You are on page 1of 9

Seventh International Conference

On High-Performance Marine Vehicles


Melbourne, Florida, USA
13-15 October 2010

CFD SIMULATION OF PLANING HULLS

Stefano Brizzolara
&
Diego Villa
brizzolara@dinav.unige.it
diego.villa@unige.it
Marine CFD Group, University of Genova, Italy

Abstract
In the last century the CFD (Computational Fluid Dynamics) methods had an exponential growth of applications,
due to the availability of user friendly codes and the increase of the computational performance, with the possibility
to generalize the solution of the N.-S equation to more complex physical problems with multiple phases and free
surfaces. Also the naval architecture community has started to intensively apply these methods for predicting steady
and unsteady performance of ships and boats. Most of the effort, though, has been made for displacements hulls,
while only few simple methods could be applied for planing hulls. That is probably due to the physical
hydrodynamic complexity odf the planing problem. The major problem still open regards the determination of the
actual possibilities of solutions and the level of confidence that can be expected on the numerical results
The paper summarize the experience gained by the Marine CFD Group of University of Genova, in these last
period, presenting a selection of significative results that cover the research followed path: from the study of simple
prismatic hull shapes, to the flow around d real hull shapes with steps and trim control appendages and propulsors.
A large database and know how has been built to predict the dynamic attitude and the resistance for different
types of full appended planing hulls.
Introduction
The solution of the hydrodynamic problem of a general planing hull advancing at stationary speed in calm water is
still nowadays tackled with simplified approximated semi-empirical methods, mainly derived from the original
Savitsky method, modified and (empirically) adjusted to allow for a dead-rise variation or to include the effect of
spray rails.

Figure 1: Experimental flow pattern around a planing hull


Few more physical, mainly based on potential flow theories, were developed and tested in this recent period, as for
instance the method of Savander et al (2002). These methods, though, are based on rather crude approximations for
what regards the shape of the hull form. They solve the problem of dealing with generally cambered (in longitudinal

Seventh International Conference


On High-Performance Marine Vehicles
Melbourne, Florida, USA
13-15 October 2010

and transversal direction) planing hull forms, but still cannot allow for important appendages like spray rails and
steps.
The flow around planing hulls, in fact, is rather complex from the hydrodynamic point of view, since involves
different physical phenomena having different length and time scales, such as thin spray flow, wave breaking,
turbulent boundary layer (see Figure 1). The free surface waves are long and often affected by overturning and
breaking phenomena. Under the bottom, in front of the stagnation line, a thin spray sheet is formed which normally
sharply separates from the chines. The rest of the incoming water flows attached under the bottom forming a
turbulent boundary layer with an oblique angle that depends on the local deadrise and trim angle of the hull and on
the proximity of the streamline to the chine.
The other problem is the attitude assumed by the hull in planing conditions; this requires a routine that searches for
the dynamic equilibrium of gravitational and hydrodynamic forces and moments acting of the hull. Virtually, all of
the above problems are nowadays solvable with volume of fluid RANSE solvers. The paper will illustrate
The research pathway and the first cases
The problem of planing hulls was faced by the authors for the first time around year 2005, encouraged by the
promising results published by Caponnetto (2001), followed by Azcueta (2003). The research program initiated with
the study of very simple planing dihedral hull forms with 20 degree constant deadrise, as systematically tested by
Kapryan & Weinstein (1952) and Chamblis & Boyd (1953), then continued with more realistic but still simple hull
forms, taken from the Series 62 whose results will be summarized in the next section, for ending at the actual state
of the art that feature general hull forms with appendages. Some of these applications will be illustrated in the last
section.

Figure 2: Dihedral planing hull with 6deg trim. Comparison numerical Lift (CL) and Drag (CR) coefficients
predicted with RANSE with experimental results and other semi-empirical formulations

Figure 3: Mesh type (up) and Free surface (low) for a Dihedral planing hull case having 20deg deadrise and 6
deg trim, at FnB=4.7. The spray area (dark yellow), the pressure area (gray) are correctly captures as well as
the overturning divergent wave.

Seventh International Conference


On High-Performance Marine Vehicles
Melbourne, Florida, USA
13-15 October 2010

In the case of constant deadrise hull forms, the CFD calculation were performed, at that time, with Star-CD. The
experimental attitude was used for RANSE simulation and predicted lift, drag and trim moment were compared with
the experimental results. An example of correlation of the numerical results obtained, entirely published in
Brizzolara and Serra (2007), is presented in Figure 2 in the case of lift and drag.
RANSE calculations were run in model scale, using an initial structured body fitted mesh (as from Figure 3), then
refined splitting the cells along their vertical dimension with two subsequent level, one close to the hull and the
other in proximity of the free surface. A total number of 300K prismatic elements, divided in 12 blocks, was
sufficient to obtain convergence of the results. The domain was rather limited, due to the proper planing condition
simulation speed range, only 1 length forward, 1 length aft, 0.8 length aside and 1 length below the hull were
sufficient to verify the sensitivity analysis of boundary conditions. On the outlet a prescribed mean piezometric
pressure was assigned. A high Reynolds k- turbulence model was used with a double later law of the wall. A The
simulation was done in non-stationary condition to solve for the deformation of the free surface up to the stationary
conditions, starting by flat initial free surface. Time step was adequately selected in order satisfy the Courant
condition everywhere in the discrete domain.
The correlation of numerical results was made with the experimental results, as noticeable rather dispersed at lowest
wetted lengths as in the example of Figure 2, and with semi-empirical formulations of Savitsky and Shufford (only
for lift and centre of pressure).
The overall correlation of RANSE results with the experiments was around 10% on drag and about 5% on lift and
trim moment. In any case a better accuracy than that of reference semi-empirical methods.
Already in these simulations unrealistic VOF distributions on the hull wetted surface were noted, namely values of
0.75/0.8 were predicted in areas where the hull should be lapped by pure water (i.e. VOF=1), but no means of
correction were adopted. Similar problems have been noted in the next series of simulations with Series 62 and real
hulls, but can be solved with a the new mesh typology and adequately corrected from the resistance point of view as
further on explained.
Series 62 with Fixed Trim
The second step of validation has been made on the parent model of Series 62 (Clement & Blount, 1963) at different
weight, speed and static trim conditions. The hull has a little more variation of the transverse sections at the bow,
keeping a constant deadrise angle of 12.5 degrees at stern and smaller chine breadth at transom.

Figure 3: Series 62 parent model


A first set of calculations, this time made with StarCCM+, have been performed on the model fixed in the
experimental attitude in the complete planing speed range.

Seventh International Conference


On High-Performance Marine Vehicles
Melbourne, Florida, USA
13-15 October 2010

Figure 4: S-62, parent model: predicted wetted pressure and spray area on the wetted part of the hull as a
function of Froude number
The mesh type was unstructured, mainly composed by Cartesian cells in the majority of the domain, by five
extruded prism layers in the hull boundary layer and by trimmed cells in the region of transition. Mesh size and
density was calibrated, by an extensive sensitivity study, resulting in a minimum number of about 700k elements.
The box like domain spans 1L forward, 1L aft, 2L below and 2L aside the hull (L = hull length). A higher mesh
density in the far field would be requested at semi-displacement speeds. A special topology of the mesh was used, as
presented in the transverse cut of Figure 7: a thin layer of cells placed in way of the chine was created in order to
capture the thin spray sheet sharp detachment. This kind of mesh permitted to save thousands of cells while
preserving good results in terms of hull flow characteristics. Still some diffusion of air below the bottom is visible
especially at lowest speed, as from Figure 6: the lower VOF strips are originated at the stagnation line and are
convected by the main flow at stern. This special mesh topology in way of the chines avoids the side to become
unrealistically wetted. In fact, as from Figure 6 the VOF on the sides is rather well behaved: at the lowest speed the
flared side above the chine is wetted at stern, while progressively it becomes dry as the speed is increased. Some
spray curles are noted at the bow, pushed on the hull by the air flow.

Figure 5: Spray and spilling wave at chine

Figure 6: Special mesh type used in way of the chine

With this kind of mesh and settings, either the lift and longitudinal moment and the total drag result within an
accuracy of 3% with respect to model tests.
Serie 62 with Free Attitude: the Marine CFD Group method
An external Java/C++ routine was developed by the Marine CFD Group (MCFDG) and linked to StarCCM+ solver
(since from version 2) to change the position of the hull during the non stationary time step iterations in order to
converge on the hull hydrodynamic equilibrium. Due to the symmetry of the problem, the equilibrium was reduced
at the sole longitudinal plane, i.e. allowing for a variation of trim angle () and draft (sinkage). The principle of the
algorithm is illustrated by the flow diagram of Figure 8.

Seventh International Conference


On High-Performance Marine Vehicles
Melbourne, Florida, USA
13-15 October 2010

The unsteady simulation is launched with the hull in a first guessed dynamic attitude and a number of time steps are
calculated with the hull in fixed position (static simulation), to avoid the initial unrealistic forces caused by the
initial impulsive acceleration. Then the algorithm calculates the deviation of vertical force and longitudinal moment
from the equilibrium condition. With the current attitude and these deviations by the Savitsky method (Savitsky
equilibrium) applied on a dihedral equivalent hull, a correction of trim and sinkage is calculated. This attitude
correction is applied smoothly to the rigid body over a number of time steps which depend on the given maximum
rotational and translational speed. Then a sufficient number of time steps is run. The algorithm in the user routine
linked to the solver automatically performs all these steps until convergence on the equilibrium. A final number of
time steps (solution smoothness) is finally imposed to reach the undisturbed stationary equilibrium flow field.
In the newest version of StarCCM+ solver there is only recently the possibility to reach this equilibrium condition
following the physics of the hull rigid body motions, launching a non stationary calculation (in any case needed for
the free surface). But this method, which follows a not meaningful physic, is inherently affected by oscillation and
can require considerable computational time to reach convergence. Special artefacts can be used also in this case,
such as to decrease the inertia of mass or introduced an artificial damping coefficient.

Figure 8: DINAV dynamic trim search algorithm

Figure 9: Convergence of dynamic trim angle: DINAV (red) vs 2DOF method


Figure 9 shows a comparison of the convergence of the MCFDG convergence algorithm against the 2DOF method.
In the presented case, the two methods are almost equivalent in terms of efficiency, but it is possible to obtain better

Seventh International Conference


On High-Performance Marine Vehicles
Melbourne, Florida, USA
13-15 October 2010

performance from the DINAV method by properly setting the number of time steps an d the under-relaxation factor
applied on the Savitsky predicted trim and sinkage corrections. In both cases the final dynamic attitude is well
predicted, as presented in the first two graphs of Figure 10. Also the accuracy of the predicted total resistance by the
DINAV automatic procedure is very satisfactory, as from the third graph of Figure 10, being in general within 3%
from the experimental results. Same order of magnitude of the errors was maintained in general for all the tested
cases of Series 62.
Real hull forms
Finally on the basis of the experience gained with the extensive systematic CFD simulation of different basic
typologies of planing hull forms, the authors have applied the illustrated method to different real hulls, as built, i.e.
with all the small geometrical details and appendages that can influence the dynamic attitude and the resistance.
An important role is normally played by the chines flare and the spray rails. An example of results obtained from the
calculations is presented in Figure 11 and corelate with the test of Figure 1. The VOF distribution highlights some
air ingestion in way of the spray rail which is only marginally present in the experiments. This convinced the authors
to apply a correction to the predicted tangential force on each cell face at the hull wall inversely proportional to the
volume of fluid calculated in the cell. For the rest either the wave and spray formation and the dynamic attitude was
very well correlated with the experimental results. Also in this case, after the friction correction, the deviation of the
numerically predicted total resistance at design speed was within the 5% against the measured value in towing tank
experiments.

Figure 10: Dynamic Draft (up right), Trim (up left) and Total Drag (bottom): comparison of obtained CFD
results on model 4667-1, with experiments and Savitsky method

Seventh International Conference


On High-Performance Marine Vehicles
Melbourne, Florida, USA
13-15 October 2010

Figure 11: Fast 20m motoryacht, free surface elevation and VOF distribution under the hull (comp. Fig.1)
Some other analysis has been carry out on realistic hull shapes, for example the effects of different configurations
and shapes of spray rails has been investigated on a full scale high speed inteceptor and a rubber boat hull. Figure 12
shows the pressure distribution calculated under the hull bottom. It can be noted that the upper part of the picture is
without the spray rails, instead the lower part is with the spray rails. The full scale validations have confirmed the
capabilities of the RANS solver to adequately predict the influence of spray rails on the resistance and dynamic
attitude of the hull. The general effect is to increase the hydrodynamic forces developed under the bottom if the ship
attitude is fixed or it can reduced the running trim and sinkage if the hull is left free to move around its centre of
gravity.

Figure 12: Appendages effect under the hull in terms of pressure coefficient.

Figure 13: Fast 12m rigid keel rib with waterjet propulsion. Pressure distribution including WJ intake
suction effect
An important feature to be included CFD model is the action of the propulsors, either submerged screw propellers or
waterjets. The action of the inclined thrust produced by waterjets or propellers can be included in the DINAV
algorithm that searches for dynamic equilibrium. In addition to the inclined thrust effect, though, the propulsors can
locally modify the flow field on the aft part of the hull, inducing additional hydrodynamic forces and moments
which can affect trim sinkage and hence the total resistance.

Seventh International Conference


On High-Performance Marine Vehicles
Melbourne, Florida, USA
13-15 October 2010

The case of Figure 13 features a fast planing hull with waterjet propulsion. The suction effect induced by the
modelled waterjet intake under the hull is strong and is able to locally modify the pressure distribution and
streamlines paths on the hull bottom. In the Figure 13 the difference in the pressure distribution between the same
hull shape with and without the waterjets can be noted. The result of this effect is a variation of dynamic trim as
large as one degree and a variation of total resistance of about 30% with respect to the bare hull, principally due to
the dynamic attitude variation.
Finally for planing hulls, another important effect is induced by stern flaps or interceptors, a very frequent practice
in the contemporary planing hulls of small to medium sized fast crafts. In this respect the authors have recently
published in Brizzolara and Villa (2009) a comprehensive CFD study for the comparison of the effects of these kind
of appendages. In particular it is been products some new relationships between the performance of the two types of
appendage, and is been given some guideline to chose witch is the ones more performances. The Figure 14 show the
effect due of the two equivalent appendages in terms of local pressure coefficient and streamline path below a
prismatic hull bottom.

Figure 14: Appendages effect under the hull in terms of pressure coefficient.
Conclusions and Further Prospects
The general conclusion, on the basis of an extensive series of simulations exemplified in this paper, is that RANSE
methods are mature to be used for accurate prediction of the resistance of planing hull in calm water including also
small appendages effects. The overall order of magnitude of the error noted after many validation studies is well
within engineering purposes for planing hull forms and can be considered equivalent to that derived from a medium
size towing tank facility.
Ad hoc meshes are often necessary to accurately resolve jet spray and wave breaking off the chines with a
reasonable number of cells (less or in the order of one million). Special care should be given to the verification of
the VOF distribution below the hull. This is very much related to the mesh quality and type and can affect the
magnitude of the friction resistance; while it does not influence significantly the pressure distribution. Some a
posteriori correction on frictional resistance might be needed in some cases.
The free surface waves, the pressure distribution and velocity flow field below the hull can be quite well predicted
by the RANSE solver with an adequate mesh and are almost unaffected by the VOF numerical diffusion under the
hull.
The running attitude of the hull can be correctly predicted, as demonstrated, either with a dedicated iterative method
based on Savitsky formulation or following the physic of the hull motion from an initial unbalanced guessed
condition. The first method can offer better margin for efficiency in terms of computational time needed to reach the
stationary condition, while the second can fall into large oscillating pitch and heave motions, physically consistent,
but practically irrelevant.
Similar studies and simulations are planned to be repeated with OpenFoam. With this new solver, it will be
interesting to compare the behaviour of the solution of the VOF advection and diffusion equations with respect to

Seventh International Conference


On High-Performance Marine Vehicles
Melbourne, Florida, USA
13-15 October 2010

the actually used commercial code. The problem for this open source solver is shifted to the generation of the mesh
which will require an ad hoc pre-processor, either commercial again or made for the purpose. In this respect the
authors are working on a method to accurately and smoothly model 3D surfaces by a so called subdivision surface
technique.
References
Caponneto M., (2001) Practical CFD Simulations for planing Hulls, HIPER 2001, High Performance Marine
Vehicles, Hamburg (D)
Azcueta R., (2003) Steady and Unsteady RANSE Simulations for planing Crafts, International Conference on
Fast Sea Transportation FAST 2003, Ischia (IT)
Clement E. P. and Blount D. L., (1963), Resistance Tests of a Systematic Series of Planing Hull Forms, SNAME
Transactions, Volume 71, pp. 491-579.
Brizzolara S., Serra F. (2007) Accuracy of CFD Codes in the Prediction of Planing Surfaces Hydrodynamic
Characteristics. 2nd International Conference on marine Research and Transportation, ICMRT07. ISCHIA. 2830 Giugno. (vol. 1, pp. A-1-A-12). ISBN: 88-901174-3-5
Brizzolara S., Villa D. (2009), A systematic CFD Analysis of Flaps / Interceptors Hydrodynamic Performance,
Fast 2009 Int. Conference on High Speed Fast Ship Design, Athens
CD-ADAPCO (2009), Star-CCM+ User and Theory Manual, version 4.04.011, 2009
Chambliss D.B., Boyd G.M. (1953), The Planing Characteristics of two V-shaped Prismatic Surfaces having angles
of Deadrise of 20 and 40 , Langley Aeronautical Laboratory, NACA Technical Note 2876, Washington 1953.
Kapryan W.J. and Weinstein I. (1952), The planing characteristics of a surface having a basic angle of dead rise of
20 degrees and horizontal chine flare, Langley Aeronautical Laboratory, NACA Technical Note 2804, 1952
Savander B.R., Scorpio S. M., Taylor R.K. (2002), Steady hydrodynamic analysis of planing surfaces, Journal of
Ship Research, vol. 46, no.4, pp. 248-279
Villa D., Vatteroni G., Brizzolara S. (2009) CFD Calculations of Planing Hulls Hydrodynamics, Star European
Conference, London, March 2009

You might also like