SSC-436 The Effect of Fabricaton Tolerances On Fatigue Life of Welded Joints

NTIS # PB2005-101516
SSC-436
THE EFFECT OF FABRICATON
TOLERANCES ON FATIGUE LIFE OF
WELDED JOINTS
This document has been approved

For public release and sale; its
Distribution is unlimited
SHIP STRUCTURE COMMITTEE
2005
SHIP STRUCTURE COMMITTEE

RADM Thomas H. Gilmour
U. S. Coast Guard Assistant Commandant,
Marine Safety and Environmental Protection
Chairman, Ship Structure Committee
Mr. W. Thomas Packard
Director,
Survivability and Structural Integrity Group
Naval Sea Systems Command
Dr. Jack Spencer

Senior Vice President
American Bureau of Shipping
Mr. Joseph Byrne

Director, Office of Ship Construction
Maritime Administration
Mr.Gerard A. McDonald
Director General, Marine Safety,
Safety & Security
Transport Canada
Mr. Thomas Connors

Director of Engineering
Military Sealift Command
Dr. Neil Pegg

Group Leader - Structural Mechanics
Defence Research & Development Canada - Atlantic
CONTRACTING OFFICER TECHNICAL REP.

Chao Lin / MARAD
Natale Nappi / NAVSEA
Robert Sedat / USCG
EXECUTIVE DIRECTOR
Lieutenant Eric M. Cooper
U. S. Coast Guard
SHIP STRUCTURE SUB-COMMITTEE

AMERICAN BUREAU OF SHIPPING
Mr. Glenn Ashe
Mr. Yung Shin
Mr. Phil Rynn
Mr. William Hanzalek
DEFENCE RESEARCH & DEVELOPMENT ATLANTIC

Dr David Stredulinsky
Mr. John Porter
MARITIME ADMINISTRATION
Mr. Chao Lin
Mr. Carlos Setterstrom
Mr. Richard Sonnenschein
MILITARY SEALIFT COMMAND

Mr. Joseph Bohr
Mr. Rick A. Anderson
Mr. Michael W. Touma
NAVAL SEA SYSTEMS COMMAND

Mr. Jeffery E. Beach
Mr. Edward E. Kadala
Mr. Allen H. Engle
Mr. Charles L. Null
TRANSPORT CANADA
Mr. Jacek Dubiel
UNITED STATES COAST GUARD

Mr. Rubin Sheinberg
Mr. Robert Sedat
Commander Ray Petow
CANADIAN COAST GUARD

Mr. Daniel Gauvin
1. Report No.
2. Government Accession No.
SSC-436
PB 2005-101516
Technical Report Documentation Page

3. Recipients Catalog No.
4. Title and Subtitle
5. Report Date
Effect of Fabrication Tolerances on Fatigue Life of Welded

Joints
December 2004
7. Author(s)
8. Performing Organization Report No.
A. Kendrick
SR-1428
9. Performing Organization Name and Address
10. Work Unit No. (TRAIS)
BMT Fleet Technology Ltd.

311 Legget Drive
Kanata, Ontario, Canada K2K 1Z8
11. Contract or Grant No.
12. Sponsoring Agency Name and Address
13. Type of Report and Period Covered
Ship Structure Committee

U.S. Coast Guard (G-MSE/SSC)
2100 Second Street, SW
Washington, DC 20593
Final Report
6. Performing Organization Code
14. Sponsoring Agency Code
G-M
15. Supplementary Notes
Sponsored by the Ship Structure Committee. Jointly funded by its member agencies.
16. Abstract
A review of available data on fabrication tolerances achieved by shipyards has been undertaken and
compared to assumptions and methodologies used in various marine fatigue analysis standards and
guidelines. New shipyard data has been gathered to supplement the extremely small amount of prior
data available in the public domain.
The project has developed procedures for tolerance data collection and analysis. It has explored the
implications of actual tolerances for fatigue life, and has compared the results with those from
standard methods using default tolerances and assumptions. A set of recommendations have been
developed, covering tolerance measurement techniques that could be used in future extensions of the
work, and areas in which the results of this project can be used in analysis and in the development of
inspection strategies.
17. Key Words
18. Distribution Statement
Distribution is available to the public through:

National Technical Information Service
U.S. Department of Commerce
Springfield, VA 22151 Ph. (703) 487-4650
19. Security Classif. (of this report)
20. Security Classif. (of this page)
Unclassified
Unclassified
21. No. of Pages
22. Price
CONVERSION FACTORS
(Approximate conversions to metric measures)
To convert from
LENGTH
inches
inches
feet
VOLUME
cubic feet
cubic inches
SECTION MODULUS
inches2 feet2
inches2 feet2
inches4
MOMENT OF INERTIA
inches2 feet2
inches2 feet2
inches4
FORCE OR MASS
long tons
long tons
pounds
pounds
pounds
PRESSURE OR STRESS
pounds/inch2
kilo pounds/inch2
BENDING OR TORQUE
foot tons
foot pounds
foot pounds
ENERGY
foot pounds
STRESS INTENSITY
kilo pound/inch2 inch(ksiin)
J-INTEGRAL
kilo pound/inch
kilo pound/inch
to
Function
Value
meters
millimeters
meters
divide
multiply by
divide by
39.3701
25.4000
3.2808
cubic meters
cubic meters
divide by
divide by
35.3149
61,024
centimeters2 meters2
centimeters3
centimeters3
multiply by
multiply by
multiply by
1.9665
196.6448
16.3871
centimeters2 meters
centimeters4
centimeters4
divide by
multiply by
multiply by
1.6684
5993.73
41.623
tonne
kilograms
tonnes
kilograms
Newtons
multiply by
multiply by
divide by
divide by
multiply by
1.0160
1016.047
2204.62
2.2046
4.4482
Newtons/meter2 (Pascals)
mega Newtons/meter2
(mega Pascals)
multiply by
multiply by
6894.757
6.8947
meter tons
kilogram meters
Newton meters
divide by
divide by
multiply by
3.2291
7.23285
1.35582
Joules
multiply by
1.355826
mega Newton MNm3/2
multiply by
1.0998
Joules/mm2
kilo Joules/m2
multiply by
multiply by
0.1753
175.3
BMT FLEET TECHNOLOGY LIMITED
5354C.FR
ABSTRACT
A review of available data on fabrication tolerances achieved by shipyards has been undertaken
and compared to assumptions and methodologies used in various marine fatigue analysis
standards and guidelines. New shipyard data has been gathered to supplement the extremely
small amount of prior data available in the public domain.
The project has developed procedures for tolerance data collection and analysis. It has explored
the implications of actual tolerances for fatigue life, and has compared the results with those
from standard methods using default tolerances and assumptions. A set of recommendations
have been developed, covering tolerance measurement techniques that could be used in future
extensions of the work, and areas in which the results of this project can be used in analysis and
in the development of inspection strategies.
The Effect of Fabrication Tolerances on Fatigue Life of Welded Joints
iii
5354C.FR
TABLE OF CONTENTS
1.
INTRODUCTION ................................................................................................................... 1
1.1 General ........................................................................................................................... 1
1.2 Report Outline................................................................................................................ 1
2.
OBJECTIVES AND APPROACH.......................................................................................... 2

2.1 Objectives....................................................................................................................... 2
2.2 General Approach .......................................................................................................... 2
2.2.1 Task 1: Data Collection..................................................................................... 2
2.2.2 Task 2: Statistical Analysis ............................................................................... 2
2.2.3 Task 3: Identification of Effect on Fatigue Life................................................ 3
3.
TECHNICAL BACKGROUND ............................................................................................. 4

3.1 Overview ........................................................................................................................ 4
3.2 Classification of Imperfections in Welded Joints .......................................................... 6
3.2.1 Planar and Non Planar Flaws ............................................................................. 7
3.2.2 Geometric Imperfections.................................................................................... 8
3.3 Use of S-N Curves in Design ......................................................................................... 8
3.3.1 Nominal Stress Approach .................................................................................. 9
3.3.2 Hot Spot Stress Approach .................................................................................. 9
3.3.3 Notch Stress Approach..................................................................................... 10
3.3.4 Axial and Angular Misalignment Imperfections - Application in S-N Curve
Analysis............................................................................................................ 18
3.4 Sensitivity in Fatigue Life Calculations....................................................................... 21
3.5 Implications for the Current Project............................................................................. 23
4.
DATA COLLECTION .......................................................................................................... 24

4.1 Literature Review......................................................................................................... 24
4.2 Field Data Collection ................................................................................................... 25
4.2.1 Summary of Shipyard Survey #1 ..................................................................... 25
4.2.2 Summary of Shipyard Survey #2 ..................................................................... 27
4.3 Survey Techniques ....................................................................................................... 29
4.3.1 Misalignment in Butt Welds ............................................................................ 30
4.3.2 Weld Reinforcement in Butt Welds ................................................................. 31
4.3.3 Weld Toe Angle in Butt Welds........................................................................ 31
4.3.4 Angular Misalignment in Butt Welds .............................................................. 32
4.3.5 Cruciform Joint Misalignment ......................................................................... 33
4.3.6 Angular Misalignment in Cruciform Joints ..................................................... 34
4.3.7 Weld Size in Cruciform Joints ......................................................................... 34
5.
STATISTICAL ANALYSIS OF SURVEY DATA .............................................................. 36

5.1 Background .................................................................................................................. 36
5.2 Methodology ................................................................................................................ 36
5.2.1 Descriptive Statistics and Histograms.............................................................. 36
5.2.2 Total Stress Concentration Factor.................................................................... 37
5.2.3 Confidence Interval on the Mean Values......................................................... 38
5.2.4 Sample Size Determination.............................................................................. 39
iv
5.3
5.4
5.5
5.6
5.7
5354C.FR
Butt-Welded Plates....................................................................................................... 40
5.3.1 Shipyard #1 Data.............................................................................................. 40
5.3.2 Shipyard #2 Data.............................................................................................. 42
Butt-Welded Stiffeners Shipyard #2 Data................................................................. 45
5.4.1 Stress Concentration Factor Kt ....................................................................... 45
5.4.2 Stress Concentration Factor Kw ....................................................................... 46
5.4.3 Stress Concentration Factor Kte....................................................................... 48
Cruciform Joints Shipyard #2 Data........................................................................... 49
5.5.1 Stress Concentration Factor Kte ....................................................................... 49
5.5.2 Cruciform Fillet Weld ................................................................................... 50
5.5.3 Cruciform Full Penetration Weld.................................................................. 51
Total Stress Concentration Factor................................................................................ 52
5.6.1 Butt-Welded Plates........................................................................................... 53
5.6.2 Butt-Welded Stiffeners..................................................................................... 55
5.6.3 Cruciform Joints............................................................................................... 56
Summary and Discussion of the Results...................................................................... 58
6.
DISCUSSION OF DATA...................................................................................................... 62
6.1 Overall Stress Concentration Factors........................................................................... 62
6.1.1 Stress Concentration Factor Kt ....................................................................... 64
6.2 Variability .................................................................................................................... 65
7.
FATIGUE LIFE CASE STUDY ........................................................................................... 68

7.1 Basis for Analysis ........................................................................................................ 68
7.2 Case Outcomes............................................................................................................. 71
8.
CONCLUSIONS ................................................................................................................... 74
9.
RECOMMENDATIONS....................................................................................................... 76
APPENDIX A: BIBLIOGRAPHY ............................................................................................ A-1

APPENDIX B: SHIPYARD DATA.......................................................................................... B-1
Shipyard 1 Butt-Welded Stiffeners
Shipyard 1 Butt-Welded Plates (Raw Data)
Shipyard 1 Butt-Welded Stiffeners (Raw Data)
Shipyard 2 Intercostal Joints (Raw Data)
Shipyard 2 Butt-Welded Plates (Raw Data)
5354C.FR
LIST OF FIGURES
Figure 3.1: Typical S-N Curve....................................................................................................... 5
Figure 3.2: Common Geometrical and Internal Weld Discontinuities .......................................... 7
Figure 3.3: Butt-Welded Plate Misalignment - Variables ........................................................... 14
Figure 3.4: Butt-Welded Plate Angular Misalignment - Variables ............................................. 14
Figure 3.5: Butt-Welded Plate Weld Toe Angle - Variables ....................................................... 14
Figure 3.6: Butt-Welded Plate Weld Reinforcement - Variables ................................................ 15
Figure 3.7: Cruciform Joint Misalignment Variables ............................................................... 17
Figure 3.8: Cruciform Joint Weld Geometry Variables............................................................ 17
Figure 3.9: Cruciform Joint Angular Misalignment Variables................................................. 17
Figure 3.10: Angularly Misaligned Plating - Seam Halfway between Supports......................... 18
Figure 3.11: Angularly Misaligned Plating - Seam at Location other than Halfway between
Supports ..................................................................................................................... 20
Figure 3.12: Effect of Misalignment on Fatigue Strength (Maddox, 1991) ................................ 21
Figure 3.13: Fatigue Life Influence of Stress Level and S-N Data for Welded Connections ..... 22
Figure 3.14: Fatigue Life Sensitivity to Stress Concentration Factor K and Weibull
Shape Factor H .......................................................................................................... 23
Figure 4.1: Measuring Alignment and/or Weld Reinforcement Height ...................................... 30
Figure 4.2: Tool for Measuring Weld Reinforcement Height and/or Plate Alignment and
Fillet Weld Leg Size ................................................................................................... 31
Figure 4.3: Measuring Weld Toe Angle ...................................................................................... 32
Figure 4.4: Measuring Angular Misalignment............................................................................. 32
Figure 4.5: Modified Protractor for Measuring Angular Misalignment ...................................... 33
Figure 4.6: Measuring Misalignment in Cruciform Joints........................................................... 33
Figure 4.7: Protractor for Measuring Angular Misalignment in Cruciform Joints...................... 34
Figure 5.1: Histogram for Kw with Extreme-Value and Normal Distribution Fits ...................... 41
Figure 5.2: Histogram for Kte with Normal and Lognormal Distribution Fits............................. 42
Figure 5.3: Histogram of Kt with Lognormal and Exponential Distribution Fits ...................... 43
Figure 5.4: Histogram of Kw with Normal and Extreme-Value Distribution Fits ....................... 44
Figure 5.5: Histogram of Kte with Lognormal and Weibull Distribution Fits ............................. 45
Figure 5.6: Histogram of Kt with Exponential and Extreme Value (Type I Gumbel largest)
Distribution Fits ......................................................................................................... 46
Figure 5.7: Histogram of Kw with Normal and Extreme Value (Type I Gumbel largest)
Figure 5.8: Histogram of Kte with Exponential and Extreme Value (Type I Gumbel largest)
Figure 5.9: Histogram of Kte with Extreme-Value and Exponential Distribution Fits ................ 50
Figure 5.10: Histogram of Kg Kw with Normal and Weibull (smallest) Distribution Fits for
Fillet Weld ................................................................................................................. 51
Figure 5.11: Histogram of Kg Kw with Lognormal and Weibull Distribution Fits for Full
Penetration Weld ....................................................................................................... 52
Figure 5.12: Histogram of total Stress Concentration Factor K with Normal and Lognormal
Distribution Fits for Butt-Welded Plates (Shipyard #1) ............................................ 54
Figure 5.13: Histogram of Total Stress Concentration Factor K with Normal and Lognormal
Distribution Fits for Butt-Welded Plates (Shipyard #2) ............................................ 55
vi
5354C.FR
Figure 5.14: Histogram of Total Stress concentration factor K with Lognormal and Normal
Distribution Fits for Butt-Welded Stiffeners (Shipyard #2) ...................................... 56
Distribution Fits for Cruciform Joints, Fillet Weld (Shipyard #2) ............................ 57
Distribution Fits for Cruciform Joints, Full Penetration Weld (Shipyard #2)........... 58
Figure 6.1: Typical Butt Weld ..................................................................................................... 64
Figure 6.2: S-N Design Curves Showing 2SD Offsets ................................................................ 66
Figure 7.1: Example VLCC (from SSC 427)............................................................................... 68
Figure 7.2: VLCC Notional Stress History.................................................................................. 69
Figure 7.3: Kt Adjusted for Thickness, Fixity............................................................................ 70
Figure 7.4: Case 1 Fatigue Life Analysis with Ktotal = Kt.Kte.Kw ............................................... 71
Figure 7.5: Case 2 Fatigue Life Analysis with Ktotal = Kte.Kw ..................................................... 72
Figure 7.6: Case 3 Fatigue Life Analysis with Ktotal = Kw ........................................................... 72
vii
5354C.FR
LIST OF TABLES
Table 3.1: Stress Concentration Factors Applicable to Butt-Welded Plates and Stiffeners of
Same Thickness .......................................................................................................... 13
Table 3.2: Stress Concentration Factors Applicable to Fillet Welded Cruciform Joints............. 16
Table 5.1: Statistics on Stress Concentration Factor Kw based on Shipyard #1 Data.................. 40
Table 5.2: Statistics on Stress Concentration Factor Kte based on Shipyard #1 Data.................. 41
Table 5.3: Statistics on Stress Concentration Factor Kt based on Shipyard #2 Data ................. 42
Table 5.6: Statistics on Stress Concentration Factor Kt based on Shipyard #2 Data ................. 46
Table 5.10: Statistics on Stress Concentration Factor Kg Kw based on Shipyard #2 Data
(for Fillet Weld) .......................................................................................................... 51
Table 5.11: Statistics on Stress Concentration Factor Kg Kg based on Shipyard #2 Data
(for Full Penetration Weld)......................................................................................... 52
Table 5.12: Statistics on Total Stress Concentration Factor K for Butt-Welded Plates
based on Shipyard #1 Data ......................................................................................... 53
Table 5.14: Statistics on Total Stress Concentration Factor K for Butt-Welded Stiffeners
Table 5.15: Statistics on Total Stress Concentration Factor K for Cruciform Joints
(Fillet Weld) based on Shipyard #2 Data.................................................................... 56
Table 5.16: Statistics on Total Stress Concentration Factor K for Cruciform Joints (Full
Penetration Weld) based on Shipyard #2 Data ........................................................... 57
Table 5.17: Summary of the Statistical Analyses for the Individual Stress Concentration
Factors Ks based on Shipyard 1 & 2 Data .................................................................. 60
Table 5.18: Summary of the Statistical Analyses for the Total Stress Concentration Factors Ks
based on Shipyard 1 & 2 Data .................................................................................... 61
Table 6.1: Comparison of Survey K Values to DNV Defaults.................................................... 63
Table 6.2: Effect of Fixed vs. Free and Assumption on SCFs..................................................... 65
Table 6.3: 2SD Equivalent SCFs ................................................................................................. 66
Table 7.1: Fatigue Life Prediction Comparisons ......................................................................... 73
viii
5354C.FR
LIST OF ABBREVIATIONS AND SYMBOLS

2SD
ABS
AHTS
AWS
BMT
BS
COV
DNV
FE
FOSM
HSE
IACS
ISO
MODU
PSF
SCF
SSC
TWI
VLCC
x
t
2
,n
K
K t
Kg
Kn
Kte
Kw
n
S
Two Standard Deviations

American Bureau of Shipping
Anchor Handling Tug Supply Vessel
American Welding Society
British Maritime Technology
British Standard
Coefficient of Variation
Det Norske Veritas
Finite Element
First Order Second Moment
Health & Safety Executive (UK)
International Association of Classification Societies
International Standard of Operation
Mobile Offshore Drilling Unit
Partial Safety Factors
Stress Concentration Factor
Ship Structure Committee
The Welding Institute
Very Large Crude Oil Carrier
sample mean
t-distribution value at an exceedance area of /2
level of significance in the range 0 to 1
population mean
population standard deviation
stress concentration factor
additional stress concentration factor due to angular mismatch (normally used for
plate connections only)
Stress concentration factor due to the gross geometry of the detail considered
additional stress concentration factor for un-symmetrical stiffeners on laterally loaded
panels, applicable when the nominal stress is derived from simple beam
additional stress concentration factor due to eccentricity tolerance (normally used for
plate connections only)
stress concentration factor due to the weld geometry
sample size
sample standard deviation
ix
1.
5354C.FR
INTRODUCTION
1.1
General
Fatigue life estimation at the design and approval stage is based on assumptions regarding
fabrication quality. Most shipbuilding standards include limits on fabrication tolerances such as
misalignment, weld profile defects, etc. However, there is limited information available on the
extent to which these standards are actually met. There is also limited understanding amongst
most designers, builders and owners regarding the extent to which fabrication tolerances may
influence life expectancy and/or through-life maintenance costs.
Fatigue life (S-N) curves used in most fatigue prediction methodologies incorporate assumptions
related to fabrication tolerances. However, these are often derived from other industries, and do
not necessarily represent shipbuilding practice. With a better understanding of actual shipyard
fabrication tolerances, it should be possible to improve the prediction accuracy of fatigue
analyses, and potentially to link the selection of construction standards to life expectancy
assessment and through-life maintenance cost.
The Ship Structure Committee has, therefore, sponsored this project to investigate actual
shipbuilding tolerances, to compare these with standard assumptions in fatigue analysis, and to
establish their significance.
1.2
Report Outline
Section 2 of this report defines the project objectives and the general approach that has been
adopted in order to fulfill these. Section 3 presents a general description of fatigue life analysis,
and how fabrication tolerances have typically been incorporated in design methodologies.
Sections 4 - 7 then describe the work undertaken in this project in order to meet the original
objectives, and to account for some of the challenges encountered in the course of the work.
Conclusions and recommendations for further work in this area are provided at Sections 8 and 9.
Detailed data and analyses are provided as a set of Appendices to this report.
2.
5354C.FR
OBJECTIVES AND APPROACH
2.1
Objectives
As outlined in the Introduction, the general objective of this project has been to conduct a study
of the effect of fabrication tolerances on the fatigue performance of welded details in ship
construction. The full extent of anticipated benefits of this could include:
More realistic fatigue life predictions, based on better information;

Better demonstration of the effects of construction tolerances on long-term vessel
performance or maintenance requirements;
Improved understanding of the costs/benefits associated with tighter fabrication
tolerances and resulting life maintenance costs; and
Improved ability to evaluate a particular yards quality control or construction tolerance
performance.
This initial project was expected to focus on the first two of these potential outcomes.
As work progressed on the project, the detailed objectives were refined to incorporate the
development of recommendations in a number of areas that will need attention before the full
range of potential benefits can be realized. These include:
a) development of improved data collection protocols and tools;
b) additional analysis of in-service experience; and
c) consideration of erection sequencing to optimize achieved tolerances.
These issues are also addressed in the project report.
2.2
General Approach
The project was planned to encompass three main tasks, as outlined below:
2.2.1 Task 1: Data Collection
This task was to encompass a literature review of applicable data for fabrication or structural
tolerance data for welded ship structural details, and limited shipyard surveys to collect new data.
2.2.2 Task 2: Statistical Analysis

This task was to compute statistical distributions describing the variability in welded ship
structural details and construction tolerances.
5354C.FR
2.2.3 Task 3: Identification of Effect on Fatigue Life

This task was to demonstrate the effects of actual and design tolerances on normal and leak
stresses and on design life.
While these principal tasks were accomplished, the more detailed sub-tasks within them were
revised as the project progressed in order to take account of early findings regarding the (lack of)
data available from prior work and the difficulty in obtaining new data. The project report
describes the work actually undertaken and the rationale for certain changes in focus during the
project.
3.
5354C.FR
TECHNICAL BACKGROUND
3.1
Overview
Many materials, including steel, when subjected to repeated strains (of sufficient magnitude) will
weaken and eventually initiate cracking. If repeated loading continues, the cracks will grow
through the member thickness and increase in length. The development of these cracks through
this process is termed fatigue crack initiation and growth. Fatigue damage in large structures
such as ships normally accumulates most rapidly at joints or discontinuities, where stresses are
raised above those in the surrounding structure by local effects.
Ships operate in environments that apply variable amplitude loading cycles, meaning that the
structural components of a ship will experience repeated loading or strain events throughout the
life of the ship. Fatigue cracking in ship structure can therefore be a serious safety and monetary
issue if it occurs. Much effort has gone into the study of fatigue problems with ship structure.
Fatigue is a complex problem primarily related to structural geometry, with secondary links to
material properties. Due to the inherent variability of the fatigue damage accumulation process,
pure analytic procedures cannot accurately predict the occurrence of fatigue failure in a real
structure. All methodologies are based upon empirical results from fatigue testing representative
samples. The normal process is to create (or extract from a real structure) a set of test specimens,
and then apply constant amplitude cyclic loads until a crack appears. The number of cycles
before failure is the fatigue life for that specimen. No two specimens are alike, thus even under
the most controlled conditions each specimen will fail after a different number of cycles. By
performing many tests however, and then fitting a curve through the data with consideration
given to the test result variance, it is possible to predict within given confidence limits when
failure will occur for a particular set of specimens. This data is normally presented in the form
of an S-N curve such as shown in Figure 3.1.
5354C.FR
Figure 3.1: Typical S-N Curve

(Reprinted from Re-evaluation of fatigue curves for flush-ground girth welds, 2003.
Prepared by TWI for HSE.)
Many structures, including ships, will not experience constant amplitude cyclic loading as was
used on the test specimens. It was established by Miner-Palmgren that the amount of fatigue
damage accumulated during each load cycle is proportional to the stress during that cycle. The
Miner-Palmgren law (hypothesis) can be stated as:
For a particular stress range, S1, the constant amplitude endurance, N1, is a
measure of the fatigue damage as a result of S1, as applied for n1 cycles, is n1/N1
of that needed to cause failure. Failure occurs under a sequence of different
stresses when the sum of all ratios n/N equals unity. That is:
n
n
n1 n2
+
+ 3 + ... = i = 1 at failure
N1 N 2 N 3
Ni
Values of Ni are taken from the appropriate design S-N curve for each value of
Si.1
S.J. Maddox, Fatigue Strength of Welded Structures, Second Edition, 1991
5354C.FR
The variability in the test data is largely a result of imperfections in the welded test specimens.
To use S-N data to predict fatigue life, the designer must assume that the structure will have
imperfections similar to the test specimens. The current project was intended to test the validity
of this assumption.
3.2
Classification of Imperfections in Welded Joints
Welded joints cannot be perfect. All joints have flaws/imperfections that reduce the structural
integrity or fatigue life of the joint. Imperfections lead to higher localized stress, thus fatigue
cracks most always start at an imperfection. The localized stress is often called the notch stress.
For fatigue design it is important to know the effect of an acceptable imperfection on the fatigue
life of the welded joint, and conversely when an imperfection becomes unacceptable within the
analytical framework used as the design basis. To do this the imperfections must be identified or
classed, and each type of imperfection must be quantified by its relevant parameters.
There are three main classes of imperfections that will decrease the fatigue life of a welded joint:
a) Planar Flaws (sometimes called Surface Weld Discontinuities)
i)
ii)
iii)
cracks
lack of fusion or penetration
undercut, root undercut, concavity and overlap. (On some occasions, undercut
and root undercut in welds are treated as shape imperfections.)
b) Non-Planar Flaws (sometimes called Embedded Weld Discontinuities)

i)
ii)
cavities
solid inclusions, e.g. porosity and slag (On some occasions cavities and solid
inclusions are treated as planar flaws.)
c) Geometrical / Shape Imperfections

i)
ii)
iii)
iv)
axial misalignment
angular misalignment
imperfect weld profile
undercut and root undercut (if it gives rise to stress concentration effects)
A comprehensive classification of the various types of weld flaws that may be encountered is
given in ISO 6520 (AWS D3.5).
It was agreed at the outset of this project that the scope would be limited to geometrical
imperfections, as these are the fabrication tolerances that can be handled explicitly using
standard design tools. However, some additional discussion of the fatigue effects of all types of
imperfections is provided in the following pages to supply context for the subsequent analyses.
5354C.FR
3.2.1 Planar and Non Planar Flaws

Non-planar flaws such as weld porosity and slag inclusion will reduce fatigue life, but are less
harmful than planar flaws for the fatigue life of a welded connection when kept below normal
workmanship levels. Planar flaws such as weld toe undercuts, cracks, overlaps, porosity, slag
inclusions and incomplete penetration can have a significant influence on fatigue life. Figure 3.2
illustrates many of the planar and non-planar flaws that can exist in a welded joint.
Figure 3.2: Common Geometrical and Internal Weld Discontinuities

(Reproduced from: Bowman D., and Munse W.H.)
Non-destructive testing is required to quantify the dimensions of planar and non-planar
imperfections. Obtaining a statistically significant sample of these types of defects requires
considerable effort.
Most fatigue assessment methodologies consider the effect of planar and non-planar
imperfections by assuming that samples upon which the S-N curves were developed contained
representative quantities of planar and non-planar imperfections. Some of the scatter in the
experimental S-N curve data is considered to be representative of planar and non-planar
imperfections contained in a typical welded joint. However, this is an assumption that may
warrant further attention in the future.
5354C.FR
More in-depth methodologies have been developed for the assessment of the effect of planar and
non-planar imperfections on fatigue life. The British Standard, BS 7910, Guide on Methods for
assessing the Acceptability of Flaws in Structure (replacing PD 6493, and PD 6539) provides an
integrity management procedure based upon a fitness for purpose philosophy.
3.2.2 Geometric Imperfections
Geometric imperfections in welded joints such as misalignment, angular misalignment, angular
distortion, excessive weld reinforcement and otherwise poor weld shapes can reduce fatigue life
by several orders of magnitude. Geometric imperfections differ from planar and non-planar
flaws in that their effect is to enhance existing regions of stress concentration in the welded joint,
chiefly the weld toe, rather than to provide additional sites for possible fatigue crack initiation.
The effect of the misalignment is to cause an increase or decrease in stress in the joint when it is
loaded, due to the introduction of local bending stresses. This applies to both butt and fillet
welded joints, but only under loading which results in membrane stresses transverse to the line of
misalignment.
For reasons outlined, it was decided at the outset to focus the study on geometric imperfections
in butt-welded plates and stiffeners, and fillet welded cruciform joints. These joints compose the
majority of welded joints in a ship and are fatigue sensitive to imperfections. An aim of the
study was to quantify the effects of geometric imperfections in terms of fatigue life. In this
regard, formulae already exist for estimating the increase in stress due to geometric
imperfections. The development of these formulae and issues concerning their application is
discussed in Section 3.3 and in subsequent sections. Finally, most geometric imperfections to
some degree can be measured relatively easily without the use of any special non-destructive
testing equipment. The figures in Section 3.3 provide a sample of the individual geometric
imperfections possible in typical butt and cruciform welded joints.
3.3
Use of S-N Curves in Design
Three different ideologies have developed with regard to the creation of S-N curves and their use
in fatigue analysis. The three approaches are commonly referred to as:
1.
2.
3.
Nominal stress approach

Hot Spot (Structural) stress approach
Notch stress approach
To evaluate fatigue strength properly, there should be consistency between the stress with which
the S-N curve is defined and the one with which fatigue strength is calculated. A brief overview
of each approach is presented in the following subsections.
5354C.FR
3.3.1 Nominal Stress Approach

The nominal stress approach uses S-N curves that have been derived using test pieces that
contain various attachments giving rise to structural discontinuity effects, and various welds, but
usually no macro-geometric effects. Fatigue strengths given in the S-N curves are nominal
stresses that exist in the structure just outside of the welded joint.
The testing procedure used to derive Nominal Stress S-N curves does not explicitly account for
the effects of imperfections. Most nominal stress procedures (such as BS5400) class each detail
according to its quality and configuration. Each individual detail is then associated with an S-N
curve that should account for expected quality and configuration in predicting the fatigue life of
the detail. The S-N curves are multiples of each other, as shown in Figure 3.3. This approach
essentially says that all welded steel has the same fatigue life, thus the same fatigue curve. The
difference in fatigue life is due to the stress concentration resulting from the weld quality,
material quality, weld geometry and the overall geometry of the detail.
This approach can work well if it can be assumed that the detail under consideration will have
the same stress pattern and imperfections as the detail upon which the S-N curve was derived.
Some potential disadvantages have been summarized as follows:
Most S-N curves proposed by international institutes such as IIW and BS 5400
are defined with the nominal stress range and the related weld-joint type. The
nominal stress excludes the stress concentration due to geometric shape such as
structural discontinuities and presence of attachments. At most of the critical
points in ship structure where fatigue strengths are concerned, there are stress
concentrations that depend not only on structural detail shapes but also on
applied loading pattern. Furthermore, it is often hard to define the nominal stress
due to the complexity of structure and loading. Accordingly, there is a high
possibility is misevaluating the fatigue strength when it is evaluated with the
nominal stress basis.2
3.3.2 Hot Spot Stress Approach
The S-N curves for this approach are based upon estimated stress at the toe of the weld. Some of
the uncertainty concerning the stress concentration due to the weld shape is removed using this
approach. Normally a coarse mesh Finite Element (FE) model, or analytic methods are used to
establish the nominal stress just outside of the weld detail. A fine mesh sub-model of the weld
detail is then created to determine the stress at the toe of the weld. This stress is then used with
the hot spot S-N curves to estimate fatigue life for that joint.
In a hot spot S-N curve testing program the nominal stress would still be measured. The stress at
the toe of the weld is determined using a FE model of the test specimen to establish a stress
concentration factor that would give stress at the toe. In the FE model the toe of a weld is a
singularity, and thus stress is normally determined at a small distance from the root.
2
Kang, W., Kim, S, A Proposed S-N Curve for Welded Ship Structures, supplement to the Welding Journal, July
2003.
5354C.FR
There are various extrapolation standards, but it is usually taken somewhere between 0.5 to 1.5
plate thicknesses from the root. Some errors are introduced using this methodology.
The resulting value of hot spot stress may differ depending on the FE program
or on the element type, although the procedure for the calculation is just the
same. It is necessary to establish a more appropriate procedure for the
calculation of the hot spot stress that may represent the state of stress in relation
to the fatigue behavior of welded joints.3
Error can also be introduced if there is no consistency in the weld profile of the test specimens,
or if the FE model does not accurately represent the test specimens. Most of the hot spot S-N
curves in use are derived largely from nominal stress S-N curves developed for BS5400. The
weld profile data for most of the nominal stress S-N curves has not been accurately recorded
(e.g. SSC-369)4. Thus in this derivation, reasonable assumptions regarding the weld profiles of
the test specimens had to be made in order to determine the stress concentration factor (SCF) for
the samples.
Even with possible derivation errors, it is generally accepted that the hot-spot stress approach is
more accurate than the nominal stress approach. Provided the hot-spot S-N curve is accurate, the
methodology allows for more freedom and accuracy in the types of structural details and weld
details that may be analyzed. The hot spot stress approach is becoming more widely accepted as a
more accurate and practical approach to fatigue analysis of ship structures. This is largely because
advances in computing mean that it is now feasible to use FE models for determination of stress.
The uncertainties concerning the planar and non-planar imperfections still exist with this
approach. For fatigue analysis it is still important that the quality of the weld detail match the
weld quality of the samples upon which the S-N curve has been derived.
3.3.3 Notch Stress Approach
The notch stress approach is based upon S-N curves derived for smooth specimens that contain
no geometric notches. A stress concentration factor is then determined to account for the
increase in stress due to each possible type of geometrical imperfection. Using this methodology
a fine mesh FE model of the welded detail is not required since the increase in stress at the weld
toe is found using an appropriate stress concentration factor. The relationship between nominal,
hot spot stress, and notch stress can be expressed as follows:
? s notch = Kw ? s hotspot = Kg Kw ? s nom
Kg = Stress concentration factor due to the gross geometry of the detail considered
Kw = Stress concentration factor due to the weld geometry
Kang, W., Kim, S, A Proposed S-N Curve for Welded Ship Structures, supplement to the Welding Journal, July
2003.
4
Stambaugh, K., et. al Reduction of S-N Curves for Ship Structural Details SSC Report 369.
10
5354C.FR
The development of empirical formulae for these SCF and for stress concentration factors
pertaining to axial and angular misalignment is addressed in the following pages. The derivation
of the notch stress S-N curves is a possible source of error for this approach. Most notch stress
S-N curves in use in the marine industry have been derived largely from the S-N curves
developed for BS5400. Again, the level of imperfections in these curves is not precisely known,
thus assumptions were made in order to develop a notch stress S-N curve. There is also
uncertainty associated with the development and application of stress concentration factors.
Most SCF factors have been developed using analytical methods, or FE analysis to predict the
stress at the notch due to a particular imperfection. With the FE approach, this is done for many
geometric variations and then regression analysis is used to determine an appropriate formula. It
is necessary to assume a weld shape for the FE analysis. Real welds will differ from the
assumed shape; thus depending on the sensitivity of the detail, significant errors are possible.
The development of the SCF factors for increase in stress due to weld profile has meant that it is
possible, within the error limitations, to address the effect of imperfect weld profile on fatigue life.
The uncertainties concerning the planar and non-planar imperfections still exist with this
approach. For fatigue analysis it is still important that the quality of the weld detail match the
weld quality of the samples upon which the S-N curve has been derived.
Det NorskeVeritas (DNV) is one of the organizations that use a notch stress approach to fatigue
life calculations. Their methodology is well suited to assess the influences of construction
tolerances on fatigue life, and is thus described in more detail below.
Under the DNV approach, the S-N curves used in notch stress analysis are based on smooth test
samples where the notch stress is equal to the nominal stress. The K-Factors used in this report
are thus defined as:
K = s notch / s nominal
Thus the notch stress range to use with the appropriate S-N curve is:
? s notch = K ? s nominal
The overall K factor is a combination of K-factors arising from different geometric imperfections.
K = Kg Kw Kte Kt Kn
Where:
Kg = Stress concentration factor due to the gross geometry of the detail considered.
Kw = Stress concentration factor due to the weld geometry.
Kte = Additional stress concentration factor due to eccentricity tolerance (normally
used for plate connections only).
K t = Additional Stress concentration Factor due to angular mismatch (normally used
for plate connections only).
Kn = Additional stress concentration factor for un-symmetrical stiffeners on laterally
loaded panels, applicable when the nominal stress is derived from simple beam
analyses (not considered in this report).
11
5354C.FR
Table 3.1 and Table 3.2 give the formulas for the individual stress concentration factors
calculated. The formulas are derived from the DNV Classification Note No. 30.7: Fatigue
Assessment of Ship Structures. The derivation, application and accuracy of some of these
formulas are discussed in Section 6.
For some geometries, DNV provides default values that have been established for normal design
fabrication of welded connections. These values are also presented in Table 3.1 and Table 3.2.
For comparison, the corresponding maximum fabrication tolerances adopted by the International
Association of Classification Societies are also presented.
The situations and parameters outlined in Tables 3.1 and 3.2 are illustrated in Figures 3.3 to 3.6
and Figures 3.7 to 3.9 respectively.
12
5354C.FR
Table 3.1: Stress Concentration Factors Applicable to Butt-Welded Plates

and Stiffeners of Same Thickness
(Published in DNV Classification Note: Fatigue Assessment of Ship Structures)
DNV Stress Concentration
Factors Formulae
Angular Misalignment
Seam Halfway Between
Supports
K t = 1 +
s
e
= 1+
4 t
t
= 6 for pinned ends

= 3 for fixed ends
IACS
Standard
e = 4 to 7 mm
depending on
location.
Limit is e = 6 to
9 mm
depending on
location.
DNV Default
e = 6 mm
For 12mm
plate,
assuming
pinned ends
Kt = 4
Angular Misalignment
e
K t = 1 + s1 1 = 1 +
t
t
Same as above.
Same as
above.
Weld Reinforcement
< 60
< 45
K w = 1.0 + 0.5(tan )
1/ 4
Misalignment
K te = 1 +
3e0
t
Gives Kw =
1.5
Notes
IACS standard is for
Fairness of plating
between frames not
necessarily for butt
joints.
e = ( s)/4 only if the
seam is at the middle
of the span.
For each butt-welded

joint there are four
Kws. One for each
weld root angle.
For high
e0 = 0.15t
strength steel e0
< 0.15t or e0 < 3
mm.
For others steels
e0 < 0.2t or
e0 < 3 mm
Gross Geometry
Kg = 1.0
Simple details with a

Kg = 1.0 were selected
for the survey.
13
5354C.FR
DATUM
e2
e0
e1
Figure 3.3: Butt-Welded Plate Misalignment - Variables
s
s2
s1
1
(angle relative to earth)
Figure 3.4: Butt-Welded Plate Angular Misalignment - Variables
Figure 3.5: Butt-Welded Plate Weld Toe Angle - Variables
14
5354C.FR
RW
RHT
Figure 3.6: Butt-Welded Plate Weld Reinforcement - Variables
15
5354C.FR
Table 3.2: Stress Concentration Factors Applicable to Fillet Welded Cruciform Joints
(Published in DNV Classification Note: Fatigue Assessment of Ship Structures)
DNV Stress Concentration
Factors Formulae
Misalignment
K te = 1 +
6 t2 e
t3 t3 t3 t3
l1 1 + 2 + 3 + 4
l1 l2 l3 l4
t
t
e = 1 + e0 2
2
2
t1 t 2
e0 0.3t1
IACS Standard
For high strength
steel.
Notes
E0 0.3t1
e0 = (5t1 3t 2 ) / 6
For other
e0 = (2t1 t 2 ) / 6
Where t3 is less than
t1, then t3 should be
substituted for t1 in
the standard.
Weld Geometry
at weld toe 90.
For axial stress in direction of

intercostal member
States that in areas of

stress concentration
and fatigue the class
society may require a
lesser angle.
K g K w = 1.2 + 1.3(tan )1 / 4
DNV Default
K g K w = 2 .5
for = 45
Gives stress at toe of weld.

Weld Geometry
For axial stress in direction of
continuous member
K g K w = 1 .8
for = 45
K g K w = 0.9 + 0.9(tan )1 / 4
Gives stress at toe of weld.
16
5354C.FR
t3
l3
e0
t3
l2
l1
t2
t1
t2
t1
t4
l4
Figure 3.7: Cruciform Joint Misalignment Variables
LEG 2b-2
t3
LEG 2b-3
NOMENCLATURE
t2
LEG 1a-1
t1
PLATE 1
SIDE A
IN DIRECTION
OF PLATE 1
LEG 1a-3
LEG 1a-1
Figure 3.8: Cruciform Joint Weld Geometry Variables
t3
t2
t1
Figure 3.9: Cruciform Joint Angular Misalignment Variables
17
5354C.FR
3.3.4 Axial and Angular Misalignment Imperfections - Application in S-N Curve Analysis
It is important to note that test specimens are normally considered free of axial and angular
misalignment, although there is some uncertainty in this assumption:
It is likely that some of the nominally-aligned test specimens used to generate
the data upon which design S-N curves are based were actually misaligned, with
the result that some misalignment is always acceptable. For example, in the case
of butt welds, there is some evidence [65] 5 to indicate that the design S-N curve
already embodies the effect of misalignment corresponding to Km = 1.3.
However, further work is needed to confirm this, and to consider cruciform
joints.6
However, assuming that the increase in local stress in butt and fillet welded joints is not
embedded in the S-N curves of either fatigue analysis approach, the stress range used in the
fatigue analysis must always be increased by a suitable factor to account for any anticipated
misalignment in the detail. Formulae have been developed for estimating this increase in stress
for certain types of misalignment, the DNV examples provided at Section 3.3.3 being one
example of this.
More generally, as shown in Figure 3.10, angular misalignment in an axially loaded joint will
induce a bending moment and a secondary bending stress in the joint.
/2
Figure 3.10: Angularly Misaligned Plating - Seam Halfway between Supports
Maddox, S.J., 1985. Fitness for purpose assessment of misalignment in transverse butt welds subject to fatigue
loading, IIW, document XII-1180-85. (unpublished).
6
BS 7910:1997, Guide on methods for assessing the acceptability of flaw in fusion welded structures Draft for
public comment.
18
5354C.FR
The stress must be resisted axially, thus the bending stress in terms of developing a SCF formula
can be thought of as an apparent increase in axial stress equal to:
K t = 1 + b
a
where the axial stress is equal to:
P
P
a = =
.
A t b
The bending stress is equal to:
My 6 M
b =
= 2
I
bt
for a flat plate. The bending moment at the joint:
s
s
s
M = P sin( ) = P
=P
2
2
22
4
for small angles. Substituting M in the formula for b gives:
3Ps
b =
.
2bt 2
Substituting a and b into the formula for K t gives:
3
K t = 1 + s .
2 t
This is the same formula7 provided by DNV in Classification note 30.7 and can be reduced to:
e
K t = 1 + 6
t
if e instead of and s is known. Also, note that for fixed end supports:
1
e
K t = 1 + s = 1 + 3 .
2 t
t
The derivation of this formula illustrates many important points regarding its application to
fatigue design of ship structures.
Generally, the plating in ship structure is lofted such that weld seams can be as close as possible
to a stiffener or support. This is done specifically to avoid the problem of secondary bending
stress due to misalignment. The formula above in terms of , angle between plates, does not
apply directly. Figure 3.11 illustrates misalignment in a more typical ship structure.
The same formula is given by Maddox in Fatigue Strength of Welded Structures, Second Edition.
19
5354C.FR
Figure 3.11: Angularly Misaligned Plating - Seam at Location

other than Halfway between Supports
The derivation of the Kt formula for this geometry is the same as above except that s/2 is
replaced by s1 and /2 is replaced by 1. Making these substitutions gives K t = 1 + 6s1 1

t
For a given value of e however, the value of Kt will be the same irrespective of the location of
the seam. For example for e = 6 mm is Kt = 4 for both geometries.
A number of systematic experimental studies of the effects of misalignment have been
undertaken (for example by the UK Department of Energy, as reported in Maddox, 1991) from
which Figure 3.12 is reproduced. Work of this type has been used more in establishing tolerance
limits than in developing more sensitive analytical treatments of the tolerances that actually exist
in fabricated structures.
20
5354C.FR
Figure 3.12: Effect of Misalignment on Fatigue Strength (Maddox, 1991)
3.4
Sensitivity in Fatigue Life Calculations
The discussion presented in DNV Classification Notes No. 30.7, Section 2.4 provides a good
overview of some of the uncertainties in fatigue life prediction, some relevant aspects of which
are presented and discussed below:
21
5354C.FR
2.4.2
Because of the sensitivity of calculated fatigue life to the accuracy of estimates of
stresses, particular care must be taken to ensure that stresses are realistic.
Fatigue damage is proportional to stress raised to the power of the inverse slope
of the S-N curve. Small changes in stress result in much greater changes in
fatigue life. Special attention should be given to stress raisers like eccentricities
and secondary deformations and stresses due to local constraints. Due
considerations should, therefore be given to the fabrication tolerances during
fatigue design.
2.4.3
There is rather a large uncertainty associated with the determination of S-N
curves. The scatter in the test results, which form the basis for the S-N curves, is
generally accepted to relate to the normal variation of weld imperfection with
normal workmanship. The ratio between calculated fatigue lives based on the
mean S-N curve and the mean minus two standard deviations S-N curve is
significant as shown in (Figure 3.13).
Figure 3.13: Fatigue Life Influence of Stress Level and S-N Data for Welded Connections
Two important aspects of fatigue life can be identified from the text and figure above. First, the
exponential nature of the relationships between stress range and fatigue life lead to dramatic
changes in expected outcome (life) for modest changes in stress. The second noteworthy aspect
is that most S-N design curves are based upon the mean minus two standard deviation curves for
the relevant experimental data set. Through this assumption, the S-N curves are associated with
a 97.6% probability of survival. This level of conservatism is intended to mitigate some of the
unknowns in fatigue life prediction. An implicit assumption is that the level of scatter in the
experimental data is representative of that in the real world. As can be seen from the example
in Figure 3.13, the scatter is normally considerable, in this case, a factor of 2.6 on design life.
22
5354C.FR
As stress range is affected by the local stress concentration factor DNV highlights the influence
of uncertainties in SCF:
2.4.4
There is also uncertainty associated with the determination of stress
concentration factors. The error introduced in the calculated fatigue life by wrong
selection of stress concentration factor is indicated in (Figure 3.14).
Figure 3.14: Fatigue Life Sensitivity to Stress Concentration Factor K and Weibull Shape
Factor H
Figure 3.14 essentially replots Figure 3.13 to emphasize that incorrect assumptions regarding
SCF can be as or more important to fatigue life assessment as is the analysis of global stress level
(or the lifetime distribution of loads). The Weibull shape parameter h is related to the expected
loading and is not addressed in the current research.
3.5
Implications for the Current Project
Based on the foregoing, it is suggested that any fatigue life analysis should be based on:
1. An understanding of the anticipated stress concentration factors, based on the design and
the construction tolerances;
2. The availability of a suitable analysis methodology, incorporating a suitable fatigue life
(S-N) curve; and
3. An ability to quantify the impacts of any differences in the assumptions or the
uncertainties underlying either (1) or (2).
The work on the current project has therefore aimed to address all of these aspects of the
problem to the extent possible, within the available level of effort.
23
4.
5354C.FR
DATA COLLECTION
Task 1 of the project included two major thrusts the collection of existing, relevant information
and data through literature surveys, and the collection of new shipyard data to supplement and
extend the literature.
4.1
Literature Review
This subtask has included two main components:
a) collection of published and other available data on construction tolerances, including:
- standards (shipyard, industry, class, etc.)
- data on achieved outcomes
b) collection of references on the effects of construction tolerances on fatigue life,
strength, and other characteristics
Sources have included previous SSC projects, published reports and studies, standards and
related documents, and other materials available to the contractors.
The three main outcomes of this subtask have been:
i) a bibliography including a summary of each significant reference, noting (inter alia)
the purpose, scope, and conclusions of the document (presented at Appendix A).
ii) data for inclusion in the project databases and analyses.
iii) definition of the parameters to be characterized in the exploration of tolerances.
Several of the main findings of the literature survey were essentially negative. Virtually no data
on actual achieved shipyard tolerances appears to have been published; an elderly and partial
exception to this rule being background material for the1975 Japanese Shipbuilding Quality
Standards, presented in SSC Report 273 (1978). Even in this case the level of detail presented is
insufficient to allow for its systematic application to further statistical analysis (see also Section
5). Thus, while classification societies, shipbuilding associations, and other bodies have
published tolerance standards, there is no body of knowledge in the public domain to relate these
standards to actual outcomes.
A second area in which data is lacking is in the definition of the data underlying design S-N
curves. As discussed in Section 3, these are assumed to incorporate some levels of
misalignment, weld imperfections, material properties etc., that contribute to their scatter.
However, as the criteria for specimen (or outcome) acceptance/rejection are not reported in any
of the reports accessed by the project team, it is not possible to assess their absolute or relative
importance.
24
5354C.FR
4.2
Field Data Collection
Prior to the initiation of the contract, three North American shipyards had indicated that they
would be prepared to permit collection of data on their actual fabrication tolerances to support
this project. However, when contacted subsequently one was no longer prepared to cooperate
(due to management changes) and a second had no significant work under way. The third yard
did allow access, and another yard eventually also cooperated. Efforts were made to extend the
surveys to other yards, but were unsuccessful due to apparent concerns over confidentiality of
the data, possible disruption to the work in the yard, etc. The recommendations at Section 9
discuss how such concerns might be mitigated more successfully in future projects.
The first field data collection effort (Shipyard #1) aimed to generate data for a reasonably wide
range of structural elements/details. It was found that (a) many of the planned measurements
were difficult to obtain, and (b) the level of scatter were such that large data sets were needed to
generate statistically reliable information. As a result, the second survey (Shipyard #2) focused
on a reduced set of parameters, and used an improved tool set to take measurements.
Overall, the survey work concentrated on geometric imperfection in two classes of welded
details, butt-welded plates and stiffeners, and cruciform joints. Figure 3.3 through Figure 3.6
illustrate the variables that were measured for butt-welded plates and stiffeners. Figure 3.7
through Figure 3.9 illustrate the variables that were measured for cruciform joints.
4.2.1 Summary of Shipyard Survey #1
The first survey was conducted at a medium sized shipyard (Shipyard #1) that builds mostly
barges and workboats. The intent of the survey was to establish measurement techniques, to
assess which details should be measured, and also to collect as much data as possible.
The visit revealed that Shipyard #1 collect very little geometric information that is useful to the
current study. According to the manager of the dimensional control department:
The yard is most interested in 1) minimizing rework, 2) classification society
acceptance, and 3) customer acceptance; the measurement and recording of
structural detail data important for fatigue (where compliance is not an issue) is
not a high priority. We collect data to monitor the early fabrication processes for
control, centering and capability, but typically collect data on the final welded
vessel only when required to confirm acceptability on a case by case basis.
The yard builds vessels mainly to American Shipping Bureau (ABS) class. The tolerance
standard they use is the ABS Guide for Shipbuilding and Repair Quality Standard, July 98
(this document is based on the International Association of Classification Societies (IACS)
Shipbuilding and Repair Quality Standard, 1996). This standard is implemented by ABS
Surveyors on-site. The ABS Surveyors inspect the Yards fit-up practices (tools, techniques,
technician skill, and final as-fitted results) and visually inspect 100% of the final product.
Measurements are generally taken of any questionable areas resulting from inspections.
25
5354C.FR
Most of the BMT Fleet Technology Limited (BMT) Surveyors time at Shipyard #1 was spent
out in the yard taking measurements. To ensure enough data for statistical analysis the survey
concentrated on measuring misalignment of two types of welded connections; butt welds at
erection joints (for both plating and stiffeners), and fillet welded intercostal joints, mainly double
bottom girder/floor intercostal joints. In the two-day visit, 39 butt weld misalignment samples
and 28 intercostal misalignment samples were taken.
Approximately 12 hours of time was spent taking measurements. Thus on average it took
(12hrs)(60min/hr)/(39+28 measurements) = 10 minutes per sample detail. The surveyor was
developing the measurement process during the visit, and as was expected, the time per sample
detail was reduced to approximately eight minutes in the next survey. The latter value can be
considered a benchmark for the level of effort that will be required in any future surveys of this
type.
Choices regarding what were measured were largely dependant on what could be measured, what
was worthwhile measuring, and what was there to measure. At the time of the visit Shipyard #1
had an 80000 BBL Oil and Asphalt Barge on the way ready to launch, and many of the Grand
Assemblies for a 120,000 BBL Oil and Asphalt Barge completed and ready for final assembly.
Even though these double-hulled vessels are quite simple in construction, they do have many
welded connections of interest, especially considering that they are large barges. For instance
the 120,000 BBL barge has a length over all of 129.6 m, breadth of 22.12 m and depth of 12.5 m.
Initially for the butt-welded plate connections the intention was to take random measurements
without much discrimination for the type of joint. It was discovered quickly that there is very
little misalignment at any welds done on the panel line (making up 95% of the longitudinal butt
welds in the final product) and that taking such measurements with the available tools would
introduce more error than is actually present in the fabrication process. Therefore, these
measurements were not pursued. Thus the focus of the survey became field welded erection
joints of major assemblies. The statistics involved in such an approach are considered in
subsequent sections.
Butt weld misalignment was measured by using a magnet/ruler combination as a datum that
would straddle the seam. Digital vernier callipers were used to measure the distance from the
datum to the plate on each side of the seam. This method did not prove to be fully adequate thus
a better and faster technique was developed for the next survey.
There were few measurable intercostal details on the vessels. Additionally, for a fully assembled
vessel such as the 80,000 BBL Barge, it was quite difficult to measure intercostal misalignment.
Major misalignment can occur at erection joints, however there is no suitable methodology that
would allow this misalignment to be accurately measured. Focus was therefore on double
bottom structure grand assemblies of the other barge. Typically, the double bottom structure
consisted of 4 longitudinal continuous girders with floors running intercostal between them.
Many of the parameters related to the weld detail, such as angle at the toe or fillet size, are quite
variable along the length of any particular weld seam or fillet.
26
5354C.FR
At the time, it was the opinion of the BMT surveyor that meaningful measurements could not be
taken. Toe angle measurements for the butt welds were taken, but none of the fillet welds at
intercostals were measured.
Drawings of the vessels under construction were provided in addition to portions of the build
strategy and accuracy control plan. These documents are confidential to Shipyard #1 and are not
available for inclusion in the report.
The data collected has been summarised in Section 5. The histograms show that sample sizes for
most tolerance measurements will have to be much larger before statistically significant
fabrication tolerance distributions can be assigned. Also a statistical prediction is predicated on
the process being under control. This is sometimes difficult to achieve where the fit-up process
is largely a function of manual efforts.
4.2.2 Summary of Shipyard Survey #2
BMT conducted the second survey at a medium sized shipyard (Shipyard #2) that builds mainly
commercial workboats and small military vessels.
At the time of the survey, the shipyard was building an 80m Anchor Handling Tug Supply
Vessel (AHTS). The steelwork for the vessel was in the final stages and most of the machinery
was in place. All of the major units, except for the bridge, had been fitted. Some of the bow and
stern assemblies were fitted but not fully welded.
The shipyard was also building assemblies for a refit on a Mobile Offshore Drilling Unit
(MODU). The assemblies inspected were to form a new pipe deck on the MODU.
Both the AHTS and the MODU were surveyed for geometric imperfections in the welded
connection details. The survey concentrated mainly on the AHTS since the focus of the study is
fatigue life of welded joints in ships, rather than offshore platforms. However, the measurements
taken from the MODU assemblies are also valid since they represent general fabrication quality
at the yard. Ease of construction, thickness of material, and many other variables will all
contribute to how much the structure at welded joints deviate from the ideal.
The focus of the survey was to measure imperfections in the joints between major assemblies. It
is more difficult to achieve close tolerances at assembly joints. The tolerances are close on joints
done on panel lines meaning the defects are difficult to measure. Focusing on assembly joints
identifies major problematic defects and provides confidence in the accuracy of the
measurements. Measurements were taken at butt welds between plating, butt welds between
stiffeners and welds at cruciform joints.
27
5354C.FR
For the vessels under construction it was difficult to find suitable cruciform joints that could be
measured, thus many samples are not at joints between major assemblies.
The items welded together and the locations within the structure are noted in the survey
summaries. In the two and half day visit, 59 samples for butt-welded plates, 29 samples for buttwelded stiffeners and 13 samples for intercostal joints were taken.
A wider range of measurements was taken during the survey at Shipyard #2 than at Shipyard #1.
The survey at Shipyard #1 illustrated the need for better tools, techniques and also that more
extensive measurements were needed to quantify the effect on fatigue that geometric
imperfections may have. The techniques employed at Shipyard #2 allowed more measurements
to be taken in the same amount of time. The accuracy was also somewhat better. A full
discussion of measurement techniques, tools and accuracy is included in Section 4.3.
Obtaining reasonable access to joints of interest was an issue during the survey. The surveyors
did not have enclosed spaces training and the shipyard would not allow entrance to any double
bottom or side tanks. Many of the joints of interest are in these locations. The shipyard would
have offered the training, but time and resources would not allow for it. Many of the joints of
interests such as butt-welded stiffeners, or cruciform joints between girders and beams would
have required the use of ladders and/or staging to take measurements. A choice was taken not to
attempt such measurements. There was much construction activity at the time of survey and the
use of a ladder would have interfered considerably with the shipyard activity. Further, the time
required per measurement increases considerably with the added complication.
Access to assemblies before they are joined to the ship will allow for more joints of interest, in
particular cruciform joints, to be easily measured since the assembly normally is upside down
before final assembly. The interest of the study however is joints between major assemblies,
which necessarily mean final ship construction.
Shipyard #2 has developed production standards that are mostly based on those of various
classification societies. The actual tolerances are proprietary to the shipyard thus BMT could not
copy their standard for publication.
Erection joint welds at the shipyard are full penetration and, in general, butt-welded plates and
stiffeners that have a thickness less than 20 mm are prepped with a single sided bevel. Most
samples had plating less than 20mm thick. The yard employed a ceramic backing technique for
many of these welds. Where ceramic backing was not used, the joint was welded from one side,
back-gouged and capped on the other side.
The visit revealed that Shipyard #2, as with Shipyard #1, collects very little imperfection
information that is useful to the current study.
The staff from the Dimensional Control department at the shipyard was BMTs liaison during the
survey. They provided valuable orientation and assistance during the visit. The BMT surveyors
for the most part worked independently in taking measurements.
28
5354C.FR
Fabrication defects are mainly a quality control concern in any yard. Yards such as Shipyard #1
combine Dimensional Control with Quality Control. Shipyard #2 has separate departments that
do this work. Larger yards will also have separate departments for this work. Future surveys
should concentrate on liaison with persons from the Quality Control Department.
Many of the parameters related to the weld detail, such as angle at the root or fillet size, are quite
variable along the length of any particular weld seam or fillet. Nonetheless, a full range of
measurements were taken at each location for the purpose of demonstrating that stress
concentration factors could be calculated at any one particular location. For consistency, at the
start of a set of measurements, a line was drawn at the random point of interest. This was to
ensure for example that angle at the weld toe and misalignment was measured at the same place.
Drawings of the vessels under construction were provided by Shipyard #2. These documents are
confidential to Shipyard #2 and not available for inclusion in the final report.
The survey data collected is summarised in Section 5.
Monitoring fabrication quality of welded joints and maintaining tolerances is important to
receiving a Fatigue Class Notation from a Classification Society. The survey work conducted
thus far illustrates the need to ascertain the quality control procedures in shipyards that build
ships for Fatigue Class Notation. Any further survey work should be conducted at shipyards that
are building such vessels for Fatigue Class Notation.
4.3
Survey Techniques
An important outcome of the project was the development of a methodology for measuring
geometrical imperfection variables in welded joints. As will be discussed in Section 6, relatively
large sample sizes are required to draw meaningful statistical conclusions regarding weld
fabrication quality in a particular ship, or shipyard. Collecting sufficient data, while minimizing
effort, requires that quick but accurate methods be employed. The development of a survey plan
covering choice of details to measure, and at what stage in the construction the detail is to be
measured will also determine the effectiveness of a survey program.
The imperfections in a joint vary along the length of the weld. The approach for the surveys was
to, as randomly as possible, pick a particular section and then take measurements that would
characterize the bulk properties of the cross section. This section is then considered
representative of imperfections anywhere along the length of the weld. The location along the
length of the weld, other than stop-start locations may or may not influence the extent of
imperfections. To ascertain this influence, the percentage distance from the end along the length
was recorded.
29
5354C.FR
Gaining a statistical picture of all the welding imperfections in a particular ship requires, at the
very least, a methodological approach for choosing sections to sample. For example, random
sections around the entire perimeter of a major erection joint are required to describe statistically
what is happening in that joint.
Obtaining access to the many areas of the vessels and limited available survey time were major
issues for the BMT surveyors. For this reason, the approach was to concentrate the
measurements in areas that were reasonably accessible.
In this regard, the surveyors concentrated mainly on erection joints in the deck and bottom of the
vessels and cruciform joints in the double bottom. Many measurements from other areas were
also taken simply because the area was accessible.
Most geometric imperfections in welded joints can be measured relatively easily and to adequate
accuracy with low cost tools. The methods and tools employed for this project are discussed in
the following sections.
4.3.1 Misalignment in Butt Welds
A dial gauge attached to a base, as shown in Figure 4.1, was used to measure misalignment of
butt-welded plates. The misalignment was determined by reading the dial on each side of the
weld and subtracting the values to obtain the misalignment. The same could be accomplished
using the weld gauge shown in Figure 4.2. Accuracy using this tool is less however. As well,
the tool may not work for welds that have excessive reinforcement or width.
Both methods are susceptible to errors resulting from angular misalignment and plate distortion
since the base of the tool must rest flush with the plate.
DIAL GAUGE
Figure 4.1: Measuring Alignment and/or Weld Reinforcement Height
30
5354C.FR
4.3.2 Weld Reinforcement in Butt Welds

The multi-purpose weld gauge shown in Figure 4.2 was used to measure weld reinforcement
height. The dial gauge could also be used. If there is misalignment in the plates, the reading will
be different from each side. A protocol must therefore be established for taking measurements
relative to the misalignment measurements. Both measurements can then be used to describe the
general profile of the joint.
Reinforcement height was only measured from one side of the weld. A true picture of the
section would require measurements from both sides.
Locating the exact underside of a weld is problematic as access is often extremely difficult. In a
true statistical analysis, this error is mitigated since the profile on one side of the plate is
supposed to be representative of the profile on the other side.
The surveys focused on erection joints between major assemblies. The gap between erection
joint plates is often quite large meaning that multi passes (done by hand) and large amounts of
weld filler are required in places. The result is that butt welds at erection joints rarely resemble
the ideal geometry upon which the stress concentration formulas have been determined.
Figure 4.2: Tool for Measuring Weld Reinforcement Height and/or Plate Alignment and
Fillet Weld Leg Size
4.3.3 Weld Toe Angle in Butt Welds

The weld toe angle was measured using a tool similar to that shown in Figure 4.3. Weld toe
angle is measured by placing the tool flat on the plate and pushing it into the toe of the weld until
the rotating piece becomes tangent.
The accuracy of this method is somewhat crude as the weld reinforcement will not often have a
circular shape. Based upon the surveyors experience it is likely that the error in measuring the
angle could be up to 5 or 10 using this methodology.
31
5354C.FR
WELD TOE
PROTRACTOR
30
20
40
50
60
70
10
Figure 4.3: Measuring Weld Toe Angle
4.3.4 Angular Misalignment in Butt Welds

The angular misalignment of horizontal plates is quite easily measured using an inclinometer as
shown in Figure 4.4. Although this method is straightforward, errors are derived from the
accuracy of the inclinometer and also from local distortions that prevent the inclinometer from
resting flat on the plate. Local distortion in the plate is in itself an error in that most fatigue
analysis methods idealize the plate as being straight.
DIGITAL INCLINOMETER
- 1.0
5.6
= 6.6
1
Figure 4.4: Measuring Angular Misalignment

Measuring alignment in vertical joints is more difficult because the inclinometer cannot be used.
Although this method was not used in the survey, it is possible to modify the tool shown in
Figure 4.5 to measure angular misalignment. Adding parallel extensions to the protractor, as
shown in Figure 4.5 allows it to straddle the weld reinforcement, thus enabling an angular
misalignment measurement to be taken. Most butt welds will also have plate misalignment. To
overcome this, one of the extensions can be made adjustable as shown in the figure.
32
5354C.FR
ADJUSTABLE HEIGHT
Figure 4.5: Modified Protractor for Measuring Angular Misalignment
4.3.5 Cruciform Joint Misalignment

Cruciform joints are susceptible to misalignment due to their nature, especially if it is not
possible for the shipwright to ensure visually that plates on each side of the joint are aligned.
Measuring misalignment to the precision of millimeters is equally as difficult for the same
reason.
The methodology used in the surveys was to find cruciform joints that had a natural datum and
then using a measuring tape, measure the distance that each plate is from that datum. Finding a
datum in the completed ship is generally difficult, however, they do exist for some assemblies.
The double bottom and deck assemblies shown in Figure 4.6 illustrate how misalignment in
cruciform joints can be measured.
e0
t3
t2
t1
USE END OF
GIRDER AS DATUM
FLOORS INTERCOSTAL
BETWEEN GIRDERS
STIFFENERS INTERCOSTAL
BETWEEN GIRDERS
USED END OF GIRDER AS DATUM
Figure 4.6: Measuring Misalignment in Cruciform Joints
33
5354C.FR
A better methodology is required if all cruciform joints in a vessel are to be surveyed. An

artificial datum could possibly be used to determine the alignment of intercostal stiffeners. The
datum would consist of a device that will clamp square over a girder. Measurements would then
be taken in reference to this device.
There are frequently no straightforward methods for measuring misalignment between, for
example, longitudinal bulkheads separated by a transverse bulkhead. Where holes are required
for system transits then these can be used to establish datum planes. In their absence, holes can
be drilled, used and repaired. For most commercial construction this is unlikely to be acceptable
on cost or schedule grounds.
4.3.6 Angular Misalignment in Cruciform Joints
This type of misalignment can easily be measured using a protractor such as shown in Figure 4.7.
A simple protractor such as shown is accurate within 0.3. Local distortion in the plates can be
a cause of error.
Figure 4.7: Protractor for Measuring Angular Misalignment in Cruciform Joints
4.3.7 Weld Size in Cruciform Joints

A set of ordinary weld gauges or the weld gauge shown in Figure 4.2 is adequate for measuring
the leg size of fillet welds in cruciform joints. It is important to use an appropriate measurement
protocol to ensure that the entire cross-section can be fully described subsequent to the
measurements being taken.
In typical shipyard construction, the leg size of a filet weld along a cruciform joint will vary
considerably. The basic philosophy of the survey is that of looking at random sections and
considering them representative of other sections throughout the vessel. Increase in local stress
due to the fillet weld is however a result of 3D geometry. In fact, each component of a joint
angles, thickness misalignment, etc., varies along the length of the joint.
34
5354C.FR
The overall fatigue resistant quality of different joints differs, depending on location, internal
support, assembly procedure, and welding types and sequences. Although beyond the scope of
this report, it is necessary to consider many parameters before reaching any conclusions on how
bulk properties should be interpreted in fatigue analysis.
35
5.
5354C.FR
STATISTICAL ANALYSIS OF SURVEY DATA
5.1
Background
As noted in Section 3, fatigue life prediction is inherently probabilistic (statistical) in nature.
Fatigue life prediction depends on both the S-N curves underlying the analyses, which include
assumed levels of scatter, and also (in the notch stress approach) on the stress concentration
factors, Ks, (or their underlying geometrical properties).
One element that is essential to perform reliability-based design for fatigue of ship structures is
the quantification of the basic fatigue damage accumulation process random variables. The
definition of these random variables requires the investigation of their variability. In reliability
assessment of any structural system, these uncertainties must be quantified. Furthermore, the
evaluation of strength and load partial safety factors (PSF) in any design format equation also
requires the characterization of these variables. For example, the First-Order Second Moment
(FOSM) method for reliability assessment and reliability-based design requires the quantification
of the mean values, standard deviations (or the coefficient of variation (COV)), and the
distribution types of all relevant random variables. They are needed to compute the safety
index or the PSFs. Therefore, complete information on the probability distributions of the
basic random variables under consideration must be developed.
Quantification of the basic random variables for the stress concentration factor K in terms of
their means, standard deviations or COVs, and probability distributions must be achieved in two
steps -- data collection and data analysis. The first step is the task of collecting as many sets of
data deemed to be appropriate for representing the random variables under study, while the
second is concerned with statistically analyzing the collected data to determine the probabilistic
characteristics of such variables.
The objective herein is to quantify the probabilistic characteristics of the stress concentration
factors Ks for use in reliability analysis and reliability-based fatigue design for ship structures.
The available statistical information and data on basic random variables for K consisted of the
data sets from two medium sized shipyards in North America, as described in Section 4. These
data were statistically analyzed and studied to quantify the probabilistic characteristics of the
various types of the stress concentration factors. These characteristics have been established and
summarized for stress concentration factors in terms of the mean values, standard deviations, and
the probability distributions.
5.2
Methodology
5.2.1 Descriptive Statistics and Histograms

The raw geometric data collected during each survey is presented in Appendix B. The figures in
Section 3 provide definitions for the parameters. The formulas used to calculate stress
concentration (K) values are those given in Table 3.1 and Table 3.2.
36
5354C.FR
The approach used to quantify the stress concentration factors K is based on the statistical first
and second moments, i.e., the mean and standard deviation of the basic variables that are used to
determine K. The data used in these analyses are those collected and measured at the Shipyard
#1 and Shipyard #2. These data represents the basic random variables that define the different
stress concentration factors as established by DNV. The steps that have been followed to assess
the different stress concentration factors are as follows:
1. For each equation that represents the stress concentration factor K, select the basic
random variables that define K.
2. Use the equation to compute K for each set of the basic variables that were collected and
measured.
3. After the K values have been computed from Step 2, perform statistical analysis to
quantify its statistical moments (i.e., mean, standard deviation, etc.) and the upper and
lower bounds on K.
4. Based on the statistical analysis of Step 3, generate a frequency histogram for each K.
5. Fit known theoretical/continuous statistical distributions for K that closely agrees with the
histogram generated in Step 4.
6. Use a commercial statistical software package or a spreadsheet to identify the two
theoretical distributions that best fit the data.
7. Document the mean value, standard deviation, upper and lower bounds, and the best two
distribution types for K.
The Chi-Squared method was used in these analyses to quantify and assess the goodness of fit of
the statistical distributions to the estimated stress concentration factor (K) data.
5.2.2 Total Stress Concentration Factor
The total (or overall) stress concentration factor Kt , as described in Section 3, is a combination
of all different factors due to various geometric imperfections. The total stress concentration
factor can be calculated as:
K = K g K w K t K te K n
(5-1)
Where:
Kg = Stress concentration factor due to the gross geometry of the detail considered.
Kw = Stress concentration factor due to the weld geometry.
Kte = Additional stress concentration factor due to eccentricity tolerance (normally
used for plate connections only).
K t = Additional stress concentration factor due to angular mismatch (normally used
for plate connections only).
Kn = Additional stress concentration factor for un-symmetrical stiffeners on laterally
loaded panels, applicable when the nominal stress is derived from simple beam
analyses (not considered in this report).
37
5354C.FR
The steps that have been followed to assess the total stress concentration factors K are similar to
those for the individual K-factors and they are given as follows:
1. For each case weld type, calculate the total concentration factor K using Eq. 1.
2. For each K computed in Step 1, perform statistical analysis to quantify its statistical
moments (i.e., mean, standard deviation, etc.) and the upper and lower bounds.
3. Based on the statistical analysis of Step 2, generate a frequency histogram for each K.
4. Fit known theoretical/continuous statistical distributions for K that closely agree with the
histograms generated in Step 3.
5. Use a commercial statistical software package such as @Risk and BestFit or a
spreadsheet to identify the two theoretical distributions that best fit the data.
6. Document the mean value, standard deviation, upper and lower bounds, and the best two
distribution types for K.
5.2.3 Confidence Interval on the Mean Values
The mean and standard deviation for the samples of the individual stress concentration factors
(e.g., Kte, Kw, etc.) represent a best estimate of the population value. However, they are only
estimates of random variables, and they do not necessarily correspond to the true values. The
accuracy of these estimates can be assessed using confidence intervals. Confidence interval
provides a range of values in which the true value of a K-factor can be expected to lie. Many
two-sided (1- )% confidence intervals have one of the following two forms depending whether
the population standard deviation () is known or not (Ayyub and McCuen 2003):
s
s
x Z
x + Z
n
n
2
2
x t
2
S
,n
x + t
2
S
,n
(5-2a)
(5-2b)
where:
x
t
2
S
n
,n
= sample mean
= t-distribution value at an exceedence area of /2.
= level of significance in the range 0 to 1.
= sample standard deviation
= sample size
= population mean.
= population standard deviation
The steps that have been followed to compute confidence interval are as follows:
1. Select a two-sided confidence interval for each K-factor.
2. Specify desired level of confidence as 99, 95, and 90%.
38
5354C.FR
3. Perform statistical calculations (e.g., mean, standard deviation, and x , n, and S of Eq. 2)
for each individual K value.
4. Determine the value of the distribution factor t / 2 based on the confidence level and K
sample size. For a sample size greater than 32, use z-statistics, otherwise use t-statistics.
5. Compute the confidence interval for each individual K value.
5.2.4 Sample Size Determination
The selection of a sample size is the first step in performing statistical analysis. In this report
sample size calculations were performed for illustrative purposes and for assessing the suitability
of the sample sizes used in this study. Sample sizes for SCF were determined herein for the
purpose of future sampling projects that might take place in other shipyards.
The sample size (n) can be computed based on two-sided confidence interval on the sample
mean as follows:
Z
n = /2
H
(5-2c)
where H = half the width of a confidence interval as defined in Eq. 5-2a. A similar equation for
n can be written to correspond for Eq. 5-2b as follows:
t
/ 2, n
n=
H
(5-2d)
The solution of these equations needs to be completed using an iterative solution approach. The
following table shows samples sizes as a function of the ratio H/ for = 0.1 (with Z/2 = 1.96)
based on Eq. 5-2c:
H/
0.1
0.2
0.3
0.4
0.5
Sample size n
384
96
43
24
15
Typical values for the ratio H/ based on the data collected in the range 0.2 to 0.5 leading to the
conclusion that the sample size should have been 15 to 96.
39
5354C.FR
5.3
Butt-Welded Plates
The stress concentration factors applicable to butt welds are those listed at Table 3.1, which for
convenience are repeated as equations 5.3 5.5 below.
K t = 1+
4 t
(5.3)
K w = 1.0 + 0.5(tan )
1/ 4
K te = 1+
5.3.1
(5.4)
3e
t
(5.5)
Shipyard #1 Data
5.3.1.1 Stress Concentration Factor Kt

The data required to calculate Kt were not measured during this initial survey.
5.3.1.2 Stress Concentration Factor Kw
The stress concentration factor Kw was computed using equation (5.4). Table 5.1 shows the
result of the statistical analysis of Kw based on the Shipyard #1 data. Figure 5.1 provides a
histogram of Kw with normal and extreme-value distribution fits. According to the Chi-Squared
goodness-of-fit test, the normal distribution is better than the extreme-vale distribution. The ChiSquared test value is 30 for the normal distribution and 117 for the extreme-value distribution.
Table 5.1: Statistics on Stress Concentration Factor K w based on Shipyard #1 Data
N
Minimum
Mean
Maximum
Standard Deviation
Coefficient of Variation (%)
99% Confidence Interval on the Mean
38
1.270
1.370
1.440
0.040
3.22
1.353 to 1.387
1.357 to 1.383
1.359 to 1.381
40
5354C.FR
12
Normal
Extreme Value
K w = 1.0 + 0.5(tan )
1/ 4
10
n = 38
Mean = 1.366
StDev = 0.0436
Frequency
0
1.24
1.27
1.30
1.33
1.35
1.38
1.41
1.44
1.46
Kw
Figure 5.1: Histogram for K w with Extreme-Value and Normal Distribution Fits
5.3.1.3 Stress Concentration Factor Kte
The stress concentration factor Kte was computed using equation (5.5). Table 5.2 provides the
result of the statistical analysis of Kte based on Shipyard #1 data. Figure 5.2 provides a
histogram of Kte with normal and lognormal distribution fits. According to the Chi-Squared
goodness-of-fit test, the normal distribution is superior to the lognormal distribution. The ChiSquared test value is 8 for normal distribution and 9 for the lognormal distribution.
Table 5.2: Statistics on Stress Concentration Factor K te based on Shipyard #1 Data
N
Minimum
Mean
Maximum
Standard Deviation
19
1.02
1.26
1.47
0.13
10.0
1.18 to 1.34
1.20 to 1.32
1.21 to 1.31
41
5354C.FR
Frequency
7
3e 0
t
K te = 1 +
n = 19
Mean = 1.258
StDev = 0.126
Normal
Lognormal
0
0.91
1.02
1.13
1.24
1.35
1.47
1.58
K te
Figure 5.2: Histogram for K te with Normal and Lognormal Distribution Fits
5.3.2
Shipyard #2 Data
5.3.2.1 Stress Concentration Factor Kt

The stress concentration factor Kt was computed using equation (5.3), and assuming free rather
than fixed end conditions. Table 5.3 provides the result of the statistical analysis of Kt based on
Shipyard #2 data. Figure 5.3 shows a histogram of Kt with lognormal and exponential
distribution fits. According to the Chi-Squared goodness-of-fit test, the lognormal distribution is
better than the exponential distribution. The Chi-Squared test value is 29 for the lognormal
distribution and 34 for the exponential distribution.
Table 5.3: Statistics on Stress Concentration Factor K t based on Shipyard #2 Data
N
Minimum
Mean
Maximum
Standard Deviation
37
1.001642
5.000042
17.54774
4.445186
88.9
3.12 to 6.88
3.57 to 6.43
3.80 to 6.20
42
5354C.FR
24
K t = 1 +
Lognormal
Exponential
20
n = 37
Mean = 5.00
StDev = 4.44
16
Frequency
4 t
12
0
1.00
3.76
6.52
9.27
12.03
14.79
17.55
20.31
K ta
Figure 5.3: Histogram of K t with Lognormal and Exponential Distribution Fits

5.3.2.2 Stress Concentration Factor Kw
The stress concentration factor Kw was computed using equation (5.4). Table 5.4 provides the
result of the statistical analysis of Kw based on Shipyard #2 data. Figure 5.4 provides a
goodness-of-fit test, the normal distribution is superior to the extreme-value distribution. The
Chi-Squared test value is 19 for the normal distribution and 43 for the extreme-value
distribution.
N
Minimum
Mean
Maximum
Standard Deviation
118
1.27
1.43
1.57
0.05
3.77
1.418 to 1.442
1.421 to 1.439
1.422 to 1.438
43
5354C.FR
35
30
K w = 1.0 + 0.5(tan )
1/4
Frequency
25
Normal
Extreme Value
n = 118
Mean = 1.43
StDev = 0.0539
20
15
10
0
1.24
1.30
1.36
1.42
1.48
1.54
1.60
Kw
Figure 5.4: Histogram of K w with Normal and Extreme-Value Distribution Fits

5.3.2.3 Stress Concentration Factor Kte
The stress concentration factor Kte was computed using equation (5.5). Table 5.5 provides the
result of the statistical analysis of Kte based on Shipyard #2 data. Figure 5.5 provides a
histogram of Kte with lognormal and Weibull (smallest) distribution fits. According to the ChiSquared goodness-of-fit test, the lognormal distribution is superior to the Weibull (smallest)
distribution. The Chi-Squared test value is 14 for the lognormal distribution and 16 for the
Weibull (smallest) distribution.
N
Minimum
Mean
Maximum
Standard Deviation
59
1.00
1.23
1.72
0.17
13.5
1.173 to 1.287
1.187 to 1.273
1.194 to 1.266
44
5354C.FR
16
3e0
t
n = 59
Mean = 1.23
StDev = 0.166
K te = 1 +
14
Lognormal
Weibull
12
Frequency
10
8
6
4
2
0
0.90
1.00
1.10
1.21
1.31
1.41
1.51
1.62
1.72
1.82
K te
Figure 5.5: Histogram of K te with Lognormal and Weibull Distribution Fits
5.4
Butt-Welded Stiffeners Shipyard #2 Data
5.4.1 Stress Concentration Factor Kt

The stress concentration factor Kt was computed using the formula in Table 3.1. Table 5.6
provides the result of the statistical analysis of Kt based on Shipyard #2 data. Figure 5.6 shows
a histogram of Kt with exponential and extreme value (Type I Gumbel largest) distribution fits.
According to the Chi-Squared goodness-of-fit test, the exponential distribution is better than the
extreme-value distribution. The Chi-Squared test value is 0.5 for the exponential distribution
and 0.5 for the extreme-value distribution.
45
5354C.FR
Table 5.6: Statistics on Stress Concentration Factor K t based on Shipyard #2 Data

N
Minimum
Mean
Maximum
Standard Deviation
8
1.00
2.38
7.65
2.28
95.96
0.30 to 4.46
0.80 to 3.96
1.05 to 3.71
K t = 1 +
Frequency
Exponential
Extreme Value
3
4 t
n =8
Mean = 2.378
StDev = 2.282
0
1.00
2.66
4.32
5.99
7.65
9.31
Kt
Figure 5.6: Histogram of K t with Exponential and Extreme Value

(Type I Gumbel largest) Distribution Fits
5.4.2 Stress Concentration Factor Kw

The stress concentration factor Kw was computed using the formula in Table 3.1. Table 5.7
shows the result of the statistical analysis of Kw based on Shipyard #2 data. Figure 5.7 provides a
goodness-of-fit test, the normal distribution is superior to the extreme-value distribution.
The Chi-Squared test value is 45 for the normal distribution and 58 for the extreme-value
distribution.
46
5354C.FR

N
Minimum
Mean
Maximum
Standard Deviation
44
1.00
1.43
1.77
0.10
6.97
1.39 to 1.47
1.40 to 1.46
1.41 to 1.45
35
30
K w = 1. 0 + 0.5(tan )
1/ 4
Normal
Extreme Value
Frequency
25
n = 44
Mean = 1.425
StDev = 0.0994
20
15
10
0
0.87
1.00
1.13
1.26
1.39
1.51
1.64
1.77
1.90
Kw
Figure 5.7: Histogram of K w with Normal and Extreme Value

(Type I Gumbel Largest) Distribution Fits
47
5354C.FR
5.4.3 Stress Concentration Factor Kte

The stress concentration factor Kte was computed using the formula in Table 3.1. Table 5.8
provides the result of the statistical analysis of Kte based on Shipyard #2 data. Figure 5.8 shows
a histogram of Kte with exponential and extreme-value distribution fits. According to the ChiSquared goodness-of-fit test, the exponential distribution is superior to the extreme-value
distribution. The Chi-Squared test value is 6 for the exponential distribution and 106 for the
extreme-value distribution.
N
Minimum
Mean
Maximum
Standard Deviation
22
1.00
1.20
1.61
0.17
14.2
1.11 to 1.29
1.13 to 1.27
1.14 to 1.26
12
Exponential
Extreme Value
10
K te = 1 +
n = 22
Mean = 1.20
StDev = 0.170
Frequency
3e0
t
0
0.70
1.00
1.15
1.30
1.46
1.61
1.76
K te
Figure 5.8: Histogram of K te with Exponential and Extreme Value

(Type I Gumbel largest) Distribution Fits
48
5354C.FR
5.5
Cruciform Joints Shipyard #2 Data
The notch stress concentration factors and formulae applicable to cruciform joints were listed at
Table 3.2 and are listed again below for convenience.
K te = 1 +
6t2 e
t3 t3 t3 t3
l1 1 + 2 + 3 + 4
l1 l2 l3 l4
(5.6)
K g K w = 1.2 + 1.3(tan )
1/ 4
(5.7)
K g K w = 0.9 + 0.9(tan )
1/ 4
(5.8)
5.5.1 Stress Concentration Factor Kte

The stress concentration factor Kte was computed using the formula Table 3.2.
Table 5.9 provides the result of the statistical analysis of Kte based on Shipyard #2 data. Figure
5.9 shows a histogram of Kte with extreme-value and exponential distribution fits. According to
the Chi-Squared goodness-of-fit test, the extreme-value distribution is better than the exponential
distribution. The Chi-Squared test value is 12 for the extreme-value distribution and 185 for the
exponential distribution.
N
Minimum
Mean
Maximum
Standard Deviation
26
1.00
1.08
1.33
0.09
8.81
1.035 to 1.125
1.045 to 1.115
1.051 to 1.109
49
5354C.FR
16
14
Extreme Value
Exponential
K te = 1 +
12
t
t3 t3 t3
l1 + 2 + 3 + 4
l1 l 2 l3 l 4
n = 26
Mean = 1.078
StDev = 0.095
10
Frequency
6t2 e
3
1
8
6
4
2
0
1.00
1.07
1.13
1.20
1.26
1.33
1.39
K te
Figure 5.9: Histogram of K te with Extreme-Value and Exponential Distribution Fits
5.5.2 Cruciform Fillet Weld

The stress concentration factor as the product Kg Kw was computed using the formula in Table 3.2.
Table 5.10 provides the result of the statistical analysis of Kg Kw based on Shipyard #2 data.
Figure 5.10 shows a histogram of Kg Kw with normal and Weibull (smallest) distribution fits.
According to the Chi-Squared goodness-of-fit test, both have equal level of fit. The Chi-Squared
test value is 5 for both distributions.
According to the Chi-Squared goodness-of-fit test, the lognormal distribution is better than the
Weibull (smallest) distribution. The Chi-Squared test value is 12 for the lognormal distribution
and 28 for the Weibull (smallest) distribution.
50
5354C.FR
Table 5.10: Statistics on Stress Concentration Factor K g K w

based on Shipyard #2 Data (for Fillet Weld)
N
Minimum
Mean
Maximum
Standard Deviation
26
2.410
2.565
2.746
0.081
3.150
2.52 to 2.61
2.53 to 2.60
2.54 to 2.59
12
K g K w = 1. 2 + 1. 3(tan )
1/ 4
10
Frequency
n = 26
Mean = 2.565
StDev = 0.0807
Normal
Weibull
0
2.34
2.41
2.48
2.54
2.61
2.68
2.75
2.81
Kg Kw
Figure 5.10: Histogram of K g K w with Normal and Weibull (smallest)

Distribution Fits for Fillet Weld
5.5.3 Cruciform Full Penetration Weld

The stress concentration factor Kg Kw was computed using Eq. 5-8. Table 5.11 provides the
result of the statistical analysis of Kg Kw based on Shipyard #2 data. Figure 5.11 shows a
histogram of Kg Kw with lognormal and Weibull (smallest) distribution fits.
51
5354C.FR
Table 5.11: Statistics on Stress Concentration Factor K g K g based on Shipyard #2 Data

(for Full Penetration Weld)
N
Minimum
Mean
Maximum
Standard Deviation
26
2.42
2.55
2.70
0.07
2.56
2.51 to 2.59
2.52 to 2.58
2.53 to 2.57
14
12
K g K w = 0. 9 + 0.9(tan )1 / 4
Lognormal
Weibull
Frequency
10
n = 26
Mean = 2.550
StDev = 0.0652
0
2.37
2.42
2.48
2.53
2.59
2.64
2.70
2.75
Kg Kw
Figure 5.11: Histogram of K g K w with Lognormal and Weibull Distribution

Fits for Full Penetration Weld
5.6
Total Stress Concentration Factor
The total stress concentration factor K was calculated based on Eq. 5-1 and on the individual
concentration factor values obtained from the statistical analyses of the previous sections.
Statistical analyses were performed as well as histograms with best-known theoretical
distributions and generated for each weld type such as butt-welded plates, butt-welded stiffeners,
and intercostals joints.
52
5.6.1
5354C.FR
Butt-Welded Plates
5.6.1.1 Shipyard #1
The total concentration factor K was computed using equation (5-1). The values for Kn and Kg in
equation (5-1) were taken as 1. Table 5.12 shows the result of the statistical analysis of K based
on the Shipyard #1 data. Figure 5.12 provides a histogram of K with normal and lognormal
distribution fits. According to the Chi-Squared goodness-of-fit test, the normal distribution is
better than the lognormal distribution. The Chi-Squared test value is 5.4 for the normal
distribution and 6.2 for the normal distribution.
based on Shipyard #1 Data
N
Minimum
Mean
Maximum
Standard Deviation
19
1.352 Kg Kn
1.712 Kg Kn
2.004 Kg Kn
0.168 Kg Kn
9.784
(1.613 to 1.811) Kg Kn
(1.637 to 1.787) Kg Kn
(1.649 to 1.775) Kg Kn
53
5354C.FR
Frequency
Normal
Lognormal
Kn = Kg = 1
n = 19
Mean = 1.352
StDev = 0.168
4
3
2
1
0
1.19
1.35
1.52
1.68
1.84
2.00
2.17
Figure 5.12: Histogram of total Stress Concentration Factor K with Normal

and Lognormal Distribution Fits for Butt-Welded Plates (Shipyard #1)
5.6.1.2 Shipyard #2
The total concentration factor K was computed using equation (5-1). The values for Kn and Kg in
equation (5-1) were taken as 1. Table 5.13 shows the result of the statistical analysis of K based
on the Shipyard #2 data. Figure 5.13 provides a histogram of K with lognormal and exponential
distribution fits. According to the Chi-Squared goodness-of-fit test, the lognormal distribution is
better than the exponential distribution. The Chi-Squared test value is 22.3 for the lognormal
distribution and 29.9 for the exponential distribution.
N
Minimum
Mean
Maximum
Standard Deviation
37
1.552 Kg Kn
9.221 Kg Kn
19.360 Kg Kn
8.574 Kg Kn
92.982
(5.590 to 12.852) Kg Kn
(6.458 to 11.984) Kg Kn
(6.903 to 11.540) Kg Kn
54
5354C.FR
25
20
Kn = Kg = 1
Frequency
Lognormal
Exponential
n = 37
Mean = 9.221
StDev = 8.574
15
10
0
1.55
7.12
12.68
18.24
23.81
29.37
34.93
Figure 5.13: Histogram of Total Stress Concentration Factor K with Normal

and Lognormal Distribution Fits for Butt-Welded Plates (Shipyard #2)
5.6.2 Butt-Welded Stiffeners
The total concentration factor K was computed for butt-welded stiffeners using equation (5-1).
The values for Kn and Kg in equation (5-1) were taken as 1. Table 5.14 shows the result of the
statistical analysis of K based on the Shipyard #2 data. Figure 5.14 provides a histogram of K
with lognormal and normal distribution fits. According to the Chi-Squared goodness-of-fit test,
the lognormal distribution is better than the normal distribution. The Chi-Squared test value is
0.4 for the lognormal distribution and 1.3 for the normal distribution.
Table 5.14: Statistics on Total Stress Concentration Factor K for Butt-Welded Stiffeners
N
Minimum
Mean
Maximum
Standard Deviation
8
1.000 Kg Kn
4.326 Kg Kn
14.758 Kg Kn
4.523 Kg Kn
104.558
(0.207 to 8.446) Kg Kn
(1.192 to7.461) Kg Kn
(1.696 to 6.957) Kg Kn
55
5354C.FR
7
6
Normal
Lognormal
Kn = Kg = 1
Frequency
n=8
Mean = 4.326
StDev = 4.523
4
3
2
1
0
1.00
7.88
14.76
21.64
Figure 5.14: Histogram of Total Stress concentration factor K with Lognormal

and Normal Distribution Fits for Butt-Welded Stiffeners (Shipyard #2)
5.6.3
Cruciform Joints
5.6.3.1 Fillet Weld

The total concentration factor K was computed for cruciform joints (fillet weld) using equation
(5-1). The values for Kn and Kt in equation (5-1) were taken as 1. Table 5.15 shows the result
of the statistical analysis of K based on the Shipyard #2 data. Figure 5.15 provides a histogram
of K with lognormal and normal distribution fits. According to the Chi-Squared goodness-of-fit
test, the lognormal distribution is better than the normal distribution. The Chi-Squared test value
is 16.5 for the lognormal distribution and 24.7 for the normal distribution.
(Fillet Weld) based on Shipyard #2 Data
N
Minimum
Mean
Maximum
Standard Deviation
26
2.442 Kn Kt
2.764 Kn Kt
3.392 Kn Kt
0.259 Kn Kt
9.354
(2.633 to 2.895) Kn Kt
(2.665 to 2.863) Kn Kt
(2.681 to2.847) Kn Kt
56
5354C.FR
12
Lognormal
Normal
10
Kn = Kt = 1
Frequency
n = 26
Mean = 2.764
StDev = 0.259
0
2.23
2.44
2.65
2.87
3.08
3.29
3.51
Figure 5.15: Histogram of Total Stress concentration factor K with Lognormal

and Normal Distribution Fits for Cruciform Joints, Fillet Weld (Shipyard #2)
5.6.3.2 Full Penetration Weld
The total concentration factor K was computed for cruciform joints (full penetration weld) using
equation (5-1). The values for Kn and Kt in equation (5-1) were taken as 1. Table 5.16 shows
the result of the statistical analysis of K based on the Shipyard #2 data. Figure 5.16 provides a
histogram of K with lognormal and normal distribution fits. According to the Chi-Squared
goodness-of-fit test, the lognormal distribution is better than the normal distribution. The ChiSquared test gave a value of 25.4 for the lognormal distribution and a value of 36.7 for the
normal distribution.
(Full Penetration Weld) based on Shipyard #2 Data
N
Minimum
Mean
Maximum
Standard Deviation
26
2.479 Kn Kt
2.750 Kn Kt
3.576 Kn Kt
0.285 Kn Kt
10.355
(2.607 to2.894) Kn Kt
(2.641 to 2.860) Kn Kt
(Kn Kt2.659 to2.842) Kn Kt
57
5354C.FR
16
14
Lognormal
Normal
12
Kn = Kt = 1
Frequency
10
n = 26
Mean = 2.750
StDev = 0.285
8
6
4
2
0
2.26
2.48
2.70
2.92
3.14
3.36
3.58
Figure 5.16: Histogram of Total Stress concentration factor K with Lognormal and
Normal Distribution Fits for Cruciform Joints, Full Penetration Weld (Shipyard #2)
5.7
Summary and Discussion of the Results
In this section, the statistical characteristics of stress concentration factors Ks for fatigue design
of ship structures are summarized and tabulated based on the data collected at the Shipyard #1
and Shipyard #2. These characteristics include the mean , the coefficient of variation (COV),
and the underlying probability distribution for each stress concentration factor component K.
However, since these results should be revised as new data (e.g., more shipyard data) and
researches on the subject emerge, and caution must be taken when using these results in
reliability assessment and reliability-based fatigue design of ship structures. They represent only
the ranges and the weighted averages of the statistical values collected from two limited sources,
Shipyard #1 and Shipyard #2; the number of samples in the data sets is not sufficient to provide a
high level of confidence in the results. Although the distributions derived above represent the
best fits to the available data, it is probable that a single set of distributions will be appropriate to
the types of phenomena involved in misalignment, rather that having different distributions for
each yard, as indicated (in some cases) by the current data sets.
The summary of the statistical characteristics of the individual stress concentration factors as
well as the total stress concentration factors Ks are provided in Tables 5.17 and 5.18. The
different stress concentration factors, as shown in Table 5.17, include factors that are
standardized and defined by the various classification societies as employed in Eqs 5-3, 5-4, 5-5,
5-6, 5-7, and 5-8, and in Figures 5.1, 5.2, 5.3, 5.12, 5.13, and 5.14.
58
5354C.FR
These factors include Kt, Kw, Kte, and KgKw that correspond to various weld types such as buttwelded plates, butt-welded stiffeners, and intercostals joints.
Table 5.17 provides statistical summary for the total concentration factor.
One reason for having statistically treated the data as separate data sets, i.e., Shipyard #1 as one
set and Shipyard #2 as another, is that the number of samples due to Shipyard #1 data set is not
statistically significant and the samples only cover one type of weld. In order to combine the two
shipyard data sets it is necessary that both should have comparable and significant samples
covering all types of welds, and better yet to have more than two shipyard data sets each with
sufficient number of data points.
59
5354C.FR
Table 5.17: Summary of the Statistical Analyses for the Individual Stress Concentration Factors K s based on Shipyard 1 & 2 Data
Weld Type
Butt-Welded Plates
Butt-Welded
Stiffeners
Descriptive Statistics
Sample Mean
Coefficient of
Variation (%)
99, 95, and 90 %
Confidence Interval
on
Kte
1.26
Kt
5.00
n/a
3.22
10.0
88.90
n/a
n/a
n/a
Distribution Type
n/a
Sample Mean
Coefficient of
Variation (%)
99, 95, and 90 %
Confidence Interval
on
n/a
mean 0.0167
mean 0.0127
mean 0.0107
Normal (30)*
Extreme Value (Type I
Gumbel largest) (117)**
n/a
n/a
Distribution Type
Cruciform Joints
Kt
n/a
Shipyard #1
Kw
1.37
Sample Mean
Coefficient of
Variation (%)
99, 95, and 90 %
Confidence Interval
on
Distribution Type
n/a
n/a
n/a
n/a
n/a
n/a
Fillet Weld
Kg Kw
n/a
mean 0.0768
mean 0.0585
mean 0.0491
Normal (8)*
Lognormal (9)**
mean 1.882
mean 1.432
mean 1.202
Lognormal (29)*
Exponential (34)**
n/a
2.38
n/a
n/a
95.96
n/a
n/a
n/a
n/a
Full Penetration Weld

Kg Kw
n/a
Kte
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
mean 2.076
mean 1.580
mean 1.326
Exponential (3)*
Fillet Weld
Kg Kw
2.57
3.15
mean 0.0455
mean 0.0346
mean 0.0290
Gumbel largest) (12)*
Exponential (199)**
Shipyard #2
Kw
1.43
3.77
mean 0.0119
mean 0.0090
mean 0.0076
Normal (19)*
1.43
6.97
mean 0.0388
mean 0.0295
mean 0.0248
Normal (45)*
Full Penetration Weld
Kg Kw
2.55
2.56
mean 0.0409
mean 0.0311
mean 0.0261
Normal (5)*
Weibull (smallest) (5)**
Kte
1.23
13.5
mean 0.0570
mean 0.0434
mean 0.0364
Lognormal (14)*
1.20
14.2
mean 0.0934
mean 0.0710
mean 0.0596
Exponential (6)*
Kte
1.08
8.81
mean 0.0354
mean 0.0269
mean 0.0226
Lognormal (12)*
(-) = Chi-Squared Test Value

* = First best fit distribution according to the Chi-Squared goodness of fit
** = Second best fit distribution according to the Chi-Squared goodness of fit
n/a = not available
60
5354C.FR
Table 5.18: Summary of the Statistical Analyses for the Total Stress
Concentration Factors K s based on Shipyard 1 & 2 Data
Weld Type
Butt-Welded
Plates
Descriptive
Statistics
Sample Mean
Coefficient of
Variation (%)
99, 95, and 90 %
Confidence
Interval on
Distribution Type
Butt-Welded
Stiffeners
Sample Mean
Coefficient of
Variation (%)
99, 95, and 90 %
Confidence
Interval on
Distribution Type
9.78
92.98
mean 3.631
mean 2.763
mean 2.318
Lognormal (22.3)*
Exponential (29.9)**
4.326
n/a
104
n/a
n/a
K
n/a
Full
Penetration
Weld
K
n/a
n/a
n/a
n/a
n/a
n/a
n/a
Fillet
Weld
Cruciform
Joints
Shipyard #2
K
9.221
mean 0.0990
mean 0.0753
mean 0.0632
Normal (5.4)*
Lognormal (6.2)**
n/a
Distribution Type
Sample Mean
Coefficient of
Variation (%)
99, 95, and 90 %
Confidence
Interval on
Shipyard #1
K
1.712
mean 4.119
mean 3.134
mean 2.630
Lognormal (0.4)*
Normal (1.3)**
Fillet Weld
Full Penetration
Weld
K
2.764
K
2.750
9.354
10.355
mean 0.1306
mean 0.0994
mean 0.0834
Lognormal
(16.5)*
Normal (24.7)**
mean 0.1439
mean 0.1095
mean 0.0919
Lognormal (25.4)*
Normal (36.7)**
(-) = Chi-Squared Test Value

* = First best fit distribution according to the Chi-Squared goodness of fit
** = Second best fit distribution according to the Chi-Squared goodness of fit
n/a = not available
61
6.
5354C.FR
DISCUSSION OF DATA
6.1
Overall Stress Concentration Factors
As noted in Section 4.2, the shipyard data analyzed in Section 5 has been taken from large block
assemblies and from completed sections of ships. The fabrication quality at the panel line level
in both yards was sufficiently good that measurements were effectively within the accuracy
limits of the data collection tools. This indicates that (subject to adequate weld properties)
fatigue problems are much more likely to manifest themselves at block/unit/assembly joints than
within panels.
Table 6.1 compares the stress concentration factors derived in Section 5 for Shipyard #2 to the
Det Norske Veritas (DNV) design default values presented in Section 3. It can be seen that for
plate connections, the overall mean values derived from the observed data are close to the
defaults. For stiffener connections, the shipyard values are close to the defaults for the cruciform
connections, but those for the butt-welded stiffeners indicate that the yard can better the default
assumptions. The situation with regard to the scatter in the results will be discussed further in
Section 6.2.
62
5354C.FR
Table 6.1: Comparison of Survey K Values to DNV Defaults

Weld Type
Shipyard #2 Data
Kw
Kte
1.43
1.23
0.05
0.17
1.48
1.4
1.53
1.57
2
3
1.5
1.45
n/a
n/a
Mean
Std. Dev.
Butt Welded Plates (Assumed Mean + 1 Std. Dev.
Mean + 2 Std. Dev.
free ends, = 6)
DNV Default Value
Std. Dev. of DNV default value
Kt
5
4.44
9.44
13.88
1
4.03
1.03
Mean
Std. Dev.
Mean + 1 Std. Dev.
Mean + 2 Std. Dev.
DNV Default Value
Kt
2.38
2.28
4.66
6.94
1
4.6
0.4
Kw
1.43
0.1
1.53
1.63
2
1.5
n/a
Kg Kw
2.57
0.08
2.65
2.73
4
2.5
Kte
1.08
0.09
1.17
1.26
5
1.18
Butt Welded Stiffeners

(Assumed free ends, = 6)
Cruciform Joints
Mean
Std. Dev.
Mean + 1 Std. Dev.
Mean + 2 Std. Dev.
DNV Default Value
Kte
1.2
0.17
1.37
1.54
3
1.45
n/a
Ktotal
9.22
8.57
17.79
26.36
8.9
Ktotal
4.33
4.52
8.85
13.37
10.18
Ktotal
2.76
0.26
3.02
3.28
3.01
1. DNV default for Kt based upon e = 6 mm. Assumed that the seam is at the middle of supports, such that
Kt = 1+l(e/t). Used the plate thicknesses in the survey to calculate Kt. the default value given is the mean of
the sample.
2. DNV default for Kw based upon q = 45
3. DNV default for Kte (butt joint) based upon misalignment = e0 = 0.15t
4. DNV default for KgKw based upon = 45
5. DNV default for Kte (cruciform joint) based on e = 0.3t. The Kte values are dependent on the distance the
joint is from supports. The values from the survey were used to determine the default K value.
It should be noted that most aspects of the comparisons above are relevant to fatigue analysis
using any classification society methodology, not only that of DNV. All the SCF numbers
calculated for axial and angular misalignment have meaning in all fatigue analysis standards. As
previously discussed, it is assumed that the test samples used to develop S-N curves had no
discernible misalignment. Therefore, in principle it does not matter if the nominal stress, hot
spot stress, or notch stress approach is being used - stress ranges used in the S-N curve
calculation should be increased to account for any misalignment. Most standards thus provide
formulae to determine the secondary bending stress caused by the axial loading of the misaligned
joints, though in practice the nominal stress approach is normally used without any correction.
63
5354C.FR
Conversely, it is important to consider that the SCF values determined for imperfect weld profile
relate only to the notch stress analysis (DNV) approach. Thus, while the overall mean SCFs
presented on the previous page are of interest, so too are the individual SCFs. This is discussed
further in Section 7.
6.1.1 Stress Concentration Factor Kt
It can be seen from the results presented in Table 6.1 that the largest individual stress
concentration factor (applicable to butt connections) in terms of both mean value and standard
deviation tends to be Kt. The calculation of Kt requires assumptions in addition to
measurements; a key assumption in the analyses above is that the plate ends are free rather than
fixed at the end supports. In both shipyards surveyed, the typical configuration was as shown in
Figure 6.1, where unit butt welds were located between transverse stiffeners. In a buckling
analysis, the plate would normally be considered as simply supported, especially for the
relatively thin plate being used in the vessels under construction. However, the assumption is
certainly conservative, especially for fatigue analysis.
Weld Seam
Direction
of Stress
Figure 6.1: Typical Butt Weld

As noted in Section 3, the formula for Kt is:
K t = 1 +
s
e
= 1+
4 t
t
where = 6 for pinned (simply supported) ends

= 3 for fixed ends
64
5354C.FR
Thus, changing the assumed fixity effectively reduces the SCF by up to 50 %. The effect on the
survey data is shown in Table 6.2 (for butt-welded plate only). It can be seen that the fixity
assumption has a major effect on fatigue life prediction, and this appears to be an area in which
more work needs to be undertaken to provide designers with useful guidance.
Table 6.2: Effect of Fixed vs. Free and Assumption on SCFs
Weld Type
Mean
Std. Dev.
free ends, = 6)
Mean + 2 Std. Dev.
DNV Default Value
Kt
5.00
4.44
9.44
13.88
4.03
1.03
Mean
Std. Dev.
fixed ends, = 3)
Mean + 2 Std. Dev.
DNV Default Value
Kt
3.00
2.23
5.23
7.46
2.52
0.52
Shipyard #2 Data
Kw
Kte
1.43
1.23
0.05
0.17
1.48
1.4
1.53
1.57
1.5
1.45
n/a
n/a
Kw
1.43
0.05
1.48
1.53
1.5
n/a
Kte
1.2
0.17
1.37
1.54
1.45
n/a
Ktotal
9.22
8.57
17.79
26.36
8.9
Ktotal
5.51
4.324
8.85
13.37
5.48
6.2
Variability
The scatter in the results for SCFs obtained from the shipyard data is considerable, and for plate
butt welds is more than is implied in standard S-N based fatigue analysis procedures. To
illustrate this, it is necessary to refer back to the discussion presented at Section 3.4.
In this DNV analysis approach (and that of other classification societies), stress ranges are based
on mean stress concentration factors, taking into account expected (default) values for
misalignment, etc. The potential variability in the actual outcomes is accounted for by displacing
the design S-N curve by two standard deviations (2SD) from the underlying curve (Figure 3.13).
The 2SD value in the DNV curves approximates to an additional stress concentration multiplier
of approximately 2.5, whereas as shown in Table 6.1 the 2SD shipyard values from this project
ranged from a low of less than 1.2 for the cruciform joints to as high as 2.9 for the butt-welded
plates (ratio of 2SD to mean total SCF).
These results imply that it may be inappropriate to use the same 2SD safety factor for all types
of joint. In fact, the DNV approach is unusual in this regard, as most design S-N curves for
specific joint types have individual 2SD values, as shown in Figure 6.2 and Table 6.3. Buttwelded plates are normally analyzed using a D, E F or F2 curve (depending on location
and quality), and cruciforms by F2. It can be seen that the variability in the shipyard plate
sample is somewhat greater than the 2SD assumption underlying any of the S-N curves.
65
5354C.FR
Mean and Design (2SD) BS7608 & DNV S-N Curves

(All curves are for welded joints in air)
Stress Range (MPa)
1000
Cmean
Cminus 2 std
Fmean
Fminus 2 std
DNV II mean
DNV II minus 2 std
F2mean
F2minus 2 std
Emean
Eminus 2 std
100
10
1.00E+05
1.00E+06
1.00E+07
1.00E+08
1.00E+09
Stress Cycles
Figure 6.2: S-N Design Curves Showing 2SD Offsets
Table 6.3: 2SD Equivalent SCFs

Curve
C
D
E
F
F2
DNV
2SD Eq
SCF
2.56
2.62
3.16
2.74
2.86
2.51
It should also be recognized that the S-N curve 2SD offsets are intended to account for scatter
from all sources, including the internal imperfections listed in Section 3.2. Therefore, the fatigue
life of the shipyard samples would be expected to display more variability than that due to the
geometric imperfections alone. Therefore, while the results for the butt-welded plates certainly
imply more scatter in fatigue life outcome than implied in the design methodology, those for the
cruciforms do not necessarily guarantee less scatter in the final outcome.
66
5354C.FR
A further point is that the plate data represents final unit assemblies, whereas equivalent
cruciform connections could not be accessed during either shipyard survey. If the final assembly
stage is most likely to show poor outcomes, the collected data set may not reflect worse case
results for cruciform connections.
67
7.
5354C.FR
FATIGUE LIFE CASE STUDY
7.1
Basis for Analysis
In order to illustrate the potential significance of the data collected in this project, it has been
used to revisit analyses undertaken in SSC Project 427 Life Expectancy Assessment of Ship
Structures8. For a typical very large crude oil carrier (VLCC) (Figure 7.1) operational profile
and scantlings a lifetime stress history was developed (Figure 7.2). Fatigue life has been
estimated using various assumptions regarding stress concentration factors for deck butt welds.
Such welds are not normally considered as the most probable locations for fatigue cracking
problems, but can still provide a useful illustration of the significance of the tolerances measured
during the project.
Figure 7.1: Example VLCC (from SSC 427)
Dinovitzer, A. Life Expectancy Assessment of Ship Structures SSC Report 427, 2003.
68
5354C.FR
Stress History in Deck of Tanker

1200000
SCF = 1
SCF = 2
SCF = 3
Cycles per year (could be per 20 years)
1000000
800000
600000
400000
200000
0
1.00
10.00
100.00
1000.00
10000.00
-200000
Stress in Deck PLating
Figure 7.2: VLCC Notional Stress History

The data for butt weld connections from Shipyard #2 have been utilized in order to develop a
picture of the expected variability in fatigue life resulting from the geometric stress concentration
factors measured in the shipyard. The outcomes have been based on two S-N curves - the
standard curve utilized with the DNV notch stress approach, and the standard curve for class F2
structural details from BS 5400 (nominal stress approach), as shown in Figure 6.2. These
outcomes have then been compared with notional design outcomes using these two
methodologies.
In a large tanker, plate thicknesses will be considerably greater than those used in most of the
assemblies surveyed in Shipyards #1 and #2. Also, such a ship would be built under fatigue
class notation, with (presumably) tighter tolerances and/or more extensive inspection. It is
therefore probable that most of the individual tolerance parameters for the tanker would differ
from those measured in the project. Therefore, the potential outcomes for the notch stress
analyses are shown for various hypothetical values of Ktotal as follows:
Case 1: Ktotal = Kt.Kte.Kw

Case 2: Ktotal = Kte.Kw
Case 3: Ktotal = Kw
The first of these assumes that no significant differences would exist between the notional tanker
fabrication tolerances and the actual measured values. The second assumes that angular
misalignment becomes negligible for the heavier plate used in large tanker construction.
69
5354C.FR
The third is used to examine the effect of weld geometry in isolation, and to illustrate
comparisons between the notch and nominal stress approach.
The angular misalignment stress concentration factor Kt has been taken from analysis using
fixed end restraint, as it has been assumed that the heavier structure is more likely to behave in
this way. Kt has also been adjusted to account for thickness effects, noting again that:
K t = 1 +
s
e
= 1+
4 t
t
The notional tanker hull plate is approximately twice the thickness values for the measured data,
and so Kt values have been recalculated on this basis. These changes to the calculation of Kt
result in a distribution of outcomes as shown in Figure 7.3, which can be compared with the
values shown at Figure 5.3 for the effect of the same absolute tolerances on the structures
actually surveyed.
24
S
4 t
n = 37
Mean = 2.00
StDev = 1.11
K t = 1 +
Lognormal
Exponential
20
Frequency
16
12
0
1.00
1.69
2.38
3.07
3.76
4.45
5.14
5.83
K ta
Figure 7.3: Kt Adjusted for Thickness, Fixity

The misalignment factor Kte also includes plate thickness in its formula. Most classification
societies express permissible misalignment as a function of plate thickness. In the absence of
better information, it has been assumed that the same proportional misalignment would be
experienced for the notional tanker as for the actual structures surveyed. This assumption may
warrant further investigation in future.
70
5354C.FR
7.2
Case Outcomes
Figures 7.4 to 7.6 illustrate the outcomes of the analyses using various combinations of the stress
concentration factors and the DNV notch stress approach. Table 7.1 then compares the outcomes
with the predictions that would be generated by standard notch and nominal stress analyses; the
former using the DNV default tolerance (and hence SCF) values for a vessel of this
configuration, and the latter only applying the F2 S-N curve to the nominal (SCF = 1) stress
history. Fatigue life is used as convenient shorthand for crack initiation and growth to the point
where some form of intervention might be anticipated.
Tanker Case Study
Fatigue Life of Butt Welds in Deck - Based on Shipyard #2 Survey Data
Nominal Stress Derived from Wave Bending Moment - 20 Year History
Using DNV Notch Stress Approach, S-N Curve II for Welded Joints in Air
Fatigue Life (years)
20
25.0
198.0
11.8
2.8
1.1
0.5
0.3
0.2
0.1
18
16
20.0
No. Samples
Fatigue Life (yrs)
12
15.0
10
8
10.0
Fatigue Life - Years
Number of Samples
14
6
4
5.0
2
0
0.0
1.55
2.79
4.03
5.27
6.51
7.75
8.99
10.23
Kt = Kta*Kte*Kw
Figure 7.4: Case 1 Fatigue Life Analysis with Ktotal = Kt.Kte .Kw
71
5354C.FR
Tanker Case Study

9
250.0
229.2
184.9
151.3
125.4
105.1
88.9
75.9
65.3
225.0
200.0
7
No. Samples
150.0
5
125.0
4
100.0
Number of Samples
175.0
Fatigue Life (yrs)
3
75.0
2
50.0
25.0
0.0
1.48
1.59
1.70
1.81
1.92
2.03
2.14
2.25
Kt = Kte*Kw
Figure 7.5: Case 2 Fatigue Life Analysis with Ktotal = Kte .Kw
Tanker Case Study


12
350.0
308
286
266
248
232
217
203
190
300.0
10
No. Samples
250.0
8
200.0
6
150.0
Number of Samples
Fatigue Life (yrs)
4
100.0
50.0
0.0
1.34
1.37
1.41
1.44
1.47
1.51
1.54
1.57
Kt = Kw
Figure 7.6: Case 3 Fatigue Life Analysis with Ktotal = Kw
72
5354C.FR
Table 7.1: Fatigue Life Prediction Comparisons

Case
Mean K (mod. Mean Fatigue DNV Default

survey data)
Life (yrs)
K
DNV Default
Fatigue Life
(yrs)
Nominal (F2)
Stress Fat.
Life (yrs)
Notch 1: K = Kta*Kte*Kw
3.66
15.1
3.82
13.3
132.7
Notch 2: K = Kte*Kw
1.80
126.4
2.18
71.5
132.7
Notch 3: K = Kw
1.44
247
1.5
219.5
132.7
The variability in outcome illustrated in these Figures and in Table 7.1 illustrates the importance
of assigning appropriate values to anticipated fabrication tolerances at the design stage, and also
of applying effective inspection and acceptance strategies during construction. While the mean
fatigue life in Case 1 might be considered marginally acceptable, approximately 20% of the
sample have predicted life expectancies of less than 1 year, which would almost certainly not be
acceptable. All of these samples in the unadjusted data set had one or more measurements
outside the nominal tolerance limits being applied by the shipyard and by the classification
societies for the structures. However, as noted the work being undertaken was not to fatigue
class notation, and the actual angular misalignment of thicker plate would be expected to be
much lower than that found in the thinner plate of the surveys. Therefore, Case 1 can be
regarded as the extreme worse case outcome, although an analysis based on the survey data
could easily be used to generate and justify the results.
Notch stress fatigue life analysis requires the designer to consider the influences of various
fabrication tolerances, although the data required to address them may not normally be available
(hence the use of defaults). Where nominal stress analysis is undertaken, designers often lack
useful guidance on which underlying S-N curves are most appropriate, and on what (if any)
situation specific factors should also be taken into account. It is normally assumed that weld
geometry effects (Kw) are incorporated in the nominal S-N curves. It is much less clear whether
any misalignment effects are included, although they are typically excluded from analysis. From
Table 7.1, a nominal stress analysis would give similar outcomes to the mean of notch stress
analyses based on actual linear misalignment and actual weld profiles.
73
5354C.FR
8.
CONCLUSIONS
1.
There is very little information in the public domain on actual fabrication tolerances
achieved by the shipbuilding industry. Neither of the two yards visited during this
project collect data of the type required to analyze fatigue performance. Classification
societies do not publish the basis for their standards or for their design guidance.
2.
Measurement of many tolerance/imperfection parameters is difficult and timeconsuming. There are no simple means of measuring many intercostals tolerances to an
adequate degree of accuracy. For many bulk parameters (for long connections) there
are no generally accepted procedures to define appropriate values for use in fatigue
analysis.
3.
Documentation of the derivation of standard S-N fatigue analysis curves does not include
sufficient information on the quality of experimental samples to allow the influence of
imperfections of various types to be isolated. In turn, this increases the difficulty of
identifying the significance of actual fabrication tolerances and imperfections.
4.
Shipyard measurements taken in this project indicate that modern automated panel lines
achieve fabrication tolerances that are very much better than those assumed in published
guidelines for fatigue analysis. This implies that production engineering should aim to
allow the maximum number of fatigue-sensitive connections to be made using automated
shop processes. Similarly, in-service inspection for fatigue cracking should concentrate
on joints made late in the erection sequence, rather than those made under shop
conditions.
5.
Measurements of block and large assembly connections indicate that the mean fabrication
tolerances are similar to default values assumed in published fatigue design guidance
notes. However, the level of scatter is higher than that implied by standard fatigue
analysis practice. This implies that more fatigue cracking may occur earlier in life than
expected, unless fabrication and inspection procedures catch and reject most samples
outside nominal tolerance limits.
6.
The level of scatter appears to be highly dependent on the type of joint. This is broadly
in line with much fatigue design practice, but the variability observed between joint types
was greater than expected. This implies that lower safety factors may be acceptable for
some types of joints, though more understanding of the contribution of internal defects to
the overall variability of fatigue outcomes would be needed to confirm this. It also
implies (more strongly) that in-service inspection should focus on joint types expected to
show high variability.
74
5354C.FR
7.
The measurements taken in this project covered a relatively small range of joints,
material properties, thicknesses, and joining (welding) techniques. It is highly probable
that more extensive data collection would allow more trends to be identified, and thus
assist designers in selecting appropriate assumptions on expected tolerances for the type
of project under development.
8.
Fatigue design guidance provided by most classification societies requires designers to

make assumptions that can have a considerable influence on predicted outcomes. For
example, in nominal stress analysis it is unclear what (if any) fabrication tolerances
should be treated explicitly. In notch stress analysis, assumptions regarding effective end
fixity of plates and stiffeners can be crucial to the acceptability of predicted results.
Predictions from fatigue analyses using both the nominal and notch stress approaches can
vary dramatically depending on the analysts approach and the relevance of the tolerance
data available.
75
5354C.FR
9.
RECOMMENDATIONS
1.
A number of the conclusions reported in Section 8 are based on relatively small samples
of data from projects that were not subject to fatigue notation quality analysis or
inspection procedures (commercial or military). Therefore, it would be highly desirable
to collect similar data for these types of project. Based on experience in this project, this
will require the support and cooperation of the owner and/or classification/inspection
authority to gain access to the shipyards and vessels.
2.
Conclusion (4) that fatigue damage is most likely at the final block or unit assemblies is
important to the development of through-life inspection strategies, and should be tested
by examination of fatigue damage records against production drawings.
3.
Future refinement of fatigue analysis, methodologies should consider the actual material
grades, thicknesses, and fabrication processes that are to be used in a project when
making assumptions regarding expected fabrication tolerances. Guidance in existing
fatigue analysis methodologies tends to be too general in nature.
4.
For design and build projects, owners and certifying authorities may wish to ensure that
design and analysis assumptions are matched to actual shipyard practices and to the
standards of fabrication achieved on previous and similar projects. This may require
many shipyards to revise their current data collection and reporting procedures.
5.
Due the difficulty in collecting many tolerance parameters and to the potential variability
in the approaches that could be taken in characterizing certain values, it would be
desirable to develop a practical guide to tools and measuring methods that could be used
by inspectors (shipyard, class, and owner representatives). Material developed in this
project and reported herein could provide some elements of such a guide.
6.
Further work by the SSC is recommended to address the end fixity issue for the
determination of stress concentration factors. This would be most effectively
accomplished by a combination of experimentation and numerical analysis.
76
5354C.FR
APPENDIX A
BIBLIOGRAPHY
A-1
5354C.FR
1. American Bureau of Shipping Shipbuilding and Repair Quality Standard for Hull Structures
During Construction 1998
2. American Society of Testing and Materials (ASTM Designation: F-1053-87)
3. Assakkaf, I., and Ayyub, B.M., 2000, Load and Resistance Factor Design Approach for
Fatigue of Marine Structures, 8th ASCE Specialty Conference on Probabilistic Mechanics
and Structural Reliability, ASCE, PMC 2000-169, University of Notre Dame, South Bend, 6
pages.
4. Atua, K.I., Assakkaf, I., and Ayyub, B.M., 1996, Statistical Characteristics of Strength and
Load Random Variables of Ship Structures, Proceedings of the 1996 ASCE Specialty
Conference on Probabilistic Mechanics and Structure Reliability, edited by Frangopol and
Grigoriu, 106-109.
5. Ayyub, B.M., and Assakkaf, I., 1998, LRFD Rules for Naval Surface Ship Structures:
Reliability-based Load and Resistance Factor Design Rules, Part IV Fatigue, Naval
Surface Warfare Center, Carderock Division, U.S. Navy, 25.
6. Ayyub, B.M., Assakkaf, I., and Atua, K., 1998, Development of LRFD Rules for Naval
Surface Ship Structures: Reliability-based Load and Resistance Factor Design Rules, Part III
Stiffened and Gross Panels, Naval Surface Warfare Center, Carderock Division, U.S.
Navy, 400.
7. Ayyub, B.M., Muhanna, R., and Daniel Bruchman, 1997, Uncertainty in Marine Structural
Strength due to Variability in Geometry and Material Properties, Technical Report
NSWCCD-TR-65-96/05, Naval Surface Warfare Center, Carderock, MD.
8. Ayyub, B.M., Muhanna, R., and Bruchman, D.D., 1996, Uncertainty in Marine Structural
Strength due to Variability in Geometry and Material Properties, CARDEROCKDIV-USSM-65-96 Technical Report, the Carderock Division of the Naval Surface Warfare Center
of U.S. Navy, Survivability, Structures and Materials Directorate, 90.
9. Basar, N.S. and Stanley, R.F. Survey of Structural Tolerances in the U.S. Commercial
Shipbuilding Industry SSC Report-273, 1978
Selected U.S. commercial shipyards, ship owner/operators, steel mills, and foreign
classification societies were surveyed or interviewed with the purpose of documenting major
deviations and recurring structural imperfections, and determining the factors leading to these
deviations. A limited amount of actual data was collected, and the report relies on general
tolerance limits reported by the shipyards involved.
10. Boresi, A.P. and Sidebottom, O.M., 1985, Advanced Mechanics of Materials, 4th Edition,
John Wilely & Sons, New York, N.Y.
11. Bowman, M.D. and Munse, W.H. The Effect of Discontinuities on the Fatigue Behaviour of
Transverse Butt Welds in Steel Institute of Highways Research Report 304, April 1981
This report presents the results of an experimental program aimed at exploring the effects of
discontinuities on fatigue life. A number of modified S-N curves are developed and
presented for use in fatigue assessments.
A-2
5354C.FR
12. Bowman, M.D. and Munse, W.H. The Effect of Discontinuities on the Fatigue Behaviour of
Transverse Butt Welds in Steel Institute of Highways Research Report 304, April 1981.
This report presents the results of an experimental program aimed at exploring the effects of
discontinuities on fatigue life. A number of modified S-N curves are developed and
presented for use in fatigue assessments.
13. British Standards Institute (BSi) Code of Practice for Fatigue Design and Assessment of
Steel Structures BS 7608, 1993
Presents a family of S-N curves, and provides guidance on how details should be classified
for the purpose of fatigue analysis.
14. Brock, J.S., 1957, Analytical Determination of the Stresses around Square Holes with
Rounded Corners, David Taylor Model Basin Report #1149.
15. Bruchman, D.D., and Ayyub, B.M., 1997, Uncertainties in Mechanical Strength Properties
of Ship Grade Steeel, CARDEROCKDIV-U-SSM-65-96 Technical Report, The Carderock
Division of the Naval Surface Warfare Centre of U.S. Navy, Survivability, Structures and
Materials Directorate, 77.
16. Bruchman, D.D.; Kihl,P.D. and Adamchak, J.C Evaluation of the Effect of Construction
Tolerances on Vessel Strength SSC Report-411, 2000
This report describes a methodology to evaluate the effect of imperfections resulting from the
fabrication process on vessel strength. It addresses both strength and fatigue issues relevant
to deformed and misaligned structure, misalignments being based on polished standards. An
approach to determine appropriate maximum misalignment amplitudes based on fatigue
considerations is described. The approach uses finite element methods to determine stress
concentration factors associated with misaligned details.
17. Butler, J.D. and Warren, T. R. The Establishment of Shipbuilding Construction Tolerances
National Shipbuilding Recovery Program NSRP-0273, 1986
18. Det Norske Veritas Fatigue Assessment of Ship Structures DNV Classification Note 30.7,
September 1998
This note provides a comprehensive description of fatigue analysis of ship structures,
focused on the notch stress approach utilized by DNV. Details are provided on acceptable
tolerances, which are taken to be the default values for fatigue analysis.
19. Dexter, R.J. and Pilarski, P.J., 2000, Effect of Welded Crack Stiffeners on Crack Growth
Rate, SSC Report 413.
20. Dinovitzer, A. Life Expectancy Assessment of Ship Structures SSC Report 427, 2003.
This report presents comprehensive analyses of two ships, taking into account anticipated
levels of structural degradation due to coating breakdown and corrosion as part of a fatigue
performance prediction.
21. Fricke, W. Fatigue analysis of welded joints: state of development Marine Structures
Volume 16, Issue 3 , May 2003
A-3
5354C.FR
22. Glen, I.F., Dinovitzer, A., Paterson, R.B., Luznik, L., and Bayley, C., 1999, Fatigueresistant Detail Design Guide for Ship Structures, SSC Report 405.
23. Horn, G.E.; Chen, Y.K. and Chen, J. M. SafeHull FEM Fatigue Assessment of Ship
Structural Details ICCAS, Cambridge, Mass, June 1999
This paper outlines the approach to fatigue assessment embedded within the SafeHull
analysis tool, covering the procedures and their implementation. The S-N curves used in
SafeHull are identified.
24. Heller Jr., S.R., Brock, J.S., and Bary, R., 1959, The Stresses around a Rectangular Opening
with Rounded Corners in a Uniformly Loaded Plate, DTMB Report 1290.
25. Hess, P.E., Ayyub, B.M., 1996, Variability in the Geometry and Imperfections of Surface
Ship Structural Scantlings, Draft Paper, Carderock Division, Naval Surface Warfare Center,
Bethesda, MD 20084-5000.
26. Hess, P.E., Nikilaidis, E., Ayyub, B.M., and Hugues, O.F., 1994, Uncertainty in Marine
Structural Strength with Application to Compressive Failure of Longitudinally Stiffened
Panels, CARDEROCKDIV-U-SSM-65-94/07 Technical Report, the Carderock Division of
the Naval Surface Warfare Center of U.S. Navy, Survivability, Structures and Materials
Directorate, 63.
27. International Association of Classification Societies (IACS) Shipbuilding and Repair
Quality Standard IACS Rec, No. 47, 1999
Part A provides guidance on acceptable quality standards for new construction, including
certain fabrication tolerances. For certain tolerances, both standard and limiting values are
provided. It is noted that actual tolerances should be distributed uniformly about the standard
value, and highly skewed distributions are unacceptable.
28. International Institute of Welding (IIW) Recommendations Concerning Stress Determination
for Fatigue Analysis of Welded Components IIW Doc. XIII-1458-92, June 1993
This document describes how fatigue analyses using nominal stress, hot spot stress, notch
stress, notch strain, and fracture mechanics should be undertaken. It also provides guidance
on how FE analysis should be used in fatigue analyses.
29. Kendrick, A., 1998, Comparison of Design Codes and Existing Criteria, Presentation,
Short Course on Fatigue and Fracture Analysis of Ship Structures, August 18-21, Troy,
Michigan.
30. Maddox, S.J. Fatigue Strength of Welded Structures Abington Publishing, 1991 (2nd Edition)
This book provides a reasonably comprehensive guide to the performance of welded joints
under fatigue loading, and provides an outline of the effects of different types of
discontinuity on fatigue life. More attention is given to weld issues than to other forms of
misalignment.
31. Munse, W.H., Wilbur, T.W., Tellalian, M.L., Nicoll, K., and Wilson K., 1982, Fatigue
Characterization of Fabricated Ship Details for Design, Ship Structure Committee Report
SSC-318.
A-4
5354C.FR
32. Roark, R.J., and Young, W.C., 1975, Formulas for Stress and Strain, Fifth Edition,
McGraw-Hill Book Company, New York, N.Y.
33. Sielski, R.A., Wilkins, J.R., Jr., and Hults, J.A., 2001, Supplemental Commercial Design
Guidance for Fatigue, SSC Report 419.
34. Sieve, M., Kihl, D.P., and Ayyub, B.M., 2000, Fatigue Design Guidance for Surface Ships,
Carderock Division Naval Surface Warfare Center, 9500 MacArthur Boulevard West,
Bethesda, MD 20817-5700.
35. Sobey, A.J., 1962, The Estimation of Stresses around Unreinforced Holes in Infinite Elastic
Sheets, Aeronautical Research Council, Reports and Memoranda No. 3354.
36. Sobey, A.J., 1963, Stress Concentration Factors for Rounded Rectangular Holes in Infinite
Sheets, Aeronautical Research Council, Reports and Memoranda No. 3407.
37. Stambaugh, K.A. et. al. Reduction of S-N Curves for Ship Structural Details SSC Report369, September 1992
This report re-analyzes data collected for SSC-318 in order to develop practical guidance for
application of S-N curves to ship structures. A limited number of recommendations is made
for how misalignments should be accounted for in nominal stress analysis.
38. The Welding Institute The stress magnification effect of angular misalignment at butt.
39. UK Department of Energy, 1990, Offshore Installation: Guidance on Design, Construction
and Certification, Fourth Edition, London, HMSO.
40. Vroman, R.H., 1995, An Analysis into the Uncertainty of Stiffened Panel Ultimate
Strength, Trident Report 234, Annapolis, MD: U.S. Naval Academy.
41. Wang, G.S., 1996, Analyzing the Fatigue Crack Growth in Structural Details, Engineering
Fracture Mechanics, Vol. 53, No. 3, 345-361, Elesvier Science.
A-5
5354C.FR
APPENDIX B: SHIPYARD DATA

SHIPYARD 1: BUTT-WELDED STIFFENERS (RAW DATA)
SHIPYARD 1: BUTT-WELDED PLATES (RAW DATA)
SHIPYARD 2: BUTT-WELDED STIFFENERS
SHIPYARD 2: INTERCOSTAL JOINTS (RAW DATA)
B-1
5354C.FR
SHIPYARD 1: BUTT-WELDED STIFFENERS (RAW DATA)

Butt Welded Stiffeners - Webs of Same Thickness
Raw Data
n=
20
One or Two
Dist.
sided
No.
along
Welding?
Length
(O or T)
1
Note: There was no data taken for the angular misalignment of these butt welds. In addition, the
welds were perpendicular to the stiffeners, thus the Kt SCF does not directly apply.
Ship No.
6094
6094
6094
6094
6094
6094
6094
6094
6094
10
6094
11
6094
12
6094
13
6094
14
6094
15
6094
16
6094
17
6094
18
6094
19
6094
20
6094
Mean
12.00
1.55
11.43
21.43
Std. Dev
0.00
1.22
4.76
4.76
Stage in
Items Butt
construction
Type of Seam
Welded together
/ GA No.
Ship, GA16
to GA17
Ship, GA16
to GA17
Ship, GA16
to GA17
Ship, GA16
to GA17
Ship, GA16
to GA17
Ship, GA16
to GA17
Ship, GA16
to GA17
Ship, GA14
to GA15
Ship, GA14
to GA15
Ship, GA14
to GA15
Ship, GA14
to GA15
Ship, GA14
to GA15
Ship, GA14
to GA15
Ship, GA13
to GA14
Ship, GA13
to GA14
Ship, GA13
to GA14
Ship, GA13
to GA14
Ship, GA13
to GA14
Ship, GA13
to GA14
Ship, GA13
to GA14
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Long. Bottom
Stiff. (Bulb PL)
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
Transverse
Erection Joint
e1
e2
e0
(mm)
(mm)
(mm)
(mm)
12
103.5 103.4
0.09
30
12
100.7 101.2
0.53
10
15
12
100.1 101.4
1.32
10
20
12
97.41 100.8
3.42
10
20
12
100.7 101.4
0.7
20
25
12
42.23 42.78
0.55
15
20
12
40.15
42.7
2.55
10
20
12
41.9
42.59
0.69
12
41.44 42.54
1.1
12
42.31 42.55
0.24
12
45.94 42.46
3.48
12
43.38
42.4
0.98
12
43.66 42.43
1.23
12
41.8
42.46
0.66
12
40.26 42.27
2.01
12
37.62 42.15
4.53
12
41.42 42.34
0.92
12
41.17 42.25
1.08
12
40.17 42.48
2.31
12
39.95 42.46
2.51
Notes
1,2,3 are on same

seam
4,5,6 are on same

seam
7,8,9 are on same

seam
10, 11, 12 are on

same seam
13, 14 are on
same seam
B-2
5354C.FR
B-3
5354C.FR
SHIPYARD 2: BUTT-WELDED STIFFENERS
B-4
5354C.FR
SHIPYARD 2: INTERCOSTAL JOINTS (RAW DATA)

Intercostal Joints
Raw Data
N=
13
Full
Dist.
Ship
No. Penetrati Along
No.
on? (y/n) Length
Stage in
construct'n /
GA No.
10%
Assembly
16%
Assembly
22%
Assembly
27%
Assembly
32%
Assembly
38%
Assembly
43%
Assembly
48%
Assembly
33%
Assembly
10
67%
Assembly
11
33%
Assembly
12
67%
Assembly
13
50%
Assembly
Costal 1
L6x4
stiffener
L6x4
stiffener
L6x4
stiffener
L6x4
stiffener
L6x4
stiffener
L6x4
stiffener
L6x4
stiffener
L6x4
stiffener
L6x4
stiffener
L6x4
stiffener
L6x4
stiffener
L6x4
stiffener
Web
Cont.
Mem.
Girder
Girder
Girder
Girder
Girder
Girder
Girder
Girder
Girder
Girder
Girder
Girder
Girder
Mean
9.85
9.85
19.23
1.73
2.95
90.50
90.22
Std. Dev
0.55
0.55
3.96
1.67
0.17
0.35
0.32
e0
0.3t1
e0=<0.3t1
(mm)
(mm)
(mm)
(deg)
(deg)
(mm)
L6x4
Stiffener
L6x4
Stiffener
L6x4
Stiffener
L6x4
Stiffener
L6x4
Stiffener
L6x4
Stiffener
L6x4
Stiffener
L6x4
Stiffener
L6x4
Stiffener
Web
l1
t1
t2
t3
(mm)
(mm)
(mm)
(cm)
10
10
22
149.8
149.6
91
90
2495
10
10
22
224.7
224.45
2.5
91
90
2495
10
10
22
299.55
299.8
2.5
91
90
2495
10
10
22
374.45
374.45
91
90
2495
10
10
22
449.45
449.45
91
90
2495
10
10
22
524.4
524.55
1.5
91
91
2495
10
10
22
600.25
600.8
5.5
91
91
2495
10
10
22
675.05
675.4
3.5
91
91
2495
10
10
16
102.3
102.2
91
90
2500
10
10
16
98.1
98
90
91
2500
10
10
16
97.6
97.9
91
90
2500
10
10
16
102.3
102.3
90
90
2500
10
58.1
58.1
2.4
90
90
2050
Costal 2
L6x4
Stiffener
L6x4
Stiffener
L6x4
Stiffener
Costal 1 Costal 2 to
Datum
to Datum
(cm)
B-5
SHIP STRUCTURE COMMITTEE PARTNERS AND LIAISON MEMBERS

PARTNERS
The Society of Naval Architects and Marine
Engineers
The Gulf Coast Region Maritime

Technology Center
Mr. Bruce S. Rosenblatt

President,
Society of Naval Architects and Marine Engineers
Dr. John Crisp

Executive Director,
Gulf Coast Maritime Technology Center
Dr. John Daidola

Chairman,
SNAME Technical & Research Steering
Committee
Dr. Bill Vorus

Site Director,
Gulf Coast Maritime Technology Center
LIAISON MEMBERS
American Iron and Steel Institute
American Society for Testing & Materials
American Society of Naval Engineers
American Welding Society
Bethlehem Steel Corporation
Canada Center for Minerals & Energy Technology
Colorado School of Mines
Edison Welding Institute
International Maritime Organization
Intl Ship and Offshore Structure Congress
INTERTANKO
Massachusetts Institute of Technology
Memorial University of Newfoundland
National Cargo Bureau
Office of Naval Research
Oil Companies International Maritime Forum
Tanker Structure Cooperative Forum
Technical University of Nova Scotia
United States Coast Guard Academy
United States Merchant Marine Academy
United States Naval Academy
University of British Columbia
University of California Berkeley
University of Houston - Composites Eng & Appl.
University of Maryland
University of Michigan
University of Waterloo
Virginia Polytechnic and State Institute
Webb Institute
Welding Research Council
Worchester Polytechnic Institute
World Maritime Consulting, INC
Mr. Alexander Wilson

Captain Charles Piersall (Ret.)
Captain Dennis K. Kruse (USN Ret.)
Mr. Richard Frank
Dr. Harold Reemsnyder
Dr. William R. Tyson
Dr. Stephen Liu
Mr. Dave Edmonds
Mr. Tom Allen
Dr. Alaa Mansour
Mr. Dragos Rauta
Mr. Dave Burke / Captain Chip McCord
Dr. M. R. Haddara
Captain Jim McNamara
Dr. Yapa Rajapaksie
Mr. Phillip Murphy
Mr. Rong Huang
Dr. C. Hsiung
Commander Kurt Colella
Dr. C. B. Kim
Dr. Ramswar Bhattacharyya
Dr. S. Calisal
Dr. Robert Bea
Dr. Jerry Williams
Dr. Bilal Ayyub
Dr. Michael Bernitsas
Dr. J. Roorda
Dr. Alan Brown
Dr. Kirsi Tikka
Dr. Martin Prager
Dr. Nick Dembsey
VADM Gene Henn, USCG Ret.
RECENT SHIP STRUCTURE COMMITTEE PUBLICATIONS

Ship Structure Committee Publications on the Web - All reports from SSC 392 and forward are
available to be downloaded from the Ship Structure Committee Web Site at URL:
http://www.shipstructure.org
SSC 391 and below are available on the SSC CD-ROM Library. Visit the National Technical
Information Service (NTIS) Web Site for ordering information at URL:
http://www.ntis.gov/fcpc/cpn7833.htm
SSC Report Number
Report Bibliography
SSC 435
Predicting Stable Fatigue Crack Propagation in Stiffened Panels R.J.

Dexter, H.N. Mahmoud 2004
SSC 434
Predicting Motion and Structural Loads in Stranded Ships Phase 1

A.J. Brown, M. Simbulan, J. McQuillan, M. Gutierrez 2004
SSC 433
Interactive Buckling Testing of Stiffened Steel Plate Panels Q. Chen,

R.S. Hanson, G.Y. Grondin 2004
SSC 432
Adaptation of Commercial Structural Criteria to Military Needs R.Vara,

C.M. Potter, R.A. Sielski, J.P. Sikora, L.R. Hill, J.C. Adamchak, D.P.
Kihl, J. Hebert, R.I. Basu, L. Ferreiro, J. Watts, P.D. Herrington 2003
SSC 431
Retention of Weld Metal Properties and Prevention of Hydrogen

Cracking R.J. Wong 2003
SSC 430
Fracture Toughness of a Ship Structure A.Dinovitzer, N. Pussegoda,

2003
SSC 429
Rapid Stress Intensity Factor Solution Estimation for Ship Structures

Applications L. Blair Carroll, S. Tiku, A.S. Dinovitzer 2003
SSC 428
In-Service Non-Destructive Evaluation of Fatigue and Fracture

Properties for Ship Structure S. Tiku 2003
SSC 427
Life Expectancy Assessment of Ship Structures A. Dinovitzer 2003
SSC 426
Post Yield Stability of Framing J. DesRochers, C. Pothier, E. Crocker

2003
SSC 425
Fatigue Strength and Adequacy of Weld Repairs R.J. Dexter, R.J.

Fitzpatrick, D.L. St. Peters 2003

SSC-436 The Effect of Fabricaton Tolerances On Fatigue Life of Welded Joints

Uploaded by

Copyright:

Available Formats

SSC-436 The Effect of Fabricaton Tolerances On Fatigue Life of Welded Joints

Uploaded by

Document Information

Original Description:

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

SSC-436 The Effect of Fabricaton Tolerances On Fatigue Life of Welded Joints

Uploaded by

Copyright:

Available Formats

NTIS # PB2005-101516

This document has been approved

SHIP STRUCTURE COMMITTEE

SHIP STRUCTURE COMMITTEE

Dr. Jack Spencer

Mr. Joseph Byrne

Mr. Thomas Connors

Dr. Neil Pegg

CONTRACTING OFFICER TECHNICAL REP.

SHIP STRUCTURE SUB-COMMITTEE

DEFENCE RESEARCH & DEVELOPMENT ATLANTIC

MILITARY SEALIFT COMMAND

NAVAL SEA SYSTEMS COMMAND

UNITED STATES COAST GUARD

CANADIAN COAST GUARD

2. Government Accession No.

Technical Report Documentation Page

4. Title and Subtitle

Effect of Fabrication Tolerances on Fatigue Life of Welded

8. Performing Organization Report No.

9. Performing Organization Name and Address

10. Work Unit No. (TRAIS)

BMT Fleet Technology Ltd.

11. Contract or Grant No.

12. Sponsoring Agency Name and Address

13. Type of Report and Period Covered

Ship Structure Committee

6. Performing Organization Code

14. Sponsoring Agency Code

15. Supplementary Notes

17. Key Words

18. Distribution Statement

Distribution is available to the public through:

20. Security Classif. (of this page)

21. No. of Pages

mega Newton MNm3/2

BMT FLEET TECHNOLOGY LIMITED

The Effect of Fabrication Tolerances on Fatigue Life of Welded Joints

BMT FLEET TECHNOLOGY LIMITED

OBJECTIVES AND APPROACH.......................................................................................... 2

TECHNICAL BACKGROUND ............................................................................................. 4

DATA COLLECTION .......................................................................................................... 24

STATISTICAL ANALYSIS OF SURVEY DATA .............................................................. 36

The Effect of Fabrication Tolerances on Fatigue Life of Welded Joints

BMT FLEET TECHNOLOGY LIMITED

FATIGUE LIFE CASE STUDY ........................................................................................... 68

APPENDIX A: BIBLIOGRAPHY ............................................................................................ A-1

The Effect of Fabrication Tolerances on Fatigue Life of Welded Joints

BMT FLEET TECHNOLOGY LIMITED

The Effect of Fabrication Tolerances on Fatigue Life of Welded Joints

BMT FLEET TECHNOLOGY LIMITED

The Effect of Fabrication Tolerances on Fatigue Life of Welded Joints

BMT FLEET TECHNOLOGY LIMITED

The Effect of Fabrication Tolerances on Fatigue Life of Welded Joints

BMT FLEET TECHNOLOGY LIMITED

LIST OF ABBREVIATIONS AND SYMBOLS

Two Standard Deviations

The Effect of Fabrication Tolerances on Fatigue Life of Welded Joints