You are on page 1of 6

Nonlinear Effects on Fatigue

Analysis for Fixed Offshore


Structures
Fatigue analysis for fixed offshore structures is an important practical issue. These struc-
tures are often drag dominated, which makes the deck response a non-Gaussian process
when it is assumed that the irregular waves are Gaussian. Incorporating nonlinear and
Oleg Gaidai non-Gaussian modeling in the fatigue analysis can be a complicated issue, cf. work of
Centre for Ships and Ocean Structures, Madhavan Pillai and Meher Prasad [2000, “Fatigue Reliability Analysis in Time Do-
Norwegian University of Science and Technology, main for Inspection Strategy of Fixed Offshore Structures,” Ocean Eng., 27(2), pp. 167–
NO-7491, Trondheim, Norway 186]. The goal of this paper is to provide evidence that for drag dominated offshore
structures it is, in fact, sufficient to perform linearization in order to obtain accurate
Arvid Naess estimates of fatigue damage. The latter fact brings fatigue analysis back into the Gauss-
Centre for Ships and Ocean Structures, ian domain, which facilitates the problem solution. Beyond straightforward linearization
Norwegian University of Science and Technology, of the exciting wave forces, this paper employs two different approaches accounting for
NO-7491, Trondheim, Norway; nonlinear effects in fatigue analysis. One is an application of the quadratic approxima-
Department of Mathematical Sciences, tion approach described in the work of Naess and co-workers [1997, “Frequency Do-
Norwegian University of Science and Technology, main Analysis of Dynamic Response of Drag Dominated Offshore Structures,” Appl.
NO-7491, Trondheim, Norway Ocean. Res., 19(3), pp. 251–262; 1996, “Stochastic Response of Offshore Structures
Excited by Drag Forces,” J. Eng. Mech., ASCE, 122, pp. 155–160]. to the stochastic
fatigue estimation of jacket type offshore structures. An alternative method proposed is
based on a spectral approximation, and this approximation turns out to be accurate and
computationally simple. The stress cycles causing structural fatigue are considered to be
directly related to the horizontal excursions of the fixed offshore structure in random
seas. Besides inertia forces, it is important to study the effect of the nonlinear Morison
type drag forces. Since no direct method for dynamic analysis with Morison type forces is
available, it is a goal to find an accurate approximation, allowing efficient dynamic
analysis. This has implications for long term fatigue analysis, which is an important issue
for design of offshore structures. 关DOI: 10.1115/1.2904584兴

Introduction ticular quadratic approximation 关1,2兴 is used. The present study


shows that the mentioned approximation is quite accurate, mean-
An important element in the design of jacket offshore structures
ing that it is a sound approach to substitute the drag forces by this
with a significant dynamic behavior is the estimation of fatigue
quadratic approximation. A major advantage of the quadratic ap-
life. The simplest approach is to restrict the study only to linear
proximation of the drag forces is that direct dynamic analysis
analysis, retaining linear, first order forces and linearizing the non-
becomes available, and thus fatigue analysis as well.
linear ones. An alternative is Monte Carlo 共MC兲 simulations, but
Fatigue analysis is well developed for narrow-band processes.
this is a very time consuming method for long term fatigue analy-
However, the horizontal deck response is not a truly narrow-
sis. It is therefore of interest to employ accurate approximations,
banded process. Detailed fatigue analysis of a non-narrow banded
taking into account nonlinear effects and allowing fast dynamic
process is quite complicated and it is not a topic of the present
analysis. A Morison type force model introduces a second order
paper. Several simplified approaches to account for bandwidth
nonlinearity, in addition to the linear inertia term. A preliminary
effects have been proposed over the years, see, e.g., Refs. 关3–5兴.
consideration indicates that higher order nonlinear effects, e.g.,
An interesting linearization technique for fatigue estimation of
inundation and nonlinear wave components, are of minor impor-
offshore jackets was presented in Ref. 关6兴.
tance for fatigue calculations.
However, since the focus of the present paper is on comparison
The purpose of the present paper is to compare the contribution
of results, the application of these methods would not change the
to long term fatigue of nonlinear effects due to Morison type
conclusions. Therefore, a narrow-band fatigue analysis is applied
forces, relative to a linearized approach. The purpose of the
here.
present paper is to confirm that linearization is accurate enough
The results of this study show that there is no significant dif-
for fatigue calculation for drag dominated offshore structures.
ference between the calculated long-term fatigue based on a lin-
First, a method that fully accounts for nonlinear effects due to
earized analysis versus a nonlinear one. Since the effect of the
Morison type forces is studied. Next, a comparison with the re-
nonlinear drag force term is largely to amplify the response peaks,
sults obtained by linearization approaches is done.
this suggests that also for a more correct wide-band fatigue analy-
Since there is no exact method available for direct dynamic
sis this difference will be small. The latter is an important practi-
analysis of the linear system response due to drag forces, a par-
cal conclusion for the fatigue design of jacket structures, since it
contributes to justify linearization, if it is done properly.
Contributed by the Ocean Offshore and Arctic Engineering Division of ASME for
publication in the JOURNAL OF OFFSHORE MECHANICS AND ARCTIC ENGINEERING. Manu-
script received July 10, 2006; final manuscript received October 9, 2007; published Morison Force Approximation
online July 18, 2008. Assoc. Editor: John M. Niedzwecki. Paper presented at the 25th
International Conference on Offshore Mechanics and Arctic Engineering Hydrodynamic forces acting on the structure are assumed to be
共OMAE2006兲, Hamburg, Germany, June 4–9, 2006. distributed over discrete nodes, located from the mean water level

Journal of Offshore Mechanics and Arctic Engineering AUGUST 2008, Vol. 130 / 031009-1
Copyright © 2008 by ASME

Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


␣ = 5.06 冉 冊
Hs
T2p
2
共1 − 0.287 ln ␥兲 共4兲

Hs = significant wave height 共m兲 and T p = 2␲ / ␻ p spectral peak


wave period 共s兲.
The power spectral density 共PSD兲 of the velocity U共t , 0兲 at the
mean water level is given by
SU共␻,z = 0兲 = ␻2S␩共␻兲 共5兲
For linear wave theory and unidirectional waves propagating in
the positive x-direction, which is assumed here, the linear transfer
function in the frequency domain from U共␻ , 0兲 to U共␻ , x兲 is given
by
cosh关k共z + d兲兴 −ikx
T共␻,x兲 = e 共6兲
cosh关kd兴
where d denotes water depth, and k = k共␻兲 is determined by a
dispersion relation. The same transformation applies to the inertia
force Fin共t , x兲. The relation between the water particle velocity
PSDs can be easily obtained from Eq. 共6兲 as follows:

SU共␻,z兲 = 冉 cosh关k共z + d兲兴


cosh关kd兴
冊 2
SU共␻,z = 0兲 共7兲

In this paper, the following deep water approximation is adopted:


cosh关k共z + d兲兴
⬇ ekz, z艋0 共8兲
cosh关kd兴
As it is seen, U共t , x兲 is linearly related to the wave elevation
process ␩共t兲. Since ␩共t兲 is a zero-mean stationary Gaussian pro-
Fig. 1 Sketch of jacket structure and nodal representation cess, U共t , x兲 is then also stationary and Gaussian. The no current
assumption leads to E关U共t , x兲兴 = 0.
Since no exact statistical analysis is available for the dynamic
response of a linear system due to Morison type forces, its qua-
to the sea bottom. This approach is known as a lumped parameter dratic approximation elaborated in Refs. 关1,2兴 is applied as
model. Figure 1 depicts the Kvitebjørn jacket platform, which follows:
could represent the structure to be investigated.
Inundation effects, i.e., drag force corrections due to varying Fd共t,x兲 ⬇ Fa共t,x兲 = kd„P共t,x兲2 − N共t,x兲2… 共9兲
water surface elevation with time, are neglected. Horizontal ex- where an auxiliary, zero-mean Gaussian process P共t , x兲 is intro-
cursions of the structure are studied; therefore, only the horizontal duced, corresponding in a sense to the water particle velocity
x-components of the hydrodynamic forces are considered. The process U共t , x兲. The process N is the Hilbert transform of P. The
standard engineering model of the hydrodynamic force F共t , x兲 at a
PSD of the processes P共t , x兲 and N共t , x兲 is chosen such that the
given location x = 共x , y , z兲 with a non-negligible drag force com-
PSDs of Fd共t , x兲 and Fa共t , x兲 are well matched. The matching
ponent is the Morison formula 关7兴
approach is described in Ref. 关1兴. It is shown in Refs. 关1,2兴 that
the statistics of Fa is similar to that of Fd. This suggests that the
F共t,x兲 = Fin共t,x兲 + Fd共t,x兲 共1兲 response statistics will also be similar. Further on in this paper,
Monte Carlo simulations shall confirm the accuracy of this ap-
where proximation for the response process.
Let us denote the PSD of the process P共t , x兲 by S共␻ , x兲, chosen
Fin共t,x兲 = kmU̇共t,x兲, Fd共t,x兲 = kdU共t,x兲兩U共t,x兲兩 共2兲 such that the PSDs of Fd共t , x兲 and Fa共t , x兲 are matched, where
Fd共t , x兲 is given by Eq. 共2兲 and Fa共t , x兲 is given by Eq. 共9兲.
Zero current is assumed. U共t , x兲 denotes the water particle veloc- First, the PSD of P共t , 0兲 is estimated with the mean water level
ity in the horizontal x-direction. Fin is the linear inertia force, at the origin x = 0. This can be done by a simple trial and error
while Fd is the nonlinear drag force. km and kd are suitably chosen procedure, cf. Ref. 关1兴. Matching at underwater nodes can be done
parameters to be specified below. The random stationary sea state automatically as an iterative procedure. The corresponding drag
is specified by a JONSWAP spectrum 关8兴, which is given for the force approximation can be obtained from Eq. 共9兲 as follows:
wave elevation ␩共t兲 as follows:
n n

再 冉冊 冋 冉 冊 册冎 兺 兺 R共␻ ,− ␻ ,x兲B B*e i共␻i−␻ j 兲t


Fa共t,x兲 = kd 共10兲
␣g 2
5 ␻p 4
1 ␻ 2 i j i j
S␩共␻兲 = 5 exp − + 共ln ␥兲exp − 2 −1 i=−n j=−n
␻ 4 ␻ 2␴ ␻ p
where
共3兲
R共␻i,− ␻ j,x兲 = 冑S共␻i,x兲S共␻ j,x兲⌬␻兵1 − ␹共␻i兲␹共␻ j兲*其 共11兲
where g = 9.81 ms−2, ␻ p denotes the peak frequency in rad/s, and
where ␹ is the Hilbert transform function, that is, ␹共␻兲 = −i for
␣, ␥, and ␴ are parameters related to the spectral shape. ␴ = 0.07
when ␻ 艋 ␻ p, and ␴ = 0.09 when ␻ ⬎ ␻ p. The parameter ␥ is cho- ␻ ⬎ 0, ␹共0兲 = 0, and ␹共␻兲 = i for ␻ ⬍ 0. The set 兵B j其nj=−n consists of
sen to be equal to 2.0. The parameter ␣ is determined from the complex Gaussian variables B j = ␣ j + i␤ j, such that 兵␣ j , ␤ j其nj=1 are
following empirical relationship: independent real Gaussian N共0 , 1 / 冑2兲 variables and B−j = B*j .

031009-2 / Vol. 130, AUGUST 2008 Transactions of the ASME

Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Dynamic Response of the Jack-Up Structure Table 1 Representative sea states
A commonly used dynamic model is a lumped parameter model k Hs 共m兲 T p 共s兲 ck
of the following type:
1 1.5 9 0.27
MẌ + CẊ + KX = Fin + Fd 共12兲 2 2.5 9 0.29
3 3.5 10 0.32
Here, M, C, and K are suitable matrices. X = 共X1 , . . . , XN兲T where 4 11.5 16 0.001
Xk = Xk共t兲 , k = 1 . . . , N, denotes the displacement response in the
wave direction at Node No. k. Fin = (Fin共t , x1兲 , . . . , Fin共t , xN兲)T and
Fd = (Fd共t , x1兲 , . . . , Fd共t , xN兲)T, where xk denote the coordinates of
Nodal Point No. k. Each of the Fin共t , xk兲 and Fd共t , xk兲 must now be to different processes P and U, respectively. The processes P and
interpreted as locally integrated point forces. Using the method U are not connected by a simple linear relationship. Therefore, in
proposed in Ref. 关5兴, the structure is modeled as a linear single order to link them together in Eq. 共14兲, we assume that they are
degree of freedom 共SDOF兲 system. The motion is considered to be independent sets of random variables. This assumption is justified
horizontal with a dominant mode shape given by the expression by the fact that the random variables U̇共t , z兲 and U共t , z兲 兩 U共t , z兲兩
␺共z兲 = 1 − cos关␲共z + d兲/2L兴, −d艋z艋L−d 共13兲 are independent variables, since U共t , z兲 is a Gaussian variable in-
dependent of U̇共t , z兲. This, however, does not mean independence
where L is the length of each platform leg, see Fig. 1. Further
of these stochastic processes. In order to check the accuracy of
simplifications are introduced by assuming that the platform is
this assumption, results of Monte Carlo simulations will be pre-
symmetrical about the xz-plane, and that the phase difference be-
sented in this paper, indicating that this introduces only minor
tween forces at the fore and aft legs can be neglected, which
differences for the response statistics.
implies that the only spatial variable to appear in subsequent ex-
pressions is z. These simplifications are introduced just to reduce Calculation of Fatigue
the complexity of expressions, and are not limitations of the
theory. It is also believed that they will not impair the ability to For the calculations, the following parameters for the linear
capture the effects of the force modeling on the fatigue life. dynamic system are chosen:
Introducing the generalized coordinate Z = Z共t兲, that is, Xk — resonance period Te = 5 s
= Xk共t兲 = Z共t兲␺共zk兲, one obtains the following SDOF equation of — relative damping ␰ = 10%. The response process Z is non-
motion: dimensionalized by a suitable constant ␮ from Eq. 共15兲.
For the long-term fatigue analysis, four representative sea
Z̈ + 2␰␻eŻ + ␻2e Z = m̃−1共F̃in + F̃d兲 共14兲
states from Table 1 are chosen.
where m̃, ␰, and ␻e denote modal mass 共including added mass
contributions兲, modal damping ratio, and undamped modal fre- The coefficients ck, k = 1 , . . . , 4 in Table 1 express the relative
quency, respectively; F̃a = F̃a共t兲 is given by the equation contributions of the particular sea states chosen. Specifically, ck is
the number of sea states within the band close to the representa-
N
tive kth sea state, divided over the total number of sea states. The
F̃d共t兲 ⬇ F̃a共t兲 = ␮ 兺 F 共t,z 兲␺共z 兲⌬z
k=1
a k k k 共15兲 full table of sea states is given in Ref. 关9兴. It is seen that moderate
sea states dominate, while the fraction of severe sea states is rather
where ␮ is a scaling factor, needed to make the response nondi- small.
The calculation of the fatigue damage is based on the
mensional. The expression for F̃i = F̃i共t兲 is entirely similar. Ap- Palmgren–Miner approach. Assuming an effectively narrow-band
proximation 共15兲 is performed pointwise, i.e., at each discrete response process for each stationary short-term sea state, the total
depth location zk, which means that accumulated fatigue damage rate can be expressed as
Fd共t,zk兲 = kdU共t,zk兲兩U共t,zk兲兩 ⬇ Fa共t,zk兲 = kd„P共t,zk兲2 − N共t,zk兲2…


f A共a兲
共16兲 DF = ␯Z+共0兲 da 共21兲
0
N共a兲
The following quadratic approximation is obtained:
where ␯Z+共0兲 = the mean zero upcrossing rate of Z共t兲, f A共a兲 = the
n n
PDF of displacement peaks 共amplitudes兲 relative to the mean re-
F̃a共t兲 = 兺 兺 G B B *e
i=−n j=n
ij i j
i共␻i−␻ j 兲t
共17兲 sponse, which is assumed equal to zero, and N共a兲 = the number of
cycles to failure corresponding to response range equal to 2a.
where Invoking the damage law, this can be expressed as


N ⬁

Gij = ␮kd 兺 R共␻ ,− ␻ ,z 兲␺共z 兲⌬z


k=1
i j k k k 共18兲 DF = c̃␯Z+共0兲2m
0
am f A共a兲da 共22兲

The exact calculation of the peak PDF f A共a兲 is generally difficult,


An analogous linear representation can be written for F̃in共t兲 as
but with the assumption of a narrow-band stationary response, the
n following equation can be assumed:
F̃in共t兲 = 兺 A B̂ e j j
i␻ jt
共19兲 1 d␯Z+共a兲
j=−n f A共a兲 = − 共23兲
␯Z+共0兲 da
where
N where ␯Z+共a兲 = the mean a-upcrossing rate of Z共t兲. If this function
can be determined, Eq. 共23兲 can be used to approximate the peak
A j = ␮km 兺 i␻ 冑S
k=1
j U共␻ j ,zk兲⌬␻␺共zk兲⌬zk 共20兲 PDF for fatigue calculation. An alternative approach is obtained
by assuming that Z共t兲 and Ż共t兲 are statistically independent, which
It is important to mention that the independent N共0 , 1兲 Gaussian is exact for a stationary Gaussian response. It then follows from
variables 兵B j其 and 兵B̂ j其 form different sets, since they correspond the Rice formula 关10兴 that

Journal of Offshore Mechanics and Arctic Engineering AUGUST 2008, Vol. 130 / 031009-3

Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use




␯Z+共a兲 = sf ZZ˙共a,s兲ds = E关Ż+兴f Z共a兲 共24兲 1
0

where Ż+ = max共0 , Ż兲, f ZZ˙共· , · 兲 = the joint PDF of Z共t兲 and Ż共t兲, and 0.8

Non−dim drag force PSD


f Z共·兲 = the PDF of Z共t兲. Hence, as shown in Ref. 关11兴, under the
assumption of a narrow-band response process, it follows from
Eqs. 共22兲–共24兲 by partial integration that 0.6



c̃␯Z+共0兲 m 0.4
DF = 2 m am−1 f Z共a兲da 共25兲
f Z共0兲 0

0.2
where c̃ is a material constant, which can be neglected since DF is
made nondimensional.
The coefficient m is also a material constant. It is taken to be
equal to 4, which is typical for offshore steel. In fatigue calcula- 0
0 0.5 1 1.5 2
tions, the stress response is considered, while in this paper the ω rad/s
dynamic analysis was carried out for jacket horizontal response.
We do not differentiate between jacket horizontal response and its Fig. 2 PSD of the normalized integrated drag force „–…; wave
stress response, for the reason that they are well approximated by spectrum „---… is scaled for comparison
a linear proportion.
The response Z共t兲 is not a narrow-band process, for reasons
discussed in the Introduction, but we apply formula 共25兲 for com- km = Cm␳␲D2/4, kd = Cd␳D/2 共27兲
parison purposes. The advantage of the approximation is that fa- Here, ␳ is the density of sea water. The inertia and drag coeffi-
tigue calculation reduces to calculation of the response PDF f Z共·兲 cients Cm and Cd, respectively, can be considered to some extent
and the mean zero upcrossing rate ␯Z+共0兲, which is, in general, as empirical parameters depending on the surface roughness of the
much easier to determine than the full upcrossing rate function jacket. In this paper, it is assumed that Cd / Cm = 1 m. Figure 2
␯Z+共·兲. An accurate numerical method for calculating the PDF and presents the nondimensional normalized PSD of the integrated
the mean upcrossing rate for combined linear and quadratic re- drag force F̃d. As an illustration, the sea state peak period T p
sponse 共14兲 has been described in Refs. 关12–14兴. This method is = 16 s, and the resonance period Te = 5 s. This sea state is severe
fast and accurate, which is of major importance for long-term and is not representative for fatigue, where moderate sea states
analysis. dominate. It is chosen since drag effects are dominating compared
In Ref. 关15兴, a very similar study of the effect of nonlinear drag to the linear ones; therefore, it clearly reflects nonlinear effects.
forces on the long-term fatigue has been presented. Also in that For comparison, Fig. 2 also presents the wave spectrum at the
work, the response process is assumed to be narrow band. More mean water level, scaled to the graph dimensions. Precise calcu-
specifically, it is assumed that the stress ranges are Rayleigh dis- lation of the integrated drag PSD is done by using analytical re-
tributed. This amounts to an assumption of a Gaussian response sults elaborated in Ref. 关16兴. Figure 3 presents the nondimensional
process; hence, the effect of the nonlinear drag forces will only normalized drag response PSD.
enter through the stress range variance. This assumption is In this section, a pointwise approximation is performed, mean-
avoided in the present paper since the fatigue calculation is based ing that at each depth point zk of the nodal representation, a sepa-
on Eq. 共25兲, which accounts for the actual distribution of the rate approximation 共9兲 is done. Exact response statistics is ob-
response. tained by MC simulations, while for the approximation it is
For the long-term fatigue analysis, the total accumulated fatigue obtained by an application of the saddle point method 关12兴. First,
damage rate DF is calculated as follows: a moderate sea state with Hs = 3.5 m, T p = 10 s is considered. Table
2 presents a comparison between the response PDFs due to only
4

DF = 兺c D
k=1
k Fk 共26兲

1
where each DFk corresponds to a particular kth sea state. To test
the accuracy of the several approximations made in the previous
Non−dim drag response PSD

section, Monte Carlo simulations have been carried out for se- 0.8
lected sea states. For the latter purpose two sea states are chosen:
moderate 共Hs = 3.5 m, T p = 10 s兲 and severe 共Hs = 11.5 m, T p
= 16 s兲. 0.6
For the integration along the jacket leg, N = 31 nonequidistant
points were chosen, −d 艋 z j 艋 0 , j = 1 , . . . , N. The first node z1 is
located at the mean water level, which means z1 = 0, see Fig. 1. 0.4
Here, the water depth is d = 190 m. Nodes z j are spaced in geo-
metric progression z j+1 − z j = bq j, such that the node density de-
creases from the mean water level to the sea bottom. Here, b and 0.2
q are suitable constants, and the last node zN is at the sea bottom,
i.e., zN = −d.
The integration along the depth coordinate −d 艋 z 艋 0 was per- 0
0 0.5 1 1.5 2
formed using Simpson’s rule. The averaged jacket pipe diameter ω rad/s
is assumed to be D = 1 m. The inertia and drag force coefficients
are expressed, respectively, as Fig. 3 PSD of the normalized integrated drag response

031009-4 / Vol. 130, AUGUST 2008 Transactions of the ASME

Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Table 2 MC and Approx PDF, drag resp., moderate sea Table 5 MC and Approx PDF, full resp., severe sea

PDF argument MC PDF Approx PDF Lin PDF PDF argument MC PDF Approx PDF Lin PDF

0 1.6⫻ 100 2.0⫻ 100 1.1⫻ 100 0 1.6⫻ 10−1 1.5⫻ 10−1 1.4⫻ 10−1
0.5 2.7⫻ 10−1 2.1⫻ 10−1 1.2⫻ 10−1 3 6.8⫻ 10−2 6.4⫻ 10−2 8.0⫻ 10−2
1 4.6⫻ 10−2 4.0⫻ 10−2 2.5⫻ 10−2 6 1.5⫻ 10−2 1.5⫻ 10−2 1.5⫻ 10−8
1.5 8.8⫻ 10−3 8.1⫻ 10−3 2.2⫻ 10−1 9 3.7⫻ 10−3 4.0⫻ 10−3 9.0⫻ 10−4
2 1.8⫻ 10−3 1.7⫻ 10−3 3.0⫻ 10−7 12 1.0⫻ 10−3 1.2⫻ 10−3 1.8⫻ 10−5
2.5 3.9⫻ 10−4 3.6⫻ 10−1 5.7⫻ 10−11 15 3.1⫻ 10−1 3.6⫻ 10−4 1.1⫻ 10−7
3 8.1⫻ 10−5 8.0⫻ 10−5 18 1.1⫻ 10−1 1.1⫻ 10−1 2.4⫻ 10−10
21 2.2⫻ 10−5 3.5⫻ 10−5

drag forces Fd and Fa. The first column is the PDF argument, i.e.,
the value of nondimensional response Z, see Eq. 共14兲. The second
one “MC PDF” is an exact Monte Carlo simulated PDF, the third duce some errors. An alternative approach is to approximate the
one “Approx PDF” is the PDF calculated using the quadratic ap- integrated drag forces F̃d analogously to Eq. 共9兲. Specifically,
proximation described above, and the fourth column “Lin PDF”
represents the PDF of the linearized response. F̃d共t兲 ⬇ F̃a共t兲 = k̃d„P̃共t兲2 − Ñ共t兲2… 共28兲
Table 3 presents comparison between response PDFs due to full
forces, including both inertia and drag. where k̃d is a suitably chosen constant. This approach is attractive
It is seen from Tables 2 and 3 that the tails of the PDF are from a computational point of view, since then the quadratic ap-
underestimated. This, however, will turn out to have only a minor proximation is done only once. Dynamic analysis of the drag
influence on the fatigue, since it is not dominated by PDF tails. force in the previous section has shown that the set of eigenvalues
The zero upcrossing rate ␯Z+共0兲 appears to be the same for the MC of the integrated force F̃a consists of pairs of eigenvalues of op-
calculations and the saddle point method. Second, the severe sea posite sign, with one dominating pair. The latter also suggests the
state with Hs = 11.5 m, T p = 16 s is considered. In this case, drag integrated approximation approach. In the column “Approx PDF,”
forces are dominant. Table 4 presents a comparison between re- Table 6 presents results of the above discussed modified quadratic
sponse PDFs due to only drag forces Fd and Fa. Table 5 presents approximation, which turn out to be more precise compared to the
a comparison between response PDFs for the severe sea state due pointwise quadratic approximation from previous sections.
to full forces, including both inertia and drag. The linearization can also be modified similarly, using an inte-
It is seen from Tables 2–5 that quadratic approximation is much grated instead of a pointwise approach 共15兲 for the total drag force
more precise in PDF tails, while for the moderate arguments both F̃d. More specifically, if S̃d共␻兲 is the PSD of the integrated drag
linear and quadratic approximations perform well. Next, the re-
force F̃d, then the corresponding linearization is as follows:
sults for fatigue are presented:
n


quadratic approximation: 94% of MC result
linearized approximation: 93% of MC result
F̃d ⬇ 兺 冑S̃ 共␻ 兲⌬␻B e
j=−n
d j j
i␻ jt
共29兲

One can conclude that it is possible to apply a linearized model, The exact expression for S̃d共␻兲 can be calculated analytically us-
without losing much accuracy. ing the approach, described in Ref. 关16兴.
The integrated approach has an advantage compared to the
Spectral Approximation pointwise one, since the total approximated drag force has exactly
the same variance as the exact one. Table 6 presents a comparison
In the previous sections, the pointwise approximation was ap-
between the response PDFs due to full forces, including both in-
plied. In this way, however, cross-correlation effects may intro-
ertia and drag. Moderate sea state Hs = 3.5 m, T p = 10 s is chosen.
Comparing Table 6 with Table 3 from the previous section, it is
seen that the accuracy of the global spectral approximation is
Table 3 MC and Approx PDF, full resp., moderate sea
better than the pointwise one. Next, preliminary results for the
PDF argument MC PDF Approx PDF Lin PDF fatigue calculation are presented:
• quadratic approximation: 99% of MC result
0 1.5⫻ 10−1 4.5⫻ 10−1 4.4⫻ 10−1
1 2.3⫻ 10−1 2.3⫻ 10−1 2.4⫻ 10−1
• linearized approximation: 98% of MC result
2 3.8⫻ 10−2 3.7⫻ 10−2 3.7⫻ 10−2
3 3.0⫻ 10−3 2.2⫻ 10−3 1.7⫻ 10−3
Comparing these results with those of the previous section, one
4 2.1⫻ 10−1 1.0⫻ 10−4 2.2⫻ 10−5 can conclude that the global spectral approximation approach per-
5 8.0⫻ 10−0 4.2⫻ 10−6 8.3⫻ 10−8 forms better than the pointwise approximation 关17兴.

Table 4 MC and Approx PDF, drag resp., severe sea


Table 6 MC and Integr. Approx PDF, full resp., moderate sea
PDF argument MC PDF Approx PDF Lin PDF
PDF argument MC PDF Approx PDF Lin PDF
0 2.8⫻ 10−1 2.5⫻ 10−1 1.7⫻ 10−1
3 4.5⫻ 10−2 4.9⫻ 10−2 7.6⫻ 10−2 0 4.5⫻ 10−1 4.5⫻ 10−1 4.4⫻ 10−1
6 8.8⫻ 10−3 8.1⫻ 10−3 7.0⫻ 10−3 1 2.3⫻ 10−1 2.3⫻ 10−1 2.4⫻ 10−1
9 2.9⫻ 10−3 3.3⫻ 10−3 1.3⫻ 10−4 2 3.8⫻ 10−2 3.8⫻ 10−2 3.8⫻ 10−2
12 9.2⫻ 10−4 9.8⫻ 10−4 4.7⫻ 10−7 3 3.0⫻ 10−3 2.4⫻ 10−3 1.9⫻ 10−3
15 2.8⫻ 10−4 3.0⫻ 10−4 3.5⫻ 10−10 4 2.1⫻ 10−4 1.2⫻ 10−4 2.7⫻ 10−5
18 9.0⫻ 10−5 9.4⫻ 10−5 5 8.0⫻ 10−6 5.8⫻ 10−6 1.2⫻ 10−7

Journal of Offshore Mechanics and Arctic Engineering AUGUST 2008, Vol. 130 / 031009-5

Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Concluding Remarks 关4兴 Larsen, C. E., and Lutes, L. D., 1991, “Prediction of Fatigue Life of Offshore
Structures by the Single-Moment Spectral Method,” Probab. Eng. Mech.,
In this paper, nonlinear effects on fatigue damage due to Mori- 6共2兲, pp. 96–108.
son type loading on a jacket structure subjected to random waves 关5兴 Wilson, J. F., 1984, Dynamics of Offshore Structures, Wiley, New York.
关6兴 Skjong, R., and Madsen, H. O., 1987, “Practical Stochastic Fatigue Analysis of
are considered. A frequency domain method is described for ac- Offshore Platforms,” Ocean Eng., 14共4兲, pp. 313–324.
curately approximating the response process due to the combined 关7兴 Sarpkaya, T., and Isaacson, M., 1981, Mechanics of Wave Forces on Offshore
inertia and drag forces in random seas. This makes it possible to Structures, Van Nostrand Reinhold, New York.
account for the nonlinear effects caused by the drag forces in an 关8兴 Hasselmann, K., Barnett, T., Bouws, E., Carlson, H., Cartwright, D., Enke, K.,
Ewing, J., Gienapp, H., Hasselmann, D., Kruseman, P., Meerburg, A., Müller,
efficient way. The major conclusion of the present study is that for P., Olbers, D., Richter, K., Sell, W., and Walden, H., 1973, “Measurements of
long-term fatigue calculations it is, in fact, sufficient to linearize Wind-Wave Growth and Swell Decay During the Joint North Sea Wave Project
the drag forces without losing too much accuracy. However, lin- 共JONSWAP兲,” Dtsch. Hydrogr. Z., Reihe A 8, No. 12.
earization can be carried out in different ways. In addition to 关9兴 Chakrabarti, S., 2005, Handbook of Offshore Engineering, Elsevier, New York.
关10兴 Lin, Y. K., 1967, Probabilistic Theory of Structural Dynamics, McGraw-Hill,
common statistical linearization, the spectral approximation is New York.
also studied, which turns out to be more accurate. This is of prac- 关11兴 Naess, A., Berstad, A. J., and Moen, L. A., 1994, “Stochastic Fatigue Analysis
tical importance for the long-term fatigue analysis. of the Tethers of a Tension Leg Platform,” in Proceedings of the Second
International Conference on Computational Stochastic Mechanics, Athens, P.
Acknowledgment D. Spanos, ed., Rotterdam, Balkema.
关12兴 Naess, A. and Johnsen, J. M., 1992, “An Effcient Numerical Method for Cal-
The financial support from the Research Council of Norway culating the Statistical Distribution of Combined 1st-Order and Wave-Drift
共NFR兲 through the Centre for Ships and Ocean Structures 共Ce- Response,” J. Offshore Mech. Arct. Eng., 114共3兲, pp. 195–204.
关13兴 Naess, A. and Karlsen, H. Chr., 2004, “Numerical Calculation of the Level
SOS兲 at the Norwegian University of Science and Technology is Crossing Rate of Second Order Stochastic Volterra Systems,” Probab. Eng.
gratefully acknowledged. Mech., 19共2兲, pp. 155–160.
关14兴 Naess, A., Karlsen, H. Chr., and Teigen, P. S., 2006, “Numerical Methods for
Calculating the Crossing Rate of High and Extreme Response Levels of Com-
References pliant Offshore Structures Subjected to Random Waves,” Appl. Ocean. Res.,
关1兴 Naess, A., and Pisano, A. A., 1997, “Frequency Domain Analysis of Dynamic 28共1兲, pp. 1–8.
Response of Drag Dominated Offshore Structures,” Appl. Ocean. Res., 19共3兲, 关15兴 Karunakaran, D., Haver, S., Baerheim, M., and Spidsoe, N., 2001, “Dynamic
pp. 251–262. Behaviour of Kvitebjørn Jacket in the North Sea,” in Proceedings of OMAE,
关2兴 Naess, A., and Yim, S. C., 1996, “Stochastic Response of Offshore Structures Rio de Janeiro, Brazil.
Excited by Drag Forces,” J. Eng. Mech., 122, pp. 155–160. 关16兴 Borgman, L., 1967, “Spectral Analysis of Ocean Wave Forces on Piling,” J.
关3兴 Benasciutti, D., and Tovo, R., 2005, “Spectral Methods for Lifetime Prediction Wtrwy. and Harb. Div., 93, pp. 129–156.
Under Wide-Band Stationary Random Processes,” Int. J. Fatigue, 27, pp. 867– 关17兴 Wirsching, P. H., and Light, M. C., 1980, “Fatigue Under Wide Band Random
877. Stress,” ASCE J. Struct. Div., 106共ST7兲, pp. 1593–1607.

031009-6 / Vol. 130, AUGUST 2008 Transactions of the ASME

Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like