Sunteți pe pagina 1din 8

19th IABSE Congress Stockholm, 21-23 September 2016

Challenges in Design and Construction of an Innovative and Sustainable Built Environment

Dynamic Response of a Floating Bridge Structure

Thomas Hansen Viuff, Bernt Johan Leira, Ole Øiseth
Norwegian University of Science and Technology, Trondheim, Norway

Xu Xiang
Norwegian Public Roads Administration, Stavanger, Norway


A theoretical overview of the stochastic dynamic analysis of a floating bridge structure is presented.
Emphasis is on the wave-induced response and the waves on the sea surface are idealized as a zero
mean stationary Gaussian process. The first-order wave load processes are derived using linear
potential theory and the structural idealization is based on the Finite Element Method. A frequency
response calculation is presented for a simplified floating bridge structure example emphasising the
influence on von Mises stress in the pontoon from low- and high frequency waves and frequency
dependence in hydrodynamic added mass and damping coefficients.
Keywords: Floating bridge; frequency response; linear dynamics; von Mises stress.

bridge and on the sea state conditions different

1 Introduction types of floating bridges are used. Only three long
Floating bridges have been around for many span floating bridges are currently located in
thousands of years and throughout the years, they difficult sea state conditions and allows for cars to
have been used as temporary supply lines or for pass. These are:
military purposes. However, it is only during the i. Hood Canal Bridge (1961) in USA a 2,398
last three decades or so that floating bridges are meter long pontoon bridge with a 1,988
being developed to the degree of sophistication, so meter long anchored floating portion, it is
they can be applied as a critical part of modern the longest floating bridge in the world
infrastructure. Still, compared with land-based located in a saltwater tidal basin, and the
bridges, including cable-stayed bridges, limited third longest floating bridge overall.
information [1] is currently available on floating ii. Bergsøysund Bridge (1992) in Norway a
bridges and even less on submerged floating 931 meter long pontoon bridge with the
tunnels for transportation. This information is longest span of 106 meters.
especially true regarding construction records, iii. Nordhordland Bridge (1994) in Norway is a
environmental conditions, durability, operations combination of a cable-stayed and
and performance of the structure. pontoon bridge. It is the longest free
The limited amount of floating bridges currently in floating bridge without anchorage.
the world is a statement to this fact. Depending on
the landscape in the proximity of the floating

19th IABSE Congress Stockholm 2016
Challenges in Design and Construction of an Innovative and Sustainable Built Environment

As the rough overview indicates, the theoretical 2 System Modelling

and practical development of floating bridges has
been carried out mainly in the USA and in Norway The linear stochastic dynamic response of a floating
with significant contributions from the industry. In bridge structure can be described using the
Norway it is mainly the Norwegian University of equation of motion to capture the dynamics of the
Science and Technology (NTNU), SINTEF the structure, potential theory to find the
research organisation and the Norwegian Public hydrodynamic added mass and damping and the
Roads Administration (NPRA). wave excitation force from the fluid-structure
interaction and stochastic theory to implement the
Pioneering studies on floating bridges was carried randomness of the wave excitation force.
by Hartz in the 1970’s. Around the same time
Holan, Sigbjörnsson and Langen carried out similar 2.1 Equation of Motion
studies on stochastic dynamics of floating bridges
[2] [3] [4] [5]. Later on in 1980 Sigbjörnsson and The equation of motion describing the linear
Langen exemplified the theory using a model of the dynamic behaviour of the floating bridge is
Salhus floating bridge [6] [7]. In recent years described in time domain as shown in (1) .
NTNU/SINTEF have led the theoretical evolution
[ M s ]{ u ( t )}  [ C s ]{ u ( t )}  [ K s ]{ u ( t )}  { q h ( t )} (1)
within structural mechanics, fluid structure
interaction and stochastic modelling of
environmental loads applied to the offshore Here, [ M ] , [ C ] and [ K ] are the frequency
s s s

industry in Norway. Many of the same theories can independent structural mass-, damping- and
be directly applied in stochastic dynamic analysis of stiffness matrices. The vector notation { u } is the
floating bridges. structural response and the dots above represents
derivatives of time t . The vector { q ( t )} represents
Recently the NPRA has started several research h

projects regarding floating bridge structures as part the hydrostatic and hydrodynamic load vector.
of a ferry-free coastal route E39 between
2.1.1 Frequency Domain Representation
Kristiansand and Trondheim in Norway, where they
aim to develop current methods of design. For a single harmonic small amplitude wave,
{ q ( t )} can be described as a harmonic wave
In the present text a dynamic analysis in frequency h

i t
domain will be given and theory on stochastic proportional to e as shown in (2).
dynamic modelling of a floating bridge is described, As an extra step in the equation, the derivatives of
including challenges regarding frequency- the structural response are derived and collected
dependent hydrodynamic added mass and within the parenthesis.
damping. Preliminary results will be given from a
frequency domain analysis of the stresses on the { q h ( t )}      [ M h (  )]  i  [ C h (  )]  [ K h ] 

pontoon. (2)
i t i t
{ Z u (  )} e  { Z q (  )} e
Although a lot of research has gone in to the topic
of floating bridges, the focus point has mostly been
Here, [ M h (  )] and [ C h (  )] are the frequency
on the structural response in terms of
displacement, velocity and acceleration of dependent hydrodynamic added mass and
structural points and as far as the author is aware, damping and  is the angular frequency. [ K h ] is
not much literature on local stress distributions for the restoring stiffness assumed frequency
floating pontoon bridges is published. This paper independent for small amplitude motion. { Z u (  )}
aims to shed some light on general stochastic and { Z q (  ) } are the complex structural response
design as well as local stress distribution. amplitude and the complex wave excitation force
amplitude, respectively, and i is the imaginary
unit. Substituting the expression for the
hydrodynamic action given in (2) into the equation

19th IABSE Congress Stockholm 2016
Challenges in Design and Construction of an Innovative and Sustainable Built Environment

of motion in (1) and rearranging the terms gives the Here,  is time lag and [ m ] and [ c ] are the time h h

frequency domain representation of the equation domain representations of the hydrodynamic

of motion. added mass and damping found from Fourier
{ Z q (  )}     [ M (  )]  i [ C (  )]  [ K ] 
 
(3) 1 
i t
{ Z u (  )} [ M h (  )] e d
[ m h (t)] 
  (10)
The inertia, damping and restoring matrices
1 
i t
include the structural terms as well as the added [ c h ( t )]   [ C h (  )] e d (11)
hydrodynamic mass and damping. The combined 2 

system matrices are hence given as.

Using the impulse response function, h (  ) , the
[ M (  )]  [ M s ]  [ M h (  )] (4) response can be obtained in time domain as a finite
sum of system responses from hydrodynamic
[ C (  )]  [ C s ]  [ C h (  )] (5) action impulses at different time steps.

[K ]  [K s ]  [K h ] (6) { u ( t )}   [ h ( t   )]{ q h ( )}d  (12)

The response induced by a single harmonic wave is The impulse response function is found from
then obtained by rearranging the terms in (3) and Fourier transform of the frequency transfer
introducing the frequency transfer function function in (8).
[ H (  )] .
1 
i t
[ h ( t )]   [ H (  )] e d (13)
{ Z u (  )}  [ H (  )]{ Z q (  )} (7) 2 

1 Several methods exist to solve (9) in time domain.

[ H (  )]     [ M (  )]  i  [ C (  )]  [ K ] 
 
Such approaches are useful if non-linear behaviour
is of interest.
By use of the principle of superposition, it is
possible within the framework of linear theory to 2.2 Description of Sea Waves
incorporate a generalized description of the
excitation represented as the sum of a finite For engineering purpose, the wind-generated
number of harmonic waves. In case of a random waves are approximated as a locally stationary and
sea state the excitation in frequency domain can be homogeneous random field and the sea surface
obtain by Fourier transform of the excitation time elevation  ({ x } , t ) becomes a function of time and
series. the two-dimensional space vector for the
horizontal surface at the mean water level.
2.1.2 Time Domain Representation

i ({  } { x }   t )
Assuming frequency independent restoring and  ({ x } , t )   e d Z  ({  } ,  ) (14)

causality the wave excitation force can be

described in the time domain as shown in (9) by use Here, Z  ( { } ,  ) is the spectral process of the sea
of the convolution integral.
surface elevation and { }  { x ,  y } is the two-
 dimensional wave number vector.
{ q h ( t )}  { q ( t )}   [ m h ( t   )]{ u ( t )} d 

(9) The spectral process is, given the assumptions of
 [ c h ( t   )]{ u ( t )} d   [ K h ]{ u ( t )} stationarity and homogeneity, related to wave

spectral density S   ( {  } ,  ) as described in (15).

r s

19th IABSE Congress Stockholm 2016
Challenges in Design and Construction of an Innovative and Sustainable Built Environment

E  d Z  ( { } ,  ) d Z  ( { } ,  ) 
T* S ( ,  )  S  (  ) C o h ( )
r s r s r s
 
r s
(15)   ( ) (19)
 S ( { } ,  ) d  x d  y d   i   x cos    y s in  
r s
( )   D ( ) e d
C o h
r s 

Here, the subscripts r and s refer to points in time

and space. The superscripts T and * refer to the Here,  x and  y are the horizontal distances
mathematical operations transpose and complex between point r and s .
conjugate, respectively. The operation E [ ] is the
2.2.1 Directional distribution
expected value.
The directional distribution is commonly
The wave spectral density is divided into a cross-
characterised by a bell shaped function centered
spectral term with r  s and auto-spectral terms
around the mean wave direction. The simplest and
with r  s . The auto-spectral density is denoted
one of the most commonly applied functional
S  ( ,  ) .
forms is the so-called cos-2s distribution, given in
The wave number vector can be described as a (20) for a specific mean wave direction.
function of the wave direction  and the modulus
2 s 1
 ( s  1)   m 
 . 2
D ( ) 
cos  
  ( 2 s  1)  2  (20)
 cos    (   m )  
{ }    (16)
 s in  
Here, s is the spreading parameter,  ( ) is the
Furthermore, within the first-order Stokes theory Gamma function and  m is the mean wave
 and  are related through the dispersion
relationship given in (17).
2.3 Fluid structure interaction
  g  ta n h (  h )
The current analysis of floating bridges is based on
Here, g is the gravitational acceleration and h is the assumption of water being incompressible,
non-viscous and irrotational. Then, within the
the water depth. In the special case of deep water
framework of potential theory, the flow field is
waves the dispersion relationship can be
governed by Laplace’s equation, given in (21) for
approximated as  2  g  . As a result of this Cartesian coordinates [8].
approximation the spectral density can be
described as a function of wave direction and  
 
 

      0
frequency. (21)
x y z
2 2 2

The auto-spectral density is generally a function of

the frequency-dependent directional distribution Here,  is the velocity potential and x , y and z
D (  ,  ) and the one-dimensional wave spectral are Cartesian coordinates. Hence, the basic
density S  (  ) . For simplicity, the directional problem at hand is to find the solution of the
distribution is normally assumed to be frequency- Laplace’s equation in terms of the velocity
independent as given in (18). potential.
Assuming no current and by virtue of the principle
S  (  ,  )  S  (  ) D ( ) (18) of superposition the velocity potential can be
obtained from the linear problem given in (22).
Due to the coherency, C o h 
( ) , between point
r s

r and s the expression for the cross-spectral

density given in (19) is a bit more complicated and
is formulated by assuming deep water waves.

19th IABSE Congress Stockholm 2016
Challenges in Design and Construction of an Innovative and Sustainable Built Environment

    7  nk dS
 i t  i t q j   i 
   0e   7e   kuk 0 (24)
k 1 (22) S0

d iffra c tio n p ro b le m ra d ia tio n p ro b le m

Here, n k represents the component of the surface
Here,  0 and  7 represents the velocity potential normal vector in the direction of the k 'th degree of
from the incident- and diffracted waves, freedom. Comparing the expression with (23) the
respectively.  k represents the velocity potential force is identified as the wave excitation force.
per unit velocity from radiated waves and u k 2.3.2 Hydrodynamic Added Mass and Damping
represents the time derivative of the complex
motion of the body in the water and together they The radiation problem describes the scenario of a
represent the velocity potential from radiated body oscillating in calm sea. Using the same
waves  k   k u k when the body is oscillating in the approach as described in section 2.3.1 the
hydrodynamic action from a body oscillating in
k 'th degree of freedom.
calm water can be found.
From first-order Stokes theory the velocity
potential for the incident wave is known. To obtain q j
  i  u j  j
nk dS
a physical legitimate solution for the other seven S0

velocity potentials in (22) the Laplace’s equation in     (25)

(21) must be satisfied together with the free-   Re  n k d S  u j    Im   nk dS  u
 j
  j
 j

surface boundary condition at the mean water  S0   S0 

level, the kinematic boundary conditions at the M h , jk ( ) C h , jk (  )

seabed and on the wetted body surface and the

radiation condition. Using the indirect boundary Comparing the expression with (23) the
integral formulation and applying Green’s second hydrodynamic added mass and damping can be
identity it is possible to obtain solutions for each of identified.
the seven velocity potentials and the pressure p
can then be obtained through Bernoulli’s equation. 3 Solution Strategy
Applying specific velocity potentials in Bernoulli’s It is commonly assumed, within the field of civil
equation and integrating the hydrodynamic engineering structural dynamics, that structural
pressure over the wetted body surface it is possible damping is very small and hence can be neglected
to obtain expressions for the wave excitation force when calculating the natural frequencies and
and the hydrodynamic added mass and damping natural modes of a classically damped system. In
when comparing to the equation for steady-state the case of fluid structure interaction there is a
harmonic rigid body motion is given in (23). significant contribution to the damping from the
hydrodynamic damping [ C h (  )] and so the system
 M 
 i t
q je  jk
( ) u k  C jk
( ) u k  K jk
uk (23) instead is categorised as a non-classically damped
k 1
system. Procedures exists to calculate this higher
order eigenvalue problem by use of the state-space
Here, the index notations of (4), (5) and (6) is
approach [9]. The solution consists of complex
eigenvalues and complex eigenvectors.
2.3.1 Wave Excitation Load In the context of this article, the dynamic response
is calculated using the direct frequency response
The diffraction problem describes the scenario of a
method with the structure subjected to a set of unit
fixed body in incident waves. By only including  0
amplitude wave with periods ranging from 1
and  7 in Bernoulli’s equation it is possible to second to 15 seconds.
obtain the hydrodynamic action by integrating the
hydrodynamic pressure over the wetted body
surface S 0 .

19th IABSE Congress Stockholm 2016
Challenges in Design and Construction of an Innovative and Sustainable Built Environment

3.1 Direct Frequency Response Method Table 1. Beam element properties

The frequency domain representation described in L Ixx Iyy Izz
section 2.1.1 applies the complex frequency [m] [m4] [m4] [m4]
transfer function [ H (  )] given in (8) to obtain
Bm1 400 1.07E-04 2.65E+00 9.21E-01
solutions in the frequency domain. The response
amplitude { Z u (  )} is a complex quantity Bm2 400 1.07E-04 2.65E+00 9.21E-01
describing the amplitude and the phase angle of
Bm3 8.02 1.17E+01 5.86E+00 5.86E+00
the dynamic response.
By splitting the load into a real part { Z q , R e (  ) } and The mass properties and dimensions of the
pontoon are listed in Table 2 and Table 3,
an imaginary part { Z q , Im (  )} as described in [10] respectively.
the solution can be as shown in (26).
Table 2. Pontoon mass properties
{ Z u ( ) }  [ H ( ) ] { Z q , R e ( ) }
(26) M rxx ryy rzz
 i [ H (  ) ] { Z q , Im (  ) } [kg] [m2] [m2] [m2]

1.37E+06 1.01E+01 6..80E+00 1.15E+01

4 Case Study
The symbols I jj
and r jj represents the second
4.1 Description of Floating Bridge Model moment of area and the radius of gyration around
the j ’th axis, respectively. I x x is the torsional
The model is a simplified floating pontoon bridge
moment of inertia.
with pontoon dimensions equal to the pontoons
used in the mid sections of the Bergsøysund Bridge Table 3. Pontoon dimensions with final draft
h d Aw
The model consists of two horizontal beams, one [m] [m] [m2]
vertical beam and a pontoon. The dimensions of
the bridge are illustrated in Figure 1 where the x - , 6.98 3.61 594
y - and z - axis corresponds to surge, sway and
Supports are located at each end of the floating
heave, respectfully.
bridge model and modelled as fixed in all degrees
of freedom.

4.2 Numerical analysis

Due to the hydrodynamic added mass and
damping, it is crucial to know the correct pontoon
draft before commencing the dynamic analysis.
Therefore, a static analysis is first carried out.

4.2.1 Static Equilibrium

From equilibrium between the pontoon mass and
buoyancy from the displaced water, the initial draft
Figure 1. Bridge schematics of the pontoon is found. The static analysis is then
carried out by replacing the pontoon with a vertical
The cross-sectional properties of the beam
spring stiffness from the waterplane area and the
elements are given in Table 1.
water density. Applying gravitational loads to the
static model the vertical displacement is
computed. The final draft d is found by combining

19th IABSE Congress Stockholm 2016
Challenges in Design and Construction of an Innovative and Sustainable Built Environment

the result from the static analysis with the initial From the analysis, information of the
draft. hydrodynamic added mass and damping as a
function of the period is illustrated in Figure 4.

4.2.3 Applied Rayleigh Damping

Figure 2. Static model with pontoon spring and Assuming a damping ratio of   0 .0 2 the Rayleigh
gravitational load damping is calibrated using the first two horizontal
undamped natural periods T n 1  4 4 .5 s and
4.2.2 Environmental Load Modelling T n 2  1 3 .9 s found from solving the classical

The hydrodynamic restoring, added mass and eigenvalue problem. From the sway response of
damping is calculated using a boundary element the midpoint of the floating bridge, it can be
method software. A panel model of the pontoon checked whether or not appropriate structural
surface as the one in Figure 3 is created and given damping is applied. It is important to have a
as input to the software. The panel model used sufficiently low mass proportional damping in
consists of 632 panel elements and is subjected to order not to damp out the wave response.
60 unit amplitude waves with periods
T  {1 : 0 .2 5 : 1 5} seconds each with a wave 5 Results
direction of 90 degrees from the global x -axis From the dynamic analysis carried out in the
corresponding to sway. The water depth is set frequency domain it is possible to obtain some
equal to 1000 meters. preliminary results of the stress distribution in the
pontoon. The stress response from a set of 60 unit
amplitude mono-chromatic beam sea waves have
been analysed and specific characteristics of the
frequency distribution of von Mises stress has been
observed. At high frequency waves (period in the
range of 1 second to 5 seconds) the largest stresses
Figure 3. Panel- and structural model of pontoon in the pontoon are located in the front part of the
The mesh size of the panel model is roughly 2 pontoon and on the corners connecting the front
meters, which according to [11] requires a vertical concrete plates to the top- and bottom
minimum wavelength of 16 meters or in this case concrete plates, see Figure 5. Maximum values are
in the range of 0.97 MPa.
an equivalent wave period of approximately 3.2

Figure 5. Von Mises stress on pontoon for

monochromatic wave excitation force with
Figure 4. Normalized hydrodynamic added mass T  2 .2 5s . The unit is Pa
and damping in y-direction (sway). Normalization
factors are f m  2 .1 9 E + 0 6 and f c  6 .8 7 E + 0 5 for At lower frequencies the largest von Mises stress is
located exclusively around the connection point
added mass and damping, respectively

19th IABSE Congress Stockholm 2016
Challenges in Design and Construction of an Innovative and Sustainable Built Environment

between the pontoon and the vertical beam as a proper connection between pontoon and bridge
illustrated in Figure 6. The stresses are in this case deck. Future work includes more pontoons and a
as high as 33.8 MPa. stochastic dynamic analysis in frequency- and time

7 Acknowledgement
The late Prof. Ragnar Sigbjörnsson has contributed
to sections 1, 2.1 and 2.2 with an initial draft.

8 References
[1] Skorpa L. Developing new methods to cross
wide and deep Norwegian fjords. Procedia
Engineeering. 2010; 4(1877): p. 81-89.
Figure 6. Von Mises stress on pontoon for [2] Holand I,LI. Salhus floating bridge: theory
monochromatic wave excitation force with and hydrodynamic coefficients. SINTEF
T  8 .5 0 s . The unit is Pa Report. Trondheim: SINTEF; 1972.
The minimum stresses at the low frequency wave [3] Sigbjörnsson R. LI. Wave-induced
excitation is roughly the same order as the stresses Vibrations of a Floating Bridge: The Salhus
from the high frequency wave excitation force. It is Bridge. Trondheim: SINTEF; 1975.
believed that the high stress is a result of the high
wave loads on the pontoon under long waves. [4] Sigbjörnsson R,LI. Wave-induced Vibrations
of a Floating Bridge: A Monte Carlo
6 Conclusion and further work Approach. SINTEF; 1975.
[5] Holand I,I,SR. Dynamic analysis of a curved
The paper has presented general theory on
floating bridge. In IABSE Proceedings; 1977.
solutions of the equation of motion in both time-
p. P-5/77.
and frequency domain and has explained how to
incorporate the randomness of the sea state into [6] Langen I,SR. On stochastic dynamics of
the design using stochastic theory. Also a brief floating bridges. Eng. Struct. 1980 October;
discussion of how potential theory and boundary 2.
element methods can be used when dealing with a
[7] Langen I. Frequency Domain Analysis of a
non-classically damped system such as a floating
Floating Bridge Exposed to Irregular Short-
bridge structure.
crested Waves. Trondheim: SINTEF; 1980.
A case study of a simplified floating bridge
[8] Munson BR, Young DF, Okiishi TH, Huebsch
structure has been presented and preliminary
WW. Fundamentals of Fluid Mechanics. 6th
results of the stress distribution on the pontoon is
ed.: John Wiley & Sons; 2010.
[9] He, J., Fu, Z. Modal Analysis: Butterworth-
From the preliminary analysis in frequency domain
Heinemann; 2001.
it can be concluded from the results given in Figure
5 and Figure 6 that the joint between the pontoon [10] DNV-GL. Sesam User Manual, Sestra, Valid
and the beam bridge structure is crucial in the from program version 8.6. 2014 October
design of the pontoon and, if not thoughtfully 31..
carried out, can generate high stresses in the [11] Faltinsen OM. Sea Loads on Ships and
pontoon surface elements. Offshore Structures: Cambridge University
Although the simplified pontoon bridge is made to Press; 1990.
resemble a realistic floating bridge structure,
many details are lost in the simplification, such as