Documente Academic
Documente Profesional
Documente Cultură
ABSTRACT
INTRODUCTION
Today it is widely accepted within the offshore industry that the jack-up
platform concept generally is sensitive to a number of nonlinear dynamic
response effects. Consequently, there is a general understanding between
structural engineers that this must be accounted for in the structural
design analysis, both with respect to which type of analysis should be used
and which mathematical models should be involved in these analyses.
A far step towards an agreed industrial practice has been taken during
71
72 N. Spidsoe, D. Karunakaran
SOURCES OF N O N L I N E A R P L A T F O R M BEHAVIOUR
I i I I
~q /1
X X X X Max LmO s e r ke8 / I X X X X ~ X kmo #let"LgS; / I
.J
- - - - - ROd L , I . ~ / I
0,9~ -
0.9990- 0.9990 -
0.9900 - 0 . 9 ~ -
0.9000 - 0.9000 -
0.5000 - 0.5000 -
O. I000 - i_ 0.I~- ~
i
, , P 3'5 ' '
0.0000 2.5555 4.8867 7.0000 0.~ 2 55 4.8667 7.0~3CI0
S T . DEV. OF PARENT S E R I E S ST. OEV. OF PAF~NT SF=J:~IES
Base shear" OTM
Bca~e Shear" O~i~lurn , . ~ momen t
TABLE 1
Statistical Parameters of Nonlinear Quasistatic Response - TPG 500 Jack-Up
discussed both as concerns the model concept and the choice of coeffi-
cients. Based on comparisons of measurements and simulations, see for
instance Refs 3 and 24, it seems reasonable to conclude that this model
gives a satisfactory representation of the forces and quasistatic response,
provided that the coefficients are properly chosen. C o m m o n practice is to
use flow independent coefficients. It is, however, clear from experiments
that the coefficients are dependent on the flow pattern around the
members and on the leg roughness, 26 and thus may vary significantly over
the structure. There are reasons to believe that the coefficients are signifi-
cantly lower in the splash zone than for the submerged part of the struc-
ture. 26 Industrial practice neglects this, and it appears thus that the normal
choices of coefficients (i.e. Cm = 2-0, Ca = 0.7-1.0) will therefore lead to
reasonable, but most probably conservative results, as concerns the
nonlinear effects to the response. A final verification of this demands full-
scale measurements, but so far necessary comparisons of such to simula-
tions has not been published, though they ought to be available from
different measurement programmes.
Transverse forces are normally not included in global response analyses.
These forces may be modelled as are the drag forces and will consequently
have the same statistical properties. Experiments have indicated that
locally these forces may also be high, but that they are not well correlated
along the legs. If so, they will not have any effect in the global analysis,
and they are therefore neglected in such analysis. However, whether this is
correct can only be confirmed through full-scale experiment.
It is also well known that, under special current conditions, transverse
forces may cause vortex induced vibrations of individual legs. Whether
this is a problem for a specific platform must be checked in each case, as
this is not included in global dynamic response analysis procedures.
Nonlinear dynamic behaviour of jack-up platforms 75
Structural properties
P-~ effects
The so called P-6 effect is directly related to the displacement and load
levels in the structural element.2 For a wave loaded structure this leads in
principle to time and load dependent structural system stiffness. Conse-
quently, both the structural flexibility and the natural frequencies may
vary with time, thus effecting both the quasistatic and resonant response.
For jack-up platforms, however, the deck loads will dominate the leg
forces. The major P-6 effect will thus be a static contribution to the linear
stiffness, which only has a minor influence to the dynamic properties of
the platform. This is demonstrated in Table 2, which gives results from a
simulation study, based on the above assumptions.
Soil-structure interaction
TABLE 2
Comparison Between Dynamic Response Estimated With and Without Including P-6
Effect TPG 500 Jack-up Platform
OS
fl "fs "'""
j
F i
TABLE 3
Ratio of Dynamic Response Using Linearized and Nonlinear Soil Stiffness
12 45, i , , , 450.
10 _ ,oL
~'3oh'
~6 ~.2501-
i 2o ~2ooI-
,,o4 Liaearizgl.... . . . .
~Lm~iz~l '
2 51-
01 ' o.,~1 ' o.~ ' o.~ 'o.o4 o o.b~"'o'.1 o.~5 o'.2 o.:5 ~ 0.11 0~2 0b3 0.h
Displacemont[m] Displacgmcnt[ml Rotation[radsl
Horizontalload-deflectioncurve Verticalload-deflectioncrave Moment-rotationcurve
Fig. 4. Foundation characteristics.
increases, the rotational stiffness decreases more than the vertical stiffness
due to their different stiffness properties, see Fig. 4. When the rotational
stiffness decreases more than the vertical stiffness, the leg bending moment
also decreases, while the axial force increases. This gives that a redis-
tribution of the moment takes place in the nonlinear analysis due to the
stiffness reduction. In the linearized analysis, there is no nonlinear soil
stiffness effect and hence there is no redistribution of moments at high
load levels. This is seen from Fig. 5, where the total overturning moment,
the axial force in the leg and the moment at the bottom of leg are shown
for both linearized and nonlinear analysis at an extreme load cycle.
It is seen from Fig. 5 that the axial force in the nonlinear analysis has
higher maxima than in the linearized analysis and that the bending moment
in the bottom of the leg has less peaked maxima in the nonlinear analysis
than in the linearized analysis. Due to this moment redistribution, the axial
force in the leg has slightly higher standard deviation of response in nonlinear
analysis than in linearized analysis. Furthermore, this redistribution produ-
ces a non-Gaussian response with larger maxima than the linearized analysis
for the axial force, see Fig. 6. On the other hand this redistribution gives
smaller maxima for bending moment in the bottom of the leg from nonlinear
analysis than the linearized analysis, see Fig. 7. The other response quantities
which have similar behaviour to the axial force are the bending moments in
the upper part of the structure. Furthermore, the shear force in the upper
part of the structure has similar behaviour to the moment at the base.
Due to this redistribution phenomenon, the extreme response predicted by
linearized analysis deviates significantly from the response derived by nonlin-
ear analysis. The axial force in the leg estimated using linearized analysis is up
to 25% smaller than the extreme response from nonlinear analysis, so is the
bending moment at the upper part of the structure where the maximum
deviation is up to 29%. Furthermore, the deviation between the nonlinear and
linear response increases as the foundation moment capacity increases.
Deck-leg interaction
The deck-leg connection is one of the highly loaded parts of a jack-up plat-
form. The interaction between the deck and the leg may be nonlinear due to
the guide clearances, system backlash, brake slippage, etc. This connection is
often modelled by nonlinear spring stiffnesses) 5 These nonlinear springs
may principally imply hysteretic behaviour, with effect both to the natural
frequencies and the structural damping. However, the deck-leg interaction
will only have significant influence on the moment distribution on the leg
section within and around the connection. Hence, for a global dynamic
analysis it may be idealized by linearized springs, which is often used.
Nonlinear dynamic behaviour ofjack-up platforms 79
1400 i i I i i
//
1200 _ e~ No~m:~---- ....
1000
8OO
60O
1
400
200
0
-200 - -tO"
-400 I I I I I
280 285 290 295 300 305 310
T i m e [sec]
Overturning moment
16
L i I i i i j
14 /~ Nordlnoar
~ 6
4
2
0
-:2
-4 I
280 285 290 295 300 305 310
Tinae [see]
I I
t~ Nonlinear
ii
Linear ....
150
i
50
t
-50 I I I
280 285 290 295 300 305 310
Time [see]
i I Z i I I I I I
. . . . Wm
tbut t ~ / /
0.9990 - 0.9990
0.9900 - 0.9900
0 9000
0.9000 -
0.5000-
0.4000- ,S
F I
4
I
8
I
t2
I
t6
I I
20
L 0 .SO00
0 . 4000
0.0
'2"
4.4
I I
8.8
I
t5.2
'
Ii
47.6 22.0
MN MN
Force Force
Non L ,ne(~r 8 0 t L mode t L tnoer" t z e d 8 0 k L mode L
I I I I I I I I I I
X X X X MOX kmo 81b"tile, /
- - -- ~ RoI=I La tgh /
WetNLL - 3 /
/ n
o.999o-
0.9990 - -~-'~
o .ssoo - I o.ssoo - x,x/~x~eX~ex ^ I
0.9000 - f X " " I
o.9ooo - i
0.5OOO -
0.9000 - I
0. I000 - 0. tO00- I
I
0 ~0 ,00 '1~0 2~0 ~-0 ZOO 0 5~0 ,00 ,;0 2~0 2~0 300
MNm MNm
Moment Moment
Non L tnaoP s o t t modg L L tnaeP t z o d s o tt mode L
Dynamic effects
Damping
ing using a Rayleigh model specified by the modal damping of the lowest
natural modes of the structure.
If a linearized model is applied for the soil-structure interaction, the
damping may be specified through the damping coefficients of dashpots.
Alternatively, the soil damping is, however, normally negligible for slender
structures such as a jack-up platform. If it is accounted for, this may be
done by including it in the Rayleigh model used for the structural damp-
ing. Tile drag damping which dominates the hydrodynamic damping is
implicitly given by the extended Morison equation used for the wave
loading.
There are three different damping mechanisms involved:
- - the linear viscous mechanism modelled by the Rayleigh model;
- the nonlinear drag damping mechanism, modelled by the extended
-
Morison equation;
- - the hysteretic soil damping mechanism, modelled by the kinematic
hardening model.
These mechanisms have different properties and effects to the dynamic
response which may have significant impact to the response extremes.
The most important and least known property of the linear viscous
damping mechanism is the so called Normalization effect, i.e. its ability to
generate a Gaussian resonant response independent of the statistical
nature of the excitation process. 4 This effect is demonstrated in Fig. 8 and
Table 4, which summarize results from a study on damping for the TPG
500 jack-up. This example shows that, even if the resonant response is very
high in the rms sense, its contribution to the extreme response is signifi-
cantly less because the total response is less non-Gaussian than the quasi-
static response because of the nearly Gaussian resonance generated by the
damping mechanism. The drag damping mechanism has a similar effect,
though the mechanism is different. As also shown by Brouwers4 this
damping mechanism generates a non-Gaussian resonant response with a
lower Kurtosis coefficient than in a Gaussian process, i.e. with a statistical
nature opposite to the quasistatic response. When the resonant and
quasistatic response is added an even stronger Normalization effect to the
total response occurs than generated by the linear viscous damping model.
The nonlinear properties of the hysteretic soil damping mechanism
depend highly on the shape of the back-bone curve. The more nonlinear
this curve is, the wider will the hysteretic curve be at the high load cycles.
Consequently, the damping in these response cycles will be higher than in
the lower. This is similar to the drag damping mechanism. Though it has
not been investigated so far, this means that one may assume that a
Normalization effect similar to what is found for the drag damping will
82 N. Spids~e,D. Karunakaran
I I I I I I I /
I
x x x x Mclx tmo Isr-Ueli X X X X M o x ~.mo 6 i s P L,e 8
-I
. . . . R o W t e Lgh . . . . R o W L Lgh
. . . . WoLbuLL-S F Lt
<= . . . . WetbuL L-3 r t l
0,9990- 0.9990-
0,9900- 0.9900-
0,9000- 0.9000-
O.,~O00- 0.S000-
O,JO00- 0.I000-
i i i i i
0 I
ST.
I
DEV. OF
;
PARENT
*'S E R I E S ,0 5
T.
4
OEV. OF
S 8
PARENT SERIES
I0
Momon t M o m o n t.
Ouoe L s t o t k c r-elserise D~,,,inomc response
I I I I I I I I
x x x x M o x Lmo l e t Les x x x x Mox Lmo e e P Le8
-I - - " ~ --'- ~ R o u l " kgh / -- -- -- RO W t e ~.gh
//
We t b u t I . - S ~
=e /
/ /
0 8990 - 0.9990
0.9900 - 0.9900
0 . SO010 - 0.9000
0 .SO00 - O.SO00
O. I000 - 0.t000
i I i l i
2 4 6 8 t0
ST. DEV. OF PARENT SERIES ST. OEV. OF PARENT SERIES
Homn Momen ~,
Dynom LC w o v e response R o e n o ni Peepnee
Fig. 8. Maxima probability distributions for overturning moment of TPG 500 jack-up
with 3% linear damping only.
also occur for this soil damping mechanism. This is probably partly the
explanation of the different statistical nature observed between linear and
nonlinear soil structure interaction models demonstrated in Fig. 6.
The different damping mechanisms are not independent, but work
together. This is demonstrated by the results given in the last part of
Table 4. From these it is seen that when the linear damping reduces the
dynamic amplification of the rms response increases. This is as expected
since the total damping level reduces. However, it is also noted that the
dynamic amplification of the extremes is almost independent of the linear
damping level. The explanation of this is that the reduction of linear
damping is compensated for by the Normalization effect both 'from the
linear damping itself and also from the drag damping which becomes
more important as the linear damping level reduces. From a design point
of view this is interesting because it implies that, as long as both these
damping models are involved, it is not important with an accurate speci-
fication of the linear damping level. However, since neither the linear
Nonlinear dynamic behaviour of jack-up platforms 83
TABLE 4
Dynamic Amplification Factors (DAF) for Various Damping Combinations TPG 500
Jack-up
Superharmonic excitation
TABLE 5
Eflbcts of Superharmonic Excitation to the Response of the TPG 500 Jack-up
Response quantities To = 3.3 s To = 5.3 s To = 6 s
i I I J
1650 1850
Bending moment Time (sec)
Monotower gravity platform
J llJ I l
t -- -t- - - --I-- --(
, ,
2250 2450
Overturning moment Time (sec)
Four-legged gravity platform
I i i i I
I i 'l i I
8600 8800
Tether tension Time (sec)
Tension leg platform
Fig. 9. Examples of ringing events observed from model tests of large volume structures.
Fig. 10. Ringing phenomenon seen from full-scale measurements - - F R I G G DP2 jacket.
the employed wave load models cause ringing effects in these cases. No
such effects have been observed. This means, however, that the problem is
not irrelevant. Improved or extended wave and wave load models may
give different result. It is therefore recommended that ringing and spring-
ing are investigated further for slender offshore structures.
Since jack-up platforms are highly dynamically sensitive, the design sea
state may be different from the extreme sea state which defines the 100
year wave height. This implies that, in principle, a long-term response
analysis which includes the contributions from all possible sea states to the
extreme response and which includes all the nonlinear effects discussed
above should be employed in dynamic response analysis of these plat-
forms. A procedure for this type of analysis tailored for drag dominated
platforms is described by Farnes 5 and Karunakaran. z2 As will be discus-
sed in the next section such analysis is, however, hardly applicable as a
practical design tool. For most jack-ups, it will not be required since the
extreme dynamic response in most cases will be dominated by extreme sea
states. This is demonstrated in Table 6 which gives results from a nonlin-
ear long-term dynamic response analysis of TPG 500 jack-up with varying
dynamic properties.
The long-term procedure is therefore not further discussed here, and it
is focused on short-term nonlinear response analysis. The following
discussions are, however, also relevant for long-term analysis as the short-
term modelling of the extreme dynamic response forms the basis for this
analysis also.
The short-term analysis of the nonlinear dynamic response requires
stochastic time domain simulation techniques. This is because the different
nonlinear phenomena involved may only be properly modelled in the time
domain as discussed above, and because realistic representation of the
dynamic behaviour of the platform demands stochastic modelling of the
88 N. Spids~e, D. Karunakaran
TABLE 6
Nonlinear Long-term Response of TPG 500 Jack-up Platform
Note: When the percentage contribution from sea states below 7 m is zero, the extreme
long term is completely dominated by the 100 year sea state.
3
>. 2
0 ~ = = , ~ v C m m ~ l dist. f r o m flttext W e i b u l l
-1 f o r 14 s e t s o f 7 s a m p l e s e a c h -
-2 I I I I I I
14 16 18 20 22 24 26 28
Base Shear [MN]
Fig. 11. Sample extremes plotted on Gumbel probability paper - - base shear.
0.16 I I I I I I I I I
22rain-sample
0.14 I'~ 4 5 r a in - s a m p le - - - - - --
II 90rain-sample ....
0.12 -~ 180min-sample ........
I1~ 360rain-sample -----
0.1
go
0.08
._= 0.06
c; 0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Number of samples
Fig. 12. C.o.v. in average extrapolated extreme response vs simulation length and number
of samples - - base shear.
0.57722 '- ]
= + [(ln ( N s ) ) h + - - fl (ln (Ns))~]-~ (1)
where
/~ -- Weibull l o c a t i o n parameter
tr -- Weibull scaling parameter
/3 -- Weibull shape factor
Ns -- number o f m a x i m a in storm duration;
1 N
-flx,= ~ .~-'~~x,)i (2)
7"7
Nonlinear dynamic behaviour of jack-up platforms 91
Cxe = ax,
_ (4)
#x,
NR = (5)
DESIGN ANALYSIS
It follows from the above discussions that the highly nonlinear dynamic
behaviour of jack-up platforms implies a wide range of significant uncer-
tainties related to structural analysis methods, to mathematical models for
structural behaviour, loading and response to the parameters involved in
these methods and models. On this background it is, at least from an
academical point of view, tempting to recommend that structural relia-
bility analysis should be included in the design process. Examples of such
analyses are presented by Jensen e t al., 8 L~seth e t al. 14 and Karunakaran
e t al. 13 Furthermore, a tailor made procedure for fixed, drag-dominated
offshore platforms based on nonlinear long-term dynamic response
analysis is established and demonstrated by Karunakaran. 12 However,
even if the tools are available and the analysis technique is proven, it
seems obvious that reliability analysis can not be a practical design tool.
This is because it requires specially qualified personnel, it is extremely
computer demanding and, maybe most important, it requires statistical
information with sufficient confidence on a wide number of models and
parameters which may not be available for a specific platform. There is,
however, a sufficient basis for extensive reliability studies of jack-ups as a
class of structures. Thus, it is possible to use this type of analysis for
92 N. Spidsoe, D. Karunakaran
TABLE q
Dynamic Response and Corresponding D A F s with C.O.V.s - - TPG 500 Jack-up
CONCLUSIONS
- - the major sources of nonlinear behaviour are the wave loading, the
damping mechanisms and the soil-structure interaction. The
Nonlinear dynamic behaviour ofjack-up platforms 95
ACKNOWLEDGEMENT
The experiences and examples presented in this paper are derived mainly
from studies carried out on contract for Statoil. The company is highly
acknowledged for release of the results. Furthermore, O. T. Gudmestad,
M. Ba~rheim and S. Haver, all from Statoil, have, through many valuable
discussions and suggestions during the projects, contributed significantly
96 N. Spidsee, D. Karunakaran
to the content o f this paper, and they are highly acknowledged for this
support.
REFERENCES
17. Olufsen, A., Uncertainty and reliability analysis of fixed offshore structures.
Dr. Ing Thesis, Norwegian Institute of Technology, 1989.
18. Penzien, J. & Tseng, S., Three-dimensional dynamic analysis of fixed offshore
platforms. In Numerical Methods in Offshore Engineering, eds O. C. Zienkie-
wicz et al., John Wiley & Sons, NY, 1978.
19. Pierson, W. J. & Holmes, P., Irregular wave forces on a pile. ASCE J.
Waterways and Harbours, WW4 (1965) 1-10.
20. Przemieniecki, J. S., Theory of Matrix Structural Analysis, Mc-Graw Hill,
USA, 1968.
21. Rainey, R. C., A new equation for calculation of wave loads on offshore
structures. J. Fluid Mech., 204 (1989).
22. Sigbj6rnsson, R., Stochastic theory of wave load processes. Engineering
Structures, 1, (1979) 58-64.
23. Skotheim, A. A., Foundation analysis of jack-up for Sleipner, B., Geovest
report, 1992.
24. Spidsoe, N., Brathaug, H-B. & Skjhstad, O., Nonlinear random wave loading
on fixed offshore platforms. In Proc. OTC, Houston, USA, 1986.
25. Spidsoe, N. & Karunakaran, D., Effects of superharmonic excitation to the
dynamic response of offshore platforms. E&P Forum Workshop on Wave and
Current Kinematics and Loading, IFP, Paris, 1990.
26. Spidsoe, N., Summary report - - effect of surface elevation. SINTEF Report,
STF'71 A89027, Trondheim, 1990.
27. Spidsoe, N., Karunakaran, D. & Gudmestad, O., Nonlinear effects of
damping to dynamic amplification factors for drag-dominated offshore plat-
forms. In Proc. 11th Int. Conf. OMAE, Calgary, 1992.
28. Svano, G., Madshus, G. & Lango, H., On the validity of nonlinear spring
idealization of soil structure interaction. In Proc. 2nd European Conf. on
Structural Dynamics, Trondheim, 1993.
A P P E N D I X A - - T P G 500 J A C K - U P
Structural model
J~
modes. The next mode is a torsional mode with a natural period of 4.3 s.
The other higher modes are found below 0-6 s,
The soil and structural damping together is assumed to be 2% and
modelled as proportional damping (Rayleigh damping) at the first and
third natural modes. The nonlinear hydrodynamic drag damping of the
structure is included by using relative velocities in the Morison equa-
tion.
Hydrodynamic coefficients
Environmental conditions
Structural model
Hydrodynamic coefficients
~ a l comlitions
C u r r e n t velocities ( c o r r e s p o n d i n g to a 10 y e a r period):
at m e a n w a t e r level 1.10 m s - 1
-- 30m 0-75 m s -1
-- 50m 0 - 7 0 m s -1
at m u d line 0.70 m s -~