Sie sind auf Seite 1von 9

Proceedings of the ASME 2012 31st International Conference on Ocean, Offshore and Arctic Engineering

OMAE2012
July 1-6, 2012, Rio de Janeiro, Brazil

OMAE2012-8

COMPARISON OF DIFFERENT EQUIVALENT DESIGN WAVES WITH SPECTRAL


ANALYSIS.

Guillaume de Hauteclocque Quentin Derbanne Amine El-Gharbaoui


Research Department Research Department Technical Offshore Department
Bureau Veritas (France) Bureau Veritas (France) Bureau Veritas (France)
Neuilly-Sur-Seine FRANCE Neuilly-Sur-Seine FRANCE Neuilly-Sur-Seine FRANCE

ABSTRACT NOMENCLATURE

Design waves are widely used to assess stress in vessels. RAO Response Amplitude Operator (complex)
Compared to a full spectral analysis, this method reduces the RAO Argument of the RAO
number of required structural calculations and is expected to pro- EDW Equivalent Design Wave
vide similar results. Moreover in a rule context, it allows loading RCW Response Conditioned Wave
a structural model with a simplified hydrodynamic loading, via DRCW Directional Response Conditioned Wave
explicit formulae for extreme hydrodynamic loads and associated LCF Load Combination Factor
load combination factors (LCF) VBM Vertical Bending Moment
Relative wave heading
In a broad sense, an equivalent design wave (EDW) is a wave Wave frequency
on which a selected response is equal to a target value (generally Tp Wave spectrum peak period
the spectral extreme response). If general practice is to define the Jonswap spectrum shape factor
EDW as a regular wave, it is possible to define it more realisti- m0 zero order spectrum moment
cally; there are also different ways to define heading and period FSRU Floating Storage and Re-gasification Unit
of the EDW which can lead to significantly different results. In
this paper the EDW are not restricted to regular waves: response
conditioned waves (RCW) and directional response conditioned INTRODUCTION
waves (DRCW) are also investigated. Those waves include more The design wave approach is widely used in the industry, it
physics and are shown to overcome some of the inconsistencies is used as an efficient engineering procedure and is the basis of
of the regular wave (Load Combination Factor above 1 for in- the hydrodynamic loads formulae in the different Rules, like
stance). the IACS Common Strucural Rules [1].
The major benefits of the EDW approach are that it signifi-
The different methods are then applied to an FSRU; hydro- cantly reduces the computational cost, and that it allows to derive
structure calculations are carried out, the resulting stress is cal- simplified formulae for hydrodynamic loading. Design waves
culated in more than 2000 relevant elements. The way to select a can be used for extreme as well as for fatigue assessment.
few design waves among a lot of individual loads from the hydro- There are two major applications of design waves, one being
dynamic calculation (motion, acceleration, internal loads, pres- to calculate the structural response based on a few waves which
sure) is discussed and finally, the stress computed by the different maximize a given set of governing loads, the other one being to
design waves is compared to the direct spectral calculation. evaluate the non-linearities in an cost-effective way. This paper

1 Copyright 2012 by ASME


focuses on the first one, all the calculations here are linear. VBM EDW
15
Compared to a full spectral analysis the design wave ap- Wave elevation DRCW
Wave elevation RCW
Wave elevation REG
proach avoids calculating the stress RAOs. Indeed, such a cal-
10
culation requires two structural calculations (real and imaginary
part) for each couple ( , ), while the design wave approach re-
5
quires only one calculation per design wave. The underlying hy-

Wave elevation
pothesis of the design wave approach are :
0

Element stresses are maximized at the same time as one of


the elemental loads applied on the vessel (internal loads, -5

acceleration, pressure...), so that the maximum stresses on


all design waves match with the spectral calculation -10

The number of loads that contributes to the stresses can be -15


-30 -20 -10 0 10 20 30
reduced to a practical number. Time

FIGURE 2: Different VBM design waves

Regular Wave

FIGURE 3: Regular wave


FIGURE 1: Calculation scheme Spectral / EDW

This paper focuses on the different ways to generate design General practice is to choose the EDW as regular. The main
waves and how it impacts the overall EDW approach validity and advantage being its simplicity. The regular design wave targeting
accuracy. a value X is defined by :

{
DESIGN WAVE TYPES X
A= |RAO( , )|
For a given response, the only mandatory design wave re- (1)
= RAO ( , )
quirement is that the linear response on this wave is equal to a
targeted value (usually taken as the long term value of the re-
sponse) at a given time (We chose t= 0 for simplicity). An infin- The parameters to define are :
ity of waves satisfy this requirement, the simplest being a regular
wave. More refined EDW are here also investigated : the uni-
directional Response Conditioned Wave (RCW) and the Direc- {
tional Response Conditioned Wave (DRCW). The application is (Heading)
(2)
here done with meteocean data provided as a scatter diagram. (Frequency)

2 Copyright 2012 by ASME


Without spreading
There are different ways to define the heading and period of With Spreading

a regular design wave: 340.0


350.0 0.0 10.0
20.0
330.0 30.0

1. The usual practice is to select the heading and period of 320.0


310.0
40.0
50.0
the wave in order to maximize the RAO. Its drawback is that the 300.0 60.0

period and heading of the selected REG-EDW does not depends 290.0 70.0

of the meteocean data. The maximum of the RAOs might be 280.0 80.0

located at a frequency / heading where there is no wave energy. 270.0 90.0

Such a EDW might thus not be very representative of the actual 260.0 100.0

waves responsible for the extreme response of the vessel. For 250.0 110.0

240.0 120.0
some response, heading would thus have to be manually tuned, 230.0 130.0
like for instance for the wave bending moment, which sometimes 220.0 140.0

presents a maximum at quite high frequency at 60 which is


210.0 150.0
200.0 160.0
190.0 180.0 170.0

a heading that actually does not contribute to the long term value.
FIGURE 5: Heading contribution to the long term value. (Verti-
2. To overcome this issue, one might suggest [3] that a way cal shear Force)
to automatically select frequency and heading could be done
by determining which frequency/heading contribute the most to
the long term extreme, and select those frequency/heading as
parameters for the REG-EDW. However, due to the spreading of
the actual sea-state spectrum, this could lead to very inaccurate
350.0360.0
0.0 10.0
results, as explained by the below scheme (figure 4). This 340.0
330.0
20.0
30.0
argument applies the same way for both heading and frequency. 320.0 40.0
310.0 50.0
This is exemplified on the vertical shear force : the heading 300.0 60.0 RAO Amplitude
contributing the most to the long term value is 180 , where the 290.0 70.0
RAO is actually quite weak (figures 5 and 6). 280.0 80.0

270.0 90.0

260.0 100.0
3. A possible improvement to avoid manual tuning of the
250.0 110.0
heading would be to select the heading contributing the most to
240.0 120.0
the long-term value, but on a calculation that would have been 230.0 130.0
previously performed without spreading (to avoid the problem 220.0 140.0
210.0 150.0
above explained). The frequency chosen remaining based on the 200.0 160.0
190.0180.0170.0
RAO maximum (the same method for both frequency/heading
would reduce to method 1 ). FIGURE 6: Vertical shear force RAO

Response Conditioned Wave

FIGURE 4: Regular EDW FIGURE 7: Response Conditioned Wave

3 Copyright 2012 by ASME


The response conditioned wave is defined as the wave train
leading to the mean of all the possible responses targeting a
given value, on an uni-directional sea-state. The RCW is thus {
Ai j = Sw(i , j ) |RAO(i , j )| mX0 d i d j
an irregular wave train, containing several components, the am- (5)
plitude/frequency/phase of these components can be computed i j = RAO (i , j )
from the RAO and wave spectrum using to equation (3).
The parameters to defined being :
{
Ai = Sw(i ) |RAO(i , )| mX0 d i
(3)
i = RAO (i , ) (Heading)


T (Peak period)
p
(6)
The parameters to define are : Spectrum shape



DirectionalSpreading


(Heading)
Tp (Peak period) (4) The choice of these parameters is here straightforward :

there are taken as the spectrum parameters of the sea-state con-
Spectrum shape tributing the most to the long term value.

Peak period and spectrum shape are chosen as the Tp and


spectrum shape of the sea-state contributing the most to the tar- HYDRO-STRUCTURAL ANALYSIS
geted value. The wave being unidirectional, the question/answer To investigate the EDW approach and to benchmark the
about the heading to choose are the same as for the regular waves. different design waves, complete spectral analysis is carried
One advantage of such an EDW is that it contains much more in- out on a FSRU. The design wave approach investigated here is
formation than a regular EDW. It contains both RAOs and wave independent from the hydro-structure calculation scheme, the
data, which should lead to a more realistic wave. An example of hydrodynamic loading could be done by several means (strip
RCW is displays on figure 7 theory, 3D BEM... ) and the same applies for the structural re-
sponse (2D model, three holds FE model, complete FE model...
Directional Response Conditioned Wave ). In the current study, the hydrodynamic part of the calculation
is done with a 3D radiation/diffraction model (HydroStar Soft-
ware), the transfer of hydrodynamic pressure to the structural
model is then done in a consistent way (Homer Software) :
instead of interpolating the pressure from the hydrodynamic
model to the structural model, the pressure is directly calculated
on the wetted part of the strucural model. This leads to perfectly
balanced model, even if the hydrodynamic and structural model
does not match exactly. More details on the methodology can be
found in [5] and [4].

The meteo data used in this calculation is the north At-


lantic scatter diagram (IACS rec.34), spectrum shape is Pierson-
Moskowtiz, the directional spreading is cosn spreading with
FIGURE 8: Direction Response Conditioned Wave n = 2, and the wave heading is taken as iso-probable. The long
term value is the value with a return period of 25 years. The ca-
pability of the EDW approach to deals with higher probability
The RCW described above is uni-directional, it can be gen- levels will be discussed in a latter section.
eralized to model irregular short-crested seas (or multi-modal). The wave data used in the present study present a direc-
The irregular response conditioned wave is thus defined the wave tional spreading. This is more general than a uni-directional
train leading to the mean of all the possible responses targeting a sea-states and it highlights some issues about the EDW approach
given value, on a directional sea-state. The different components that would hidden on uni-directional sea-states (selection of the
can be computed thanks to equation 5. EDW heading for instance).

4 Copyright 2012 by ASME


Equivalent Design Waves calculaion
Figure 11 presents the methodology used to assess long term
stresses with an EDW approach. The wave data used are of
course the same as the ones used in the full spectral analysis.
In order to avoid any a priori assumption, the number of loads
defining design waves is not limited at first. The different design
waves will be based on following loads (47 EDWs):
Acceleration (ACCX, ACCY, ACCZ) at various positions, in-
cluding the center of gravity of the internal tanks.
Roll motion RAO
FIGURE 9: FSRU Finite Element Model
Internal loads (MX, MY, FZ) at various positions along the
ship.
Pressure at water-line

Full spectral analysis


In order to validate the EDW approach, a full spectral anal-
ysis is performed and considered as the reference. Figure 10
presents the different steps of the spectral calculation.

FIGURE 11: EDW Methodology

The generation of the design waves is performed using the


FIGURE 10: Full spectral analysis formulae presented in the previous section (equations 1, 3 and 5).

RESULTS, Comparison spectral/EDW


Stress RAOs were computed on more than 2 000 relevant The following graphs present results of the different EDW
elements (the overall FE model is made of 250 000 elements). types, with several choices in the heading definition. Each point
Among those 2 000 elements (upper/lower planes , xx, xy and represents one of the 1000 elements, the value plotted on the
yy components) the 1000 stresses with the highest spectral long Y-axis is the maximum response over the 47 EDWs, while the
term values were selected to perform the comparison with the X-axis is the spectral value (The reference is the line y=x).
EDW approach.

5 Copyright 2012 by ASME


Regular wave The results plotted on figure 13 are not acceptable at all,
the error is huge! On this particular case, some design waves
are completely unrealistic, over 50 meters. The results can be
A/ REG-EDW, heading and frequency based on the maximum of improved by removing the unappropriated EDW afterwards (i.e.
the RAO the design waves which present unphysical steepness), but the
accuracy would remain far behind the other method. This method
200
has been suggested in [3] on uni-directional sea-states, while the
Quadratic error = 29.74%
Long term stress present case deals with a short-crested sea, so that it is fair to test
180 Mean error = 19.82% it on uni-directional sea-state, the results are indeed better but
Standard deviation = 22.18%
160 the problem of the frequency spreading remains and the error
is still significant.
140
Max EDWs value

120

100

80 C/ REG-EDW, heading based on the sea-state that would con-


60
tribute the most to the extreme on calculation without spreading,
frequency based on the RAO maximum
40

20 200
20 40 60 80 100 120 140 160 180 Long term stress
Quadratic error = 7.74%
Spectral value 180 Mean error = -2.51%
Standard deviation = 7.32%
FIGURE 12: Regular EDW - Max RAO (A) 160

140
Max EDWs value

120

The results from figure 12 show a significant error (quadratic 100


error = 30%), which is not acceptable.
80

60

B/ REG-EDW, heading and frequency of the sea-state that ( T2p ) 40


contribute the most to the extreme, as suggested by [3]. For
20
reference, a case without any-spreading is also tested. 20 40 60 80 100 120 140 160 180
Spectral value

FIGURE 14: Regular EDW - Max Contribution / No spreading


200
Quadratic error = 230.85%
Long term stress (C)
Long term stress unidirectional
180 Mean error = 166.73%
Standard deviation = 159.7%
160

140
Max EDWs value

120
The results from figure 12 to 14 show that the way of se-
100
lecting the heading of the EDW has to be defined with care :
80 the quadratic error varies from 7.7% to 230% depending on how
where chosen heading and peak period.
60

40
Choosing the heading/frequency from the RAO maximum
leads to inaccurate results. Selecting the heading and frequency
20
20 40 60 80 100 120 140 160 180 200 220
of the sea-state contributing the most to the long term value make
Spectral value things a lot worst (fig 13), due to spreading (in both frequency
and heading), as shown on figure 4. This way of selecting head-
FIGURE 13: Regular EDW - Max Contribution (B) ing and frequency is discarded from now on.

6 Copyright 2012 by ASME


Response Conditioned Wave F/ RCW-EDW, Heading of the sea-state that would contribute the
most to the extreme on a calculation without spreading. Tp of the
sea-state contributing the most at this heading.

200
D/ RCW-EDW, Heading based on the maximum of the RAO, Tp Quadratic error = 7.34%
Long term stress

of the sea-state contributing the most to the long term value at 180 Mean error = -2.93%
Standard deviation = 6.73%
this heading 160

140

Max EDWs value


200
Long term stress
Quadratic error = 23.91% 120
180 Mean error = 17.23%
Standard deviation = 16.57% 100
160

80
140
Max EDWs value

60
120

40
100

20
80 20 40 60 80 100 120 140 160 180
Spectral value
60

FIGURE 17: Response Conditioned Wave - Max Contribution no


40
spreading (F)
20
20 40 60 80 100 120 140 160 180
Spectral value

FIGURE 15: Response Conditioned Wave - Max RAO (D)


As for the regular wave, choosing the EDW heading as the
one contributing the most on a calculation without spreading
leads to the most accurate results.

Directional Response Conditioned Wave

G/ DRCW, Heading and Tp of the sea-state contributing the


E/ RCW-EDW, Heading and Tp of the sea-state contributing the most to the long term value
most to the long term value

200 200
Long term stress Long term stress
Quadratic error = 30.87% Quadratic error = 8.17%
180 Mean error = 9.57% 180 Mean error = -5.72%
Standard deviation = 29.35% Standard deviation = 5.84%
160 160

140 140
Max EDWs value

Max EDWs value

120 120

100 100

80 80

60 60

40 40

20 20
20 40 60 80 100 120 140 160 180 20 40 60 80 100 120 140 160 180
Spectral value Spectral value

FIGURE 16: Response Conditioned Wave - Max Contribution(E) FIGURE 18: Directional Response Conditioned Wave (G)

7 Copyright 2012 by ASME


Conclusions on design waves type
REG
RCW
14 % DRCW
The best choice with regards to direction for uni-directional
case is to select the heading of the sea-state that would con-
tribute the most on a calculation without spreading, this method 12 %

works quite well for both REG EDW and RCW. The Directional

Quadratic Error
Response Conditioned Wave also provides satisfactory results,
10 %
without having to wonder about the EDW direction.

Having discarded inaccurate ways of selecting heading and 8%


frequency (A, B, D and E leads to a quadratic error above 20% !),
three acceptable design wave types remain (C, F, and G). Figure
6%
19 plots the error ratio distribution for those three kinds of design
waves. The RCW seems to be the best compromise (sharpest 45 40 35 30 25 20 15 10 5
peak with correct location.) Number of design waves

FIGURE 20: EDW selection

12
Although reducing the number of design waves can only
REG
RCW make the EDW response decrease, figure 20 shows that the num-
DRCW

10
ber of design waves can significantly be reduced, without losing
too much accuracy. According to figure 20 and 21, a good com-
promise would be around 10 design waves, reducing further than
8
10 makes the error rise up very quickly. 10 EDW can be consid-
Probability density

ered as a reasonable number in practice, thus validating the EDW


6
approach.

10
2 RCW 47 EDW
RCW 10 EDW
9 RCW 8 EDW
RCW 4 EDW
0 8 RCW 3 EDW
0.7 0.8 0.9 1 1.1 1.2 1.3
Ratio EDW/Spectral 7
Probability density

6
FIGURE 19: Error distribution
5

Design wave selection 0


0.7 0.8 0.9 1 1.1 1.2 1.3
Ratio EDW/Spectral
In previous section all the design waves were used to deter-
mine the long term stresses. This section aims at reducing the FIGURE 21: Error distribution when reducing the number of
number of design waves to a few ones, without losing too much EDW (Response Conditioned Wave)
accuracy. The EDW selection will be performed for the three
types of design wave. (Using the choice of heading based on
the contribution to the long term value based on the calculation
that would be performed without spreading). The selection of the The design waves selected by the 3 different approaches are
design wave is done so that the quadratic error is minimized. not strictly the same, but similar.

8 Copyright 2012 by ASME


Probability Quadratic Mean Standard the damage ( 102 ). The EDW approach being also accurate at
those probability levels, it can be used for fatigue assessment.
level error error deviation Another way to compute fatigue without a full spectral anal-
ysis would be to build the long term term distribution using EDW
targeting several probability levels.
25 years 7.34% 2.93% 6.73%
1e-2 9.13% 0.41% 9.12%
CONCLUSIONS
TABLE 1: Accuracy of the EDW approximation for different Different variations of design wave approach have been per-
probability levels formed on an FSRU. The way to select the heading of the applied
EDW has been shown to have significant impact on the final ac-
curacy. For uni-directional EDW, choosing the heading based
Probability levels on preliminary calculation without spreading has been demon-
Design waves are often used for low probability event like strated to be much more accurate than the other methods; this
extreme design values. By increasing the probability level, one method leads to satisfactory results, is robust, and does not re-
can fear that the contribution to the long term values would be- quire any tuning.
come more spread, and that modeling this scattered contributions As far as the kind of EDW is concerned, RCW and DRCW
by a few waves might not be valid anymore. To check the validity include more physics than a regular wave, the accuracy of the
of the EDW approach at a higher level of probability the design design wave approach being thus improved. Although the gain is
wave approach is performed targeting the value at probability 1e- not spectacular, the additional computational cost is negligible,
2. (Which is the level of probability contributing the most to the so that there is no reason not to use it.
damage on the North Atlantic scatter diagram [2]) This study has also allowed to quantify the errors done
through the design wave approximation, which is very important
to properly assess the overall in-certainties in the whole calcula-
Heading contribution to the Long term value tion procedure.
226506-lower-yy
350.0 0.0
340.0 Probability 1e-2
330.0 Probability 25 Years
320.0
310.0 50.0

300.0 60.0 REFERENCES


290.0 70.0 [1] IACS. Common structural rules, 2005.
280.0 80.0 [2] Derbanne Q., Rezende F., de Hauteclocque G., and Chen
270.0 90.0
X.B. Evaluation of rule-based fatigue design loads associ-
260.0 100.0
ated at a new probability level. ISOPE, 2011.
[3] RESMI SARALA, Mohammad Hajiarab, and Richard Bam-
250.0 110.0
ford. Equivalent design wave approach for calculating site-
240.0 120.0

230.0 130.0
specific environmental loads on an fpso. OMAE, 2011.
220.0 140.0 [4] Malenica S, Stumpf E., Sireta F.X., and Chen X.B. Consis-
210.0
200.0 160.0
150.0
tent hydro-structure interface for evaluation of global struc-
190.0 180.0 170.0
tural responses in linear seakeeping. OMAE, 2008.
[5] Chen X.B. Hydrodynamics in offshore and naval applica-
FIGURE 22: Heading contribution to different probability level
tions - part i. Keynote lecture of 6th Intl. Conf. HydroDy-
namics, Perth (Australia), 2004.

If indeed, the contribution of the different heading is more


spread (figure 22), the error of the design wave approach remains
almost the same.
For fatigue calculation, the overall long-term distribution is
required. General rules practice is to use a Weibull distribu-
tion, based on a assessed value at a given probability level and
given Weibull coefficients. As shown in [2], the sensitivity to
the Weibull coefficients is very weak when the long-term distri-
bution is based on a probability level which contribute a lot to

9 Copyright 2012 by ASME

Das könnte Ihnen auch gefallen