Sei sulla pagina 1di 7

Use of simplified structural models to predict dynamic response to

wave-in-deck loads
Katrine Hansen
Stavanger University College, Stavanger, Norway

ABSTRACT: Forces arising from wave slapping at the deck of offshore platforms, in particular fixed steel frame
(jacket) platforms, have been attended to during the latest decade for two main reasons. Firstly, at some offshore
hydrocarbon producing fields the sea bed is subsiding. Fixed platforms located in such areas will experience
a decreasing deck clearance relative to the sea surface. Secondly, environmental design criteria, i.e. design
wave height, have been tightened due to observations of wave data, a desire for a higher level of safety or as a
result of more careful analysis of statistical data (Gudmestad and Hansen 2000). This paper will focus on how
an impulse like (peak) load, resulting from deck inundation, slightly prior to the wave force on the steel frame
will influence the dynamic response of the structure. Single degree of freedom (SDOF) models and full finite
element models will be used. It will be illustrated that simplified models, which are more resource economical
than complex models, are useful in the assessment of existing platforms.

1 INTRODUCTION the static capacity of the platform, leading to struc-


Wave-in-deck forces are of highly transient nature, tural collapse. Dynamic amplification might worsen
and therefore the displacement response of the plat- the situation, but dynamic attenuation could actually
form must be assessed considering dynamic effects. lead to an improved situation in some instances. Fig-
Despite this, state-of-practice is currently to use static ure 1 shows a principle sketch of an assumed force
analysis methodology (more specifically non-linear time history (total base shear or overturning moment)
static pushover analysis) to determine the capacity of resulting from a wave whose crest inundates the deck
the structure as a whole, disregarding important ef- of a jacket platform. The sharp peak, which is of large
fects of inertia and damping. magnitude and short duration and thus of transient na-
ture, represents the wave impact on the deck. This im-
Force pact event will probably have passed before the wave
force on the steel frame reaches its maximum value,
meaning that the impulse-like force is followed by
a continuing increasing force in the same direction.
Loads close to or exceeding the static capacity of the
platform might imply permanent structural deforma-
Time tion. However, having in mind the short duration of
ements the peak, a ductile structure will not necessarily reach
total structural collapse.
The general equation of dynamic equilibrium

Figure 1: Typical load history for platform subject to m u(t) + c u(t) + k u(t) = Fex (t) (1)
wave that reaches the deck
where u is displacement, t is time, m is mass, c
In this traditional static approach the peak wave is damping coefficient, k is stiffness and Fex is ex-
force on the deck is applied simultaneously with the ternally applied force, is valid in the linear (elas-
maximum wave force on the jacket. For most exist- tic) domain only. However, Biggs (1964) analyzed
ing jackets, even a small deck inundation could un- non-linear SDOF models by substituting a non-linear
der such assumptions result in load effects exceeding static resistance function (load - displacement curve)
for the ku-term, and illustrated how the shape of the
resistance function influences the dynamic response.
For a jacket, the global resistance function, denoted 2 u(b) 4 u(b)
(m + ma ) + EI
R, can be established by use of a non-linear static t2 z 4
pushover analysis.
Distributed parameter (DP) systems have, to the au- u(b)
thors knowledge, not previously been evaluated in +c = 0, d z 0 (3)
t
this context.
Dynamic response of jackets to extreme waves has where m denotes distributed mass along the beam,
previously been investigated, and dynamic effects are ma is added mass, u(a) and u(b) denotes the displace-
found to possibly be favourable for structures possess- ment of the beam above and below the SWL, respec-
ing certain qualities, e.g. high residual capacity (at tively, E is modulus of elasticity, I is moment of in-
least 80%) compared to the ultimate capacity (Stew- ertia, t is time, z is vertical position, h is deck clear-
art 1992; Dalane and Haver 1995; Schmucker 1996; ing above SWL and d is water depth. To solve these
Moan et al. 1997; Emami Azadi 1998; HSE 1998). equations, 8 boundary conditions are needed; two at
Because of the resource consuming nature of non- the sea bed, four at SWL and two at the top of the
linear dynamic analyses, approximation of dynamic beam. These boundary condition incorporate the vari-
behaviour of jackets by use of SDOF models has ables not mentioned above: lumped mass at the tip of
also to some extent been studied (Schmucker 1996; the cantilever (md ), rotational moment of inertia of
Moan et al. 1997; Emami Azadi 1998). However, this mass (J) and externally applied force and mo-
none of these studies consider explicitly the topic of ment (F (ex t) and Mex (t) respectively).
the present paper: wave-in-deck load as an additional In order to introduce non-linear behaviour, E can
impact force that acts while the wave load on the be defined as a function of the displacement u, such
jacket
PSfrag still increases.
replacements that Equations 2 and 3 reads
Global load level
2 CONCEPTUAL Rult MODEL DISTRIBUTED 2 u(a) 4 u(a)
PARAMETERS m + E(u) I = 0, 0 z h (4)
Rel t2 z 4
A distributedRparameter
res
system (DP) can be useful
and informative
Displ.to study idealized systems. A can-
tilever model will be the simplest way to represent 2 u(b) 4 u(b)
a jacket as a uDP system, see Figure 2. The struc- (m + ma ) + E(u)I
ult
tural propertiesuelof the cantilever (such as diameter, t2 z 4
modulus of elasticity,
ures wall thickness, distributed and
concentrated umass) must be chosen to represent real u(b)
max +c = 0, d z 0 (5)
jacket behaviour.
ucap t
x, u Note that u, and thus E, are functions of t and z.
EI The author has worked on solving this problem in
pc EI Mex (t) the case of a pure elastic model, i.e. when E is con-
Fex (t) stant. The effort used for an elastic model can not be
z h directly transferred a non-linear model, which will be
J, md
very complicated, and it was thus decided to leave this
model for a possible later occasion.
1 SWL EI, m
3 CONCEPTUAL MODEL SDOF
If extending the considerations on dynamic behaviour
Elastic into the non-linear region by use of the static resis-
rebound tance function as will be described subsequently, it
Figure 2: Distributed parameter model is important to verify that the dynamic deformation
mode is the same as the static collapse mode that
The effect of water is included only below still wa- forms the basis for the resistance function.
ter level (SWL), and the time variation of the surface A single degree of freedom (SDOF) mass - spring
elevation cannot be taken into account in a model like system is chosen to idealize a real structural system,
this one. The differential equations of motion for the in this case a jacket, see Figure 3. A realistic resis-
model is tance function is approximated by a few straight lines.
2 u(a) 4 u(a) Some parameters are given in the figure, note in par-
m + EI = 0, 0 z h (2) ticular that umax is the largest actual displacement for
t2 z 4
a given load history, while ucap is the largest permissi- In practical engineering it is a main problem that
ble displacement, or the displacement capacity. This carrying out non-linear finite element analyses is ex-
resistance function R(u) now describes the stiffness tremely time consuming. Results for many different
term, which in the elastic range is R(u) = k u, in the wave heights and periods are needed. In addition, the
dynamic equilibrium equation 6: maximum response is sensitive to the structures nat-
ural period. Therefore it is desirable to carry out even
mu + R(u) = Fex (t) (6)
more analyses with slightly different natural periods
In order to simplify, no damping is included in this to ensure that the response obtained is not unreal-
paper. If assessing a real structure, realistic damp- istically small due to a unfortunate combination of
ing parameters must be included. Note also that for natural period and load history (Skallerud and Am-
ements
a real structural system, the resistance function must dahl 2001). The presented concept makes it possible
be obtained for the load in question, in our case the to take advantage of the static pushover analysis, and
transverse wave load including wave-in-deck force. based on data from this, carry out (lots of) SDOF
The degree of freedom u(t) represents the displace- analyses to assess the dynamic behaviour. Finally, the
ment of the top of the structure (the deck). kel , kpe most unfavourable case(s) of response should be veri-
and kpu are the terms used to describe the elastic stiff- fied by use of non-linear finite element analyses in the
ness, post-elastic stiffness and post-ultimate stiffness, time domain.
respectively.
Global load level 4 COMPARISON WITH FINITE ELEMENT
Elastic ANALYSIS
Rult rebound
Rres A number of full finite element analyses have been
kpu Rel carried out using the FE program USFOS (Hellan
k
et al. 1998) in order to illustrate the simplified models
ability to approximate the largest displacement.
1 m
uel uult ures umax ucap 4.1 Models
u
Fex (t) Displ. Two typical North Sea jacket models are available for
investigation. The water depth, d, is varied in order to
vary the deck inundation for a wave height H = 33
Figure 3: SDOF model
m, which is a typical 10000 years wave in the North
Examples on how to solve equation 6 is given in Sea. The wave period, T , varies from 14 to 16 s.
the literature, e.g. Biggs (1964). Here the 2. cen-
tral difference formulation is used, with initial value
u(t = 0) = 0. The acceleration depends on the ex-
ternal force and the static resistance, and is different
from zero as long as the force is different from zero:
1
u = (Fex (t) R(u)) (7)
m
These initial values might influence the maximum re-
sponse. It would therefore be appropriate to slowly in-
crease the wave over a few wave cycles before the ex-
treme wave is applied (Stewart 1992; SINTEF 1998).
However, this requires that damping is included, oth-
erwise the vibration amplitude will accumulate for ev-
ery wave cycle. For the purpose of evaluating the use-
fulness of a SDOF model, the assumptions used here Figure 4: Model D
regarding the load history are satisfactory.
The resistance curve is expressed as follows: Model D (Figure 4) is a X-braced jacket which is
R(u) = kel u, 0 < u < uel 92.75 m high from the seabed to the underside of the
deck. The water depth is varied from 72 to 76 m. The
R(u) = Rel + kpe (u uel ), uel < u < uult piles are 38 m long, and the fundamental period of
R(u) = Rult + kpu (u uult ), uult < u < ures vibration is 1.44 s. A typical resistance function for
model D is shown in Figure 5. The dash-dot line is
R(u) = Rres , ures < u < ucap the result from the static pushover analysis in USFOS,
R(u) = 0, u > ucap while the solid line is the approximation to be used for
replacements
the SDOF model. This particular resistance function The load on the deck is automatically generated in
is obtained for a wave with H = 33 m, T = 15 s and the following way:
d = 74 m.
1. Vertical load generating elements are located at
80 the edges of the deck, two per edge, see Figure
4.
60
2. The two elements on the edge facing the wave
40 have Cd = 3 times the deck width they cover
Force [MN]

and Cm = 0. The rest have Cd = Cm = 0. This


20 is a simplification based on DNVs slamming ap-
0 proach (Det Norske Veritas 1991), where a slam-
0 0.2 0.4 0.6 0.8 1 1.2 ming factor for smooth circular members not less
Displacement [m] than 3.0 is recommended. This approach might
not be completely realistic, but makes it conve-
Figure 5: Model D, typical resistance function nient to calculate force time histories. Since the
objective of this work is to reveal to what extent
Model M is a K-braced jacket which is 90.55 m SDOF models are able to predict the dynamic re-
high from the seabed to the underside of the deck, sponse of a jacket to a given time force history,
located in 69 - 70.6 m water depth. The piles are 43.22 it is not important that the load history is 100%
m long, and the natural period 2.05 s. The resistance realistic.
curve for H = 33 m, T = 16 s and d = 69.5 m is
replacements
shown in Figure 6. 3. Given the values of the hydrodynamic factors for
the different elements, the forces are generated
automatically in USFOS for both static and dy-
60 namic analyses.
Force [MN]

40 4.3 Comparison procedure and results


Static nonlinear pushover analysis is carried out to es-
20 tablish the resistance curve of the model, subjected
to the actual wave. Further, non-linear dynamic time
history analysis is carried out, from which the result
0
0 0.2 0.4 0.6 0.8 1 is the displacement time history for a chosen location,
Displacement [m] preferably the center of the deck.
Figure 6: Model M, typical resistance function The non-linear resistance curve established by the
static pushover analysis is approximated by a few
straight lines and used in the SDOF model described
previously for calculation of displacement time histo-
4.2 Loads histories ries. The procedure does thus consist of the following
The wave load histories are generated by USFOS steps:
based on Stoke 5th order theory (Skjelbreia and Hen-
drickson 1960). The models are subjected to one wave establish load history (USFOS)
cycle. An identical load history is used when running carry out static nonlinear pushover analysis in order
the SDOF model. The load history is based on a 33 m to establish resistance curve (USFOS)
high wave having a period of 14 - 16 s. The current
speed is 1 m/s. An example is shown in figure 7. determine eigenfrequencies of linear model (e.g.
USFOS) and use it to calculate equivalent mass for
SDOF model
Force [MN]

40 apply load history and resistance curve in SDOF


analysis as described in section 3
20
compare with non-linear dynamic pushover analysis
0
The results are presented in Figures 8 and 9 for plat-
0 2 4 6 8 10 12 14 16
form D and M respectively. Maximum displacement
Time [s]
umax is plotted as a function of deck submergence sd .
Figure 7: Load history generated for model D for H = The parameter sd is in this case varied by varying the
33 m, T = 15 s and d = 74 m water depth, as explained previously.
1 1 The resistance curve for platform M is much sim-
umax [m] umax [m] pler to approximate with straight lines, and one should
0.8 0.8 thus expect good agreement in the results. This is the
SDOF, T = 15 s SDOF, T = 16 s case for wave period T = 16 s. For T = 14 s, how-
USFOS, T = 15 s USFOS, T = 16 s
0.6 0.6 ever, the SDOF model overpredicts the displacements
with 8 - 25% for all analyzed cases. Despite much ef-
0.4 0.4 fort, no obvious reasons for this are found. Since the
deviations are to the safe side, they are considered ac-
0.2 0.2
ceptable, but only so far. There are obviously other
parameters, in addition to the representation of the re-
sd [m] sd [m] sistance curve, that have a significant influence on the
0 0
0 1 2 3 4 0 1 2 3 4 results from the SDOF analyses. This will be closer
Figure 8: Platform D: Max. displacements from looked into in the near future.
SDOF model vs. finite element model, for wave pe-
riod (T) 15 s and 16 s, respectively 5 A FEW WORDS ABOUT AN EARLIER
STUDY
The results for platform D for wave period T = 15 To the authors knowledge, there exists one parame-
s have reasonably good agreement. For T = 16, ter study investigating the influence of the static re-
the deviations are somewhat larger, especially for the sistance curve (Schmucker 1996). Schmucker used
large values of sd . However, for these cases the SDOF a squared sinusoidal (SQS) load history, see Figure
results are generally on the safe side. The largest 10. In this paper, the load history (denoted SD) de-
deviation is for T = 16 and sd = 3.88 m, where scribed in section 4.2, and this in the authors opin-
the SDOF model overpredicts the maximum displace- ion more realistic for the wave in deck problem. The
ment with 37%. conclusions from Schmuckers parameter study is not
At the highest levels of deck submergence the max- repeated here. However, a very brief comparison of
imum displacement enters the non-linear area of the some results obtained by Schmucker and the ones ob-
resistance curve, or even exceeds the displacement tained with the SD load history is given in the further.
corresponding to ultimate capacity. In this area the
real resistance function is curved, while the approx-
Force [MN]

40
imated function used in the SDOF analysis consists
of only a few straight lines. That means that in US- 20
FOS, the transition into the post-ultimate region is 0
smoother, whereas in the SDOF model the transition 0 5 10 15
happens instantly, and in the post ultimate region the Time [s]
displacement then increases quickly.
Figure 10: Typical load history analyzed by
It turned out that it is difficult to correctly approx- Schmucker
imate the resistance curve with only a few straight
lines, and the SDOF-results are sensitive to the ap-
proximation of the resistance curve. 5.1 A brief comparison
Schmucker investigated three non-dimensional prop-
0.7 0.7 erties of the static resistance function, see also figure
umax [m] umax [m] 3:
0.6 0.6
SDOF, T = 14 s SDOF, T = 16 s kpe
USFOS, T = 14 s USFOS, T = 16 s Relative post elastic stiffness,
0.5 0.5 kel
Rres
Relative residual strength, rres =
0.4 0.4 Rult
kpu
0.3 0.3 Relative post ultimate stiffness,
kel
sd [m] sd [m]
0.2 0.2 He presented the normalized maximum displacement
1 0 1 1 0 1
(ductility demand) d = umax /uel versus the dynamic
Figure 9: Platform M: Max. displacements from overload r = fmax /Rult .
SDOF model vs. finite element model, for wave pe- This comparison is only carried out for different
riod (T) 14 s and 16 s, respectively values of the rres , that is kpe /kel and kpu /kel are not
varied. The SD load histories discussed in this pa- at Delft University of Technology in the Nether-
per generally show ductility demands in the range lands. At Stavanger University College, the support
of 1/3 of Schmuckers. The load histories are fun- from Professor Ivar Langen and Professor Ove To-
damentally different, since in the SQS load history bias Gudmestad is very much appreciated. The author
studied by Scmucker the forces stay large for a much would also like to thank Professor Jrgen Amdahl at
longer time. For an elastic-perfectly-plastic (EPP) Norwegian University of Science and Technology for
model with r = 1.3, the squared sinusoidal load his- useful comments on this paper.
tory demands a ductility of more than 13, while the
load history with an additional deck load impulse de- REFERENCES
mands only some 4.7. The results are summarized in Biggs, J. M. (1964). Introduction to structural dy-
Table 1. namics. McGraw-Hill.
Table 1: Ductility demands d vs. relative residual Dalane, J. I. and S. Haver (1995, May). Requali-
strength rres for different values of dynamic overload fication of an unmanned jacket structure using
r reliability methods. In Proceedings of the 27th
rres = 0.8 rres = 0.9 rres = 1.0 Annual Offshore Technology Conference 1995,
r SQS SD SQS SD SQS SD Houston, Texas, USA. OTC 7756.
1.3 34.4 11.9 22.7 7.7 13.5 4.7 Det Norske Veritas (1991, March). Environmen-
1.2 22.6 7.3 13.8 4.7 7.3 2.8 tal conditions and environmental loads. Oslo,
1.1 13.1 4.3 7.2 2.7 3.2 1.7 Norway. Classification note No. 30.5.
Emami Azadi, M. R. (1998, May). Analy-
Schmucker did not attempt to include wave-in-deck sis of static and dynamic pile-soil-jacket
loads explicitly in his SQS load history. However, the behaviour. Ph. D. thesis, Norges teknisk-
above results clearly illustrates the importance of do- naturvitenskapelige universitet (NTNU),
ing so, and the importance of doing it realistically. Trondheim, Norway. No. 1998:52.
Different load histories give widely different results. Gudmestad, O. T. and K. Hansen (2000, May). On
some research issues related to requalification
5.2 The influence of impulse duration and wave pe- of fixed steel jacket structures. In Proceedings
riod of the 10th International Offshore and Polar
Schmucker found no significant dependence on the Engineering Conference 2000, Seattle, Wash-
duration of the load for TTNw 3, where Tw is the ington, USA.
duration of the load and TN is the natural period of Hellan, ., J. Amdahl, B. Brodtkorb, T. Holmas,
the structure. There are, however, reasons to believe and E. Eberg (1998). USFOS Users Manual.
that the duration of the load on deck, will influence Trondheim, Norway: SINTEF report STF71
the ductility demand significantly, since this load du- F88039, rev. 98-04-01.
ration might be more in the area of the natural period
of the platform (a few seconds). HSE (1998, November). Dynamic push-over anal-
ysis of jacket structures. Technical Report OTO
6 CONCLUSIONS 98 092 (AME/ 37037B01/R/02.2), Health &
In this paper an investigation on the use of SDOF Safety Executive, United Kingdom. Prepared
models to predict dynamic response of a jacket plat- by Mott MacDonald Oil Gas & Maritime Di-
form has been presented. The SDOF model gives a vision.
reasonable estimate of the maximum displacement re- Moan, T., . Hellan, and M. R. Emami Azadi
sponse. However, there is still much work remaining (1997). Nonlinear dynamic versus static anal-
to improve the representation of jacket systems with ysis of jacket systems for ultimate limit state
SDOF models, and to further reveal the implications check. In Proceedings from International Con-
of using this very simple type of model to represent ference on Advances in Marine Structures III,
such complex systems as jackets. Dunfermline, Scotland.
The importance of including wave-in-deck forces
explicitly and at the correct time has been illustrated. Schmucker, D. G. (1996). Near-failure behavior
of jacket-type offshore platforms in the extreme
7 ACKNOWLEDGEMENT wave environment. Ph. D. thesis, Stanford Uni-
versity, Stanford, California, USA.
This work was initiated during the spring 2001, when
the author visited Professor J.H.Vugts and the Off- SINTEF (1998, September). Ultiguide phase 2.
shore Group at the Civil Engineering Department evaluation report on dynamic effects. Technical
Report SINTEF: STF22 F98685, DVN: DNV
98-3097, SINTEF and DNV.
Skallerud, B. and J. Amdahl (2001). Nonlinear
analysis of offshore structures, Chapter 13, Dy-
namic effects. To be published by Research
Studies Press Ltd.
Skjelbreia, L. and J. Hendrickson (1960, August).
Fifth order gravity wave theory. In Proceedings
of Seventh Conference on Coastal Engineering,
the Hague, the Netherlands, pp. 184 196.
Stewart, G. (1992, 14 - 19 June). Non-linear struc-
tural dynamics by the pseudo-force influence
part II: Application to offshore platform col-
lapse. In Proceedings of the 2nd International
Offshore and Polar Engineering Conference
1992, San Francisco, California, USA.

Potrebbero piacerti anche