Вы находитесь на странице: 1из 7

1st Civil and Environmental Engineering Student Conference

25-26 June 2012


Imperial College London

Wave-in-Deck Loading
M.A. Libot

ABSTRACT

In this study a numerical scheme has been developed that expands the momentum flux formulation developed by van de
nd
Graaf et al. (1995) to include a variety of wave models, ranging from Airy to 2 order random wave theories. The scheme
is also capable of assessing a variety of parameters, the most important of which is the deck orientation relative to the
incoming wave event. This paper will highlight the key features of this scheme. This is followed by a comparison between
the experimental work and the numerical scheme, examining the effect of the wave model and of varying the orientation
of the deck structure with respect to the wave field. The study has found that the momentum flux formulation is an upper
bound solution to the wave in deck problem, and that it is of equal importance to ensure the correct wave theory is
employed to model the sea state.

1. INTRODUCTION UNOCAL platform was sheared from the jacket


The offshore industry is one in which the due to wave crests impacting the deck horizontally
boundaries of mankind’s understanding of the seas (Bea et al., 1999). Whereas for other loading
is consistently pushed to the limit. Through phenomena, such as jacket loads, there are
exploration in deeper waters as well as harsher industry-wide standards, there is currently no
environments such as the Barents Sea and general consensus in the industry over how to
potentially the Arctic Sea there has been a drive to model these loads accurately.
develop increasingly accurate and complex wave Furthermore, there exist very few critical
models. These environments have led an reviews of the various wave-in-deck models, and
increased probability of offshore structures being little work has been done on extending the problem
impacted by extreme wave events (Roos, 2011). to incorporate 3D effects. The model presented in
Consequently this needs to be accommodated for this paper is the momentum flux formulation due to
in future design of offshore structures in order to its relative ease of application as well as sound
avoid potential wave-in-deck impacts. These occur theoretical background.
when the wave crest impacts the deck of the Therefore, it is clear that there needs to be a
structure. Furthermore, it has become increasingly considered review of the momentum flux
important to consider wave-in-deck loads when formulation through experiments. In addition an
reassessing existing offshore structures due to extension of the model to allow for a variety of
recent developments in extreme wave modelling potential sea states is required. This will allow an
as well as subsidence of existing fixed structures insight for other academics or persons interested in
due to extensive oil & gas extraction. Many existing further developing the momentum flux formulation
platforms have recently been found to have an of its accuracy and its drawbacks, as well as
inadequate, or out-dated, air gap and hence there providing an approach to extending the problem to
has been increased interest from both the industry a 3D environment.
and academia to enhance their understanding of
wave in deck impacts 1.2 Objectives
Two methods for calculating wave-in-deck The main objectives of this paper are presented
loads exist; the component approach and the below:
silhouette approach. This present paper will  Produce a numerical scheme that solves
present a numerical scheme that extends and the momentum flux formulation as
adapts the momentum flux formulation presented presented by van de Graaf et al. (1995).
by van de Graaf et al. (1995), a silhouette type
approach. This paper also examines the validity of  Extend application of the momentum flux
the momentum flux formulation through method to incorporate various wave
comparison to experimental work conducted in the theories as well as a varying deck
wave basin in the Hydrodynamics lab in Imperial orientation with respect to the wave event.
College.
 Investigate the validity of the momentum
1.1 Rationale flux formulation through experimental work
The potential dangers with which wave-in- in the Imperial College Wave Basin.
deck events are associated have been
demonstrated on numerous occasions. For
example, during Hurricane Hilda, the deck of the
1
2. PREVIOUS WORK load, however an additional forcing term related to
There is no general consensus within the the rate of change of the added mass is included:
offshore industry on which method to employ to
best calculate horizontal wave-in-deck loads. A 𝑑
𝐹𝑖 = (𝑀 𝑉 ) (2)
variety of models exist, some of which have been 𝑑𝑡 𝑖 𝑖
more extensively examined than others (van Raaij
& Gudmestad, 2007). The analytical models which Where 𝐹𝑖 is the impact force on member i, 𝑀𝑖
have been published can be split into two broad is the added mass acting on member i, and 𝑉𝑖 is
categories: the impact velocity on the member i. It should be
noted that work done using fully numerical models
 Silhouette Approach – models based on has shown that the assumption of no interference
the definition of a projected area. Includes is flawed, causing the Kaplan model to be
those described in API, DNV and BSI conservative compared to the API model (Chen &
standards as well as the momentum flux Yu , 2009). In addition, these methods are
formulation. computationally expensive.

 Detailed Component Approach – these 2.3 Momentum Flux Formulation


models examine loading on each individual The overarching principle behind the
component on the topside structure. This momentum flux method is the instant and complete
category includes the model proposed by dissipation of the horizontal momentum travelling
Kaplan (1976) as well as the Amoco and into the deck. In essence, this implies that the
Chevron in house models. portion of the crest of the wave that impacts the
deck is ‘shaved off’ (van de Graaf et al., 1995). A
The silhouette approach can be sub-divided critical advantage of the momentum flux method is
into two categories; drag formulation and that it does not rely on any empirical factors or
momentum formulation. The method proposed by coefficients to define the forcing on the deck.
van de Graaf et al. (1995) falls into the latter. Therefore it allows the formulation of a general
model which is applicable in various sea states and
2.1 Drag Formulation different environments without the need of
The models which fall under this category are calibration with different coefficients. Furthermore,
the simplest, and subsequently are the ones it allows for the calculation of the load time
proposed in the three industry standards; API, BSI histories, which are crucial to evaluating the
and DNV. The model proposed in API is based on dynamic response. The equations that make up
Morison’s drag load, but requires an empirical the method are shown in eq (3) and (4).
constant 𝐶𝑑 based on the porosity of the deck
𝜂
structure. The API model is given by:
𝐹𝑓𝑎𝑐𝑒 (𝑡) = ∫ 𝜌𝑏(𝑢(𝑡) + 𝑢𝑐 )2 𝑑𝑧 (3)

2
𝐹 = 0.5𝜌𝐶𝑑 𝐴𝑢 (1)

Where 𝜌 is the density, 𝐴 is the wetted area 𝑥


and 𝑢 is the horizontal particle velocity. It has been 𝐹𝑏𝑎𝑠𝑒 (𝑡) = ∫ 𝜌𝑏𝑣𝑢𝑑𝑥 𝑓𝑜𝑟 𝑣 > 0 (4)
𝑥
shown by both Roos (2011) and HSE (2001) that
the API formulation is a good method of quickly
estimating the order of magnitude of a wave in Where ℎ𝑑 is the height of the deck above
deck event but does not consider inertial, impact or SWL, 𝜂𝑓 is the height of the crest impacting the
pressure gradient effects (HSE, 2001). deck, 𝑏 is the deck width, the x co-ordinate of the
Furthermore, its inability to produce a time load crest is defined by 𝑥𝑐 (𝑡) and the intersection of the
history means it is of little value when examining front of the wave with the deck is defined as 𝑥𝑓 (𝑡).
dynamic response, something which carries great The face load is the load acting on the front face of
significance when considering events of this the structure and the base load is the load due to
nature. the vertically transported horizontal momentum
coming through the base of the deck as shown in
2.2 Detailed Component Approach figure 1.
As described by its name, this approach
involves the calculation of the load on each
component individually. The major assumption
underlying these models is that there is no
interference between the components and the
wave. The general consensus is that the model
proposed by Kaplan (1976) is the most detailed of
the three previously mentioned. The loads on each Figure 1: Sketch showing face and base load components
component are calculated using a Morison type taken from Roos (2011)

2
particle velocities need to be resolved into
The momentum flux method is considered to components perpendicular to the face. When
be an upper bound solution to the wave-in-deck finding the total face load, the loads on face 1 and
problem due to the assumption that all momentum face 2 are computed individually using trapezoidal
is instantaneously destroyed. This implies there is integration and subsequently summed. Due to the
no diversion of momentum in the form of run up, nature of the scheme, it had to be ensured that
something which is likely to happen in a real sufficient deck points were used to ensure
environment. convergence. It was found that in the vertical
direction convergence was instantaneous,
3. NUMERICAL MODELLING whereas 30 points/m were needed to model
The main focus of this section is the variation in the (𝑥, 𝑦) directions. The same is true
adaptation of the momentum flux formulation to a for the base load.
numerical scheme. Although the initial work done
by van de Graaf et al. (1995) was limited to an Airy 3.2 Base Loading
wave train impacting normal to the deck, the Although both face and base loads have been
numerical scheme aims to incorporate a wide incorporated into one scheme, they also exist
range of wave theories and conditions as well as independently due to their nature. This is because
deck orientations and inundations in an attempt to base loading is only prevalent in platforms which
predict the loading time history for wave in deck allow the ingress of water through the base of the
events. deck. Therefore this component may not apply to
Another key limitation of the method as certain deck structures. The user therefore has the
proposed by van de Graaf et al. (1995) is the option to choose what components of the loading
definition of the deck co-ordinates, allowing only for to investigate.
regular wave theory. Therefore, it was chosen to The scheme for the base load involves the
change the co-ordinate system to accommodate definition of a wave field which is 50% bigger than
for random waves with any focus position a: the deck to allow for rotation. The wave field data
is then interpolated to the co-ordinates on the deck
𝑎 ≤𝑥 ≤ 𝑎+𝐿 (5) grid using linear interpolation to facilitate
calculations. For a given time 𝑡𝑖 the wetted points
This paper will only present the main steps (𝑥𝑚 , 𝑦𝑚 ) for which 𝜂(𝑡𝑖 ) > ℎ𝑑 are computed. An
involved in modelling the loads for a deck structure example of the wetted base area for a directional
with varying orientation to the wave field. wave is given in figure 3.
40

3.1 Face Loading 30

The face loading is relatively straightforward 20

to convert into a numerical scheme; all that is 10

required is the definition of a deck face as well as


Y (m)

the calculation of horizontal particle velocities. The -10

calculation of particle velocities can be done -20

through the use of any analytical or numerical -30

wave theory, or can be input from fully nonlinear -40


-10 0 10 20 30 40 50 60 70

data obtained either empirically or numerically. X (m)

The first step is to define the deck grid. This Figure 3: Position of directionally spread NewWave event
𝝈𝜽 = 𝟑𝟎° at t=0.6s for a rotation of 𝟐𝟓°
deck grid will also be used to calculate the base
loads. An example deck grid at an orientation of Each point in the wetted area is treated as a
22.5° is shown in figure 2. ‘square’ with an area of 𝑑𝐴 = 𝑑𝑥 ∗ 𝑑𝑦. The
horizontal and vertical particle velocities on each of
these ‘squares’ are then calculated. The face load
0.4
for each wetted point at 𝑡𝑖 is then calculated using
0.3
Face 1 equation 5.
0.2

0.1 𝐹𝑏𝑎𝑠𝑒 (𝑡𝑖 ) = ∑ 𝜌𝑑𝐴 ∗ 𝑢(𝑥𝑚 , 𝑦𝑚 , ℎ𝑑 , 𝑡𝑖 )


(6)
∗ 𝑣(𝑥𝑚 , 𝑦𝑚 , ℎ𝑑 , 𝑡𝑖 )
y (m)

-0.1

This is then carried out for all time steps to


-0.2
generate the loading time history of the event.
-0.3
Face 2
-0.4
-0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
3.3 Validation of Scheme
x (m) In order to provide a validation of the
Figure 2:Discretised 2D mesh of deck structure at 𝟐𝟐. 𝟓° numerical scheme, a comparison to the results of
It is clear that the deck grid is rotated rather an analysis on the NRA structure by van de Graaf
than the wave field, which means the horizontal et al. (1995) is presented in figure 4.
3
time scale of 𝑡𝑠 = 1: 10 were adopted. It should be
noted that no substructure was included to avoid
any diffraction effects. The loads were measured
through the use of four 5kN piezoelectric load
sensors (Kistler 9313) which measured the
horizontal load.

Airy Wave Theory Wave-in-Deck Total Load with time at  =0


9

7
Figure 7: Photograph of model deck structure
6
Load (MN)

5
4.2 Unidirectional Broadside Impact
4 The numerical scheme was adapted to
3 incorporate the dimensions of the model deck as
2
well as the input parameters for the waves used in
the wave basin. It should be noted that regular
1
wave theory was not considered due to the effect
0
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 of water ramping in the basin when regular wave
Time (s)
trains were run. Figure 7 presents a comparison
Figure 4: Comparison between van de Graaf et al (1995) st
between the load time histories predicted by 1
predicted loads and loads predicted by numerical scheme nd
order and 2 order random wave theory and the
one observed experimentally for an inundation of
4. COMPARISON OF NUMERICAL SCHEME 34mm.
TO EXPERIMENTAL RESULTS
4.1 Experimental Setup 8

All experimental work was undertaken in the


‘wave basin’ at the Hydrodynamics Laboratory in 7

the Skempton Building at Imperial College London. 6


The basin has a working length of 20m, a width of
12m and a water depth of 1. 5m as shown in figure 5

5. However, for the experiments undertaken, the


Load (N)

4
basin was equipped with a faux bed, reducing the
effective water depth to 1.265m. 3

0
31.2 31.3 31.4 31.5 31.6 31.7 31.8 31.9 32 32.1 32.2
Time (s)
9

5
Load (N)

3
Figure 5: Layout of Wave Basin (Roos, 2011)
2
The deck structure used in the experiments
1
was a 0.5𝑚 ∗ 0.65𝑚 ∗ 0.12𝑚 structure made of
green acoustic foam. A photograph is provided in 0
31.2 31.3 31.4 31.5 31.6 31.7 31.8 31.9 32 32.1 32.2
figure x. The bottom of the deck was situated at Time (s)

0.1m above SWL. A length scale 𝑙𝑠 = 1: 100 and a Figure 6: Comparison of load time histories for inundation of
34mm, (a) LRWT (red), (b) 2nd order random (red)
4
For an inundation of approximately 34mm it with broadside impacts. However, in a real sea
can be seen that the numerical scheme under state the wave event is unlikely to strike the deck
predicts the load when using linear random wave perpendicular to its direction of travel. Figure 9
theory. This would appear to be in contrast with the shows the trends associated with varying the deck
notion that the method is conservative. However, orientation relative to the wave field for both the
this concept is based on the implicit principle that numerical scheme and the experiments conducted.
the kinematics associated with the event are
20
correctly modeled. Figure 8 shows that there is an
increase of 34% in crest elevation in the generated 19
wave event when compared to linear random wave
18
theory due to nonlinear terms. It is therefore
reasonable to assume that the kinematics will not 17

Peak Face Load (N)


be correctly modelled when employing linear 16
random wave theory for the event described in
figure 7 (a). 15

14

14
13
12
12
10
11
8 0 10 20 30 40 50 60
Deck Orientation (deg)
6
Figure 9: Variation in peak face load with deck orientation,
4 linear random wave theory (blue), 2nd order random wave
 (m)

theory#(green) and experimental results (black)


2

0
The effect of varying the deck orientation from
-2 0° to 56.25° for an inundation of 56mm, with the
-4 black line representing the experimental result is
-6 clearly visible. It can be seen that the
-8
experimentally observed loads show good
30 30.5 31 31.5 32 32.5 33 33.5 34 nd
Time (s)
correlation with 2 order random wave theory. It
Figure 8: Surface elevation time-histories, η(t), for 1st Order should be noted that the experimental results are
NewWave theory (black), 2nd Order NewWave Theory (red) originally higher than the numerically predicted
and Experimental results (blue) for input amplitude 0.1m
ones, only to drop below them as the deck
orientation varied from 0° to 56.25°. This is due to
This is validated when modelling the same an additional feature that is not modelled in the
nd
event with a 2 order random wave theory, as momentum flux formulation method; namely the
shown in figure 7 (b). In this case, the numerical diversion of momentum around the structure.
predictions are in excess of the experimentally When impacting the face perpendicular to the
observed loads, confirming the conservative nature direction of the group velocity, the momentum can
of the method. This is further highlighted when only be diverted in the form of run up, but this is
considering figure 8. The wave event that causes a limited. However, as the structure tends towards
34mm deck inundation has a crest elevation 24% 45° the momentum can be diverted through both
nd
higher than the one predicted by 2 order random run up as well as around the faces of the structure.
theory. Some of this is alleviated by the fact that This causes the load to decrease with respect to
the wave used in the numerical scheme has an the numerically predicted peak load, and further
input amplitude of 134mm, however the kinematics strengthens the argument that the momentum flux
are still underestimated in comparison to those formulation is an upper bound solution to the wave
underneath the wave crest in the basin. This is in in deck problem when modelled with the correct
contrast with what is shown in figure (b), wave theory.
confirming the conservative nature of the It is also apparent that once the structure has
momentum flux formulation. This was also rotated past 45° there is a dramatic increase in the
confirmed throughout the experiments by the peak load. This is due to face 2 (see figure 2)
presence of momentum diversion in the form of run becoming the dominant face, and due to it being
up, with the waves often splashing over the top of 30% longer than face 1 the loads are expected to
the structure. This diversion of momentum implies be higher at 56.25° than at 33.75°.
that not all the momentum is instantaneously and
totally dissipated. 4.4 Effect of Deck Orientation (Directional)
As with unidirectional waves, the effect of the
4.3 Effect of Deck Orientation angle at which the wave enters the deck structure
(Unidirectional) is of importance to adequately model a real world
One of the main limitations of the van de situation. The effect of deck orientation was only
Graaf et al. (1995) method is that it can only deal
5
investigated for the 𝜎𝜃 = 30° and an inundation of The conclusion that the choice of wave theory
56mm case due to time constraints, and the is of crucial importance is strongly linked to the
variation in peak load with deck orientation is conclusion that the momentum flux formulation is
compared to the numerical predictions in figure 10. conservative in describing horizontal wave in deck
loading. All the waves produced in the basin have
18
been shown to require higher order terms than
nd
17 provided by 2 order random wave theory to
16 accurately match measured surface profiles.
However, for an inundation of 34mm the
15
momentum flux formulation over predicts the peak
load, indicating the conservative nature of the
Peak Load (N)

14

13 model. Nevertheless, if the correct wave model is


12
not employed the scheme can provide a lower
bound to problem, something which is undesirable.
11
In conclusion, the momentum flux formulation
10 is conservative in nature, primarily due to the
9
assumptions of complete and instantaneous
momentum dissipation as well as 100% porosity of
8
0 10 20 30 40 50 60 the deck structure, but attention has to be paid to
Deck Orientation (deg)
the wave model employed.
Figure 10: Comparison of numerical scheme (blue) and
experiments 𝝈𝜽 = 𝟑𝟎° (green) with varying deck orientation for
a 56mm inundation
6. ACKNOWLEDGEMENTS
First and foremost I would like to thank my
The trend of a decreasing peak load with
supervisor Prof. C. Swan for the opportunity to
deck orientation tending towards 45° is also visible
conduct this work as well as the guidance he
with directional waves, but is much less
provided throughout. Special thanks also need to
pronounced than with unidirectional waves. This
be given to M. Latheef and J. Haley for assistance
can most probably be associated with the
with technical and other queries. I would also like
transverse components of the velocity, which are
to thank Bill and Dave, without whom I would not
zero in a long crested sea state, acting on the
have been able to conduct the laboratory
more exposed face. Studies done on large offshore
experiments.
structures by de St. Q. Isaacson and Sinha (1986)
7. REFERENCES
have demonstrated that the transverse
components of the force are substantial and
increase with directional spreading. This can API, 2000. API RP2A-WSD: Recommended practice for
explain the reason for the experimentally observed planning, designing and constructing fixed offshore
loads being higher than the loads observed for the platforms - working stress design. 2nd ed. Washington
same unidirectional effect. Although the D.C.: American Society of Petroleum Engineers
unidirectional waves also predicted an increase in Publishing Services.
the peak load once the structure has rotated past Bea, R.G., Iversen, R. & Xu, T., 2001. Wave-in-Deck Forces on
45°, the directional experiments show a larger Offshore Platforms. Journal of Offshore Mechanics and
Arctic Engineering, 123, pp.10-21.
increase. These transverse components need to Chen, H.-C. & Yu , K., 2009. CFD simulations of wave-current
be resolved through less of an angle on the long body interactions including greenwater and wet deck
face when it is at 56.25°, thereby increasing the slamming. Computers & Fluids, 38(5), pp.970-80.
load. These transverse components have not been de St. Q. Isaacson, M. & Sinha, S., 1986. Directional wave
effects on large offshore structures. Journal of
included in the numerical scheme, leading to the Waterway, Port and Coastal Engineering, 112.
increase in peak load at 56.25° to be less defined. HSE, 2001. Offshore Technology Report OTO 2001 034:
This has also caused the numerically predicted Assessment of the effect of wave-in-deck loads on a
directional loads to always be lower than the typical jack-up. Sheffield: HSE Prepared by MSL
Engineering on behalf of HSE.
predicted unidirectional, although it was observed Kaplan, P. & Silbert, M., 1976. Impact forces on platform
in the basin that at 45° the 𝜎𝜃 = 30° produces a horizontal members in the splash zone. In Proceedings
higher peak load than the same unidirectional of 8th Offshore Technology Conference, Paper No.
event. OTC 2498. Houston, Texas, 1976. Offshore Technology
Conference.
Roos, J., 2011. Wave-structure interaction: The effective
5. CONCLUSIONS prediction of wave-in-deck loads. PhD Thesis. Imperial
The two main conclusions drawn from this College London.
study is that the choice of wave model is of great van de Graaf, J.W., Tromans, P. & Vanderschuren, L., 1995.
Wave loads on decks. Offshore Structures Engineering
significance and that the momentum flux Newsletter, No. 10.
formulation correctly captures the load time trace, van Raaij, K., 2005. Dynamic behaviour of jackets exposed to
even though it has been shown to be conservative. wave-in-deck forces. PhD Thesis. University of
The trends of decreasing peak load with changing Stavanger, Norway.
van Raaij, K. & Gudmestad, O., 2007. Wave-in-deck loading on
deck orientation between 0° and 45° have also fixed steel jacket decks. Marine Structures, 20(3),
been corroborated by the experimental results. pp.164-84.

6
7

Вам также может понравиться