Você está na página 1de 8

A method for predicting the influence of an additive bulb on ship

resistance.
G.D. Tzabiras
School of Naval Architecture and Marine Engineering
National Technical University of Athens
9 Heroon Polytechniou str., Zografos 15773, Athens, Greece

ABSTRACT: The present work is concerned with the numerical prediction of the ship resistance by combining two CFD approaches: an inviscid code and a RANS solver. The main purpose is to decide effectively and rapidly about optimized bulb configurations for performing
model experiments. A potential flow method has been developed to solve the non-linear freesurface problem that calculates the geometry of the free-surface and, in a first approximation,
the wave resistance. Then, by applying the RANS code beneath the predetermined surface, the
total resistance is calculated. The method is applied for a model of a passenger-ferry that has
been tested in the towing tank of NTUA. In order to accelerate the whole procedure in the design stage, the effect of an additive bulb, that is equivalent to the original implicit, is studied.
1 INTRODUCTION
The hydrodynamic design of bulbs, being of major importance for many types of ships, is a frequent activity in towing tanks. However, the selection of the proper bulb is not a trivial task,
since it is related to the complicated flow phenomena around the bow that are more or less difficult to predict. Usually, such a design is based on previous experience or on systematic series of
experiments, e.g. Kracht (1978), but there are many cases where both practices may fail. Therefore, new configurations have to be tested in order to achieve an effective reduction of the ship
resistance. In these cases, the drawing and the manufacturing of various bulbs that are adjusted
to the original hull is not only expensive but also time-consuming.
An alternative, promising way to accelerate the whole procedure and reduce the required cost
is to apply CFD tools that may provide useful information about the effectiveness of a new bulb
design. The aim of the present work is to investigate the possibility of relevant methods to predict the hydrodynamic behavior of such configurations. Although the application of advanced
RANS methods that solve the real free-surface problem are still very demanding, it is noticeable
that the flow around the bow can be effectively simulated regarding the fluid as inviscid (unless
wave-breaking or spray occur). In this respect, a potential flow code has been developed to calculate the free-surface geometry and next, an existing RANS solver is applied to calculate the resistance. This is not a new concept and various methods have been developed to face the problem, each one with its own advantages and shortcomings. In addition, a fast representation
(based on the conformal mapping technique) is applied to generate an additive bulb with the
purpose of comparing its behavior with an implicit one having the same general characteristics.
2. THE NUMERICAL METHOD
2.1 The potential flow solver
The numerical method that has been developed to solve the potential flow around ships is based
on an iterative procedure which, basically, decouples the two free-surface boundary conditions.

Unlike most of the non-linear codes (e.g. Janson, 1997, Raven, 1993) which employ potential
formulations that involve implicitly these conditions, primitive variables are introduced here
that is, a velocity-pressure coupling solution is applied. To solve for the kinematic boundary
value problem (Laplace equation) the ship hull and the real free-surface are covered with quadrilateral panels, Figure 1.

Figure1. Free-surface panels.


Two co-ordinate systems (x, y, z) are introduced, i.e. the ship reference system which is employed to construct the panels and the absolute system with z=0 on the undisturbed free-surface
where all the flow equations refer to. The co-ordinates are transformed successively between the
two systems taking also into account the possible sinkage and trim corrections. Both the kinematic and the dynamic boundary conditions are satisfied on the free-surface at the end of the iterative algorithm which follows the three main steps:
1. For a known free surface geometry solve the Laplace problem which satisfies the kinematic condition on both the hull and the free-surface
2. Solve the vertical momentum equation by introducing the normal pressure gradient on
the free-surface which is calculated from step 1 using the Bernoulli equation and calculate the corresponding velocity component.
3. Update the free surface by a two-step Lagrangian-Eulerian procedure using the vertical
velocity components calculated in step 2.
The above steps are repeated until convergence, which is effectively obtained when the dynamic
boundary condition is satisfied under a specified criterion.
The first step follows the classical methodology of Hess & Smith (1968), that is a constant
source distribution is assumed on each quadrilateral panel. This approach has also been used in
previous works that adopt different formulations. The unknown sources have to satisfy the kinematic condition, i.e, the normal velocity at the null point of each panel equals zero. They
are calculated by solving the corresponding linear system by the Gauss-Seidel iterative algorithm which is quite efficient, since the relevant matrix is diagonally dominant.
After the calculation of sources, the velocity components (ux, uy, uz ) are calculated on the null
points of the surface elements and the total pressure p* is derived from the Bernoulli equation.
The total pressure is the sum of the static pS and the hydrostatic term gz. Obviously, for an arbitrary surface, pS is different than the ambient pressure (which is assumed equal to zero). This
difference is introduced as a source term to calculate the vertical velocity uz* on the free surface
by solving the corresponding inviscid momentum equation:

u xu z u y u z u z2
p

y
z
z
x

(1)

Equation (1) is solved numerically after integrating on the surface control volumes, as shown
in Figure 2. The result of this integration is a set of non-linear algebraic equations corresponding
to the surface panels having the form

APu zP AE u zE
AW uzW
AU u zU AD uzD
( x y )( p*P gz P )

(2)

Figure 2. Surface control volume


The vertical side z* of the control volume in Figure 2 is an arbitrary parameter that controls
the convergence of the procedure. It appears in the convective terms Ai of equation (2), but essentially determines the influence of the pressure gradient. By performing numerical experiments on various ship forms and Froude numbers, it has been found that efficient values for z*
should be of the order (xy). The convective coefficients can be approximated by any higher
order upstream difference scheme. However, in the present investigation, the zero-order upstream scheme is applied in the stream-wise UD direction, in order to avoid as far as possible
downstream reflections. Then, only one sweep of the computational domain is needed to solve
for the vertical velocity component, but fine discretization in the longitudinal direction is required to achieve satisfactory results for the free-surface geometry. In addition, to avoid numerical diffusion in the lateral direction, an alternative option can also bee applied which is
based on the integration along local flow-lines. The latter can be easily determined around the
surface control points using the potential flow results. In this case, equation (2) reduces to the
simple form:

APu zP AU u zU
( x )( p*P gz P )

(3)

where uzU* is calculated by linear interpolation on the upstream face of the control volume
along the local streamline passing through point P.
After the vertical velocity components have been calculated, the new free surface is computed by performing two corrective steps as described by Tzabiras (2004). First, following the
local flow-lines, points on a transverse surface line define corresponding points on the next,
downstream line and a new transverse profile is generated. Next, this new cut is corrected so
that the flow rate through the corresponding surface panel vanishes. This way, the surface flowlines associate the dynamic with the kinematic condition.
The application of the method to multiple-hull forms or transom sterns (without dead water) is straightforward. Besides, since accuracy depends strongly on the number of panels, to reduce the computational effort the solution may start with a coarse grid which is successively refined to the maximum number of panels that are defined in the input data. Except the
geometrical interpolations, there is no difficulty to pass from the one grid resolution to another
due to the decomposition which is followed. The wave resistance as well as vertical forces and
moments are calculated by integrating the static pressures on the hull panels. The latter are used
to calculate the dynamic sinkage and trim whenever this is required and the center of gravity is
known. This procedure is carried out by changing sinkage and trim values after a certain number
of iterations is completed, until final convergence is achieved.

2.2 Panel construction


The number of panels that are required to make accurate predictions for the free-surface geometry and the wave resistance component depends on the examined ship form as well as on the
Froude number. Low Froude number flows or vessels with bulbous bows at high speeds require
fine longitudinal and transverse discretization which needs a fast method for generating the panels. The conformal mapping technique, e.g. Tzabiras (1997), has been employed in the present
investigation which transforms ship or bulbous bow sections onto the unit circle through the
general transformation:
N

z a1 an 1n

(5)

n 2

In transformation (5) z represents the complex plane on the unit circle, the arbitrary hull or
bulb section, while coefficients an are calculated iteratively for an initial set of data sections.
The intermediate sections that are needed for fine longitudinal grids are generated straight away
by calculating the corresponding coefficients applying cubic interpolation. Then, the panels on
the hull or the bulb are directly constructed by finding the intersection of the free surface with
the analytical representation of the sections and using any desired distribution of grid points
girth-wise (equal arcs or equal angles on the circle plane). The points on a transverse cut of the
updated free surface are found by interpolation following an exponential arrangement.
The geometry of an additive bulb can be defined by introducing the six geometrical parameters of Figures 3a, b, e.g. Kracht (1978), Tzabiras (1997). Three of them are linear, i.e. the
length LB of the bulb up to FP, its maximum breadth BB and the vertical distance ZB of the leading edge from the base line. The remaining are: the transverse area ABT of the bulb section at FP,
the area ALB of the longitudinal projection of the bulb contour and the volume of the bulb VB.
Evidently, there are many ways to design a bulb which satisfies these parameters. In the present
work the geometry of an additive bulb is produced using a on a three-parametric Lewis representation defined by the two principal dimensions and the area of any section, longitudinal or
transverse (Tzabiras, 1997). Assuming a subdivision of ALB in the upper (U) and the lower (D)
regions, the longitudinal bulb profile can be directly generated. The same procedure holds for
the basic section ABT. The other transverse sections along the bulb vary proportionally to their
principal dimensions T(x)B(x) where B(x) is calculated by an iterative Newton-Raphson method
so that the total volume of the bulb equals VB. To fit the additive bulb after FP it is assumed that
its basic section remains constant downstream. Then, the panel construction on the hull is based
simply on keeping the maximum breadth of the bulb or the hull, at a certain depth. To obtain a
smooth distribution of points girth-wise, the intersection of the bulb with the corresponding hull
section is firstly defined and, then, points are redistributed in both directions. This procedure is
shown in Figures 4a, b where an original hull is transformed by applying an additive bulb.

Figure 3a,b. Left (a): transverse section at FP of an additive bulb


. Right (b): Longitudinal profile

Figure 4a,b: Left (a): original bow and additive bulb


Right (b): Transformed bow lines
2.4 The RANS solver
The RANS multi-block code for ship flows developed at NTUA (e.g. Tzabiras, 2004) is used to
solve the viscous flow under the specified free surface. The numerical solution is based on the
finite volume approach, employs curvilinear co-ordinate systems and staggered meshes and
solves the velocity and the pressure fields by a pressure correction technique. Two turbulence
models can be applied that are based on k-, and k--SST variants. The code may also be applied to solve independently the free surface flow following a surface-tracking process.
3 TEST CASES
In order to test the developed potential flow method, a first application for the well documented
Series-60, cB=0.6 hull is presented in the sequel. Computations were carried out for two Froude
numbers, i.e. Fr=0.25 and 0.35. In both cases the grid was refined sequentially every 100 steps,
starting from 5000 panels to 25000 panels at the finest mesh. The computation domain extended
from L PP to 3L PP in the longitudinal dimension and 1.5LPP on both sides of the model, where
LPP=3.048m and x=0 coincides with FP. The calculated results for the wave profile on the hull
are compared to the experimental data by Garofallidis (1996) in Figures 5a, b. In general, the
calculated profile follows the experimental one, except the region close to AP, where the potential theory predicts higher values. The convergence history of the wave resistance coefficient CW
is presented in Figure 6, showing that after 300 steps CW has practically converged, while the
procedure ended in 450 iterations where the dynamic condition is satisfied. The results for CW,
or (CR), presented in Tables 1&2, show that the potential flow predicts lower values than the
measured, which is an expected result since the viscous effects are ignored.

Figure 5a, b. Left (a): Comparison of calculated vs. experimental wave profiles at Fr=0.25
Right (a): Comparison of profiles at Fr=0.35 (Series-60 model)

Figure 6. Convergence history of CW


Table 1. Comparison between calculated and experimental values (S-60, Fr=0.25)
________________________________________________________________
CRx103
sinkage(m)
trim(deg)
_____________________________________________________________________________________________________

Experimental
Calculated

0.8
0.45

0.0061
0.0079

-0.12
-0.07

______________________________________________________________________________________________________

Table 2. Comparison between calculated and experimental values (S-60, Fr=0.35)


_________________________________________________________________
CRx103
sinkage(m)
trim(deg)
______________________________________________________________________________________________________

Experimental
Calculated

1.26
1.23

0.0142
0.0148

-0.07
-0.05

______________________________________________________________________________________________________

Next, the method was applied for a 1:30 model of a passenger-ferry ship with transom stern
and L PP=115.5m. The model had initially a conventional bow and an implicit bulb was designed
to improve its resistance. According to Figure 3, the main characteristics of the bulb at full scale
were: LB=4.9m, BB=1.62m, ZB=2.75m, TB=5m. The ships draft was equal to 5.2m and the design speed 19 knots, corresponding to Fr=0.29. The resistance experiments were carried out at
the Laboratory for Ship and Marine Hydrodynamics (LSMH) of NTUA. Potential flow results
were obtained by applying the aforementioned grid refinement procedure and a total of 5000
(hull)+22000(free-surface) elements were applied. Besides, computations were carried out with
an additive bulb that had equivalent parameters to the implicit. The integrated wave resistance
components are compared to the experimental CR in Table 3. Calculations do not include sinkage and trim corrections. A first conclusion, derived from the experimental results, is that the
designed bulb will not improve drastically the resistance of the ship, provided also that the nondimensional coefficients are divided by the wetted surface which is higher when the bulb is
added. The potential flow predicts higher values in both cases, which shows a rather unexpected
behaviour because the viscous pressure is not taken into account. Apart from any numerical
problems, probably due to the adopted quadrilateral panels, it should be noted here that a serious
wave-breaking and spray was observed during experiments, which was more intense with the
conventional bow. Since the present method cannot take into account these phenomena, such
differences may be expected. However, it is noteworthy that the calculated results follow the
correct trends, while the additive bulb shows better behaviour with respect to the implicit. The
wave profiles on the hull, plotted in Figure 7, show that both bulbs produce essentially the same
shape, which shifts the crest of the original hull towards FP. There are no substantial differences
at the stern region, where the flow separates at the transom, as shown ib Figure 8.

Table 3. Comparisons of CR between experiments and potential flow results


______________________________________________________________
CRx103
conventional bow implicit bulb
additive bulb

__________________________________________________________________________________________________

Experimental
Calculated

1.32
1.38

1.26
1.27

1.21

__________________________________________________________________________________________________

Figure 7. Calculated wave profiles on the hull.

Figure 8. Free-surface around transom

The computed free-surface in all cases was, then, taken as a known boundary to apply the
RANS code underneath. The computations domain was divided in three blocks covering respectively the bow, the stern and the transom and wake regions. The k- model with wall functions
was adopted as the turbulence model, taking into account that ship had a very low block coefficient. The external boundary conditions for the flow variables were calculated from the potential
flow solution as described by Tzabiras & Garofallidis (2005) and, therefore, a restricted domain
could be used for viscous calculations that require very fine grids to obtain reliable solutions. A
mesh-refinement technique was applied (e.g. Tzabiras 1997) to perform grid dependence tests.
Typical results are presented in Table 4 for the case of the implicit bulb, where the first number
denotes the number of nodes in the longitudinal direction, the second refers to nodes girth-wise
and the last, nodes normal to a ships section. As observed, the results obtained with the finest
grid can be considered as reliable for practical conclusions.
Table 4. Grid dependence tests with the RANS solver
______________________________________________________________
Bow Grid
111x27x31
221x54x81
327x81x91
CTx102
3.658
3.625
3.628

__________________________________________________________________________________________________

Stern & Transom


CTx103

126x27x31
4.534

251x54x62
4.437

371x81x91
4.389

___________________________________________________________________________________________________

The calculated total resistance coefficient CT (non-dimensionialized by the wetted surface in


still water in each case) is compared to the experimental values in Table 5. Although the results
show a surprisingly good agreement, it should be considered that the input free-surface in calculations has two main drawbacks: first, it cannot take into account the wave breaking and, second, the inviscid stern waves are normally higher than the real. Nevertheless, the comparisons

demonstrate that RANS computations are substantially more reliable than the initial potential
flow predictions of CR because they include all the viscous influences on the velocity and pressure fields. The benefit from the potential solution relies on the prediction of the free-surface
geometry which seems to be satisfactory for taking decisions about the proper design of bulbous
bows. In addition, a significant conclusion is that the additive bulb behaves just as the implicit
one and may be a strong alternative due to the easier adjustment of bow lines. From both the
experiments and calculations, it is also obvious that the employed bulb does not reduce significantly the total resistance and, consequently, is not suggested for the real ship.
A final run was carried out for the full scale ship with the implicit bulb under the calculated
free-surface. The computed total resistance coefficient was CT=2.485x10-3, while the viscous
pressure coefficient CP=0.962x103. As expected, the latter is lower than the model scale
CP=1.23x103, owing to the reduction of the viscous pressure component. This result verifies that
the Froude hypothesis is not valid for the particular hull.
Table 5. Comparisons of CT between experiments and RANS calculations
______________________________________________________________
CRx103
conventional bow implicit bulb
additive bulb

__________________________________________________________________________________________________

Experimental
Calculated

4.529
4.475

4.444
4.470

4.451

__________________________________________________________________________________________________

CONCLUSIONS
The combined application of the potential and the viscous codes to predict the total resistance of
models or ships with bulbous bows seems to be a reliable tool in the design stage. The additive
bulb produces the same results with an implicit bulb with the same geometrical parameters.
RANS computations are needed to obtain reliable results, while they can be applied to full scale
without any problem. Further comparisons with experimental data are needed to ascertain the
efficiency of the particular procedure.
REFERENCES
Garofallidis, D.A. 1996. Experimental and numerical investigation of the flow around a ship
model at various Froude numbers. Ph.D. Thesis, Department of Naval Architecture and Ma
rine Engineering, National Technical University of Athens, Greece.
Hess, J.L. and Smith A.M.O. 1968. Calculation of potential flow about arbitrary bodies, Prog.
Aeraunaut. Sci., vol 8, 1-136
Janson, C.E. 1997. Potential flow panel methods for the calculation of free-surface flows with
lift. Ph.D. Thesis, Department of Naval Architecture and Ocean Engineering, Chalmers
Technical University, Gothemburg, Sweden.
Kracht, A.M. 1978. Design of bulbous bows. SNAME Transactions
Raven, H.C. 1992. Nonlinear ship wave calculations with the RAPID method. 6th International
Conference on Numerical Ship Hydrodynamics, Iowa City, USA
Tzabiras, G.D. 1997. A numerical study of additive bulb effects on the resistance and selfpropulsion characteristics of a full ship form. Ship Technology Research, vol. 44(1), 98-108
Tzabiras, G.D. 2004. Resistance and self-propusion calculations for a Series-60, cB-0.6 hull at
model and full scale. Ship Technology Research, vol. 51(1), 21-34
Tzabiras, G. And Garofallidis, D. 2005. Computation of the resistance of a Series-60, cB-0.6
model under a measured free-surface. Proc. Marine Transportation and Exploitation of
Ocean and Coastal Resources, Lisbon, Portugal, 295-300

Você também pode gostar