Вы находитесь на странице: 1из 122

TECHNICAL REPORT

FPSO FATIGUE CAPACITY JIP

FATIGUE DESIGN RECOMMENDATIONS


FOR FPSOS

P FPS
I
O
J

i ty
Fa t

ig
u e Ca p a c

REPORT NO. 2003-0582


REVISION NO. 02

DET NORSKE VERITAS/BUREAU VERITAS


DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Table of Content Page

1 INTRODUCTION ....................................................................................................... 4
1.1 Scope of document 4
1.2 Criticality of fatigue cracks in FPSOs 4
1.3 The fatigue process and significant factors influencing fatigue in FPSOs 5
1.4 Significant hot spot areas in FPSO's 5
1.5 Loads to be considered for fatigue analysis 6

2 DEFINITIONS............................................................................................................. 7
2.1 Definition of stresses 7
2.1.1 Nominal stress 7
2.1.2 Structural stress or hot spot stress 8
2.1.3 Definition of stress concentration factor 10
2.1.4 Stress for analysis of fillet welds and partial penetration K-welds 11
2.1.5 Stress for analysis of base material 13
2.1.6 Effect of mean stress 13
2.2 The Palmgren-Miner fatigue damage accumulation rule 13
2.3 Definition of Design Fatigue Factors 14
2.3.1 Design Fatigue Factor 14
2.3.2 Implied probability of fatigue cracks 14

3 OVERVIEW OF FATIGUE ANALYSIS METHODS FOR FPSOS ....................... 16


3.1 Introduction 16
3.2 Full Spectral Fatigue Analysis (SFA) 16
3.3 Other Methods of Analysis 17
3.3.1 Deterministic (One Wave) FE Analysis 17
3.3.2 “Component Stochastic Analysis” 18
3.3.3 Simplified (rule based) analysis 19
3.4 Conditions of analyses 19
3.4.1 Draught (loading condition) 19
3.4.2 Wave heading 20

4 HYDRODYNAMIC ANALYSIS.............................................................................. 22
4.1 Introduction 22
4.2 Still water Model - Mass distribution 22
4.2.1 Still-water model 22
4.2.2 Mass Model 22
4.3 Hydrodynamic Model 23
4.4 Hydrodynamic Analysis 23
4.5 Intermittent wetting 24

Page i
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

5 STRUCTURAL ANALYSIS..................................................................................... 25
5.1 Modelling 25
5.1.1 Strategy 25
5.1.2 Full Ship Model 27
5.1.2.1 Model 27
5.1.2.2 Boundary conditions 27
5.1.3 FE Model of cargo holds (3 holds model) 27
5.1.3.1 Three holds model 27
5.1.3.2 Boundary conditions and end loads 28
5.1.4 Sub-modelling 29
5.1.5 Stiffener Model 29
5.1.5.1 Explicit modelling, using plate elements of appropriate size. 29
5.1.5.2 Modelling by beam elements 30
5.1.5.3 Analytical approach 31
5.1.6 Stress Concentration Model 31
5.2 Loading 31
5.2.1 Overview of load transfer 31
5.2.2 Wave pressure loads 33
5.2.3 Intermittent wetting pressure 33
5.2.4 Internal fluid pressure loads 34
5.2.5 Inertia loading 34
5.2.6 Dynamic equilibrium 34
5.2.7 Still Water loads 34
5.2.8 Wave pressure loads; component method and deterministic analysis 35
5.2.8.1 Deterministic Analysis 35
5.2.8.2 Component method 35
5.3 Analysis 35
5.3.1 Verification 35
5.3.2 Results 36
5.4 Long term distribution of stresses and calculation of fatigue damage 37
5.4.1 Metocean conditions 37
5.4.2 Response spectrum 37
5.4.3 Spectral moments 38
5.4.4 "Short-Term" stress distribution 39
5.4.5 "Long-Term" stress distribution (Spectral analysis) 39
5.4.6 "Long-Term" stress distribution (Deterministic Fatigue analysis) 40
5.4.7 Calculation of fatigue damage 40
5.4.7.1 Numerical integration 40
5.4.7.2 Closed Form Approach 41
5.4.7.3 Summation and fatigue life 41
5.4.7.4 Evaluation 41
5.4.8 Description of waves as Swell and Wind Seas 41
5.4.9 Structures supporting loads from Mooring system or risers. 42
5.4.10 Analysis of fatigue due to loading/unloading process 42

6 FINITE ELEMENT ANALYSIS OF DETAILS....................................................... 44


Page ii
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

6.1 The hot spot stress methodology 44


6.1.1 The hot spot stress concept 44
6.1.2 Methods for hot spot stress derivation 44
6.1.3 Link between hot spot stress and S-N curve 47
6.2 FE modelling and hot spot S-N curve 48
6.2.1 General 48
6.2.2 Mesh size using 4-Node Shell Elements 52
6.2.3 Mesh size using 8-Node Shell Elements 52
6.2.4 Mesh size using Three-Dimensional Solid Elements 52
6.3 Derivation of effective hot spot stress from FE analysis 53
6.4 Thickness effect 54
6.5 Expected accuracy of finite element analysis for hot spot stress analysis 54
6.6 Validation of computer programs and finite elements for fatigue analysis 54
6.6.1 General 54
6.6.2 Details to be analysed for validation 54
6.6.3 Target hot spot stress values 54
6.7 FE analysis of hopper corner details 55
6.7.1 Welded structure 55
6.7.2 Effect of tolerances for hopper corners 56
6.7.3 Bent hopper corner 57

7 DESIGN OF FILLET WELDED CONNECTIONS ................................................. 58


7.1 Recommended design criteria 58
7.2 Effect of residual stress and mean stress on fatigue life 61
7.3 Design of fillet welded penetrations in plated structures 61
7.3.1 Critical hot spot areas 61
7.3.2 Principal stress direction relative to weld toe 62
7.3.3 Stress concentration factors for holes with reinforcement 63
7.3.4 Procedure for fatigue design of penetrations with welded reinforcement 64
7.3.5 Example of design analysis 67

8 S-N CURVES ............................................................................................................ 69


8.1 Definition of S-N curves and failure criterion 69
8.2 S-N curves for different details and environments 70

9 REFERENCES........................................................................................................... 73
Appendix A Effect of tolerances
Appendix B Mean stress effects
Appendix C Design of cut-outs in plated structures
Appendix D Intermitting wetting pressure and internal fluid pressure
Appendix E Damage (Closed form)

Page iii
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

1 INTRODUCTION

1.1 Scope of document


The scope of this document is to provide recommendations on fatigue analysis of FPSO’s
with the purpose of achieving a reliable long term operation of structures that are permanently
installed on a field.
This document is the end product of the FPSO – Fatigue Capacity Joint Industry Project that
was performed with a Phase I in 1998 – 2000 and a Phase II in 2001 – 2003.
Reference is also made to the Phase II Summary Report, ref./1/.
The main findings from the project that are considered to be of significant information for a
reliable fatigue design of FPSOs are included in these Recommendations.
A fatigue design of an FPSO requires a number of different considerations. It is realised that
only selected areas were investigated in this JIP. Therefore it is found appropriate to make
references to other standards and recommendations from the Classification Companies such
as refs. /2/- /6/ and /25/ for items in the fatigue assessment procedure that was not included in
the JIP.
Where detailed recommendations are presented herein in this document this information is
considered to be governing over the others.
It is assumed that this document will be used by designers of FPSOs. It should be noted that
Design Fatigue Factors need to be specified before the design is started. (It is assumed that
values of the Design Fatigue Factors will be dependent on Company philosophy with respect
to risk for shut down and repair).

1.2 Criticality of fatigue cracks in FPSOs


Fatigue and corrosion are recognised as the predominant factors which contribute to the
structural failures observed in ships in service as well as FPSOs that are either converted
tankers or new-buildings with a typical tanker structure.
However, FPSOs differ significantly from sailing ships for their service conditions:
• Site specific as opposed to ocean trading
• Continuous loading and offloading operations at sea
• Topsides facilities in continuous operations
• Inspection, repair and maintenance at sea, with generally no dry-docking during the
intended service life.
Repair during service life of FPSO is difficult and may result in economic loss due to
interruption of the field production. Consequently, FPSOs must be designed with respect to
the fatigue limit state in a reliable manner in order to be safe during their service life.
Most of the existing FPSOs have not been in service for a sufficient time that their
appropriateness with respect to fatigue design can be verified. Therefore, the experience from
tanker structures with similar details is of importance when hot spot areas to be analysed are
considered. Fatigue damages are known to occur frequently for some ship types and
Page 4
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

categories of hull structure elements. The fatigue life, is, in particular, related to the
magnitude of the dynamic stress level, the number of load cycles, the corrosiveness of the
environment and the magnitude of stress concentration factors for the structural details.

1.3 The fatigue process and significant factors influencing fatigue in


FPSOs
Fatigue may be defined as a process of cycle by cycle accumulation of damage in a structure
subjected to fluctuating stresses, going through several stages from the initial “crack-free”
state to a “failure” state. For welded structures, the fatigue process includes three main phases
:
• initiation of crack at critical location,
• propagation or crack growth, and
• final failure by ductile tearing or brittle fracture, when crack has reached a critical size.
There are two different types of fatigue:
• low-cyclic fatigue occurring for a low number of cycles, less than 5x103, in the range of
plastic deformations,
• high-cyclic fatigue occurring for a large number of cycles in the range of elastic
deformations.
Fatigue observed on ship building structures is generally of the second type and is mainly
governed by the crack growth. However, a combination of the two fatigue types may occur at
some hot spots due to low cycle fatigue during loading and unloading and high cycle wave
loading that occurs in each loading condition.
Many factors affect the fatigue behaviour of FPSO`s:
1. sea conditions and resulting wave induced loads,
2. loading conditions of the vessel,
3. geometry of the members or configuration of the weld details producing stress
concentrations,
4. materials and welding procedures,
5. workmanship,
6. environmental factors e. g. sea water environment.
Loading and geometry (1, 2, 3) are addressed by fatigue analysis. The others are implicitly
embedded in fatigue resistance criteria (S-N curves). Therefore, it is assumed that the welding
procedures and workmanship are carried out according to the rule standards and state-of-the-
art in such a way that, with the exception of particular designs, their influence on the fatigue
life need not be explicitly considered.

1.4 Significant hot spot areas in FPSO's


A major fraction of the total number of fatigue damages on ship structures occurs in panel
stiffeners on the ship side and bottom and on the tank boundaries of ballast- and cargo tanks.
However, the calculated fatigue life depends on the type of stiffeners used, and the detail
Page 5
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

design of the connection to supporting girder webs and bulkheads. In general asymmetrical
profiles will have a reduced fatigue life compared to symmetrical profiles unless the reduced
efficiency of the asymmetrical profile is compensated for by an improved design for the
attachment to transverse girder webs and bulkhead structures.
Structural elements in the cargo area being of possible interest for fatigue evaluation are listed
in Table 1-1.

Table 1-1 Example of hot spot areas to be considered


Structural Structural detail Load type
member
Side-, bottom- Butt joints, doubling plates, module Hull girder bending, stiffener
and deck plating support stools, deck openings and lateral pressure load and
and longitudinals attachment to transverse webs, transverse support deformation from
bulkheads and intermediate longitudinal topside inertia loads
girders
Transverse girder Bracket toes, girder flange butt joints, Sea pressure load combined
and stringer curved girder flanges, knuckle of inner with cargo or ballast
structures bottom and sloped hopper side and other differential pressure load
panel knuckles including intersection
with transverse girder webs. Single lug
slots for panel stiffeners, access and
lightening holes
Longitudinal Bracket terminations of abutting Hull girder bending, and
girders of deck transverse members (girders, stiffeners) bending / deformation of
and bottom longitudinal girder and
structure considered abutting member
FPSO specific Moonpool and turret Hull girder bending
details Mooring system and hull interface Mooring forces and
structure. FPSO accelerations
Riser porches and hull interface structure. Riser loads
Turret/ hull interface structure. Crane loads
Topside supports and hull interface Wind loads
structure. Topside loads
Flare tower/ hull interface structure.
Crane pedestals
Caisson
Helideck

1.5 Loads to be considered for fatigue analysis


All dynamic loads need to be considered in fatigue analysis of FPSOs. As shown in Table 1-1
there are a number of different types of dynamic loads on FPSOs. These are:
• Global hull girder bending loads due to wave action
Page 6
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

• Deformation loads on girders and stiffeners as a result of global deformation


• Pressure loads from wave actions and cargo and ballast water and differential pressure
from water outside/inside and cargo
• Anchoring forces
• Risers loads
• Wind loads
• Operational loads due to filling and emptying of tanks
• Operation of cranes
• Rotating machinery
In the following, mainly wave induced loads from wind sea wave and swell are considered to
determine the fatigue strength of ship structures. Swell response can be calculated similarly to
that of wave frequency response. See also section 5.4.8.
This includes both Global loads (Hull girder loads) and local loads such as sea pressure,
internal fluid pressure and other inertia loads (e. g. from topside).
Loads induced by mooring, risers, etc may induce fatigue damage in specific areas of the
structure (turret, mooring and risers porches). Description of these loads and related analysis
is not covered in this document (loading side). For guidance see ref. /4/.
Moreover, dynamic loads due to natural vibration or impact loads which also may induce
fatigue damage are not taken into account in the analyses approaches that are described in this
document. These items are not considered significant with respect to fatigue provided that the
FPSO is designed similar to that of a tanker.
Static loads and residual stresses from welding, considered as mean stresses, are normally not
explicitly considered for FPSO structures. Reference is also made to section 2.1.6.

2 DEFINITIONS

2.1 Definition of stresses


2.1.1 Nominal stress
Nominal stress is the stress calculated in the sectional area under consideration, disregarding
the local stress raising effects of the structural detail, i. e. the stress increase due to the
structural joint configuration (structural effect) as well as that of the weld toe (notch effects).
Other macro-geometric stress raising effects are, however, to be included, such as reduced
effective plate breadth or large cut-outs in the vicinity of the detail considered. (It is important
that the definition of nominal stress corresponds with the nominal stress S-N data used for
fatigue analysis).
In FPSO and ship structures, nominal stresses are usually subdivided into:
• global stresses acting in larger structural components such as the whole hull girder as well
as primary structural members, and
• additional local bending stresses acting in secondary structural members such as stiffeners
and plate fields under local loads, see also ref. /3/ and /25/.

Page 7
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Nominal stresses are usually associated with well-defined stress components e. g., directional
stresses or shear stresses.
Nominal stresses are derived from beam element models or from global coarse mesh FE
models. Stress concentrations resulting from the gross shape of the structure, e.g. shear lag
effects, are included in the nominal stresses derived from coarse mesh FE models (provided
that the model contains several elements between the longitudinal bulkheads and the ship
sides). Reference is also made to Chapter 5.
When nominal stress is used for fatigue analysis it has to be combined with appropriate stress
concentration factors, ref. Section 2.1.3, before the stress is used together with the S-N curves
defined in Chapter 8.
A nominal stress approach is considered to be efficient for fatigue analysis of butt welds with
transition in thickness and for cruciform joints. This is explained more in detail in Appendix
A. Reference is also made to Chapter 8 for definition of S-N curves to be used for these
connections.

2.1.2 Structural stress or hot spot stress


Effects from all stress raisers that are not implicitly included in fatigue test data and the
corresponding S-N curves must be taken into account in the stress analysis. In order to
correctly determine the stresses to be used in fatigue analyses, it is important to note the
definition of the different stress categories (see also Figure 2-1).
The hot spot stress at a weld toe is defined as the geometric stress that includes stress rising
effects due to structural discontinuities and presence of attachments, but excluding the
localised stress due to the presence of the weld itself. (The local notch stress increase is
normally excluded from the analysis by calculating the stress at a surface location away from
the hot spot). Effects caused by fabrication imperfections such as misalignment of structural
parts, are however normally not included in FEM analyses, and must be separately taken into
account. (Reference is made to last paragraph in this section for a practical design advice on
the issue of tolerances to be considered). The greatest value of the extrapolation to the weld
toe of the geometric stress distribution immediately outside the region affected by the
geometry of the weld is commonly denoted the hot spot stress.
The notch stress is the maximum stress at the weld toe and includes the geometric stress and
the stress due to the presence of the weld. The local notch effect is accounted for in the hot
spot S-N curve that is being used for welded connections in this Fatigue Design
Recommendation, and the notch stress will not be further referred to herein for purpose of
fatigue analysis.
The relevant hot spot stress at a weld toe is the maximum range of normal stress range
adjacent to the potential crack location with stress concentrations being taken into account. In
case of complex stress state, the principal stress in the plate surface within ± 60° of the normal
to the weld toe should be used as the relevant hot spot stress. This principal stress should be
calculated at the hot spot from the extrapolated values of the stress components.

Page 8
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Notes:
1) In many case, the principal stress is approximately the same as the stress perpendicular
to the weld.
2) In some particular situations, the direction of principal stresses may not be stable over
a wave cycle; then the range of normal stress is to be evaluated for a set of directions
within 60° of the normal to the weld toe, to find the maximum.
3) When stresses are outside of this range, the detail is treated as a longitudinally loaded
weld, see section 8.2.

Stress Notch stress

Extrapolation of geometric stress to


derive the hot spot stress

Hot spot stress

σt/2 σ3t/2

t/2 3t/2
Distance from hot spot

Region effected by
the notch stress

Figure 2-1 Definition of stress categories

Page 9
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

σ nominal σ nominal

σ nominal * K
A A

σ nominal
60°
60°

σ nominal A σ nominal A
A-A

Figure 2-2 Explanation of local stresses

2.1.3 Definition of stress concentration factor


Stress concentration factors or K-factors may be determined based on fine mesh finite element
analyses as described in Chapter 6 . Alternatively, K-factors may be obtained from selection
of factors for typical details in ships.
The fatigue life of a weld toe detail is governed by the hot spot stress range. For welded
components other than butt welded connections the hot spot stress is obtained by
multiplication of the nominal stress by geometric K-factors, Kg. The K-factors in this
document are thus defined as
σ hot spot (2.1)
Kg =
σ nominal

The relation between the hot spot stress range to be used together with the hot spot stress S-N-
curve and the nominal stress range is thus
∆σ hot spot = K g ∆σ nominal (2.2)

Page 10
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

All stress risers (excluding the localised stress concentration due to the weld profile itself)
have to be considered when evaluating the hot spot stress. This can be done by multiplication
of K-factors arising from different causes.
As an example, for the weld shown in Figure 2-2 a), the relevant hot spot stress for fatigue
design would be the nominal tensile stress times the stress concentration due to the attachment
(σhot spot = Kg attachment σ).
For the weld shown in Figure 2-2 b), the stress concentration factor for the local geometry
must in addition be accounted for, resulting in a relevant hot spot stress equal to σhot spot = Kg
attachment Kg holeσ, where Kg hole is the stress concentration factor due to the hole.

Besides, as the hot spot stress is usually calculated on the basis of an idealised, perfectly
aligned structure, any possible misalignment needs to be taken into account separately. This is
normally necessary for butt joints, cruciform joints and transverse fillet welds depending on
loading, see also Appendix A.
The resulting K-factor to be used for calculation of hot spot stress is derived as

K = K g K te K tα K n (2.3)

where
Kg = stress concentration factor due to the gross geometry of the detail considered
Kte = additional stress concentration factor due to production tolerance (eccentricity) (normally
used for plate butt weld connections and cruciform joints only. See also section below)
Ktα = additional stress concentration factor due to angular mismatch (normally used for plate
connections only).
Kn = additional stress concentration factor for asymmetrical stiffeners on laterally loaded panels,
applicable when the nominal stress is derived from simple beam analyses. The value of this
factor may be significant and should be included in the analysis, ref. /3/ and /25/.

For fatigue design of butt welds and cruciform joints it may be most efficient to use the
concept of nominal stress, see section 2.1.1. Then stress concentration factors not accounted
for in the S-N data need to be included in the stress range to be entered into the nominal stress
S-N curve. For butt welds and cruciform joints the following equation can be used for
derivation of this stress range:
∆σ nomS − Ncurve = K te K tα ∆σ nominal (2.4)

where Kte and Ktα are stress concentration factors due to eccentricity and angular mismatch
exceeding the tolerances included in the S-N curve. Reference is also made to Appendix A.

2.1.4 Stress for analysis of fillet welds and partial penetration K-welds
It should be noted that cracks may also start from the root of fillet welds or partial penetration
K-welds (if used).

Page 11
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

For potential cracks in the weld throat (growing from the weld root) of load-carrying fillet-
welded joints, the relevant stress range is the maximum range of nominal shear stress in the
weld metal. Reference is also made to Chapter 7.
The relevant stress range for potential cracks in the weld throat of load-carrying fillet-welded
joints and partial penetration welded joints may be calculated as:

∆σ w = ∆σ ⊥2 + ∆τ ⊥2 + 0.2 ∆τ 2// (2.5)

See Figure 2-3 for explanation of stress components. (σw is denoted engineering shear in
some countries).
The stresses are mean values over the throat thickness (nominal values). It is difficult to
derive this stress from a finite element analysis even with the fillet weld included in the
model. Therefore an alternative approach is recommended for derivation of the engineering
stress in the fillet weld, see section 7.3.
The total stress fluctuation (i.e. maximum compression and maximum tension) should be
considered to be transmitted through the welds for fatigue assessments. See also section 7.

σ τ
τ
Throat
section τ p

σnp

Figure 2-3 Explanation of stresses on the throat section of a fillet weld

Page 12
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

2.1.5 Stress for analysis of base material


The largest range for the cyclic stress (i. e. principal stress, see also 2.1.2) is to be used
together with the actual S-N curve for the base material. This stress is to include geometric
stress concentration factors due to cut-outs etc.
2.1.6 Effect of mean stress
It is recommended that the mean stress effect is neglected for fatigue assessment of welded
connections for new built FPSOs.
The reason for this is that it is difficult to fully utilise the positive effect of mean stress at
welded connections because of uncertainty with respect to tensile residual stresses after
construction. The mean stress effect is further discussed in Appendix B. It is likely the effect
can be more easily accepted at the roots of fillet welds than for weld toes. One might consider
taking the effect into account in planning in-service inspection in order to direct the inspection
to details where fatigue cracking most likely will occur.
For fatigue analysis of regions in the base material not significantly influenced by residual
stresses due to welding, the stress range may be reduced depending on whether mean cycling
stress is tension or compression. This may be performed by reducing the stress cycle that is
compressive by a factor 0.6. Then the resulting stress range is obtained by adding this reduced
stress to that of the tensile part of the cyclic stress.

2.2 The Palmgren-Miner fatigue damage accumulation rule


The fatigue life may be calculated based on the S-N fatigue approach under the assumption of
linear cumulative damage (Palmgren-Miner rule). For definition of S-N curves see section
8.1.
When the long-term stress range distribution is expressed by a stress histogram, consisting of
a convenient number of constant amplitude stress range blocks ∆σi each with a number of
stress repetitions ni the fatigue criterion reads:

nk
1 k
m (2.6)
D = ∑ i = ∑ ni ⋅ (∆σ i ) ≤ η
i =1 N i a i =1
where
D = accumulated fatigue damage
a = intercept of the design S-N curve with the log N axis
m = negative inverse slope of the S-N curve
k = number of stress blocks
ni = number of stress cycles in stress block i
Ni = number of cycles to failure at constant stress range ∆σi

η = usage factor

Page 13
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

= 1 / Design Fatigue Factor (see Section 2.3).


The fatigue damage is thus calculated without taking the stress cycles sequence into account.

2.3 Definition of Design Fatigue Factors


2.3.1 Design Fatigue Factor
The Design Fatigue Factor (DFF) is a Safety Factor on the calculated lifetime of structural
details, with respect to the intended operating lifetime of the vessel.
In general the safety factor on fatigue lives in the rules for trading ships is equal 1.0. An
expected long-term stress range is aimed for in the design response analysis.
For offshore structures, most/several standards, Rules of Classification Societies, or Company
specifications require a minimum DFF of 2, or higher, taking into account the criticality of
potential fatigue cracks in a FPSO, as discussed in section 1.3 above.
The implication of a selected DFF on the probability of fatigue cracks is discussed in Section
2.3.2 below.
In any case, Design Fatigue Factors beyond regulatory requirements should be defined in
Company specifications.
2.3.2 Implied probability of fatigue cracks
A safety factor on life equal 1.0 implies a probability of a fatigue crack during service life
equal 2.3 % due to the safety in the S-N curve if uncertainties in other parameters are
neglected. However, this probability number is considered to be a nominal value. In ships
there are a large number of similar details and additional uncertainties in the load and
response that also contributes to the probability of a fatigue failure. A conservative
assessment of one of the assumptions made for the fatigue assessment, may on the other side,
reduce this nominal probability of failure. In general increasing the safety factor may be an
efficient way to reduce the probability of fatigue cracking during service life.
Reference is made to Figure 2-4 where uncertainties on the most important parameters in the
fatigue design procedure are accounted for when probability of fatigue failure is calculated.
This figure is derived by probabilistic analysis where the uncertainty in loading is included in
addition to uncertainties in S-N data and the Palmgren-Miner damage accumulation rule
under variable amplitude loading. Accumulated probability of a fatigue failure during 20
years service life is shown as function of Design Fatigue Factor (DFF) used for design. (The
DFF is defined as a safety factor on fatigue life). Also the annual probability of a fatigue
failure the last year in service is shown in the same figure.
The accumulated probability of a fatigue crack as function of time in service is shown in
Figure 2-5. Here only uncertainty in S-N data is accounted for in the analysis. The figure
shows the results for DFF = 1.0. The results for other values of DFFs can simply be obtained
by multiplication of the time scale on the abscissa axis by the actual DFF that is considered
used.
Calculated fatigue life forms the basis for assessment of probability of fatigue cracking during
service life. Thus, it implicitly forms the basis for requirement to in-service inspection. For
details showing a short fatigue life at an early design stage, it is recommended that the
Page 14
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

considered details are evaluated in terms of improvement of local geometry to reduce its
stress concentration. At an early design stage it is considered more cost efficient to prepare for
minor geometric modifications than to rely on methods for fatigue improvement under
fabrication and construction, such as grinding and hammer peening.

Design Fatigue Factor vs. PoF; Design life=20 yr


1.0E-01
Accumulatet Failure prob.
Annual PoF at the last year
Failure probability

1.0E-02

1.0E-03

1.0E-04

1.0E-05
0 1 2 3 4 5 6 7 8 9 10 11 12
Design Fatigue Factor (DFF)

Figure 2-4 Failure probability as function of design fatigue factor

1.00
0.90
0.80
Accumulated probability

0.70
0.60

0.50
0.40
0.30
0.20
0.10
0.00
0 20 40 60 80 100 120 140 160 180 200
Time in service (years)

Figure 2-5 Accumulated probability of fatigue crack as function of service life for (20
years design life and statistical scatter in S-N data only)
Page 15
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

3 OVERVIEW OF FATIGUE ANALYSIS METHODS FOR FPSOS

3.1 Introduction
A procedure for fatigue assessment of fatigue strength requires determining:
• the long term distribution of stresses resulting from the action of the cyclic loads applied
on the structure,
• the fatigue capacity of the structure, characterised by S-N curves.

A spectral analysis methodology based on linear frequency domain analysis is the reference
fatigue analysis approach for FPSOs.
Linear spectral analysis can be used for analysis of global ship response (due to global wave
loads). Also most of the local structural responses (due to local pressure) can be assumed
linear for the range of wave heights that are giving the largest contribution to fatigue damage.
In some areas, the local response is not linear: this is the case of side shell longitudinal
stiffener close to water line, due to intermittent wetting effect. This can be addressed (see
section 4.5) through a suitable linearisation of the pressure history or by post-processing of
the result of linear analysis.
Effects from other load effects such as slowly varying response, or the Loading-Unloading
cycles may be important in some environments and should be included if it influence the
fatigue life. Reference is made to section 5.4.8 and 5.4.10.
Other fatigue analysis may also be used such as the deterministic (One Wave) approach or the
“Component stochastic analysis”. A brief overview of these approaches is given below in
section 3.3.
Simplified fatigue analysis methods may be combined with advanced analysis. Different
methods may be used at different stages in the design loop, or for verification of a fatigue
design. Reference is e. g. made to ref. /5/ and /6/.

3.2 Full Spectral Fatigue Analysis (SFA)


A Spectral Fatigue Analysis (SFA) is based on four interlinked activities (see Figure 3-1):
• Wave Loads Analysis
Loads are determined by a hydrodynamic analysis in frequency domain. This includes loads
directly applied by the wave (sea pressure) and loads resulting from the vessel motions
induced by the wave (internal fluid pressure and inertia forces).
• Structural Analysis
Loads are applied to the structural model. Stresses (RAO) are calculated at the region of
interest.
• Statistics

Page 16
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

The response spectrum of stresses at each location is obtained by the technique of spectral
analysis. The "short-term" distribution of stresses is obtained by statistics over short-term sea
states and condition of vessel. Long-term distribution of stresses is then obtained by
summation, over the wave scatter diagram at vessel site of the "short term" distributions.
• Fatigue damage
The fatigue damage at given locations in the structure is determined from the long-term
distribution of stress ranges by the Miner sum and appropriate S-N curve. The Damage can
also be calculated by summation of fatigue damage for each short-term response.
Several models linked to these activities have to be prepared: Static (still-water) model
corresponding to the loading condition being analysed (draught, weight distribution),
Hydrodynamic Model, Structural Model. The objective is to apply to the structural model the
loads obtained from the hydrodynamic model. Therefore, special attention has to be paid to
the consistency between these models and the interfaces between hydrodynamic and structural
model for the transfer of loads.
Full SFA approach (see Figure 3-3) thus requires the following steps:
• Calculation of RAO’s of loads (end loads, pressure fields), in frequency domain;
• Calculation of hot spot stress RAO at location of interest (from FE model), for each wave
loading;
• Calculation of short stress distributions for each vessel condition and sea state in the wave
scatter diagram representing the wave environment at vessel site.
Then:
• Calculation of fatigue damage using Miner Rule
• summation of the "short-term" damages over all sea states and vessel conditions
Or:
• Calculation of the long term stress distributions by summation of the short term
distribution over all sea states and vessel conditions
• Calculation of fatigue damage using Miner Rule, from the long-term stress distribution.

3.3 Other Methods of Analysis


3.3.1 Deterministic (One Wave) FE Analysis
In a Deterministic (One Wave) Analysis (derived from the rules methodology for ships) (see
), only loads parameters are determined in frequency domain from a hydrodynamic analysis,
then statistically processed. The structural response and stresses are then derived from a
limited number of representative loading conditions.
Deterministic approach (see Figure 3-2) requires the following steps:
• Calculation of load parameters ( RAO’s, then long-term distribution)
• Determination for each load of the reference value corresponding to a suitable reference
probability of exceedance (typically10-2 -10-3).
• Combination of loads in a set of "crest" and "trough" loadings
• Calculation of stresses (from FE model)
• Calculation of damage obtained using the "closed form" Weibull approach with:
Page 17
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

¾ Calculated stresses that are assumed having the same reference probability as loads
¾ a Weibull shape parameter taken from the dominating load long term distribution.
These steps correspond to those of the Rule approach for trading ships except for the
calculation of the load parameters. Instead of values specified in rules, loads are taken from
the actual wave conditions of the vessel at site.
Deterministic approach may be used to get approximate evaluation of stresses and of fatigue
damage, that will be generally conservative.

LOADS Short/long term distribution


of stresses

STRESSES
(frequency domain) DAMAGE

Figure 3-1 Methodology for Fatigue Analysis of FPSO hull based on full spectral
analysis (SFA)

Long term distribution


LOADS
of loads parameters

« One Wave »
DAMAGE
STRESSES

Figure 3-2 Methodology for Fatigue Analysis of FPSO hull based on deterministic (one
wave) analysis

3.3.2 “Component Stochastic Analysis”


The “Component Stochastic Analysis” (see /3/) is, in its principle, a variant of the full spectral
analysis, in which the description of loads on the structure is simplified. Thus accuracy of this
method is very much depending on the degree of simplification.
Page 18
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

The “component” approach requires the following steps:


• Calculation of the RAO’s of load parameters , then of the long-term distribution for
selected components used to define load cases;
• Calculation of structural response (from FE model), for a set of unit load cases;
• Building of the hot spot stress RAO’s from the RAO’s of each component and the
corresponding stress transfer function;
• Calculation of stress distributions and damage as in section 3.2.

3.3.3 Simplified (rule based) analysis


In Simplified (rule based) analysis, the calculation of stresses is based on analytical methods
using elastic beam theory analysis. Three components of the stress response (hull girder
bending, stresses induced by the relative deformation of bulkheads and transverse frames, and
bending of stiffeners under the local pressure) are calculated and combined analytically. Such
type of analysis is described in /3/ and /25/. This document does not cover such approach.

3.4 Conditions of analyses


The Two principal parameters for selecting conditions of analysis are:
• vessel draught and loading condition,
• wave heading.
3.4.1 Draught (loading condition)
Two basic design cases are usually considered for an oil tanker: i.e. the full and empty
(ballast) conditions. This is not very relevant for an FPSO. Depending on the production rate,
it takes a few days or weeks to fill up all the storage tanks, and one or two days to offload to a
shuttle tanker. The vessel is seldom in the full and ballast conditions.
For a FPSO, representative loading conditions of the loading/unloading sequences are to be
selected from the loading manual or from operating data. Each loading condition is
characterised by the levels in oil and water ballast tanks and the associated FPSO draught. A
probability of occurrence is assigned to each loading condition.
The structural response is first depending on the intensity of hydrodynamic loads that is not
much depending on draught and loading conditions. Then a single draught and load condition
might be used in a first approach.
On the other hand, structural response is also depending on the foot print of load on structure
(principally wave pressure on side shell) , that varies with vessel draught :
• As guidance, two loading conditions at least have to be used if the difference in draught
between the loaded and unloaded condition does not exceed 4 metres.
• An intermediate loading condition may have to be considered if the difference in draught
is larger than 4 meters. More than three loading conditions might be necessary if the
difference in draught is larger, and to enable proper integration of the damage over
draught variations (see 5.4.7.3).

Page 19
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Note: In most cases, the cyclic internal pressure is not inducing much fatigue in comparison
with the effect of external (wave induced) pressure. This might however need an assessment
in some particular conditions.
Thus, the number of loading conditions will depend on
• the objectives of the analysis
• the effort and time which are allocated
• the criticality on the results.
Full SFA is requiring a substantial computational effort. At the other end, deterministic
analysis may provide useful tools for a more comprehensive scanning of vessel loading
conditions and for screening analysis of significant hot spot areas.
3.4.2 Wave heading
The range of headings to be considered depends on the type of vessel mooring.
For a trading ship (e.g an FPSO before conversion) or a spread moored vessel, the range of
headings has to cover 360°.
When the vessel is turret moored, the heading of interest is the relative wave/vessel heading
that can be generally taken as [-90°, 90°] from "head sea".
The heading interval is often taken as 45°, in relation with commonly available directional
information. This is however not sufficient to catch the directional response of
vessel/structure to wave (e. g. horizontal wave bending moment) and precludes the use of
directional spectrum to describe waves.
A heading interval in the range of 15° to 22° will be adequate to get accurate directional
response using directional spectrum.
In case of turret moored, where the data to evaluate a distribution of relative heading is
generally not available, a simplification would consist to consider a single off-head heading,
and taking damages only from the windward side of the vessel.

Page 20
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

HYDRODYNAMIC ANALYSIS (1) STRUCTURAL ANALYSIS (2)

Hydrodynamic Model
3D FE Model of 3 Holds

HYDRODYNAMIC RESPONSE
R LOADS
•Wave Pressure
•Wave pressure
A
•Relative Wave Elevation
•Intermittent Wetting Pressure
O
•Accelerations
s •Inertia loads

•Hull Girder Wave loads


•Internal pressures in cargo and ballast tanks

•Hull Girder Wave loads at model ’s ends


FATIGUE DAMAGE (3)

Wave Scatter diagram Wave spectrum Stress RAO GLOBAL ANALYSIS

« Short » Term stress distribution LOCAL ANALYSIS


« Short » Term
« Long » Term stress distribution Damage Hot spot stress RAO

Damage

Figure 3-3 Flow chart for Full Spectral Fatigue Analysis of FPSO hull

Page 21
Reference to part of this report which may lead to misinterpretation is not permissible.
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

4 HYDRODYNAMIC ANALYSIS

4.1 Introduction
The objective of the hydrodynamic analysis is to provide RAOs (response to a harmonic wave
with unit amplitude) of vessel motions, resulting accelerations, wave pressures on hull and global
wave loads such as bending moments and shear forces.
For each draught corresponding to each loading condition, a hydrodynamic model of the wet
surface is built, in order to evaluate the parameters mentioned above.
The corresponding mass distribution is to be defined.
The modelling strategy will depend on the overall arrangements and capabilities of software for
hydrodynamic analysis, for structural analysis and on available interfaces routines.
The hydrodynamic analysis is performed by means of the 3D 1st order diffraction-radiation
method (linear potential theory).

4.2 Still water Model - Mass distribution


4.2.1 Still-water model
A 2D model composed of the geometrical lines of the hull (body plan) and including the position
of holds and ballast tanks is prepared using a standard ship loading condition/ stability software.
Each loading condition is described in terms of weight distribution and/or filling of the
compartments.
Ship equilibrium (i.e. draught and trim) is then obtained.
Still water bending moments and shear forces of the hull girder are also calculated.
4.2.2 Mass Model
The mass model has to reflect the steel weight distribution, the weight distribution of cargo and
ballast and the weight distribution of topsides and others items in all 3 directions.
When the structural model is a full ship FE model (see 5.1.2.1) or hybrid model, the mass matrix
is obtained from the structural model which includes a description of the mass distribution.
When the structural model is a 3-hold model (see 5.1.3.1) , the mass matrix of the modelled part
is obtained in the same way as above. The mass matrix of the fore and aft parts has also to be
evaluated, so that the aggregate mass matrix of all 3 parts is matching the global mass matrix of
vessel.
Any inaccuracy on total mass (matrix) and/or on the 3 partial mass (matrix) may result in either
an unbalance of the structural model or incorrect end shear forces/ bending moments (see 5.2.1).

Page 22
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

4.3 Hydrodynamic Model


For each draught (loading condition), the hydrodynamic model is cut at the waterline as only the
wet surface is considered in linear 3D analysis.
The dimensions of panels of the hydrodynamic model have typically to follow the requirements
below in order to ensure reasonable accuracy of hydrodynamic analysis:
• Panel size has to be uniform over the vessel's length, but finer in some areas such as near
waterline, near bilge, at ends.
• Panel size has to be fit for representing spatial variations of pressure field over the considered
domain of wave frequencies.
A common criterion for minimum length of the panel is Lmin = λ/6 to 10, where λ is the shortest
wave length. A typical size is in the range 5 m. Some inaccuracy at highest frequencies is thus
accepted.
In addition, depending on the overall arrangements and capabilities of software and on available
interfaces between hydrodynamic and structural models (see 5.2.2), the hydrodynamic model
may have to also satisfy constraints imposed by structural models and interfaces.
Three typical situations may be found:
• Hydrodynamic model is in agreement with the structural model i.e. each panel is matched to
the set of structural elements representing the same area of hull.
• Hydrodynamic model is identical to the structural model i.e. all panels are matched to the
structural elements in the region of the FE model. The hydrodynamic model is then generally
much finer than needed.
• Hydrodynamic model and structural model can be independent (subject to adequate
interface). It is however recommended that hydrodynamic model have two sections matching
the end sections of the structural model.
Static equilibrium of the hydrodynamic model has to be ensured by checking the consistency
between submerged volume and centre of buoyancy on one hand, mass and centre of gravity on
the other hand.

4.4 Hydrodynamic Analysis


The hydrodynamic analysis is based on linear theory. The vessel motions are calculated without
taking into account non-linearities due to intermittent wetting effect in the splash zone, flare and
water on deck.
The motion analysis will account for such loads as the external loads by incident and diffracted
wave, the inertial loads, the hydrodynamic added mass and damping loads and the hydrostatic
restoring forces.
The effect of mooring is generally neglected, as having limited influence on motions behaviour,
when the area of the structure under study doesn’t include mooring attachment. Similarly, riser
influence is generally neglected. In some cases riser may have contribution to roll damping.
Non-linear roll damping is usually included in order to avoid unrealistic amplitude of roll
motion. A common practice is to consider a wave amplitude at the same probability of
exceedance as the stress range that contribute the most to fatigue damage:

Page 23
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

• about the significant value for a single sea state,


• the value with a probability of exceedance between 10-2 to 10-3 when considering long term
distribution.
Typically, the wave frequencies are taken in the range of 0.1 to 1.5 or 2.0 rad/s.

4.5 Intermittent wetting


In linear diffraction-radiation analysis, hydrodynamic pressure is obtained over the wetted
surface at rest as a harmonic pressure variation at each point. Therefore, the pressure on the side
shell between the surface at rest and a wave crest is not modelled and, in a wave through below
the surface at rest an unrealistic negative pressure is generated.
This intermittent wetting is a non-linear effect. An accurate model would require a 2nd order or
higher pressure/motions model. In lieu of such models (that are currently not available),
intermittent wetting is taken into account as a correction of the linear pressure field.
The correction is local and is assumed not to affect significantly the vessel motions, nor to
modify the first order pressure field in the permanently submerged part. It will however result in
a more accurate structural response in the area affected by intermittent wetting.
Alternative methods to include the effects of intermittent wetting are presented and discussed in
Annex D.
The model for the local pressure and resulting pressure range have been calibrated with results of
full size measurements and model tests, and a reasonable agreement was found for the purpose
of a fatigue analysis.
Reference is made to ref. /30/.

Page 24
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

5 STRUCTURAL ANALYSIS
The purpose of the structural analysis is to determine the transfer functions (i.e the response on
regular waves) of nominal stress (or hot spot stress) or stress components /3/ at every locations
of interest of the FE model.

5.1 Modelling
5.1.1 Strategy
The structural model has to catch the structural response:
a. Hull girder loads (Wave bending moments and shear forces acting on "hull girder"
section),
b. Overall Hull Deformation not included in a), such as shear lag, differential displacement
between longitudinal bulkhead(s) and side shell,
c. Global deformations of cargo holds primary framing,
d. Local response (e.g. stiffener) to local loads and the effect of the above.
The structural model may be a full ship model (Figure 5-1) or includes several cargo holds
(generally 3) centred on the area of interest for the analysis.
In the latter case, the following methods can be used for the modelling of the rest of vessel in
order to get an overall balance of loads of the model:
• Ends loads (from the non-modelled end parts of the vessel) applied to the 3-hold model
Figure 5-2
• Hybrid model (Figure 5-3) composed of the 3-hold model and hull girder models of vessel
end parts, connected at both ends of the 3-hold model

Figure 5-1 Full Ship Model

Page 25
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Figure 5-2 3-Hold Model

Figure 5-3 Hybrid Model: 3-Hold Model + Hull Girder

The following aspects of structural modelling are discussed below:


• Full Ship Model
• FE Model of cargo holds
• Sub-Modelling of a part of the above models
• Stiffener model
• SCF Model used to determine local hot spot stresses in structural details.

A new built FPSO is assumed to be provided with an efficient corrosion protection system. All
finite element models may therefore be based on gross scantlings (i. e. nominal plate and bracket
thickness).
Guidance on structural modelling may also be found in Appendix F of /4/.

Page 26
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

5.1.2 Full Ship Model


5.1.2.1 Model
The Full Ship model is ideal to evaluate the total structural response including all structural
effects but it requires more modelling work than other options below.
The full structure of the vessel shall be included in the model. Structural components not
contributing to the global stiffness, such as superstructures and topsides, etc, are not normally
included in the global analysis. However, the mass of these elements should be correctly
included in the model (see 5.2.5).
The full ship model can be a relatively coarse mesh in order to get nominal stresses (from global
effects). Element sizes and types are selected for this purpose (typically one element between
frame). Stiffened panels may be modelled by means of anisotropic elements or hybrid elements
combining beams for stiffeners and semi-membrane plate elements for shell.
However, such coarse mesh may not be sufficient to catch framing deformation in the frame
adjacent to a bulkhead: relatively large element size in the global model may lead to a stiffer
frame system and consequently too small deflections in the frame.
Besides coarse mesh will not give reliable results in plates with lateral loads.
Then full ship model has to include sub-models with a medium coarse mesh covering a large
enough area of the global model in order to capture correct frame deflection.
Further local structure sub-models (e. g. one transverse frame or a stringer with adjacent
structures) have to be made in order to determine stresses in areas exposed to local lateral loads
in addition to global loads and deformations (see 5.1.4).
5.1.2.2 Boundary conditions
The boundary conditions for the global structural model should consist in simple supports that
will avoid built-in stresses.
The fixation points should be located away from areas of interest, as the load transfer from the
hydrodynamic load analysis may lead to some unbalance in the model.
5.1.3 FE Model of cargo holds (3 holds model)
5.1.3.1 Three holds model
The model extends over 3 holds, with end sections corresponding to transversal bulkheads, and
is used to obtain deformation response and nominal stress of the primary hull structural members
in the central hold of the model.
Such model will catch the structural response to wave loads, but may miss some part of the
deformations of type b) in 5.1.1 above due to shear lag as quoted there.
The current method to model the 3 holds consists to use plate elements for plating, transverse
and longitudinal bulkheads, web frames, stringers and beam elements for flange of web frames
and for stiffeners (see 5.1.5).
As shown on Figure 5-2, the 3 holds model includes a central area where the element size should
be as given in 5.1.4 below, and model ends, where a coarser mesh may be used.

Page 27
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

The central area shall include the central hold and a few frames fore and/or aft of central hold
transverse bulkheads, so as to cover bulkhead stringers and adjacent frames on one end at least.
In the coarse mesh part of the model, element size may be taken as one or two elements per
frame spacing, in the longitudinal direction, and one element per stiffener spacing, in the
transverse direction.
An alternative is to use a coarse mesh model for this 3 holds model, and fine mesh sub-models
(see 5.1.4) in areas where stresses are to be obtained.

5.1.3.2 Boundary conditions and end loads


The boundary conditions for the 3 hold model should consist in simple supports within end-
sections of model so as to ensure model stability and avoid built-in stresses.
Hull girder loads are to be applied at the end sections in order to recreate interactions of the non
modelled parts of the vessel on the 3-hold model.
The following techniques can be used for these purposes:
• displacements on one end section applied by fixing all degree of freedom over the full end-
section (one section only).
• stresses (forces distribution) applied along the end-sections, considering hull girder stress
assumptions (St Venant principle)
• loads or displacements of the end-sections applied assuming a plane and rigid hull girder
section (Navier & Bernoulli). Loads or displacements may be applied by using constraints of
the end section degrees of freedom to a master node. Loads or displacements are applied to a
master node and are thus transmitted to the entire hull girder.

Typically, the following arrangements of boundary conditions and end loads are used:
• The 3 holds model is fixed at one end. To ensure equilibrium, shear forces and bending
moments (vertical and horizontal) are applied at the other end at either a support node with
joints constraints or as loads on the section.
• Two support nodes are introduced, one at each end-section of the model, and located on the
centreline, at the elevation of the neutral axis of the structural model.
• Joint constraints are applied between these support (master) nodes and the nodes of the end-
sections. The model is simply supported through these nodes
• Global hull girder loads are applied to the master nodes along the degree of freedom that are
not defined as supports (typically the Hull girder bending moments at each end and the axial
force at one end).
The second arrangement is more convenient, both to specify end loads and to limit the impact of
eventual unbalance (see 5.2.6)
When a hybrid model is used, master nodes in both end-sections are fitted as above, and the
beams representing fore and aft parts of the hull girder are connected to these master nodes. The
end loads are thus automatically applied in the analysis.

Page 28
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

5.1.4 Sub-modelling
When necessary (see 5.1.2.1 and 5.1.3.1 above), fine mesh models corresponding to one or
several parts of a coarser mesh model (e.g. one transverse frame or a stringer with adjacent
structures) have to be prepared.
Element size is typically in the order of one stiffener spacing in all directions so as to include
three elements at least in the web height of primary framing. Such element size will lead to three
or more elements between web-frame. This will be sufficient to catch effective flange/shear lag
effect for web-frame (see also 5.1.5 below for stiffener modelling).
In brackets of primary framing, several elements have to be included (at least three in smaller
bracket).
Notes:
A. Guidance on element size is based on four nodes elements. Using higher order elements may
permit somewhat coarser meshing.
B. In order to get correct deformations and resulting stresses in small or slender parts such as
web frames or bracket quoted above, when using 4-node elements, elements with improved
in plane bending (i.e. elements with constant (in-plane) shear or additional (internal) degrees
of freedom) should be used.

Boundary of the sub-model should coincide with areas of the parent model where
displacements/forces are correct and be located "far enough" of analysed area to avoid undue
constraints onto a finely meshed part.
All loads (local pressures, inertia loads) acting on the corresponding part of the coarser model is
to be applied in a manner consistent with the fineness of this sub-model.

5.1.5 Stiffener Model


The following approaches can be used to model a stiffener.
5.1.5.1 Explicit modelling, using plate elements of appropriate size.
When modelling by plate elements, a current practice is to model stiffeners with 4-noded shell
elements, with typically 3 elements over the web height and 2 elements across the flanges. End
span connecting elements (Flat bar, brackets) are explicitly modelled in same way.
Note: Elements with improved in plane bending should be used (see note in 5.1.4 above).
The model must include several spans of the stiffener on each side of the area of interest. Such
technique will generally be used within a fine mesh sub-model.
Nominal stresses are obtained at stiffeners. They include the contributions of hull girder bending,
global hold deformation and local pressure effect, all three together.
The determination of nominal stresses may involve some extrapolations to avoid using stresses at
point of discontinuity. Alternatively, a SCF model may be included to directly get the hot spot
stresses.

Page 29
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

5.1.5.2 Modelling by beam elements


Stiffeners may be modelled using beam elements.
A first possibility would be to combine beam elements (with eccentricity) for stiffener itself and
plate elements for shell. However, there is incompatibility between the displacements along weld
connecting the web and associated plating, i. e. between formulation of in plane displacements of
the associated plating elements and edges displacements of a beam element. Therefore, this
modelling with plate elements for shell and beam elements (with eccentricity) for stiffener will
not perform correctly unless a large number of elements along one span are used. This is not
consistent with the intent of such modelling and this difficulty can be avoided using a hybrid
model.
The hybrid model is combining beams for stiffeners and semi-membrane plate elements for
shell. Beam elements provide the out of plane bending stiffness, including the effect of
associated plating, whereas plate elements only have in-plane stiffness. A small out of plane
stiffness must however be specified to avoid instability and transmit pressures to the stiffener.
The local pressure being transmitted as concentrated loads at the nodes of the model, instead of
distributed load over the span, the number of elements along one span should be six at least. This
number may be limited to four provided an analytical correction on bending component of stress
is performed when extracting stresses from the analysis.
This technique is an option that will permit to model large areas, and to check fatigue in a
number (all if desired) of connections of longitudinals, whilst helping to keep model size within
practical limits
Nominal (axial and bending) stresses in beam elements include the contributions of ship
bending, global hold deformation and local pressure effect.
The stiffener is modelled as a beam until the intersection point. In case of a bracket and a flat
bar, the nominal stress should be taken at 2/3 of the distance between frame and toe (of bracket
or stiffener), to get nominal stress at toe (Figure 5-4).

Page 30
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Frame or
bulkhead

Flat
bar
L

Bracket
2
L
3

Side
Shell

Figure 5-4 Location for calculation of nominal stresses

5.1.5.3 Analytical approach


The analytical approach is based on rule-based assessment procedure. Assumptions made in such
model and span effects (such as 3.5 in /3/ and 3.2 in chapter 4, section 3, in /25/) cannot provide
an accuracy similar to above models.
5.1.6 Stress Concentration Model
SCF Model is used to determine hot spot in details such as bracket connection, stiffener to web
frame connections or local design of frame/girder.
The principles for modelling used in the SCF model are given in Section 6. Use of submodelling
technique is assumed. The same principles also apply to analysis where the local SCF model is
included in the global FE model.
The boundaries of the sub-model (SCF model) should coincide with locations of the global
model where the displacements/ forces are calculated, and with “hard” points in the global
model.
The deformations of the global model FE model are transferred to the SCF model as boundary
displacements.

5.2 Loading
5.2.1 Overview of load transfer
The loads to be applied to the structural model are the dynamic loads resulting from wave
actions and vessel motions.
In hydrodynamic analysis, the loads acting on the whole vessel are described and resolved to get
the vessel motions.

Page 31
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

The equation of motion may be written as:

(M + M a )⋅ X & + QD + (K + km ) X = F(incident + diffracted wave)


&& + B ⋅ X (7)

where
X: vector of motions (6 components)
M: mass matrix of the vessel including solid mass and liquid in tanks
Ma: hydrodynamic added-mass matrix of the vessel
B: hydrodynamic linear damping matrix
QD : quadratic (roll) damping & ⋅X
QD 4 = C ⋅ X &
4 4

K: hydrostatic restoring matrix


km : stiffness matrix from mooring or risers
F exciting force and moment

In structural analysis, the load components corresponding to each term of this equation have to
be distributed over the model.
These loads may be split into the following terms:
• Solid mass and liquid in tanks submitted to accelerations, resulting in inertia forces and
&& ) ;
internal fluid pressures ( M X
• Incident, diffracted and radiated wave pressure and hydrostatic restoring force, resulting in
external pressure on hull ( M a X&& + B X
& + K X − F) ;
• Roll damping effect ( C X & X & ) , that is generally not taken into account in the load
4 4

model (see 5.2.6 below) ;


• Mooring loads ( km X ) : Generally, when the area of the structure under study does not
include mooring attachment, mooring stiffness does not affect substantially motions and
resulting loads, and is neglected.

Wave pressures and inertia loads obtained from the hydrodynamic analysis are transferred from
the hydrodynamic model into the structural model. So, hydrodynamic and structural models have
to be consistent, both in geometry (see 5.2.2) and in the description of mass (see 4.2 above).
Under the assumptions of a linear spectral analysis (see 3.2), overall loading for each wave
frequency is harmonic, and can be described by real and imaginary (cos and sin) parts, i.e. the
snapshots of loads at two instants over one period. Generally, structural resonance is not
anticipated, and these two load cases can be resolved by static analysis.
Depending on particulars of the structural analysis software, the whole loadings have to be
assembled, then resolved, or all parts can be resolved as “elementary load cases”, then the result
assembled by summation. When an elementary loading is a function of a single intensity
parameter, a “unit load case” can be defined.
Page 32
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

5.2.2 Wave pressure loads


The generation of pressures can be performed according to several methods with an impact on
modelling requirements as discussed in 5.1.1:
• Transfer of a uniform pressure on each panel (calculated at centre of facet of hydrodynamic
model) onto the corresponding set of structural elements. The structural model has to be
consistent with the hydrodynamic mesh).
• Transfer of a uniform pressure on each panel (calculated at centre of facet of hydrodynamic
model) onto each corresponding element of structural model. This requires that the
hydrodynamic mesh is completely identical to the structural model of outside shell.
• Interpolation of pressure between values of wave pressures on panels that is adjacent to each
structural element. This method is not recommended (difficulty in interpolation).
• Direct evaluation of wave pressure at each node of the structural model (Figure 5-5). In this
method which is based on recent theoretical and numerical developments, the wave pressure
is determined accurately at each node, without any interpolation /29/. This method permits to
have a hydrodynamic mesh independent of the structural model.

Note: for pressure description in the component and deterministic methods, see 5.2.8 below)

Figure 5-5 Wave pressure loads on structural model (example)

5.2.3 Intermittent wetting pressure


The modified pressure in the intermittent wetting area has to be described in the same way as
above (see 5.2.2).

Page 33
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Alternatively, the intermittent wetting correction can be entered as separate loading for later
combination (see Appendix D).
The modified pressure or the correction can be limited to the area of interest and the linear
pressure is used for the rest of the vessel. This will limit loads unbalance due to e. g. intermittent
wetting effect on shaped vessel ends.
5.2.4 Internal fluid pressure loads
Cargo and ballast tanks of the 3 holds model (or the area of interest) are loaded by internal fluid
pressure calculated from accelerations in tank.
A quasi-static approximation is used, under the assumption of small movements of the vessel
where the pressure is defined based on the intensity and angular variations of the total
acceleration (including gravity). In this approach, eventual resonant motion or sloshing effect is
neglected. (See D9 in Appendix D)
For tanks outside the area of interest, modelling by solid mass will generally be sufficient.
5.2.5 Inertia loading
The inertia loading is generated from the accelerations of the vessel acting on the steel mass of
the vessel, and other masses (e.g. topsides equipment).
The steel mass on the modelled part of the vessel is distributed through plate and beam elements
by means of adjusted density. When a hybrid model is used, the accelerations acting on ends
(modelled by hull girders) are applied to the masses associated to these parts.
Mass of topsides modules can be introduced as distributed mass on beam elements with no
rigidity.
5.2.6 Dynamic equilibrium
Dynamic equilibrium at each frequency is checked by comparing total load on structural model
with the results of hydrodynamic analysis.
In the case of a 3 holds model, total loads have to be equal to the difference between force values
calculated at end sections of the model from the hydrodynamic analysis.
The condition of dynamic equilibrium requires consistent mass distributions between
hydrodynamic and structural models.
Some unbalance may exist for some wave frequency (e.g. around roll natural frequency) in case
of off head wave due to some loads terms (e.g. roll quadratic damping see 5.2.1). This can reflect
on several components of total loads due to coupling between vessel motions. However, the
effect of such unbalance may be often considered as negligible when the resulting stresses are
moderate and the frequency is far enough from the range leading to maximum response.
5.2.7 Still Water loads
Generally, still water loads need not to be described.
However, for some structural details (e.g. at bulkheads), analysis in considering the combination
of wave loads and the loading/unloading loads has also to be studied (see 5.4.10 below)

Page 34
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

5.2.8 Wave pressure loads; component method and deterministic analysis


5.2.8.1 Deterministic Analysis
In a deterministic analysis, crest and through load cases are defined in the same way as for a
strength analysis (see /25/), i.e. through pressure diagrams including the total (hydrostatic plus
wave) external pressure. Alternatively, a pressure range loading can be obtained by difference.
The intermittent wetting is thus taken into account.
The external pressure at waterline is scaled with the relative wave elevation at a specified
probability level (10-2 to 10-3), instead of the Rules value.
Same approach is followed for the internal pressure in tanks that are scaled with the vertical
acceleration at same probability level as RWE.
Combinations of external pressure, internal pressure, and wave bending moments are made (see
/25/), from which the stress ranges at the reference probability level can be extracted.
5.2.8.2 Component method
In the component method (see /3/), the outside shell is split in a number of areas (e.g. side-shell,
bilge, bottom, on each side), and the pressure over each area is applied as a separate unit load
case. Internal pressures and end loads are also described as unit load cases.
The external pressures are derived from the pressure diagram over a mid-section of the model at
a 10-4 probability level, obtained from hydrodynamic analysis. This pressure diagram is modified
around the waterline to account for intermittent effect, then normalised by the pressure at a
reference panel for each area.
By this approach, the number of load cases to be computed in structural analysis can be
significantly reduced, with however a direct impact on the accuracy on the spatial description
(and the frequency dependence) of the implied pressure field, and therefore a direct impact on
the resulting structural response.

5.3 Analysis
5.3.1 Verification
Model and results should be verified at each step of analysis, as quoted in the above, , to ensure
that the derived results are reasonable. Simplified methods and/or experience from similar
analysis may be used for such assessment. The checklist below gives a summary of aspects to be
considered in verifications (see also 5.4.7.4 below).
• Hydrodynamic model:
- Mesh size/shape
- Buoyancy (COB, volume, trim)
- Mass data (COG, roll/pitch radius of gyration)
- Mass distribution (induced Still water bending/shear)
- Transfer functions (all relevant transfer functions are correctly described and seem
physical)

Page 35
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

- Roll eigen period and amplitude


- Correct scatter diagram and wave spreading
- Comparison with Rule loads

• FE- models :
- Mesh size/shape
- Plate thickness
- Material parameters
- Boundary conditions
- Mass distribution/balance
- Mesh density/shape
- Actual section values (area, section modulus) of the model
- Load application and load sums

• Structural analysis :
- Load transfer (pressures, acceleration, load sums)
- Global/local displacement patterns/magnitude
- Global section forces (compared to hydrodynamic analysis)
- Stress level and distribution
- Sub model boundary displacements/stresses

5.3.2 Results
From analysis, RAO of stresses are obtained.
A sufficient number of frequencies and headings is to be analysed to get meaningful RAO’s for
spectral integration (see 5.4.3 below): Generally, 30 to 35 frequencies will be adequate. 20 is a
minimum. For headings reference is made to 3.4.2 above.
Generally, unitary amplitude loading is used, then RAO of stress at any location is the
corresponding amplitude of stress.
When calculation is made for a finite amplitude wave, RAO of stresses is defined as:
SA SR (8)
RAO = =
A H
with:
SA = amplitude of stress variation

Page 36
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

A = wave (single) amplitude


SR = stress range
H = wave height
Thus an implicit linearisation of the response is made.
In the component method, the results of analysis are the transfer functions of stresses to each
load component. The RAO’s of stress at each location of interest are then obtained by
summation, over all load components, of the stress transfer functions times the RAO’s of the
corresponding load component.
In deterministic analysis, stress ranges are obtained from specified combinations.

5.4 Long term distribution of stresses and calculation of fatigue damage


The procedure of statistical analysis and calculation of damage can be summarised as follows:
• From RAO of stresses, short-term distributions each corresponding to a given sea state, a
wave heading and a condition of vessel are obtained by the technique of Spectral Analysis.
• The long-term stress distribution is obtained by summation of the short term distributions
over the wave scatter diagram at vessel site, the set of wave headings and loading conditions.
• Fatigue damage is obtained using Miner Rule, from long term stress distribution and S-N
data.
• Alternatively, "short-term" damages are calculated from the short-term distributions and the
total fatigue damage is obtained by summation of the "short term" damages.
• In simplified Fatigue analysis, long-term stress distribution is modelled by a Weibull
distribution.
5.4.1 Metocean conditions
Wave Scatter Diagram
The long-term description of seas at the FPSO's site is usually provided in terms of a scatter
diagram of sea-states, giving the probability of occurrence (Pi) of discrete sea-states: i, each
defined by the significant wave height : Hs and a period (e.g. the zero crossing period: Tz).
Wave Spectrum
Each sea-state is described by a wave energy spectrum of appropriate shape for the site under
consideration (e.g. Pierson-Moskowitz, JONSWAP, Ochi Hubble).
5.4.2 Response spectrum
The stress response is obtained from the wave spectrum and the stress RAO ( RAO σ ( ω θ w ) )
for each wave heading (θW) and each loading condition.

2
Sσσ (ω H S , Tz , θ W ) = RAO σ ( ω θ w .Sη η (ω H S , TZ ) (9)

Page 37
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

When directional spreading is considered, the stress response S σσ ( ω H s , Tz , θ w ) is


obtained as:
π
2 (10)
Sσσ (ω H S , Tz , θ W ) = ∫ RAO
−π
σ ( ω θ w ) .Sη η (ω H S , TZ , θ W ) ⋅ D (ω, θ − θ W ) dθ

where
Sηη (ω H S , Tz , θ W ) is short crested wave spectrum

D (ω, θ) is spreading function.

5.4.3 Spectral moments


The spectral moments of order n of the stress process are given by:
∞ (11)
m n = ∫ ω n ⋅ S σσ (ω H S , TZ , θ w ) dω
0

Assuming that the stress process is stationary and gaussian, the statistical parameters of the stress
process can be described below from the spectral moments:
• RMS (Root mean square): RMS = m o
m0
• Mean zero-upcrossing period of stress cycle: Tz = 2 ⋅ π⋅
m2
• ( ν Z = 1/ TZ is then the mean zero-crossing frequency)

The following observations can be made concerning the integration of spectral moments:
To obtain m0 (the intensity of the response), integration over a range of frequency based on the
energy in the wave spectrum (typically 0.45 to 2.2 the mean zero-crossing frequency of the wave
spectrum) would be sufficient.
However, a larger range is required in many cases, toward high frequencies, to get accurate
enough estimate of m2, and thus avoid to underestimate the number of cycles. This implies
generally an extrapolation of the stress response towards high frequencies.
Note:
A bandwidth parameter ε of the stress process can be defined, that is depending on the moment
m4. However a difficulty arises with m4, due to the current formulation of wave spectra (e.g. P-
M or JONSWAP), that have a m4 infinite : Some local response will exhibit same property, and
the calculated of the response will be spurious : therefore, the use of ε (e.g. in a bandwidth
correction) should be avoided.

Page 38
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

5.4.4 "Short-Term" stress distribution


For ship's structure, the stress range is generally assumed narrow banded and then assumed to be
Rayleigh distributed during each short-term loading state (i.e. the combination of a sea state, a
wave heading and a loading condition).
The probability of exceedance of the stress range level ∆σ for a loading state is given by:

∆σ 2 (12)
Pe (∆σ ) = exp(− )
8 m0

Note:
The narrow band assumption is realistic when wave loads do not induce significant resonant
response of the structure. The effect of corrections to account for bandwidth has been checked
for offshore structures, and found not so large compared with other uncertainties in fatigue
assessment such as uncertainties in loading, S-N data and Miner rule. Therefore, it is
recommended that any “wide band correction” be ignored.
5.4.5 "Long-Term" stress distribution (Spectral analysis)
The long-term stress distribution is the sum of the short-term distributions, weighted by the
probability of occurrence of each loading state, considering the number of stress cycles in this
loading state.
The number of exceedance “ne” of the stress range level ∆σ over reference duration Dref is given
by:
n e (∆ σ ) = D ref ⋅ ∑P ijk ⋅ Pe ijk ( ∆ σ ) ⋅ ν ijk (13)
i , j ,k

where:
Pijk Probability of occurrence of the loading state (ijk)
Peijk(∆σ) Probability of exceedance of the stress response (range ∆σ) to each loading state (ijk)
νijk Average zero-crossing frequency of the stress response to each loading state (ijk)
The probability of exceedance of the stress range level ∆σ is then given by:
ne (∆σ ) (14)
Pe (∆σ ) =
NT

where NT is the total number of stress cycles in the reference duration Dref , given by:

N T = ∑ Pijk ⋅ Peijk (∆σ ) ⋅ν ijk ⋅ Dref (15)


i , j ,k

Note: Alternatively, a long-term stress distribution is established in the same way for each
loading condition (see 5.4.7.3 below).

Page 39
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

5.4.6 "Long-Term" stress distribution (Deterministic Fatigue analysis)


In a deterministic (one-wave) analysis, the long-term stress range distribution is modelled by a
Weibull distribution, having shape parameter c and scale parameter b.
The number of exceedance of the stress range level ∆σ over a reference duration Dref is given by:

⎛ ⎛ ∆σ ⎞
h
⎞ (16)
ne ( ∆σ ) = N T ⋅ exp⎜⎜ − ⎜ ⎟ ⎟
⎟⎟
⎜ ⎜⎝ q ⎟

⎝ ⎠

The shape parameter h is estimated. h can be taken from the long term distribution of the
dominant wave effect (e.g. RWE for side shell stiffeners).
The scale parameter q is given by
∆σ 0 (17)
q=
(− ln p0 )
1/ h

with:
∆σ 0 = stress range value at the reference probability of exceedance po
N T = total number of stress cycles in the reference duration Dref ; NT is usually taken equal to
108 for 20 years.
5.4.7 Calculation of fatigue damage
Fatigue damage using Miner Rule may be obtained by numerical integration or by means of
"closed form" equations where applicable.
The damage can be calculated directly from the long term distribution of stresses or calculated
for each short term distribution, then summed over all loading states.
5.4.7.1 Numerical integration
The stress distribution is converted into a histogram.
A minimum of 40 to 50 classes at least are to be used, taking into account that only a limited part
of these classes will effectively contribute to the damage and to limit numerical inaccuracy e.g.
around change in slope of the S-N curve. 100 or 200 classes are typically used to eliminate such
problem and get results that are consistent with analytical integration in the close form approach.
The damage is given by:
i =c (18)
ni
D=∑
i =1 Ni

where:
c: number of classes
ni : number of cycles of stress ranges ∆σi in class i
Ni : number of cycles to failure at constant stress ranges ∆σi
Page 40
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

5.4.7.2 Closed Form Approach


Closed form approach can be used when the distribution is available in an analytical form.
A closed form approach is given in annex E for the short-term damage (Rayleigh distribution)
considering one or two slopes S-N curve.
A closed form equation is given in annex E for the long-term damage based on the "Weibull"
distribution used in a deterministic fatigue analysis.
5.4.7.3 Summation and fatigue life
The damage calculated for each loading condition is finally summed in order to obtain the total
damage and resulting fatigue life for each detail analysed.

DLT = ∑ Pk ⋅ Dk (19)
k

where Pk is the probability of occurrence of the loading condition k


The fatigue life (assuming DLT =1 at failure) is then given by:
D ref (20)
FL =
D LT

where Dref is the reference duration for which DLT have been calculated.
FL can be compared with the target design life, taking into account the relevant DFF (see
sections 2.2 and 2.3).
To perform the summation of damage, the distribution of draughts has to be specified based on
particular vessel operating profile.
Generally a limited number of loading conditions (draughts) is considered for the analysis, and
an interpolation of the damage over draughts is therefore necessary.
5.4.7.4 Evaluation
The results have to be thoroughly evaluated with respect to the assumptions/simplifications that
have been made in the course of the analysis and revisit some of them if necessary.
For this evaluation, it will be useful to identity the loading states that contribute the most to
fatigue damage (draught, heading, Hs/Tp combination)
An alternative is to do a sensitivity study on the assumptions made for the calculation (e.g.
spectral peakness, directional spreading, draught distribution).
5.4.8 Description of waves as Swell and Wind Seas
In some areas of the world, the wave conditions are better described as a combination of (long
distance) swell and (local) wind seas, resulting in fatigue damages (from each) having possibly a
similar magnitude.
The response from swell and waves may be calculated separately from the RAO’s of stresses,
and combined by the “combined spectrum” method:

Page 41
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

In the “combined spectrum” method, the two response spectra are added together, and the
response addressed as in the 5.4.2 to 5.4.4 above.
The standard deviation of the combined stress process is given by
2
s C = s1 + s 2
2 (21)

where s1 and s 2 are the standard deviations of the two stress processes.
The up-crossing rate of the combined stress process through the mean level is given by
1 2 2 (22)
νC = s1 ν12 + s 2 ν 22
sC

where ν1 and ν 2 are the up-crossing rates of the two processes.


This procedure tends to be conservative (with respect to a Rainflow counting), but is convenient
to use.
Note:
the “simple summation” method , in which the fatigue damages from the two processes are
simply added, is non-conservative in all cases (the two processes, in a considered time interval,
are acting one on top of the other) and should not be used.
the “dual narrow band” method quoted in 5.4.9 below is not applicable to such case.
5.4.9 Structures supporting loads from Mooring system or risers.
The load variations induced by vessel slow drift result in low-frequency stress, that have to be
evaluated, and can be combined with wave frequency stresses using the above “combined
spectrum”.
If both frequency components of stress are significant, then the conservatism of the combined
spectrum method can be reduced by using the “dual narrow-band” method. Reference is made to
ref. /24/ for detailed analytical equations of correction factor on fatigue damage.
5.4.10 Analysis of fatigue due to loading/unloading process
a) General
The number of Loading-Unloading Cycles of a FPSO (in the range of 1000 in 20 years) is much
larger than for a trading tanker and, although moderate, has been found to give a contribution to
fatigue in some cases.
For some structural details (e.g. at bulkheads), analysis in considering the combination of wave
loads and the loading/unloading loads has to be studied.
b) Still water stress range
For the analysis of fatigue due to loading/unloading process, a sequence of still water loads
conditions, representative of a full loading and unloading cycle should be described and
analysed. The successive steps, assumed repeated indefinitely, will then form a periodical signal

Page 42
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

of the stress at any location of interest, from which stress ranges can be extracted by Rainflow
counting, by starting from the maximum (or the minimum) stress over a full cycle.
Attention should be given that changes in the sequence can significantly alter the stress ranges
(e.g. whether two adjacent tanks are filled and emptied in reverse order or in same order would
change the count from two stress cycles to a single one with double amplitude).
When the sequence is not well defined, it might then be more appropriate to extract maximum
and minimum stresses at each location from a series of design “strength check” loading
conditions.
c) Combination with wave cycle
Due to the superposition of (slowly varying) still water stress cycles and wave cycles, the
damage at a given location is given, following the principle of Rainflow counting, by the sum of:
• the damage due to wave cycles alone,
• the damage due to the envelope of still water plus wave cycles, that may be taken as the
damage due to still water, with stress range(s) equal to the sum of the still water range(s) and
the wave stress range with a return period of one or two days.
d) Note:
Still water stresses might lead to high hot spot stresses at some details:
When in compression, a beneficial effect can be expected in parent material (see section 2.1.6)
When it tension, high local stresses may induce tearing in combination with fatigue crack
propagation. This should be avoided by proper detailing.

Page 43
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

6 FINITE ELEMENT ANALYSIS OF DETAILS

6.1 The hot spot stress methodology


6.1.1 The hot spot stress concept
The hot spot stress concept has a long history in fatigue design of tubular joints. The same
concept is also becoming part of a fatigue design methodology for plated structures in ships and
FPSOs. A main reason for this is that stresses from finite element analysis have to be compared
with some capacity for assessment of acceptance. Here a hot spot stress S-N curve is used as a
design capacity curve.
The hot spot stress can be directly derived from the performed finite element analysis, or it can
be derived through use of appropriate stress concentration factors and the nominal stress from
the analysis data. If fine element meshes are used for modelling a structure it is questioned how
to derive the nominal stress. For this reason a procedure for derivation of hot spot stress from FE
analyses is required where the hot spot stress can be linked to one hot spot S-N curve for
calculation of fatigue damage.
6.1.2 Methods for hot spot stress derivation
Due to the nature of the stress field at a hot spot region there are questions on how to establish
the hot spot stress, see Figure 6-1. For the hot spot concept the notch effect due to the weld is
included in the S-N curve and the hot spot stress is derived by extrapolation of the structural
stress to the weld toe as indicated in Figure 6-1.
It is observed that the stress used as basis for such an extrapolation should be outside that
affected by the weld notch, but close enough to pick up the structural stress.
Two methods may be used for hot spot stress derivation at plate surfaces from the calculated FE
results in this recommendation:
• Linear extrapolation over reference points 0.5 and 1.5 x plate thickness away from the hot
spot (technique preferred by most Classification Societies).
• No extrapolation, but considering the stress value at 0.5 x plate thickness away from the hot
spot as the relevant hot spot stress.
The reference points 0.5 and 1.5 x plate thickness away from the hot spot is also denoted read
out points. For analysis by solid elements where the weld is included in the model the distance to
the stress read out points is measured from the weld toe. For analysis by shell elements the
distance to the stress read out points is measured from the intersection lines as the weld is
normally not included in the finite element model.
The hot spot stress methodology is based on a linear stress distribution through the plate
thickness in the way that the finite element modelling is performed:
• One 20-node isoparametric element over the thickness (second order displacement function
giving linear stress when calculating its derivative).
• Shell elements that represent axial force and bending moment over the plate thickness.
In addition one may calculate the “structural stress” (also denoted geometric stress) from that of
a linear stress distribution through the plate at the hot spot. The methods described above imply
also a linear stress distribution through the thickness away from the hot spot.
Page 44
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

The hot spot stress can be calculated directly from the nodal forces from a FE analysis. The
nodal forces can be calculated from the element stiffness matrix and the corresponding nodal
displacement (and rotations for shell elements). Whether this feature is available or not depends
on the actual computer program that is used.
For three-dimensional analysis an integration (or linearization) of the stress may be performed as
an alternative. For shell elements the stresses are normally presented as membrane stress and
surface stress (membrane + bending) at the Gaussian points.
The “structural stress” or the hot spot stress can be calculated as follows:
σG = σm +σb (6.1)

where
σ m = membrane stress
σ b = bending stress
The membrane stress at a section through the thickness at the hot spot is calculated as

1
t (6.2)
t ∫0
σm = σ y ( x) dx

The bending stress is calculated from moment equilibrium as

6
t (6.3)
σ b = 2 ∫ σ y ( x) x dx − σ m
t 0

A similar integration of the nodal forces may be performed.


It should be noted that the hot spot stress derived from calculation of stresses through the
thickness of the plate need not be exactly the same as that derived by linear extrapolation on the
surface, ref. Figure 6-1. This is dependent on finite element type used and finite element
modelling.

Page 45
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Notch stress
Stress

Hot spot stress


Surface stress

Attachment
plate t/2 3t/2
Hot spot stress
Notch stress
Fillet weld

Membrane stress

Stress evaluation plane

Figure 6-1 Schematic stress distribution through plate thickness at a hot spot

Nominal stress

Fillet weld
Attachment plate

A A
Ref Figure 6-1

Figure 6-2 Schematic hot spot stress field in plate at attachment


Page 46
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

6.1.3 Link between hot spot stress and S-N curve


It should be noted that the finite element modelling might influence the calculated stress at the
hot spot region. Parameters affecting this are:
• type of element used
• size of elements at the hot spot region
• how the stresses are derived from the analysis (Gaussian stress, nodal stress etc.).
Therefore the same procedure for analysis should be used for design as that is used for
assessment of a design hot spot stress S-N curve. In principle different methods for hot spot
stress derivation may be linked to different S-N curves. However, it is essential that the same
methodology is used in design as that used in derivation of the design S-N curve.
The following considerations with respect to hot spot stress calculation and link to S-N curve
were made in the JIP:
• A number of Classification Societies have recommended derivation of hot spot stress
from linear extrapolation using the stress values calculated at the read out points 0.5t and
1.5t from the weld toe (or intersection line) from FE analysis with a well defined mesh
size at the hot spot region. This procedure is found acceptable with respect to accuracy
and robustness. However, it is considered to be time-consuming in terms of engineering
time. This is due to manual work involved in the derivation of the hot spot stress. It is
also observed that engineers perform this different if it is not properly described in the
guideline how to establish the stress at 0.5t and 1.5t. Some derive these stresses from the
Gaussian stresses and other directly from the nodal stresses. The procedure for
calculation of elements nodal stresses are often not clearly described in computer
programs and this introduces some uncertainty when unproven programs are used for hot
spot stress derivation. An example of hot spot stress derivation is shown in Figure 6-3 for
reduced integration order (2x2). First the Gaussian stresses at the top surface are
extrapolated to the section line for hot spot stress analysis. Then the stresses at the read
out points 0.5t and 1.5t are calculated and finally the hot spot stress is derived by linear
extrapolation. An alternative is to use a number of stress points along line A-B and fit
some function for derivation of stress at 0.5t and 1.5t. Another alternative is to use the
element stresses at the nodal points along line A-B and taking their mean value as stress
at the mid-side points.
• From an engineering point of view a simple derivation of the hot spot stress at a position
0.5t from the weld toe is attractive. It also shows small scatter in the calculated results.
Normally it provides somewhat lower hot spot values than the method of extrapolation
from 0.5t and 1.5 t. However, this difference can be compensated for by use of a lower
hot spot S-N curve or simply by increasing the 0.5t hot spot values by a factor 1.11 which
is approximately the ratio in stress between the FAT90 curve (applicable for 0.5t and 1.5t
extrapolation) and the FAT 80 curve (applicable for the 0.5t hot spot value). For S-N
curves reference is made to Section 8. Thus in the following the FAT90 curve will be
denoted the hot spot S-N curve.

Page 47
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Extrapolated
hot spot stress

Intersection
line

A
B
0.5 t 1.5 t
Hot 4 3 4 3
spot

1 2 1 2

Gaussian integration
point

Figure 6-3 Example of derivation of hot spot stress for 8-node shell elements

6.2 FE modelling and hot spot S-N curve


6.2.1 General
The following guidance is made to the computation of hot spot stresses with local models using
the finite element method:
Hot spot stresses are calculated assuming linear material behaviour and using an idealized
structural model with no fabrication-related misalignment. The extent of the local model has to
be chosen such that effects due to the boundaries on the structural detail considered are
sufficiently small and reasonable boundary conditions can be formulated.
In plate structures, three types of hot spots at weld toes can be identified as exemplified in
Figure 6-4:
a) at the weld toe on the plate surface at an ending attachment
b) at the weld toe around the plate edge of an ending attachment
c) along the weld of an attached plate (weld toes on both the plate and attachment surface).

Models with thin plate or shell elements or alternatively with solid elements are normally used.
It should be noted that on the one hand the arrangement and type of elements have to allow for
steep stress gradients as well as for the formation of plate bending, and on the other hand, only
the linear stress distribution in the plate thickness direction needs to be evaluated with respect to
the definition of structural stress.

Page 48
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

The following methods of modelling are recommended:


• The simplest way of modelling is offered by thin plate and shell elements which have to be
arranged in the mid-plane of the structural components, see also Figure 6-5.
• 8-noded elements are recommended particularly in case of steep stress gradients. Care should
be given to possible stress underestimation especially at weld toes of type b) in connection
with 4-noded elements. 4-node elements should contain improved in-plane bending modes.
4-node elements with improved in-plane bending modes are considered to provide reliable
hot spot stresses.
The welds are usually not modelled except for special cases where the results are affected by
high local bending, e. g. due to an offset between plates or due to a small free plate length
between adjacent welds such as at lug (or collar) plates. Here, the weld may be included by
vertical or inclined plate elements having appropriate stiffness or by introducing constrained
equations for coupled node displacements.
A thickness equal 2 times the thickness of the plates may be used for modelling of the welds
by transverse plates.
An alternative particularly for complex cases is offered by solid elements which need to have a
displacement function allowing steep stress gradients as well as plate bending with linear stress
distribution in the plate thickness direction. This is offered, e. g., by isoparametric 20-node
elements (with mid-side nodes at the edges) which mean that only one element in plate thickness
direction is required. An easy evaluation of the membrane and bending stress components is then
possible if a reduced integration order with only two integration points in the thickness direction
is chosen. A finer mesh sub-division is necessary particularly if 8-noded solid elements are
selected. Here, at least four elements are recommended in thickness direction. Modelling of the
welds is generally recommended and easily possible as shown in Figure 6-6.
Derivation of stress at read out points 0.5t and 1.5t
The element sizes depend on the method of stress evaluation. For this reason, the evaluation
methods are described before summarizing recommendations on element sizes. The stress
components on the plate surface should be evaluated along the paths shown in Figure 6-5 and
Figure 6-6 and extrapolated to the hot spot. The average stress components between adjacent
elements are used for the extrapolation.
Recommended stress evaluation points are located at distances 0.5 t and 1.5 t away from the hot
spot, where t is the plate thickness at the weld toe. These locations are also denoted as stress read
out points.
If the element size at a hot spot region of size txt is used, the stresses may be evaluated as
follows:
• In case of plate or shell elements the surface stress may be evaluated at the corresponding
mid-side points. Thus the stresses at mid side nodes along line A-B in Figure 6-3 may be
used directly as stress at read out points 0.5t and 1.5t.
• In case of solid elements the stress may first be extrapolated from the Gaussian points to
the surface. Then these stresses can be interpolated linearly to the surface centre or
extrapolated to the edge of the elements if this is the line for hot spot stress derivation.
For meshes with 4-node shell elements larger than t x t it is recommended to fit a second order
polynomial to the element stresses in the three first elements and derive stresses for extrapolation
from the 0.5 t and 1.5 t points. An example of this is shown schematically in Figure 6-7. This
Page 49
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

procedure may be used to establish stress values at the 0.5 t and 1.5 t points.
For 8-node elements a second order polynomial may be fitted to the stress results at the mid-side
nodes of the three first elements and the stress at the read out points 0.5 t and 1.5 t can be
derived.

Derivation of hot spot stress


Two alternative methods can be used for hot spot stress derivation:
• A linear extrapolation of the stresses to the hot spot from the read out points at 0.5t and 1.5t.
The principal stress at the hot spot is calculated from the extrapolated component values.
• The hot spot stress is taken as the stress at the read out point 0.5 t away from the hot spot and
multiplied by 1.11.

Hot spot S-N curve


It is recommended to link the derived hot spot stress to that of FAT90 curve IIW (1996), see also
section 8.

c
b
c
a

Figure 6-4 Different hot spot positions

Page 50
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Figure 6-5 Stress extrapolation in a three-dimensional FE model to the weld toe

Figure 6-6 Stress extrapolation in a three-dimensional FE model to the weld toe

Page 51
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

6.2.2 Mesh size using 4-Node Shell Elements


Use of 4-node shell elements with improved in plane bending gives reliable hot spot stress based
on the hot spot stress methodology described in section 6.2.1.
It is recommended to use a mesh size from 0.5t x 0.5t up to 2t x 2t. Larger mesh sizes at the hot
spot region may provide non-conservative results.

6.2.3 Mesh size using 8-Node Shell Elements


Use of 8-node shell elements with internal degrees of freedom gives reliable hot spot stress based
on the hot spot stress methodology described in section 6.2.1.
It is recommended to use a mesh size from 0.5t x 0.5t up to 2t x 2t. Smaller and larger mesh sizes
at the hot spot region may provide non-conservative results. This depends on detail to be
analysed.

6.2.4 Mesh size using Three-Dimensional Solid Elements


Solid modelling is time consuming and more laborious than shell modelling that is normally the
preferred method when a plated structure is to be analysed. However, in some cases of more
complex structures solid elements are considered to give more reliable results and are
recommended used.
For modelling with three dimensional elements the dimensions of the first two or three elements
in front of the weld toe should be chosen as follows. The element length may be selected to
correspond to the plate thickness. In the transverse direction, the plate thickness may be chosen
again for the breadth of the plate elements. However, the breadth should not exceed the
“attachment width”, i. e. the thickness of the attached plate plus 2 x the weld leg length (in case
of type c: the thickness of the web plate behind plus 2 x weld leg length), see Figure 6-6.
However, based on the performed analyses in this project it is likely that the length of the
elements can be increased to 2t without significant loss of reliability.
For 20-node hexahedral elements it is sufficient with one element over the thickness to pick up a
linear stress distribution. For simple 8-node brick elements at least 4 elements are required for
the same purpose.
In cases where three-dimensional elements are used for the FE modelling it is recommended that
also the fillet weld will be modelled to achieve proper local stiffness and geometry. (Some
simplifications always have to be assumed when fillet welds are neglected in the modelling).
In general finite element meshes should be made without significant distortion of the mesh and
width length width ratio within 1:4. As the fillet weld will normally be included in a model based
on three-dimensional elements the general requirements to FE mesh will likely limit the size of
the mesh also the hot spot region.
In order to capture the properties of bulb sections with respect to St Venant torsion it is
recommended to use several three-dimensional elements for modelling of a bulb section. If in
addition the weld from stiffeners in the transverse frames is modelled the requirements with
respect to element shape will govern the FE model at the hot spot region.

Page 52
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Hot spot 0.5 t 1.5 t


stress
Second order polynomial

Results from FE analysis

0 t 2t 3t 4t 5t 6t
Distance from hot spot
Figure 6-7 Derivation of hot spot stress for element size larger than t x t

6.3 Derivation of effective hot spot stress from FE analysis


At hot spots with significant plate bending one might derive an effective hot spot stress for
fatigue assessment based on the following equation:
∆σ e , spot = ∆σ a , spot + 0.60 ∆σ b , spot (6.4)

where
∆σ a , spot = membrane stress

∆σ b , spot = bending stress.


The reason for a reduction factor on the bending stress is load shedding effect during crack
growth. The effect is limited to areas with a localised stress concentration, which occurs for
example at a hopper corner, see section 6.7.1. However, in a case where the stress variation
Page 53
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

along the weld is small, the difference in fatigue life between axial loading and pure bending is
much smaller. Therefore it should be noted that it is not correct to generally reduce the bending
part of the stress to 60 percent. This has to be restricted to cases with a pronounced stress
concentration.

6.4 Thickness effect


It is recommended to use the S-N curves defined in Section 8 for thickness effects. For the hot
spot stresses derived using the recommended methodology a thickness exponent k = 0.25 should
be used.

6.5 Expected accuracy of finite element analysis for hot spot stress analysis
Linear finite element analysis provides rather accurate hot spot stresses provided that the
recommendations with respect to finite element modelling and read out of stresses are followed.
Reference is made to /9/. Quite a number of different analyses by different finite element
programs and users provided hot spot stress results that were “good”. “good” was defined to be
in the region -5 to +10 % of target hot spot stress value. A number of analyses were within
“acceptable” that was defined as being within -10 to -5 and +10+ to +20 of the target hot spot
stress. The methodology was not allowed to provide more non-conservative results than
approximately 10% below target. However, it was accepted that methodology for some types of
specimens provided results more conservative than +20% of target. See also ref./9/.

6.6 Validation of computer programs and finite elements for fatigue analysis
6.6.1 General
In this project a number of finite elements and computer programs have been validated for a
number of different structural details. See also refs. /10/ and /11/.
If other programs are used for hot spot stress analysis it is recommended to test out the program
as well as the user on well known details and verify the results against proper target values.

6.6.2 Details to be analysed for validation


The HHI details 1-5 have frequently been referred to in this project as they are considered as
representative specimens for ship and FPSO design. They are also appropriate for validation as
target hot spot stress values are defined. See reference /9/.
If the validation results in hot spot stresses different from target values a calibration factor might
be introduced to account for this discrepancy as proposed in IIW (1996). See also DNV-RP-
C203 (2001).
6.6.3 Target hot spot stress values
Reference is made to /9/ for target hot spot stress values from the FPSO Fatigue Capacity JIP.

Page 54
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

6.7 FE analysis of hopper corner details


6.7.1 Welded structure
At hot spots of welded hopper corners there will be significant local bending stress over the plate
thickness. Use of the surface stress as a hot spot stress without taking the stress gradient through
the thickness into account leads to a conservative approach.
Thus, details with the same hot spot stress but with different stress gradient will show the same
fatigue lives based on S-N data while the actual lives might be very different. This may be
illustrated by crack growth analysis using fracture mechanics. An example of a situation with the
same hot spot stress, but with different stress gradients is shown in Figure 6-9. Crack growth of a
semi-elliptic crack is considered as shown in Figure 6-8.
For a situation with some bending over the plate thickness the calculated fatigue life becomes
different as seen from the same Figure 6-9. And in case of pure bending the difference in fatigue
life is increased further.
The effect of bending stress at a hopper corner can be taken into account in design by calculating
an effective hot spot stress as shown in section 6.3.
It should be noted that eccentricity may have significant influence on the calculated hot spot
stress, ref. /18/. Therefore eccentricities at hopper corners should be included in the finite
element model for assessment of local stresses.

Initiation point Semi-elliptic crack

Section through beam at weld toe

Figure 6-8 Crack growth analysis of hopper corner

Page 55
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

10
9
8
Crack depth (mm)

7
6
5
4
3 PureMembrane
2 SCF=1.8
1 Pure Bending

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Normalised Time
Figure 6-9 Crack growth curves for same hot spot stress with different stress gradients

6.7.2 Effect of tolerances for hopper corners


The effect of fabrication tolerances for a hopper knuckle was investigated in ref /18/. A load case
with motion induced pressure acting inside the ballast tank was analysed. Based on this analyses
it was found that a nominal eccentricity of 0.25 t would result in the lowest stress concentration
factor and the longest fatigue life, see Figure 6.

Page 56
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Hopper sloping Eccentricity


plate (Misalignment)

Inner bottom plate

Girder

Figure 6-10 Eccentricity of hopper plate connection

6.7.3 Bent hopper corner


At a bent hopper corner there is less local plate bending than at a welded hopper corner, see also
ref./19/.

Page 57
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

7 DESIGN OF FILLET WELDED CONNECTIONS

7.1 Recommended design criteria


The reference stress in fillet welds to be used together with design S-N curve is calculated
according to section 2.1.4. This stress shall be used together with the FAT36 curve (or W3
curve), see Chapter 8.
For a continuous fillet weld (such as section B-B in Figure 7-1) the reference stress to be used
together with the design curve is obtained from
Floc section (7.1)
σw =
2ale

where the notations are explained in Figure 7-2.


Floc section = axial force in the element (t = thickness and le = element width)
a = throat thickness.
For the ends of the attachment the reference stress to be used together with the design curve is
obtained from
Floc end (7.2)
σw =
2ta + a 2
where the notations are explained in Figure 7-2.
Floc end = axial force in half the element txt/2 (t = thickness)

Figure 7-2 shows the forces and the reference areas used for derivation of the local nominal weld
throat stresses (or reference stress) to be used together with the S-N curve. The stress is assumed
to be derived from a fine mesh finite element analysis using a “hot spot stress finite element
model” in a section at the fillet weld being investigated. The element stresses in the elements at
the fillet welds are used for calculation of forces that are being transferred through the fillet
welds. Even if three-dimensional elements are used for the analysis it is recommended to follow
this procedure as it is difficult to derive reliable reference stresses directly from the stress
analysis of the fillet welds.

Page 58
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

200

A R=30 A

A-A

Figure 7-1 Fatigue tested specimen, see also Figure 7.2

Page 59
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Reference weld force area

Reference weld stress area

a a

t t

a a
t/2 a

Element at Element at
a section t end of t
B-B bracket

le t

σ1 + σ 2 + σ 3 + σ 4 σ 1 + 3σ 2 + 3σ 3 + σ 4
Floc section = le t Floc end = t2
4 16

At a section B-B At the end of bracket

Figure 7-2 Methodology for calculation of local forces for fillet weld

Page 60
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

7.2 Effect of residual stress and mean stress on fatigue life


The fatigue design procedure presented in section 7.1 is likely conservative due to residual
stresses in compression at the ends of the attachments. Reference is made to refs. /15/ and /16/.
The methodology is considered to be conservative for the whole stress cycle in the plate being in
tension. Therefore a bonus factor of 1.2 for the weld root at the attachment end is proposed. This
means that the reference stress in the weld throat can be reduced by this factor. A bonus factor of
1.1 is proposed for the weld toe at the attachment end. There is no bonus factor for the fillet weld
some distance away from the attachment end. Before using these bonus factors in a design it
should be verified that the residual in an actual fabrication is similar to that of the considered test
specimens. It should me remembered that taking out more of the reserve capacity there is less
reserve capacity to that of actual tolerances that becomes more important.
It should also be added that it is likely that the design procedure becomes more conservative
when part of the stress cycle is in compression.

7.3 Design of fillet welded penetrations in plated structures


7.3.1 Critical hot spot areas
Fatigue cracking around a circumferential weld may occur at several locations on reinforced
rings in plates depending on geometry of ring and weld size, ref. Figure 7-3.
1. Fatigue cracking transverse to the weld toe in a region with a large stress concentration
giving large stress parallel to the weld (Flexible reinforcement). See Figure 7-3a.
2. Fatigue cracking parallel to the weld toe (Stiff reinforcement with large weld size). See
Figure 7-3b.
3. Fatigue cracking from the weld root (Stiff reinforcement with small fillet weld size). See
Figure 7-3c.

All these potential regions for fatigue cracking should be assessed in a design with use of
appropriate stress concentration factors for holes with reinforcement.

Page 61
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Position of fatigue crack Comment


Fillet weld

σp Fatigue crack growing normal to the weld toe


due to large stress concentration when insert
tubular is thin.
Then σ hot spot = σ p

a)

σn
α Fatigue crack initiating from the weld toe for
σ1 thicker insert tubular. The principal stress σ1 is
the crack driving stress.
45°

Then σ hot spot = σ 1


Also the region at crown position to be checked.
Insert Then σ hot spot = σ n
Tubular

b)

τ σn Fatigue crack in the fillet weld (initiating from


p
the weld root) at region with large normal stress
45°

and shear stress (Small fillet weld size in relation


to thickness of insert tubular or stiffening ring).

c)

Figure 7-3 Potential fatigue crack locations at welded penetrations

7.3.2 Principal stress direction relative to weld toe


The procedure for fatigue analysis of weld toes is based on the assumption that is only necessary
to consider the ranges of the cyclic principal stresses in determining the fatigue life. When the
potential fatigue crack is located in the parent material at the weld toe, the relevant hot spot
stress is the range of maximum principal stress adjacent to the potential crack location with stress
concentrations being taking into account. The principal stress within ± 60o normal to the weld
toe together with an S-N curve for the weld for fatigue assessment is shown in 2.1.2. Then the
principal stress is linked to the FAT90 following the IIW (1996) notation for fatigue assessment
Page 62
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

of the weld toe location (or the D-curve following DNV – RP- C203 (2001)).
If the angle between the principal stress and the normal to the weld toe is larger than 60o the
fatigue life with a stress component parallel with the weld together with an appropriate S-N
curve that depends on the welding process should be used, ref. IIW (1996), see also Chapter 8.

7.3.3 Stress concentration factors for holes with reinforcement


Stress concentration factors for holes with reinforcement are given in Appendix 3 of DNV-RP-
C203 (2001).
8-node shell elements were used for analysis of tubulars through plates, ref. Figure 7-3. The
welds were not included in the shell models. 20-node three-dimensional isoparametric elements
were used for analysis of penetrations stiffened by flat rings. Typical weld sizes were included in
the three-dimensional models. A large width of the plate were included in the FE model for
simulation of “infinite” width. The size of the element mesh at the hot spot region was txt where
t = plate thickness. One element was used over the main plate thickness for one sided flat ring
stiffener. For double sided flat ring stiffeners the condition of symmetry was used which implied
2 elements over the main plate thickness. The hot spot stress in the shell model was derived from
extrapolation of the surface stress at point t/2 and 3t/2 to the intersection line. The hot spot stress
in the three-dimensional model was derived from stresses at the elements nodal points that were
derived in the computer program from the gaussian stresses. Then the stresses at points t/2 and
3t/2 from the weld toe were used for a linear stress extrapolation for derivation of the hot spot
stress at the weld toe. Then the SCFs in the graphs of DNV-RP-C203 were derived by
calculating the ratio of the hot spot stress and the nominal stress.

Page 63
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

tr

tp
r r
A A H A A
tr tr
A-A

a) Maximum stress parallel with weld b) Stress normal to the weld


at the 45° position

σ1 θ

r r r
A A A A A A
tr tr tr

c) Principal stress at the d) Shear stress in the plate e) Stress normal to the weld
45° position at 45° position

Figure 7-4 Type of stress and positions for stress calculation at a tubular through a plate

7.3.4 Procedure for fatigue design of penetrations with welded reinforcement


The fatigue assessment procedure is described as follows:
Potential fatigue cracking transverse to the weld toe, ref Figure 7-3a and Figure 7-4:
For stresses parallel with the weld the local stress to be used together with the C curve is
obtained with SCF from Appendix 3 of DNV-RP-C203.

Potential fatigue cracking parallel to the weld toe, ref Figure 7-3b:
For stresses normal to the weld the resulting hot spot stress to be used together with the D curve
is obtained with SCF from Appendix 3 of DNV-RP-C203. Then the hot spot stress is derived as
σ = SCF σ (7.3)
Hot spot Nominal

Potential fatigue cracking from the weld root, ref. Figure 7-3c:

Page 64
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

The relevant stress range for potential cracks in the weld throat of load-carrying fillet-welded
joints and partial penetration welded joints may be calculated from:
(7.4)
∆σ w = ∆σ ⊥2 + ∆τ ⊥2 + 0.2 ∆τ 2//
The stress components in the fillet weld is shown in Figure 7-5.
At some locations there are stress in the plate perpendicular to the fillet weld, σ n , see Figure
7-3b, and a shear stress in the plate parallel with the weld τ // p , see Figure 7-3d.
The 45o position is the region selected for assessment.
Equilibrium of plate in section parallel with the weld gives:
τ // 2a = τ // p t p (7.5)
where
τ // = shear stress in the weld as shown in Figure 7-5
τ // p = the shear stress in the plate as shown in Figure 7-3d
a = throat thickness of weld
tp = plate thickness
The shear stress in the weld is then obtained from eq. (7.3) as
τ // p t p (7.6)
τ // =
2a
Equilibrium of plate in section normal to the weld (See Figure 7-6) gives:
(7.7)
(τ ⊥ + σ ⊥ ) 1 2 2 a = σ n t p
2
And assuming a reaction force on the weld throat in the direction of σn then gives:
σn tp (7.8)
τ⊥ =σ⊥ =
2 2a
Then from eq. (7.2), (7.4), and (7.6) the fillet weld is designed for a combined stress obtained as
t (7.9)
∆σ w = ∆σ n + 0.2∆τ // p
2 2

2a
where
t = plate thickness
a = throat thickness for a double sided fillet weld.
The resulting stress range shall be used together with the W3 curve (= FAT36 in IIW for air
environment).
The total stress range (i.e. maximum compression and maximum tension) should be considered
to be transmitted through the welds for fatigue assessments.

Page 65
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

σ = σz
τ = τyz
τ = τxy
τ
σ
τ
Throat
section y
x

Figure 7-5 Explanation of stresses on the throat section of a fillet weld

σ τ

tp σn

Fillet weld
a

Figure 7-6 Stress in section normal to the weld

Page 66
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

h T
t P
L

Figure 7-7 Calculation of nominal stress in fillet weld

7.3.5 Example of design analysis


A tubular φ 800x15 is used as a sleeve through a deck plate of thickness 20 mm. The tubular will
be welded to the deck plate by a double sided fillet weld.
Assume Weibull parameter h = 0.90 and that the deck has been designed such that S-N class F3
details (= FAT56) can be welded to the deck plate and still achieve a fatigue life of 20 years.
From DNV-RP-C203 Table 2.14-2 a maximum stress range of 199.6 MPa during 108 cycles is
derived for the F3 detail.
Questions asked:
1. Is the fatigue life of the penetration acceptable with respect to fatigue cracking from the weld
toe?
2. How large fillet weld is required to avoid fatigue cracking from the weld root?

Page 67
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

The following assessment is made: r/tp = 20, tr/tp = 0.75. Then from DNV-RP-C203:
Figure 4 SCF = 2.17 applies to position Figure 7-3a.
Figure 6 SCF = 0.15 applies to position Figure 7-3c and weld root.
Figure 8 SCF = 1.07 applies to position Figure 7-3b and weld toe.
Figure 10 SCF = 0.46 applies to position Figure 7-3b and weld root.
Figure 12 SCF = -0.75 applies to position Figure 7-3b and weld toe.
The C-curve or FAT125 applies to position Figure 7-3a.
The D-curve or FAT90 applies to weld toes of Figure 7-3b.
The W3-curve or FAT36 applies to weld root of Figure 7-3c.
Check of fatigue cracking at the Figure 7-3a position:
∆σ = ∆σ0 x SCF = 199.6 x 2.17 = 433.13 MPa which is just within the acceptable value of 445.5
MPa for a C detail, ref. DNV-RP-C203 Table 2.14-2.
Check of fatigue cracking at the Figure 7-3b position:
∆σ = ∆σ0 x SCF = 199.6 x 1.07 = 213.57 MPa which is well within the acceptable value of 320.8
MPa for a D detail, ref. DNV-RP-C203 Table 2.14-2.
Thus the fatigue life of weld toe is acceptable.
The required throat thickness is calculated as follows. From DNV-RP-C203 Table 2.14-2 a
maximum stress range of 128.2 MPa for a W3 detail (Weibull shape parameter = 0.90).
Then from equation (6):
σ no min al SCF t p 199.6 * 0.46 * 20 211.71
τ⊥ =σ⊥ = = =
2 2a 2 2a a

τ // p t p σ no min al SCF t p 199.6 * 0.46 * 20 918.16


τ // = = = =
2a 2a 2a a
Then from equation (7.2):
2 2 2
⎛ 211.71 ⎞ ⎛ 211.71 ⎞ ⎛ 918.16 ⎞
128.2 = ⎜ ⎟ +⎜ ⎟ + 0.2 ⎜ ⎟
⎝ a ⎠ ⎝ a ⎠ ⎝ a ⎠
From this equation a required throat thickness is a = 4.0 mm for both sides of the plate. This is a
required weld size that is well below the minimum required weld size specified in ship
classification rules.

Page 68
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

8 S-N CURVES

8.1 Definition of S-N curves and failure criterion


The design S-N curves in this document are based on the mean-minus-two-standard-deviation
curves for relevant experimental data. The S-N curves are thus associated with a 97.7%
probability of survival.
The failure criterion in the S-N curves is crack growth through the thickness of small scale
specimens. For specimens where the crack is initiated at a plate edge the crack length
corresponds to approximately 20 mm.
Experience from tanker structures is that fatigue cracks can grow to a significant size without
leading to a catastrophic failure. However, this depends on considered detail. Cracks as large as
1 metre have been discovered in tankers during survey. It is considered difficult to document that
the structure is safe with so large cracks. Also leakage may become a severe problem at a much
earlier stage of crack growth. Therefore maximum size of cracks to be detected during in-service
at different locations should be assessed when planning inspection programmes for an FPSO.
The basic design S-N curve is given as
log N = log a − m log ∆ σ (8.1)
N = predicted number of cycles to failure for stress range ∆σ
∆σ = stress range
m = negative inverse slope of S-N curve
log a = intercept of log N-axis by S-N curve
loga = log a − 2 s (8.2)

where
a = constant relating to mean S-N curve
s = standard deviation of log N.

The fatigue strength of welded joints is to some extent dependent on plate thickness. This effect
is due to the local geometry of the weld toe in relation to thickness of the adjoining plates. It is
also dependent on the stress gradient over the thickness. The thickness effect is accounted for by
a modification on stress such that the design S-N curve for thickness larger than the reference
thickness reads:
⎛ ⎛ t ⎞
k
⎞ (8.3)
log N = log a − m log⎜ ∆σ ⎜⎜ ⎟⎟ ⎟
⎜ ⎝ t ref ⎠ ⎟
⎝ ⎠
where
tref = reference thickness equal 25 mm for welded connections.
t = thickness through which a crack will most likely grow. t = tref is used for thickness less
than tref.
k = thickness exponent on fatigue strength as given in Table 8-1, Table 8-2 and Table 8-3.
Thickness effects are mainly used for welded joints and not for the base material, ref. also /2/.
(However, there is also a thickness effect factor used for cast joints and bolts).
Page 69
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

8.2 S-N curves for different details and environments


Fatigue design S-N curves are presented in Table 8-1 for air environment, in Table 8-2 for
seawater with cathodic protection and in Table 8-3 for seawater with free corrosion. These S-N
curves are from DNV-RP-C203 (2001). However for air environment below 5*106 cycles they
are similar to that of IIW(1996).
The reason for referring to the S-N curves of DNV-RP-C203 rather than IIW is that only S-N
curves for air conditions is presented in IIW; also the transition in slope is at 5*106 cycles and
with a cut off level at 108 cycles in IIW. In DNV-RP-C203 the transition in slope is at 107 cycles
for air condition and 106 cycles for seawater with cathodic protection. There is no cut off limit in
DNV-RP-C203. Reference is made to ref. /12/ for basis of S-N curves in DNV-RP-C203.
For offshore structures in the North Sea it has been normal practice to use S-N curves with
transition in slope at 107 cycles. The S-N curves related to type of environment should be
selected based on the actual corrosion protection system planned for the FPSO. For a newbuilt
FPSO a good corrosion protection system may be used. For a converted tanker the quality of the
corrosion protection may be more dubious. Then one might consider to split the service life into
one interval with good protection and one interval at the end with a less good system and
calculate fatigue damage from different S-N curves for the two intervals as is used in ship
design, ref. /3/.
It should be noted that the content of the cargo tanks are likely corrosive and the S-N curve
should be selected based on a proper assessment of the corrosion protection that will be used.
The effect of different environments is shown in Figure 8-1 for the D-curve. The D-curve is
similar to FAT90 in air environment below 5*106 cycles. This S-N curve is also the basic hot
spot S-N curve. Note that the FAT notation in IIW is related to S-N curves in air environment
only.
For stress parallel with a weld the C-curve may be used (= FAT125 below 5*106 cycles).
For design of fillet welds the W3 curve may be used for design (= FAT36 below 5*106 cycles).
S-N curves B1 and B2 apply to the base material. Curve B1 can be used for rolled plates and
sections. Curve B2 can be used for machine gas cut surfaces and curve C should be used for
manually gas cut surfaces.
The other S-N curves apply to different details using the nominal stress approach. For further
description of classification of details see Appendix 1 of DNV-RP-C203.
For design of butt welds FAT90 may be used when welded from sides. For design of cruciform
joints based on the nominal stress approach FAT 71 (Curve F) can be used for full penetration
and FAT 63 (curve F1) for fillet welds. Reference is also made to DNV-RP-C203 (2001),
sections 2.1.3 and Appendix A.

Page 70
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Table 8-1 S-N curves in air


S-N curve S-N curve log a1 log a 2 Fatigue Thickness Stress
DNV-RP- notation N ≤ 107 N > 10 7 limit at 107 exponent k concentration
C203 IIW (1996) cycles cycles cycles inherent the S-N
m1 = 3.0 detail as derived by
m2 = 5.0
the hot spot
method
B1 FAT160 12.913 16.856 93.57 0
B2 FAT140 12.739 16.566 81.87 0
C FAT125 12.592 16.320 73.10 0.15
C1 FAT112 12.449 16.081 65.50 0.15
C2 FAT100 12.301 15.835 58.48 0.15
D FAT90 12.164 15.606 52.63 0.20 1.00
E FAT80 12.010 15.350 46.78 0.20 1.13
F FAT71 11.855 15.091 41.52 0.25 1.27
F1 FAT63 11.699 14.832 36.84 0.25 1.43
F3 FAT56 11.546 14.576 32.75 0.25 1.61
G FAT50 11.398 14.330 29.24 0.25 1.80
W1 FAT45 11.261 14.101 26.32 0.25 2.00
W2 FAT40 11.107 13.845 23.39 0.25 2.25
W3 FAT36 10.970 13.617 21.05 0.25 2.50

Table 8-2 S-N curves in seawater with cathodic protection


S-N curve log a1 loga2 Fatigue limit at Thickness Stress concentration
DNV-RP- N ≤ 10 cycles 107 cycles exponent k inherent the S-N detail
6
N > 10 cycles
6
C203 as derived by the hot
m1 = 3.0 m2 = 5.0 spot method
B1 12.513 16.856 93.57 0
B2 12.339 16.566 81.87 0
C 12.192 16.320 73.10 0.15
C1 12.049 16.081 65.50 0.15
C2 11.901 15.835 58.48 0.15
D 11.764 15.606 52.63 0.20 1.00
E 11.610 15.350 46.78 0.20 1.13
F 11.455 15.091 41.52 0.25 1.27
F1 11.299 14.832 36.84 0.25 1.43
F3 11.146 14.576 32.75 0.25 1.61
G 10.998 14.330 29.24 0.25 1.80
W1 10.861 14.101 26.32 0.25 2.00
W2 10.707 13.845 23.39 0.25 2.25
W3 10.570 13.617 21.05 0.25 2.50

Page 71
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Table 8-3 S-N curves in seawater for free corrosion


S-N curve log a Thickness Stress
DNV-RP- For all cycles m = exponent concentration
C203 3.0 k inherent the S-N
detail as derived
by the hot spot
method
B1 12.436 0
B2 12.262 0
C 12.115 0.15
C1 11.972 0.15
C2 11.824 0.15
D 11.687 0.20 1.00
E 11.533 0.20 1.13
F 11.378 0.25 1.27
F1 11.222 0.25 1.43
F3 11.068 0.25 1.61
G 10.921 0.25 1.80
W1 10.784 0.25 2.00
W2 10.630 0.25 2.25
W3 10.493 0.25 2.50

1000
Air

Seawater with cathodic protection

Seawater free corrosion


Stress range (Mpa)

100

10
1.00E+04 1.00E+05 1.00E+06 1.00E+07 1.00E+08 1.00E+09
Number of cycles

Figure 8-1 S-N curve D as function of environment (hot spot stress S-N curve)

Page 72
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

9 REFERENCES
/1/ Lotsberg, I.: Phase II Summary Report. DNV Report 2003-0579. June 2003.
Also Bergan, P. G. and Lotsberg, I.: Advances in fatigue Assessment of FPSOs.
OMAE-FPSO’04-0012. Int. Conf. Houston.
/2/ DNV-RP-C203 Fatigue Strength Analysis of Offshore Steel Structures. October 2001.

/3/ CN 30.7 Fatigue Assessment of Ship Structures. DNV, 2003.

/4/ DNV-RP-C102 Design of Offshore Ships February 2002.

/5/ Kuo, J.-F., Lacey, P. B., Zettlemmoyer, N. and MacMillan, A.: Fatigue Methodology
Specification for New-Built FPSO. OMAE Paper no 3016, Rio de Janeiro, June 2001.
/6/ Na, J. H., Lee, I. H., Sim, W. S. and Shin, H. S.: Full Stochastic Fatigue Analysis for
Kizomba ‘A’ FPSO-Hull Interface Design. Proceedings 22nd Int. conf. on Offshore
Mechanics and Arctic Engineering, Cancun Mexico, June 2003.
/7/ IACS Part A Shipbuilding and Repair quality standard for New Construction.

/8/ Hobbacher, A.: Fatigue Design of Welded Joints and Components. IIW. XIII-1539-96/
XV-845-96.
/9/ Lotsberg, I.: Assessment of Hot Spot Stress Methods for finite Element Analysis. DNV
Report No. 2002-0556 April 2003.
Also: Lotsberg, I. (2004b), “Recommended Methodology for Analysis of Structural
Stress for Fatigue Assessment of Plated Structures”. OMAE-FPSO'04-0013, Int. Conf.
Houston.
/10/ Fricke, W. (2001). Recommended Hot Spot Analysis Procedure for Structural Details
of FPSO’s and Ships Based on Round-Robin FE Analyses. Proc. 11th ISOPE,
Stavanger. Also International Journal of Offshore and Polar Engineering. Vol. 12, No.
1, March 2002.
/11/ Fricke, W. and Säbel, A.: “Hot Spot Stress Analysis of Five Structural details and
Recommendations for Modelling, Stress Evaluation and design S-N Curve. GL Report
No FF99.188.A, Rev 02, June 2000.
/12/ Lotsberg, I. and Larsen, P. K. (2001). Developments in Fatigue Design Standards for
Offshore Structures, ISOPE, Stavanger.
/13/ Lotsberg, I and Lindberg Bjerke, S.: Tested Fatigue Capacity of Fillet Welded
Specimens. DNV Report No. 2002-3124. April 2003.
And: Lotsberg, I. (2003), “Fatigue Capacity of Fillet Welded Connections subjected to
Axial and Shear Loading”. IIW Document no XIII-2000-03 (XV-1146-03). Also to be
published in Journal of Fatigue.
/14/ Lotsberg, I: Fatigue Capacity of Fillet Welded Penetrations. DNV Report No. 2002-
0255. October 2002.
Also: Lotsberg, I. (2004a), “Fatigue Design of Welded Pipe Penetrations in Plated
Page 73
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

Structures”. Marine Structures, Vol. 17/1 pp 29-51.


/15/ Doerk, O.: Experimental Investigation of Root Cracking at Fillet welds Around
Attachment Ends. TUHH April 2003.
Also: Fricke, W., Doerk, O. and Gruenitz, L. (2004), “Fatigue Strength Investigation
and Assessment of Fillet-Welds around Toes of Stiffeners and Brackets. OMAE-
FPSO'04-0010. Int. Conf. Houston.
/16/ Doerk, O.: Report on Design Criterion for Root Crack Assessment at Fillet Welds
Around Attachment Ends. TUHH May 2003.
/17/ Chen, W. and Landet, E.: Stress Analysis of Cut-outs with and without Reinforcement.
OMAE Rio de Janeiro, June 2001.
/18/ Polezhaeva, H. and Chung, H.: Effect of Misalignment on the Stress Concentration of a
Welded Hopper Knuckle for a Typical FPSO. OMAE Rio de Janeiro, June 2001.
/19/ Kim, Y. I. and Jeon, J. C.: Fatigue Strength of Bent Hopper Corner Detail. DSME June
2003.
/20/ Gurney, T. R.: Fatigue Design Rules for Welded Steel Joints, the Welding Institute
Research Bulletin. Volume 17, No. 5, May 1976.
/21/ Maddox, S. J.: Developments in Fatigue Design Codes and Fitness-for-service
Assessment Methods. International Conference on Performance of Dynamically
Loaded Welded Structures. Welding Research Council, Inc. 1997.
/22/ Yoneya, T., Kumano, A., Yamamoto, N. and Shigemi, T.: Hull Cracking of Very Large
Ship Structures. IOS’ 93, Glasgow 1993.
/23/ Lotsberg, I., Askheim, D. Ø., Haavi, T. and Maddox, S. ( 2001) Full Scale Fatigue
Testing of Side Longitudinals in FPSOs. ISOPE Stavanger.
/24/ DNV-OS-E301 Position Mooring. DNV June 2001.
/25/ Bureau Veritas, "Fatigue Check of Structural Details" in Part B, Chp7 of "Rules and
Regulations for the Classification of Steel Ships, May 2003.
/26/ Cramer, E.H., Loseth, R., Bitner-Gregersen, E., "Fatigue in Side Shell Longitudinal due
to External Wave Pressure", OMAE Vol.2 pp.267-272, 1993.
/27/ Pastoor, L.W., Pinkster, J.A., Krekel, M., "Local and Global Hydrodynamic Loads on
the FPSO 'Glass Dowr' for Fatigue Assessments", ISOPE 1998.
/28/ Van der Cammen J., "Model-test validation of Intermittent Wetting Model", Doc No.
FFC-G-100-RP-2004/Rev.A, August 2003.
/29/ Blandeau, F., Francois, M., Malenica, S., Chen, X.B., "Linear and non-linear wave
loads on FPSOs", ISOPE, 1999.
/30/ Van der Cammen J., Kaminski M. L., Francois M., (2004) " Is uncertainty of wave
induced fatiguing loading on FPSO's uncertain?", OMAE-FPSO’04-0055. Int. Conf.
Houston.
/31/ Agerskov, H., Petersen, R.I., Lopez Martinez, L., " Fatigue in high-strength steel
offshore structures", 9th International Symposium on Offshore Engineering, Brazil,

Page 74
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS FPS
IP

O
J
Report No: 2003-0582, rev. 02

i ty
Fa t
TECHNICAL REPORT ig
u e Ca p a c

September 1995.
/32/ Lotsberg, I. and Sigurdsson, G. (2004), “Hot Spot S-N Curve for Fatigue Analysis of
Plated Structures”. OMAE-FPSO'04-0014, Int. Conf. Houston.
/33/ Kim, W.S. and Lotsberg, I. (2004), “Fatigue Test Data for Welded Connections in Ship
Shaped Structures”. OMAE-FPSO'04-0018. Int. Conf. Houston.
/34/ Lotsberg, I. and Rove, H., “Stress Concentration Factors for Butt Welds in Stiffened
Plates”. OMAE New Orleans, (2000).

- o0o -

Page 75
Reference to part of this report which may lead to misinterpretation is not permissible.
FatiguedesignrecommendationsRev02October2004.doc
DET NORSKE VERITAS/BUREAU VERITAS

APPENDIX
A
EFFECT OF TOLERANCES

Page A-1
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

A.1 Effect of misalignment on butt welds in plates


A1.1 Introduction
During the last years the reliability of equations for stress concentration factors for butt welds in
design rules for floating production vessels (FPSOs) have become an important issue for fatigue
design. Block sections welded from one side only and buttwelds at thickness transitions are
found to be critical areas in terms of calculated fatigue life. A representative stress distribution at
butt welds is required in order to perform a reliable fatigue design of these areas and to establish
a sound basis for planning in-service inspection.
A stress concentration factor can be defined as a stress magnification at a detail due to the detail
itself or due to a fabrication tolerance with the nominal stress as a reference value. The
maximum stress is often referred to as the hot spot stress that is used together with S-N data for
fatigue life calculation. This hot spot stress is thus derived as the stress concentration factor
times the nominal stress.

A1.2 Physical explanation of increased stress at an eccentric butt weld in plates


The stress at a plate butt weld between two plates as shown in Figure A-1 is considered. It is
assumed that the plates are welded together with an eccentricity δ (and without angular
mismatch). The plates are subjected to a loading per unit width N = σnominal * t, where σnominal is
stress and t = thickness of the plates. Due to the asymmetry of the static system shown in Figure
A-1d the moment will be zero at the middle of the weld (inflection point). Then moment
equilibrium yields that the maximum moment becomes M = N δ /2 as shown in Figure A-1f. The
section modulus for a unit plate width is W = t2/6. Then the bending stress over the plate is
obtained as

δ δ (1)
N σ nominal t
σb =
M
= 2 = 2 =3δ σ
2 nominal
W W t t
6
and the stress concentration frequently referred to at an unstiffened plate weld joint is obtained
from the definition given in the introduction as

σ nominal + σ b δ (2)
SCF = = 1+ 3
σ nominal t

Page A-2
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

a)

b)

t
Notch region

A-A

Static system:
N

Deflected shape:
N

Bending moment:

Axial + bending stress:

σ σ δ

Figure A-1 Stress at an eccentric butt weld


Page A-3
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

The following equation applies for eccentricity with shift in neutral axis due to eccentricity of
plates in addition to shift in neutral axis due to joining of plates with different thickness:

t 20.5 t 20.5 (4)


SCF = 1 + 6 δ 1.5 1.5 + 3(t1 − t 2 ) 1.5 1.5
t1 + t 2 t1 + t 2

Tolerances are important for calculation of stress concentration factors for butt welds and
cruciform joints. It is not quite obvious what tolerances are accounted for in the S-N data, as
eccentricities were not measured for most of the test data that are used as a basis for derivation of
the design S-N curves that are used today. In IIW (1996) it is stated that an eccentricity of 0.10t
is included in the test data for butt welds and 0.15t for cruciform joints for the S-N curves to be
used for these connections based on a nominal stress approach, ref. 2.1.1. In testing of cruciform
joints the transverse plate is free in terms of boundary conditions. In a real structure it is
restrained with respect to rotation in a similar manner as the two other plates. Therefore this
corresponds to a tolerance of 0.3t that is accounted for in the S-N data for cruciform joints. This
is in the range that normally is being accepted as a fabrication tolerance for cruciform joints.
For butt welds an effective eccentricity in equations 2, 3 and 4 can be calculated as
δ = δ max fabrication code − 0.1t (6)

The maximum value of eccentricity in the IACS construction standard is the minimum of 0.15t
and 3 mm for butt welds and t/3 for cruciform joints.
Thus when the IACS standard is fulfilled with respect to these tolerances mainly butt welds at
connections with different plate thickness such as shown in Figure 10 need to be investigated
with respect to additional stress resulting from eccentricity. Then an effective eccentricity from
equation (6) is used for calculation of stress concentration factor to be multiplied with the
nominal stress before the design S-N curve is entered.

t2
t1

Figure A-2 Worst case combination of eccentricity and transition in thickness

Page A-4
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

A.2 Root gap in fillet welds


Test specimens with root gaps up to 3 mm have been fatigue tested in the laboratory. The test
results show that the fatigue capacity increase with increasing gap. The reason for this is that the
weld goes into the root gap and a larger throat thickness is achieved.
In ship design the root gap is limited to that of 2 mm following IACS, ref. /7/.

g ≤ 2 mm

Figure A-3 Gap at tee fillet, from ref. /7/.

- o0o -

Page A-5
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

APPENDIX
B
MEAN STRESS EFFECTS

Page B-1
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

Residual Stress and effect of mean stress


The procedure for fatigue analysis using S-N curves originating from The Welding Institute
(Gurney (1976)) is generally based on the assumption that only the cyclic stress ranges need to
be considered for fatigue assessment of welded structures, and the mean stresses can be
neglected. This has been an efficient design approach for bridges and offshore structures for the
last 25 years.
It is likely that some of the residual stress as introduced by the welding is removed by repetitive
load cycles. However, a significant reduction in residual stress should be demonstrated in order
to rely on mean stress effect in design. The largest contribution to fatigue damage for ships and
offshore structures is related to the small stress ranges in the S-N curve. As pointed out by
Maddox (1997) the residual stress has to be very low in order that part of the cycle comes into
compression at these small stress ranges.
Most fatigue tests are performed under pulsating tension loading. Thus the S-N curves are
applicable for this loading condition.
Most ship structures are subjected to some permanent loads, which in some cases results in
significant mean stress. When the mean stress is high and in compression, the calculated fatigue
damage might be over estimated.
The Japanese Ship Classification Society (NK) performed a survey of 30 second generation
VLCCs (very large crude oil carriers) built in 1987-90 where a number of side longitudinals had
cracked during the first years of service. NK (Ref. Yoneya et al.1993) pointed out that in their
study of damages, most cracks were found in cargo tanks near the loaded water line, with high
stress concentration between 2 to 5 m below the loaded water line. NK stated that there was
practically no damage in wing ballast tanks. As explained by NK, in a full load condition, the
end connections of the side longitudinals in the empty wing ballast tanks would only be
subjected to compressive stresses under the action of sea loads acting on ship’s sides. While the
end connections of the side longitudinals in the loaded cargo tanks would be subjected to
constant tension under internal pressure, and the side pressure from sea would induce stress
fluctuation on the tension side. As a result, the end connections of the side longitudinals in cargo
tanks became more susceptible to fatigue damage.
Thus in many Classification Companies there has been a strong wish to include a mean stress
effect in the fatigue design procedure. At present the mean stress effect has been reflected in the
fatigue assessment procedure by Bureau Veritas, the China Classification Society, DNV,
Germanischer Loyd, Korean Register of Shipping and RINA. The mean stress effect is not been
included by ABS and Lloyd’s Register.
The effect of mean stress was also investigated in the full scale fatigue tests of side longitudinals
reported by Lotsberg et al (2001).
Two specimens with double sided brackets were tested both in ballast and loaded conditions. The
test results from the loaded condition, giving compression at the hot spot, are well above the S-N
curve for tensile load cycling. The capacities are shown in Figure B-1. It is noted that the ballast
condition is significantly more severe than the loaded condition in terms of fatigue damage. This
may be explained by the different mean stresses at the hot spots for the two loading conditions:
• In specimen no 1, the ballast condition (water filled ballast tanks) leads to tensile stresses at
the hot spots (neglecting the residual stress). (R = σmin/σmax = 0.16)
Page B-2
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

• In specimen no 2, the loaded condition (empty ballast tanks) leads to compressive stresses at
the hot spots (neglecting the residual stress). (R = - ∞).
The test results from specimens 1 and 2 are presented in terms of reduction factors on
compressive stress in Figure B-2. It is shown that the effect of mean stress in these tests is larger
than that given in the DNV CN 30.7 (2001), ref. /3/, for fatigue assessment.
Most design standards for land and offshore structures allow for mean stress effects for base
material and some for Post Weld Heat Treated (PWHT) structures. The Japanese Society of Steel
Construction is one of few standards that include a beneficial effect of compressive mean stress
in fatigue assessment in addition to the aforementioned classification companies.
A different practice with respect to mean stress effects have developed in the ship industry as
compared with the offshore industry. One reason for this is difference in material strength used
in the two industries. A typical material yield strength for offshore structures is in the range 350
– 500 MPa. The material yield strength used in ships is in the range 235-320 MPa. A lower yield
strength in ship structures implies that local yielding a hot spots will more likely occur than if a
higher material yield strength is used and that residual tensile stresses are removed during early
service life. Thus, it is considered acceptable to include some beneficial effects from
compressive load cycles when a fatigue assessment of a ship is performed. Also the consequence
of a fatigue crack is traditionally viewed different in a sailing ship compared with an offshore
platform as a repair can easily be performed after each 5-year survey in a dock.

2
Specimen no

0 1000000 2000000 3000000 4000000 5000000 6000000 7000000


Fatigue life (Number of cycles)

Figure B-1 Comparison of the fatigue lives obtained from specimens in the loaded and
ballast conditions for the same load range

Page B-3
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

1.2
Reduction factor on stress range

1
Compressive stress
at hot spot
0.8
Curve for design used by one
Classification Company
0.6

0.4
Full scale fatigue
test result
0.2

0
-1 -0.5 0 0.5 1
Mean stress per stress range

Figure B-2 Reduction factor on stress range from test compared with a design curve

- o0o -

Page B-4
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

APPENDIX
C
DESIGN OF CUT-OUTS IN PLATED STRUCTURES

Page C-1
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

C.1 Geometry
The following hole geometries are considered see also Figure C-1:
1) Circular cut-out with diameter = 600 mm
2) Rectangular cut-out 600 x 800 mm with rounded corner R = 300 mm
3) Rectangular cut-out 600 x 1200 mm with rounded corner R = 300 mm

1) Cut-out Ø600 2) Cut-out Ø600x800 3) Cut-out Ø600x1200

Figure C-1 Cut-out geometry

For the three cut-out geometry six different edge reinforcements is applied, see Figure C-2. The
reinforcement details are described below, see also Figure C-2:
(A) Cut-out alone (no reinforcement) (Figure C-2 (A))
(B) Cut-out with inserted plate (15 mm thick, 300 mm wide) around the edge (Figure C-2
(B))
(C) Cut-out with double side reinforcement 50 mm away from the edge (Figure C-2 (C))
(D) Cut-out with single side reinforcement 50 mm away from the edge (Figure C-2 (D))
(E) Cut-out with double side reinforcement 100 mm away from the edge (Figure C-2 (E))
(F) Cut-out with single side reinforcement 100 mm away from the edge (Figure C-2 (F))

Stress concentrations factors are presented for the hot-spots marked in Figure C-2.

Page C-2
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

(A) (C) (E)

(B) (D) (F)

Figure C-2 Cut-out, reinforcement and hot-spot positions

For geometry (D) and (F), the maximum stresses of the bottom or the top surface in the 20 mm
plate at the cut-out edge are given in the plots. For the other geometry the stresses are
symmetrical about the mid-plane of the plate.
C.2 Applied stresses
The following stresses has been considered:
• longitudinal stress, σx
• transverse stress, σy
• shear stress, τ
The stresses in the longitudinal and the transverse directions are applied separately but are
combined with shear stress. The shear stress is varied between zero and up to the value of the
normal stress.
C.3 Stress Concentration Factor Definition
The definition of the stress concentration factors presented for cut-outs are the maximum
principal stress divided by the nominal normal stress, σx or σy, (not the nominal principal stress).
The maximum principal stress in the hot-spot is selected as the maximum of | σ1| and | σ2|.
The stress concentration factor (Kg) is then:

K g , x( y ) =
(
max σ 1 ; σ 2 )
σ x( y )

Page C-3
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

C4 Results
In general, stress concentration factors are given at 5 points (see Figure C-2) except for the cases
shown in Figure C-2 (A) and (B).
The following should be noted:
• Maximum principal stresses are parallel to the weld toe (hot-spots 2 to 5) with only one
exception:
for double reinforcement and point 2 (see Figure C-2,(C) and (E), the maximum principal
stress is normal to the weld toe.

C4.1 SCFs for Point 1, ref. Figure C-2

6.50

6.00

5.50

5.00
K g= σ 1max /σ x

A1
4.50
B1

4.00 C1
D1
3.50 E1
F1
3.00

2.50

2.00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σx

Figure C-3 Circular Cut-out Ø=600 mm, σx and τ

Page C-4
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

7.00

6.50

6.00

5.50

5.00
K g = σ 1max /σ x

A1

4.50 B1
C1
4.00 D1
E1
3.50 F1

3.00

2.50

2.00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σx

Figure C-4 Rectangular Cut-out with Rounded Corners: 600 x 800 mm, σx and τ

7.50

7.00

6.50

6.00

5.50
K g = σ 1max /σ x

A1
5.00 B1
C1
4.50 D1
E1
4.00
F1

3.50

3.00

2.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σx

Figure C-5 Rectangular Cut-out with Rounded Corners: 600 x 1200 mm, σx and τ
Page C-5
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

7.50

6.50

5.50
K g = σ 1max /σ y

A1
4.50 B1
C1
D1

3.50 E1
F1

2.50

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/σy

Figure C-6 Rectangular Cut-out with Rounded Corners: 600 x 800 mm, σy and τ
7.00

6.50

6.00

5.50

5.00
K g = σ 1max /σ y

A1
4.50
B1

4.00 C1
D1
3.50 E1
F1
3.00

2.50

2.00

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
τ/ σy

Figure C-7 Rectangular Cut-out with Rounded Corners: 600 x 1200 mm, σy and τ

Page C-6
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

C4.2 SCFs for Point 2, ref. Figure C-2

5.50

5.00

4.50

4.00
K g = σ 1max / σ x

B2

3.50 C2
D2
E2
3.00
F2

2.50

2.00

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σx

Figure C-8 Circular Cut-out Ø = 600 mm, σx and τ, stresses for C and E are normal to the
weld

6.00

5.50

5.00

4.50
K g = σ 1max /σ x

4.00 B2
C2
3.50
D2
E2
3.00
F2
2.50

2.00

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σx

Figure C-9 Rectangular Cut-out with Rounded Corners: 600 x 800 mm, σx and τ

Page C-7
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

6.50

6.00

5.50

5.00

4.50
K g = σ 1max /σ x

B2
4.00
C2
D2
3.50
E2

3.00 F2

2.50

2.00

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σx

Figure C-10 Rectangular Cut-out with Rounded Corners: 600 x 1200 mm, σx and τ

6.00

5.50

5.00

4.50
K g = σ 1max /σ y

4.00 B2
C2
3.50 D2
E2
3.00 F2

2.50

2.00

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σy

Figure C-11 Rectangular Cut-out with Rounded Corners: 600 x 800 mm, σy and τ
Page C-8
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

7.00

6.50

6.00

5.50

5.00
K g = σ 1max /σ y

4.50
B2
4.00
C2

3.50 D2
E2
3.00
F2
2.50

2.00

1.50

1.00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σ y

Figure C-12 Rectangular Cut-out with Rounded Corners: 600 x 1200 mm, σy and τ

C4.3 SCFs for Point 3, ref. Figure C-2


5.00

4.50

4.00
K g = σ 1max / σ

3.50
B3
C3

3.00 D3
E3
F3
2.50

2.00

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σx

Figure C-13 Circular Cut-out Ø = 600 mm, σx and τ


Page C-9
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

5.50

5.00

4.50

4.00
K g = σ 1max / σ

B3
3.50 C3
D3
3.00 E3
F3

2.50

2.00

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σ x

Figure C-14 Rectangular Cut-out with Rounded Corners: 600 x 800 mm, σx and τ

6.50

6.00

5.50

5.00
K g = σ 1max /σ x

4.50
B3
4.00 C3
D3
3.50
E3

3.00 F3

2.50

2.00

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σ x

Figure C-15 Rectangular Cut-out with Rounded Corners: 600 x 1200 mm, σx and τ

Page C-10
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

6.00

5.50

5.00

4.50
K g = σ 1max / σ y

4.00
B3
C3
3.50 D3
E3
3.00 F3

2.50

2.00

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σ

Figure C-16 Rectangular Cut-out with Rounded Corners: 600 x 800 mm, σy and τ

7.50

7.00

6.50

6.00

5.50

5.00
K g = σ 1max / σ y

B3
4.50
C3
4.00 D3
E3
3.50
F3
3.00

2.50

2.00

1.50

1.00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
τ/ σ y

Figure C-17 Rectangular Cut-out with Rounded Corners: 600 x 1200 mm, σy and τ

Page C-11
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

C4.4 SCFs for Point 4, ref. Figure C-2


3.20

3.00

2.80

2.60

2.40
K g = σ 1max /σ x

2.20 C4
D4
2.00
E4
1.80 F4

1.60

1.40

1.20

1.00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σx

Figure C-18 Circular Cut-out Ø = 600 mm, σx and τ

3.50

3.30

3.10

2.90

2.70
K g = σ 1max /σ x

2.50 C4
D4
2.30 E4
F4
2.10

1.90

1.70

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σx

Figure C-19 Rectangular Cut-out with Rounded Corners: 600 x 800 mm, σx and τ

Page C-12
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

4.00

3.50

3.00
K g = σ 1max / σ x

C4
D4
2.50
E4
F4

2.00

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σ x

Figure C-20 Rectangular Cut-out with Rounded Corners: 600 x 1200 mm, σx and τ

3.50

3.00

2.50
K g = σ 1max /σ y

C4
D4
2.00
E4
F4

1.50

1.00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σy

Figure C-21 Rectangular Cut-out with Rounded Corners: 600 x 800 mm, σy and τ

Page C-13
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

4.00

3.50

3.00
K g = σ 1max /σ y

2.50 C4
D4
E4
2.00 F4

1.50

1.00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σy

Figure C-22 Rectangular Cut-out with Rounded Corners: 600 x 1200 mm, σy and τ

C4.5 SCFs for point 5, ref. Figure C-2

3.40

3.20

3.00

2.80
K g = σ 1max / σ x

2.60

C5
2.40
D5
2.20 E5
F5
2.00

1.80

1.60

1.40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σx

Figure C-23 Circular Cut-out Ø = 600 mm, σx and τ


Page C-14
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

4.00

3.50

3.00
K g = σ 1max / σ x

C5
D5
2.50 E5
F5

2.00

1.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σx

Figure C-24 Rectangular Cut-out with Rounded Corners: 600 x 800 mm, σx and τ

4.00

3.80

3.60

3.40

3.20
K g = σ 1max / σ x

3.00
C5
2.80 D5
E5
2.60
F5
2.40

2.20

2.00

1.80
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σx

Figure C-25 Rectangular Cut-out with Rounded Corners: 600 x 1200 mm, σx and τ
Page C-15
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

3.60

3.40

3.20

3.00

2.80
K g = σ 1max / σ y

2.60
C5
2.40 D5
E5
2.20
F5

2.00

1.80

1.60

1.40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σy

Figure C-26 Rectangular Cut-out with Rounded Corners: 600 x 800 mm, σy and τ
4.00

3.50

3.00
K g = σ 1max /σ y

2.50 C5
D5
E5
2.00 F5

1.50

1.00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

τ/ σy

Figure C-27 Rectangular Cut-out with Rounded Corners: 600 x 1200 mm, σy and τ
- o0o -

Page C-16
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

APPENDIX
D
INTERMITTING WETTING PRESSURE AND INTERNAL FLUID
PRESSURE

Page D-1
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

CONTENT

D.1 Introduction 1
D.2 Pressure Model 1
st
D.2.1 Linear theory (3D 1 order diffraction-radiation) pressure 2
D.3.2 Modified pressure to include intermittent wetting 4
D.3.3 Time history of pressure 4
D.3.4 Pressure Range 5
D.3.5 Notes 5
D.3 Analysis of Local Pressure 6
D.3.1 Time domain analysis 6
D.3.2 Frequency domain and statistical analysis 6
D.3.3 Statistical distribution 7
D.4 Pressure Model for FE Analysis 7
D.4.1 Linearisation of pressure loading 8
D.4.2 Simplified (cylindrical) correction 9
D.4.3 Extent of intermittent wetting foot print 10
D.9 Internal fluid pressure in tanks 12
D.9.1 Model 6
D.9.2 Implementation 6
D.9.3 Notes 7

Page D-2
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

D.1 Introduction

In linear diffraction-radiation analysis, hydrodynamic pressure is obtained over the wetted


surface at rest as a harmonic pressure variation at each point. Therefore, the pressure on the side
shell between the surface at rest and a wave crest is not modelled and, in a wave through an
unrealistic negative pressure is generated in an area below surface at rest.
This intermittent wetting is a non-linear effect. In lieu of a more exact model (currently not
available), intermittent wetting is taken into account as a correction of the linear pressure field.
The correction is taken as local and is assumed not to affect significantly vessel motions and the
resulting first order pressure field in the permanently submerged part. It will however result in a
more accurate evaluation of the structural response in the area affected by intermittent wetting.

The pressure model to account for the intermittent wetting is first presented.
An outline of the methods for the analysis of local pressure and the resulting stresses and damage
is given, and the methods for the implementation of this pressure model into a 3D FE based
spectral analysis are described.
An overview of the problem of intermittent wetting and the background to the present
recommendations is given in /30/ and /28/.

D.2 Pressure Model


D.2.1 Linear theory (3D 1st order diffraction-radiation) pressure
The linear pressure model based on 1st order diffraction-radiation method (linear potential
theory) that forms the basis of the pressure model in the intermittent wetting area is first briefly
summarised.

According to the linear potential theory, the linear hydrodynamic pressure, at any point on vessel
hull below the waterline at rest is composed of two parts:
• A purely hydrodynamic part, associated with the time derivative of the velocity potential Φ,
that can be decomposed into three components :
- incident wave,
- diffraction by vessel hull (at rest)
- radiation components resulting from vessel motions
• hydrostatic variation due to ship motions
The two later term being obtained after solving of the equation of motions.

Notes:
a) pressures are obtained from the analysis as the pressure p1 (x,y,z,ω), for a wave of unit
amplitude with a circular frequency ω.
The pressure in waves of finite amplitude A is then given by:
p( x, y, z, t ) = A ⋅ ℜe {p1 ( x , y, z, ω) ⋅ e − iωt }

Page D-1
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

The pressure in irregular waves will be then obtained by summation over all wave
components, taking respective phases into account.

b) The pressure in meters of water height at the altitude z is written:


~
p z = p( x, y, z, t ) / ρg

c) The relative wave elevation (RWE) is equal to ( ~


p 0 ), i.e. the pressure ~
p z at z = 0.

D.2.2 Modified pressure to include intermittent wetting


The above pressure is modified as follows around the waterline, in order to represent intermittent
wetting, based on the total (hydrodynamic plus hydrostatic) pressure:

• In the wave crest, a hydrostatic pressure variation is assumed between instant water surface
(z= ~
p 0 > 0) and mean waterline (see Figure D-1).
• In the wave through, total pressure is set to zero when the sum of hydrodynamic pressure and
hydrostatic pressure is negative (see Figure D-2).

The total pressure is thus:

• For a positive relative wave elevation ~p 0 > 0 :


~ ⎧⎪~ p0 − z for 0 < z < Z + = ~p0
P ( z, t ) = ⎨~ z
⎪⎩ p − z for z < 0
Note: z positive upwards.

• For a negative relative wave elevation ~p 0 < 0 ,:


~ ⎧⎪0 for 0 > z > Z −
P ( z , t ) = ⎨~ z
⎪⎩ p − z for z < Z −

where Z − is given by the equation z = ~


p z (see note §D.3.2 below)

D.2.3 Time history of pressure


The time history of pressure, at different elevation z in a given section, is illustrated in Figure D-
3 below, for regular waves.
In the lower region (permanently wetted), the sea pressure is equal to the linear pressure.

Page D-2
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

Figure D-1 Pressure distribution for ~


p 0 > 0 (Wave crest)

Figure D-2 Pressure distribution for ~


p 0 < 0 (Wave through)

Page D-3
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

Pressure History

pa

Wave Profile ‘

Figure D-3 time history of pressure


D.2.4 Pressure Range

In regular waves, the range of pressure at different elevation z in a given section is :


(see also Figure D-4).
⎧~p0 − z for 0 < z < Z max
~

⎪ ~z ⎧Z max = ~p0

∆P(z, t ) = ⎨ p − z for Z min < z < 0 with ⎨
~z
⎪ ~z ⎪⎩Z min given by − z + p = 0
⎪⎩2 ⋅ p for z < Z min

where Zmin and Zmax define the foot-print of the area that is affected by the effect of intermittent
wetting.

Z max
Z max Zmax Zmax

z=0
Z min RANGE Maximum
~
total pressure
( p o > 0) 2 ⋅ Z min
Zmin Zmin

Hydrostatic pressure

Minimum total pressure


~
( p o < 0)

(a ) (b) (c)
Figure D-4 Total Pressure (a), pressure range (b) and mean pressure (c)
Page D-4
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

Besides, for the purpose of frequency domain analysis, and subsequent statistical analysis, the
actual pressure signal in the intermittent wetting area (as shown in Figure D-3) is approximated
by an harmonic variation with same pressure range (pa) (see Figure D-5).

pa

Figure D-5 Modified (Harmonic) pressure signal

In a given section, this pressure variation is in phase with ~p z i.e. almost in phase with ~
p 0 : the
relative wave elevation.
As shown on Figure D-4c, the mean pressure includes an additional term upon hydrostatic
pressure. This term will be generally omitted for fatigue analysis where mean load is not
described.

D.2.5 Notes
1) the pressure ~
z
p (and similarly the pressure range) can be also written as
~
p = ~p ⋅ f ( z ) where f(z) is a decay function.
z 0

In analytical work, this decay have been often taken as the same (exponential) decay as
the incident wave : this assumption is not always valid (see /30/), and is not necessary
when the actual pressure field is available from the results of linear sea-keeping
analysis.

2) The lower bound of the intermittent wetting area Zmin is given by an implicit equation, and is
thus depending on wave frequency. Taking Zmin = - Zmax , or a constant value, is often
considered, but result in a unrealistic pressure jump in the area between these two points,
with a discontinuity at the -Zmax level (see e.g. fig. 3 in /27/)

D.3 Analysis of Local Pressure and resulting stresses and damage


D.3.1 General
From the pressure model as defined above, the bending stresses in side shell longitudinals
induced by local pressure (i.e. stresses due to bending over each individual stiffener span) and
the resulting fatigue damage can be obtained in several different ways. However such derivation
will generally omit the dependence of stresses on wavelength, thus frequency, for shortest
waves.
Furthermore, these local pressure stiffener bending stresses should be combined with :
• the axial stresses due to hull girder bending,

Page D-5
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

• the additional stiffener bending stresses due to the relative displacements of stiffeners
support points, induced by global deformations of vessel cargo hold framing (having a
prevailing contribution for e.g. stiffener’s connections at transverse bulkheads).

This will generally assume in addition that the stresses induced by global hold deformation are
not dependent on intermittent wetting. This assumption is valid for FPSO in mild conditions,
where the footprint of IWE is a few stiffener spacing, but becomes inaccurate in harsher
conditions.

D.3.1 Time domain analysis


A time domain implementation of equations in §D.3.4 is proposed in /27/, consisting in
following steps:
• generating a time domain signal of wave elevation and then the corresponding pressure
signal,
• creating a modified pressure signal following equations in §D3.4,
• computing the local pressure stiffener bending stresses, and combining with other stresses,
• processing of the modified signal, by Rainflow counting.

This combination can be performed provided separate RAO’s for each contribution are available.
The remarks above are thus applicable to this approach.

D.3.3 Frequency domain and statistical analysis

The pressure variation at any point, as specified in §D.2.4 is harmonic, and, for the purpose of
stiffeners local strength evaluation, can be treated by the spectral method, from which short term
or long term statistics of pressure local stress response can be made, and the resulting damage
evaluated.
The derivation of short-term local stress response, leading to closed form expressions of the
damage, has been proposed in refs. /26/ and /27/, based however on the simplifications quoted in
D2.5 above.
In these approaches, the combination of the local effect of pressure with stresses induced by the
hull girder bending moments is made analytically. The combination with global hold
deformations is not addressed.
Thus these models are more relevant for Rule based evaluations than in the context of a full FE
based spectral analysis.

The equations in §D.3.4 can be also used also to derive the long-term distribution of local stress
ranges from the pressure range calculated by linear analysis, as illustrated in Figure D-6 below.
The resulting fatigue damage can be obtained by e.g. numerical integration, with however same
limitations as above.

Page D-6
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

pressure range (linear)

p( z ) ; z=0

p( z ) ; z positive

p( z ) ; z negative
Pressure range

|Z|

Log probability of exceedance

Figure D-6 long-term distribution of pressure range

Page D-7
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

D.4 Pressure Model for FE Analysis


By FE analysis, the combined effect, including correct phasing, of all contributions to cyclic
stresses can be modelled. However, the effect of intermittent wetting, combining variable
intensity and footprint, is a non-linear effect.
The most consistent way to model the pressure field as described in D2.2 would be a time
domain simulation in irregular waves. Alternatively, in regular waves, the pressure history
including the intermittent wetting effect could be described by as a periodic (but non-harmonic)
loading. Such methods are however not commonly available.
For spectral analysis based on a 3D model, a linearisation has to be made.

D.4.1 Linearisation of pressure loading


The pressure variation at any point as defined in D.2.4 is harmonic, but not proportional to the
wave height.
The linearisation required for spectral analysis is achieved by considering the response of the
structure in regular waves of finite height (i.e. not unitary), and defining the RAO of stresses as:
SA SR
RAO = =
A H
with:
SA = amplitude of stress variation at a point of interest,
A = wave (single) amplitude,
SR = stress range,
H = wave height.

This is actually equivalent to fixing the upper and lower bounds Zmax and Zmin of the intermittent
wetting area to those in regular waves of height H. The amplitude of pressure at any point along
hull is then a harmonic signal proportional to selected wave height H*, with a range:
⎧H *
⎪ 2 [~p 0
−z ] for 0 < z < Z max

~ ⎪H *
∆P(z) = ⎨ [~p z
− z] for Z min < z < 0
⎪ 2
⎪H *. ~
pz for z < Zmin


H * ~0 H * ~z
with Zmax = p , and Zmin given by −z = p
2 2

1) The modified pressure range can also be written as the sum of the 3D hydrodynamic pressure
and an “intermittent wetting correction”(see Figure D-7 below for illustration), such that:
~
∆P = ∆p lin + ∆p corr
The pressure range of the intermittent wetting correction is then:

Page D-8
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

⎧~ p0 − z for 0 < z < Zmax




∆p corr = ⎨− ~ pz − z for Zmin < z < 0

⎪⎩ 0 for z < Z min

Note:
This loading is in phase with ~
p z i.e almost in phase with ~
p 0 : the RWE in the section under
consideration.

z z z

Z max
Free
surface 1/2
RWE
-1/2
RWE
Z min
Wave
pressure

Linear pressure Wave induced Intermittent Wetting


amplitude (A) pressure variation (B) Correction
(B-A)

Figure D-7 Pressure field and “intermittent wetting correction”

D.4.2 Simplified (cylindrical) intermittent wetting correction


1) If the wavelength is large versus the length-wise dimension of the area under study (e.g. the
length of a cargo tank), the structural response will be almost in phase with wave crest and
trough passing over the area, i.e. with RWE.

Around these instants, the intermittent wetting correction will be almost constant over the area
and can be modelled by a cylindrical pressure distribution with following properties:
• Pressure variation is harmonic,
• The intensity of pressure variation follows same diagram as in §D.5 above, in the vertical
direction, but is identical in all sections, with a constant phase,
This modelling is approximate but can be solved as a single unitary loading, provided Zmax and
Zmin are taken the same for all frequencies.
The correction on stresses resulting from the “intermittent wetting correction” can be used to
correct RAO of stresses, considering an intensity of the correction proportional to RWE at each
frequency:
RAO corr = RAO + RWE ⋅ σ iwc
Where σ iwc is the stress resulting from this loading, with the appropriate sign (most often, the
correction is a reduction).

Page D-9
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

However, in this approach, the intermittent wetting correction tends to be overestimated for short
waves, as the length-wise variations of pressure are not included. This can be avoided by
evaluating a decay of the stress correction, based on wavelength.

2) Note
In Rules, pressure diagrams are defined for a crest and a trough situation of the area of interest.
(see e.g. references /3/ and /25/ )
The resulting pressure range diagram, as illustrated in Figure D-8, thus includes the effect of
intermittent wetting.
As this Rule-based pressure diagram is usually modelled as cylindrical, there is consistency
between the above “cylindrical correction” and the Rule-based pressure diagram.

Rule – pressure ½ range

0.5 S

0.75 S

~ 0.8 S

0.5 S

Figure D-8 Rule-based Pressure Diagram

3) Simplified (cylindrical) correction (Small Foot Print)


For mild environment, the relative wave elevation is small, e.g. in the range of a few stiffener
spacing.
The approach presented above can be further simplified into a quasi-static approach /26/, /29/
neglecting the decay of pressure with depth, over the considered area.
The foot print of the additional pressure application is included between fixed boundaries, Zmax
and Zmin= - Zmax .

Page D-10
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

z z z z
Zmax Zmax
Free
surface
½ RWE - ½ RWE

Wave -Zmax
pressure Zmin
2a

Linear pressure Wave induced pressure Range of corrected Intermittent wetting


(A) variation (B) pressure Correction (B-A)

Figure D-9 Simplified “cylindrical” correction, small foot print

D.4.3 Extent of intermittent wetting footprint


In regular waves of finite amplitude, the positions Zmax and Zmin defining the intermittent wetting
footprint are depending on the (finite) wave amplitude selected for analysis.
Zmax is equal to the relative wave elevation RWE at the considered location.

1) For a short term analysis on a given sea state, selecting Zmax as the significant RWE, i.e. H as
the significant height Hs of the sea state, will provide a representative profile of the pressure
around waterline for the evaluation of global cargo-hold deformation and of around-waterline
stiffener bending. /30/
Zmin is then obtained as defined in D.4.1, but Zmax and Zmin will both depend on wave frequency
and on Hs. A practical simplification will be to consider a single Hs.

2) For the application of a cylindrical correction, it will be practical to consider a single value of
Zmax, i.e. of the RWE :
Zmax, may be taken from the long term distribution of the relative wave elevation, as the value
with a probability of exceedance in the range of 10-3 to 10-5 .
The same assumption is implicit in the Rule-based pressure diagram of the simplified method.

Then, Zmin can be selected based on Zmax and the pressure decay at the frequency for which RWE
is maximum.
Z
In Rule diagrams, the ratio min is taken as 0.75.
Z max
3) Note
The effect of intermittent wetting correction depends on the selected footprint.
Its effect is most often a reduction of stresses with respect to those given by the First order
Diffraction-Radiation. Then, selecting a large footprint might be non-conservative.

It is also worth noting that large correction might go beyond the scope of a spectral fatigue
analysis.

Page D-11
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

D.9 Internal fluid pressure in tanks


D.9.1 Model
A quasi-static approximation is used, under the assumption of small movements of the vessel,
where the pressure is defined based on the intensity and angular variations of the total
acceleration (including gravity).
In this approach, eventual resonant motion or sloshing effect is neglected.
This approach has been found verified by the full scale measurements of FPSO Integrity, when
the natural period of liquid in tank is small.

The total (static + dynamic) pressure at a point P in a tank is given by:


→ →
Ptot = p o − ρ Γ ∗ OP
M
where:

O is a point where the (total) pressure at the free surface is assumed invariant and equal to
po (see below).
* stand for scalar product,
r
ΓM is the (translational) acceleration of a reference point M (see below), taking into account
the angular variation of gravity direction versus the vessel axis.
r
With the assumption of small movements of the vessel, the components of ΓM , in vessel axis,
are :
ΓMX = (− g ψ + γ X )
ΓMY = ( g φ + γY )
ΓMZ = (g + γ Z )
with:
γ x , γ y , γ z : accelerations of tank's reference point M, obtained from the motion analysis,
φ, ψ : roll and pitch angle.

D.9.2 Implementation
The total pressure loading can be split in elementary loading cases : one for the still water
loading (that will be generally omitted), and one for each of the three components of the dynamic
acceleration of point M, as illustrated on Figure D-10.

a) When the tank is full, (as on Figure D-10)


• Dynamic pressure is applied to all tank boundaries, including tank top, thus assuming a
sufficient setting pressure p o ,
• O may be taken as the centre of tank top surface (the variation of the position of O, from
one end to the other, that is included in Ship Rules tank pressures, is not considered as
relevant here; this is in addition a non-linear effect),
• M is the Centre of Volume of the tank.
b) In case of partial filling :
Page D-12
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

• Dynamic pressure is applied to tank boundaries, up to the level of liquid surface at rest
(linear approximation),
• O is the centre of the free-surface,
• M is then the Metacentric point of the Volume of liquid. (the transverse metacentric point
may be used).
v
In both cases, the integration of pressure over tank boundaries gives F = ρ ⋅ V ⋅ ΓM , following the
law of Archimedes, and the line of action of F will be passing by the point M.
Then, the resulting inertia load will be equivalent to that of a concentrated mass at point M.
This will be consistent with the mass model, where tanks is described by a mass and the Centre
of Volume of the tank and a “free surface correction” of the vertical position of the Centre of
Gravity of the vessel is made for partially filled tank (as in Stability calculations).

D.9.3 Notes
In this model, the accelerations γ x , γ y , γ z in M are combining both translational and rotational
accelerations of vessel cog (or motion reference point), but the effect of the accelerations
induced by rotational motions around point M are not considered.
In the case of a tank with substantial internal partitioning (as usual side hull ballast tank), it
would be more appropriate to apply this model to a series of sub-set of the tank.
Then pressures will appear on internal walls that will better distribute the total inertia load.
There will be however a difficulty if openings in these walls are modelled.

2L 1
2L 2

H Z
Height of
liquid
X
Y

Page D-13
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

γX γy γ
‘ z
(in Z-X plane) (in Z-Yplane) (in Z-X plane or Z-Y plane)
Z Z Z

X Y Y

L1 L2
H H

ρ.L1 ρ.L2 ρ.H

Figure D-10 Pressure diagrams in tanks

- o0o -

Page D-14
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

APPENDIX
E
DAMAGE (CLOSED FORM)

Page E-1
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

Closed form expressions of damage considering a Rayleigh or Weibull distribution for the stress
range are given below for two cases: one slope and two slopes S-N curve:

E.1 Damage considering a single slope S-N curve

a) “Rayleigh “ distribution
Dref ⋅ ν Z ⋅ 2( 2 Sd )
m
⋅ Γ(1 +
m
2
)
D =
a

with:
Dref Reference duration
νZ Average zero-crossing frequency of the stress response
Sd Standard deviation of the stress process (RMS)
m, a : S-N curve parameters of the single slope S-N curve
Γ : Gamma function

b) “Weibull “ distribution
m
NT ⋅ q ⋅ Γ( 1 +
m
)
D = h
a

with:
NT Total number of stress cycles in the reference duration Dref (5.4.5)
h Weibull shape parameter of the stress distribution
∆σ 0
q Weibull scale parameter of the stress distribution given by: q =
p0 ) (− ln 1/ h

∆σ 0 Stress range value at the reference probability of exceedance po


m, a : S-N curve parameters of the single slope S-N curve
Γ : Gamma function

E.2 Damage considering a two slopes S-N curve

a) “Rayleigh “ distribution
(
Dref ⋅ ν Z ⋅ 2 2 Sd )
m1
⋅ µ ⋅ Γ(1 +
m1
2
)
D =
a1

with:
Dref Reference duration
νZ Average zero-crossing frequency of the stress response

Page E-2
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

Sd Standard deviation of the stress process (RMS)


m1 , a 1 : S-N curve parameters above change of slope
m2 , a 2 : S-N curve parameters below change of slope
µ Coefficient taking into accounts the change in slope of the S-N curve;

⎧ ⎛ m1 ⎞ ⎛a ⎛ m ⎞⎞ ⎫
⎨ γ ⎜1 + ; κ ⎟ −⎜⎜ 1 ⋅ (2 ⋅ 2 ⋅ Sd ) m2 − m1 ⋅ γ ⎜1 + 2 ; κ ⎟ ⎟⎟ ⎬
⎝ 2 ⎠ ⎝ a2 ⎝ 2 ⎠⎠ ⎭
µ = 1 .0 − ⎩
m
Γ ( 1+ 1 )
2
2
⎛ SQ ⎞
κ =⎜ ⎟
⎜ 2 ⋅ 2 ⋅ Sd ⎟
⎝ ⎠
SQ : Stress range at change of slope on the S-N curve
Γ: Gamma function
γ: Incomplete Gamma function as given in table 4-V /25/ or by the following
x
−t
formula: γ (a + 1 , x ) = ∫ t ⋅ e
a
dt
0

b) “Weibull “ distribution
m1
NT ⋅ q ⋅ µ ⋅ Γ( 1 +
m
)
h
D =
a1

with:
NT Total number of stress cycles in the reference duration Dref (5.4.5)
h Weibull shape parameter of the stress distribution
∆σ 0
q Weibull scale parameter of the stress distribution given by: q =
(− ln
p0 )
1/ h

∆σ 0 Stress range value at the reference probability of exceedance po


m1 , a 1 : S-N curve parameters above change of slope
m2 , a 2 : S-N curve parameters below change of slope
µ Coefficient taking into account the change in slope of the S-N curve;

⎧ ⎛ m1 ⎞ ⎛ a1 ⎛ m ⎞⎞ ⎫
⎨ γ ⎜1 + ; κ ⎟ − ⎜⎜ ⋅ (q ) ⋅ γ ⎜1 + 2 ; κ
m1 − m2
⎟ ⎟⎟ ⎬
⎝ h ⎠ ⎝ a2 ⎝ h ⎠⎠ ⎭
µ = 1.0 − ⎩
m
Γ (1 + 1 )
h
h
⎛ SQ ⎞
κ = ⎜⎜ ⎟⎟
⎝ q ⎠

SQ : Stress range at change of slope on the S-N curve

Page E-3
Report No. 2003-0582, rev. 02
DET NORSKE VERITAS/BUREAU VERITAS

Γ: Gamma function
γ: Incomplete Gamma function (see above)

- o0o -

Page E-4
Report No. 2003-0582, rev. 02

Вам также может понравиться