fpi, and (4) the spectral width parameter for each
of the M cells. Let ni be the n-th spectral moment of the i-th cell. General reference for the
following discussion is Wirsching, Paez, and Ortiz (1995).
RMS
i = 2i (6.6)
53
Section
Rate of Zero Crossings (equivalent frequency)

The rate of zero crossings (in Hertz) for the i-th cell is:
f oi
1
2
2i
...........................................................................................................................(6.7)
oi
The expression for foi is based on a stress spectral density that is a function of (rad/sec) per Equation 6.4.
If the spectral density function is in terms of f (Hertz), then the factor (1/2) should be missing. See
Subsection 6/8.
If the decision is made to also apply the wide band correction factor, the following must be computed.
Ignoring this factor produces conservative results.
Rate of Peaks
The rate of peaks (number per second) for the i-th cell is:
f pi
4i
...................................................................................................................................(6.8)
2i
The expression for fpi is based on a stress spectral density that is a function of (rad/sec) per Equation 6.4.
If the spectral density function is in terms of f (Hertz), then the factor (1/2) should be missing. See
Subsection 6/8.
Spectral Width Parameter

The spectral width parameter is an index of the spectrum (i.e., the stress spectral density function). For the
i-th sea state:
i 1 i2
0 < i < 1.0 .......................................................................................................(6.9)
where i is the irregularity factor defined as:
f oi
f pi
0 < i < 1.0 ...........................................................................................................(6.10)
A process can be assumed to be narrow band if i will be close to 1.0 (say 0.95) and i close to zero. i
becomes smaller as the process becomes more wide band. The spectral shape will differ with each sea state
and, thus, i must be defined for each sea state.
Fatigue Damage Expression for Wide Band Stress
8.1
As a general reference, consider Wirsching, Paez, and Ortiz (1995).
The expression for damage used by ABS is based on the assumption of a narrow band stress process in which
the stress ranges have a Rayleigh distribution. However, the stress process will, in general, be wide band.
A rainflow correction factor, an index of the wide band spectrum, is applied to the narrow band equation
for damage to form a damage expression for wide band stress. For the i-th sea state, the rainflow correction
factor is denoted as i and the frequency of loading is taken to be foi.
Realizations of a narrow band and a wide band process are shown in Section 6, Figure 2. Each has the
same RMS and rate of zero crossings.
54
Section
FIGURE 2
Realizations of a Narrow Band and Wide Band Process
(Both Have the Same RMS and Rate of Zero Crossings)
Stress
Narrow band
process
One cycle
Wide band
process
time
For the narrow band process, stress cycles are easily identified and the application of Miners rule is
straightforward. While it is not immediately obvious how to identify individual stress cycles in a wide band
process, the rainflow algorithm provides a formal method for stress cycle counting. 6/8.4 describes the
rainflow method and presents a closed form expression for computing damage based on a correction factor
applied to the equivalent narrow band process.
8.2
Definitions
The spectral density function of the stress process, S(t), is given as WS(f) where frequency, f, is in Hertz.
An index of the frequency spectrum of the stress process, S(t), is the irregularity factor:
fo
fp
0 < <1.0 ...................................................................................................... (6.11)
where fo is the rate of zero crossings of S(t) with positive slope and fp is the rate of peaks of S(t). In terms
of moments of the spectral density function of S(t):
fo =
2
o
fp =
4
....................................................................................................... (6.12)
2
where the moments of the spectral density function, WS(f), are defined as:
j =
f j W S ( f )df ............................................................................................................... (6.13)
An alternate index is the spectral width parameter:
= 1 2 .............................................................................................................................. (6.14)
For a narrow band process, = 1.0 and = 0. For values close to 1.0, the process has an approximate narrow
band appearance. As decreases, the process becomes more wide-band, but there is no formal definition
of as to what defines a wide band stress.
Finally, note that the standard deviation (or RMS, root mean square if the process has a zero mean) is:
= o .................................................................................................................................. (6.15)
55
Section
8.3
The Equivalent Narrow Band Process

Consider a stationary Gaussian narrow band process. The damage at any cycle life n (see Equation 6.2) is
(the subscript NB denotes narrow band stress damage):
D NB =
n
2 2
A
m
+ 1 .................................................................................................... (6.16)
2
DNB can be written in terms of frequency and time. To compute damage at time t note that the number of
cycles is:
n = fot ......................................................................................................................................... (6.17)
where fo is the rate of zero crossings or equivalent frequency of the process.
Thus, DNB becomes:
D NB =
( f ot )
A
(2
m
+ 1 ................................................................................................ (6.18)
An approximation to damage under wide band stresses can be made by assuming that the damage is the
same as that caused by a narrow band stress having the same RMS, , and rate of zero crossings, fo. This
is called the equivalent narrow band process. Thus, for a wide band process, and fo are computed from
the spectral density function and Equation 6.18 is used to compute the damage.
The equivalent narrow band damage model provides a conservative estimate, but generally by less than a
factor of 10% on damage or life, depending upon the shape of the stress spectral density, and the slope of
the S-N curve.
However, it is generally considered that the rainflow method [Dowling (1972)] is the most accurate of several
models for counting stress cycles in a wide band process.
8.4
The Rainflow Method

Consider the segment of stress of Section 6, Figure 3. Rotate the picture 90 as shown. The method as
originally presented will be summarized.
FIGURE 3
Segment of Stress Process to Demonstrate Rainflow Method
S(t)
1
S(t)
2
6
8
3
1
11
S2
4
13
S1
9
5
12
10
11
13
10
12
56
Section
i)
Imagine each trough has a water source and water flows downward from off of the roofs.
ii)
The water path is interrupted when the path passes a trough which is more negative (e.g., 5) than
the original (e.g., 1). This path defines stress range S1 as shown. Note that the mean value of this
stress cycle is also defined.
iii)
A path (e.g., starting at point 3) ends when it hits another path as shown. This defines another stress
range S2.
iv)
The same process is pursued throughout the length of the available record.
v)
The process can be repeated by considering the peaks as water sources. The stress cycles generated
by the peak process should match the cycles of the trough generated process.
The result of rainflow counting is a random sample, Si; i = 1, P, where P is the number of peaks in the
process. Then:
S em = E ( S m ) =
S
i =1
m
i
.............................................................................................................. (6.19)
For additional information on the rainflow method, see Fatigue Design Handbook (1988) and BS 7608
(1993) for the reservoir method, equivalent to the rainflow.
8.5
A Closed Form Expression for Wide Band Damage

To estimate rainflow damage, models which employ a modified form of the equivalent narrow band process
have been developed, as described here. Three models have been proposed in the literature. The initial
model was presented by Wirsching and Light (1980). IN ADDITION, Two other models were presented
by Ortiz and Chen (1987) and Lutes and Larson (1990). The three models give comparable results for the
parameters of interest of welded joints. The models are based on the form:
D = DNB .................................................................................................................................. (6.20)
where , called the wide band correction factor, is a function of the indices of the spectral density function
and also the slope of the S-N curve, m. Thus, plays the role of a correction factor to narrow band damage.
The following expressions for rainflow damage is based on the assumptions:
i)
The S-N curve, NSm = A is defined down to S = 0 (i.e., there is no break in the curve).
ii)
The rainflow method defines fatigue stress cycles.
iii)
Mean stress is not a factor in welded joint fatigue. Mean stress associated with each cycle is ignored.
Wirsching and Light (1980)
i = a(m) = [1 a(m)](1 i)b(m) .............................................................................................. (6.21)

where
a(m) = 0.926 0.033m ............................................................................................................. (6.22)
b(m) = 1.587m 2.323 ............................................................................................................. (6.23)
Ortiz and Chen (1987)
km
.................................................................................................................................... (6.24)
where
k =
2k
........................................................................................................................ (6.25)
o k +2
and
k=
2.0
..................................................................................................................................... (6.26)
m
57
Section
Lutes and Larson (1990)
( 2 / m )m / 2
.......................................................................................................................... (6.27)
o+
Example: Consider a wide-band stress process defined by its spectral density function:
WS(f) = 0.05 0 < f < 20 Hz
0
otherwise
For the models described above, the following moments are calculated using Equation 6.13:
0 = 1.0
2 = 133.32
4 = 32,000
k = 2/m = 4.424
k+2 = 804.51
The rate of zero crossings and the rate of peaks are computed by Equation 6.12:
fp = 15.49
fo = 11.55
The irregularity factor and the spectral width parameter are given computed by Equations 6.11 and 6.14:
= 0.745
= 0.667
The three rainflow correction factors are:
= 0.839 (Wirsching and Light)
Equation 6.21
= 0.843 (Ortiz and Chen)
Equation 6.24
= 0.806 (Lutes and Larsen)
Equation 6.27
The Damage Calculation for Single Segment S-N Curve

Damage will be calculated in terms of the design life, NT (cycles) or T (time)
Consider a single sea state. Let ni be the number of cycles in this sea state over the service life. The damage
produced by this sea state is given by Equation 6.28, a restatement of Equation 6.2.
Di =
ni
2 2 i
A
m
+ 1 ..................................................................................................... (6.28)
2
Total damage can be written in terms of NT (cycles) or T (time) as follows. The number of cycles in the i-th
sea state can be expressed as the product of the time in the i-th sea state times the equivalent frequency.
ni = (pi T) foi (6.29)
pi is the fraction of time in the i-th sea state. Note that units of time must be compatible. If foi is in Hertz,
then T must be in seconds.
Thus, when the wide band correction is applied, the damage in the i-th sea state is derived from Equation
6.28 as:
Di =
( pi T )
A
i f oi 2 2 i
m
+ 1 ........................................................................................ (6.30)
2
The total damage at the design life is the sum of the damages in each of the M sea states.
D=
( )
T
2 2
A
m
+ 1
i f oi p i i ..................................................................................... (6.31)
2
i =1
An alternative form can be derived in terms of NT by noting that:

T=
58
NT
..................................................................................................................................... (6.32)
fo
Section
where fo is the average frequency over the service life. This is calculated as:
M
p f
fo
.......................................................................................................................... (6.33)
oi
i 1
Then total damage over the design life is:
NT
2 2
Af o
m
1
i f oi pi i ................................................................................. (6.34)
2
i 1
The safety check expression is then:
10
1.0
................................................................................................................................. (6.35)
FDF
The Damage Calculation for Bi-Linear S-N Curve

For fatigue strength defined by the bilinear curve of Section 5, Figure 4, damage in the i-th sea state can be
derived from Equation 5.23. Note that the Raleigh distribution is a special form of the Weibull distribution
with parameters as defined by Equation 6.1. Making the substitution of parameters, damage for the i-th sea
state is:
Di
r
N
N
2 2 i 0 b, z i
2 2 i
C
A
a, z i .................................................................... (6.36)
where
a
m
1 ................................................................................................................................. (6.37)
2
r
1 ................................................................................................................................... (6.38)
2
SQ
z
2 2
i
.......................................................................................................................... (6.39)
This damage expression is reformulated by making the following substitutions:

m = r m ................................................................................................................................ (6.40)
(a,z) = (a) 0(a,z) .............................................................................................................. (6.41)
A
1
................................................................................................................................. (6.42)
C S Qm
The latter is derived by writing Equations 5.21 at the common point (NQ, SQ). The gamma functions are
defined by Equations 5.3, 5.24, and 5.25. After making the above substitutions, using Equation 6.36 and
applying the wide band correction, damage in the i-th sea state is:
Di
pi T
A
i f oi 2 2 i
m
1 i ................................................................................... (6.43)
2
where
m / 2
1
o a, z i
zi
i 1
a
0 b, z
...................................................................................... (6.44)
59
Section
The total damage over the design life is obtained by summing over all M sea states:
D=
( )
T
2 2
A
m
+ 1 i i f oi p i im ................................................................................ (6.45)
i =1
2
As in the single segment case, T can be replaced by NT/fo.

The safety check is defined by Equation 6.35.
60
Section 7: Deterministic Method of Fatigue Assessment
SECTION
Deterministic Method of Fatigue Assessment
General
In the deterministic method of fatigue assessment, the long-term distribution of stresses is defined by what
is known historically as block loading. The long-term distribution of stress ranges is decomposed (either
by analysis or by signal processing equipment) into blocks of constant amplitude stresses with a corresponding
number of cycles. Section 1, Table 1, repeated here, illustrates deterministic stress spectra. The choice of
J blocks is arbitrary.
TABLE 1
Deterministic Stress Spectra
Stress Range
Si
Number of Cycles
ni
S1
n1
S2
n2
S3
n3
.
.
SJ-1
nJ-1
SJ
nJ
Applying the Palmgren-Miner linear cumulative damage:
D=
ni
N
i =1
And the safety check expression is:

D
1.0
................................................................................................................................... (7.1)
FDF
Application to a Self-Elevating Unit

Typically, the fatigue assessment of a self-elevating drilling unit (MODU) is approached on a deterministic
basis. The lack of specific knowledge of the conditions in which the unit will operate during its service life
indicate the use of the deterministic approach, which can produce notional, comparative results.
A format for a deterministic fatigue assessment method for a self-elevating unit that may be suitable for
Classification is the weather box method. In this deterministic method, the expected wave conditions that
the unit will be exposed to during its life are split into a finite number of regular wave height/ period
combinations. The minimum acceptable number of combinations would normally be five, but more would
be preferable. The unit is then assessed under these conditions for a number of different headings and
water depths. An acceptable set of assessment parameters is given in Section 7, Table 2.
61
Section
Deterministic Method of Fatigue Assessment
TABLE 2
Wave and Other Parameters to be Used in the Fatigue Assessment
Item
Number of wave height/period combinations
Number of Cases
At least 5 over a realistic range for anticipated areas
of operation
Wave steepness (Wave length divided by height)
Should normally be in the range of 12 to 15
Number of water depths
Normally 3 over a realistic range would be

acceptable
Number of weather approach directions
Normally 8, but 3 may be possible with symmetry
Leg fixity conditions
2: Pinned for upper leg, high fixity for lower leg
The calculation method can be summarized as:
62
Determine the wave/water depth/direction combinations for study.
Calculate the wave load ranges on the self-elevating unit for each of the chosen combinations. Given
the magnitude of most of the waves that inflict significant fatigue damage, it is usually acceptable to
use a linear wave theory.
Determine a suitable Dynamic Amplification Factor (DAF) for each wave/period and water depth
combination. It is anticipated that the Single Degree of Freedom approach (SDOF) would be used.
Apply the wave loads and suitable inertial loads to a detailed structural model of the unit and calculate
the stress ranges throughout the structure.
Increase stress ranges by a suitable Stress Concentration Factor (SCF), which is determined from either a
finite element analysis or suitable parametric equations. It is often adequate to use a single SCF for
each member end at the joint, and to effectively assume that the most highly stressed material is always
the same, regardless of loading details. Alternatively, if a more conservative yet representative fatigue
life is desired, then it is acceptable to keep track of the location of all of the stress ranges around each
joint.
Use Palmgren-Miner rule to sum the fatigue damage from each of the individual water depth/wave
direction combinations.
Repeat for changed foundation fixity, if warranted by preliminary estimates using simplified techniques.
Section 8: Fracture Mechanics Fatigue Model
SECTION
Fracture Mechanics Fatigue Model
Introduction
While ABS specifies the S-N curve method for fatigue assessment, the fracture mechanics fatigue model is
commonly used to assess remaining life after a crack is discovered. This model is useful for evaluating crack
growth and for inspection planning, but it also could be employed for design purposes. A Section in the Guide
is provided for basic reference to the fracture mechanics fatigue model. The intent of this Commentary
Section is to provide some background and supporting information.
The essence of the fracture mechanics approach is that there is a pre-existing crack or flaw. Subsequent
sub-critical crack growth occurs under oscillatory stress. The goal of analysis is to estimate the number of
stress cycles until the crack grows to such a size that instantaneous fracture occurs. A secondary goal,
important for developing inspection and repair strategy, is to estimate the crack depth for a given cycle life.
Crack Growth Model (Fatigue Strength)
2.1
Stress Intensity Factor Range

The fracture mechanics model for fatigue strength is based on crack growth data. As shown in Section 8,
Figure 1, the strength model is a relationship between the crack growth rate, da/dN (N is the number of
cycles and a is the depth of the crack) and the stress intensity factor range, K.
K = Y (a) S a ........................................................................................................................ (8.1)
where Y(a) is the geometry factor and S is the stress range. A constant amplitude stress process is assumed.
As the name implies, the geometry factor depends upon the geometry in the neighborhood of the crack and
the direction and type of loading. Catalogs of Y(a) are available. For example, tables of Y(a) are included
in IIW (1996), BSI 7910 (1999), BS 7608(1993), and API 579 (2001).
2.2
The Paris Law

A simplified description of the crack growth process is shown in Section 8, Figure 1.
The relationship between da/dN and K is generally observed to be a sigmoid curve on a log da/dN versus
log K plot. At low values of K (Region I), the rate of growth falls off rapidly to a threshold stress
intensity factor, Kth, below which no significant crack growth will occur. At high values of K (Region
III), when the maximum stress intensity factor in the cycle approaches the critical stress intensity factor for
failure under static load, the rate of crack growth accelerates rapidly.
There is a quasi-linear central portion (Region II). For engineering purposes, it is usually sufficiently accurate
to ignore the existence of the threshold and assume that the linear relationship applies for all values of K
from zero to failure.
63
Section
Fatigue Strength Based on Fracture Mechanics
FIGURE 1
A Model of Crack Propagation Rate versus Stress Intensity Factor Range
da
dN
Upper knee
typically
about 10-3 in/cycle
REGION III
High Crack
Growth Rate
Lower knee
typically
about 10-7 in/cycle
REGION II
Governed by
Paris Equation
da
= C(K)m
dN
REGION I
Low Crack
Growth Rate
K
Kth
Historical crack growth data has indicated that the Paris law is generally a reasonable model for the linear
central portion.
da
= C (K ) m ............................................................................................................................ (8.2)
dN
where C and m are the Paris coefficient and exponent, respectively, as determined from crack growth data.
2.3
Determination of the Paris Parameters, C and m

Crack growth data are widely published. The Paris parameters C, m depend on the material and the applied
conditions, such as stress ratio, environment, test frequency and waveform in the crack growth test. Whenever
possible, data relevant to the particular material under service conditions should be used and, where any
doubt exists concerning the influence of the environment, such data should be obtained. If the available
data are sufficient to define C and m, the chosen values should correspond to the mean plus two standard
deviation of log da/dN.
Gurney (1979) published an empirical form for the parameters of the Paris equation. The general relationship
for the median curve for steels is:
C=
64
1.315E 4
(895.4)m
(N, mm units) ............................................................................................ (8.3)
Section
For other units, the conversions are:

C (ksi units) = 0.0394 (34.75)m C (N, mm units)
C (ksi units) = 39.37 (1.1)m C (MPa units)
Most of the data on crack growth in structural steels and welds indicate that a default slope of m = 3.0 is
reasonable. Crack growth rates that are found in relevant engineering literature are presented in Section 8,
Table 1. Values are listed for structural steel in-air, except where indicated.
TABLE 1
Paris Parameters for Structural Steel
C (N, mm units) corresponding to m = 3.0
Notes
3.66E-13
Gurney (1979); median value of Equation 8.3 is multiplied by 2.0 to

account for scatter [BS 7608(1993)]
3.00E-13
Upper bound to many published data for welds and HAZ in which crack
growth was the striation mechanism [BS 7608(1993)]
6.00E-13
Normal crack growth accompanied by cleavage or microvoid

coalescence. [BS 7608(1993)]
2.30E-12
Steels operating in a marine environment. [BS 7608(1993)]
6.74E-13
Maddox (1974); upper bound for weld material
2.17E-13
Barsom (1971); upper bound for A36 steel
3.00E-13
Department of Energy Review Panel (otherwise undocumented)
1.00E-12
In seawater; Department of Energy Review Panel (otherwise

undocumented)
These values are provided only for reference purposes.
Life Prediction
3.1
Relationship Between Cycles and Crack Depth

The analysis objective is cycles to failure, or alternatively, the crack size associated with a given life. In
doing so, it is assumed that the real flaws can be idealized as sharp-tipped cracks.
Integration of Equation 8.2 gives the number of cycles, N, required for a crack to propagate from an initial
size, ai, to a crack depth a:
NS m =
1
C
ai
1
m
( x)(x) m / 2
dx ..................................................................................................... (8.4)
When a = ac, the critical crack depth, failure is assumed, and N would be cycles to failure.
Note that this form is identical to the characteristic S-N model of Chapter 3, NSm = A. The fatigue strength
coefficient, A, will be equal to the right hand side of Equation 8.4.
3.2
Determination of Initial Crack Size, ai

This Section provides information for those special cases where the fracture mechanics model would be
employed for design.
The fracture mechanics model critically depends upon the value of the initial flaw size ai. It is therefore
important that ai is not underestimated. When a crack is discovered by one of the inspection methods, its
measurement is subject to uncertainty and it is important for the analyst to account for the uncertainty.
65
Section
In the context of design, an assumption of initial crack size must be made. From the codes:
[DNV(2000)] For surface cracks starting from transition between weld/base material, a crack depth of 0.5 mm
(e.g., due to undercuts and microcracks at bottom of the undercuts) may be assumed if no other documented
information about crack depth is available.
[BS 7608(1993)] For normally flaw-free welded joints failing from the weld toe, it is recommended that ai
be in the range 0.10 mm to 0.25 mm unless a larger size is known to be relevant.
With regard to flaw shape, also from BS 7608 (1993) In the absence of definite information of the shape
of the initial flaws, it is also recommended that, for joints/details with welds transverse to the direction of
stress, the flaw at the weld toe is assumed to be long and continuous, i.e.:
ai
0 ........................................................................................................................................ (8.5)
2c
At the ends of longitudinal welds, however, it would be realistic to assume that the initial flaw is a semielliptical in shape with:
ai
0.1 ...................................................................................................................................... (8.6)
2c
The initial crack size to be used in the calculation shall be evaluated in each case, taking into account the
defect size for various weldments, geometries and the inspection accuracy.
Because life is sensitive to the assumed value of ai, it is important that ai not be underestimated.
3.3
66
Determination of the Failure (Critical) Crack Length, ac.

The analyst has will use judgment with regard to the choice of ac. Each of the following failure modes
should be considered carefully in making this choice:
i)
Unstable brittle fracture
ii)
Yielding of the remaining section
iii)
Leakage
iv)
Stress corrosion
v)
Instability
vi)
Creep
Section 9: References
SECTION
References
1. ABS (2001a) Rules for Building and Classing Steel Vessels, (2001), American Bureau of Shipping,
Houston, TX
2. API 579 (2001), API Recommended Practice for Fitness for Service, (2000) The American Petroleum
Institute, Houston, TX
3. API (2001a), RP2A-WSD: Planning, Designing and Constructing Fixed Offshore Platforms
Working Stress Design, The American Petroleum Institute, Houston, TX
4. API (2001b), RP2A-LRFD: Planning, Designing and Constructing Fixed Offshore Platforms Load
and Resistance Factor Design, The American Petroleum Institute, Houston, TX
5. API (2001c), RP2T, Recommended Practice for Planning, Designing, and Constructing Tension Leg
Platforms, The American Petroleum Institute, Houston, TX
6. AWS (2002); Structural Welding Code-Steel, American Welding Society Specification ANSI/AWS
D1.1M, 18th Edition
7. Barsom, J.M. (1971), Fatigue Crack Propagation in Steels of Various Yield Strengths, Journal of
Engineering for Industry, ASME, Vol 7, No. 4
8. BSI 7910 (1999), Guidance on Methods of Assessing the Acceptability of Flaws in Fusion Welded
Structures, Health and Safety Executive (HSE), British Standards Institute
9. BS 7608 (1993), Code of Practice for Fatigue Design and Assessment of Steel Structures, British
Standards Institution, 2 Park Street, London, W1A 2BS
10. DEn (1990); Offshore Installations, Guidance on Design, Construction, and Certification, Department
of Energy, UK, 4th Edition, London, HMSO
11. DNV (2000) Fatigue Strength Analysis of Offshore Steel Structures, RP-C203, Det Norske Veritas,
Hovik, Norway
12. DNV (1998), Fatigue Assessment of Steel Structures, Classification Notes No. 30.7, Det Norske
Veritas, Hovik, Norway
13. Dowling, N.E., (1972), Fatigue Failure Predictions of Complicated Stress-Strain Histories, Journal
of Materials, ASTM, Vol. 7, No.1
14. Eurocode 3, (1992); Design of Steel Structures, European Committee for Standardization, Brussels,
Belgium
15. Fatigue Design Handbook, AE-10, (1988), Society of Automotive Engineers, Warrendale, PA
16. Gurney, T.R., (1979), Fatigue of Welded Structures, 2nd Ed, Cambridge Univ. Press, New York
17. HSE (1995), Offshore Installations Guidance on Design, Construction, and Certification, Third Amendment
to Fourth Edition, Health and Safety Executive British Standards Institute London
18. IIW (1996); Fatigue Design of Welded Joints and Components, Report XIII-1539-96/XV-845-96, The
International Institute of Welding, Abington Publishing, Cambridge, England
19. ISO/CD 19902 (2000). Petroleum and Natural Gas Industries Offshore Structures Fixed Steel
Structures, Section A16, International Standards Organization. (Draft Document)
20. Lotsberg, I. and Larsen, P. K. (2001), Developments in Fatigue Design Standards for Offshore Structures,
ISOPE 2001, Stavager, Norway
67
Section
References
21. Lutes, L. and Larson, M.L., (1990), Improved Spectral Method for Variable Amplitude Fatigue Prediction,
Journal of the Structural Division, ASCE, Vol. 116, pp 1149-1164
22. Maddox, S.J. (1974), Assessing the Significance of Flaws in Welds Subject to Fatigue, Welding
Journal 53, No. 9
23. Nolte, K.G. and Hansford, J.E., (1976), Closed Form Expressions for Determining the Fatigue Damage
of Structures Due to Ocean Waves, Proceedings of the Offshore Technology Conference, OTC Paper
2606,
24. Ortiz and Chen (1987), Fatigue Damage Prediction for Stationary Wide Band Stresses, 5th International
Conference on Applications of Statistics and Probability in Civil Engineering (ICASP), June 8-12.
Vancouver, BC, Canada
25. Stacey, A. and Sharp, J. V. (1995) The Revised HSE Fatigue Guidance, 1995 OMAE Volume III,
Material Engineering, ASME
26. Wirsching, P.H. and Light, M.C. (1980), Fatigue Under Wide Band Stresses, Journal of the Structural
Division, ASCE, Vol 106, pp 1593-1607
27. Wirsching, P.H. and Chen, Y.N. (1988), Considerations of Probability Based Fatigue Design for
Marine Structures, Marine Structures, 1, pp. 23-45
28. Wirsching, P.H., Paez, T.L., and Ortiz, K., (1995), Random Vibratons, Theory and Practice, Wiley
Interscience
68

Fatigue Assessment of Offshore Structures Commentary On The Guide For The

Uploaded by

Copyright:

Available Formats

Fatigue Assessment of Offshore Structures Commentary On The Guide For The

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

Fatigue Assessment of Offshore Structures Commentary On The Guide For The

Uploaded by

Copyright:

Available Formats

Commentary on the Guide for the Fatigue Assessment of Offshore Structures (April 2003)

COMMENTARY ON THE GUIDE FOR THE

FATIGUE ASSESSMENT OF OFFSHORE STRUCTURES

JANUARY 2004 (Updated July 2014 see next page)

American Bureau of Shipping

COMMENTARY ON THE GUIDE FOR THE

FATIGUE ASSESSMENT OF OFFSHORE STRUCTURES

Deterministic Stress Spectra ..................................................... 2

Tubular Joints: Statistics on Damage at Failure,

Plated Joints: Statistics on Damage at Failure,

Fatigue Strength Based on S-N Curves General Concepts ............. 4

Details of the Basic In-Air S-N Curves....................................6

An Example of S-N Fatigue Data Showing the Least

Weld Toe Extrapolation Points for a Tubular Joint ................... 8

S-N Curves .............................................................................................. 9

Tubular Intersection Connections ..................................................... 11

Without Weld Profile Control ......................................................... 11

With Weld Improvement ................................................................ 12

Plated Connections ........................................................................... 13

Discussion of the Thickness Effect ................................................... 14

Fatigue Test Data on Plated Joints ............................................... 15

Design F-Curves with Thickness Adjustment ................................ 15

Thickness Adjustments to Test Data and Their Regressed

Effects of Corrosion on Fatigue Strength.......................................... 28

Preliminary Remarks ..................................................................... 28

A Summary of the Results ............................................................. 28

The Summaries ............................................................................. 28

Coverage of the Two Main Sources of S-N Curves Used

AWS-HSE/DEn Curves for Similar Detail Classes ................. 13

Parameters of Plate Thickness Adjustment for Plated

Parameters of Plate Thickness Adjustment for Tubular

Parameters of F-curves .......................................................... 15

Details of Basic Design S-N Curves HSE(1995) .................... 29

Life Reduction Factors to be Applied to the Lower Cycle

Life Reduction Factors to be Applied to the Lower Segment

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

F-Curves with Thickness Adjustment and Test Data;

Test data with DEn(1990) Thickness Adjustment and their

Test Data with HSE(1995) Thickness Adjustment and their

Test Data with DNV(2000) Thickness Adjustment and their

Regressed S-N Curves and Design F-curves ......................... 27

Fatigue Design Factors ........................................................................ 31

The Simplified Fatigue Assessment Method ..................................... 34