Dynamic Fatigue Assessment of Fixed Offshore Platform

Azin Azarhoushang and Hamid Nikraz

Curtin University of Technology
Perth, Australia
ABSTRACT

In this paper a practical method for dynamic fatigue assessment of
jacket-type offshore structure is investigated. Fatigue is a primary mode
of failure for steel structures that are subjected to dynamic loads. For
offshore structures the main source of fatigue inducing stress range are
ocean waves. The spectral approach to fatigue analysis is an attempt to
account for the random nature of a confused sea in a rational manner.
An accurate procedure for the random vibration computation of
structure is developed using frequency domain techniques. To illustrate
the aforementioned methodology, a jacket type offshore platform in the
Persian Gulf has been selected.

KEY WORDS: dynamic-response; jacket-type; offshore-platform;
fatigue assessment; frequency domain.

INTRODUCTION

The primary structural components of jacket type offshore structures
including topsides, jacket, piles and the surrounding soil are considered
using SACS software (refer to Fig. 1). For Platform description and
environmental data refer to Table 1.

Spectral-based Fatigue Analysis is a complex and numerically intensive
technique. The method is most appropriate when there is a linear
relationship between wave height and the wave-induced loads, and the
structural response to these loads is linear. Adaptations to the basic
method have been developed to account for various non-linearity.

A frequency domain-spectral method is based on linear concept
(transfer function) and clearly able to reflect the random nature of the
wave loading via the combination of the structural transfer functions
with a wave spectrum. In the spectral method a relationship to
characterize the expected energy in individual sea states is employed
(such the Pierson-Moskowitz or J ONSWOP spectral formulations);
with a scatter diagram that describes the expected long-term
probability of occurrence information for sea-states at a platforms
installation site. The suitability of a method as a practical assessment
tool is investigated as a relevant response process.










Table 1. Platform description and environmental data



























Fig. 1 General view of SACS structural model
Overall J acket Size at work point (EL 7.3 m) 18 (m)x36 (m)
Overall J acket Size at mudline (EL -41.3 m) 30.15 (m)x36 (m)
Topside Deck dimension (EL 35.5 m) 41.2 (m)x23.2 (m)
Topside Deck dimension (EL 13.3 m) 51.6 (m)x23.2 (m)
Topside Operating Weight 1140 (MT)
Water Depth 41.6 m
1142
Proceedings of the Twenty-second (2012) I nternational Offshore and Polar Engineering Conference
Rhodes, Greece, June 1722, 2012
Copyright 2012 by the I nternational Society of Offshore and Polar Engineers (I SOPE)
ISBN 978-1-880653-944 (Set); ISSN 1098-6189 (Set)
www.isope.org


TRANSFER FUNCTION

The transfer function (frequency response function) is the amplitude of
the sinusoidal response when the excitation is a sinusoid of unit
amplitude.The main task of a spectral fatigue analysis is the
determination of the stress range transfer function which expresses the
relationship between the stress at a particular structural location per
unit wave amplitude and wave of frequency and heading.

For fatigue analysis of offshore structures, the excitation is the
elevation of the seas surface at a point as a function of time, and the
responses of interest or stationary random function are the hot spot
stress ranges at the connections. Stress range is defined as the
difference between successive maximum and minimum peaks in the
plot of stress vs. time.

To generate a transfer function for a particular fatigue case (wave
direction), several waves of various heights but constant steepness are
used to load the structure. To achieve this, waves of constant steepness
is considered. Wave steepness is a simple relation between wave height
(H) and wave length (L), refer to equation 6. Typically, wave
steepnesss of 1:15, 1:20 and 1:25 are considered. However to select
the actual wave height which causes maximum damage an appropriate
calibration procedure is adopted to derive a design constant steepness.

In general, a spectral analysis should be used to properly account for
the actual distribution of wave energy over the entire frequency range.
Typically, the user need not generate loading for all possible stress
ranges. It is necessary to select only the loading required to yield an
accurate and sufficiently detailed transfer function.

In order to generate the peaks and valleys of the transfer function
accurately, a group of waves with frequencies around the primary and
secondary modes of the structure should be selected. For the wave
response analysis, initial frequency grid is modified in the vicinity of
the first natural period and the revised frequency grid is considered
which adequately describes the frequencies of the platform,
cancellation and enhancement effects and peak and valley period of
transfer function curve as shown in Fig. 2.


Fig. 2 Selection of Frequencies for Detailed Analyses

It is demonstrated that revised frequency grid produce similar transfer
function to the one developed for calibration purposes and no peaks on
the transfer functions are left out (refer to Fig. 3).







Fig. 3 static transfer function plot

CALIBRATION METHOD

The calibration is based on matching of global response parameters.
For matching purposes the base shear wave load is commonly used.
The seastate which contribute most to fatigue damage is considered.
For this purpose the center of fatigue damage scatter diagrams along
each direction are considered, as shown in Table 2.

Table 2. Directional wave scatter diagram 0 degrees

TOTAL 0 1 100 1968 1555 619 80 5 0 0 4328
4.25 1 1
3.75 1 1
3.25 1 1 2
2.75 5 1 6
Hs 2.25 6 34 40
1.75 1 165 39 1 206
1.25 1 246 394 1 642
0.75 705 1141 54 1900
0.25 1 100 1262 167 1530
0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5
Tp 4328


Transfer functions are developed for the wave steepness related to the
weighted center of damage in the scatter diagram. The weighing factor
is the mean value of the normalised static base shear transfer function
over the width of each cell in the scatter diagram. This weighing factor
introduces a stress period dependency which allows the cancellation
effects to be reflected in the fatigue damage scatter diagram. The
weighted damage in each cell of the scatter diagram is given as;


(1)


Where D
ij
is the damage in each cell of the fatigue damage scatter
diagram, W
j
is the period dependant weighing factor. From the
weighted fatigue damage calculations, the wave height (H
c
) and wave
period (T
c
) at corresponding center of fatigue damage scatter diagram is
calculated by the following equations:


H
c
=D
i
H
si
/D
i
(2)

T
c
=D
i
T
pi
/D
i
(3)

)
1 j
T
j
(T * 0.5
m
)]
1 i
H
i
(H * 0.5
j
[W
*
ij
N
ij
D
+
+
+
+
=
1143



From these values, the most probable maximum wave in the seastate at
the center of damage of the scatter diagram is derived as below:

H
ref
=1.86 H
c
(eg 3.754 for 0 deg)

(4)

T
ref
=T
c
(eg 5.821for 0 deg) (5)


This wave period is compared with the static base shear transfer
function for the direction of interest to confirm that it does not fall in
the valley of the transfer function. In case if it falls in the valley, the
same is shifted to the adjacent peak which is 3.75 sec depending on
wave energy distribution as shown below.



Fig. 4 Static transfer function


To account for the sensitivity of the structure, typically a broadside,
end-on and diagonal direction are considered for calibration and a
single representative steepness (In this case an average design steepness
of 1:16 is used) chosen from these. The wave steepness thus
determined is used to establish the wave height to be used with each
wave frequency.
Use of constant wave steepness will give unrealistically large wave
heights at small wave frequencies. Therefore a maximum height equal
to the wave height with a one year return period and ending with a
period corresponding to a wave height of 1.0 ft (0.33 m) should
normally be used (API, 2007). The wave height to be considered for the
analysis and design is calculated from the following equation for metric
units:

L =1.56xT
2
(6)
H =L x steepness (7)

Root mean square (RMS) base shear wave loads are calculated for the
seastate at the centre of the fatigue damage scatter diagram, using
standard spectral analysis techniques. For this SACS (EDI, 2011)
modules are used by making the mass equal to zero (to eliminate
inertial loads) and fully damping the structure (to make a low forcing
frequency) and hence the response calculated is the static response of
interest. This will provide an RMS base shear value for each wave
steepness and direction as shown below.







Fig. 5 Response function base shear

A stationary process is one whose statistics do not change with time.
For every RMS stress there exists an average time, Tz, between zero
crossings with a positive slope for a stationary Gaussian process with
zero mean. This period called the Zero Crossing Period. For a narrow
band process this is the average period or the reciprocal of the average
frequency of the process.

Assuming zero mean and a Rayleigh distribution of peaks, the most
probable maximum [MPM] base shear wave load is found using the
following equation (Zentech, 2000):

MPM
RMS
=[2ln
T
T
z
]
0.5
(8)

Where T is the 3 hours duration of the storm in seconds (3 x 60 x 60 =
10800 sec) and Tz is the mean zero crossing period of wave in the
seastate at the center of damage of fatigue scatter diagram. This will
provide the maximum base shear wave load from the spectral
calculation for each wave steepness and direction. The corresponding
base shear wave load range is double the MPM base shear wave load.

Wave power spectra can be generally classified into two characteristic
forms. These are related to either the fetch-limited, developing sea
conditions without swell, or the fully developed seas, and can be
reasonably described by the J ONSWAP and the Pierson-Moskowitz
spectra, respectively. The shape of the stress (strain) power spectra is
important because corrosion crack growth of the structural steels in
tubular joints is sensitive to frequency content. Therefore, it is
necessary to simulate the correct frequency content, and hence power
spectra, of typical wave loading (Kam, 1990). For seastates described
by a J ONSWAP spectrum, Tz is assumed to be equal to Tp/1.28.

The spectral base shear is matched against the deterministic base shear.
The drag and inertia coefficients are increased to match the spectral
base shear range with the deterministic base shear range as per
following procedure.
1144



Since the dynamic module in SACS uses the linear theory (i.e. modal
superposition), linearized foundation super elements are automatically
created at each pile-head by the PSI program. To linearise the
foundation for evaluation of fatigue damage, pile stubs of appropriate
stiffness are modeled.

An equivalent solution to pilesoil interaction is considered using the
pile stub to account for finite deflections of the pile (the P-delta effect).
For derivation of the pile stubs, the most probable maximum wave
corresponding to the center of fatigue damage scatter diagram is
considered.

A normal inplace analysis with these waves is carried out to obtain the
pilehead forces. Using these forces, pile stub properties is determined
by using SINGLE PILE module in SACS suite of programs. Different
load cases are used to calculated stiffness in each orthogonal direction
as well as the diagonal direction as below:

Table 3. Centre of fatigue damage wave

Wave 0 deg 45 deg 90 deg
Href = 1.371 2.438 2.438 m
Tref = 3.750 5.000 5.000 sec
Tz = 2.93 3.91 3.91 sec


Values of Hc and Tc are used to determine the static RMS (kN) values
for the transfer function plots (refer to Fig 5) along the three wave
directions. The values are tabulated as below:


Table 4. RMS values

Wave Wave Steepnesss
Direction 1:25 1:20 1:15
0
0
23.987 23.900 23.658
45
0
23.655 23.212 22.383
90
0
89.360 88.930 88.381


For above values of Tz and RMS, MPM in kN is given by the equation
8 as below:

Table 5. MPM values

Wave Wave Steepnesss
Direction 1:25 1:20 1:15
0
0
97.213 96.860 95.879
45
0
94.174 92.411 89.110
90
0
355.756 354.044 351.858


Spectral Base Shear Range is given as 2MPM for each wave direction
and wave steepness in the Table 6:




Table 6. Spectral base shear

Wave Wave Steepnesss
Direction 1:25 1:20 1:15
0
0
194.426 193.720 191.758
45
0
188.348 184.822 178.220
90
0
711.512 708.088 703.716


Spectral Base Shear is compared against the deterministic base shear
range (272.626kN along 0 deg, 192.458kN along 45 deg and
696.324kN along 90 deg). In order to linearise the wave effects, the
drag and inertia coefficients are enhanced by the ratio of deterministic
base shear to spectral base shear, refer to the Table 7 and 8:

Table 7. Enhanced drag coefficient

Cd
Cd for Smooth
members (initial =0.6)
Cd for Rough members
(initial =0.7)
Wave Wave Steepness Wave Steepness
Direction 1:25 1:20 1:15 1:25 1:20 1:15
0
0
0.841 0.844 0.853 0.982 0.985 0.995
45
0
0.613 0.625 0.648 0.715 0.729 0.756
90
0
0.587 0.590 0.594 0.685 0.688 0.693
Cd 0.853 0.995

Table 8. Enhanced mass coefficient

Cm
Cm for Smooth
members (initial =2.0)
Cm for Rough members
(initial =2.0)
Wave Wave Steepness Wave Steepness
Direction 1:25 1:20 1:15 1:25 1:20 1:15
0
0
2.804 2.815 2.843 2.804 2.815 2.843
45
0
2.044 2.083 2.160 2.044 2.083 2.160
90
0
1.957 1.967 1.979 1.957 1.967 1.979
Cm 2.843 2.843


Enhancement and Cancelation wave period (T
ep
and T
cp
) correspond
with wave lengths which are specific multiples or a fraction of key
dimensions such as the main leg spacing, are given by the following
equations respectively:
T
ep
d
1.56n
:=
(9)
T
cp
d
1.56 n 0.5 ( )
:=
(10)
Where d is the leg spacing at work point in metric unit and n =1, 2, 3

Using above formulas, the initial frequency grid considered for
calibration is modified with additional wave steps to account for
enhancement and cancelation effects as follows:

10 steps from 10.0 to 5.0 seconds at an interval of 0.5 seconds.
4 steps from 5.0 to 4.0 seconds at an interval of 0.25 seconds.
20 steps from 4.0 to 3.0 seconds at an interval of 0.05 seconds.
21 steps from 2.5 to 1.0 seconds at an interval of 0.1 seconds.
1145

DYNAMIC ANALYSIS

With the spectral analysis method, dynamic response can be included
by using a frequency domain dynamic solution instead of a quasi-static
solution, for the structural analyses used to determine member nominal
stresses. The steady state dynamic response to each of the regular
waves stepped through the structure is thus included in the resulting hot
spot stress range transfer functions.

In order for the frequency domain formulation and the associated
probabilistically based analysis to be valid, load analysis and the
associated structural analysis are assumed to be linear. Hence scaling
and superposition of stress range transfer functions from unit amplitude
waves are considered valid.

The linearised (Airy) wave theory is adopted to model the wave
elevation, water particle velocity and accelerations. This modeling is,
however, limited by the linearity of wave component superposition
assumptions and does not include nonlinear/non- Gaussian wave
characteristics. While since the fatigue damage accumulation
experienced by jackets is mainly caused by wave loadings due to small
and medium sized waves, this linear assumption can then normally
yield sufficiently accurate wave loading calculation results (J ia, 2008).

DYNAMIC CHARACTERISTICS

After calibrating the spectral response, dynamic characteristics (mode
shapes and frequencies) of the structure are generated with reduced
structural stiffness matrix and consistent mass approach. When dealing
with very large structures (particularly lattice structures) the technique
of mass condensation may be used (Guyan et al, 1965).

A set of master (retained) degrees of freedom are selected at each
horizontal elevation on the main legs to extract the Eigen values
(periods) and Eigen vectors (mode shapes), which includes all stiffness
and mass properties related to the reduced degrees of freedom. After
the modes are extracted using the master degrees of freedom, they are
expanded to include the full 6 degrees of freedom for all joints in the
structure. A consistent mass approach is considered since it is more
desirable for structures immersed in the fluid. The added mass is
generated automatically by SACS and depends on the size, orientation
and proximity of the member to the free surface.

The eigenvalue parameter and mass participation factor for the first 10
mode shapes are given in Table 9 and 10 respectively as the results
from the dynamic analysis. The mass participation factor should be
above 90% in at least two lateral directions.

Table 9. Frequencies and generalized mass



Table 10. Mass participation factor

MODE X Y Z X Y Z
1 0.0000003 0.8399167 0.000003 0 0.839917 0.000003
2 0.8884136 0.0000001 0.0000023 0.888414 0.839917 0.000005
3 0.0001832 0.0000478 0.0000009 0.888597 0.839965 0.000006
4 0.0000361 0.0095535 0.0003985 0.888633 0.849518 0.000405
5 0.0000149 0.1432379 0.0000059 0.888648 0.992756 0.000411
6 0.0934848 0.000001 0.0002555 0.982133 0.992757 0.000666
7 0.0001383 0.0000036 0.0000218 0.982271 0.992761 0.000688
8 0.0042 0.0004932 0.0005601 0.986471 0.993254 0.001248
9 0.0040244 0.0001033 0.5436326 0.990496 0.993357 0.544881
10 0.0040383 0.0000593 0.4086777 0.994534 0.993417 0.953558


Entrapped mass is calculated for members designated as flooded in the
model file. Hydrodynamic effects of marine growth are included in the
program to account for the density and effective diameter due to marine
growth. Mode shape for the first lateral mode of vibration is presented
below.










































Fig. 6 First mode shape of the structure
MODE
FREQ.
(CPS)
GEN.
MASS EIGENVALUE
PERIOD
(SECS)
1 0.297 5.78E+03 2.88E-01 3.37
2 0.526 4.99E+03 9.17E-02 1.90
3 0.728 3.81E+03 4.78E-02 1.37
4 1.479 1.26E+02 1.16E-02 0.68
5 1.547 4.55E+02 1.06E-02 0.65
6 1.645 2.61E+03 9.36E-03 0.61
7 2.037 2.58E+03 6.10E-03 0.49
8 2.434 8.12E+03 4.28E-03 0.41
9 2.658 3.43E+03 3.59E-03 0.38
10 2.690 5.61E+03 3.50E-03 0.37
1146




































Fig 7. Fatigue life of the joints (years) with calibration


CONCLUSIONS

Fatigue degradation has been a main cause of damage to fixed offshore
platforms, resulting in a consequent reduction in overall structural
integrity. When the Fatigue Demand and Fatigue Strength are
established, they are compared and the adequacy of the structural
component with respect to fatigue is assessed using a Damage
Accumulation Rule and a Fatigue Safety Check. This approach is
referred to as the Palmgren-Miner Rule. Calculated Fatigue Life is the
computed life, in units of time (or number of cycles) for a particular
structural detail considering its appropriate S-N curve (ABS, 2010).
It can be clearly seen that the fatigue life is significantly higher (about
four times) ignoring calibration process, refer to Fig 7 and 8 above.
This means that for platforms with the period above 3 seconds (refer to
Table 9) the effects of dynamics and structural response which is
incorporated through calibration process become more important and
must be adequately modeled to ensure that realistic results were
obtained.

ACKNOWLEDGEMENTS

The authors would like to thank WorleyParsons for the support
provided to use SACS software during the research reported in this
paper.





































Fig 8. Fatigue life of the joints (years) ignoring calibration


REFERENCES

ABS Fatigue assessment of offshore structures, American Bureau of
Shipping (2010)

API RECOMMENDED PRACTICE 2A-WSD (RP 2A-WSD)
Recommended Practice for Planning, Designing and Constructing
Fixed Offshore Platforms-Working Stress Design, American
Petroleum Institute (2007)

EDI. SACS Users manual, Engineering Dynamics Inc. (2011)

Guyan, R. J . (1965), Reduction of stiffness and mass matrices

J ia, J . An efficient nonlinear dynamic approach for calculating wave
induced fatigue damage of offshore structures and its industrial
applications for lifetime extension, Appl Ocean Research 30
(2008) 189-198

Kam, J .C.P. Wave action standard history (wash) for fatigue testing
offshore structures, Appl Ocean Research 14 (1992) 1-10

Zentech Inc. Validation of wave response analysis for jack-up rigs,
Health and Safety Executive (2000)


1147