• / Marine Structures 8 (1995) 359-383

Elsevier ScienceLirnitvd
Printed inGreat Britain.
0951-8339/95/$9.50
ELSEVIER 0951-8339(94)00026-3

Fatigue Assessment of Ship Structures

Espen H. Cramer,* Robert Loseth & Kjell Olaisen

Det Norske Veritas ClassificationAS, 1322 Havik, Norway
*Det Norske Veritas Industry AS, 1322 Hovik, Norway
(Received 7 December 1994)

ABSTRACT

T~is paper summarizes the procedures and recommendations given in the

recently developed guideline from Det Norske Veritas on fatigue assess-
ment of ship structures. The procedures defined are included in the D N V
computational package Nauticus Hull.
The accumulated fatigue damage under time varying loading is estimated
based on the S - N fatigue approach assuming linear cumulative damage.
Master S - N curves have been established for welded and unweided details
in corrosive and non-corrosive environments with the basis in a smooth
geometry. Hence, the S - N curves are defined for the local notch stresses,
being the nominal stresses combined with the stress concentration factors
for the local geometry. The stress concentration factors are derived either
from local finite element stress analysis or through empirical expressions.
The long term stress range distribution acting on a detail is defined from
either pre-defined distributions with reference stress levels from empirical
formulas in the D N V Rules or through long term frequency response
analyses. For the accurate assessment of the stress response in the hull
structure, recommendations are given for the application of simple beam
models and the use of finite element methods for the global hull analysis,
the analysis of frames and girders and for the local stress analysis.
A closed form approach for estimating the probability of fatigue failure
of a detail within the design life of the ship is further defined.

Key words." ships, fatigue, stresses, loads.

359
360 E. H. Cramer, R. Loseth, K. Olaisen

1 INTRODUCTION
1.1 Background

A growing number of fatigue incidents in local structural details, espe-

cially when using HTS steel, have demonstrated that a more direct control
of fatigue is required. DNV introduced the fatigue requirements in the
Rules for Classification of Ships in 1991, 3 and has recently developed a
guideline containing recommended procedures for simplified and direct
fatigue analysis of ship structures. ~ The guideline addresses the fatigue
assessment of ship structures in particular, compared to the DNV Classi-
fication Note for fatigue strength analysis of mobile offshore units from
1984.4 This article presents the main content of this guideline and discus-
ses some of the background for the procedures suggested.
The developed guideline contains practical procedures for fatigue
assessment of ship structures that;
- - can be applied for different ship types and structural members
-- allow the designers flexibility in the choice of design solutions
- - account for the level of workmanship and corrosion protection
- - provide consistency in fatigue prediction for different analysis
techniques
- - provide results that agree with experience from ships in operation.

1.2 Overview
The aim of the fatigue assessment is to ensure that all parts of the hull
structure subjected to dynamic loading have an adequate fatigue life. To
ensure that the structure will fulfil its intended function, a fatigue assess-
ment should be carried out for each individual type of structural detail
which is subjected to extensive dynamic loading. It should be noted that
every welded joint and attachment, or other form for stress concentration,
is potentially a source of fatigue cracking.
The calculated fatigue life provides a basis for the structural design
(steel selection, scantlings and modelling of local details). Furthermore, it
can form the basis for an efficient inspection program during fabrication
and throughout the life of the structure.
In the following, the fatigue design procedure is described using the S-N
data defined for the local notch stress, allowing for a continuous evalua-
tion of the fatigue capacity based on incremental changes in the local
geometry. Credit is given for higher standards of workmanship and the use
of higher quality coating systems.
The long term distribution of the stress range response, being a funda-
mental requirement for the fatigue analysis, is determined in various ways.
 Fatigue assessment of ship structures 361

The simplest way is to estimate the reference stress range level at a defined
occurrence rate from empirical rule formulas for the wave induced loading
combined with empirical expressions for the shape of the stress range
distribution. The more accurate way of estimating the stress range distri-
bution is to carry out a direct long term frequency response analysis.
From the wave induced loading, the nominal stresses are derived
applying either specified empirical and analytical formulas, 2- or 3-
dimensional beam element models, or for more accurate estimation, finite
element analyses. The local stress response at the considered hot-spot is
obtained by applying stress concentration factors, defined through
empirical expressions or through the use of local finite element analyses.

2 FATIGUE ANALYSIS

2.1 General

The fatigue lives are calculated based on the S-N fatigue approach under
the assumption of linear cumulative damage, Palmgren-Miner rule. The
S-N data are defined for smooth geometries, implying that the notch
stress range, being the maximum principal stress range adjacent to the
potential crack location including stress concentrations, are applied in the
fatigue assessment. The stress concentration is due to the gross shape of
the structure and the local geometry of the weld.
For fatigue analyses of regions in the base material not significantly
affected by residual stresses due to welding, the effective stress range for
the fatigue analysis may be reduced depending on the mean cycling stress.

2.2 S--N curves

The S-N curves recommended for the fatigue assessment are obtained
from experience and fatigue tests. 5 The S-N curves are to be applied
together with the notch stress, the local stress at the weld toe due to
structural discontinuities (hot-spot stress) and the weld geometry. Differ-
ent S-N curves are defined for welded joints and base material in air/
cathodic protected environments and for corrosive environments. The
recommended S-N curves to be applied with the derived notch stresses are
given in Table 1. Special considerations should be made for the plate
thicknesses larger than 22 mm.
The design S-N curves are based on the mean-minus-two-standard-
deviation curves for relevant experimental data, and are thus associated
362 E. H. C r a m e r , R. L o s e t h , K. Olaisen

with a 97.7% probability of survival. The basic design S-N curve is given
as,

log N = log ~ - m. log Atr

where N is the predicted number of cycles to fatigue failure for the notch
stress range Aa.

2.3 Cumulative damage

The cumulative fatigue damage is derived from the local stress range
response over the design life of the ship, having an acceptable usage factor
for the fatigue damage capacity equal to or less than one. The design life
for ship structures is normally not taken to be less than 20 years.
The expression for the accumulated fatigue damage depends on the
formulation of the long term stress range response. When the local long
term stress range response is defined through histograms with k stress
blocks with n; stress cycles in stress block i, the cumulative damage is
defined as,
D = I - - . ~-~ ?li • m f f im
a
i=1
Defining the long term stress range distribution through a Weibull
distribution with scale and shape parameter q and h, respectively, the
cumulative damage is expressed as,
D =--ndq'a • P(1 + h )

where nd is the number of stress cycles over the design life, estimated from
the product of the design life of the ship, the sailing rate and the average
occurrence rate of the stress response. As a first approximation, the aver-
age occurrence rate per second on ship structures may be taken as
v0 = (4. lOgl0 L) -1, where L is the ship rule length in m.

TABLE 1
S-N parameters, DNV l

S-N Curve Material Environment N ~ 10 7 N > 107

log~ m log~ m

Type-I Welded joint Air-cathodic prt. 12.65 3.0 16-42 5.0

Type-II Welded joint Corrosive 12.38 3.0 12-38 3-0
Type-III Base material Air-cathodic prt. 12.89 3.0 16.81 5.0
Type-IV Base material Corrosive 12.62 3.0 12.62 3.0
 Fatigue assessment o f ship structures 363

The cumulative fatigue damage can also be estimated from the accu-
mulated damage within each short term condition. The cumulative
damage is then the weighted sum over all the sea states and wave heading
directions the ship is encountering over the design life, weighted with the
occurrence rate r of the different sea states and heading directions,
all seastates
all headings

a
(m)
D = n~. 1" 1 + ~- • E r U • (2 ~ . ) "
i=l,j=l

where m0 is variance of the stress response process.

The long term stress response is usually derived for specified loading
conditions. The total cumulative damage is the weighted sum of the
damage contributions within the different loading conditions, weighted
w.r.t, the fraction of the sailing time within the different loading condi-
tions. To account for the durability limitations of the corrosive protection
system, the design life may be divided into two separate time intervals,
applying non-corrosive and corrosive S-N data.
The above expressions for cumulative damage are given for one-slope
S-N curves. Applying the two-slope S-N curves, equivalent expressions
on a slightly more complex form are valid.

3 SIMPLIFIED STRESS ANALYSIS

3.1 General

The simplified approach for estimating the long term stress range distri-
bution is based on the assumption of Weibull distributed stress ranges.
The approach defines simplified formulas for estimating the individual
stress response components, the combination of global and local stress
response components and the modelling of the shape of the stress response
distribution.
The wave induced dynamic loading is estimated from empirical rule
formulations or dynamic pressure load analyses. The corresponding stress
response is derived applying empirical and analytical expressions, or for
more accurate estimation, the use of frame analyses or finite element
analyses.

3.2 Long term stress distribution

The long term distribution of stress ranges can be shown to be well

described through a Weibull distribution, having cumulative probability,
364 E. H. Cramer, R. Loseth, K. Olaisen

F(Aa) = 1 - exp -

The shape parameter depends on the prismatic parameters of the ship,

the location of the considered detail and the sailing route over the design
life. In lieu of more accurate calculations, the shape parameter may be
taken as, 2 (all units in m)

h = h0 for deck longitudinals

h = ho + ha(D - z ) / ( D - Tact) for ship side above the water line
h = ho + haz/Zact - O'O05(Tact - z) for ship side below the water line

h = h0 - 0"005Tact for bottom longitudinals

h = h0 + ha for longitudinal and transverse
bulkheads

where the basic shape parameter ho is defined as a function of the ship rule
length L,
h0 --- 2-21 - 0-54 x lOgl0(L)
ha is the additional factor depending on the motion response period, equal
to 0.05 in general and 0.0 for plating subjected to roll motions for vessels
with roll period over 14 seconds. D is the moulded depth of the ship, Tact
is the actual draught and z is the location height above the keel.
The scale parameter is derived from the shape parameter and a refer-
ence stress response, Aa0, exceeded once out of the corresponding refer-
ence number of stress cycles, no, defined as the no -1 probability level.
Atr 0
q -- (In no) 1/h

For ship structures, the main contribution to the cumulative fatigue

damage is caused by waves of moderate size. The number of stress cycles
for which the reference stress response is derived should therefore not be
too high. In order to avoid a strong influence on the fatigue damage by
the estimated shape parameter, the number of stress cycles for which the
reference stress response is derived should not be higher than 104.
In the evaluation of the dynamic stress response, both global-, Aag
(primary) and local-, Atrt (secondary and tertiary) dynamic stress compo-
nents should be considered. The global stress components include wave
induced vertical and horizontal hull girder bending stresses, and for
vessels with large deck openings also wave induced torsional stresses.
 Fatigue assessment o f ship structures 365

The local stress components include stress responses due to dynamic

external sea pressure and dynamic internal inertia pressure loads, being
secondary stress responses from bending of girder systems and stiffeners
between girder supports, and tertiary stress responses from bending of
unstiffened plate elements between longitudinals and transverse frames.

3.3 Combination of stresses

For each loading condition, the local dynamic stress components due to
internal and external pressure loads are to be combined with the global
stress components induced by hull girder wave bending. The stress
components to be combined are the notch stresses.
If a combined long term stress response analysis is not carried out, the
combined stress range response from the combined global and local stress
range responses is the largest of: 2
AO'g -1- 0-6. Aat
A a = f e "f,n" max 0.6. Atrg + Atrl

where fe is the operation route reduction factor and f,n is the mean stress
reduction factor. A reduction in the effective estimated stress response is
achieved for vessels that for longer periods operate in environments not as
harsh as the North Atlantic. For world wide trade, the reduction factor
may be taken as 0.8.
The combined global stress range response is estimated from the wave
induced vertical hull girder bending stresses, Aav, and the horizontal hull
girder bending (and torsion as relevant) stresses, Aah,
Atrg = V / A a { + A a ~ + 2 . Pvh " mtTh " tTv
where the long term correlation of Pvh is defined as 0.10.
The combined local stress response is estimated assuming an average long
terrn correlation between the external sea pressure induced stress amplitude,
try, and the internal inertia pressure induced stress amplitude, ai,
at = 2 . V/a2e + tr 2 + 2 . Pei " fie" tri
The correlation between the external and internal pressures induced
stresses is a function of the location of the considered detail,

0-5 _ 0"2-~a~t+0"6
z
02 lyl
I~ . . . .
4 Ixl.z
O" L . Tac t
z~Tact

Pei =
0 . 3 + 0 . 2 I--~--
'"~ 0.2 1]Y~
"" z> Tact
L, /~
366 E. H. Cramer, R. Loseth, K. Olaisen

where B is the greatest moulded breadth of the ship, and x, y, z are the
longitudinal, transverse and vertical distance from the origin at (midship,
centreline, baseline) to the load point of the considered structural detail.

3.4 Calculation of stress components

The global and local stress components are derived from the wave induced
bending moments and the external and internal wave induced pressure
loads.
The global stress components are based on gross (as build) scantlings,
whereas the local stress components are based on net scantlings, i.e. gross
scantlings minus corrosion addition. The corrosion addition depends both
on the type of tank and the location of the considered detail within the
tank. Each of the individual stress components should be estimated
accounting for the relevant stress concentration factors.
The stress contributions are estimated applying simple analytical and
empirical expressions, accounting for the effective span of longitudinal/
stiffeners, the effective breadth of plate flanges and the relative deflection
between transverse bulkheads and adjacent web frames, l

4 S I M P L I F I E D C A L C U L A T I O N S OF LOADS

The linear dynamic load components for which the individual stress
contributions are estimated are calculated from empirical rule expressions
as defined in the DNV Rules. 3 The load components considered are the
global wave induced bending moments the external sea pressure acting on
the hull and the internal inertia pressure acting on the tank boundaries.
The rule expressions are adjusted for an excess probability of 10-4 per
wave cycle.
The fatigue damage should in general be calculated for all representa-
tive load conditions combined with the expected operation time within
each of the considered conditions.
In the calculation of the effective dynamic external pressure at the ship
side, the effective pressure amplitude will be reduced due to intermittent
wet and dry surface in the surface zone.
The dynamic internal inertia pressure loads should be calculated based
on the combined acceleration in longitudinal, transverse and vertical
direction. As an approximation, however, the inertia pressure can be esti-
mated as the maximum inertia induced pressure in the longitudinal,
transverse or vertical direction.
 Fatigue assessment of ship structures 367

For vessels that only occasionally operate with partly filled tanks, the
effect of liquid sloshing pressures may be neglected in fatigue computations.

5 WAVE L O A D I N G BY DIRECT COMPUTATIONS

5.1 General

The long term load (or stress) response distribution can be directly derived
applying a long term linear frequency response analysis, accounting for
the ,different sea states the ship will encounter over the design life. The
loads computed by direct computations may substitute the simplified load
components, and can, combined with a strength model, substitute the
simplified expressions for the individual stress components, or the
combined stress response. If a full long term frequency response analysis is
carried out, the long term distribution of stress ranges can be calculated
directly for each load condition.
A linear modelling of the ship response is, in general, sufficient for
fatigue assessment purposes as most of the fatigue damage is related to
moderate wave heights. The fatigue damage should be calculated for all
the main load conditions, applying the cruising speed within each load
conditions in the response modelling.

5.2 The long term distribution

The long term distribution of stresses for fatigue analyses may be esti-
mated using the estimated long term frequency of occurrence of different
sea states defined in Table 2 or 3, where each sea state is described through
a significant wave height Hs and zero crossing period Tz. Table 2 repre-
sents, the North Atlantic wave climate (Marsden squares 8, 9 and 15)6 and
Table 3 the wave climate for world wide operation. The world wide sailing
route will normally apply, however, for shuttle tankers and vessels that
will frequently sail on the North Atlantic or in other harsh environments,
the wave data for the North Atlantic should be chosen.
The distribution of wave energy for different frequencies within each sea
state is defined through the Pierson-Moskowitz spectrum. The response
spectrum is defined from the modulus square of the transfer function.
The long term distribution of the response is established as the weighted
sum of the individual short term response distributions over all the sea
states and heading directions, weighted with the relative occurrence rate of
response cycles. The long term distribution is fitted to a Weibull distribu-
tion, defining the distribution parameters, q and h.
368 E. H. Cramer, R. Lsseth, K. Olaisen

TABLE 2
Scatter Diagram for the North-Atlantic for use in Fatigue Computations

0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 IO.5 11.5

13.50 000001100000
12.50 0 0 1 2 2 2 2 1 1 1 0 0
11.50 0146754221 10
10.50 0 4 14 21 19 13 8 4 3 2 1 1
9.50 1 16 43 48 34 20 11 5 3 1 1 0
8.50 4 47 80 65 38 19 8 4 2 1 0 0
7.50 13 75 79 46 21 9 3 1 1 0 0 0
6.50 21 54 33 14 5 2 1 0 0 0 0 0
5.50 14 13 5 1 0 0 0 0 0 0 0 0
4.50 300000000000
3.50 000000000000
2.50 000000000000

TABLE 3
Scatter Diagram for World Wide Trade for use in Fatigue Computations

0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 IO.5 Il.5

13.50 000000000000
12.50 0 0 1 1 1 1 0 0 0 0 0 0
11.50 012332111000
10.50 0 3 9 11 8 5 2 1 1 0 0 0
9.50 1 13 27 24 15 8 4 2 1 1 0 0
8.50 4 39 57 38 19 8 3 1 1 0 0 0
7-50 13 80 76 37 14 5 2 1 0 0 0 0
6.50 32 99 61 21 6 2 1 0 0 0 0 0
5.50 44 70 28 7 2 0 0 0 0 0 0 0
4.50 32 22 6 1 0 0 0 0 0 0 0 0
3-50 820000000000
2.50 100000000000
 Fatigue assessment of ship structures 369

5.3 Transfer function

The transfer function, defining the response to a sinusoidal wave with unit
amplitude for different frequencies and wave heading directions, is
obtained applying linear potential theory and the equation of motions of
the ship. Transfer functions should be defined for the vertical-, horizontal-
and torsional bending moments, the external sea pressure and the internal
inertia pressure from cargo.
In the evaluation of the response in the waterline region, a reduction of
the pressure range applies due to intermittent wet and dry surface.
A consistent representation of phase and amplitude for the transfer
functions are necessary in order to achieve the correct modelling of the
combined local stress response. The combined local stress response is
obtained by summation of the local stress transfer functions for all the
relevant load components, conditioned on the wave heading direction
0,
all load
components
H,(o l 0) = Ai" H,(col O)
i=l

Hi(~l 0) is the load transfer function for the sectional load component i
and Ai is the stress response in the considered detail, including the stress
concentration factors, due to a unit load. The A; factors are determined by
finite element calculations, or alternatively, by simplified methods.

5.4 Design wave approach

As a simplification to the frequency domain analysis, a design wave

approach may be applied. In this approach, the reference load response
at a selected level, e.g. 10-4, is determined. The derived reference
response is combined with an assumed Weibull shape parameter h in
order to define the long term stress range distribution. The shape
parameter can be determined from a long term distribution of the
dominating load contribution, according to the procedure described
earlier, or alternatively, based on the empirical expression given in the
simplified stress analysis.
The selection of wave height and wave period should in general aim at
determining the load response effect at the 10-4 probability level. This
may also be determined by choosing the 10-4 combination of wave height
and period that gives the most severe load response effect. Guidance on
choice of wave steepness is given in Ref. 1.
370 E. H. Cramer, R. L~seth, K. Olaisen

6 FINITE ELEMENT ANALYSIS

6.1 Finite element models

The main aim of applying a finite element model in the fatigue

analysis is to obtain a more accurate assessment of the stress response
in the hull structure. The finite element analysis may be divided into
three levels:
1. global hull analysis
2. analysis of frames and girders
3. local stress analysis.

6.2 Global hull analysis

The purpose of the global hull analysis is to obtain a reliable description

of the overall stiffness and global stress distribution in the hull. The
following effects should be taken into account:
-- vertical hull girder bending including shear lag effects
-- vertical shear distribution between ship side and bulkheads
-- horizontal hull girder bending including shear lag effects
-- torsion of the hull girder (if open hull type)
-- transverse bending and shear.
The extent of the model is dependent on the type of response to be
considered and the structural arrangement of the hull. Generally, a
complete model of the hull is desirable in order to obtain a good repre-
sentation of the stiffness and load transfer in the aftbody and forebody.
The loads can be produced by direct computation of hydrodynamic
pressures and motions applied on the model. Simplified load models
(pressures/line loads) producing specified hull girder loads may also be
applied.
The global analysis may be carried out with a relatively coarse mesh.
Stiffened panels may be modelled by means of anisotropic elements. Alter-
natively, a combination of plate elements and beam elements, may be used.
It is important to have a good representation of the overall membrane panel
stiffness in the longitudinal/transverse directions and for shear.

6.3 Analysis of frame and girder structures

Frame and girder models should be capable of analyzing deformations as

well as stresses in the framing/girder system. Typical results derived will be
 Fatigue assessment of ship structures 371

membrane stresses caused by bending, shear and torsion for example in a

double skin construction.
The structural modelling shall generally be such as to represent the
deformation response sufficiently. The model for analysis of frames and
girders should be compatible with the global model if forced deformations
should be applied. If a separate analysis of frames and girders is carried
out without any finite element calculation of the global stress response, the
extent of the model, boundary conditions and load distribution should be
carefully evaluated in order to obtain an acceptable global support and
stiffness for the frame/girder model.
Similarly to the global hull model, the frame/girder model is primarily
used for calculation of nominal stresses.

6.4 Local finite element analysis

Local finite element analyses may be used for calculation of local stresses
for determination of the stress concentration factors, the K-factors. These
analyses involve use of fine element mesh models of details such as bracket
connections, stiffener to web frame connections or local design of frames/
girders.
The K-factor is calculated in two steps, K = Kg. Kw, being the product
of the; stress concentration factor due to the geometry effect and the stress
concentration factor due to the weld itself.

1. By means of a fine mesh model using shell elements or solid

elements, the stress concentration due to the geometry effect of the
actual detail is calculated resulting in a geometric stress concentra-
tion factor, the Kg-factor.
2. The stress concentration due to the weld itself, the Kw-factor, may
be based on standard values from tables, ~ or direct finite element
calculations with very fine mesh of solid elements (the weld radius
has to be modelled).

The aim of the local finite element analysis is normally not to

calculate directly the stress at a detail, but to calculate the stress
distribution in the region of the hot spot such that these stresses can
be used as a basis for derivation of the geometric stress concentration
factor's.
It is important to have a continuous and not too steep change in the
density of the element mesh in the area where the local stresses are to be
analyzed. The following element sizes are recommended for a welded plate
attachment, where t is the plate thickness:
372 E. H. Cramer, R. Loseth, K. Olaisen

2 0 - n o d e solid e l e m e n t s t x t × t

8 - n o d e shell e l e m e n t s 2t x 2t

4 - n o d e shell e l e m e n t s t × t

Normally the element stresses are derived at the nodes or at the Gaus-
sian integration points. Depending on element type one may thus have to
perform several extrapolations in order to determine the stress at the weld
toe. Principal stresses are used for the extrapolation. It is recommended
that the final extrapolation of stresses is performed by a linear extrapola-
tion of the surface stresses at a distance t/2 and 3t/2 from the weld toe.
Elements containing discontinuities should not be included in the extra-
polation.
The region between t/2 and 3t/2 is found to be sufficiently far away to
avoid pick-up of the peak of the notch stress resulting from the weld, and
close enough to the notch that it can pick up the gross geometry behavior
in the joint region. The geometric stress concentration factor is then
defined as the increase in the stress level at the weld toe relative to the
nominal stress.

7 STRESS C O N C E N T R A T I O N F A C T O R S

7.1 General

The stress concentration factors may be determined based on finite

element analyses or through predefined empirical expressions based on
experience and previous analyses.
The fatigue life of a detail is governed by the notch stress range, defined
as the product of the nominal stress range and the stress concentration
factor, the K-factor.
All stress risers have to be considered when evaluating the notch stress.
The resulting K-factor to be used f o r calculation of the notch stress is
defined as

K = gg" gw" gt e" gt~" gn

where
Kg= stress concentration factor due to the gross geometry of the
detail considered
Kw= stress concentration factor due to the weld geometry.
(Kw = 1-5 if not stated otherwise).
 Fatigue assessment of ship structures 373

gte additional stress concentration factor due to eccentricity

tolerance (normally used for plate connections only)
Kt~---- additionally stress concentration factor due to angular
mismatch (normally used for plate connections only)
additional stress concentration factor for un-symmetrical
stiffeners on laterally loaded panels, applicable when the
nominal stress is derived from simple beam analyses.

In Ref. 1, a selection of stress concentration factors for typical ship struc-

tural details is specified.

7.2 Workmanship

The fatigue life of a welded joint is highly dependent on the local stress
concentrations factors arising from surface imperfections during the
fabrication process, consisting of weld discontinuities and geometrical
deviations.
Surface weld discontinuities are weld toe undercuts, cracks, overlaps,
porosity, slag inclusions and incomplete penetration. Geometrical imper-
fections are defined as misalignment, angular distortion, excessive weld
reinforcement and otherwise poor weld shapes.
Embedded weld discontinuities like porosity and slag inclusions are
harmless for the fatigue strength when kept below normal workmanship
levels
Empirical expressions for calculation of the geometric stress concentra-
tion factor due to the local weld geometry and fabrication tolerances for
alignment of butt joints and cruciform joints are given.1 Normally default
value.,; based on geometrical imperfections within limits normally accepted
according to good shipbuilding practices should be applied, if not other-
wise specified.
In special cases, K-factors may be calculated based on specified, higher
standards of workmanship. However, care should be taken not to under-
estimate the stress concentration factors, assuming a quality which is
difficult to achieve and follow up during production.
It is recognized that the fatigue life of steel structures is reduced
under corroding conditions. For steel which is fully cathodic protected,
approximately the same fatigue life as in dry air is obtained.
In estimating the efficient life time of coating- and cathodic protection
systems, due consideration is to be given to specification, application and
maintenance of the systems. A guideline for effective life times for
common corrosion protection systems is given.
374 E. H. Cramer, R. Loseth, K. Olaisen

8 NUMERICAL EXERCISE

8.1 Description

A numerical exercise showing the procedures for estimating the accumu-

lated fatigue damage in the side shell longitudinal-transverse frame
connection is presented, see Fig. 1. The simplified approach for calcula-
tion of loads and load effects is applied. The considered detail is located in
a coated side-shell ballast tank with cathodic protection in the mid-ship o f
a double hull tanker.
The considered ship has the following characteristics:

Principal dimensions

Length of ship L = 221.0 m

Breadth o f ship B = 42.0 m
Block coefficient Cs = 0.83
D e p t h of ship D = 20.3 m
Neutral axis above keel no -- 8.2 m
Vertical sectional modulus at deck line Zv -- 23-1 m 3
Horizontal sec. modulus in ship side Zh = 40.1 m 3

Load condition

L o a d e d draught Tt = 14-2 m
Metacentric height GM = 5.0 m
Roll radius of gyration KR = 16-4m
Ballast draught Tf = 7-2m
Metacentric height GM -- 13-9 m
Roll radius o f gyration KR = 16.4m
Cruising speed V = 13-0 knots
Fraction of time loaded Ptoad = 0"5
Fraction o f time in ballast Pball : 0"5
Fraction o f time at sea Ps ---- 1.0

Local geometry of stiffener

Stiffener sectional modulus at top of flange Zs = 0-461 10 -3 m 3

Stiffener spacing s = 0-790 m
W e b frame spacing ls = 3.420 m
Distance above keel z -- 13.06 m
 Fatigue assessment of ship structures 375

i~ Effectivespanlength2753 ~ I ~ ~ x350x10

Consideredpoint
Skinplate 200x10
/

q
790x16.5 250x90x9/14

SectionA-A I I

Fig. l. Descriptionof consideredship structuraldetail.

Thickness o f flange tf = 14.0 mm

Thickness of bracket tb = 10.0 mm
Width o f flange br -- 90.0 mm
Height of stiffener h = 250.0 mm
Thickness of web t,,, = 9.0 mm
Thickness of plate tp = 16.5 mm

8.2 Parameter estimation

S - N curve

The S - N curve type I in air is applied over the whole design life as the
considered welded detail is located in a coated ballast tank with cathodic
protection.

Stress concentration factors (K-factors)

The stress curve concentration factor is defined due to the local geometry
and the geometry o f the weld. The considered geometry is an un-symmet-
ric L-profile exposed to lateral loading. For laterally loaded panels there is
an additional stress concentration for un-symmetrical stiffeners compared
376 E. H. Cramer, R. Loseth, K. Olaisen

to symmetric beam result. The combination with a triangular bracket will

result in the following stress concentration factors.
For supported members welded to a stiffener flange the combined local
K-factor due to the gross geometry and the geometry of the weld is (from
Table 7.21)
Kw = 1.50
Kg = 1.47
where it is assumed that the bracket weld is kept clear of the flange edge
and that the stiffener is straight.
The K-factor for a laterally loaded un-symmetrical stiffener is (7.2.12),1
Kn = 1.49
The total stress concentration factor for axial and lateral loading is then,
Kax s = Kg" Kw = 2-20
gla t : K g . Kw. K, = 3.27

8.3 Simplified loads

The numerical values are taken according to the described simplified

method, here described with basis in the fully loaded condition.

Internal pressure loads

The longitudinal is located in a ballast tank with no local bending in the

fully loaded condition due to internal pressure loads.
The loads due to cargo pressures will, however, act on the double hull
and cause double hull stresses and frame deflections. The transverse
acceleration is found to give the largest inertia pressure load, resulting in
an internal pressure load amplitude at the 10-4 level equal to,
Pi = f a "P" at. ly~l = 16.6 kN/m 2
wherefa ~ 0.5 is the scaling factor transforming the pressure from 10-8 to
the 10 -4 level, p is the density of the liquid cargo in the tank, a t is the
transverse acceleration and Ys is the transverse distance from the centre of
the tank to the point considered.

E x t e r n a l sea pressure loads

The dynamic external sea pressure is calculated as,

Pe = rp . Pd = 38.2 kN/m 2
 Fatigue assessment of ship structures 377

where Pd is the dynamic pressure amplitude below the water line in

accordance with expressions taken from the D N V Rules 3 and rp is a
reduction factor for the effective pressure amplitude in the surface zone
(intermittent wet and dry surface). The dynamic pressure amplitude is
computed according to (4.3.1) 1 for a single amplitude roll angle of
0-377 radians, Pd = 64.7 k N / m 2. The effective pressure reduction factor
is 0.59.
For the simplified estimation of double hull stresses a uniform pressure
distribution is assumed. The effective average lateral pressure below the
waterline is scaled according to actual draught,

Tact = 26-7 k N / m 2
P c - dh = Pe " - -
D

Global hull bending m o m e n t s

The vertical wave induced dynamic sagging and hogging bending

moments and the horizontal bending moments are estimated in accor-
dance to the D N V Rules 2 (probability level 10-8),
M w , vsagg = -3469-9 kN/m

M w , vhogg = 3250.4 k N / m
Mw, H = 1843.1 k N / m

8.4 Simplified stresses

Stresses due to stiffener bending

The stress per unit pressure (kN/m 2) due to stiffener bending is computed
as described in (3.5.5), 1 accounting for the effective span of the long-
itudinal stiffener and the reduction in the effective bending m o m e n t at the
weld toe,

ff 2A-1 = Ktat • p "12Z

s" t 2 .rp = 2.67N/mm z

Stresses due to relative deflection between f r a m e a n d b u l k h e a d

Stresses will also be introduced due to deflection of the nearest frame

relative to the bulkhead. In order to determine the correct magnitude of
deflection, an element analysis is required. For a double hull vessel the
378 E. H. Cramer, R. Loseth, K. Olaisen

deflection should be seen in connection with the double hull bending

stresses. As an approximation, however, the minimum deflection along
the frame at the probability level 1 0 . 4 c a n be estimated based on
simplified formulas as a function of the frame-bulkhead distance, the
maximum dynamic pressure at the waterline and the material factor for
transverse frame, (3.5.6). l For unit dynamic pressure, having material
factor J~ -- 1.28 (NV32 steel), the deflection is

___P "/~
--"fl ' 10-4 = 2-19.10-2 mm
20

The estimated maximum deflection results in the following stress per unit
maximum dynamic local pressure at the stiffener,

6EI
O'2A-2 = Kaxs" ,-q--~r6 • ~ = 1.65 N/mm 2

Note that for the bracket close to the bulkhead a negative sign applies for
external pressure (compression).

Double hull stresses

The longitudinal secondary bending stresses, in the double hull panels at

the intersection with the transverse bulkhead, can be assessed applying 3-
dimensional frame or finite element analyses. In the following, however,
the stress in the top of the flange of the stiffener considered, caused by
double hull bending, is computed according to the approximate expres-
sions in Appendix B.I For unit average lateral pressure p at the middle of
the ship, the double hull stress is

if2 -~" g a x s " g b "p " b2 " ra = 1.34 N/mm 2

x/ia ib

ia and ib are the smeared out stiffness per girder about transverse and
longitudinal neutral axis of double bottom, b is transverse width of double
bottom panel, r a distance from detail considered to neutral axis of panel
and Kb is a coefficient dependent on apparent aspect ratio and the actual
boundary condition.

Stresses f r o m global loads

The vertical and horizontal wave induced bending moments results in the
following stresses,
 Fatigue assessment of ship structures 379

z
A t r , = f a " Kaxs " ( M w, Vhogg - M w, Vsagg ) " Ivv : 133.8 N / m m 2

A a h = fa " Kaxs " M w , 14 " ~Y = 97-0 N / m m 2

8.5 Combination of stress components

The stress amplitudes due to internal and external pressures are combined
accounting for the sign.

L o c a l stresses

Stresses due to internal pressure loads:

Stiffener bending 2.67. (p; = 0.0) 0.0 N / m m 2
Stiffener deflection 1-65. (Pi = 16-6) 27.2 N / m m 2
Double hull bending: - 1 . 3 4 . (Pi = 16.6) -22.0 N / m m 2
Stresses from internal pressure loads, a; 5-0 N / m m 2

Stresses due to external pressure loads:

Stiffener bending - 2 - 6 7 . (Pc = 38.2) - 102-0 N / m m 2
Stiffener deflection - 1 - 6 5 . (Pe = 38.2) - 62.5 N / m m 2
Double hull bending: 1-34. (Pc-dh = 26-7) 35-8 N / m m 2
Stresses from external pressure loads, tre - 128.7 N / m m 2
The combined stress range is then determined, having the average local
stress correlation Pei = 0-23,

A t r t = 2 . ~/tr2e + tr2i + 2 . Pei " 0"e" 0"i ~--- 255.3 N / m m 2

Global stresses

The ,combined global stress is determined using the average global corre-
lation Pvh : 0 . 1 0 ,

Acre = V / A a 2 + Aazh + 2 . Pvh . A a v . A a h = 172.9N/mm 2

C o m b i n e d notch stresses

The effective combined stress is corrected for an assumed zero mean stress
level at the welded detail, giving fm = 0.9, and assuming world wide trad-
380 E. H. Cramer,R. Leseth,K. Olaisen
ing, resulting in fe = 0"8. The combined hot spot stress range at the 10-4
level for loaded condition is then determined as,

I /~0"g-~-0.6. m0"l
Aa0 = f e "fro" max 0.6. Aa e + Aat = 258.5N/mm 2

8.6 Long term distribution

For the specified location of the considered detail, the long term Weibull
shape distribution parameter is h = 0-94.

8.7 Fatigue damage

The fatigue damage in fully loaded condition over a design life of 20 years
is investigated. The number of stress cycles in the fully loaded condition
over the design life is:
ntoad =Ps "Ptoad" Td" V0 = 0.9 X 0"5 X 6.7 X 108 = 3.0 X 107
The Weibull scale parameter is:

Aa0 -- 24.2 N/mm 2

q - (lnno)X/h

The part accumulated fatigue damage for the loaded condition over the
design life is then,
_ [. ml fSo'~hl qm2[ (~)h]
Dt°aa=nt°ad'qmlr01 1 + T ; ( , q ) ] +--~-2Y 1 + h ; =0.54.

where y ( ) and I" ( ) are the Incomplete Gamma function and Comple-
mentary Incomplete Gamma function.
By a similar analysis of the ballast condition, the part fatigue damage
over the design life is computed to be, Dbalt = 0.06.
The total fatigue damage over a design life of 20 years for world trade is
then,
D = Dtoad + Dbau = 0.06
which is within the acceptable usage factor r / < 1. The equivalent fatigue
life of the welded detail is,
Ta
Tfatig,,~ = -~- = 33 years
 Fatigue assessment of ship structures 381

The detail is acceptable for normal operation according to the described

procedure.

9 SUMMARY

A procedure for fatigue assessment of ships is presented. The procedure

accounts for two different levels of refinement for the load assessment, a
simplified approach with basis in the DNV Ship Rules, and a direct
approach based on a long-term frequency response analysis. Both
approaches, or combinations of these, can be applied in the load model-
ling. The procedure also defines two different levels of refinement in the
stress response modelling, a simplified procedure and a procedure based
on Finite Element analyses.
The fatigue assessment is based on the S--N approach under the
assumption of linear cumulative damage, where master S-N curves have
been defined for welded and unwelded details in corrosive and non-
corrosive environments for smooth geometries. The stresses to be consid-
ered are the local notch stresses, being the nominal stresses combined with
the stress concentration factors, K-factors, for the local geometry and the
geometry of the weld.

ACKNOWLEDGEMENTS

The presented work on fatigue assessment of ship structures has been

carried out under the DNV Classification project, Maintenance Friendly
VLCCs, headed by W. Magelssen. The following have contributed from
o

DNV: A. Boe, E. H. Cramer, S. Gran, G. Holtsmark, L. Hovem,

P. Lersbryggen, I. Lotsberg, R. Loseth, K. Olaisen, C. Miirer and S.
Valsg~rd. Valuable comments have been received from DNV Technical
Committees.

REFERENCES

° Det Norske Veritas, Fatigue Assessment of Ship Structures, Draft Classifica-

tion Note, DNVC Report No. 93-4032, Hovik, 1994.
2. Det Norske Veritas, Loads and Load Combinations for Fatigue Calcula-
tions- Background for the Wave Load Section of the DNVC Classification
Note: Fatigue Assessments of Ship Structures, DNVC Report No. 93-4032,
Hovik, 1994.
382 E. H. Cramer, R. Loseth, K. Olaisen

3. Det Norske Veritas, Rules for Classification of Ships, Hull Structural Design,
Ships with Length 100 Meters and above, Hovik, January 1993.
4. Det Norske Veritas, Classification Note no. 30.2, Fatigue Strength Analysis
for Mobile Offshore Units, Hovik, 1984.
5. BS 5400 Part 10, Code of Practice for Fatigue, Steel, Concrete and Composite
Bridges, British Standard Institute, 1980.
6. British Maritime Technology, BMT (Primary Contributors Hogben, H., Da
Cunha, L. F. and Olliver, H. N.), Global Wave Statistics, Unwin Brothers
Limited, London, 1986.
7. Det Norske Veritas, PROBAN-Theory Manual, DNVR Report No. 89-2023,
H~vik, 1989.
8. Det Norske Veritas, Classification Note no. 30.6, Structural Reliability
Analysis of Marine Structures, H~vik, 1991.

APPENDIX A F A T I G U E RELIABILITY ANALYSIS

There are a number of different uncertainties associated with fatigue life

predictions. The calculated loading on the ship is uncertain due to uncer-
tainties in the wave climate description, the sailing route and the
manoeuvering philosophy. The stress response in the ship is uncertain due
to uncertainties in the response modelling and in the evaluation of the
local stress concentrations. The fatigue capacity is uncertain due to
uncertainties in the local weld geometry, the fatigue capacity of the weld
material and the size of initial imperfections.
These uncertainties can be assessed directly through a stochastic
modelling of the uncertain variables, leading to probabilistic fatigue
analysis, defining the probability of having fatigue failure within the
design life of the ship.
In order to conduct a probabilistic fatigue analysis having a general
probabilistic modelling of the uncertain variables, more advanced prob-
abilistic analysis programs must be applied. 7 However, if the choice of
probabilistic models for the stochastic variables is limited to normal and
log-normal distributions, closed form analytical expressions for the fatigue
failure probability can be defined. In Ref. 1 procedures for such a model-
ling have been shown.
The outcome of a fatigue reliability analysis is;

- - the probability that fatigue failure shall occur in the considered

structural element within the design period considered,
- the design point, that is the combination of values for the stochastic
-

variables which is most likely to provoke a fatigue failure,

- - a set of partial coefficients for the stochastic variables, relating the
design point value to a characteristic value of the variable,
 Fatigue assessment of ship structures 383

--- a set of parametric sensitivity factors which quantify the change in

the reliability by a unit change in the statistical parameters of the
stochastic variables.
The target annual probability of failure should comply with target values
derived from relevant standards. 8 The target values should be selected
after due consideration to the possible consequence of a fatigue failure in
the structural element considered, e.g.
--- loss of stability due to flooding,
--- pollution potential caused by cargo leakage,
--- explosion risk due to gas leakage,
--- initiation of progressive collapse.