FATIGUE ANALYSIS

ON OCEAN STRUCTURES

Eko B Djatmiko
Department of Ocean Engineering
Faculty of Marine Technology ITS
Surabaya January 2012

1. INTRODUCTION
Damage and failure on ocean steel structures (OSS: offshore platforms,
ships, etc): mainly due to fatigue; at primary, secondary or tertiary
structural elements (intensity increases with corrosion)
Fatigue damage is one of the most important failure modes in OSS,
which are subject to continuous dynamic variable amplitude loading,
comprises of:

Low frequency (quasi-static) cyclical load brought about the wave excitation,
at the rate of some 107 ~ 108 times during the operational life of the OSS (abt
20 years)
High frequency (dynamic) cyclical loads which can be classified into transient
loads (slamming, wave slapping, hull whipping) and steady loads (engine
vibration, propeller, hull springing), at the rate of 106 times during the
operational life of the OSS (abt 20 years)
Very low frequency (static) cyclical load brought about the variation of logistic
loads and hydrostatic loads (tidal), at the rate of 4000 ~ 8000 times during the
operational life of the OSS (abt 20 years)
Cyclic loads due to the irregular thermal gradient brought about the climate
and cargo temperatures, at the rate of 7000 times during the operational life
of the OSS (abt 20 years)

1-1

 A large number of factors affect the fatigue damage on OSS, like:

Mean stresses and their redistribution
Residual stresses
Loading of the structure including load sequences
Thickness of the structural joints
Corrosive environments and temperature of the surroundings
Design
Fabrication and methods for improving fatigue performance
Sensitivity of the material
Fatigue occurs mostly on the weld joints and structural elements
where stress concentration develops
Failure commences with crack initiation (fatigue) followed with crack
growth up to fracture take place on the structure.
Repair and maintenance costs for OSS: large proportion is
allocated to tackle the failure and damage due to fatigue (mostly
related also to corrosion)

 Fatigue analysis at the design stage is mainly directed towards

identification of the structural parts which has a high probability to
suffer fatigue failure and further considered as the basis for
redesign of the corresponding structures
Result of final fatigue checks is necessary to establish an
inspection strategy for OSS
Differences between analyses of fatigue limit state (FLS) and
working stress design (WSD), ultimate limit state (ULS), or
accidental limit state (ALS):

FLS

WSD, ULS, ALS

Takes into account all levels of load

intensities

Takes into account only the

maximum loads, eg. Highest wave
of 1-year period and/or 100-year
period

Takes into account total cycle

occurences of all load levels during
the operational life

Takes into account only one cycle

occurrence of maximum load.
1-2

 Fatigue Design Criteria of Offshore Structures according to API RP 2AWSD and API RP 2A-LRFD:

A detailed fatigue analysis should be performed for template type structures.

It is recommended that a spectral analysis technique be used.
In general the design fatigue life of each joint and member should be at
least twice the intended service life of the structure (ie. SF = 2.0)
For the design fatigue life, the design value damage ratio (damage index) D
should not exceed unity (<1.0)
For critical elements whose sole failure could be catastrophic use of larger
SF should be considered (eg. up to 5.0) applied also for members where
access for inspection and repair is restricted

Example of the requirement on the design fatigue life of the Belanak

FPSO:

Design service life 30 years

Fatigue life of 60 to 300 years (10 times of service life !!)

1-3

Alexander L. Kielland Rig Accident

(Source: wikipedia)
The Alexander L Kielland was a Norwegian semi-submersible drilling rig that
capsized whilst working in the Ekofisk oil field in March 1980 killing 123
people. The capsize was the worst disaster in Norwegian waters since World
War II. The rig, located approximately 320 km east from Dundee, Scotland,
was owned by the Stavanger Drilling Company of Norway and was on hire to
the U.S. company Phillips Petroleum at the time of the disaster.
In driving rain and mist, early in the evening of 27 March 1980 more than 200
men were off duty in the accommodation on the Alexander L. Kielland. The
wind was gusting to 40 knots with waves up to 12 m high. The rig had just
been winched away from the Edda production platform. Minutes before 18:30
those on board felt a 'sharp crack' followed by 'some kind of trembling'.
Suddenly the rig heeled over 30 and then stabilised. Five of the six anchor
cables had broken, the one remaining cable preventing the rig from
capsizing. The list continued to increase and at 18.53 the remaining anchor
cable snapped and the rig turned upside down.
1-4

A year later in March 1981, the investigative report concluded that the rig
collapsed owing to a fatigue crack in one of its six bracings (bracing D-6),
which connected the collapsed D-leg to the rest of the rig. This was traced to
a small 6 mm fillet weld which joined a non-load-bearing flange plate to this D6 bracing. This flange plate held a sonar device used during drilling
operations. The poor profile of the fillet weld contributed to a reduction in its
fatigue strength.
Further, the investigation found considerable amounts of lamellar tearing in
the flange plate and cold cracks in the butt weld. Cold cracks in the welds,
increased stress concentrations due to the weakened flange plate, the poor
weld profile, and cyclical stresses (which would be common in the North Sea),
seemed to collectively play a role in the rig's collapse.

1-5

27 March 1980 at 18.30

Casualty: 123 men out of 212 men onboard
Production loss

1-6

Fractures on the right side of the Alexander L. Kielland rig

1-7

Alexander L Kielland after accident

Broken support bracing

Broken support bracing

Alexander L Kielland salvaged

1-8

2. S-N CURVES AND GRAPHS

S-N Graphs contain scatter data obtained from fatigue tests on certain
structural joints (carried out in the laboratory) see Fig. 2.1 and 2.2

Figure 2.1. Test on structural configuration

on a load frame with actuators

Figure 2.2. Test on a specimen using

universal testing machine (UTM)
2-1

Figure 2.3. Fatigue test on a complex structural configuration

(aeroplane structure)
2-2

 There are two general types of fatigue tests conducted. One test focuses on
the nominal stress required to cause a fatigue failure in some number of
cycles. This test results in data presented as a plot of stress (S) against the
number of cycles to failure (N), which is known as an S-N curve. A log scale is
almost always used for N.
The data is obtained by cycling smooth or notched specimens until failure. The
usual procedure is to test the first specimen at a high peak stress where
failure is expected in a fairly short number of cycles. The test stress is
decreased for each succeeding specimen until one or two specimens do not
fail in the specified numbers of cycles, which is usually at least 107 cycles. The
highest stress at which a runout (non-failure) occurs is taken as the fatigue
threshold. Not all materials have a fatigue threshold (most nonferrous metallic
alloys do not) and for these materials the test is usually terminated after about
108 or 5x108 cycles.
Since the amplitude of the cyclic loading has a major effect on the fatigue
performance, the S-N relationship is determined for one specific loading
amplitude. The amplitude is expressed as the R ratio value, which is the
minimum peak stress divided by the maximum peak stress. (R=min/max). It is
most common to test at an R ratio of 0.1, but families of curves with each
curve at a different R ratio are often developed.
(source: www.ndt-ed.org)

2-3

 A variation to the cyclic stress controlled fatigue test is the cyclic strain controlled
test. In this test, the strain amplitude is held constant during cycling. Strain
controlled cyclic loading is more representative of the loading found in thermal
cycling, where a component expands and contracts in response to fluctuations in
the operating temperature.
It should be noted that there are several short comings of S-N fatigue data.

First, the conditions of the test specimens do not always represent actual
service conditions. For example, components with surface conditions, such
as pitting from corrosion, which differs from the condition of the test
specimens will have significantly different fatigue performance.

Furthermore, there is often a considerable amount of scatter in fatigue data

even when carefully machined standard specimens out of the same lot of
material are used. Since there is considerable scatter in the data, a
reduction factor is often applied to the S-N curves to provide conservative
values for the design of components.

(source: www.ndt-ed.org)

2-4

log N

b)

log S

log S

log S

a)

c)

log N

log N

Figure 2.4. S-N Graph: a) uncertainty due to the slope, b) uncertainty

due to intercept, c) total uncertainty

S-N Graphs shows the correlation between stress range, designated S or

S (in MPa or N/mm2), and the number of cycles, N, of the load excitation
which causes fatigue failure.
S-N Graphs is given in a log log scale (due to a large range of S and N)
S-N Curve (also known as Whler curve) is the mean line of the scatter
data, derived from regression analysis
The level of reliability on the accuracy in the determination of S-N curve is
influenced by the slope parameter and the intercept (or position) of the
curve within the graph. Both parameters have peculiar uncertainties.
Combination of uncertainties of the two parameters produces a total
uncertainty for the S-N curve (see Fig. 2.4)

2-5

log S

In practical design it is suggested to choose the curve having 95% level of

confidence (ie. approximately equal to lowering the mean curve by 2 times
of the standard deviation), as shown in Fig. 2.5.

a
b
c
log N

Figure 2.5. Regression of S-N curve for: a) Mean fatigue life

b) Mean minus 1x standard deviation, c) Mean minus 2x standard deviation

2-6

log S

S-N curve for structural joint configuration with shorter fatigue life tend
to be leaner/lower slope (see Fig. 2.6)

N2 > N1

Si

N1

N2

log N

Figure 2.6. Comparison of S-N curves with lower and higher slope

2-7

2.1 Analytical Expression of S-N Curve

Considering the form of the S-N curve, hence the appropriate equation
to be used correspondingly is:

NS m A
or

(2.1)

log N log A m log S

where:
N = cycles to failure
S = stress range
A = intercept of the log axis
m = slope of the S-N curve

2-8

DATA ON S-N CURVES

(see Fig. 2.7, Fig. 2.8 and Appendix A)

CLASS log A
B
15.369
7
C
14.043
2
D
12.600
7
E
12.516
9
F
12.237
0
F2
12.090
0
G
11.7525
W
11.5662

m
4.0
3.5
3.0
3.0
3.0
3.0
3.0
3.0

Department of Energy, Guidance Notes Revision

Drafting Panel, August 1983, Issue N
Offshore Installations: Guidance on Design and
Construction. New Fatigue Design Guidance for
Steel Welded Joints in Offshore Structures
Det norske Veritas, Fatigue Strength Analysis for
Mobile Offshore Unit, Classification Notes No. 30.2,
1984
T : All tubular joints
B,C,D,E,F,F2,G,W : All other joints depending on
a) Geometrical arrangement of the detail
b) The direction of fluctuating stress relative to the detail
c) The method of fabrication and inspection of detail
(see also Appendix A)
2-9

Figure 2.7. S-N curves for non-tubular structural joints

2-10

Figure 2.8. S-N curves for tubular structure joints

2-11

Figure 2.9. S-N curve for brittle aluminum with a UTS of 320 Mpa
(peculiar curve pattern in comparison to steel structure)
2-12

2.2 Effect of Thickness on the S-N Curve

Data of S-N curve is derived from the material test with thicknesses of:
t = 32 mm for tubular joints (T class)
t = 22 mm for other joints (B,C,D,E,F,F2,G and W classes)
If a standard S-N will be used for a structure having different plate
thickness then a correction should be made as follows:

t
N N0 0
t

m/4

A
because N 0 m
S
hence N

A t0

m
S t

(2.2)
m/4

2-13

3. FATIGUE ANALYSIS BY USING

DETERMINISTIC APPROACH
Computation of fatigue on a structural joint is carried out on the basis of
Palmgren-Miner (1945) cummulative damage hypothesis, expressed as:

ni
n3
nm
n1
n2
D

.........
N
N
N
N
Nm
i 1
i
1
2
3
m

(3.1)

where:
ni = number of cycles of stress range at intensity Si (N/mm2) which actually
occur on the structural joint brought about external load excitation (wave)
Ni = number of cycles of stress range at intensity Si (N/mm2) which will yield
fatigue failure on the joint in question. The figure may be obtained from an
S-N curve for an appropriate joint
Si = stress range (or Si); twice of stress amplitude that is experienced by the
joint (N/mm2)
In accordance to Palmgren-Miner hypothesis, the failure of the joint takes
place when the damage index D approaches value of 1.0.
3-1

 The value of Si accounted for in the computation is the maximum stress

range on a certain location within the joint (ie. hot spot stress) which can be
derived by magnifying the nominal stress range, Si(nom), by considering the
stress concentration factor (SCF). Hence the maximum stress range is
calculated as follows:

S i S i ( nom) SCF

(3.2)

The nominal stress range, Si(nom), is obtained from the analysis of regular
wave load (deterministic analysis) to generate internal forces and/or
moments on the structural components in question, appropriately
correspond to the wave in the Metocean data.
The wave load so obtained is further accounted for in the structural analysis,
for instance global analysis by using a conventional stress analysis or by
means of global FEM (eg. SAP, GTSTRUDL, etc) to derive the nominal
stress range, Si(nom).
The value of SCF for a joint may be found by adopting peculiar formulae as
can be found in references by Almar-Naess (1985), API (1980), Munse
(1984), etc.
SCF is not necessary to be computed when the FEM could directly produce
stresses on the detail structure (eg. NASTRAN, ANSYS, ABACUS)
3-2

 The number of cycles ni for any stress range Si which arises due to the
wave load is characterized by wave height Hi (m) and period Ti (sec) can
be calculated by using the following equation:

Pi T
ni
Ti

(3.3)

Pi is the relative frequency of occurrence of each wave, having

characteristic height Hi (m) and period Ti (secs) which causes a stress Si
to develop.
Variable T is the fatigue life of the structure after counting all stress cycles
By substituting eq. (3.3) into eq. (3.1), the equation of structural fatigue
failure becomes:
D

PT
PT
P1T
PT
2 3 ........ m 1
N 1T1 N 2T2 N 3T3
N mTm

(3.4)

The fatigue life T is finally found by solving the above eq. (3.4) by taking
into account Pi , Ni and Ti as shown in the example contained in Table
3.1.
3-3

Table 3.1. Example of fatigue calculation by deterministic method

S-N Curve:
NS3=1.41x1012
Hi(m)

Ti(det)

Pi

Si(N/mm2)

Ni

Pi/(NixTi)

0.0 1.5

0.8781

11

1.059E+09

2.763E-10

1.5 3.0

0.1035

32

4.303E+07

4.811E-10

3.0 4.5

0.0124

79

2.860E+06

6.194E-10

4.5 6.0

0.0042

124

7.395E+05

6.310E-10

6.0 7.5

10

0.0011

158

3.575E+05

3.077E-10

7.5 10.0

11

0.0005

191

2.024E+05

2.246E-10

10.0 12.5

12

0.0001

226

1.222E+05

6.822E-11

0.999

Total =

2.608E-09

T(sec) = 3.834E+08
T(yrs) = 12.157

Known (Metocean Data)

Calculated from regular wave load analysis for
Hi and Ti (deterministic method) continued with
Stress analysis to obtain Si

Calculated from
S-N curve equation:
Ni = A/Sim

Calculated by:
m

T D
i 1

N i Ti
Pi

1 year = 31,536,000 secs

3-4

As a summary, the procedure in accomplishing the structural fatigue

calculation by the deterministic method is performed as follows:
a. Find the wave distribution data containing the values of Hi (m), Ti
(secs) dan Pi
b. Calculate the wave load at any concerned joint as a function of wave
height Hi (m) and period Ti (secs) (by adopting the regular wave
theory deterministic method)
c. Calculate the nominal stress range Si(nom) (N/mm2) for any concerned
joint (by means of stress analysis, or FEM)
d. Calculate the SCF appropriately for the type of joint so concerned
e. Calculate the maximum stress range Si (N/mm2) at the hot spot
f.

Select an appropriate S-N diagram related to the type of joint and

calculate values of Ni as functions of Si (N/mm2); this can be read
from the graph or derived from equation NSm=A
g. Calculate all Pi /(NixTi) and subsituted those into eq. (3.4) to obtain
fatigue life of the joint, ie. T (with final result in years) by inversing the
summation T = D (Ni x Ti)/Pi
3-5

It should be noted that the deterministic approach has some drawbacks, as

follows:
a. The wave loads applied on the structure are generated by regular
wave, which is not essentially true in real operation.
b. The wave at any height interval Hi (m) corresponds only to a single
period Ti (secs). Hence this is not appropriate to be implemented on
structures which are sensitive to the wave period (or reversely
frequency) variation, where resonant might develop.
c. Point b) also implies that the method is appropriate only if the
structure having natural period (or frequency) outside the wave
periods commonly occur at sea. Such structures are inherently stiff,
eg. fixed jacket platform and other robust offshore structures.
d. For period (or frequency) sensitive structures the deterministic
method might give an over- or under-estimate results. Hence it
should be used only for rough estimate of fatigue life in early design
stage.

3-5

4. FATIGUE ANALYSIS ON OSS

BY FULL SPECTRAL METHOD
There are a number of aspects need to be comprehended as basic thoughts on
the necessity in performing fatigue analysis on OSS by employing the Full
Spectral Method, as follows:
OSS are designed to be operated in real seas with the primary (dominant)
environmental loads due to the wave excitation.
The real sea waves are random in nature, hence the responses of an OSS
due to the wave load excitation will also be random;
Therefore the number of load cycles as well as intensities should be
computed by applying an appropriate method and procedures so that
accurate results of the analysis will be attained.

Figure 4.1. An example of a random wave time history

4-1

Random waves (as in the case of any random signal), as shown by the time history
in Fig. 4.1, by means of Fast Fourier Transform could then be presented in the form
of wave spectra (see references on sea waves).

Following this, the random responses of an OSS could also be presented

in the form of response spectra. This is obtained by correlating the
responses in regular waves and the wave spectra. (Note: random waves as well
as random responses are composed by the superposition of a large/infinite number of regular
components)

From a spectra (either wave or response), and by applying certain

formulations and algorithms, one may derives the statistical value
(including the distribution) of the intensities as well as number of cycles
that could possibly develop during the lifetime of an OSS.
The technical detail of the full spectral analysis is as described in the
following.

4-2

 First Stage: the distribution of fatigue load is computed on the basis

of loads on OSS due to regular waves excitation, similar to that
carried out in the deterministic approach (by employing Morisons
theory, 2-D strip theory, or 3-D diffraction theory), depending on the
OSS configuration.

The difference in comparison to analysis in Chapter 3 is, here the load

computation is performed by varying the wave frequencies
(commonly between 0.1 ~ 2.0 rad/s at interval of 0.2 rad/s).
Results of the analysis are then presented in the form of transfer
function graph, which correlate between the ratio of load amplitude
(bending moment, shear force, etc) to wave amplitude, designated as
the RAO (Response Amplitude Operator), for each incremental
frequency . (see Fig. 4.2)
RAOs are computed for a number of wave directions (appropriately
representing the occurrence in the operational site of the OSS).
In some cases RAOs are also computed for a number internal load
conditions. For example analysis related to FSO may include: ballast
condition, 10% storage load, 50% storage load, and 100% storage
load.

4-3

Stress Analysis,
(FEM)

RAO = S/w

Wave Load
Analysis
(Regular)

RAO = Fw/w

Second Stage: transforming the Load RAO into Stress RAO for
particular joints under observation by performing stress analyses
(most appropriately applying FEM) see Fig. 4.2

STAGE I

STAGE II

Figure 4.2. Load analysis to derive the Stress RAO

4-3

 Third Stage: conducting the full spectral analysis, comprises of the

following steps:

Step 1: the definition of operational scenario (operational box).

This step include also the determination of probability related to the
occurrence of any sub-operational box, eg. probability of wave
occurrences, probability of wave directions, probability of load
conditions, probability of advancing speeds, etc)
Probability of wave occurrences may be obtained from the wave scatter
diagram, as exhibited in Fig. 4.4. The (joint) probability of any single
combination of H and T is the fraction of its occurrence relative to the
total wave occurrences.
Probability of wave direction may be obtained from Metocean data,
which in most cases very much related to the geographical
chracteristics, climate and (dominant) wind directions.
Probability of load conditions is obtained from the fraction of certain
load condition (eg. 50% cargo load) relative to the overall load
operation (say: ballast, 10%, 50%, and 100% load)
Probability of advancing speeds is related to the sea going ships, which
are operated at different speed levels. For instance military vessel may
be operated at speeds correspond to harboring, surveilance, chasing,
combat, etc.
4-2

Wave
(H & T)

3m
40o
2m

30o

Load Cond.
Full Load

Wave Dir.
Figure 4.3. Operational scenario (box of operation) of OSS

pi = probability of wave intensities (joint prob of H & T)

pj = probability of wave directions
pk = probability of load conditions
4-3

Table 4.1. ABS Wave Scatter Diagram for Unrestricted Service Classification
[ABS, GUIDE FOR: SPECTRAL-BASED FATIGUE ANALYSIS FOR FLOATING PRODUCTION,
STORAGE AND OFFLOADING (FPSO) INSTALLATIONS, May 2010]
Wave Periods
Wave
Heights
(m)

3.5

4.5

5.5

6.5

7.5

8.5

9.5

10.5

11.5

12.5

13.5

Sum
Over All
Periods

0.5

260

1344

2149

1349

413

76

10

5610

55

1223

5349

7569

4788

1698

397

69

21158

406

3245

7844

7977

4305

1458

351

65

10

25670

113

1332

4599

6488

4716

2092

642

149

28

20161

30

469

2101

3779

3439

1876

696

192

43

12625

156

858

1867

2030

1307

564

180

46

7016

52

336

856

1077

795

390

140

40

3688

18

132

383

545

452

247

98

30

1906

53

172

272

250

150

65

22

990

22

78

136

137

90

42

15

5224-4

1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
9.5

8
7
1

1
1

Hs (m)
4

1 2

3 2

1 4

6 11 10

11 23 31 40 59 47 14

3 44 58 49 61 69 144 48 21

1
1

1
2

4 41 53 65 82 117 186 193 97 35 12

12 20 9 27 55 63 104 66 43 18 20
2

10

12

14

16

Tp sec)

Figure 4.4. Wave scatter diagram (total wave data = 2196)

4-5

 Third Stage: conducting the full spectral analysis, comprises of the

following steps:
Step 2: conducting computations to generate the stress range spectra on the
basis of data yielded from the RAOs and wave spectra for every element in the
box of operation. This step is basically identical with changing the information of
stresses in regular waves into stresses in random waves (sea waves).

SS()

S()

2
RAO = S/w

RAO2(stress) x Wave Spectra = Stress Spectra

Figure 4.5. Computation to obtain the stress spectra
4-6

 Third Stage: conducting the full spectral analysis, comprises of the

following steps:
Step 3: conducting computations to obtain the number of cycles and followed by
computations of the distribution in the short-term scope, as shown in Fig. 4.6.

Number of cycles
Short-term
1
n0
(1/det)

m2

m0

m0 SS ( ) d
0

pS(S)

ps (S )

S 1 S2

S3 S4

Sm

S S 2 / 2 m0
e
m0

m2 2SS ( ) d
0

Figure 4.6. Computation of stress range short-term distribution

(Rayleigh distribution)

4-7

Step4: Computation of number of cycles in the long-term (operational life of

the structure) by:

nL ( n0 pi p j pk ) xTL
i

(4.1)

where :
nL = number of stress cycles in the long-term
TL = lifetime of the structure (secs)
no = number of cycles per unit time (1/sec); can be found from the interval
operation in the short-term (see Fig. 4.6):
n0

1
2

m2
m0

(4.2)

mo = area under the spectral curve (of stress response)

m2 = second moment of area under the spectral curve (of stress
response)
4-8

 Step 5: Determination of the probability density function (PDF) for the

stress range S, in the long-term period (lifetime of the structure); which
could be approximated by Weibull distribution, as follows (see Fig. 4.7):

p L (S )

n
i

pi p j p k p s ( S )

(4.3)

n0 pi p j pk
i

where ps(S) is the PDF in the short-term; which is approximated by

the Rayleigh distribution, as follows:

S S 2 / 2 m0
ps (S )
e
m0

(4.4)

4-9

Hi & Ti

n0m-2

n0m-1

n0m

Load cond

Wa
ve
Dir
.

n01

n02

n03

+
x pi . pj . pl
t io n

u
Weib

u
s t r ib
i
d
l
l

100

101

102

103

104

105

106

107

108

Total number of occurence = nL

Figure 4.7. Computation of the long-term stress range distribution

(Weibull distribution)

4-10

SUMMARY ON
THE PROCEDURE OF FULL-SPECTRAL FATIGUE ANALYSIS
1. Performing the regular wave load analysis to derive RAO of the structural responses
(Bending Moment, Shear Force); carried out for a number of appropriate wave directions
(eg. 0, 45, 90, 135, 180 degs)
2. Transforming the structural response RAOs into the stress range RAOs (using stress
analysis, or FEM)
3. Defining the operational scenario of the OSS by considering among others: the metocean
data (wave scatter diagram, joint probability of H & T, wave direction), loading conditions,
advancing speeds (for travelling ships), spectral variation (if any), and so on
4. Computing the stress spectra for all mode of operation as defined in point 3)
5. Computing number of cycles (eq. 4.2) and distribution of stress cycles in the short-term as
can be represented by Rayleigh distribution (eq. 4.4) for each operational mode in point 3)
6. Computing the distribution of stress cycles in the long-term (which is the summation of all
the short-term stress cycle distributions) by considering the designed operational life T (in
years seconds) and all the probabilities of elements within operational mode in point 3),
and solving eqs. (4.1) and (4.3) the long-term stress range distribution will follow the
theoretical Weibull distribution
7. Correlating results of the analysis and computation of stress cycle distribution in the longterm as obtained in point 6) with the fatigue data represented by S-N curve by
implementing the Palmgren-Miner rule (eq. 4.1) to finally determine the fatigue life of the
structural joint under observation.

4-11

5. CLOSED-FORM FATIGUE EQUATION

Method of predicting the fatigue damage on steel structure has been

introduced by Palmgren-Miner through the accumulative damage
hypothesis:

ni
n3
nm
n1
n2
D

.........
N1 N 2 N 3
Nm
i 1 N i
m

(5.1)

If p(S) is the stress pdf which can be defined in such a way hence p(Si)dS is
equivalent to the number of oscillation of stress component with the peak
value lies within an interval dS and with the mean value of Si. Further by
taking f and T as the mean frequency which vary randomly and the overall
operational time, respectively, hence the increase of the damage due to Si
which will take place during an interval T is:

T f p ( S i )dS
dD
N Si

(5.2)
5-1

N(Si) is number of cycle which would bring about damage at stress level Si.
From eq. (5.2) the expected damage that would take place in a certain
period T, could then be obtained by integrating contributions of all cycles of
the stress components, that is:

E ( D) T f
0

p L (S )
ds
N (S )

(5.3)

If T is the operational life of the structure as initially designed (ie. TL), hence
TL= nL/f. So eq. (5.3) could be rewritten by substituting N=A/Sm , as follows:

n
E ( D) L S m p( S )ds
A 0

(5.4)

5-2

It is important to notice that p(S) is the continuous form

(theoretical) of pL(S) in eq. (4.3), which represents a discrete
distribution. From various investigation on the long-term wave
distribution it has been concluded that PL(S) could be closely
approximated by Weibull pdf, namely:

S
p L (S )

S
exp

(5.5)

where:
= scale parameter
= form parameter
The value of is a function of the extreme stress range.
The value of is a function of the structural configuration and operational
sea site; for the general equation may range between 0.75 up to 2.0; for
ocean structure may range between 0.9 (mostly for large structures) up
to 1.1. (mostly for small structures).
5-3

If parameter Se is defined as the extreme stress occurs once during overall

cycles nL , thence could be calculated as:

ln nL
Se

(5.6)

or

Se( ln nL ) 1/

Parameter will be obtained from iteration on the basis of results from

pL(S) and nL .

By substituting eq. (5.5) into eq. (5.4) we can find:

n
S
D L Sm
A 0

S
exp ds

(5.7)

5-4

Integral in eq. (5.7) could be simplified by implementing the gamma

function, (n) and substituting into eq. (5.6) as follows:

S
Taking

will give

nL m (1 m / ) 1
D x
exp x dx
A 0

(5.8)

If the gamma funtion is defined as :

(n) e t t n 1 dt

(5.9)

Approximation of gamma
function:

Stirling Formula :

( x) 0.0076 exp (1.6 x) 1.26

( x 1) 2x x / e

5-5

Analogy of the factor within the integral in eq. (5.8) and (5.9) yields:

nL m
(1 m / )
A

(5.10)

or, by substituting eq. (5.6) into eq. (5.10) finally we find:

nL Se m
D
(1 m / )
m /
A (ln nL )

(5.11)

this is referred to as
the CLOSED-FORM

FATIGUE LIFE EQUATION

5-6

6. RELIABILITY AGAINST FATIGUE FAILURE

Margin of Safety :

M RL

(6.1)

R = resistance or strength factor

L = load factor

Definition of margin of safety with regards to fatigue (by considering D from

eq. 5.11) is :

MF D
NL
Se m
MF
(1 m / )
m
/

A (ln NL )

(6.2)

6-1

Table 6.1. Some references on ftigue failure index

COV -

Shin & Lukins

0.90

0.67

Shutz

1.00

0.60

Schilling

0.70

0.60

Gurney

0.85

0.22

Eide & Berge

0.78

0.19

Holmes & Kerr

0.69

0.61

Table 6.2. Examples of uncertainty on variables

VARIABLE

MEAN

COV

DIST TYPE

NL

8.4x107

0.05

Log-Normal

1.51x1012

0.31

Log-Normal

3.0

0.03

Normal

Se

200

0.20

Log-Normal

0.94

0.05

Log-Normal

6-2

 Solutions : MVFOSM (mean value first-order second moment),

AFOSM (advanced first-order second moment), or Monte Carlo
Simulation
MVFOSM : R & L are independents with normal distribution

pf P ( L R ) ( )

M
1

M VM

(indeks keselamatan)

M x i
2
VM

x
M

RL
2
R
L2

(6.3)

Vi2

1
2VR2 VL2

(6.4)

Vi

(6.5)

(6.7)

i
COV
xi

(6.6)

R/L

(6.9)

6-3

APPENDIX A.
CLASSIFICATION OF STRUCTURAL JOINTS (refer also to Fig. 2.8)

A-1

A-2

A-3

A-4

A-5

A-6

A-7

A-8

A-9

A-10

APPENDIX B
STRESS CONCENTRATION FACTOR

Faktor konsentrasi tegangan atau stress concentration factor (SCF)

merupakan perbandingan antara tegangan tertinggi di suatu posisi pada
sambungan (hot spot stress) dan tegangan nominal pada brace
(Gibstein, 1985):
SCF

maks
n

Besarnya SCF untuk tiap sambungan akan berbeda tergantung pada

geometrinya dan SCF ini merupakan parameter yang dapat
mengindikasikan kekuatan sambungannya. Konsentrasi tegangan
menggambarkan suatu kondisi dimana telah terjadi tegangan lokal
yang tinggi akibat dari geometri sambungan tersebut, sehingga
dibutuhkan keakuratan yang tinggi dalam penentuan nilai tegangan hot
spot, dan juga pennetuan SCF untuk jenis sambungan yang berbeda.

B-1

Pada titik yang berdekatan di suatu sambungan antara chord dan brace
nilai SCF yang terjadi akan berbeda, karena kedua member mempunyai
parameter-parameter dan orientasi yang berbeda. SCF untuk brace diberi
notasi SCFb dan untuk chord diberi notasi SCFc.
Hot spot adalah lokasi pada suatu sambungan (tubular) dimana terjadi
tegagan tarik/tekan maksimum. Secara umum diidentifikasi sda tiga tipe
tegangan dasar yang menyebabkan munculnya hot spot (Becker, et al.,
1970):
1. Tipe A, disebabkan oleh gaya-gaya aksial dan momen-momen yang
merupakan hasil dari kombinasi frame dan truss jacket.
2. Tipe B disebabkan detail-detail sambungan struktur seperti geometri
sambungan yang kurang memadai, variasi kekakuan yang bervariasi
disambungan dan lain-lain.
3. Tipe C, disebabkan oleh faktor metalurgis yang dihasilkan dan kesalahan
pengelasan, seperti undercut, porosity, dan lain-lain.

Untuk mencari besar dari SCF dapat dilakukan dengan pengukuran

langsung yaitu dengan ekspenimen dengan menggunakan strain gage
atau dengan menggunakan rumus-rumus pendekatan (kuang, semedley
dil). Beberapa rumus pendekatan yang diberikan oleh Kuang dan
Smedley sebagai berikut
B-2

BEBAN
AKSIAL

BEBAN
AKSIAL

IN PLANE
BENDING

IN PLANE
BENDING

BRACE

BRACE

OUT OF
PLANE
BENDING

g
T

OUT OF
PLANE
BENDING

D
L

Gambar B.1. Parameter tubular joint

Parameter Utama:
L= panjang Chord
D
= diameter terluar chord
d= diameter terluar brace
T= tebal chord
t = tebal brace
g= jarak ujung ke ujung antara
brace

Parameter Turunan:
= 2L/D
= t/T
= d/D
= g/D
= D/2T
= sudut antara brace dan chord

B-3

1. Persamaan SCF Smedley

Persamaan Smedley diberikan untuk batasan parameter sebagai berikut:

8 40
0.25 1
0.13 1.0
30 0 90 0
12 32

Persamaan
Smedley
untuk
posisi
0.5 sadel
1.7 pada
0.7 3 chord:
SCFAX / T ,Y 6.78 6.42
sin

2
SCFOPB / T ,Y 1.6 1.15 5 sin 1.35

Dari hasil eksperimen menunjukkan bahwa bahan baja mencapai

kelelahan ditentukan oleh besaran beda tegangan maksimum terhadap
minimum yang berulang-ulang seperti dinyatakan pada kurva S-N.
B-4

2. Persamaan SCF dari Naess (sambungan tubular Y)

K.Ks = SCF at chord saddle
K.Kc = SCF at chord crown
K.Kb = SCF at brace
K = Weld correction factor = 0.794
Gambar B.2. Sambungan tubular tipe Y

K c 0.75. 0.6 . 0.8 .(1.6. 0.25 0.7. 2 ). sin (1.5 1.6. )

Untuk Aksial:
K s . . .(6.78 6.42.

1/ 2

). sin

(1.7 0.7. 3 )

K b 1 0.63.K c

K c K 'c K o .K "c
K 'c (0.7 1.37.

Untuk In-plane Bending :

0.5

. .(1 )).(2. sin

0.5

sin
.( 0.5. / ).( / 2 / sin ). sin
Ko
1 1.5 /

Untuk Out-plane Bending :

2

Ks . . .(1.6 1.15. 5 ). sin (1.35 )

K b 1 0.63.Ks

K "c 1.05 1.30. 1.5 .(1.2 ).(cos 4 0.15)

K b (1 0.63.K c or s )

B-5

3. Persamaan Kuang

Persamaan Kuang untuk parameter dengan batasan sbb:

7 40
0.2 0.8
0 .3 0 .8
0.02 1.0
8.3 33.3
30 0 90 0

Untuk Chord:
3
SCFAX / T ,Y 1.981 0.057 1.2 0.08 1.33 sin 1.694
SCFIPB / T ,Y 0.702 0.4 0.6 8.860 sin 057
SCFOPB / T ,Y 1.020 0.787 1.014 0.889 sin 1.557
SCFOPB / T ,Y 0.462 0.619 1.014 0.889 sin 1.557

Untuk Brace:
3
SCFAX / T ,Y 3.751 0.120 1..35 0.550 1.33 sin 1.94
SCFIPB / T ,Y 1.301 0.23 0.6 0.38 sin 021
SCFOPB / T ,Y 1.522 0.801 0.852 0.543 sin 2.033
SCFOPB / T ,Y 0.796 0.281 0.852 0.543 sin 2.033
B-6