You are on page 1of 14

Applied Ocean Research 142 (2024) 103802

Contents lists available at ScienceDirect

Applied Ocean Research


journal homepage: www.elsevier.com/locate/apor

Research paper

Stochastic dynamic analysis of floating bridges exposed to inhomogeneous


and irregular waves
Knut Andreas Kvåle a ,∗, Bernt Leira b , Ole Øiseth a
a Department of Structural Engineering, Faculty of Engineering, NTNU, Norwegian University of Science and Technology, Trondheim, Norway
b
Department of Marine Technology, Faculty of Engineering, NTNU, Norwegian University of Science and Technology, Trondheim, Norway

ARTICLE INFO ABSTRACT

Keywords: As future floating bridges gets longer, the likelihood of significant inhomogeneous wave conditions across them
Floating bridge increase. Consequently, a proper assessment of the effects of inhomogeneous wave conditions is needed. In this
Inhomogeneity paper, we present an approach to model inhomogeneous sea states by representing the stochastic process using
Wave excitation
generalized harmonic decomposition, retaining the coherency inherent in Airy wave theory. The methodology
Coherency
is applied to a concept model of a floating bridge to cross the Bjørnafjord in Norway, which is more than five
Frequency domain
Time domain
kilometres wide. The paper emphasizes on frequency-domain simulation, but the required steps to generate
time-domain representations are given. To suggest what might be critical considerations in design, the response
effects due to some selected conceptual inhomogeneities are highlighted. For varying wave heights alone, it is
found that it is likely conservative to approximate the inhomogeneous sea state as homogeneous by assuming
the harshest conditions across the full width of the strait. By modulating the mean wave angle of a 100-
year swell sea state, a spherical wave front matching the curvature of the bridge girder is found to produce
significantly more severe response than the homogeneous reference swell; an increase of 14% of the maximum
axial force response is found. Furthermore, a proper treatment of the coherency for swell sea states is found
to be crucial, in contrast to typical wind sea conditions.

1. Introduction assumed across the bridge spans. As the benefit of floating bridges over,
e.g., suspension bridges, in general is larger for longer spans, methods
The assumption of homogeneous wave conditions have been the to account for the effect of inhomogeneous wave conditions are highly
de facto standard for engineers when designing floating structures. relevant. An approach relying on the assumption of homogeneous wave
This implies that the properties associated to the sea state, such as, conditions would have to introduce unnecessary conservativism to
e.g., wave height, peak period, spectral shape and wave direction, have ensure designs with sufficient safety; excessive safety margins and large
been assumed identical across the span of the structure. Inhomogeneity costs may be the consequences.
is a topic more commonly considered in wind engineering, as long- Floating bridges are currently highly relevant in a Norwegian con-
spanning structures, such as suspension bridges, traditionally have text, as the Norwegian Public Roads Administration (NPRA) is planning
been mainly exposed to wind excitation. Wave inhomogeneities have, and building the Coastal Highway Route E39 along the west coast, to re-
however, been of less importance, as ocean engineering as a field has place currently operating ferries with bridges or tunnels. The deep and
been dominated by structures with limited spatial extents. As floating wide fjords along the planned route call for record-breaking and novel
bridges now are considered as viable options for several very wide fjord bridge designs. The channelling effects in and the local obstructions
crossings, inhomogeneous waves are more relevant than ever. present around the fjords combined with the long spans make assess-
Currently, there exist only a handful of long-span floating bridges ments of the effects of inhomogeneity vital. The inhomogeneity across
in the world, and only three of these are located in harsh marine one of the straits involved in the NPRA-project, the Bjørnafjord, was
environments (Watanabe, 2003; Kvåle, 2017). Floating bridges may thoroughly investigated in Cheng et al. (2019). This paper concluded
serve as cost-efficient permanent links for very wide crossings, due to that the wave conditions are significantly inhomogeneous in the fjord.
the vertical support provided by the buoyancy. Despite the high vertical Wei et al. (2017, 2018) suggest methods to estimate the response of
stiffness, floating bridges may be prone to relatively lively lateral dy- floating structures exposed to inhomogeneous waves, in time domain
namic behaviours. Most commonly, homogeneous wave conditions are

∗ Corresponding author.
E-mail address: knut.a.kvale@ntnu.no (K.A. Kvåle).

https://doi.org/10.1016/j.apor.2023.103802
Received 16 June 2023; Received in revised form 7 October 2023; Accepted 12 November 2023
Available online 27 November 2023
0141-1187/© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

and frequency domain, with emphasis on structures consisting of dis- (1979)):


crete panel modules. There also exist several papers in the literature
discussing the inhomogeneity of wave conditions in fjords and their 𝜂({𝑥}, 𝑡) = e𝑖(𝜔𝑡−{𝜅}⋅{𝑥}) 𝑑𝑍𝜂 ({𝜅}, 𝜔) (1)
∫{𝜅},𝜔
effects on the dynamic behaviour of floating bridges, e.g., Cheng et al.
where {𝜅} = [𝜅𝑥 , 𝜅𝑦 ]𝑇 is the wave number vector, 𝜔 is the circular
(2018), Dai et al. (2020, 2021). In these papers, the coherency between √
the wave elevation at the different pontoon locations are not fully frequency, 𝑖 = −1, and 𝑍𝜂 is the spectral process describing the
considered, but the wave excitations of the different pontoons are either wave elevation. The wave number vector can be rewritten as {𝜅} =
assumed fully correlated or uncorrelated. Cui et al. (2022) suggested a 𝜅[cos 𝜃, sin 𝜃]𝑇 , where 𝜃 is the wave angle and 𝜅 = |{𝜅}| is the wave
methodology to account for the measured coherence from measurement number. The spectral process is related to the cross-power spectral
campaigns at the planned location for the Bjørnafjord crossing, with density (CPSD) in the following manner (Sigbjörnsson, 1979):
( )
emphasis on time-domain simulation, and applied the methodology 𝐸 𝑑𝑍𝜂𝑝 (𝜔, 𝜃)𝑑𝑍𝜂𝑞 (𝜔, 𝜃)∗ = 𝑆𝜂,𝑝𝑞 (𝜔, 𝜃)𝑑𝜃𝑑𝜔 (2)
to predict the response of a conceptualized model of the Bjørnafjord
Bridge. Here, 𝐸[⋅] is defined as the mathematical expectation; the indices 𝑝 and
Herein, the stochasticity of the inhomogeneous wave process is 𝑞 refer to two points on the water surface; ∗ is used to denote complex
introduced in a format using generalized harmonic decomposition, conjugacy; and 𝑆𝜂,𝑝𝑞 (𝜔, 𝜃) is the two-dimensional CPSD between the
and the wave kinematics are defined from linear wave theory. Con- sea surface elevations at points 𝑝 and 𝑞, which is a function of the
sequently, the coherency inherent in the linear wave description is wave angle and the frequency. For homogeneous waves, the following
retained. A proper representation of coherency is considered crucial for expression is used to establish the CPSD between wave elevations at
long-span floating bridges. Available literature focuses mainly on time- positions 𝑝 and 𝑞:
domain descriptions, whereas the current paper will rely on frequency- 𝜋

domain representations of the wave-induced processes until the final, 𝑆𝜂,𝑝𝑞 (𝜔) = 𝑆𝜂 (𝜔, 𝜃)𝑒𝑖{𝜅}⋅{𝛥𝑥} 𝑑𝜃 (3)
∫−𝜋
optional, step to simulate wave excitation time histories using Monte
where {𝛥𝑥} = {𝑥𝑞 } − {𝑥𝑝 }, and {𝑥𝑝 } and {𝑥𝑞 } are the coordinate
Carlo simulation. The objective of the current paper is twofold. First,
vectors of points 𝑝 and 𝑞, respectively. The exponential term, termed the
we will introduce a systematic methodology to model inhomogeneous
two-dimensional coherency function, describes to what degree a wave
sea states for both frequency-domain and time-domain analysis of component with a given frequency and direction is correlated between
floating bridges. The frequency domain approach is considered very points 𝑝 and 𝑞:
valuable in both early exploratory stages of design, screening of crucial
environmental conditions, and for applications requiring numerous 𝛾(𝜃, 𝜔) = 𝑒𝑖{𝜅}⋅{𝛥𝑥} (4)
simulations, such as fatigue and extreme response analyses. Second, we
Note that this term is unity and that the resulting CPSD of wave
will demonstrate some of the potential consequences of inhomogeneity
elevation given in Eq. (3) reduces to the one-dimensional wave PSD
with some principal and extreme-scenario examples. This will both
when the two points 𝑝 and 𝑞 coincide.
showcase the methodology and aid to suggest what aspects to consider For most engineering applications, it is assumed that the two-
when establishing design bases for future bridges. dimensional wave spectral density can be decomposed as follows:
The methodology applied to treat the fluid–structure interaction is
the same as in Kvåle et al. (2016); it is based on the combination of 𝑆𝜂 (𝜔, 𝜃) = 𝑆𝜂 (𝜔)𝐷(𝜃) (5)
a finite element (FE) model of the structural components of the bridge
where 𝑆𝜂 (𝜔) is the one-dimensional wave spectral density and 𝐷(𝜃) is
with hydrodynamic properties obtained from a potential theory solver.
the directional distribution, which are characterizing how the wave
The analysis relies on the decomposition of the equation system into
energy is distributed over frequency and wave directions, respectively.
the coordinate basis given by the mode shapes of the structural FE
Note that this decomposition is not a prerequisite for the methodology
model, and thus enables a very efficient simulation. The major part of presented, but rather a convenient step for the modelling of the sea
the results presented herein are based on simulations in the frequency states.
domain. As demonstrated, the methodology can easily be applied in the In the derivation of Eq. (3), homogeneous wave conditions are
time domain by the virtue of Cholesky decomposition and Monte Carlo assumed. If no such assumption is introduced, a similar expression
simulation of the spectral process of the wave excitation. can be established for inhomogeneous conditions, by exploiting the
Due to the very flexible behaviour of bridges with these span structure of Eq. (2):
lengths, wind excitation will also be an important effect to be consid- 𝜋√ √
ered in design analyses, but not considered to be within the scope of 𝑆𝜂,𝑝𝑞 (𝜔) = 𝑆𝜂,𝑝 (𝜔, 𝜃)𝑒−𝑖{𝜅}⋅{𝑥𝑝 } 𝑆𝜂,𝑞 (𝜔, 𝜃)𝑒𝑖{𝜅}⋅{𝑥𝑞 } 𝑑𝜃
∫−𝜋
this paper. Furthermore, nonlinear wave loads are not considered. A 𝜋√ √
similar treatment of inhomogeneous wave excitation by using quadratic = 𝑆𝜂,𝑝 (𝜔, 𝜃) 𝑆𝜂,𝑞 (𝜔, 𝜃)𝑒𝑖{𝜅}⋅{𝛥𝑥} 𝑑𝜃 (6)
transfer functions (QTFs) could be envisioned, but would likely be ∫−𝜋
rather computationally heavy, also in the frequency domain. Tradition- where 𝑆𝜂,𝑝 (𝜔, 𝜃) and 𝑆𝜂,𝑞 (𝜔, 𝜃) correspond to the two-dimensional wave
ally, for engineering applications, Newman’s approximation is applied spectral densities at points 𝑝 and 𝑞, respectively. Furthermore, this can
to account for the difference effects from quadratic wave loads. As be rewritten as:
this only considers the amplitude and does not account for the phases 𝜋
𝑆𝜂,𝑝𝑞 (𝜔) = 𝑍𝑝 (𝜔, 𝜃)𝑍𝑞 (𝜔, 𝜃)∗ 𝑑𝜃 (7)
of the excitation, it might be a problematic approach when applied ∫−𝜋
to long-span rather than single-point structures, as it was originally √
Here, 𝑍𝑛 (𝜔, 𝜃) = 𝑆𝜂,𝑛 (𝜔, 𝜃)𝑒−𝑖{𝜅}⋅{𝑥𝑛 } with 𝑛 = 𝑝, 𝑞. As the points 𝑝 and 𝑞
intended for. Therefore, the current paper does not account for wave
are arbitrary, this may be expressed as a function of position, i.e., 𝑍𝑛 =
drift excitation based on Newman’s approximation, even though it
𝑍(𝜔, 𝜃, {𝑥𝑛 }). Furthermore, the correlation of the wave elevations at
would be straightforward to include.
positions 𝑝 and 𝑞 can be obtained as follows:

2. Theoretical outline 𝜎𝑝𝑞 ∫ 𝑆𝑝,𝑞 (𝜔)𝑑𝜔


𝜌𝑝,𝑞 = = √ (8)
𝜎𝑝 𝜎𝑞
∫ 𝑆𝑞,𝑞 (𝜔)𝑑𝜔 ∫ 𝑆𝑝,𝑝 (𝜔)𝑑𝜔
2.1. Water surface elevation
As demonstrated in Kvåle et al. (2016), this can be used to establish
The random-field sea surface can be expressed by the following correlation fields of wave elevation, to assess the extents of correlated
Riemann–Stieltjes integral (see, e.g., Kinsman (1965), Sigbjörnsson wave elevations.

2
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

2.2. Hydrodynamic action

The linear hydrodynamic excitation acting on pontoon 𝑞 can, by


using generalized harmonic decomposition, be expressed as:

{𝑃𝑞 (𝑡)} = 𝑒𝑖𝜔𝑡 {𝑑𝑍𝑃𝑞 (𝜔)} (12)
∫−∞
where {𝑍𝑃𝑞 (𝜔)} is the spectral process associated with the excitation of
pontoon 𝑞. The force components in {𝑃𝑞 (𝑡)} are stacked in a typical FE
format, i.e., with three forces followed by three moments. Furthermore,
the wave process can be described through:

{𝑑𝑍𝑃𝑞 (𝜔, 𝜃)} = {𝑄𝑞 (𝜔, 𝜃)}𝑑𝑍𝜂𝑞 (𝜔, 𝜃) (13)


where {𝑄𝑞 (𝜔, 𝜃)} is the directional hydrodynamic transfer function
relating the wave amplitude to wave-induced forces on pontoon 𝑞.
The zero-mean Gaussian hydrodynamic excitation acting on the two
pontoons 𝑝 and 𝑞 can be fully characterized by the CPSD matrix (Langen
and Sigbjörnsson, 1980). By combining Eq. (13) with Eqs. (2)–(7), we
can establish the following wave excitation CPSD matrix:
[𝑆𝑝,𝑞 (𝜔)]
𝜋√ √
= 𝑆𝜂,𝑝 (𝜔, 𝜃)𝑒−𝑖{𝜅}⋅{𝑥𝑝 } {𝑄𝑝 (𝜔, 𝜃)} 𝑆𝜂,𝑞 (𝜔, 𝜃)𝑒𝑖{𝜅}⋅{𝑥𝑞 } {𝑄𝑞 (𝜔, 𝜃)}∗𝑇 𝑑𝜃
∫−𝜋
(14)
𝑇
where is the matrix transpose. Note that for 𝑝 = 𝑞, this reduces to
the CPSD matrix (including auto terms) between all force and moment
Fig. 1. Contour plot of two-dimensional wave spectral density, 𝑆𝜂 (𝜔, 𝜃), due to 100- components acting on a single pontoon:
year wind sea. The square in the central plot indicates the area (variance) of a single
harmonic contribution to the full time-domain representation. [𝑆𝑞,𝑞 (𝜔)] = 𝑆𝜂,𝑞 (𝜔, 𝜃){𝑄𝑞 (𝜔, 𝜃)}{𝑄𝑞 (𝜔, 𝜃)}∗𝑇 𝑑𝜃 (15)

Eq. (14) can be rewritten, for efficient software implementation, as
follows:
2.1.1. Time-domain simulation of water surface elevation 𝜋
The expression given in Eq. (6) can be used as basis for time-domain [𝑆𝑝,𝑞 (𝜔)] = {𝑍𝑝 (𝜔, 𝜃)}{𝑍𝑞 (𝜔, 𝜃)}∗𝑇 𝑑𝜃 (16)
∫−𝜋
simulation of the water surface elevation, to yield the following expres- √
where {𝑍𝑛 (𝜔, 𝜃)} = 𝑆𝜂,𝑛 (𝜔, 𝜃)𝑒−𝑖{𝜅}⋅{𝑥𝑛 } {𝑄𝑛 (𝜔, 𝜃)} with 𝑛 = 𝑝, 𝑞. This
sion describing the wave elevation at location {𝑥} = [𝑥, 𝑦]𝑇 (Naess and
can straightforwardly be defined differently for different positions and
Moan, 2012): pontoon types. The global FE-format excitation spectral density matrix
(𝑅 ) can finally be constructed as:

𝑁 ∑
𝜂({𝑥}, 𝑡) = 𝐴𝑘,𝑟 𝑒𝑖𝜔𝑘 𝑡 (9)
⎡ [𝑆1,1 (𝜔)] [𝑆1,2 (𝜔)] … [𝑆1,𝑀 ] ⎤

𝑘=1 𝑟=1 ⎢ ⎥
⎢ [𝑆2,1 (𝜔)] [𝑆2,2 (𝜔)] … [𝑆2,𝑀 ] ⎥
𝐴𝑘,𝑟 = 2𝑆𝜂 (𝜔𝑘 , 𝜃𝑟 )𝛥𝜔𝑘 𝛥𝜃𝑟 𝑒−𝑖{𝜅𝑘 }⋅{𝑥}+𝑖𝛼𝑘𝑟 (10) [𝑆𝑃 (𝜔)] = ⎢ ⎥ (17)
⎢ ⋮ ⋮ ⋱ ⋮ ⎥
which is the equivalent to the inverse discrete Fourier transform, easily ⎢ ⎥
⎣[𝑆𝑀,1 (𝜔)] [𝑆𝑀,2 (𝜔)] … [𝑆𝑀,𝑀 ]⎦
solved by applying the inverse fast Fourier transform (IFFT), of the
where each submatrix [𝑆𝑝,𝑞 ] is established from Eq. (16) and 𝑀 is the
following expression (assuming correct use of scaling factors):
total number of pontoons.
𝑅 √

𝐵𝑘 (𝑥, 𝑦, 𝜔) = 2𝑆𝜂 (𝜔, 𝜃𝑟 )𝛥𝜔𝛥𝜃𝑟 e−𝑖{𝜅𝑘 }⋅{𝑥} ⋅ e𝑖𝛼𝑘𝑟 (11) 2.2.1. Time-domain representation
𝑟=1 Although not considered within the scope of the current paper,
Here, 𝑟 is the angle index ranging from 1 to 𝑅, 𝛥𝜃 is the angle nonlinear analysis is normally needed in design. A prerequisite to
increment, 𝛥𝜔 is the frequency increment, and 𝛼𝑘𝑟 is the uniformly this is a time-domain representation of the hydrodynamic action de-
scribed in Eq. (16). Following this approach, any relative motion of
distributed random phase angle associated with frequency 𝜔𝑘 and wave
the structure with respect to the wave field is not accounted for,
angle 𝜃𝑟 . The two-dimensional wave spectral density of an example
but the nonlinear behaviour of the structure itself can be. The time-
sea state (100-year wind sea) is shown in Fig. 1 with an annota- domain representation of the hydrodynamic action can be established
tion indicating how this could be sampled, one frequency-direction either based on a closed-form technique in line with the simulation
combination at the time, to obtain a realization of the spectral pro- of wave elevation in Eq. (9) or through matrix decomposition of the
cess from the summation of all harmonic contributions. Note that the cross-spectral density matrix (Shinozuka, 1972; Shinozuka and Wai,
elevation field simulated by these expressions does not maintain its 1979; George, 1996). The Cholesky decomposition can be exploited to
ergodic properties (Jefferys, 1987). Furthermore, as the waves from decompose the spectral density matrix:
all directions are assumed at the identical discrete frequency values, [𝑆𝑃 (𝜔)] = [𝐵(𝜔)][𝐵(𝜔)]∗𝑇 (18)
it is known to produce fields with potential issues with interference
patterns. However, the gridded approach is a prerequisite for enabling Using this, the excitation vector time history can be simulated as
follows (Shinozuka, 1972; Langen and Sigbjörnsson, 1980):
the use of the FFT or IFFT for efficient simulation, and by using a fine ( 𝑗 𝑁 )
resolution this is a minor concern as it is merely used for illustrational ∑ ∑√
𝑝𝑗 (𝑡) = R 2𝛥𝜔𝑘 𝐵𝑗𝑚 (𝜔𝑘 )e𝑖(𝜔𝑘 𝑡+𝜙𝑚𝑘 ) (19)
purposes herein. 𝑚=1 𝑘=1

3
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

Fig. 2. Map section showing location and local topography around the bridge site. Map sections: © Kartverket.

2.3. Response prediction

Herein, we relied on a frequency-domain representation and the


power-spectral density method to solve the response quantities induced
by the prescribed sea states. Pontoons are considered rigid bodies, as
the flexibility of the girder is dominating. The equation of motion is
assumed constructed as follows in hybrid time–frequency domain (see,
e.g., Langen and Sigbjörnsson (1980)):
( ) ( )
[𝑀𝑠 ] + [𝑀ℎ (𝜔)] {𝑢}̈ + [𝐶𝑠 ] + [𝐶ℎ (𝜔)] {𝑢}
̇ (22)
( )
+ [𝐾𝑠 ] + [𝐾ℎ ] {𝑢} = {𝑃 (𝜔)}e𝑖𝜔𝑡

Here, [𝑀𝑠 ], [𝐶𝑠 ] and [𝐾𝑠 ] are the structural mass, damping and stiffness
matrices, respectively; {𝑢} is the displacement response vector; [𝑀ℎ (𝜔)]
and [𝐶ℎ (𝜔)] are the hydrodynamic added mass and damping matrices;
Fig. 3. Render of the concept bridge developed by the AMC consultant group.
[𝐾ℎ ] is the hydrostatic restoring stiffness matrix; and {𝑃 (𝜔)} is the
Illustration by Dissing+Weitling (part of AMC). The render corresponds to an updated complex-valued amplitude vector corresponding to the excitation of fre-
concept model. quency component 𝜔. The inertia of the pontoons is assumed included
in [𝑀𝑠 ]. Finally, the response induced by the given excitation can be
established through the power-spectral density method:

[𝑆𝑢 (𝜔)] = [𝐻(𝜔)]𝑇 ∗ [𝑆𝑃 (𝜔)][𝐻(𝜔)] (23)


where R denotes the real part; 𝛥𝜔 is the frequency resolution; 𝑁 is
where [𝐻(𝜔)] is the discrete frequency response function (FRF) matrix,
the number of discrete frequency components; 𝜙𝑚𝑘 is the uniformly
defined through [𝐻(𝜔)]−1 = −𝜔2 [𝑀(𝜔)] + 𝑖𝜔[𝐶(𝜔)] + [𝐾]. Note that the
distributed random phase angle between 0 and 2𝜋 corresponding to
frequency dependency of the matrices [𝑀(𝜔)] = [𝑀ℎ (𝜔)] + [𝑀𝑠 ] and
degree of freedom (DOF) 𝑚 and frequency component 𝑘, i.e., 𝜔𝑘 ; 𝐵𝑗𝑚 (𝜔)
[𝐶(𝜔)] = [𝐶ℎ (𝜔)] + [𝐶𝑠 ] does not affect the implementation or efficiency
is component 𝑗, 𝑚 of the lower triangular matrix [𝐵(𝜔)]; and 𝑝𝑗 (𝑡) is
of this approach.
component 𝑗 of the excitation vector {𝑝ℎ (𝑡)}. This can be rewritten as:
( 𝑗 𝑁 )
∑ ∑√ 3. Bjørnafjorden bridge case study
𝑝𝑗 (𝑡) = R 2𝛥𝜔𝑘 𝐵𝑗𝑚 (𝜔𝑘 )e𝑖𝜙𝑚𝑘 e𝑖𝜔𝑘 𝑡 (20)
𝑚=1 𝑘=1
The Bjørnafjord is located just south of Bergen, and is one of the
After assuming that 𝛥𝜔 = 𝛥𝜔𝑘 for all frequency components, the most challenging crossings along the planned Coastal Highway Route
equation above can be furthermore rewritten by the use of the IFFT: E39. As it is situated between the major cities of Stavanger and Bergen,
( 𝑗 )
∑ (√ ) it is clear that a bridge across it would play a significant role to
−1 𝑖𝜙𝑚𝑘
𝑝𝑗 (𝑡) = R F 2𝛥𝜔𝐵𝑗𝑚 (𝜔)e (21) transport and industry in the region. The location of the fjord and the
𝑚=1
planned position of the bridge is depicted in Fig. 2. To cross the fjord
where F−1 denotes inverse Fourier transform, in practice solved by the along the planned route, a bridge with a length well above 5000 m is
IFFT. This implies the generation of 𝑁 × 𝑀 random phase angles in needed. Regardless of the specifics of the solution, a bridge like this
total; one for each combination of DOF and frequency component. will break records and require extension of current bridge technology.

4
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

Fig. 4. Overview of components and geometry of the bridge model. Note that the pontoons use a local coordinate system that differs from the local coordinate system of the
girder.

Fig. 5. Technical drawing of the bridge girder cross section. Units are in mm.
Source: Reproduced from (Norconsult AS, 2017).

Table 1
girder to the pontoons, whereas bar elements are used to represent the
Structural properties of girder. cables. The structural properties of the girder are summarized in Ta-
Property Value ble 1. The FE model does not include any hydrodynamic contributions
Area, 𝐴 1.43 m2
from the pontoons; only the inertia from pontoons and the hydrostatic
Weak-axis second moment of area, 𝐼𝑦𝑦 2.67 m4 restoring stiffness are included. The hydrodynamic transfer functions
Strong-axis second moment of area, 𝐼𝑧𝑧 114.83 m4 relating the wave elevations with forces on the pontoons, added mass
Second polar moment of area, 𝐽 6.88 m4 and added damping for all pontoons were established by using DNV
Young’s modulus, 𝐸 210 GPa
HydroD WADAM. The method for combination of the multiple sub-
Shear modulus, 𝐺 80 GPa
structures are described in Kvåle et al. (2016), and the reader is referred
to that paper for specifics. By using the mode shapes obtained from
the FE model, excluding the hydrodynamic added mass and added
The current state of the development stages can be found in the reports damping, as a coordinate basis, the results from the hydrodynamic
made by the consultant companies involved in the project (AMC, 2019; analyses can easily be included to form a complete model. The most
Dr.techn. Olav Olsen and Norconsult, 2019). relevant properties of the pontoons and the connected columns are
Through several development stages, the NPRA and the consultant given in Table 2. It is noted that the hydrostatic stiffness component
companies involved have developed several bridge concepts. Herein, 𝐾𝜃𝑥 ,𝜃𝑥 is negative for pontoons 11–46. This does not imply that the
only an end-supported floating bridge concept is considered (see Fig. 3). pontoon itself is unstable. However, when combined with the relevant
Although there exist several variants of this type, in essence, the di- superstructure, the centre of gravity shifts upwards due to the mass of
mensions and design of all the end-supported bridge concepts designed the girder and column; which in turn causes the contribution to the
during the development stages are close to the depiction in Fig. 4. As stiffness about the pontoon local x-axes (see Fig. 4) to be negative.
the figure shows, the considered bridge concept consists of a cable- This corresponds to rotation about the long axis of the pontoon. The
stayed high bridge on the south end, which gradually descends into hydrodynamic added mass and damping obtained for the pontoon type
a low floating bridge with a constant elevation between pontoons 11 used for the major part of the bridge, i.e., ranging pontoons 11–46, are
and 46. The cross section of the girder assumed for the low bridge is illustrated in Figs. 7 and 8. The wet mode shapes of the first eight modes
depicted in Fig. 5. The pontoon and column geometry corresponding to of the bridge are depicted in Fig. 9, and the corresponding damped
the same segment are shown in Fig. 6. natural periods indicated in the figure caption. As the hydrodynamic
added mass and damping are frequency dependent, the shown mode
3.1. Model setup shapes are obtained from an iterative procedure and are strictly only
able to represent the system at resonance. It is referred to Kvåle et al.
In this paper, we use an Abaqus FE model developed by the NPRA. (2016) for more details on the applied procedure.
The model is relatively detailed and consists of beam elements to repre- Reported section forces in the bridge girder are following the right-
sent the bridge girder, the tower, and the columns attaching the bridge hand rule convention and the local axis system depicted in Fig. 4;

5
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

Table 2
Properties of pontoons and columns. The counting starts on the left in Fig. 4, near the high bridge, and increases towards right. The columns are of the same steel
quality as the girder, and posses identical material properties. Note that the pontoon properties refer to the local coordinate system of the pontoons, depicted in
Fig. 4, which differs from the local coordinate system of the girder.
Pontoon 1 Pontoons 2–4 Pontoons 5–10 Pontoons 11–46
Area [m2 ] 872.940 769.807 664.969 558.414
Pontoon dimensions Length × width 58 m × 16 m 58 m × 14 m 58 m × 12 m 58 m × 10 m
Height 9 m 9 m 9 m 9 m
𝐾𝑧𝑧 [N/m] 8.78 ⋅ 106 7.74 ⋅ 106 6.68 ⋅ 106 5.61 ⋅ 106
𝐾𝜃𝑥 ,𝜃𝑥 [Nm/rad] 8.27 ⋅ 107 3.79 ⋅ 107 3.22 ⋅ 106 −1.66 ⋅ 107
Pontoon properties
𝐾𝜃𝑦 ,𝜃𝑦 [Nm/rad] 2.11 ⋅ 109 1.89 ⋅ 109 1.65 ⋅ 109 1.40 ⋅ 109
Mass, 𝑀𝑧 , 𝑀𝑦 , 𝑀𝑥 [kg] 1970.0 ⋅ 103 1441.9 ⋅ 103 1273.9 ⋅ 103 958.9 ⋅ 103
𝐴 [m2 ] 1.75 1.15 1.15 0.67
Column properties 𝐼𝑦𝑦 and 𝐼𝑧𝑧 [m4 ] 27.22 12.014 12.014 4.27
𝐽 [m4 ] 54.436 24.03 24.03 8.54

Fig. 6. Technical drawing of the pontoon type and column corresponding to pontoons 11–46. Units are in mm.
Source: Reproduced from (Norconsult AS, 2017).

Fig. 7. Added mass (pontoons 11–46) due to radiation forces, established using the
potential theory solver DNV HydroD WADAM. (a) Translational terms. (b) Rotational Fig. 8. Wave radiation damping (pontoons 11–46), established using the potential
terms. theory solver DNV HydroD WADAM. (a) Translational terms. (b) Rotational terms.

6
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

Fig. 9. Horizontal projection of the mode shapes of the combined system, including hydrodynamic radiation forces. (a) Mode 1 (𝑇𝑑 = 117.0 s). (b) Mode 2 (𝑇𝑑 = 61.9 s). (c) Mode
3 (𝑇𝑑 = 34.7 s). (d) Mode 4 (𝑇𝑑 = 24.1 s). (e) Mode 5 (𝑇𝑑 = 16.6 s). (f) Mode 6 (𝑇𝑑 = 14.2 s). (g) Mode 7 (𝑇𝑑 = 11.8 s). (h) Mode 8 (𝑇𝑑 = 9.9 s).

Fig. 11. Directional distributions of the considered 100-year wind sea and swell sea
Fig. 10. JONSWAP spectra describing the one-dimensional wave spectral density of state.
the considered 100-year wind sea and swell sea state.

sea states are thereafter constructed by defining a selected subset of


𝐹𝑥 is the axial force, 𝑀𝑥 is the torsional moment, 𝑀𝑦 is the bending the homogeneous set of parameters as functions of the location. The
moment about the axis with the lowest resistance (weak-axis bending JONSWAP spectrum is used to model the wave spectral density, and
moment), and 𝑀𝑧 is the bending moment about the axis with the reads out (Hasselmann et al., 1973; Stansberg et al., 2002):
highest resistance (strong-axis bending moment). ( )
( ( )−4 ) (𝑓 −𝑓𝑝 )2
exp −
5 𝑓 2𝜎 2 𝑓𝑝2
3.2. Homogeneous base states 𝑆(𝑓 ) = 𝛼𝑔 2 (2𝜋)−4 𝑓 −5 exp − 𝛾 (24)
4 𝑓𝑝

The homogeneous conditions for wind sea and swell, provided in Here, 𝑓 is the circular frequency, 𝑇𝑝 = 1∕𝑓𝑝 is the peak period, 𝛾 is
the design basis of stage 5 of the development project run by the NPRA, the peak enhancement factor, 𝛼 is the Philips parameter, and 𝜎 = 0.07
are used as a basis for wave conditions considered. The inhomogeneous for 𝑓 < 𝑓𝑝 and 𝜎 = 0.09 for 𝑓 ≥ 𝑓𝑝 . The energy of the waves is

7
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

Fig. 12. Wave elevation simulation snapshot of homogeneous (reference) wind sea state. The black dots indicate pontoon locations, and thus points of excitation, and the red dots
indicate support points.

2. Inhomogeneous 100-year, varying 𝐻𝑠 (Section 3.3.2)

The suggested inhomogeneity case is not considered to necessarily


represent realistic sea states at site, but still, it has some conceptual
interest.

3.3.1. Homogeneous conditions


The considered homogeneous wind sea conditions with a return
period of 100 years, described in Figs. 10 and 11, were used to simulate
the wave elevation over the area of the bridge, according to Eq. (9).
A snapshot of the simulated wave elevation is shown in Fig. 12. The
Fig. 13. Correlation to wave elevation at central pontoon, corresponding to reference wave elevation correlation field, describing the correlation to the wave
wind sea state represented in Fig. 12. The black dots indicate pontoon locations, and
thus points of excitation.
elevation at the central pontoon of the bridge, is depicted in Fig. 13.
The figure reveals a very limited correlated wave elevation region. The
low correlation hints about a highly limited, and intuitive, effect of
inhomogeneity; only a direct reduction of excitation would significantly
reduce the response as the coherence between neighbouring pontoons
is very low. Any potential effects that covariance would have, due to
global vibration modes of the bridge, would not manifest clearly as the
excitation on different pontoons would be basically independent.

3.3.2. Linearly varying wave height


To conceptually mimic the boundary effects in the fjord, we intro-
duced a linear variation of the wave height of the wave field, as seen
Fig. 14. Functional variation of 𝐻𝑠 used for inhomogeneous wind sea. Note that 𝐻𝑠 in Fig. 14. Note that the value of 𝐻𝑠 at the midspan of the bridge is
is not zero at the ends.
identical to the homogeneous value of 𝐻𝑠 applied above. The values of
𝐻𝑠 at the two edges are set to half of the midspan value. The introduced
Table 3
Sea state parameters for chosen 100-year conditions.
inhomogeneity is intended to conceptually model the likely reduced
wave heights near the shore due to boundary effects in the fjord. This
JONSWAP parameters Directional distribution
will give insight into the behaviour of the bridge when exposed to such
𝐻𝑠 [m] 𝑇𝑝 [s] 𝛾 𝜃0 [◦ ] 𝑠
conditions and help to illustrate that the suggested framework provides
Swell 0.34 16.0 4 −90 31
reasonable results. In reality, the peak period of the waves would be
Wind sea 2.1 5.5 2.05 −90 12
affected as well. However, this is deliberately chosen to be disregarded
as the results would become much more chaotic and harder to interpret.
The water elevation was simulated according to Eq. (9), based on the
distributed over multiple wave angles, which can be described by the varying 𝐻𝑠 as well as the other sea state parameters described in the
cos2𝑠 directional distribution, written as (Longuet-Higgins et al., 1963): preceding sub-section. The resulting wave elevation at an arbitrary time
( ) instance is depicted in Fig. 15. By studying the figure it is clear that the
𝜃 − 𝜃0 2𝑠
𝐷(𝜃) = 𝐶 cos (25) amplitudes of the waves are lower near the two ends of the bridge, as
2
given by the variation of 𝐻𝑠 .
2𝜋 Figs. 16 and 17 show standard deviations of selected components
where 𝐶 is a factor introduced to ensure that ∫0 𝐷(𝜃)𝑑𝜃 = 1, s is
the spreading parameter, and 𝜃0 is the mean wave direction. The one- of displacement and section forces, respectively, due to the described
dimensional wave spectral densities and directional distributions are inhomogeneous sea state. The corresponding response components ob-
given in Figs. 10 and 11, respectively, for both wind sea and swell tained from the homogeneous base sea state are also given in the
conditions with 100 years return period. The corresponding sea state figures. The reduced wave energy near the ends of the bridge does
parameters are summarized in Table 3. not increase the response of any components at any location, and
the homogeneous condition is conservative. Reduced 𝐻𝑠 towards the
3.3. Wind-sea-induced response ends of the bridge scales the response of the vertical components
correspondingly; due to the limited stiffness contribution provided by
The following wind sea states are considered: the girder compared to the pontoons, the girder response is dominated
by local effects. Components with lateral components, however, are
1. Homogeneous 100-year (Section 3.3.1) dominated by the stiffness contribution from the girder, which manifest

8
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

Fig. 15. Wave elevation simulation snapshot of inhomogeneous wind sea state with varying 𝐻𝑠 . The black dots indicate pontoon locations, and thus points of excitation, and the
red dots indicate support points.

in a response dominated by the girder behaviour, without any clear


local effects.

3.4. Swell-induced response

The following swell sea states are considered:

1. Homogeneous 100-year (Section 3.4.1)


2. Inhomogeneous 100-year, linearly varying 𝐻𝑠 (Section 3.4.2)
3. Inhomogeneous 100-year, spherical wave front matching bridge
curvature (Section 3.4.3)
4. Adjusted 100-year with period matching a vibration mode,
where 𝐻𝑠 is varied to mimic a shadowing effect reducing po-
tential cancellation of modal excitation (Section 3.4.4)

The suggested inhomogeneities are not considered to necessarily


represent realistic sea states at site, but are rather cases of conceptual
interest.

Fig. 16. Effects on standard deviations of selected displacement response components 3.4.1. Homogeneous conditions
along the bridge girder due to the inhomogeneous wind sea state with a linearly varying
Fig. 18 shows a snapshot of the simulated wave elevation due
𝐻𝑠 .
to the homogeneous swell wave conditions, based on Eq. (9). The
corresponding correlation between wave elevations in the region en-
compassing the bridge and the wave elevation at the central pontoon
is shown in Fig. 19. In contrast to the extent of the correlated regions
observed for the wind sea state, the correlated region due to this swell
condition is relatively large; the excitation of 5 pontoons (4 spans) can
be considered to be, to some degree, correlated.

3.4.2. Linearly varying wave height


Realistically, swell waves would likely cause boundary effects close
to the sides of the fjord. As for the wind sea in Section 3.3.2, we mod-
elled this effect very simplistic by a bilinear variation of the significant
wave height along the strait. 𝐻𝑠 at the two edges are set to half of that
at the midspan, which is equal to the homogeneous 𝐻𝑠 . The functional
value of 𝐻𝑠 is presented in Fig. 20. By applying Eq. (9), these conditions
result in a water elevation at an arbitrary time instance as depicted
in Fig. 21. The snapshot of the water elevation clearly represents the
requested variation of wave amplitude along the strait.
The displacement and section force response due to the described
swell sea state are presented in Figs. 22 and 23, respectively. The effect
of the inhomogeneity is similar to that seen for the wind sea state;
vertical response components show local effects due to the scaling along
the bridge, whereas lateral response components get a global reduction
because the girder stiffness controls the lateral behaviour.

3.4.3. Spherical wave front


A fundamental assumption in linear wave theory is the fact that
waves are plane. In coastal waters, it is not unlikely that the wave
front is, in fact, spherical. To mimic this, the mean wave angle was
modulated across the strait. To maximize the effect on the response, we
Fig. 17. Effects on standard deviations of selected section forces and moments along assumed that the curvature of the wave front matched the curvature of
the bridge girder due to the inhomogeneous wind sea state with a linearly varying 𝐻𝑠 . the bridge girder. It is noted that this has to be considered an extreme

9
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

Fig. 18. Wave elevation simulation snapshot of homogeneous (reference) swell sea state. The black dots indicate pontoon locations, and thus points of excitation, and the red
dots indicate support points.

Fig. 19. Correlation to wave elevation at central pontoon, corresponding to reference swell sea state represented in Fig. 18. The black dots indicate pontoon locations, and thus
points of excitation, and the red dots indicate support points.

are significantly increased by the inhomogeneity: an increase of 14% of


the largest standard deviation is observed. The figures also depict plots
of the response disregarding the CPSD of wave elevation between the
pontoons, implying they are considered independent processes; this is
denoted as 𝛾 = 0 in the figure legends as it can be achieved by defining
the coherency function as zero for all combinations where 𝑝 ≠ 𝑞 (equal
to 1 for 𝑝 = 𝑞). The homogeneous and inhomogeneous sea states result
in almost identical response when disregarding the coherency. Thus,
these plots indicate that it is the coherent excitation of the modes of
Fig. 20. Functional variation of 𝐻𝑠 used for inhomogeneous swell. Note that 𝐻𝑠 is not
the bridge rather than the changed projection areas on pontoons that
zero at the ends. cause the increased response. It is also worth noting that the inclusion
and proper treatment of the coherency is crucial when simulating swell
conditions. This is in contrast to what is the case for wind sea excitation,
situation. However, we consider it to be relevant as a conceptual case. which in many cases can fairly accurately be treated without account-
The mean wave angle is defined as equal to the angle perpendicular ing for coherency (Kvåle et al., 2016). However, if the true correlation
to the bridge girder along the span length. When the 𝑥-axis is defined at site is lower than what is assumed in linear wave theory, the figure
tangential to the local bridge axis and equal to 0 at the middle of the indicates that the weak-axis moments (𝑀𝑦 ) might be underpredicted.
bridge, the following equation describe the normal angle of the arc This is again due to the fact that the pontoon stiffness is dominating
representing the bridge: the response in weak-axis bending, causing a proneness to local load
( ) effects. Also, the figures reveal that the weak-axis bending moment is
𝑥 − 𝐿∕2 larger for the homogeneous sea state than the inhomogeneous sea state,
𝜃0 (𝑥) = − cos−1 (26)
𝑅 some distance from the midspan.
where 𝐿 = 5239 m and 𝑅 = 5000 m. This functional relationship is
shown in Fig. 24. An arbitrary time instance of the water elevation 3.4.4. Shadowing effect
simulation corresponding to this inhomogeneous sea state is shown in For narrow-banded excitation around the frequencies of asymmetric
Fig. 25. modes, some degree of cancellation could be expected for long wave
In Figs. 26 and 27, chosen components of the displacement response crests. Thus, in theory, sea states with local regions with a lower exci-
and section forces due to the presented sea state are depicted in terms tation amplitude than the homogeneous conditions could cause harsher
of standard deviations along the bridge girder. The figures also show effective excitation than fully homogeneous sea states. Such a situation
the response due to the homogeneous swell condition, for reference. It could occur for instance if vessels or local topography (e.g., islands
is clear that the concavity of the wave front affects the response sig- or promontories) is obscuring the wave propagation, i.e., is causing a
nificantly. Specifically, the components related to the lateral response shadowing effect.
of the bridge, i.e., strong-axis moments and lateral displacements, are To assess this effect, the peak period of the 100-year swell sea
increased by the introduction of the inhomogeneous conditions. Equally condition was increased from 16.0 s to 16.4 s, to match the damped
important, especially in the context of parametric excitation induced by natural period of the asymmetric mode 5. Thereafter, by considering
axial force variation (Kvåle et al., 2019), is the fact that the axial forces the mode shape, we constructed an inhomogeneous sea state with

10
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

Fig. 21. Wave elevation simulation snapshot of inhomogeneous swell sea state, with linearly varying 𝐻𝑠 . The black dots indicate pontoon locations, and thus points of excitation,
and the red dots indicate support points.

Fig. 24. Functional value of 𝜃0 based on Eq. (26).

varying SWH, with the goal to avoid the potential cancellation effects.
Fig. 28 show the lateral component of the mode shape of mode 5 and
the suggested function of SWH, 𝐻𝑠 (𝑥).
The choice of the function is based on setting 𝐻𝑠 to 0 for all
mode shape displacements below 75% of the maximum mode shape
displacement, but only for peaks on the side of 0 with the fewest peaks
Fig. 22. Effects on standard deviations of selected displacement response components (< 0 in Fig. 28). The resulting simulated water surface elevation, at an
along the bridge girder due to the inhomogeneous swell sea state with a linearly varying arbitrary time instance, is depicted in Fig. 29.
𝐻𝑠 .
The displacement response and section forces in the girder due to
the given inhomogeneous sea state is compared with the corresponding
homogeneous sea state, with constant values of 𝐻𝑠 , in Fig. 30 and
Fig. 31, respectively. In general, the locally lowered 𝐻𝑠 results in a
lower response. Some local effects are present in components where
the stiffness of the girder is not dominating; for instance, the weak-
axis bending moment 𝑀𝑦 which is related to response mainly driven
by the pontoons and not the girder. For other components, such as the
strong-axis bending moment 𝑀𝑧 , the girder stiffness is the dominating
contribution and as a consequence the lower-order mode shapes of the
girder will ensure a correlated response across the girder, with few local
effects.
To explain why the mode-shape-based functional description of 𝐻𝑠
does not increase the global response, the correlation plot of the wave
elevation of the homogeneous swell sea condition shown in Fig. 19
is useful. A very similar correlation plot is resulting from the sea
state under investigation, as the peak period is only slightly shifted.
As seen in the figure, the correlation length of the wave elevation
with the chosen conditions is in the order of the length of 4 spans
(encompassing 5 pontoons). For any cancellation effects to manifest, it
is reasonable that a significant amount of both positive and negative
mode contributions requires correlated excitation. This is therefore
only likely to occur for higher-order modes or a very longcrested sea
state, which makes this type of problem irrelevant for these types of
structures undergoing realistic excitation scenarios.

3.5. Verification of time-domain simulation procedure

This sub-section aims to exemplify and verify the time-domain


procedures given in Section 2.2.1. We used Eq. (21) to simulate the
Fig. 23. Effects on standard deviations of selected section forces and moments along time histories of the wave excitation due to the swell sea state given in
the bridge girder due to the inhomogeneous swell sea state with a linearly varying 𝐻𝑠 . Section 3.4.3. Figs. 32 and 33 show the CPSDs representing the lateral

11
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

Fig. 25. Wave elevation simulation snapshot of swell sea state with spherical wave front matching curvature of the bridge girder. The black dots indicate pontoon locations, and
thus points of excitation, and the red dots indicate support points.

Fig. 28. Functional value of 𝐻𝑠 applied in an attempt to reduce cancellation effects.


The markers indicate the lateral displacements of the considered mode 5 at the pontoon
locations.

Fig. 26. Effects on standard deviations of selected displacement response components


and vertical excitations, respectively, acting on pontoons 1 and 4. The
along the bridge girder due to the inhomogeneous swell sea state with a spherical wave spectral density estimates from the simulated excitations are also shown
front (varying 𝜃0 ). in the figures; these are based on the ensemble averages of the FFT-
based spectral densities obtained from 200 random realizations. Despite
a large distance, and a resulting low coherence, between the two pon-
toons, the simulated excitation agrees well with the reference spectral
densities. By applying a more sophisticated method for the spectral
density estimation, e.g., Welch’s method, noise would be reduced and
the agreement improved.
The resulting excitation time histories for each of the 200 real-
izations are used in the following multi-step procedure to predict the
time-domain response:

1. Generate realization 𝑛 of excitation spectral density in Eq. (17),


{𝑝(𝑡)}𝑛
2. Fourier transform time histories to frequency domain by FFT,
{𝑃 (𝜔)}𝑛 = F{𝑝(𝑡)}𝑛
3. Conduct frequency-domain response prediction through {𝑅(𝜔)}𝑛 =
[𝐻(𝜔)]{𝑃 (𝜔)}𝑛
4. Inverse Fourier transform frequency-domain response to time do-
main, {𝑟(𝑡)}𝑛 = F−1 {𝑅(𝜔)}𝑛

The ensemble average of the FFT-based cross-spectral density esti-


mates between the lateral response on pontoon 11 and 23 is shown in
Fig. 34. The plot shows a very good agreement, which indicates that
the time-domain representation is valid.

4. Conclusions

Herein, we presented a methodology consistent with linear wave


theory, to treat the inhomogeneous wave excitation on floating struc-
tures, including a proper treatment of the wave coherency. The method-
ology is fundamentally based on a frequency-domain representation of
Fig. 27. Effects on standard deviations of selected section forces and moments along
the random wave field, but the required steps for simulation in time
the bridge girder due to the inhomogeneous swell sea state with a spherical wave front
(varying 𝜃0 ). domain are given.

12
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

Fig. 29. Wave elevation simulation snapshot of swell sea state with defined regions with 𝐻𝑠 = 0. The black dots indicate pontoon locations, and thus points of excitation, and the
red dots indicate support points.

Fig. 30. Effects on standard deviations of selected displacement response components


along the bridge girder due to the inhomogeneous swell sea condition with peak period
corresponding to mode 5, with homogeneous and inhomogeneous 𝐻𝑠 as depicted in
Fig. 28.

Furthermore, selected conceptual inhomogeneous wind sea and


swell sea conditions were compared with the homogeneous base states
with 100-year return period. The following findings are of particular
interest and are considered to have transferable value:

• Coherency (and thus cross-spectral density) between wave eleva- Fig. 31. Effects on standard deviations of selected section forces and moments along
tion at pontoon locations must be properly accounted when sim- the bridge girder due to the inhomogeneous swell sea condition with peak period
ulating swell excitation scenarios; an increase of the strong-axis corresponding to mode 5, with homogeneous and inhomogeneous 𝐻𝑠 as depicted in
Fig. 28.
bending moment of approximately 100% is observed compared to
fully non-correlated wave excitation for the chosen 100-year swell
sea state. Coherency can likely be neglected when simulating
wind sea for structures with similar pontoon distances and wave
conditions as the case study.
• Sea states with inhomogeneous significant wave height can likely
be approximated by assuming the largest significant wave height
for the full wave field, conservatively. Very narrow-banded exci-
tation or higher-order modes could violate this assumption, but
these special cases are considered unlikely.
• If the coastal topography generates a spherical wave front across
a strait, an inhomogeneous wave condition using a varying mean
wave angle could be used to model this behaviour. For the cross-
ing of typical fjords, with steep and pronounced topography, this
is considered crucial. For the studied bridge, exposed to a swell
sea condition with a return period of 100 years, a 14% increase
of the largest axial force is observed due to this effect.
Fig. 32. Cross power spectral density between lateral excitation on pontoons 1 and
4, from direct solution and time-domain simulation. The CPSD estimated from the
Certain response effects of inhomogeneous wave conditions are excitations simulated in time domain is based on the ensemble average of 200
significant and need to be included in the assessment of future floating simulations.

13
K.A. Kvåle et al. Applied Ocean Research 142 (2024) 103802

Data availability

No data was used for the research described in the article.

Acknowledgements

This research was conducted with financial support from the Nor-
wegian Public Roads Administration, Norway. The authors gratefully
acknowledge this support.

References

AMC, 2019. SBJ-33-C5-AMC-90-RE-100 - Preferred Solution, K12 - Main Report. tech.


rep., Statens vegvesen, Oslo, p. 181.
Fig. 33. Cross power spectral density (CPSD) between vertical excitation on pontoons
Cheng, Z., Gao, Z., Moan, T., 2018. Wave load effect analysis of a floating bridge in a
1 and 4, from direct solution and time-domain simulation. The CPSD estimated from
fjord considering inhomogeneous wave conditions. Eng. Struct. 163, 197–214.
the excitations simulated in time domain is based on the ensemble average of 200
Cheng, Z., Svangstu, E., Gao, Z., Moan, T., 2019. Field measurements of inhomogeneous
simulations.
wave conditions in Bjørnafjorden. J. Waterw. Port Coast. Ocean Eng. 145 (1),
5018008.
Cui, M., Cheng, Z., Moan, T., 2022. A generic method for assessment of inhomogeneous
wave load effects of very long floating bridges. Mar. Struct. 83, 103186.
Dai, J., Leira, B.J., Moan, T., Kvittem, M.I., 2020. Inhomogeneous wave load effects
on a long, straight and side-anchored floating pontoon bridge. Mar. Struct. 72,
102763.
Dai, J., Stefanakos, C., Leira, B.J., Alsos, H.S., 2021. Effect of modelling inhomogeneous
wave conditions on structural responses of a very long floating bridge. J. Mar. Sci.
Eng. 9 (5), 548.
Dr.techn. Olav Olsen, Norconsult, 2019. SBJ-33-C5-OON-22-RE-100 - K12 – Summary
report. tech. rep., Statens vegvesen, Oslo, p. 247.
George, D., 1996. Simulation of ergodic multivariate stochastic processes. J. Eng. Mech.
122 (8), 778–787.
Hasselmann, K.F., Barnett, T.P., Bouws, E., Carlson, H., Cartwright, D.E., Eake, K.,
Euring, J.A., Gicnapp, A., Hasselmann, D.E., Kruseman, P., Enke, K., Ewing, J.A.,
Gienapp, H., Hasselmann, D.E., Kruseman, P., 1973. Measurements of Wind-Wave
Growth and Swell Decay During the Joint North Sea Wave Project (JONSWAP).
tech. rep..
Jefferys, E.R., 1987. Directional seas should be ergodic. Appl. Ocean Res. 9 (4),
186–191.
Kinsman, B., 1965. Wind Waves: Their Generation and Propagation on the Ocean
Surface. Courier Dover Publications.
Kvåle, K.A., 2017. Dynamic Behaviour of Floating Bridges Exposed to Wave Excitation.
A Numerical and Experimental Investigation (Ph.D. thesis). NTNU.
Kvåle, K.A., Heiervang, M.F., Giske, F.-I.G., 2019. SBJ-33-C5-AMC-90-RE-119 - Preferred
Solution, K12 – Appendix S - Parametric Excitation. tech. rep., p. 214.
Kvåle, K.A., Sigbjörnsson, R., Øiseth, O., 2016. Modelling the stochastic dynamic
Fig. 34. Cross power spectral density of the lateral response on pontoons 11 and behaviour of a pontoon bridge: A case study. Comput. Struct. 165, 123–135.
23. Reference refers to the frequency-domain representation. The CPSD estimated from Langen, I., Sigbjörnsson, R., 1980. On stochastic dynamics of floating bridges. Eng.
the excitations simulated in time domain is based on the ensemble average of 200 Struct. 2 (4), 209–216.
simulations. Longuet-Higgins, M.S., Cartwright, D.E., Smith, N.D., 1963. Observations of the direc-
tional spectrum of sea waves using the motions of a floating buoy. In: Proc. Conf.
Ocean Wave Spectra. Prentice-Hall, pp. 111–132.
Naess, A., Moan, T., 2012. Stochastic Dynamics of Marine Structures. Cambridge
bridges. This is particularly important for swell sea conditions. Other University Press, New York.
types of inhomogeneous wave conditions can likely be conservatively Norconsult AS, 2017. K7 Bjørnafjorden End-Anchored Floating Bridge. Appendix A –
approximated as homogeneous through intuitive assumptions. Drawings binder. tech. rep., p. 54.
Shinozuka, M., 1972. Monte Carlo Solution of Structural Dynamics. Columbia Univ,
New York.
CRediT authorship contribution statement Shinozuka, M., Wai, P., 1979. Digital simulation of short-crested sea surface elevations.
J. Ship Res. 23 (01), 76–84.
Knut Andreas Kvåle: Conceptualization, Methodology, Software, Sigbjörnsson, R., 1979. Stochastic theory of wave loading processes. Eng. Struct. 1 (2),
Writing – original draft, Writing – review & editing, Visualization, 58–64.
Stansberg, C.T., Contento, G., Won Hong, S., Irani, M., Ishida, S., Mercier, R.,
Investigation. Bernt Leira: Investigation, Writing – review & editing.
Wang, Y., Wolfram, J., 2002. The specialist committee on waves: Final report and
Ole Øiseth: Conceptualization, Investigation, Project administration, recommendations to the 23rd ITTC. In: Proceedings of the 23rd ITTC. pp. 505–736.
Funding acquisition, Writing – review & editing, Supervision. Watanabe, E., 2003. Floating bridges: past and present. Struct. Eng. Int. 13 (2),
128–132.
Declaration of competing interest Wei, W., Fu, S., Moan, T., Lu, Z., Deng, S., 2017. A discrete-modules-based frequency
domain hydroelasticity method for floating structures in inhomogeneous sea
conditions. J. Fluids Struct. 74, 321–339.
The authors declare that they have no known competing finan- Wei, W., Fu, S., Moan, T., Song, C., Ren, T., 2018. A time-domain method for
cial interests or personal relationships that could have appeared to hydroelasticity of very large floating structures in inhomogeneous sea condition.
influence the work reported in this paper. Mar. Struct. 57, 180–192.

14

You might also like