SPECIAL ISSUE PAPER 377

Estimation of hull girder vertical bending moments

including non-linear and flexibility effects using closed
form expressions
P T Pedersen* and J J Jensen
Department of Mechanical Engineering, Technical University of Denmark, Lyngby, Denmark
The manuscript was received on 6 January 2009 and was accepted after revision for publication on 15 May 2009.
DOI: 10.1243/14750902JEME143

Abstract: A simple but rational procedure for prediction of extreme wave-induced hull girder
bending moment in slender mono-hull displacement vessels is presented. The procedure takes
into account main ship hull characteristics such as length, breadth, draught, block coefficient,
bow flare coefficient, forward speed, and hull flexibility. The wave-induced loads are evaluated
for specific operational profiles.
Non-linearity in the wave bending moment is modelled using results derived from a secondorder strip theory and water entry solutions for wedge-type sections. Hence, bow flare
slamming is accounted for through a momentum type of approach. The stochastic properties
of this non-linear response are calculated through a monotonic Hermite transformation. In
addition, the impulse loading attributable to, for example, bottom slamming or a rapid change
in bow flare is included using a modal expansion in the two lowest vertical vibration modes.
These whipping vibrations are added to the wave frequency non-linear response, taking into
account the rise time of the impulse response as well as the phase lag between the occurrence
of the maximum non-linear load and the maximum impulse load. Previous results for the
sagging bending moment are validated by comparison with fully non-linear strip theory
calculations and supplemented with new closed form results for the hogging bending moment.
Focus is on the extreme hull girder hogging bending moment. Owing to the few input
parameters, this procedure can be used to estimate the wave-induced bending moments at the
conceptual design phase. Another application area is for novel single-hull ship types not
presently covered by the rules of the classification societies. As one application example the
container ship MSC Napoli is considered. Further validations are needed, however, in order to
select proper values of the parameters entering the analytical form of the slamming impulse.
Keywords: wave-induced bending moments, slamming loads, whipping loads, hogging
bending moments, short-term loads, long-term loads

INTRODUCTION

Non-linear wave-induced moments and slamminginduced hull girder whipping bending moments are
important for the ultimate hull girder sectional
bending moment and therefore play a significant
role in hull girder structural reliability (see Fig. 1). At
the same time the long-term prediction of the non*Corresponding author: Department of Mechanical Engineering,
Technical University of Denmark, Section of Maritime Engineering, Building 403, Nils Koppels Alle, Lyngby, DK-2800, Denmark.
email: ptp@mek.dtu.dk
JEME143

linear sectional forces for flexible hulls is complicated by the fact that this depends on a large
number of parameters, which at the design stage are
largely unknown. In the present paper a simplified
procedure is presented for calculation of the shortand long-term distribution of wave-frequency-induced and whipping-induced hull girder stresses.
Much research has been presented on slamming
pressures and the resulting hull girder response.
Starting in the 1970s the research group of Bishop
and Price, then at University College of London [24]
developed a theory for a unified treatment of the

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

378

P T Pedersen and J J Jensen

Fig. 1

A measured time signal of the longitudinal strain in the deck amidships of a naval vessel
(Lpp 5 130 m) containing both wave-frequency parts and high-frequency whipping part
[1]. The maximum strain shown is approximately 100 microstrain, corresponding to
20 MPa. Both port and starboard signals are included. Oblique sea and moderate sea state
(Hs 5 34 m), based on visual observations. Service speed.

combined effect of the wave-frequency hull girder

response and the high-frequency whipping response. These theoretical studies were based on a
non-uniform beam theory with a strip theory to
represent the fluid actions. The present paper will be
based on a similar approach.
There are still no well-established tools available
for determining the total response of ships to
slamming loads combined with non-linear lowfrequency loads such as vertical bending moments.
The random nature of these loads means that the
solutions become complex and are not suitable for
routine engineering evaluations.
In recent years progress has been made on
numerical calculations based on comprehensive
three-dimensional (3D) time simulations coupled
with finite element models; see, for example,
references [5] and [6]. Even if coupled 3D hydroelastic codes exist, however, they are not yet
sufficiently mature for design calculations based on
long-term operational profiles.
At present the main design loads on merchant
ship hulls are based on rules developed by the
classification societies. These rules do not explicitly
take into account such factors as draught, hull girder

flexibility, operational procedures, and so on. There

is therefore a need to develop new design-oriented
prediction tools, which can be used in the initial
design phase and if needed can also serve as a gauge
for detailed numerical calculations using advanced
numerical methods.
With this purpose in mind the overall aim of this
paper is to present closed form estimates for the
wave-frequency hull girder sectional bending moments for ships with fine forms, such as container
ships, and to combine this non-linear response with
the hull girder whipping bending moments caused
by bottom slamming. The procedure derived for
calculation of the response contribution due to
bottom slamming is based on knowledge of the
lowest hull girder natural frequencies and approximate hull girder vibration modes. This structural
part of the approach is verified through comparisons
with results obtained from detailed numerical finite
element analyses. The wave bending moments
caused by whipping are combined with the wavefrequency wave-induced bending moment, taking
into account the phase angles and the effect of
damping. The resulting short- and long-term distributions for the hogging peak values are presented

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

JEME143

Estimation of hull girder vertical bending moments

for two different container ships. One of these is the

container ship MSC Napoli, for which comprehensive numerical load calculations have been performed (see references [7] and [8]).
It must, however, be emphasized that further
validations with full-scale and model tests should
be performed in order to select proper values of
especially the parameters entering the analytical
form of the slamming impulse. With the present
rather limited validation, the procedure is more
aimed at parameter studies and sensitivity analyses
than for accurate extreme load predictions. In that
sense the procedure is to be considered more an
extension of the direct closed form rule-based load
formulas than competing with extensive non-linear
load calculations.

SIMPLIFIED LINEAR AND NON-LINEAR WAVE

FREQUENCY RESPONSES

The basis for calculation of the continuous waveinduced hydrodynamic forces F(x,t) follows from
references [9] to [11] and is based on a strip theory
with time derivation of the momentum of the added
mass of water surrounding the hull. The xyz
coordinate system is fixed in the vessel, with x in
the direction of ship longitudinal axis pointing
forwards and z pointing upwards. In addition to
forces due to a change in momentum of the added
mass of water, a damping term and a restoring term
are included, both dependent on the relative motion.
Thus the force per unit length of the hull acting at
position x is taken in the form



D
D~z
D~z
~
F x,t ~{
mh z,x
zN ~z,x :
Dt
Dt
Dt
{~
z

z
{T

9
=
Lpx,z,t
jzzw Bz,xdz
;
Lz

where the difference between the absolute displacement of the ship in the vertical direction, w(x,t), and
the modified surface elevation of the ocean, h(x,t), is
denoted by z(x,t), the added mass per unit length by
mh, and the damping by N. The modification of the
real wave elevation h(x,t) takes into account the
exponential decay of the wave particle motion
through the Smith correction factor k. The use of
the Smith correction factor rather than direct
pressure integration of the hydrodynamic pressure
is, of course, an approximation that could be
JEME143

379

avoided. Here it is retained as it leads to closed

form expressions for the wave-induced loads [12], to
be used in the present procedure. The operator D/Dt
is the total derivative with respect to time t; that is
D
L
L
: {V
Dt Lt
Lx
where V is the forward speed of the ship. The
breadth of the ship is denoted by B(z,x) and the draft
by T(x). The FroudeKrylov fluid pressure is denoted
by p(x,z,t). The integration of the pressure over the
section accounts for the vertical motion w of the
section. It might be more logical to define z(x,t) with
the opposite sign, but the sign convention here is
taken as the same as in the original publication of
this procedure [10].
If the z dependence in mh, N, and B is neglected,
the force expression corresponds to a linear strip
theory. As described previously by Jensen and
Pedersen [10], it is possible to evaluate F(x,t) by a
perturbation method, taking into account linear and
quadratic terms in the relative displacement z and
thereby in the displacement of the hull w and the
wave surface elevation h.
Neglecting the bottom slamming pressure but
including the linear and the quadratic terms in
equation (1) the wave-induced bending moment
Mrigid body,quadratic has been calculated for a number
of different ship types assuming rigid hulls. Based on
these comprehensive numerical calculations an
estimation procedure of the long-term waveinduced bending moments based on closed form
expressions has been suggested by Jensen and
Mansour [12]. The expressions are based on analytically determined response operator functions for
simplified ship hull geometries; i.e. the same type of
expressions as those sought in the present paper.
The closed form expressions for the standard
deviation of the linear part of the wave-induced
bending moment amidships for ships sailing in longcrested irregular waves modelled by the Pierson
Moskowitz (ISSC) wave spectrum were in reference
[12] derived in the following form
sM ~rgBL2 HS WTz ,b,Fn ,CB ,d

where the significant wave height is denoted HS, the

function W depends on the zero-up-crossing period Tz
and the course angle b. The ship velocity is given by the
Froude number Fn, the hull form through the block
coefficient CB, and the draft d. For other sea spectra
similar results can be obtained from reference [13].

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

380

P T Pedersen and J J Jensen

The non-linear part represented by the quadratic

terms in equation (1) gives separate predictions for
hogging and sagging bending moments; i.e. the
contribution from momentum slamming caused by
immersion and re-emergence of the linearly flared
forward and aft part of the vessel is included in the
quadratic strip theory described by equation (1) and
therefore also in the resulting non-linear sectional
forces.
Non-linear numerical calculations in reference
[10] show that the wave frequency non-linear
hogging bending moment peaks become slightly
lower than the linear prediction. On the other hand,
the sagging bending moment peaks become much
larger and increase with the bow flare. In reference
[12] an empirical procedure is developed whereby
first the skewness of the response is approximated
by an analytical expression in the wave spectrum
parameters HS, Tz, the flare coefficient Cfsag and the
Froude number Fn



{10Fn
k3 ~0:26HS Cfsag 1{exp
jcosbj


Tz {5
,1
|min
5

The flare coefficient Cfsag is here defined as the

difference between the upper deck area and the water
plane area in the forward 20 per cent of the ship hull
normalized by the freeboard in this area and with the
ship length. Together with the analytical approximation for the linear standard deviation of the bending
moments sM, this skewness is used to derive the
probability distribution for the peaks of the waveinduced sagging bending moments by application of
a second-order Hermite series, see references [12]
and [14]. Hence, the probability P that an individual
peak M exceeds a given level m becomes


1 2
P Mom~exp { u m
2

where u is given by
u~

{1z

p


1z4xxzm
2x

with
x~

k3
m

~
,m
sM
6

It is here assumed that x % 1 in accordance with the

numerical results in references [10] and [11].

For the hogging bending moment a second-order

transformation cannot be applied as it is nonmonotonic. Therefore, a third-order transformation
is used, as it is monotonic for all values of the
(negative) skewness provided that the kurtosis is
larger than 3 with an amount depending on the
skewness. Previous second-order strip theory calculations, e.g. references [10] and [11], have shown that
the calculated kurtosis values are generally only
marginally larger than this threshold level. Hence, as
it has not been found possible to estimate the kurtosis
by a closed form expression similar to the skewness,
equation (3), the third-order Hermite transformation
will apply this threshold level for the kurtosis and
reads, again assuming x % 1
u~

1=3
{1z1z3xxzm
x

The skewness k3 is again estimated from equation (3),

but with the flare coefficient Cfhog, now defined as the
difference between the bottom area and the water
plane area in the forward 20 per cent of the ship hull
normalized by the draft in this area and with the ship
length. From a theoretical standpoint this approach
can certainly be questioned as the wave bending
process, sag and hog, should have the same statistical
moments except for a sign change in the odd
moments (considering both sag and hog peaks to be
positive). Extreme responses are sought, however,
and these have to be modelled here solely through the
standard deviation and an effective skewness. The
linear part of the hogging and sagging moments are
the same and therefore also the standard deviations.
The non-linearities are then to be reflected through
the skewness alone. As the even and odd statistical
moments counteract each other in the hogging
moment process the non-linearities and hence the
effective skewness in hogging will be smaller than in
sagging. Therefore, the hogging and sagging bending
moments are considered as two partly independent
processes, each with their own skewnesses, but with
the same standard deviation. The probability of
exceedance is still given by equation (4).
In order to validate this approach the results from
equations (3) to (6) are compared with results
obtained previously by Jensen [15] using time
domain non-linear strip theory calculations together
with extreme value calculations based on the firstorder reliability method (FORM). The container ship
considered in reference [15] has a length of 276.38 m
and a breadth of 32.2 m. The ship is sailing at 5 m/s
in head sea in a sea state represented by long-crested

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

JEME143

Estimation of hull girder vertical bending moments

waves and a PiersonMoskowitz spectrum with a

significant wave height of 10 m and a mean wave
period of 11.35 s. Extreme values are in reference
[15] presented in terms of the mean out-crossing
rate n(m) related through the assumption of statistically independent peaks to equation (4) by


1
n m
P Mom~exp { u2 m ~
2
n 0

From the body plan for this container vessel the flare
coefficient is found to be 0.2 and 20.07 for the
sagging and hogging wave bending moment, respectively. The corresponding skewnesses from equation
(3) are k3 5 0.35(sag) and k3 5 20.12(hog). Figure 2
shows the comparison between the results from the
time domain simulations [15], and the present
closed form expressions. For sag both the secondand third-order Hermite transformations are applied
and it is seen that they differ only marginally. An
excellent agreement is in general obtained except in
sagging at very low probabilities of exceedance.
Here, however, neither the time domain simulations
(which here include water on deck) nor the closed
form expressions can be considered as the correct
result owing to the strongly non-linear wave kinematics in these situations. The implicit use of the

minimum kurtosis (~3z2k23 9) required for a
monotonic transformation is seen to provide a

381

uniform agreement with the non-linear calculations

for all probability levels. For comparison the International Association of Classification Societies
(IACS) rule sagging wave bending moment is close
to 4000 MNm for this vessel.
The results shown in Fig. 2 are in line with the
general observation that the sagging peaks are higher
than the linear peaks, whereas the opposite is the
case in hogging as long as the ship hull girder is
considered rigid, i.e. neglecting the flexibility of the
hull girder. Hull girder deformations will be considered in the next section.

DYNAMIC RESPONSE OF FLEXIBLE HULLS

As it is generally accepted that the lower modes of

ship hull vibrations can be determined quite accurately through modelling the hull by a few global hull
girder modes, the whipping response in the vertical
plane will be based on a modal formulation and the
vertical hull girder deflection w(x,t) will be approximated as a series in the form
wx,t ~

N
X

ui t vi x

i~0

where ui(t) are the modal coordinates to be

determined and where vi(x) are approximations to
eigenvectors for free vibrations of the homogeneous,
self-adjoining part of the hull girder beam or finite
element models. These trial functions vi(x) are
assumed orthogonalized so that
L



ms vi vj dx~dij Mi

Fig. 2

JEME143

Comparison between the mean out-crossing

rate determined by numerical simulations [15]
and the present closed form expressions,
equations (2) to (7). Short-term results
(HS 5 10 m and a mean wave period equal to
11.35 s): the thin dashed lines are the linear
results (Rayleigh distribution), the thick continuous curves are the results from reference
[15], the dasheddotted curves are the results
from the third-order transformation, equation
(6), and the dotted line (sag only) represents the
results from the second-order Hermite transformation, equation (5)

where ms(x) is the mass per unit length excluding the

added mass contribution, dij is Kroneckers delta,
and Mi are the generalized masses. Furthermore, the
trial functions are assumed to be ordered so that v0 is
the heave mode, v1(x) is the pitch mode, v2(x) is the
approximation to the two-node vibration mode
2, and so on.
associated with the natural frequency V
If the series (8) is substituted in the differential
equations governing the structural response, and the
orthogonality relations, equation (9), are used, the
following set of equations of motion can be obtained

2 uj

j u_ j zV
j zjj V
u
j
!
#
L "
n
X
1
~
vj F x,t,
ui vi zqx,t dx
Mj 0
i~0

10

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

382

P T Pedersen and J J Jensen

where j 5 0,1,2,N, jj is a damping coefficient, F(x,t)

is the low-frequency external force per unit length
given by equation (1), and q(x,t) denotes the transient
slamming load per unit length along the hull girder.
For normal ships the lowest hull girder frequencies are so high that the wave-induced forces given
by equation (1) do not excite hull girder vibrations.
Therefore, a simplified solution to this problem can
be obtained by breaking the response down into two
parts.
The first part is the basic closed form rigid hull
solution (j 5 0 and 1) to the quadratic problem
formulated above, neglecting transient loads, i.e.
q(x,t) 5 0, but including momentum slamming due
to more slowly varying hull geometries, i.e. bow flare
slamming. Equation (10) is of course an approximation, but a unique or better separation between
momentum slamming and impulse loads has not
been identified. As shown above this results in an
analytical expression for the stochastic distribution
of wave-frequency sectional hull girder forces represented by equations (4) to (6).
The second part associated with j 5 2, , N will be
a closed form approximation to the transient slamming induced hull girder sectional forces, i.e.
neglecting the wave-frequency load F(x,t) but including the transient slamming loads through the
load term q(x,t).
As trial functions vi(x) polynomials will be used,
fulfilling the orthogonality conditions, equation (9)
and fulfilling the freefree beam boundary conditions.
In addition the functions are chosen to be symmetric
or anti-symmetric about amidships and normalized
such that the maximum deflection at x 5 0 and L
equals 1. The total mass distribution m(x) is taken as a
symmetric mass distribution, increasing linearly from
zero to 0.3L forward and aft and constant on the
remaining 40 per cent middle section. With these
constraints the least-squares fitting approximation to
the two-node vibration mode becomes
x
3
x
2
x

{1:23
{4:58
z1;
v2 x~7:95
L
L
L
x
0 0:5
L




L
L
v2 xz ~v2
{x
2
2

11
and shown in Fig. 3. With these assumptions the
modal mass in equation (9) M2 5 0.089M0, where M0 is
the total wet mass.
Similar expressions are derived for the three-node
vibration form. It would be straightforward to derive

Fig. 3 Approximation to the vertical two-node vibration mode as a function of x/L, equation (11)

and apply mode shapes based on a real and

asymmetric mass distribution, but as the procedure
is intended for use in the early design phase, this is
not done here.
Provided that the duration of the slamming force
q(x,t) is short compared with the considered hull
girder vibration frequencies, then the associated
approximate whipping bending moment can be
found as
Mwhipping x,t ~{

N
X
j~2

:{

N
X

j
u



x{fms vj f df

j S j x
u

12

j~2

By considering the dynamic load as an impulse a

slightly unconservative estimate might be obtained
in some cases, but generally the approximation is
good when compared with the exact dynamic
amplification found in a single degree-of-freedom
system subjected to a triangular load: f(t) 5 t/Tc;
0 ( t ( Tc; 0 elsewhere, c.f. reference [16]. Based on
the approximation to the two-node vibration mode
given by equation (11), the variation of S2(x)/LM2 is
shown in Fig. 4. It is noted that the variation S2(x)/
LM2 for the non-uniform beam approximation of the
hull rather well resembles the numerical results for
the spatial distribution of the wave-induced bending
moments obtained from the second-order strip
theory, see reference [17].
In order to obtain simplified expressions for the
modal amplitudes ui(t) caused by slamming loads,
equation (10) will be considered without the forcing
function F(x,t) but with a generalized slamming load
Pi t ~vi x0 bLqx0 ,t

13

where bL is the longitudinal extent of the slamming

area centred around x0.

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

JEME143

Estimation of hull girder vertical bending moments

Fig. 4

Normalized bending moment variation S2(x)/

LM2 as function of x/L for a non-uniform beam
vibrating in two-node mode, given by equation
(11)

Normally the time variation of the slamming load

is not known in any detail. Therefore, the slamming
impulse will be introduced instead
Ii ~

ts

Pi t dt

14

and it is assumed that the slamming load P(t) . 0 is

acting during a time ts, which is short compared with
one-quarter of the vibration period Ti. With these
assumptions it is shown in reference [9] that an
approximation to the maximum modal displacement for t 5 Tc can be derived as
uimax ~

Ii
e{ji Vi Tc
V i Mi

15

where Vi are the wet natural frequencies. Similarly,

the modal contributions to the maximum whipping
bending moments are found as
Mimax x~Ii Vi L

Si x {ji Vi Tci
e
LMi

16

where Tc denotes the rise time from the time of

impact to the maximum value of the response. For
superposition with the wave-frequency-induced
loads derived by equation (1), the phase angles for
the time of maximum impact pressure also need to
be considered. Here these phase lags are taken
relative to peaks in the wave-induced sagging
bending moments. Positive means that the whipping
bending moments occur before these peak values.
With this definition the phase lag for maximum
bottom slamming pressure is taken as 2p/3 and the
phase lag for the momentum bow flare slamming is
taken as p/6. These phase lags are based on
numerical calculations in regular and irregular
waves using a non-linear strip theory program, see
JEME143

383

reference [16], and are subjected to large uncertainties. When considering the peak hogging bending
moment p should be added to these phase lags. The
phase lags only enter equation (16) in the exponential term and, as seen for example in Fig. 1, the effect
of the phase lag is limited if there is a large difference
between the encounter and the two-node periods.
Hence, fixed values for the phase angles are
considered to be acceptable in the present context.
For derivation of these extremely simple expressions for the whipping-induced responses a number
of simplifying assumptions have had to be introduced in order to obtain formulas which can be
applied easily at the early design stages. It has been
assumed that the response can be determined by a
small number of modes, that the actual time
variation of the slamming load can be represented
by the impulse of the load, that the actual mass
distribution can be approximated by a symmetric
idealized distribution, that the actual eigenvectors
can be approximated by some polynomials, and so
on. In order to show that the procedure still has
some relevance, the results of a response calculation
will now be compared with the results of comprehensive numerical finite element analyses.
As an example a Panmax container ship with high
sides will be considered. The main particulars are
given in Table 1. The ANSYS finite element model of
the ship structure can be seen in Fig. 5. The added
mass of the surrounding water on the lowest natural
frequencies is modelled by use of three parameter
Table 1

Main particulars for the Panmax container

ship

Length o.a. (m)

Breadth (m)
Draught (m)
Depth (m)
Deadweight (deadweight tons)
Container capacity (twenty foot equivalent units)

Fig. 5

294.0
32.2
12.5
30.2
55 000 DWT
4600 TEU

Element mesh used for dynamic finite element

analysis

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

384

P T Pedersen and J J Jensen

conform transformations (Lewis forms) of the ship

cross sections.
The modal analysis carried out for frequencies up
to 3 Hz reveals two vertical vibrating mode shapes,
and three coupled horizontaltorsion mode shapes.
The two lowest natural frequencies in the vertical
plane are 0.62 Hz and 1.07 Hz. The corresponding
mode shapes are shown in Fig. 6.
The numerical finite element model is now subjected to an idealized slamming load at x 5 290 m
in the form of a load qbL which increases linearly
during 0 , t , 0.068 s to 9 MN and thereafter decreases to zero, also linearly from 0.068 s , t , 0.136 s.
The magnitude, 9 MN, of this load is purely artificial,
but as the flexible analysis is linear the result
anyway serves to validate the whipping procedure
used here.
First the longitudinal stress response amidships
near the deck is calculated in the time domain using
the complete finite element model. This is a
comprehensive calculation, which here is used for
comparison with the results of two different modal
analyses in order to obtain some evidence on the
accuracy of the proposed procedure. For the undamped structure the calculated stress variation
with time is shown in Fig. 7.
Thereafter a modal analysis procedure is applied to
calculate the stress at the same location but now using
only the two vertical vibration modes depicted in
Fig. 6. This procedure is based on the eigenvectors
derived from the finite element model; i.e. the impulse
procedure derived above is applied but using the
actual eigenvectors determined by the 3D finite
element model. Thus, for instance, the stress response
at some location in the hull is determined from
sx,t ~

3
X

ui t si x

i~2

where si(x) is the stress eigenvector for mode number

Fig. 6

Fig. 7 Numerical comparison of whipping stress in a

Panmax container ship at the deck level amidships. Results are shown for a full numerical
time integration using a 3D finite element
model, a two-mode solution procedure using
the actual eigenvectors from the 3D finite
element model, and finally a one-mode solution based on an assumed mode shape,
equation (11)

i, and where the modal amplitudes are determined

from equation (15)
uimax ~

Ii
e{ji Vi Tc
Vi Mi

where the impulse Ii is calculated from equation (14)

using the numerically determined modal deflections
at the location of the slamming impact. The result is
included in Fig. 7. It is seen that this modal expansion
gives a good approximation to the maximum whipping stresses even if the higher mode contributions
found by the time domain analysis cannot be
represented. It is also seen that, owing to the
assumptions associated with representation of the
time variation of the slamming force as a simple

Lowest vertical hull girder vibration modes: the two-node mode with frequency 0.62 Hz
(left) and the three-node mode with frequency 1.07 Hz (right)

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

JEME143

Estimation of hull girder vertical bending moments

impulse acting at time t 5 0, there is a slight phase

difference.
Finally, the same longitudinal stress will be
calculated using the simple analytical expressions
determined above; i.e. instead of using the actual
modes found by the finite element analysis, the
response is determined based on the assumed mode
shapes and mass distributions.
As the assumed three-node vibration mode is
taken to be symmetric about amidships, the dynamic hull girder bending moment will always be
zero amidships for this mode. Therefore, for calculation of whipping stresses amidships only the
assumed two-node vibration mode will give a
contribution to the longitudinal stress at the point
considered.
The calculation is very simple. At x0 5 290 m the
two-node mode deflection calculated from equation
(11) is v2 5 0.937 and the impulse then becomes, see
equation (14)
1
I2 ~ |9|0:136|0:937~0:5734 MN s
2
and the amplitude of the whipping bending moment
is determined from equation (16) and Fig. 4
Mwhipping, amidship ~I2 V2 L

S2
LM2

~0:5734|2p|0:62

SLAMMING-INDUCED LOADING

In order to relate the impulse I to the slamming

pressure, two types of slamming loads are considered. The first is bottom slamming, which occurs
when the relative displacement at any section
exceeds the draught so that the local section
emerges and then re-enters the water. The magnitude of the bottom pressure impulse is here
estimated from the slamming load per unit length
on a wedge derived from calculations performed by
Zhao and Faltinsen [18]
qt ~3

From standard beam theory the moment of resistance at the considered location is determined as
Wdeck 5 32.0 m3 and thus the maximum stress
amplitude is found to be

smax ~4:1 MN m2
The time variation of this simplified solution is also
shown in Fig. 7.
The fact that the maximum stress value obtained
from the solution of the simple analytical expression
is rather close to the one obtained from the full finite
element numerical integration indicates that the
simple analytical expression can give a reasonable
estimate of the response/stress and therefore can be
a strong tool in the determination of impact
responses.
For sections away from amidships similar comparisons have not been made, but as the results here
depend strongly on the three-node mode shape the
accuracy is probably smaller.

p2
rz_ 3 t
4 tan2 a

17

where the wedge angle is denoted a and with r, z, t

being the density of water, the vertical relative
velocity at the slamming location, and the time
measured from the wedge apex hitting the water,
respectively. The relative vertical velocity is considered to be constant in this case when dealing with
short impulses. Furthermore, as impact with a flat
surface is assumed, any effect of wave slope is
ignored. For bottom slamming the time at maximum
immersion including account for the water up-rise
on a section with dead rise angle a and local breadth
Bl is approximately
ts ~

|294|0:2~131:3 MN

JEME143

385

2 Bl tan a
3 2z_

18

taking into account that the water up-rise is about

half the immersion of the section according to the
results shown in reference [18].
Denoting the effective longitudinal extent of the
slamming area bL and the longitudinal centre of
pressure x0 then the weighted impulses of the
bottom slam is
Ii ~vi x0 bL

ts
0

qx,t dt~

p2
bLrvi x0 B2l z_ x0
24

19

The second type of slamming load is caused by a

rapidly decreasing angle of the bow sections with the
horizon, both in the transverse and in the longitudinal directions; i.e. the angle at which the wave
hits the shell plating. It is noted that the quasi-static
calculation procedure presented above includes the
two-dimensional (2D) effect of a linear increase and
decrease of the waterline breadth with immersion
and that the momentum bow flare slamming
considered here is confined to the effect of the large
waterline increase that most container vessels have

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

386

P T Pedersen and J J Jensen

in the fore-ship, above the still water level, in order

to increase deck area and to protect cargo on the
deck from water impact. A mathematical representation of this momentum slamming load can be
derived considering the bow forward of FP (forward
perpendicular) as a simple half cone (which often
will model the bow fairly well, as it is a surface that
can be developed from flat plates). Then, the
momentum impact of that body can be determined
by a von Karman approach, where the added mass of
a cone is taken to be equal to the added mass of a
disc having the same radius as the cones penetration of the still water plane, i.e. r 5 z cot b, where z is
the relative immersion and b is the angle between a
horizontal plane and the cone surface. The added
mass of a disc in an infinite fluid is 8/3 r3r. Thus for a
half cone neglecting the Wagner correction, the
following is found
4
M~ rr 3
3
According to Ramos et al. [19] the slamming load
terms proportional to the forward velocity V play a
less significant role for the whipping loads. This
finding should of course be supported by further
validations before it is taken for granted. Generally,
large horizontal wave forces are experienced in the
bow area in severe sea states, especially when the
wave slope approaches the angle of the bow profile,
so their influence can be neglected only if they do
not lead to significant vertical hull bending moments.
Neglecting the effect of forward speed V and wave
slope on the slamming load for a given vertical
velocity z, the impact of the bow flare slamming load
is found as
t

M2 z_

M1 z_

Ii, impact ~vi x0 F t dt~vi x0

dM z_

in order to cause dynamic amplification. Hence, for

bottom slamming


p 1
Vcr, i ~max 0:093 Lg, Bl Vi tana
6

21

In the cases considered in the present paper, the

second criterion in equation (21) is the active one. As
the whipping bending moment depends linearly on
the impulse I and thereby on the relative vertical
velocity, then the standard deviation of the whipping
bending moment
sm, whipping x,Tc ~

p2 2 2
bL Bl r sz_ x0
24
N
X
Si x {ji Vi Tc 22
|
vi x0 Vi
e
LMi
i~2

can be added directly to the standard deviation of

the wave-induced bending moment.
Here sz is the standard deviation at the location x0
of the relative vertical velocity. It is noted that the
time correlation is taken care of by the exponential
decay factor in the whipping response. The relative
occurrence of whipping events due to bottom
slamming is taken as
"




 #
1 Vcr, i 2 1 dl x0 2
{
gwh, i ~exp {
2 sz_ x0
2 sz x 0

23

where dl is the local draft at the slamming position,

see Fig. 8. The last term follows from the requirement that the angled section must be out of the
water prior to a slam. The standard deviations sz, sz
are here determined by the closed form expressions
given in reference [16]. Finally, the probability
P(Mw + wh . m) that the combined wave and whipping bending moment exceeds a value m in a
stationary stochastic condition is written in the
following form

or

4 
~z_ vi x0
Ii, impact ~ r z23 {z13 cot3 b
3

4 
~ r r23 {r13 z_ vi x0
3

20

In the following it is assumed that in order to

develop a slamming impact, the vertical velocity
shall be larger than the classical threshold value by
Ochi. In addition the duration of the slam ts must be
small compared with the relevant whipping period
Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

Fig. 8

Wedge impact into water

JEME143

Estimation of hull girder vertical bending moments

387



P Mwzwh wm~ 1{gwh,2 P Mw wm

 

z gwh,2 {gwh,3 P Mwh,2 wm


24
zgwh,3 P Mwh,3 wm
Here P(Mw . m) is calculated using the standard
deviation sM and skewness k3 as calculated by
equation (4), whereas P(Mwh . m) applies the same
skewness, but a standard deviation equal to sM +
sm, whipping. The subscript i 5 2, 3 signifies the
relevant two- and three-node properties (natural
frequency, damping factor, mode shape, and bending moment variation). P(Mwh,i . m) is calculated
with the two-node (i 5 2) or the two- and three-node
modes (i 5 3) included in equation (22).

5
5.1

Fig. 9

VERIFICATION AND APPLICATION EXAMPLES

Effect of whipping on hogging bending
moments

The proposed procedure for inclusion of whipping in

the stationary-wave-induced hogging bending moment distribution can be compared with the shortterm non-linear results in reference [15]. Similar
comparisons are made in Fig. 2 for the rigid vessel
and the same ship and operational condition are used
again. The only change is that now the hull flexibility
is included. The two-node period of the vessel is
assumed to be 1.5 s and the damping to be 1 per cent
of the critical damping. Both values are somewhat
arbitrary, but within the range usually found from
full-scale measurements, c.f. reference [6].
The slamming extent is taken as bL 5 0.07L and
Bl 5 15 m. The other slamming-related parameters,
Fig. 8, are set to a 5 5u and dl 5 9 m. The quality of
these four input data are crucial for the accuracy of
the whipping predictions and only comparison with
model test and full-scale data can indicate whether
proper physical plausible values can be defined. A few
comparisons are given in the following, but they are
far from sufficient to give any firm guideline on how
to define them. The results are shown in Fig. 9 in
terms of the probability of exceedance. Using equation (7) and n(0) 5 0.076s21, these calculated mean
out-crossing rates n(m) can be compared with the
corresponding comprehensive numerical results in
reference [15]. Both sets of results show that whipping starts to become important when the probability
of exceedance is below 1022 in the present severe sea
state (Hs 5 10 m) and that the whipping contribution
to the hogging bending moment implies that the
JEME143

Probability of exceedance for the short-term

(Hs 5 10 m) non-linear wave-induced hogging
bending moment for a flexible ship. Data as for
Fig. 2, except that whipping is now included.
Flare coefficient Cfhog 5 20.07. Thin curve:
linear rigid hull result (Rayleigh distribution);
thick curve: non-linear flexible hull result,
equations (2) to (4), equation (6) and equations
(21) to (24)

probability of exceedance for the hogging bending

moment for the flexible hull now exceeds that from
the corresponding linear rigid hull calculations. A
more detailed comparison between the results presented in reference [15] and the present results is not
meaningful as the parameters determining the slamming impulse are not known accurately. The main
conclusion is that the sole use of linear rigid hull
results for the wave-induced hogging bending moment can lead to unconservative extreme wave
loading estimates owing to the effect of the whipping-induced hull girder vibrations.
5.2

MSC Napoli, a short-term condition

In January 2007 the 4400 TEU container ship MSC

Napoli suffered extensive hull damage during a
voyage in the English Channel. The hull girder
damage in front of the engine room was attributed
to a hogging bending moment exceeding the
ultimate hull girder capacity. MSC Napoli has the
overall length of 275.6 m, breadth moulded 37.1 m,
mean moulded draught 12.63 m, and block coefficient 0.609. The vessel was sailing with a speed of 11
knots against a sea with Hs 5 7.59 m and Tz 5 10
11 s. The lowest wet hull girder natural frequencies
for vibrations in the vertical plane are 0.594 Hz and
1.223 Hz. More details of the vessel and the environmental conditions are given in references [7] and [8].
For this vessel the available body plan does not
provide enough information for a detailed calculation
of the flare coefficients Cf. It must, however, be
expected to be more negative than for the smaller
Panmax vessel (with Cfhog 5 20.07) considered in
Fig. 2 and is thus taken as 20.1. The bending moment

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

388

P T Pedersen and J J Jensen

is calculated approximately at the location where the

damage occurred. The very precise location is not
relevant owing to the rather rough model, Fig. 4, of
bending moment variation used here.
The effect of whipping is calculated using the
following slamming-related parameters: longitudinal
extent bL 5 0.07L, Bl 5 16 m, a 5 5u, and dl 5 12 m.
The results are shown in Fig. 10 for long-crested
head waves in a sea state given by a Pierson
Moskowich spectrum with Hs 5 9 m and Tz 5 11 s.
The reducing effect on the hogging bending moment
of the non-linearity due to the flare is seen in Fig. 10
for exceedance levels above 1022, whereas whipping
becomes very important thereafter. The most probable wave-induced hogging bending moment during
3 h (1000 peaks) of operation under these conditions
becomes 2049 MNm (linear, rigid hull), 1861 MNm
(non-linear, rigid hull), and 2592 MNm (non-linear,
flexible hull). The same overall behaviour, albeit with
somewhat lower bending moments, is seen for the
other sea state, Hs 5 7.5 m and Tz 5 10 s, considered
in relation to this accident. The calculated values can
be compared with 1900 MNm for the non-linear
rigid hull and 2470 MNm for the non-linear flexible
hull estimated in reference [8] for the same probability level and to a rule hogging bending moment
of about 2000 MNm.

5.3

Long-term predictions

The basic assumption in the long-term analysis is

that the process can be modelled as a sequence of
stationary processes with independent peaks. With a
given operational profile and a relevant wave scatter
diagram the long-term distribution of the non-linear

Fig. 10

Probability of exceedance for the short-term

(Hs 5 9 m, Tz 5 11 s) non-linear wave-induced
hogging bending moment for MSC Napoli.
Flare coefficient Cf 5 20.1. Whipping is included. Thin curve: linear rigid hull result
(Rayleigh distribution); thick curve: non-linear
flexible result, equations (2) to (4), equation
(6) and equations (21) to (24)

wave-induced wave-frequency loads and accelerations and the whipping loads can be calculated. The
operational profile specifies the relative occurrence
of short-term conditions with constant forward
speed, heading, significant wave height, and upcrossing period. Wave scatter diagrams for the North
Atlantic, the North Pacific, or the Asia/Europe trades
are generally used. These long-term scatter diagrams
can be modified in order to account for the effect of
weather routing. Significant effects of weather routing on the long-term wave scatter diagram have
been found in references [9], [13], [20], and [21],
reducing the predicted extreme values significantly.
In reference [9] it is shown how the long-term
prediction of sagging bending moments is influenced by the shipping route, application of weather
routing, the longitudinal extent of the slamming
area, etc. Even if the effects of non-linearities are
most significant for the sagging bending moments,
the practical importance of wave-induced sagging
bending moments are often of minor importance for
the hull girder owing to high still water hogging
bending moments in slender ships.
Here the long-term hogging bending moment is
first shown to be also influenced by whipping. This
has been ignored in reference [9], which focuses only
on the sagging bending moments. Figures 11 and 12
show such results for a container vessel (length
overall of 283.8 m, breadth moulded 42.8 m, mean
moulded draught 14 m, and block coefficient 0.60)
where extensive full-scale measurements have been
collected and analysed by Okada et al. [22]. Similar
results for the sagging bending moment were presented in reference [9], where the assumed long-term
operational conditions are also described. The operational conditions model on average the actual voyage
recordings for this vessel and the scatter diagrams
model the actual route. Thus the calculations are

Fig. 11 Probability of exceedance for the long-term

non-linear wave-induced hogging bending
moment for a rigid ship. Flare coefficient
Cfhog 5 20.12. Thin curve: linear rigid hull
results; thick curve: non-linear rigid hull result

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

JEME143

Estimation of hull girder vertical bending moments

Fig. 12

Probability of exceedance for the long-term

non-linear wave-induced hogging bending
moment for a flexible ship. Flare coefficient
Cfhog 5 20.12. Longitudinal extent of slamming bL 5 0.07L. Thin curve: linear rigid hull
results; thick curve: non-linear flexible hull
result

intended to model the actual measured data rather

than being design calculations, where head sea and
the IACS scatter diagram should be applied.
For this vessel the available body plan does not
provide enough information for a detailed calculation of the flare coefficient Cfhog. It must, however,
be expected to be more negative than for the smaller
Panmax vessel (with Cfhog 5 20.07) considered in
Fig. 2 and also for MSC Napoli and therefore the
value Cfhog 5 20.12 is used. For the rigid hull, Fig. 11
shows that non-linearities slightly reduce the extreme hogging bending moment when compared
with the linear results. This is in accordance with the
short-term results shown in Fig. 2. A more negative
Cfhog value than 20.12 will lead to smaller hogging
bending moments.
In order to analyse the effect of whipping, the
slamming-related parameters are chosen as in reference [9], i.e. longitudinal extent bL 5 0.07L 5
19.9 m, Bl 5 20 m, a 5 5u, and dl 5 10 m. The longterm hogging bending moment result, but now
including the whipping effect, is shown in Fig. 12.
This figure clearly shows again that whipping greatly
changes the extreme hogging bending moments,
which now become much larger than the corresponding linear rigid hull results for low exceedance
probability levels. It is also noticed that whipping
first becomes important at long-term probability of
exceedance levels between 1024 and 1025. An extrapolation to an exceedance level of 1028 does not seem
reasonable. The reason for this can be attributed to
the four input parameters, given above, for the wedge
configuration, but also to the quality of the closed
form expression [13], for the vertical motion in the
fore part of the vessels, entering equation (22)
through the standard deviation sz(x0) of the relative
JEME143

Fig. 13

389

Probability of exceedance for the long-term

non-linear wave-induced hogging bending
moment for a flexible ship. Flare coefficient
Cfhog 5 20.12. Longitudinal extent of slamming bL 5 0.07L. Thin curve: linear rigid hull
results; thick curve: non-linear flexible hull
result. Bl 5 L/20, a 5 30u, and dl 5 0 m

vertical motion. More accurate motion estimates,

however, probably require more knowledge of the
hull geometry and loading condition than those
described by the main particulars of the vessel.
All the preceding calculations have been done
assuming that bottom slamming at the fore body is
the dominating source for the impulse load. If the
impulse load instead occurs on the bow flare,
however, reasonable parameters could be, for example, bL 5 0.07L, Bl 5 L/20, a 5 30u, and dl 5 0 m.
The results then predict that whipping generally
starts to occur at higher probability levels (owing to
the change in a, dl), but with smaller values (owing to
a and Bl). This possibility will be considered in a
future study, but the result corresponding to Fig. 12
is shown in Fig. 13.

CONCLUSIONS

It has been the purpose of the present paper to

present a probabilistic procedure that can be used at
the early design stage to determine the sectional
forces on ships, taking into account the non-linear
contributions due to bow flare and hydroelastic
effects associated with slamming loading. The procedure is fast and easy to apply and the only input data
are those available at the design stage, i.e. main
dimensions, flare coefficients, and estimates of the
lowest natural frequencies besides the environmental
conditions for which the ship is to be designed. All the
comparisons performed with the presented procedure indicate a reasonable engineering accuracy.
At present the wave-induced hull girder sectional
loads on, for example, large container ships are

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

390

P T Pedersen and J J Jensen

based on common rules derived by the classification

societies; these rules only take into account the
length, breadth, block coefficient, and a flare
coefficient. The draft, hull girder flexibility, operational areas, weather routing, etc. cannot be explicitly considered. Therefore, the rules are not well
suited to give correct estimates of the most probable
largest sagging bending moments in ships with bow
flare; in addition full-scale measurements and
numerical simulations show that whipping may lead
to a 30 per cent increase of the sagging wave loading
[9]. Such important load effects are not presently
covered by the rules, so there is a need to develop
new design rules that adequately account for these
effects. The main motivation for the present paper
has been to contribute towards this goal. More
research is still needed, however, to obtain better
probabilistic description of the impulsive loads
caused by slamming.
ACKNOWLEDGEMENTS
The authors wish to thank MSc student Christian
Klimt Nielsen for performing the finite element
analysis of the model shown in Figs 5 and 6. The
authors also wish to thank the reviewers for their
very thorough comments.
F Authors 2009

8
9

10

11
12

13

14
15

16

REFERENCES
1 Nielsen, J. K., Hald, N. H., Michelsen, J., Nielsen,
U. D., Baatrup, J., Jensen, J. J., and Petersen, E. S.
SeaSense Real-time onboard decision support. In
Proceedings of the World Maritime Technology
Conference, London, 610 March 2006.
2 Belik, O., Bishop, R. E. D., and Price, W. G. On the
slamming response of ships to regular head waves.
Trans. RINA, 1980, 122, 325337.
3 Belik, O., Bishop, R. E. D., and Price, W. G.
Influence of bottom and flare slamming on
structural responses. Trans. RINA, 1988, 130,
261253.
4 Hirdaris, S. E., Price, W. G., and Temarel, P. Twoand three-dimensional modeling of a bulker in
regular waves. Mar. Structs, 2003, 16, 627658.
5 Malenica, S., Stumpf, E., Sireta, F.-X., and Chen,
A.-B. Consistent hydro-structure interface for evaluation of global structural responses in linear
seakeeping. In Proceedings of the ASME 27th
Conference on Offshore mechanics and arctic
engineering, Estoril, Portugal, 2008.
6 Iijima, K., Yao, T., and Moan, T. Structural
response of a ship in severe seas considering global

17

18
19

20

21

22

hydroelastic vibrations. Mar. Structs, 2008, 21,

420445.
MSC Napoli report: Load and strength assessments, 18 January 2007 incidents, Det Norske
Veritas report no. 2007-1928, DNV, Norway, 2007.
MSC Napoli ultimate strength analysis, report
NT: 2995/DR; ATA:1234A, Bureau Veritas, Marine
Division, 2007.
Jensen, J. J., Pedersen, P. T., Shi, B., Wang, S.,
Petricic, M., and Mansour, A. Wave induced
extreme hull girder loads on containerships. In
Proceedings of SNAME Annual Meeting, Houston,
October 2008.
Jensen, J. J. and Pedersen, P. T. Wave-induced
bending moments in ships a quadratic theory.
Trans. RINA, 1979, 121, 151165.
Jensen, J. J. and Dogliani, M. Wave-induced ship
hull vibration in stochastic seaways. Mar. Structs,
1996, 9, 353387.
Jensen, J. J. and Mansour, A. E. Estimation of ship
long-term wave-induced bending moment using
closed-form expressions. Trans. RINA, 2002, 144,
177191.
Jensen, J. J., Mansour, A. E., and Olsen, A. S.
Estimation of ship motions using closed-form
expressions. Ocean Engng, 2004, 31, 6185.
Winterstein, S. R. Nonlinear vibration models for
extremes and fatigue. ASCE, J. Engng Mech. Div.,
1988, 114(10), 17721790.
Jensen, J. J. Stochastic procedures for extreme
wave load predictions wave bending moment in
ships. Mar. Structs, 2009, 22, 194208.
Jensen, J. J. and Mansour, A. E. Estimation of the
effect of green water and bow flare slamming on
the wave-induced vertical bending moment using
closed-form expressions. Proceedings of the Third
International Conference on Hydroelasticity in
marine technology, (Ed R Eatock Taylor), 2003,
pp. 155161 (The Oxford University Press, Oxford).
Jensen, J. J. and Pedersen, P. T. Bending moments
and shear forces in ships sailing in irregular waves.
J. Ship Res., 1981, 24(4), 243251.
Zhao, R. and Faltinsen, O. Water entry of twodimensional bodies. J. Fluid Mechanics, 1993, 246,
593612.
Ramos, J., Incecik, A., and Guedes Soares, C.
Experimental study of slam-induced stresses in a
containership. Mar. Structs, 2000, 13(1), 2551.
Olsen, A. S., Schrter, C., and Jensen, J. J.
Encountered wave height distributions for ships
in the North Atlantic sea. In Proceedings of
nde, September 2004, pp.
PRADS04, Travemu
10431050.
Olsen, A. S., Schrter, C., and Jensen, J. J. Wave
height distribution observed by ships in the North
Atlantic. J. Ships Offshore Structs, 2005, 1(1), 112.
Okada, T., Takeda, Y., and Maeda, T. On board
measurement of stresses and deflections of a postPanamax containership and its feedback to rational
design. Mar. Structs, 2006, 19, 141172.

Proc. IMechE Vol. 223 Part M: J. Engineering for the Maritime Environment

JEME143