MODELLING OF PROPULSION SHAFT LINE AND SHAFTING ALIGNMENT...

Nenad VULI1
Ante ESTAN2
Vedrana CVITANI3

629.5.026
N. VULI, A. ESTAN, V. UDC
CVITANI

Modelling of Propulsion Shaft

Line and Basic Procedure of
Shafting Alignment Calculation
Original scientic paper
The main propulsion shafting is exposed to various operating conditions throughout the entire
lifetime of a modern ship. The necessary condition for the shafting to withstand and survive all
possible situations is its proper dimensioning and manufacture, as well as its assembly and testing
onboard. Its alignment is of utmost importance during the assembly process itself.
The aim of this paper is to present the shafting alignment calculation procedure in order to help
the designer to understand the whole alignment process. Calculation presumptions, modelling of
shafting parts, material properties and loading are given in detail. The advantages of the transfer
matrix methods over the nite element methods in this particular case have been described.
The important part is to establish the designed shafting elastic line onboard the ship, during the
outtting in the shipyard. It is proposed in the conclusion that the presented matter be included
into a future edition of the CRS Technical Rules.

Authors' addresses:
Croatian Register of Shipping,
Marasovieva 67, HR-21000 Split,
Croatia,
E-mail: nenad.vulic@fesb.hr
2
Ante estan, Faculty of Mechanical
Engineering and Naval Architecture,
University of Zagreb, I. Luia 5,
Zagreb, Croatia
E-mail: ante.sestan@fsb.hr
3
Vedrana Cvitani, FESB, University
of Split, Ruera Bokovia bb, HR21000 Split, Croatia
E-mail: vedrana.cvitanic@fesb.hr

Keywords: propulsion system, propulsion shafting, shaft alignment, transfer matrix method

Received (Primljeno): 2007-12-17

Accepted (Prihvaeno): 2008-02-11
Open for discussion (Otvoreno za
raspravu): 2009-30-09

Modeliranje i osnove prorauna centracije brodskog porivnog vratilnog

voda
Izvorni znanstveni rad
Tijekom ivotnog vijeka suvremenog broda, porivni vratilni vod izloen je vrlo promjenjivim
radnim stanjima. Osnovni su uvjeti da vratilni vod ispuni svoju funkciju u svim moguim radnim
uvjetima pravilno dimenzioniranje i izrada, kao i montaa i ispitivanje na brodu. U provoenju
montae posebnu vanost ima postupak centracije.
Cilj je ovoga rada prikazati metodologiju prorauna centracije vratilnog voda, sa svrhom da se
projektantima omogui lake razumijevanje cjelovitoga postupka centracije. Potanko su prikazane
proraunske pretpostavke, modeliranje dijelova vratila, kao i znaajke materijala i optereenja.
Opisana je prednost primjene metode poetnih parametara u matrinom prikazu (tzv. metode prijenosnih matrica) u odnosu na metodu konanih elemenata u ovom specinom sluaju. Naglaena
je vanost postizanja projektne elastine linije vratilnog voda, tijekom opremanja broda. U zakljuku
se predlae da se prikazani pristup ukljui u budua izdanja Tehnikih pravila HRB-a.
Kljune rijei: porivni sustav, porivni vratilni vod, centracija vratilnog voda, metoda prijenosnih
matrica

1 Introduction
The purpose of the propulsion machinery (main engine,
gearbox, propulsion shaft line, propeller and pertinent auxiliary
systems) is to propel the ship and to control manoeuvring, thus
enabling the navigator to be in control of the ships speed and
course. The main propulsion shaft line is the essential part of a
modern ship propulsion system, exposed to various conditions
throughout the ships lifetime. It has to function properly under
all possible operating conditions. Consequently, the shaft line
preliminary and final design, its static and dynamic behaviour
shall be carefully considered by the designer and by the classification society.
Shafting alignment procedure considers static and pseudostatic loading of the shafting in order to determine its static
response. This procedure consists of three phases: calculation,

assembly and validation of the assembled shaft line onboard

the ship. The main goal of this procedure is to determine and
ensure onboard achievement of the bearings designed positions
in athwart direction in order to comply with the loading criteria
for propulsion system and shafting parts. For this purpose the
shaft line is usually modelled as a continuous multi-span beam
on several supports. They may be modelled as absolutely stiff or
linearly elastic (in the case of static and pseudo-static response),
or even as real radial journal bearings (in the case of dynamic
response).
The goal of this paper is to provide designers with the basic
information how to model real shafting systems in order to
perform shaft alignment calculations. The paper aims to present
the conventional shafting alignment calculation procedure and
its presumptions. Modelling of shafting parts, material properties and loading is given in detail, in order to help the designer

59(2008)3, 223-227

223

N. VULI, A. ESTAN, V. CVITANI

MODELLING OF PROPULSION SHAFT LINE AND SHAFTING ALIGNMENT...

understand the whole calculation process. The advantage of

the transfer matrix method over the finite element method in
this particular case is briefly described. The important part is
to establish the designed shafting elastic line onboard the ship,
during the outfitting phase in the shipyard. The results of a real
life calculation example are presented in the end.

2 Shafting alignment calculation procedure

The shafting alignment calculation comprises evaluation of
the shafting elastic line and the reaction forces of supports for
the pre-determined offsets of supports. In case of propulsion
systems with gearboxes (mainly small, medium and high-speed
four-stroke diesel engines) the scope of the analysis is restricted to
the shaft line from the propeller to the output shaft of the gearbox,
together with its bearings and the bull gear. The remaining shaft
line parts (clutches, input shaft, as well as the engine itself) need
not be taken into account. A typical shaft line layout of this kind
is schematically shown in Figure 1.
In the case of directly coupled engines (mainly large slowspeed two-stroke diesel engines) the shafting alignment analysis
takes into account the model and the static behaviour of the engine
crankshaft. The complete crankshaft need not be modelled in
detail, as almost every slow-speed diesel engine manufacturer
provides drawings describing this model as a girder system available to the shaft line designers.

external concentrated force F in the centre of the cross section

of the left element end, [N];
external concentrated moment T in the centre of the cross
section of the left element end, [Nm];
external uniformly distributed load q along the element (owing to other possible forces, additional to the shaft self weight
and buoyancy), [N/m].
A general element model, together with the support at its right
end is shown in Figure 2.

Figure 2 General model of shaft line element [3]

Slika 2 Openiti model elementa vratilnog voda [3]

All the calculations are to be performed for the vertical plane,

where the influence of self-weight and buoyancy shall be taken
into consideration within the loading of the model. In the case
of propulsion systems with gearboxes, where gearing forces in
horizontal direction have a significant influence, the separate
calculations for the horizontal plane are also needed.
2.2 Calculation presumptions

Figure 1 Schematic of a typical marine shaft line including a

gearbox [1]
Slika 1 Shematski prikaz tipinog brodskog vratilnog voda s
reduktorom [1]

2.1 Input data and modelling of the system

The data describing dimensions, material and loading of the
shafts, together with the data describing the bearings concept
(slide or roller), bearing clearances and lubrication means are to
be available for shafting alignment calculations. This real system
is modelled as a statically indeterminate system of variable section beams with multiple supports. The shaft line elements are
modelled by means of circular section model elements, and the
shaft line bearings are modelled by means of absolutely stiff or
linearly elastic supports. In general, the cross-section varies from
one beam to another.
In general, model elements are of conical shape. A special
case of conical element is the element of cylindrical shape, as a
cone with equal diameters on both ends.
Elements are made of homogenous material, of specific
density , submerged (completely, partially, or not at all) into
sea-water of specific density w. The shaft material elastic properties are described by means of Young modulus of elasticity E
and shear modulus G.
As the calculation presumes the ship afloat, after assembling
all the parts of shaft line, loading of elements consists of:
self-weight of the element;
buoyancy in sea water (for submerged elements);

224

59(2008)3, 223-227

The calculations are based upon the real element dimensions,

and the following presumptions:
Propeller is completely or semi-submerged into water;
Volumetric forces (self-weights and buoyancy) are uniformly
distributed along each element;
All the bearings may be modelled by means of absolutely
rigid or linearly elastic supports;
The influence of shear forces and deformations is to be taken
into account;
The axial position of each reaction force is on the half way
of the bearing length.
If necessary, the inclination of shafting with respect to the
ship waterline (horizontal plane) may be taken into account by
calculating of components (for concentrated forces) and correction of gravity constant (for volumetric forces).
2.3 Selection of calculation method (FEM vs. transfermatrix method)
The most appropriate modelling and calculation procedures
in this case are the method of initial parameters in its matrix form
(the so called: transfer-matrix method) and finite element method
(FEM). Practically equivalent results may be obtained by means
of either of these two methods, except in the case of trapezoidal
loading along the element itself.
However, the transfer matrix method is chosen and preferred,
as it requires linear systems of significantly smaller ranges to be
solved. Particularly, FEM requires solving of 2m equations (where
m is the number of shaft line elements). On the other hand, the

MODELLING OF PROPULSION SHAFT LINE AND SHAFTING ALIGNMENT...

transfer-matrix method requires solving only of z+2 equations

(where z is the total number of stiff supports between the system
ends). In addition to this, the transfer matrix method is purely
analytical, implementing the solutions to differential equations
for beams in bending and shear.
There is an additional advantage of the transfer-matrix method
over FEM in this case. FEM calculation results (solutions) are
valid in the nodes only, whereas the transfer-matrix method allows the user to obtain deflections, slopes, bending moments and
shear forces along the element itself (i.e. between the nodes) on
the basis of calculated results in the nodes.
Consequently, the calculation model based on transfer matrices in a single (e.g. vertical) plane is chosen and described
further on.
2.4 Element transfer matrices and selection of initial
parameters
For calculation purposes the whole shaft line is modelled as a
system of multi-span beams, supported in rigid (absolute stiff) or
linearly elastic supports. Each beam has a uniform circular crosssection (solid or hollow). Conical shafting elements are modelled
as cylindrical with mid-section diameters, for the evaluation of
stiffness and loading by volumetric forces.
The basic goal of the transfer matrix method is to determine
the state vectors vi in each section of the whole system. It is
necessary to determine these vectors at each end section of each
element:

v i = [ wi

Mi

Qi 1]

(1)

Considering the system element (i), the state vector (vi+1)

at the right section of the element right end is related with the
state vector (vi) at the right section of the element left end as
follows:
)
v (i +right
= Li ,support v (i +left1 ) = Li ,support Li ,elem v i(left ) = Li v (i left ) (2)
1

In the equation (2) Li = Li,support Li,elem denotes the total transfer

matrix of the element i (including the support at its right end). It
may also be written in the expanded form (3):

Li =
0

0
0

2
i
i

1
0

2 EI i
i

EI i
1

3
i

6 EI i

i i
GAi

2
i

2 EI i
i

4
T F
q 2
q

i
i + i i + i i + i i Fi + i i
EI i 2
6
24 GAi
2
2

F
q

i Ti + i i + i i
EI i
2
6

qi 2i

Ti + Fi i +

( Fi + qi i + Ri +1 )

(3)

The quantities in the equations (1) to (3) have the following

meaning:
li element length, [mm]
EIi element bending stiffness, [Nm2]
GAi element shear stiffness, [N]
i = f (du/dv) shear form factor for the circular (solid or hollow)
section, i =1,11 1,45
qi uniform distributed external loading along the element (see
Figure 2), [N/m]

N. VULI, A. ESTAN, V. CVITANI

Fi concentrated force at the element left end (see Figure 2),

[N]
Ti concentrated moment at the element left end (see Figure
2), [Nm]
Ri+1 reaction of the support (if any), at the element right end,
positive downwards, [N].
w, displacement components (deflection, [m] and slope,
[m/m]),
M, Q internal forces (bending moment, [Nm] and shear force,
[N]).
Note: In case there is no support at the element right end, the
transfer matrix Li,elem = Li for the sole element is obtained from
(3), taking Ri+1=0.
The initial parameters to be selected are the two additional
unknowns at the whole system left end. They are finally determined from the two known parameters at the system right end,
together with the reactions in all rigid supports. The system ends
may either be free, propped, or fixed. Any case may be chosen,
however, the most common situation is that both of the ends are
free. In the case of free left end of the system the unknown initial
parameters are:
w1 deflection of the system left section;
1 slope of the system left section.
These parameters, together with all the reaction forces in rigid
supports (R1, R2, , Rz) are determined from the known boundary
conditions at the right end of the system, i.e.
Mm+1=0; Qm+1=0 in case of free right end;
wm+1=0; Mm+1=0 in case of propped right end;
wm+1=0; m+1=0 in case of fixed right end.
The total number of equations to be solved is thus z+2 only.
2.5 Calculation of inuence coefcients, initial reactions of supports and system response
The whole elastic system is described by means of the system
matrix A, and the system vector b. Both of them are assembled
on the basis of the boundary conditions at each fixed support and
at the rightmost end of the system, by means of span transfer
matrices. For each span these span transfer matrices are simply
matrix products of transfer matrices that relate the state vector
in the section of one stiff support (or system leftmost end) to
the next one:

v (jR+1) = L span, j v (jL ) = L k L k 1 L1 v (jL )

(4)

where:
k number of elements in the present span.
In case of both the left and right ends free, the vector of unknowns k consists of the two initial parameters (w0 and 0) and
of the reaction forces in stiff supports (Rj), as follows:

k = w0

R1

R2 Rz

(5)

In this particular case, the boundary conditions used to assemble matrix A and vector b are:

The zero displacements of the fixed supports (forming the

first z equations). This condition may also be expressed by

59(2008)3, 223-227

225

N. VULI, A. ESTAN, V. CVITANI

MODELLING OF PROPULSION SHAFT LINE AND SHAFTING ALIGNMENT...

the null-vector p0 of initial offsets of supports (p0=0).

Mm+1=0 and Qm+1=0, used to form the remaining 2 equations
(i.e. the 2 rows of A and the 2 components of b).
The best practice is to calculate the influence coefficients prior
to the calculation of components of k. That is the essential part of
the complete analysis. The influence coefficient hij quantitatively
expresses the change of reaction force (in N) in the movement
direction of the support i, when the support j moves for 1 mm
in that direction.
The matrix of influence coefficients H, which is independent
of the actual support offsets, is determined as:

H = A 1

(6)

The vector of unknowns k0, containing the initial reactions

in the stiff supports (i.e. those for zero support offsets) then
becomes:

k0 = H b0

(7)

Once the components of the vector k0 are known, the state

vectors in each section of the system may be easily found by a
simple matrix multiplication, beginning from the known state vector at the leftmost end of the system. This is the system bending
and shear response in terms of deflection, slope, bending moment
and shear force at both ends of each element.
2.6 Calculation of bearing reactions and the system
response for designed support offsets
Designed support offsets are to be determined in advance
so that the system response satisfies certain criteria for the final
calculated case. This final case may be the static response of
the assembled shaft line during the ship outfitting, or even the
pseudo-static response of the shaft line in operation. If the external
forces have not changed meanwhile, and transferring from the
system with zero support offset to the present one, the vector of
unknowns k will be:

k = H (p p0 ) + k0

(8)

Bending and shear response of the present system with the

designed support offsets is determined according to the same
procedure described in 2.4 for the system with zero support
offsets.
2.7 Design acceptance criteria and their verication
Detailed description of the design acceptance criteria would
be beyond the scope of this paper, so they will be just briefly
outlined here. These criteria are to be met for the pseudo-static
response of the shaft line in operation, both for cold and hot
working conditions, as follows [3]:

The stresses in shafts are to be below the prescribed permissible limits. This criterion may be applied either to the normal
stresses or the equivalent stresses.

Loading of the bearings is to be within prescribed limits. In

case of vertical plane calculations, bearing reactions are to be
directed upwards (to avoid overloading of the neighbouring
bearings) with the rule of thumb criterion for the minimal

226

59(2008)3, 223-227

reaction value as 20% of the left and right total load of the
span. Maximal reaction values shall not exceed the ones allowable by the specific pressure in the bearings, dependent
upon the bearing material in question.

Shaft line slope in the bearings is to be within allowable limits

dependent upon the bearing pre-selected clearances, to avoid
metal contact between the bearing and the shaft at the bearing
ends. Otherwise, slope boring of the bearings will be unavoidable. The rule of thumb states that the slope may spend up
to 50% of the bearing clearance. Some classification societies,
e.g. [4], prescribe that the relative slope between the propeller
shaft and the aftermost sterntube bearing is, in general, not
to exceed 0.3 mm/m in the static condition.

The shaft line shall not overload the gearbox itself, in case of
propulsion systems with gearboxes. The gearbox manufacturers usually prescribe the maximal allowable load transmitted
by the shaft line to the gearbox. In the absence of this data, the
rule of thumb will be to limit the difference in reaction forces
in the two bearings of the gearbox output shaft to maximum
20% of the weight of the bull gear.

The shaft line is not to overload the main propulsion engine

crankshaft or thrust shaft, in case of propulsion systems
with directly coupled main engines. As a rule, the engine
manufacturers usually prescribe the maximal allowable load
transmitted by the shaft line to the engine flange in terms of
shear forces and bending moments allowable range.
The stated design acceptance criteria shall be explicitly verified in the calculation phase, after all the results (system response
values) have been obtained.

3 Calculation example
The presented calculation procedure has been implemented
in the computer program MarShAl (Marine Shafting Alignment),
coded in MS Excel/VBA, dedicated to the presented calculations. For illustration a few characteristic diagrams obtained by
this computer code, which have been implemented and verified
on an inland navigation ship, are presented hereafter (Figure 3).
The propulsion system consists of a four-stroke engine (279 kW),
connected to the shafting by a reduction gearbox.

4 Conclusion
Shaft line is to be properly aligned in order to ensure its reliable functioning throughout the complete ship lifetime. Careful
calculation, as well as setting up of its results onboard (for the
ship afloat), as well as their validation during the assembly and
the testing phase is essential. This paper describes details of the
calculation procedure, promoting the advantages of (somewhat
forgotten) transfer matrix methods.
Details of the design acceptance criteria, a real model of
radial journal bearings and the review of classification society
requirements are beyond the scope of this paper. However, from
the authors long-term experience with this matter, it should be
important to enhance the existing technical rules requirements
of classification societies for shafting alignment to cover also the
case of small size ships. A proposal of this kind is expected to be
the matter of further work.
It is to be pointed out once again that this paper presents only
the basic information related to shaft alignment calculations, in

MODELLING OF PROPULSION SHAFT LINE AND SHAFTING ALIGNMENT...

N. VULI, A. ESTAN, V. CVITANI

order to help designers, shipbuilders, or

even engineering students understand the
essential calculation concepts. The authors
experience shows that it is important to
make such information, presented in simple
terms, widely available to the public. Comprehensive further information regarding the
complex subject of shaft alignment calculation, validation and criteria may be found
elsewhere in literature, e.g. [4], [5], [6] and
[7], together with the detailed information
about specialised and extremely powerful
software.

References
[1] KOZOUSEK, W.M., DAVIES, P. G.:
Analysis and Survey Procedures of
Propulsion Systems: Shafting Alignment, LR Technical Association,
Paper No. 5, London 2000.
[2] VULI, N.: Advanced Shafting
Alignment: Behaviour of Shafting in
Operation, Brodogradnja 52(2004)3,
p. 203-212.
[3] VULI, N.: Shaft Alignment Computer Calculation (in Croatian), Bulletin
Jugoregistar, No. 1, 1988, p. 1-13.
[4] : Rules and Regulations for the
Classification of Ships, Part 5, Chapter 8 Shaft Vibration and Alignment
(edition July 2006), Lloyds Register,
London, 2006.
[5] : Guidance Notes on Propulsion
Shafting Alignment, American Bureau of Shipping, Houston, 2004.
[6] : Guidelines on Shafting Alignment, Nippon Kaiji Kyokai ,Tokyo,
2006.
[7] BATRAK, Y.: Intellectual Maritime
Technologies - ShaftMaster Software,
Web page: www.shaftsoftware.com

Figure 3 Example of calculation results

Slika 3 Primjer dijagramskog prikaza rezultata prorauna

59(2008)3, 223-227

227

I. SENJANOVI, T. SENJANOVI, S. TOMAEVI, S. RUDAN

Ivo SENJANOVI
Tanja SENJANOVI
Stipe TOMAEVI
Smiljko RUDAN

UDC: 629.544:629.5.023.463
CONTRIBUTION OF TRANSVERSE
BULKHEADS...

Contribution of Transverse
Bulkheads to Hull Stiffness
of Large Container Ships
Preliminary communication
Ultra large container ships are rather exible and exposed to signicant wave deformations.
Therefore, the hydroelastic strength analysis is required for these types of ships. The coupling of
a beam structural model and a 3D hydrodynamic model is preferable for reasons of simplicity. In
this paper, the contribution of large number of transverse bulkheads to general hull stiffness is
analysed. The prismatic pontoon with the cross-section of a large container vessel is considered
for this purpose. The 3D FEM torsional analysis is performed with transverse bulkheads included
and excluded. The correlation analysis of the obtained deformations indicated the inuence of
transverse bulkheads on the ship hull stiffness. The analysis is done by employing the torsional
theory of thin-walled girders.
Keywords: container ship, nite element method, stiffness, thin-walled girder, torsion

Doprinos poprenih pregrada krutosti trupa velikih kontejnerskih brodova

Prethodno priopenje
Authors address:
Faculty of Mechanical Engineering
and Naval Architecture, University of
Zagreb,
Zagreb, Croatia
E-mail: ivo.senjanovic@fsb.hr
Received (Primljeno): 2007-04-04
Accepted (Prihvaeno): 2007-04-27
Open for discussion (Otvoreno za
raspravu): 2009-09-30

Vrlo veliki kontejnerski brodovi prilino su elastini i stoga podloni velikim valnim deformacijama. Zato se danas njihova vrstoa istrauje metodama hidroelastinosti. Pritom se radi
jednostavnosti spreu gredni strukturni model i 3D hidrodinamiki model. U ovom lanku istraen
je doprinos velikoga broja poprenih pregrada kontejnerskih brodova opoj krutosti trupa. Za
te potrebe razmatran je prizmatini ponton s poprenim presjekom velikoga kontejnerskog
broda. Konstruiran je 3D model konanih elemenata i provedena je analiza uvijanja s ukljuenim
i iskljuenim poprenim pregradama. Korelacijskom analizom deformacija za ova dva sluaja
ustanovljen je utjecaj poprenih pregrada na krutost brodskoga trupa. Pritom se koristila teorija
uvijanja tankostjenih nosaa.
Kljune rijei: kontejnerski brod, metoda konanih elemenata, krutost, tankostjeni nosa,
uvijanje

1 Introduction
Nowadays sea transport is rapidly increasing and ultra large
container ships are built [1]. Since they are rather flexible, their
hydroelastic response becomes an imperative subject of investigation. In the early design stage, the coupling of a FEM beam
structural model with a 3D hydrodynamic model based on the
radiation-diffraction theory is reasonable [2], [3].
The 1D FEM structural model is quite sophisticated since it
takes into account the bending and shear stiffness, as well as the
torsional and warping stiffness [4], [5]. The general hull stiffness
is increased due to the large number of transverse bulkheads in
holds. There are two types of bulkheads, i.e. ordinary watertight
bulkheads and grillage ones. The distance between them is determined by the container length.
Transverse bulkheads stretch within one web frame spacing
and are quite stiff. They can be directly included in the 1D FEM
model as a short beam element with a closed cross-section [6], [7].

228

59(2008)3, 228-238

However, due to the large number of transverse bulkheads and

to the model discontinuity, it is more practical and reasonable
to take into account their continuous contribution to the general
hull stiffness.
Different attempts to take the influence of transverse bulkheads into account have been made. One of the first approaches
was to increase the deck thickness based on the equivalence of
the deformation energy of transverse bulkhead girders and the
increased deck energy [8]. Today, the usual way is to model
transverse bulkheads by axial elastic springs at their joints to the
ship hull. The spring effect is condensed in lumped bimoments
[9]. Furthermore, in the case of a large number of transverse
bulkheads, the lumped bimoments might be distributed along
the hull girder [10]. The distributed bimoments are manifested
as additional torque load, which depends on the variation of
the twist angle as pure torsional torque. Therefore, only the
torsional stiffness of the ship hull is increased due to the bulkhead influence.

CONTRIBUTION OF TRANSVERSE BULKHEADS...

I. SENJANOVI, T. SENJANOVI, S. TOMAEVI, S. RUDAN

The effect of transverse structure on the deformation of

thin-walled girders is a challenging subject of contemporary
investigations [11], [12], [13]. Recent literature shows that the
problem is rather complex and the complicated solutions offered
there reduce the applicative advantages of the combined beam
theory and the thin-walled girder theory (1D + 2D) with respect
to the direct 3D FEM analysis. In any case, the reliability of the
1D + 2D theory has to be checked by the correlation analysis
with 3D FEM solutions.
In the light of the above circumstances, especially of the needs
of the ship hydroelasticity analyses, where results of dry natural
vibrations of the ship hull are required (modes, frequencies, modal stiffness, modal mass), a simpler solution is preferable and
more convenient. That was the motivation for the investigation
of this challenging problem. Thus, the 3D FEM analysis of the
prismatic hold structure with and without transverse bulkheads
is performed. The equivalence of the maximum twist angle in
the 1D and 3D models is used as a condition for determining the
change of torsional beam stiffness. The reliability of approach
is checked by the correlation for 1D and 3D warping functions
and stress distribution.

2 Outline of the thin-walled girder theory

The thin-walled girder torsional theory is developed under
assumptions that a considered structure is of membrane type (only
in-plane deformation occurs) and that there is no distortion of
the cross-section (twist angle is constant along the cross-section
contour).

where
E, G Youngs modulus and shear modulus
It, Iw torsional and warping modulus

twist angle
Substitution of (2) into (1) leads to the ordinary differential
equation of the fourth order

EI w

d4
d2

GI
= x .
t
d x4
d x2

(3)

Its solution reads

= A0 + A1 x + A2 ch x + A3 sh x + p ,

(4)

where

GI t
EI w

(5)

and Ai are integration constants, while p represents a particular

solution which depends on x.
Let us consider the twisting of the girder shown in Figure 1,
which is loaded by torque Mt at the ends, while x = 0. The warping of the girder ends is suspended. In this case the twist angle
is an anti-symmetric function and therefore A0 = A2 = 0. The
remaining constants A1 and A3 are determined by satisfying the
boundary conditions

x = l : T = Mt , u =

d
u = 0,
dx

(6)

where u is the warping function (axial displacement) and u is

the relative sectional warping due to the unit beam deformation,
d
/dx, defined according to the theory of thin-walled girders [15],
[16]. The final expressions for the twist angle reads

Mtl x
sh x

GI t l l ch l

(7)

Now, it is possible to determine sectional forces, i.e. pure

torsional and warping torques (2)

ch x
ch x
Tt = M t 1
, Tw = M t

ch l
ch l

Figure 1 Beam torsion

Slika 1 Uvijanje grede

(8)

and warping (sectorial) bimoment

A prismatic girder exposed to torsion is shown in Figure 1.

The equilibrium of sectional torque, T, and the distributed external
torsional load, x, yields

d T = x d x.

d
d3
EI w
,
dx
d x3

d 2
sh x
= Mt
.
2
dx
ch l

(2)

(9)

Furthermore, the warping function (6) takes the form of

(1)

According to the theory of thin-walled girders, the sectional

torque consists of a pure torsional part and a warping contribution [14]

T = Tt + Tw = GI t

Bw = EI w

u=

Mt
GI t

ch x
1 ch l u .

(10)

Torques Tt and Tw are the result of shear stresses t and w due

to pure torsion and suspended warping, respectively. The warping
bimoment Bw represents the work of axial normal stress on the
displacement u at a cross-section, i.e.

59(2008)3, 228-238

229

I. SENJANOVI, T. SENJANOVI, S. TOMAEVI, S. RUDAN

Bw = t u d s,

(11)

where

d 2
u.
d x2

=E

(12)

Thus, by substituting (12) into (11) one finds the expression

for warping modulus in (9)

I w = u 2 t d s.

(13)

3 Modelling of transverse bulkheads

The length of transverse bulkheads in large container ships
is equal to one web frame spacing. They are of grillage type and
therefore quite stiff. As a result, the bulkhead influence on the
hull warping reduction is significant.
In the torsional thin-walled girder theory, bulkheads can be
modelled by axial elastic foundation at their joint to the hull structure [9], [10]. In the case of a large number of bulkheads, the line
foundation can be spread to the area foundation of the hull shell.
The corresponding axial (tangential) surface load yields

q = u = u

d
,
dx

(14)

where is the spread bulkhead stiffness and u is the warping

function (6).
Axial load q causes an additional bimoment per unit length
on the relative sectional warping u

b = qu d s.

CONTRIBUTION OF TRANSVERSE BULKHEADS...

4 Effect of transverse bulkheads

In order to take the influence of bulkheads into account, another approach can also be applied. A ship hull consists of a large
number of open cross-section segments (holds) and of closed ones
(bulkheads). For the open section, the torsional modulus It is quite
small, and therefore the warping modulus Iw plays the main role.
In a short bulkhead area, the torsional modulus of closed section
I t0 is one order of magnitude higher than It, while I w0 is of the
same order as Iw. For the reason of simplicity we can consider a
uniform girder with the equivalent torsional modulus I t* , where
I t < I t* < I t0 , and the equivalent warping modulus I w* equal to
Iw. In this case, the differential equation (3) takes the form of

EI w

d 4
d 2
GI t*
= 0,
4
dx
d x2

(20)

GI t* = GI t + GI b .

(21)

where

GIb is the additional hull torsional stiffness due to bulkheads

as closed cross-section segments. Parameters k and GIb, in (19)
and (21) respectively, are equivalent quantities.
Instead of bulkhead modelling by equivalent axial elastic
foundation, as it is usually done in literature, it is possible to
determine the contribution of bulkheads by the 3D FEM analysis,
as it is elaborated in Section 7. Let us assume, for the time being,
that the end twist angles of a prismatic girder without and with
transverse bulkheads are known, (l) and *(l) respectively.
Referring to (7), one writes

(l ) =

(15)

* (l ) =

By substituting (14) into (15) one writes

b=k

d
,
dx

(16)

(22)

Mtl
th y*
1

,
GI t*
y*

(23)

where

y = l = l

where

k = u2 d s

(17)

y* = * l = l

is the sectional bulkhead stiffness.

According to the theory presented in [9] and [10], the bulkhead bimoment causes a distributed torque

b =

db
d2
,
=k
dx
d x2

(18)

where relation (16) is used for b. The torque b is the transformed

bulkhead load and has to be equilibrated by sectional torques Tt
and Tw (2). Thus, by substituting (18) into (3), the differential
equation for girder torsion with bulkhead influence is obtained:

d
d
( GI t + k ) 2 = 0.
4
dx
dx
4

EI w

(19)

Generally, the value of k has to be calculated for a given

bulkhead structure. It is a result of flexural bulkhead stiffness.

230

Mtl
th y
1

GI t
y

59(2008)3, 228-238

GI t
EI w

(24)

GI t*
.
EI w

(25)

Ratios of Eqs (23) and (22) lead to the transcendental equation

for determining the unknown parameter y*

1
y*2

th y* 1
=
1

y* y 2

th y * (l )

1
.

y (l )

(26)

Now, the new value of torsional modulus can be determined

by employing (24) and (25), i.e.
2

I t* y*
=
.
I t y

(27)

According to (21), the contribution of bulkheads to torsional

stiffness is
2

I b y*
=
1.
I t y

(28)

CONTRIBUTION OF TRANSVERSE BULKHEADS...

I. SENJANOVI, T. SENJANOVI, S. TOMAEVI, S. RUDAN

The twist angle (7) and the warping function (10) in nondimensional form read respectively:

GI t* * x sh * x
= *
Mtl
l y ch y*

(29)

GI u
ch x
= 1
.
Mt u
ch y*

(30)

*
t

Referring to (8), the twisting and warping torques take the

following form:

Tt*
ch * x
= 1
,
Mt
ch y*

Tw* ch * x
=
.
Mt
ch y*

(31)

An additional way to check the obtained results is to compare

the normal stress ratio at girder ends represented by the warping
bimoments (9)

Bw* (l ) y th y*
=
.
Bw (l ) y* th y

(36)

Actually, ratios (35) and (36) are related to the first and second
derivative of the twist angle, (6) and (9) respectively.

5 Ship particulars
A 7800 TEU container vessel of the following main particulars
is considered, Figure 2.

Furthermore, the twisting torque can be split into the hull part
and the bulkhead contribution

Tt* Th* Tb*

=
+
,
Mt Mt Mt

(32)

where, proportionally to their torsional moduli (21),

Th*
I T*
= t* t ,
M t It M t
*

Tb* I b Tt*
=
.
M t I t* M t

(33)

Ratios I t I t and I b I t are defined by (27) and (28).

Finally, the warping bimoment (9) takes the following nondimensional form:

sh * x
Bw*
= *
.
Mtl
y ch y*

(34)

The presented approach is based on the known ratio of end

values of the twist angle for a girder without and with transverse
bulkheads. Its reliability can be checked by the known ratio of
warping functions in the middle of the girder. According to (10),
it follows that

1
u (0 ) y
ch y*
.
= *
u (0 ) y 1 1
ch y
*

(35)
Figure 3 Midship cross-section
Slika 3 Glavno rebro

Figure 2 A 7800 TEU container vessel

Slika 2 Kontejnerski brod nosivosti 7800 TEU

59(2008)3, 228-238

231

I. SENJANOVI, T. SENJANOVI, S. TOMAEVI, S. RUDAN

Length overall
Length between perpendiculars
Breadth
Depth
Draught
Displacement

Loa = 334 m
Lpp = 319 m
B = 42.8 m
H = 24.6 m
T = 14.5 m
= 135530 t

CONTRIBUTION OF TRANSVERSE BULKHEADS...

The midship cross-section is shown in Figure 3, while

Figure 4 shows the transverse bulkhead. Properties of the open
cross-section are determined by the STIFF program [17].
Cross-section area
Horizontal shear area
Vertical shear area
Vertical position of neutral line
Vertical position of shear - torsional centre
Horizontal moment of inertia
Vertical moment of inertia
Torsional modulus
Warping modulus

A = 6.394 m2
Ash = 1.015 m2
Asv = 1.314 m2
zNL = 11.66 m
zD = 13.50 m
Ibh = 1899 m4
Ibv = 676 m4
It = 14.45 m4
Iw = 171400 m6

Position of deformation centre, zD, is rather low due to the

open cross-section. The relative warping of cross-section, u, is
illustrated in Figure 5. Youngs modulus, shear modulus and
Poissons ratio are: E = 2.06 108 kN/m2, G = 0.7923 108 kN/m2,
= 0.3, respectively.

6 FEM models of a hull segment

Figure 4 Transverse bulkhead

Slika 4 Poprena pregrada
Figure 5 Warping of cross-section, u
Slika 5 Vitoperenje poprenog presjeka, u

Figure 6 Prismatic FEM model of a hull part

Slika 6 Prizmatini model konanih elemenata dijela trupa

The front holds of the ship as a prismatic thin-walled girder

with the length of L = 2l = 174 m are considered. The FEM
model is generated by the software [18]. It is constructed of
four different types of superelements, and includes the total of
13 superelements, Figures 6 and 7. The shell finite elements are
used. The model is clamped at the fore end and the only warping is suspended at the aft end. The vertical distributed load
is imposed at the aft cross-section, generating the total torque
Mt = 40570 kNm, Figure 8, [19].
There are two types of transverse bulkheads within the ship
hold space, i.e. the ordinary watertight bulkheads and bulkheads
of grillage construction. Both types stretch within one web frame
spacing. The bulkhead top ends with the stool. Such a bulkhead
design makes them quite strong and therefore the general hull
stiffness is increased.

232

59(2008)3, 228-238

CONTRIBUTION OF TRANSVERSE BULKHEADS...

I. SENJANOVI, T. SENJANOVI, S. TOMAEVI, S. RUDAN

Figure 10 Superelement No. 3 with a transverse bulkhead

Slika 10 Superelement br. 3 s poprenom pregradom
Figure 7 Superelement No. 3
Slika 7 Superelement br. 3

The FEM model of the ship segment with transverse bulkheads and a typical superelement with the watertight bulkhead are
shown in Figures 9 and 10 respectively. The boundary conditions
and imposed load are the same as in the case of prismatic model
without transverse bulkheads.

Figure 8 Load at the aft end

Slika 8 Optereenje kraja modela
Figure 9 Prismatic FEM model of a hull part with transverse
bulkheads
Slika 9 Prizmatini model konanih elemenata dijela trupa s
poprenim pregradama

Figure 11 Deformation and stresses of the model without transverse bulkheads, x [N/mm2]
Slika 11 Deformacije i naprezanja modela bez poprenih pregrada, x [N/mm2]
Figure 12 Deformation and stresses of the model with transverse
bulkheads, x [N/mm2]
Slika 12 Deformacije i naprezanja modela s poprenim pregradama, x [N/mm2]

59(2008)3, 228-238

233

I. SENJANOVI, T. SENJANOVI, S. TOMAEVI, S. RUDAN

Deformed models, without and with transverse bulkheads,

are shown in Figures 11 and 12 respectively. Distortion of the
cross-section is negligible as a result of a double skin cross-section with very strong web frames. Due to the same reason, the
bending stresses are negligible in comparison to the membrane
stresses; therefore, the structure behaves as a membrane one. Different colours in Figures 11 and 12 denote the levels of von Mises
membrane stress. High stress concentration in the hatch coaming
and the upper deck at the model ends confirms the well-known fact
caused by the suspended warping of the cross-section, [20].

7 Inuence of transverse bulkheads

Since the 3D FEM model behaves as a membrane structure
without distortion of the cross-section, the obtained results are
comparable to those of the 1D analysis.

CONTRIBUTION OF TRANSVERSE BULKHEADS...

3 D 0.27690 10 3
=
= 1.2055.
1D 0.22969 10 3

(a)

Vertical position of the deformation centre in the 3D model

is above that of the 1D model, points D3D and D1D in Figure 13,
respectively. The influence of transverse bulkheads on the position
of the deformation centre is quite weak, point D3D. The warping
of cross-section determined by the 3D analysis is rather close to
that of the 1D analysis, Figure 5. Therefore, the warping correlation could be done only for one representative point of extreme
displacement value. Let us chose the joint of the bilge and the
inner bottom, Figure 5, where

u 3 D 1.02594 mm
=
= 0.9753.
u1D 1.05192 mm

(b)

Thus, the correlation of 3D and 1D analyses results is quite

good concerning warping, while the 3D FEM model is more
elastic than the 1D model from the twisting point of view. This
could be caused by the shear influence on torsion which is not
taken into account in the beam analysis, [10]. That fact might be
the subject of further investigations. However, it does not have
a significant influence on the relative bulkhead contribution to
the hull stiffness.
The twist angle ratio of the model with and without transverse
bulkheads reads

3*D 0.24876 10 3
=
= 0.89837.
3 D 0.27690 10 3

(c)

The warping ratio in the bilge point is

u 3*D 0.90013 mm
=
= 0.87737.
u 3 D 1.02594 mm

(d)

The axial normal stress ratio in the hatch coaming, Figures

11 and 12, yields

3*D 5.93623 N / mm 2
=
= 0.93365.
3 D 6.35805 N / mm 2

(e)

In the considered numerical example, according to (24),

y = 0.49541, while the solution of Eq. (26) gives y* = 0.7464.
The variation of torsional stiffness, Eq. (27), is I t* I t = 2.27 . It
means that the bulkhead contribution is I b I t = 1.27 .
Figure 14 Twist angle
Slika 14 Kut uvijanja

Figure 13 Rotation of the cross-section

Slika 13 Zakret poprenog presjeka

Rotation of the model free cross-section, determined analytically by the beam theory (1D) and numerically by FEM (3D) for
the pontoon with and without transverse bulkheads, is shown in
Figure 13. The twist angle of the 3D analysis is somewhat higher
than that of the 1D analysis.

234

59(2008)3, 228-238

CONTRIBUTION OF TRANSVERSE BULKHEADS...

I. SENJANOVI, T. SENJANOVI, S. TOMAEVI, S. RUDAN

u * (0 )
= 0.89388.
u (0 )

(f)

Discrepancy between the 1D analysis and the 3D FEM analysis, value (d), is only 1.9%.
The warping bimoment shown in Figure 17 is also reduced
due to bulkheads. The 1D ratio, Eq. (36), yields

Bw* (l )
= 0.91633.
Bw (l )

Figure 15 Deck warping

Slika 15 Vitoperenje palube

(g)

By comparing it to the 3D FEM stress ratio (e), a discrepancy

of -1.9% is obtained. This fact confirms quite good simulation of
the bulkhead effect in the thin-walled girder theory.
The torques distributions are shown in Figure 16. The hull
twisting torque in the case with bulkheads, Th, is very close to
the pure twisting torque without bulkheads, Tt. The warping
torque Tw is now reduced in comparison to Tw due to the bulkhead
contribution, Tb.
The influence of the increased value of torsional stiffness on
vibrations can, for instance, be analysed in the case of uncoupled
natural vibration of a free thin-walled girder with suspended
boundary cross-section warping. The corresponding formula for
natural frequencies derived in Appendix reads, (A16)

n =

n
L

GI t
n EI w
1+
, n = 0, 1, 2 ...
0
L GI t
Jt

(37)

The following relation between natural frequencies of a hull

segment with and without transverse bulkheads exists:

Figure 16 Twisting and warping torques

Slika 16 Torzijski momenti uvijanja i vitoperenja

n*
=
n

2 y*
1+
n

2y
1+
n

(38)

For the first natural frequencies of elastic modes one finds

1* 1 = 1.05594. Thus, a 127% torsional stiffness increase due

to bulkheads results in a 5.6% increase in the first frequency [M3]

in the considered case. It is evident from (38) that the variation
of higher mode natural frequencies is decreased.

8 Bending stiffness analysis

8.1 Horizontal bending

Figure 17 Warping bimoment

Slika 17 Bimoment vitoperenja

The girder displacements and sectional forces are determined

for both cases, i.e. without and with transverse bulkheads, and
are shown in Figures 14, 15, 16 and 17 in non-dimensional
form. The corresponding formulae from Sections 2 and 4 are
used. The twist angle is reduced according to given values of
the 3D FEM analysis, Figure 14. The warping of the cross-section u is also reduced, Figure 15. Its variation defined by the 1D
analysis, Eq. (35) is

Horizontal bending is analysed by the FEM model adapted

for this purpose. The model aft end is entirely free and loaded
by distributed loads, as shown in Figure 18. The vertical load
generates a torque of Mt = 40570 kNm, while the total horizontal
force Fy, acting about the deformation centre, equilibrates it. In
this way, the girder is only exposed to horizontal bending.
In the cases of the model without and with transverse
bulkheads, the horizontal force of pure bending takes values
of Fy = 1500 kN and Fy* = 1565 kN, respectively. The corresponding maximum deflections yield y = 19.8654 mm and
y* = 20.4069 mm, Figures 19 and 20. Thus, the moment of inertia
of the cross-section of the reinforced model can be expressed by
that of the model without bulkheads:

59(2008)3, 228-238

235

I. SENJANOVI, T. SENJANOVI, S. TOMAEVI, S. RUDAN

CONTRIBUTION OF TRANSVERSE BULKHEADS...

I z* =

y / Fy
I = 1.01565 I z .
y* / Fy* z

(h)

The correction is rather small, approximately 1.56% and its

influence on vibration is almost negligible. Stress concentration
in the bilge area at the fixed model end due to bending is evident,
Figures 19 and 20.
Since pure torque Mt and horizontal forces Fy and Fy* for the
model without and with transverse bulkheads are known, it is
possible to determine the vertical position of the deformation
centre:

zD =
Figure 18 Load at the model free aft end in the case of horizontal
bending
Slika 18 Optereenje horizontalnog savijanja na slobodnom
stranjem kraju modela

h Mt
h M

, zD* = *t ,
2 Fy
2 Fy

(i)

where h is the double bottom height. The obtained results are

compared with those of pure torsion determined in Section 7,
in Table 1. The 3D FEM analyses show that the torsional centre
and the shear centre are not the same points. In the thin-walled
girder theory these two centres are not distinguished, and the
unique deformation centre is determined. Probably, the suspended
warping in the 3D FEM torsional analysis has some influence on
the vertical position of the torsional centre. We can see that the
transverse bulkheads also influence the position of the torsional
and shear centres.
Table 1 Vertical position of the deformation centre, zD [m]
Tablica 1 Vertikalni poloaj sredita deformacije, zD [m]

Figure 19 Horizontal bending of the model without transverse

bulkheads, vM [N/mm2]
Slika 19 Horizontalno savijanje modela bez poprenih pregrada,
vM [N/mm2]
Figure 20 Horizontal bending of the model with transverse bulkheads, vM [N/mm2]
Slika 20 Horizontalno savijanje modela s poprenim pregradama, vM [N/mm2]

Torsional centre,
3D FEM twisting
Shear centre,
3D FEM bending
Deformation centre,
2D strip theory

Without bulkheads

With bulkheads

-10.60

-10.25

-12.52

-11.96

-13.50

8.2 Vertical bending

A similar FEM analysis is performed for the investigation of
vertical bending stiffness. The total vertical force, imposed at the
free model end, is Fv = 2000 kN. The corresponding deflection at
the same place, in the case of the model without and with transverse bulkheads, yields z = -28.3706 mm and z* = -28.1697 mm,
respectively. Thus, for the corrected vertical moment of inertia
of the cross-section one finds:

I y* =

z
I = 1.00713 I y .
z* y

(j)

It is obvious that the influence of transverse bulkheads on the

vertical stiffness is even lower than on the horizontal stiffness.

9 Conclusion
Hydroelastic analysis of large container vessels becomes an
actual problem. For the reason of simplicity, a beam model of
hull girder is coupled with a 3D hydrodynamic model. Instead
of calculating the transverse bulkhead stiffness, the contribution
of bulkheads to the global stiffness of the ship hull is determined
by the 3D FEM analysis of a prismatic ship-like pontoon. This

236

59(2008)3, 228-238

CONTRIBUTION OF TRANSVERSE BULKHEADS...

I. SENJANOVI, T. SENJANOVI, S. TOMAEVI, S. RUDAN

is a simple and reliable engineering approach. It was found that

the increase in torsional stiffness is considerable in the illustrated
numerical example. The influence of this fact on the resonant ship
hull response to wave excitation is significant and therefore has
to be taken into account. On the other hand, the influence of the
transverse bulkheads on vertical and horizontal bending stiffness
is rather small and may be neglected.
However, some discrepancies between the thin-walled girder
theory and the 3D FEM still exist. In the analysed numerical
example of a ship hull segment, the twist angle determined by
the beam analysis is significantly smaller than that obtained by
the 3D FEM analysis. Even the twist angle of the beam without
bulkheads is still lower than that of the 3D FEM model reinforced
by bulkheads. Also, there are some discrepancies of the shear
centre position between the 1D and 3D models without bulkheads. On the other hand, agreement between the cross-section
warping is excellent. This problem will be the subject of further
investigation.
Most of present papers dealing with problems of thin-walled
structures are concentrated on the investigation within the thinwalled girder theory. The validation of results should be based
on the correlation analysis with 3D FEM models which simulate the structure behaviour in a more realistic way. Also, some
model tests and full scale measurements are very valuable for
this purpose.

[11] BOSWELL, L. F., LI, Q.: Consideration of relationship between

torsion, distortion and warping of thin-walled beams, International
Journal of Solids and Structures 21(1995), p. 147-161.
[12] KIM, J. H., KIM, Y. Y.: Thin-walled multi-cell beam analysis for
coupled torsion, distortion and warping deformation, ASME Journal
of Applied Mechanics 68(2001), p. 260-269.
[13] RENDEK, S., BLA, I.: Distortion of thin-walled beams, ThinWalled Structures 42(2004), p. 255-277.
[14] SENJANOVI, I., FAN, Y.: A higher-order torsional beam theory,
Engineering Modelling, 10(1997)1-4, p. 25-40.
[15] SENJANOVI, I., FAN, Y.: A higher-order theory of thin-walled
girders with application to ship structures, Computers & Structures,
43(1992)1, p. 31-52,
[16] SENJANOVI, I., FAN, Y.: A finite element formulation of ship crosssectional stiffness parameters, Brodogradnja, 41(1993)1, p. 27-36.
[17] ...: STIFF, Users Manual, FSB, Zagreb, 1990.
[18] ...: SESAM, Users Manual, Det Norske Veritas, Hvik, 2003.
[19] SENJANOVI, T.: Utjecaj skladine konstrukcije na krutost trupa
kontejnerskih brodova, Diplomski rad, FSB, Zagreb, 2007.
[20] SENJANOVI, I., FAN, Y.: Pontoon torsional strength analysis
related to ships with large deck openings, Journal of Ship Research,
35(1991)4, p. 339-351.
[21] SENJANOVI, I., ATIPOVI, I., TOMAEVI, S.: Coupled flexural and torsional vibrations of ship-like structures, (in preparation).

Acknowledgement

The differential equation of uncoupled torsional beam vibrations can be written as an extension of the static equation (3),
[4], [21]

The authors would like to express their gratitude to Prof.

Radoslav Pavazza from the Faculty of Electroengineering,
Mechanical Engineering and Naval Architecture, University of
Split, for his useful consulting during the investigation of this
challenging problem.

References
[1]

Proceedings of International Conference on Design & Operation of

Container Ships, RINA, London, 2006.
[2] MALENICA, ., SENJANOVI, I., TOMAEVI, S., STUMPF,
E.: Some aspects of hydroelastic issues in the design of ultra large
container ships, The 22nd International Workshop on Water Waves and
Floating Bodies, IWWWFB, Plitvice Lakes, Croatia, 2007.
[3] TOMAEVI, S.: Hydroelastic model of dynamic response of container
ships in waves, Ph. D. Thesis, FSB, Zagreb, 2007. (in Croatian).
[4] SENJANOVI, I.: Ship Vibrations, 2nd Part, University of Zagreb,
1990.
[5] SENJANOVI, I., GRUBII, R.: Coupled horizontal and torsional
vibration of a ship hull with large hatch openings, Computers &
Structures, 41(1991)2, p. 213-226.
[6] HASLUM, K., TONNESSEN, A.: An analysis of torsion in ship hulls,
European Shipbuilding, No. 5/6 (1972), p.67-89.
[7] PEDERSEN, P. T.: Torsional response of containerships, Journal of
Ship Research, 29(1985), p.194-205.
[8] SENJANOVI, I.: A solution to the torsion problem on container ships,
1st and 2nd Parts, Brodogradnja, No. 1 and 2, 1972, (in Croatian).
[9] PAVAZZA, R., PLAZIBAT, B., MATOKOVI, A.: Idealisation of
ships with large hatch openings by a thin-walled rod of open section
on many elastic supports, Thin-Walled Structures 32 (1998), p. 305325.
[10] PAVAZZA, R.: Bending and torsion of thin-walled beams of open
section on elastic foundation, Ph.D. Thesis, University of Zagreb,
1991.

Appendix
Torsional beam vibrations

EI w

2
4
2
0
GI
J

+
= x (t ),
t
t
x4
x2
t2

(A1)

where the twist angle and the distributed torque x are time
dependent quantities. The symbol J t0 denotes the polar mass
moment of inertia. Natural vibrations are harmonic and Eq. (A1)
is reduced to the homogeneous form

EI w

d4
d2
GI t
2 J t0 = 0,
4
dx
d x2

(A2)

where and are the natural mode and the natural frequency,
respectively.
Solution of (A2) is assumed in exponential form

= e x .

(A3)

By substituting (A3) into (A2) one finds the following biquadratic characteristic equation:

GI t 2 2 J t0

= 0.
EI w
EI w

(A4)

Its four roots yield

j = , i,

(A5)

where

GI t
2 EI w

1 + 4 2

J t0 EI w
+1
(GI t )2

59(2008)3, 228-238

(A6)

237

I. SENJANOVI, T. SENJANOVI, S. TOMAEVI, S. RUDAN

GI t
2 EI w

1 + 4 2

J t0 EI w
1.
(GI t )2

(A7)

CONTRIBUTION OF TRANSVERSE BULKHEADS...

For a nontrivial solution, determinants of (A11) and (A12)

have to be zero. That leads to the frequency equations

Thus, the solution of (A2) takes the following form:

= A1 sh x + A2 ch x + A3 sin x + A4 cos x.(A8)

Let us consider vibrations of a free beam with suspended
warping at its ends. The corresponding boundary conditions
read

x = l : T = 0, u = 0

(A9)

that leads to

d
x = l :
= 0,
dx

d3
= 0.
d x3

(A10)

In the case of symmetric modes, A1 = A3 = 0, while for antisymmetric modes A2 = A4 = 0. The corresponding eigenvalue
problems yield

238

sh l sin l A2
3 sh l 3 sin l A = {0}

(A11)

cos l A1
ch l
3 ch l 3 cos l A = {0} .

(A12)

59(2008)3, 228-238

( 2 + 2 ) sh l sin l = 0

(A13)

( 2 + 2 ) ch l cos l = 0

(A14)

with the same eigenvalue formula for the symmetric (n = 0, 2)

and anti-symmetric (n = 1, 3) modes

l =

n
, n = 0, 1, 2 ...
2

(A15)

Taking (A15) into account, one finds the following expression

for natural frequencies of torsional vibrations from (A7)

n =

n
2l

GI t
n EI w
, n = 0, 1, 2 ... (A16)
1+
2l GI t
J t0

Integration constants A2 and A4, and A1 and A3 are determined

from (A11) and (A12), respectively. Symmetric and anti-symmetric natural modes according to (A8) yield

n = n sin n l ch n x + n sh n l cos n x, n = 0, 2 ... (A17)

n = n cos n l sh n x n ch n l sin n x, n = 1, 3 ... (A18)

In case n = 0, the natural frequency 0 =0, Eq. (A16), and
the natural mode 0 = 1, Eq. (A17), that is related to the rigid
body rotation.