See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/225370823

Numerical simulation of interference effects for a high-speed catamaran

Article in Journal of Marine Science and Technology · September 2011

DOI: 10.1007/s00773-011-0132-3

CITATIONS READS

75 5,533

3 authors:

Riccardo Broglia Stefano Zaghi

Italian National Research Council Italian National Research Council
130 PUBLICATIONS 2,500 CITATIONS 43 PUBLICATIONS 634 CITATIONS

SEE PROFILE SEE PROFILE

Andrea Di Mascio
Università degli Studi dell'Aquila
134 PUBLICATIONS 3,238 CITATIONS

SEE PROFILE

All content following this page was uploaded by Andrea Di Mascio on 22 May 2014.

The user has requested enhancement of the downloaded file.

J Mar Sci Technol (2011) 16:254–269
DOI 10.1007/s00773-011-0132-3

ORIGINAL ARTICLE

Numerical simulation of interference effects for a high-speed

catamaran
Riccardo Broglia • Stefano Zaghi • Andrea Di Mascio

Received: 19 March 2010 / Accepted: 29 May 2011 / Published online: 30 June 2011
 JASNAOE 2011

Abstract The simulations of the flow around a high- military purposes; in particular, catamaran configurations
speed vessel in both catamaran and monohull configura- are very attractive because of their excellent performance
tions are carried out by the numerical solution of the with respect to speed, safety, resistance and transversal
Reynold averaged Navier–Stokes (RANS) equations. The stability.
goal of the analysis is the investigation of the interference As a consequence, a large number of theoretical,
phenomena between the two hulls, with focus on its experimental and numerical studies were carried out in
dependence on the Reynolds number (Re). To this aim, recent years. Experimental studies were primarily dedi-
numerical simulations are carried out for values of Re cated to the analysis of the effect of the hull separation on
ranging from 106 to 108 for two different values of the both resistance components and wave interference. Resis-
Froude number (Fr = 0.30, 0.45). Wave patterns, wave tance components of high-speed catamarans were analyzed
profiles, limiting streamlines, surface pressure and velocity by Insel and Molland [20] and by Molland et al. [26], the
fields are analyzed; comparison is made between the cat- investigation being focused on the effect of the separation
amaran and the monohull configurations. Dependence of distance, length over beam, length over displacement and
the pressure and viscous resistance coefficients, as well as breadth over draught ratios for a systematic series of high-
of the interference factor, on the Reynolds number is speed displacement catamarans; valuable experimental
investigated. Verification and validation for both resistance results were reported in terms of viscous and non viscous
coefficients and wave cuts is also performed. resistance components, as well as wake contours. A similar
analysis was carried out in shallow water by Molland et al.
Keywords Catamaran  RANS based simulations  [27]. Millward [25] used the thin ship theory to investigate
Interference  Scale effects the effects of demihull separation in both deep and shallow
water over a range of Froude number.
In order to quantify the viscous effects on resistance and
1 Introduction interference, Armstrong [1] tested a double model in a
wind tunnel; it has been shown that there is a significant
The demand for high-speed multihull vessels has strongly interference of purely viscous nature between hulls in close
increased during the last decades for both commercial and proximity; however, it has to be pointed out that in this
experiment the influence of factors such as wind tunnel
blockage effects, model mounting effects, etc. could have
R. Broglia (&)  S. Zaghi  A. Di Mascio
affected the results. A rather complete analysis of the
Istituto Nazionale per Studi ed Esperienze di Architettura
Navale, CNR, via di Vallerano, 139, 00128 Rome, Italy resistance components of high-speed catamarans, as well as
e-mail: [email protected] of suitable form factor to be used for resistance scaling has
S. Zaghi been provided by Couser et al. [10]. In a recent paper,
e-mail: [email protected] Souto-Iglesias et al. [32] made an accurate experimental
A. Di Mascio analysis of the effect of separation distance between the
e-mail: [email protected] demihulls; resistance, trim and sinkage, as well as inner

123
J Mar Sci Technol (2011) 16:254–269 255

and outer wave cuts were measured and an interesting analysis for a catamaran advancing in head waves, while
correlation between the interference factor with wave cuts Zlatev et al. [42] have made a combined experimental and
was highlighted. numerical analysis of the manoeuvring characteristics for
Theoretical and numerical approaches based on the the same catamaran.
potential flow theory have also been used for the investi- The aim of the present work is the analysis of the vis-
gation of the flow field around multihull vessels; most of cous flow field around a fast catamaran. The geometry
these works were devoted to the analysis of wave inter- considered is the Delft catamaran model 372, a high-speed
ference effects, seakeeping performances or evaluation of catamaran designed and tested experimentally at the
the optimal separation distance between the hulls, as for Technical University of Delft (see van’t Veer [39, 40]) and
example in Moraes et al. [28], Colicchio et al. [9], Lugni recently at the Istituto Nazionale per Studi ed Esperienze di
et al. [21], Doctors and Scrace [16], Tarafder and Suzuki Archettettura Navale (CNR-INSEAN) (Broglia et al. [5]).
[37] and Maki et al. [22] (in this paper the authors have The analysis is carried out by RANS simulations around
also used a RANS approach). A theoretical approach has both catamaran and monohull configurations for a
also been used to quantify the influence of the viscosity on Reynolds number ranging from *106 up to *108 at two
the wavemaking capability of a catamaran by Doctors [15], different values of the Froude numbers. In this paper the
in both deep and restricted water, for different demihull analysis is undertaken using a simulation code developed at
separations. CNR-INSEAN (Di Mascio et al. [11–14], Broglia et al.
However, in spite of the large increase of the interest on [3]), which is a finite volume second order accurate solver
multihull vessels, the use of computational fluid dynamics for the unsteady incompressible Navier–Stokes equations.
(CFD) for the analysis of the viscous flow past catamarans The work has been done in the framework of a NICOP
is still very limited. As widely accepted, CFD can be a collaboration project between CNR-INSEAN and the Iowa
useful support to model tests in a towing tank; moreover, a Institute of Hydraulic Research (IIHR). In this project, both
CFD analysis would allow a complete investigation of the experimental and numerical activities are scheduled and in
flow field details, which can help the physical under- progress; the main topics of this research activity are:
standing of the phenomena involved. Indeed, the hydro- (a) the analysis of the viscous and non viscous effects on
dynamics of a catamaran presents complex problems. For interference for a fast displacement catamaran and (b) the
example, the estimation of resistance requires the analysis analysis of the hydrodynamics nonlinearities in the sea-
of boundary layer and wake deformation (caused by the keeping performances of a fast catamaran in rough sea.
asymmetry of the flow around the demihulls), that have Only the first topic is reported and discussed in the present
relevant effects on the development of longitudinal vortices paper.
and on the viscous component of the drag; on the other The paper is organized as follows: in the next section,
hand, for the estimation of wave resistance, an accurate the mathematical models with the proper boundary and
investigation on the interaction between the two wave initial conditions will be recalled, and the numerical
systems generated from the demihulls is mandatory. From schemes will be briefly described. Numerical results will
this point of view, the use of numerical simulations, based follow: the flow field around the catamaran will be inves-
on the Reynolds averaged Navier–Stokes equations, could tigated, and dependencies on the Reynolds number will be
be of great help in understanding both inviscid and viscous shown. Conclusions and perspectives will close the paper.
effects on the hydrodynamic performances of multihull
vessels.
Viscous analysis of the flow field has already been 2 Mathematical model
carried out by Pinto et al. [31], Stern et al. [34], Campana
et al. [6] and Miller et al. [24], among others. From these The governing equations for the unsteady motion of an
papers and from the results reported in the present paper, it incompressible viscous fluid can be written in integral form
can be seen that the flow field around the catamaran as:
completely differs from that around the monohull. In par- I
ticular, distorted streamlines are observed around each U  ndS ¼ 0
demihull, and also the wave pattern is altered; conse-
SðVÞ
quently, the pressure distributions along the hull is con- Z I ð1Þ
siderably changed, the boundary layer is modified, and o
UdV þ ðFC  FD Þ  ndS ¼ 0
zones of flow separation can appear. All of these effects are ot
V SðVÞ
strongly dependent on the Reynolds number.
Other interesting studies have been proposed very where V is a control volume, S(V) its boundary, and n the
recently: Castiglione et al. [7] have performed a CFD outward unit normal. In the general formulation, the

123
256 J Mar Sci Technol (2011) 16:254–269

equations are written in an inertial frame of reference, in Initial conditions have to be specified for the velocity
order to take into account the possibility of grid motion to field and for the free surface configuration:
follow possible moving boundaries. The equations are
ui ðx; y; z; 0Þ ¼ ui ðx; y; zÞ i ¼ 1; 2; 3
made non-dimensional with a reference velocity U? and a ð5Þ
 y; zÞ:
Fðx; y; z; 0Þ ¼ Fðx;
reference length L and the water density q.
In Eq. 1, FC and FD represent inviscid (advection and
pressure) and diffusive fluxes, respectively: 3 Numerical model
FC ¼ pI þ ðU  VÞU
  The numerical solution of the governing equations (1) is
1 ð2Þ
FD ¼ þ mt ðrU þ rUT Þ computed by means of a simulation code developed at
Re
CNR-INSEAN; the code yields the numerical solution of
In the previous equation, p = P ? z/Fn2 is the the unsteady Reynolds averaged Navier–Stokes equations
hydrodynamic pressure (i.e. the difference between with proper boundary and initial conditions. The algorithm
the total P and the hydrostatic pressure -z/Fn2, Fn = is formulated as a finite volume scheme, with variable co-
U?/(gL)1/2 being the Froude number and g the acceleration located at cell centres. Turbulent stresses are taken into
of gravity parallel to the vertical axis z, positive upward), account by the Boussinesq hypothesis, with several turbu-
V is the local velocity of the control volume boundary, lence models (both algebraic and differential) imple-
Re = U?L/m the Reynolds number, m the kinematic mented. Free surface effects are taken into account by a
viscosity, and mt the non-dimensional turbulent viscosity; single phase level-set algorithm. Complex geometries and
in the present work, the turbulent viscosity has been multiple bodies in relative motion are handled by a suitable
calculated by means of the Spalart–Allmaras one-equation dynamical overlapping grid approach. High performance
model [33]. In what follows, ui is the ith Cartesian computing is achieved by an efficient shared and distrib-
component of the velocity vector (the Cartesian uted memory parallelization.
components of the velocity will be also denoted with u, In the following subsections the main features of the
v, and w). numerical algorithm are briefly recalled; the interested
The problem is closed by enforcing appropriate condi- reader is referred to Di Mascio et al. [11–14] and Broglia
tions at physical and computational boundaries. On solid et al. [3] for details.
walls, the relative velocity is set to zero (whereas no
condition on the pressure is required); at the (fictitious) 3.1 Spatial discretization
inflow boundary, velocity is set to the undisturbed flow
value, and pressure is extrapolated from inside; on the For the numerical solution of the Eq. 1, the fluid domain D
contrary, pressure is set to zero at the outflow, whereas is partitioned into Nl structured blocks Dl, each subdivided
velocity is extrapolated from inner points. into Ni 9 Nj 9 Nk disjoint hexahedrons Dlijk. In the
At the free surface, whose location is one of the numerical scheme adopted here, the blocks are not neces-
unknowns of the problem, the dynamic boundary condition sarily disjoint, but can be partially overlapped, as it will be
requires continuity of stresses across the surface; if the explained. Conservation laws are then applied to each finite
presence of the air is neglected, the dynamic boundary volume:
condition reads: 6 Z
X
z j U  n dS ¼ 0
p ¼ sij ni nj þ 2 þ
Fn We2 s¼1
Sls
1
sij ni tj ¼ 0 ð3Þ Z ð6Þ
6 Z
X
o
sij ni tj2 ¼0 U dV þ ðFC  FD Þ  n dS ¼ 0
ot s¼1
l
Vijk Sls
where sij is the stress tensor, j is the average curvature,
We = (qU2?L/r)1/2 is the Weber number (r being the where Sls is the sth face of the finite volume Dlijk , whose
surface tension coefficient), whereas n, t1 and t2 are l
measure is Vijk .
the surface normal and two tangential unit vectors, To obtain second order accuracy in space, convective and
respectively. viscous fluxes in the momentum equations, as well as surface
The actual position of the free surface F(x, y, z, t) = 0 is integral of the velocity in the continuity equation, are com-
computed from the kinematic condition: puted by means of the midpoint rule, and therefore all the
DFðx; y; z; tÞ quantities are evaluated at cell faces centroids. A simple
¼0 ð4Þ second order centred scheme is used for the computations of
Dt

123
J Mar Sci Technol (2011) 16:254–269 257

the viscous terms, whereas high order Godunov type order to improve the convergence rate of the sub-iteration
schemes [18] can be applied for the computation of the algorithm. For steady problems, the solution can be
eulerian terms (see Di Mascio et al. [13], Harten et al. [19], achieved either as asymptotic solution of the unsteady
Van Leer [38]). In the computations which follow the third problem or neglecting the physical time derivative in (10)
order scheme has been used. and iterate in the pseudo-time only.

3.2 Temporal integration and dual time stepping 3.3 Single phase level set

The semi-discrete system of equations can be rewritten in The presence of the free surface is simulated by means of a
vector form as: single-phase level-set approach developed in [12], which is
briefly recalled here. In classical level set approaches (see
oðVqÞlijk for example Osher and Sethian [30] and Sussman et al.
K þ Rlijk ¼ 0 ð7Þ
ot [36]), a smooth function u(x, y, z, t), whose zero level
coincides for t = 0 with the free surface, is defined in the
where K = diag(0, 1, 1, 1),
Z whole physical domain (i.e. in both liquid and air phases);
1 the kinematic boundary condition (4) is extended to all the
qlijk ¼ l ðp; u; v; wÞT dV ð8Þ
Vijk points in the domain, yielding a transport equation for the
l
Vijk level set function:
is the volume average of the unknowns and Rlijk represents ouðx; y; z; tÞ
þ ðU  VÞ  ruðx; y; z; tÞ ¼ 0
the flux balance on the current cell. In order to remove any ot ð11Þ
stability constraint on the time step, an implicit scheme is uðx; y; z; 0Þ ¼ dðx; y; zÞ
considered: the time derivative in the previous equation is U being the velocity of the underlying flow, V the grid
approximated by a second order accurate three-points velocity and d(x, y, z) the signed distance from the free
backward formula surface at t = 0. u(x, y, z, t), being a material surface, its
nþ1 n n1 zero level represents the free surface location for t [ 0;
  
3ðVqÞlijk  4ðVqÞlijk  þðVqÞlijk  nþ1
 moreover, as u(x, y, z, t) is initialized to be the signed
K þ Rlijk  ¼ 0
2Dt distance from the surface of discontinuity, the sign of the
ð9Þ level set function remains unchanged at material points. In
the single-phase algorithm adopted here, the Navier–Stokes
where the superscript n denotes the time level and Dt is the
equation are solved only in the liquid phase, identified by
physical time step. The previous equation represents a
the negative values of the level set function. In the outside,
system of coupled non-linear algebraic equations, that are
the velocity and pressure are simply extrapolated along the
solved iteratively. This is achieved by means of a dual (or
normal direction to the isosurface of u(x, y, z, t). The
pseudo) time integration approach (see Merkle and
function u(x, y, z, t) is reinitialized as a distance function at
Athavale [23] for more details); to this end, a pseudo-
each time step in order to improve accuracy.
time derivative is introduced in the discrete system of
equations as
3.4 Overlapping grids approach
mþ1 m
 
ðVqÞlijk  ðVqÞlijk 
~ In this section the basic elements of the overlapping grid
K
Ds discretization (or ‘‘chimera’’ method) for both fixed grids
mþ1 n n1
   and its extension to moving grids will be briefly recalled.
3ðVqÞlijk  4ðVqÞlijk  þðVqÞlijk 
þK For more details and examples of applications the reader is
2Dt referred to Muscari et al. [29], Di Mascio et al. [14] and
mþ1
l  Broglia et al. [4].
þ Rijk  ¼ 0 ð10Þ
The introduction of chimera capabilities in the RANS
being the pseudo-time step, K ~ ¼ ð1=b; 1; 1; 1Þ, and b the code is made through a modification of both the boundary
pseudo-compressibility factor (see Chorin [8]). The solution conditions and internal point treatment for those zones
is iterated to steady state with respect to the pseudo-time where overlapping appears. In particular, besides the nat-
s for each physical time step; the pseudo-time integration is ural type (e.g. wall, symmetry, inflow, outflow,…), a new
made by means of the approximated factorization scheme type of boundary condition, the chimera type is added for
by Beam and Warming [2]. Local dual time step and an those boundaries where the solution must be interpolated
efficient multi-grid technique (Favini et al. [17]) are used in from other blocks. For these cells, the first step is to find a

123
258 J Mar Sci Technol (2011) 16:254–269

‘‘donor’’ cell, i.e. a cell that contains the face centre for tuning is left to the user. Communication between proces-
which an interpolation is needed. In order to retain the best sors for the coarse grain parallelization is obtained by using
possible approximation, if a boundary cell falls within standard Message Passing Interface (MPI) library, whereas
more than one donor cell, the smallest one is chosen as fine grain (shared memory) parallelization is achieved by
basis for interpolation. Once the donor is identified, a means of the Open Message Passing (OpenMP) library. The
convex set of eight donor cell centres is searched, and a tri- efficiency of the parallel code has been investigated in
linear interpolation is used to transfer the solution from the Broglia et al. [3] (to which the reader is referred for details)
donor block to the boundary of the one under analysis. and satisfactory speed up performances have been shown in
As to internal points, we look at first for possible different test cases.
overlapping for each cell centres. If this overlapping is
found, the cell is marked as a hole only if the donor cell is
smaller than the one we are considering. As for boundary 4 Computational parameters
cells, if more than one possible donors are found, the
smallest one is chosen. Differently from standard chimera Numerical simulations are carried out around the Delft 372
approaches, however, the cells marked as holes are not catamaran model (Fig. 1), whose main particulars are
removed from the computation but the interpolated solu- given in Table 1. The model has a nominal length of
tion is enforced by adding a forcing term to the Navier– Lpp = 3 m. The analysis is carried out for two values of the
Stokes equations, in a body-force fashion: Froude number, namely Fr = 0.30 and 0.45 and several
mþ1 m values for the Reynolds number. All the computational
 
ðVqÞlijk  ðVqÞlijk  parameters are summarized in Table 2, the values corre-
K~
Ds sponding to towing test results being in bold; numerical
mþ1 n n1 computations for the monohull have also been carried out
  
3ðVqÞlijk  4ðVqÞlijk  þðVqÞlijk 
þK at the same conditions.
  2Dt  The frame of reference adopted is shown in Fig. 1: the
mþ1 k  m m 
l   longitudinal axis is aligned with the free stream velocity,
þ Rijk  þ ðVqÞlijk  ðVqÞint  ¼0 ð12Þ
d positive backward; the z-axis is vertical, positive upward;
In the previous equation qlijk is the vector of the
dependent variables at the cell ijk (of the block l) marked
as a hole, Rlijk is the vector of residuals, k is a parameter
O(10), and d is the minimum between the cell diameter and
the physical time step. This approach is particularly useful
when using multi-grid and approximate factorization, as it
allows to maintain a structured data set. For points that fall
within a rigid body, the forcing term in the equation is
analogous to the previous one, but the velocity in the
forcing term is equal to the local velocity of the body.
Of course, when dealing with moving grids, the grid
topology must be re-computed at each time step. In order to
speed up the algorithm, a nested search that exploits the Fig. 1 Delft 372 catamaran: geometry
multigrid structure is performed.
Table 1 Delft 372 catamaran: main particulars (nominal length)
3.5 Code parallelization
Dimension Symbol Value
The coarse/fine grain parallelization of the unsteady RANS Length between perpendiculars Lpp 3.00 m
code has been achieved by distributing the structured blocks Beam overall B 0.94 m
among the available distributed (nodes) or shared memory Beam demihull b 0.24 m
(threads) processors, and by spreading the computational
Distance between centre of hulls H 0.70 m
work to be done (mostly in terms of do loop inside each
Draught T 0.15 m
block among available shared memory processors). Useful
Displacement D 87.07 kg
pre-processing tools, which allows the splitting of the
Vertical centre of gravity KG 0.34 m
structured blocks and the distribution of them among the
Longitudinal centre of gravity LCG 1.41 m
processors, are available for load balancing, whereas fine

123
J Mar Sci Technol (2011) 16:254–269 259

Table 2 Computational parameters an unbalance in the vertical force of at most 2% and in an

Fr Lpp (m) U (m/s) Re We
error in the position of the centre of buoyancy (which gives
an estimation on the unbalance for pitch) below 1%.
0.30 1.5 1.151 1.726 9 106 164.80 Finally, the fact that the computed variation of vertical
3.0 1.627 4.882 9 106 329.61 forces and pitching moment with the Reynolds number is
6.0 2.301 1.381 9 107 659.61 less than 0.1 and 0.3% respectively, justifies the assump-
7
12.0 3.254 3.905 9 10 1318.44 tion made on the independence of the attitude on the length
20.0 4.201 8.403 9 107 2197.40 scale for fixed Froude number.
0.45 1.5 1.726 2.589 9 106 247.21
3.0 2.441 7.322 9 106 494.41 4.1 Computational grid
6.0 3.452 2.071 9 107 988.83
12.0 4.882 5.858 9 107 1977.66 The computational domain is discretized by means of a
multi block grid; the mesh around the catamaran consists of
about 6 million cells distributed in 39 patched and over-
Table 3 Trim (in degrees) and sinkage (non dimensionalized with lapped blocks; for the grid around the monohull a total of
Lpp) about 3.4 million volumes are distributed in 23 blocks.
Fr Trim Sinkage Because of the symmetry about the vertical plane, for the
catamaran only one hull is discretized, whereas for the
0.30 -0.091 -1.96 9 10-3 monohull only half of a hull is considered; in both cases,
0.45 -1.390 -5.19 9 10-3 symmetry boundary conditions are enforced on the plane
y = 0. Overlapping grids capabilities are exploited to attain
high quality meshes and for refinement purposes. An
y-axis completes a right hand system of reference. The overview of the mesh around the catamaran is given in
origin is placed at the centre of gravity of the catamaran. Fig. 2, where, for the sake of clearness, only every fourth
In order to perform steady state computations, trim and point has been shown and chimera cells have been hidden.
sinkage for both the catamaran and the monohull are fixed The background grid, which covers the whole compu-
at the dynamical positions taken from the measurements by tational domain, is composed of four Cartesian blocks with
Van’t Veer [39] at the Technical University of Delft, a total of about 50,000 volumes; in particular, it is divided
reported in Table 3. Sinkage and trim are assumed positive into two buffer zones at the inflow and outflow, a central
when the centre of gravity moves upward and when the zone and a coarse bottom zone where no relevant phe-
bow moves downward with respect to the centre of gravity, nomena are expected to take place. The boundary condi-
consistently with the frame of references given in Fig. 1. In tions enforced on the blocks of the background grid are:
the simulations, assumptions were made in order to sim- inflow on the front face, outflow on the lateral, on the
plify the analysis: the attitude is considered independent on bottom and on the rear faces, symmetry on y = 0 and
the Reynolds number, and the same attitude is set for the extrapolation at the top.
monohull. In a recent work, Stern et al. [34] and Souto The physical domain close to the hull surface is dis-
et al. [32] have shown that sinkage and trim are larger for cretized by means of 16 body-fitted patched and over-
the catamaran than for the monohull and it has been shown lapped blocks, with a total of about 2.1 million volumes.
that the difference is proportional to the interference factor. Cells are clustered toward the hull surface, so that the
Nevertheless, in the present work the values of sinkage and spacing normal to the wall is less by the one wall unit;
trim are assumed to be the same, because the investigation moreover, no less than 32 cells are within the boundary
here is focused on the interference phenomenon for waves layer thickness (for the highest Reynolds number). The
and wake; the study of the interference effects on trim and blocks adjacent to the hull surface are immersed in a group
sinkage is postponed to future research. of four Cartesian blocks with a total of about 1.8 million
The validity of the assumption made for sinkage and volumes; when the grid is adapted to the prescribed trim
trim was checked a posteriori by evaluating the unbalance and sinkage, the blocks adjacent to the catamaran move
on forces and moments and estimating the change in atti- rigidly with the body inside this grid. The Cartesian blocks
tude from hydrostatic considerations. For the low Froude are also used to increase the resolution around the water
number, the unbalance would cause a change of about 30% level in the near field and in the inner region. Another
in sinkage and of about 15% in trim, whereas for the high Cartesian block with 550,000 volumes is employed to
Froude number the variation is about 15 and 3% for increase the resolution around the water level in the outer
sinkage and trim respectively. Although these values seem region. Finally, 14 blocks are used for refinement purpose
rather high, it has to be notice that these are consequence of in the wake, with about 1.5 million control volumes.

123
260 J Mar Sci Technol (2011) 16:254–269

Fig. 2 Overview of the

overlapping grid around the
Delft catamaran 372

Fig. 3 Wave patterns for the

monohull (top) and the
catamaran (bottom); left
Fr = 0.3, right Fr = 0.45;
Lpp = 3.0 m

4.2 Iteration parameters comparison between the flow field around the catamaran
and the monohull is carried out. The analysis of the com-
Numerical simulations have been carried out by means of puted hydrodynamic loads, for both the monohull and the
the Full Multi Grid-Full Approximation Scheme (FMG- catamaran will follow. Verification for resistance coeffi-
FAS) [17], with four grid levels. In this approximation cients is carried out; validation is made for those cases for
procedure, the solution is computed on the coarsest grid which experimental data are available.
level first; then it is approximated on the next finer grid and
the solution is iterated by exploiting all the coarser grid 5.1 Wave patterns
levels available with a V-cycle. The process is repeated up
to the finest grid level. On each grid level the iterative Wave patterns for both the monohull and the catamaran are
solution is carried on until the L2-norm of the residuals depicted in Fig. 3 for both values of the Froude number.
drops of four orders of magnitude and the variation on the When comparing the solutions for the monohull and the
forces and moment is at most on the fourth digit. catamaran, it can be seen that the outer wave field is only
slightly influenced by the presence of the twin hull, which,
on the contrary, strongly affects the wave pattern in the
5 Numerical results whole inner region. Because of the interference between
the transversal bow wave systems, a deep wave trough
In this section the numerical simulations carried out are appears; this becomes deeper and moves astern as the
presented: an analysis of the flow field around the cata- Froude number increases. It can be observed that the
maran, in terms of wave pattern, surface pressure, limiting resistance coefficient, as well as the attitude of the cata-
streamlines and velocity fields is shown first; then a maran (see, for example, Stern et al. [34], Broglia et al.

123
J Mar Sci Technol (2011) 16:254–269 261

Fig. 4 Wave profiles for the monohull (black) and the catamaran (inner in blue, outer in red) at Fr = 0.30 (left) and Fr = 0.45 (right), model
length Lpp = 3.0 m (color figure online)

Fig. 5 Wave cut along centerline for the catamaran (dashed line) and twice the monohull (solid line) at an equivalent distance. Left Fr = 0.30;
right Fr = 0.45 (model length Lpp = 3.0 m)

[5]), is directly related to the magnitude and the position of the wave profile on the outer side of the catamaran and of
this wave trough. Moreover, as the Froude number the monohull are close to each other, the major differences
increases, the resistance coefficient (in particular the being located around the longitudinal position of the wave
pressure contribution, as it will be also shown in the fol- trough in the inner side (i.e. around mid-ship for Fr = 0.30
lowing) increases; this happens up to values of the Froude and more astern for Fr = 0.45). As it will be shown in the
number for which the wave trough does not overtake the following, in this area a low pressure region appears and
stern of the catamaran. At very high Froude number (not induces a cross flow around the hull from the outer to the
computed in this work) the wave trough overtakes the stern inner side, related to which a lower wave in the outer
and the pressure resistance coefficient (and the total resis- region is observed (Fig. 4). The difference between the
tance coefficient) drops. Sinkage and trim are also strongly outer wave profiles for the catamaran and the monohull
related to the position of this wave trough and their maxima seems independent of the Reynolds number.
are attained when the wave trough reaches the stern (Stern On the other hand, wave interference is large in the inner
et al. [34], Broglia et al. [5]). region; as a matter of fact, differences between inner and
In Fig. 3, a strong interference between the stern wave outer wave profiles for the catamaran, as well as between
systems is also highlighted, the catamaran showing a much the catamaran inner profile and the monohull profile can be
higher rooster tail than the monohull. The computed clearly observed (Fig. 4). When inner/outer side profiles
solutions show that the transom stern is not completely dry for the catamaran are compared, a rather small asymmetry
and a small region of dead water has been observed for in the bow region can be inferred for both values of the
both values of the Froude number. As expected, the wave Froude number; this difference, such as the induced cross
pattern is only slightly affected by the Reynolds number, at flow, can be ascribed to the asymmetry of the pressure field
least for this range of values. between the inner and outer regions. Larger differences
appear starting from x/Lpp = 1/3, i.e. around the longitu-
5.2 Wave profiles and wave cuts dinal position where a large interference between the bow
wave systems in the inner region occurs.
Wave interference, as well as inner/outer asymmetry for In Fig. 5 wave cuts along the centre line for the cata-
the wave system can be analyzed by comparing wave maran at both speeds are compared with twice the longi-
profiles and wave cuts for the catamaran and the monohull. tudinal wave cuts for the monohull at an equivalent
In Fig. 4 both inner and outer wave profiles on the cata- distance; the fact that wave elevation for the catamaran is
maran and on the monohull are shown. As already noted, larger than twice the monohull, reveals a non-linear

123
262 J Mar Sci Technol (2011) 16:254–269

Fig. 6 Wave cuts at y = ±1/4H: catamaran internal (red) and external (blue), monohull (black). Left Fr = 0.30; right Fr = 0.45 (model length
Lpp = 3.0 m) (color figure online)

Catamaran Monohull Catamaran Monohull moves downstream; the same happens to the trough, which
lowers and moves amidships. The same situation is
observed for the stern wave systems, that grows and moves
downstream as the Froude number increases.
Similar conclusions can be drawn from the observation
of the non-centred wave cuts. In Fig. 6 internal and
external wave cuts for the catamaran and for the monohull
are compared at y = ±1/4H, where H is the distance
between the demihulls (Fig. 1). The external catamaran
and the monohull wave cuts are similar, the differences
being concentrated in the transom region and in the wake:
weak interaction effects between the twin hulls in the outer
region is, therefore, confirmed. When looking at the wave
cuts in the inner region and at an equivalent distance for the
monohull, it can be observed that the heights of the first
crest are rather similar, which indicates that the interaction
between the bow wave system has not completely occurred
yet. On the contrary, the difference for the following wave
trough clearly indicates a strong interaction of the trans-
versal waves; moreover, for the amplitude of the wave
for the catamaran, being greater than twice the one for
the monohull, unfavourable (i.e. positive) interference is
observed. The difference between the catamaran wave cuts
and the monohull increases with the value of the Froude
number.

Fig. 7 Surface pressure at Fr = 0.30 (left) and Fr = 0.45 (right); 5.3 Surface pressure, limiting streamlines
Lpp = 3.0 m and velocity field

interaction between the wave systems generating from the The pressure distribution on the hull surface is strongly
two hulls. This causes an increase in the (wave) resistance: affected by the interference phenomena; moreover, as it
the total resistance for the catamaran is larger than twice will be shown in this section, the resulting transversal
the resistance for the monohull and, consequently an pressure gradient drives a cross flow which cause an
unfavourable interference occurs. Moreover, by inspection increase of the total resistance. The pressure distributions
of these wave cuts, it can also be observed that the non- on the surface of the catamaran and the monohull are
linear interference increases when moving downstream, the shown in Fig. 7 for both values of the Froude number (in
difference between monohull and catamaran wave cuts at the figure c.p. denotes the position of the catamaran centre
the first crest being negligible at both Froude numbers. plane). It can be seen that the high pressure region is rather
Finally, the position and the heights of the troughs and small, the shape of the hull being streamlined and with a
the crests of the inner wave field can be highlighted: as the sharp trailing edge. Nevertheless, a port/starboard asym-
Froude number increases, the crest at the bow grows and metry at the bow appears, with a low pressure region in the

123
J Mar Sci Technol (2011) 16:254–269 263

inner side (which is lower for the higher Froude number). of the cross flow depends on the Froude number; at the
From the figures it is clear that, even for the highest Froude higher speed, the cross flow starts almost amidship and is
number considered, the minimum of the pressure occurs stronger. The resulting distortion of the flow is shown in
well upstream the stern; at higher Froude number, this Fig. 9, where the limiting streamlines on the hull are
minimum overtakes the stern, with a large variation on the depicted. A weak cross flow from inner to outer side occurs
attitude of the catamaran and a decrease of the total just after the forward perpendicular; this cross flow is
resistance coefficient. related to the wave height reduction in the outer region.
The high port/starboard asymmetry of the pressure field Downstream, the cross flow changes direction, being from
induces a cross flow (both from inner to outer side and vice the outer to the inner side; at the lower Froude number, the
versa) around the keel of the catamaran, that leads to an low pressure region is located almost amidships, and
increase of the total resistance. In Fig. 8, the lateral therefore, the transverse flow takes place just after the bow.
velocity contours on the vertical symmetry plane are pre- For the higher Froude number, the low pressure region is
sented; negative values represent an outward flow, whereas located almost at the stern and the transverse current
positive values an inward motion. As it can be clearly seen, toward the centre plane starts around amidships. This cross
the cross flow is toward the outer side after the bow and flow (like the pressure distribution) seems not to depend
from the outer to the inner side downstream, where the on the Reynolds number. This can be seen from the
wave interaction generates a low pressure region between cross sections of the axial velocity contours (placed at
the twin hulls. The position, the extension and the strength x/Lpp = 0.20, 0.50, 0.85 from the fore perpendicular)
reported in Fig. 10, for Fr = 0.30 and Fr = 0.45. In these
pictures, for the sake of clarity, only three contour lines
(u = 0.70, u = 0.90 and u = 0.99) are shown for the
Reynolds numbers investigated. Because of the outward
transverse flow at x/Lpp = 0.20 the inner boundary layer is
thinner, and the contour lines are bulged outward. At the
higher Froude number the distortion of the contour lines at
this cross section is less evident; this is because, as already
explained, the low pressure region, and therefore, the
transversal flow, moves downstream. In this section, very
close to the bow, the dependence on the Reynolds number
is negligible. The distortion of the contour lines is clearly
Fig. 8 Lateral velocity contours on the symmetry vertical plane of observable for both Froude numbers from section x/Lpp =
the demihull. Top Fr = 0.30; bottom Fr = 0.45. Lpp = 3.0 m 0.20. For the higher values of Fr, the transversal flow is

Fig. 9 Limiting streamline at Catamaran Monohull Catamaran Monohull

Fr = 0.30 (left) and Fr = 0.45
(right), Lpp = 3.0 m. Top
pictures front view; bottom
pictures view from astern

123
264 J Mar Sci Technol (2011) 16:254–269

Fig. 10 Longitudinal velocity Monohull Catamaran Monohull Catamaran

contours on three cross sections
and for different model length.
Left Fr = 0.30; right Fr = 0.45

Fig. 11 Total CT, viscous CV and pressure CP resistance coefficients versus Reynolds number

still toward the outer region and consequently the boundary 5.4 Global loads and interference factor
layer thickness in the inner side is thicker than on the outer
side. The situation for the lower Fr is different: from the The computed resistance coefficients, as a function of the
lateral velocity contours (Fig. 8) and the limiting stream- Reynolds number, are shown in Fig. 11, where the
lines (Fig. 9) it is evident that, in this region, the cross flow extrapolated values are reported for both the catamaran and
is toward the centre plane; as a matter of fact, the contour the monohull (multiplied by two); the total resistance
line for u = 0.9 shows an inward distortion. However, the coefficient (CT = R/0.5qU2L2pp) is decomposed into its
boundary layer thickness (at least above half the drought) is pressure (CP) and viscous (CV) contributions. As expected,
still thicker on the outer side than on the inner side. Indeed, the viscous component decreases with the Reynolds num-
the cross flow has not yet recovered the deformation ber, whereas pressure contribution is practically constant
attained in the fore part of the hull. Moving downstream with Re; therefore, a global decrease for the total resistance
(section at x/Lpp = 0.85), due to the strong transverse flow coefficient with the Re is observed. For the low Fr test, a
toward the centre plane, the contour lines are definitely slight increase of the viscous resistance coefficient at the
bulged inward and the boundary layer is clearly thicker highest Reynolds number is observed; this unexpected
inside than outside. At the last two cross sections, Reynolds behaviour is probably due to a poor grid resolution, and
number dependence is quite evident, the boundary layer therefore, further investigations are required. It is inter-
being thinner as the Reynolds number increases; however, esting to note that, for a fixed model length, the viscous
since the shape of the contour lines are similar, it seems resistance coefficient decreases with the Froude number,
that the cross flow does not have a strong dependency on whereas the pressure contribution for the high Fr is around
the Reynolds number. three times larger.

123
J Mar Sci Technol (2011) 16:254–269 265

Comparing the resistance coefficients for the catamaran catamaran, for this separation and for this range of Froude
and the monohull, it is evident that the total resistance numbers. Different results have been observed for other
coefficient for the catamaran is larger than twice the catamaran and trimaran geometry, and with a clear
monohull; this means that at both Froude number and for dependence on the separation distance of the hulls (see for
the whole Reynolds number range, an unfavourable inter- example Pinto et al. [31]). Non zero viscous interference is
ference between the twin hulls takes place. This behaviour expected for geometry where either the change in the
is the result of the flow field distortion, the more pro- wetted surface is large or where flow separation occurs, or
nounced wavemaking capability and the change in the also when, due to the cross flow, a large amount of vorticity
pressure distribution on the demihulls, which have been is shed from the keel.
shown in the previous sections. In summary, low values for the total interference are
By inspection of the viscous and the pressure contribu- computed; by comparison with previous works (see for
tions, it can be observed that this increase is due to the example Souto et al. [32]), it can be observed that the value
pressure contribution, the viscous part being practically the obtained for the low Froude number case seems to be in
same (similar results have been shown for the BIW-A agreement with other authors, whereas the value provided
catamaran by Stern et al. [34]). Larger differences in the for the higher Froude number seems too low. These results
viscous resistance coefficient between the mono- and the could be influenced by the initial hypothesis of fixed trim
multihulls were to be expected only if the differences in the and sinkage for the catamaran and the monohull. In fact, it
wetted surface were relevant, or if other viscous phenom- has been shown that, for Froude numbers less than 0.45, the
ena, such as flow separation or detachment of vorticity difference in the attitude between the monohull and cata-
from the keel, had appeared, which is not the case. maran configurations is negligible, whereas larger differ-
In Fig. 12 the interference factor, defined as Ix ¼ ences are measured for Froude numbers from 0.45 (Souto
ðCxCAT  2CxMONO Þ 2CxMONO (where x denotes any given et al. [32], Broglia et al. [5]). At this Froude number, both
resistance component, either total, pressure or viscous), is sinkage and trim of the catamaran are normally larger than
reported as function of the Reynolds number. The inter- those for the monohull in the same conditions; this lead to
ference factor for the total resistance is almost constant an increase in both the viscous (for the increase of the
with the Reynolds number, whereas it increases with the wetted surface) and pressure (due to the augmented wave
Froude number; for Fr = 0.30 the value is around 4%, making capability) resistance coefficient, and therefore to
while for Fr = 0.45 it is around 10%. At low speed, the higher values for the interference factor. By an a posteriori
viscous interference factor is practically negligible, analysis of the unbalance on forces and moments and the
whereas negative values (i.e. favourable interference), consequent change in attitude from hydrostatic consider-
increasing in magnitude with the Reynolds number is ations, the differences in sinkage and trim between the
observable at higher speed. For the pressure interference monohull and the catamaran has been confirmed. More-
factor the behaviour with the Reynolds number is not clear; over, it is also confirmed that the difference is larger at
IP can be roughly estimated around 20 and 30% for the higher Froude number.
lower and the higher value of the Froude number,
respectively. 5.5 Verification and validation
On the other hand, an almost negligible value for the
viscous interference factor is computed, that suggests that Verification, i.e. the evaluation of the order of convergence
viscous effects have a very small influence on the inter- and the assessment of numerical uncertainty, has been
ference phenomena. This is true, of course, only for this carried out for the total resistance coefficient (as well as for

Fig. 12 Interference factors;

total IT, viscous IV and pressure
IP versus Reynolds number

123
266 J Mar Sci Technol (2011) 16:254–269

Table 4 Simulation uncertainty

Lpp (m) 103S3 103S2 103S1 pG 103S UG (%S1)
for total, viscous and pressure
resistance coefficients Total resistance coefficient CT
(Fr = 0.30)
1.5 1.593 1.384 1.341 2.28 1.327 0.24%
3.0 1.485 1.270 1.215 1.97 1.197 0.05%
6.0 1.449 1.162 1.092 2.04 1.069 0.07%
12.0 1.462 1.151 1.001 1.05 0.951 13.96%
20.0 1.286 1.179 1.023 -0.54 1.023 –
Viscous resistance coefficient CV
1.5 1.207 1.096 1.087 3.62 1.084 0.48%
3.0 1.121 0.985 0.965 2.77 0.958 1.02%
6.0 1.091 0.880 0.854 3.02 0.845 1.60%
12.0 1.106 0.871 0.767 1.18 0.732 10.76%
20.0 0.929 0.893 0.778 -1.68 0.778 –
Pressure resistance coefficient CP
1.5 0.385 0.288 0.254 1.51 0.243 7.22%
3.0 0.364 0.285 0.250 1.17 0.238 11.14%
6.0 0.358 0.282 0.238 0.79 0.223 25.42%
12.0 0.356 0.280 0.235 0.76 0.220 27.80%
20.0 0.357 0.286 0.245 0.79 0.231 22.87%

Table 5 Simulation uncertainty

Lpp (m) 103S3 103S2 103S1 pG 103S UG (%S1)
for total, viscous and pressure
resistance coefficients Total resistance coefficient CT
(Fr = 0.45)
1.5 1.794 1.785 1.774 -0.29 1.774 –
3.0 1.658 1.641 1.649 0.00 1.649 –
6.0 1.665 1.494 1.464 2.51 1.454 0.93
12.0 1.572 1.521 1.355 -1.70 1.355 –
Viscous resistance coefficient CV
1.5 1.181 1.076 1.058 2.54 1.052 0.78
3.0 1.081 0.955 0.929 2.28 0.920 0.21
6.0 1.111 0.852 0.806 2.49 0.791 2.57
12.0 1.029 0.906 0.727 -0.54 0.727 –
Pressure resistance coefficient CP
1.5 0.613 0.709 0.716 3.78 0.718 0.57
3.0 0.577 0.686 0.719 1.72 0.730 0.46
6.0 0.554 0.640 0.656 2.43 0.661 1.07
12.0 0.543 0.613 0.623 2.81 0.626 0.80

its viscous and pressure components) and for a longitudinal observed, it is considered ‘‘confident’’ when 0.7 B CG B
wave cut. The analysis has been carried out at both Froude 1.3, and ‘‘lack of confidence’’ otherwise, CG being the
numbers, and for the whole Reynolds number range correction factor; the extrapolated value and grid uncer-
investigated. Results are summarized in Tables 4, 5 and 6 tainty (expressed in percentage of the extrapolated value)
and in Fig. 13. are computed following again the procedure described in
The analysis for the resistance coefficients has been [35, 41]. When oscillatory convergence or divergence is
performed following the procedure described in Stern et al. observed, uncertainty is not estimated and the computed
[35] and Wilson et al. [41]. Since negligible iterative values on the finest mesh are reported.
uncertainty has been observed, grid uncertainty is consid- At the lower Froude number, the estimated order of
ered as the only contribution to the numerical uncertainty. accuracy is close to the theoretical value of two for the total
As suggested in [35, 41], when monotonic convergence is resistance coefficient of the models with length up to 6 m;

123
J Mar Sci Technol (2011) 16:254–269 267

Table 6 Validation for the predicted total resistance coefficients

Fr Re 103S TU Delft CNR-INSEAN USN (%S)
3 3
10 D e (%D) 10 D e (%D)

0.30 4.882 9 106 1.197 1.072 11.63 1.111 7.83% 0.05%

0.45 7.322 9 106 1.649 1.560 5.71 1.622 1.63% –

Fig. 13 External wave cuts at y/b = 1.146 from the centre of the hull; solid line, numerical results; error bar, numerical uncertainty;
experiments, symbols. Left Fr = 0.30; right Fr = 0.45 (model length Lpp = 3.0 m)

moreover, low numerical uncertainty is estimated for these reported. At Fr = 0.45 the agreement with experiments is
computations. For the longer models a degradation of the good with both TU Delft data and CNR-INSEAN mea-
convergence is observed (for the longest one a divergent surements, the error being less than 6 and 2%, respectively.
condition is shown); the numerical uncertainty is conse- For the lower Froude number the agreement is less satis-
quentially high (around 14%). The actual order of accuracy factory, the error being around 12 and 8% with respect to
for the prediction of the viscous and pressure contributions TU Delft and CNR-INSEAN data, respectively. Unfortu-
is somewhat far from the theoretical value. A significant nately, uncertainty analysis for experimental data being not
increase of the numerical uncertainty is observable at available, validation could not be completed.
higher Reynolds number for the pressure resistance coef- The comparison between the numerical and the experi-
ficient. This seems to indicate that the grid sequence is not mental (from [39]) longitudinal wave cuts is shown in
in the asymptotic range; computations on finer grid could Fig. 13; the wave cut is taken in the outer region at a
answer these issues. distance of y/b = 1.146 from the longitudinal axis of the
At the higher Froude number, the estimated order of demihull. In the same figure numerical uncertainty is also
accuracy is generally satisfactory for the computation of reported. As a general comment, the agreement between
the pressure and the viscous coefficients, at least for the computed and measured data, for both Froude numbers, is
lower Reynolds numbers (i.e. with the model length up to rather satisfactory; larger discrepancies can be observed in
6 m). The measured convergence order (pG) for the total the far wake (larger for the lower Froude number) where
resistance coefficient is far from the theoretical value, the grid resolution in the longitudinal direction is
divergent or oscillatory solutions being attained for it. decreasing. Anyhow, the difference between numerical and
However, for the cases where monotonic convergence is experimental data is within the numerical uncertainty (or at
achieved, the resulting grid uncertainty is generally small; least of the same order) almost everywhere. The observed
and for the cases where oscillatory or divergence condition convergence order is around one for both Froude numbers
is observed, it can be noted that the difference between the if measured in L2-norm; at the crest and trough, local
values obtained on the medium and the finest meshes convergence order is close to two. However, as for the
(which can be considered as an estimation of the grid resistance, since the uncertainty analysis for experimental
uncertainty) is rather small. Large variation is noted for the wave cuts is not available, the validation cannot be
computation at the higher Reynolds number. completed.
Validation has been made by comparison with experi-
mental data from [39] and from calm water tests carried out
at CNR-INSEAN [5]; both experimental campaigns were 6 Conclusions
performed for a model with Lpp = 3 m. The comparison, in
terms of the total resistance coefficient, is shown in The simulations of the flow around the high-speed Delft
Table 6, where simulations at both Froude numbers are catamaran model 372 have been carried out by means of

123
268 J Mar Sci Technol (2011) 16:254–269

the numerical solutions of the RANS equations. By the 3. Broglia R, Di Mascio A, Amati G (2007) A parallel unsteady
analysis of the computed flow field, relevant interference RANS code for the numerical simulations of free surface flows.
In: Proceedings of 2nd international conference on marine
effects such as cross flows and distorted streamlines along research and transportation, Ischia, Naples, Italy
the hull surface, port/starboard side asymmetry of the 4. Broglia R, Di Mascio A, Muscari R (2007) Numerical study of
pressure filed on each hull from bow to stern, and inner/ confined water effects on a self-propelled submarine in steady
outer wave patterns and profiles asymmetry have been manoeuvres. Int J Offshore Polar Eng 17:89–96
5. Broglia R et al (2011) Calm water tests for the DELF 372 Cat-
shown. The attention has also been devoted to the scale amaran model at several hull separations. CNR-INSEAN Report
(Reynolds number) effects on the flow field in general, and (in preparation)
on the interference in particular. To this aim, solutions at 6. Campana E, Peri D, Pinto A, Tahara Y, Kandasamy M, Stern F,
different Re in the range between *106 to *108 and for Cary C, Hoffman R, Gorski J, Kennel C (2006) Simulation based
design of fast multihull ships. In: Proceedings of 26th symposium
two different Froude numbers have been computed. The on naval hydrodynamics, Rome, Italy
analysis of the results has shown that the dependence on 7. Castiglione T, Stern F, Kandasamy M, Bova S (2009) Unsteady
the Reynolds number is rather weak. RANS simulation for a catamaran advancing in regular waves. In:
Validation has been also carried out for resistance as well Proceedings of the 10th international conference on fast and sea
transportation, FAST 2009, Athens, Greece, October 2009
as for a longitudinal cut. For resistance, satisfactory con- 8. Chorin A (1967) A numerical method for solving incompressible
vergence properties have been shown for the low Froude viscous flow problems. J Comp Phys 2:12–26
number case for the total resistance coefficient, and for the 9. Colicchio G, Colagrossi A, Lugni C, Faltinsen OM (2005)
viscous resistance coefficient at high Froude number, up to Experimental and numerical investigation of a trimaran in calm
water. In: Proceedings of 8th International Conference on Fast
medium/high Reynolds number; a degradation of the con- Sea Transportation, Sain Petersburg, Russia, June 2005
vergence is noted for the other resistance components. 10. Couser PR, Molland AF, Armstrong NA, Utama IKAP (1997)
Numerical uncertainty is generally rather small, apart for Calm water powering predictions for high-speed catamarans.
the higher Reynolds number and for the pressure coefficient FAST 1997, Sydney, Australia, July 1997
11. Di Mascio A, Broglia R, Favini B (2001) A second order godu-
at the lower Froude number. The agreement with available nov-type scheme for naval hydrodynamics. In: Godunov meth-
experimental data is rather good for the higher Froude ods: theory and applications. Kluwer Academic/Plenum
number, while is less satisfactory for the lower Froude Publishers, Dordrecht/New York, pp 253–261
number. Validation for the wave cuts has also shown sat- 12. Di Mascio A, Broglia R, Muscari R (2007) On the application of
the one-phase level set method for naval hydrodynamic flows.
isfactory convergence properties, low uncertainty and a Comput Fluids 36(5):868–886
rather good agreement with available experimental data. 13. Di Mascio A, Broglia R, Muscari R (2009) Prediction of
This work has been done in the framework of an hydrodynamic coefficients of ship hulls by high-order Godunov-
ongoing NICOP collaboration project between CNR-INS- type methods. J Mar Sci Tech 14(1):19–29
14. Di Mascio A, Muscari R, Broglia R (2006) An overlapping grids
EAN and IIHR. The activities scheduled for this project approach for moving bodies problems. In: Proceedings of 16th
include experimental tests in both calm water and in waves, international offshore and polar engineering conference, San
as well as complementary numerical simulations. An Francisco, California (USA)
experimental and numerical analysis of the demihull sep- 15. Doctors L (2003) The influence of viscosity on the wavemaking
of a model catamaran. In: Proceedings of 18th IWWWFB, Le
aration effects on the interference and in general on the Croisic (France), April 2003
flow field around the Deft 372 catamaran will be the topic 16. Doctors L, Scrace R (2003) The optimization of trimaran sidehull
of future activities. position for minimum resistance. In: Proceedings of the FAST
2003, Ischia, Italy
Acknowledgments This research was supported by the U.S. Office 17. Favini B, Broglia R, Di Mascio A (1996) Multi-grid acceleration
of Naval research, through Dr. L. Patrick Purtell in the framework of of second order ENO schemes from low subsonic to high
the NICOP project ‘‘Complementary EFD and CFD Analysis of Calm supersonic flows. Int J Numer Methods Fluids 23:589–596
Water Hydrodynamics and Large Amplitude Motion for High-Speed 18. Godunov SK (1959) A finite difference method for the numerical
Catamarans’’, grant N00014-08-1-1037. Numerical computations computation of discontinuous solutions of the equations of fluid
presented here have been performed on the parallel machines of dynamics. Mat Sbornik 47:271
CASPUR Supercomputing Center (Rome); their support is gratefully 19. Harten A, Engquist B, Osher S, Chakravarthy SR (1987) Uni-
acknowledged. formly high order accurate essentially non-oscillatory schemes.
J Comp Phys 71:231–303
20. Insel M, Molland AF (1992) An investigation into the resistance
components of high speed displacement catamarans. Trans RINA
References 134:1992
21. Lugni C, Colagrossi A, Landrini M, Faltinsen OM (2004)
1. Armstrong T (2003) The effect of demihull separation on the Experimental and numerical study of semi-displacement mono-
frictional resistance of catamarans. In: Proceedings of the FAST hull and catamaran in calm water and incident waves. In: Pro-
2003, Ischia, Italy ceedings of 25th symposium on naval hydrodynamics, St. John’s,
2. Beam RM, Warming RF (1978) An implicit factored scheme Canada, August 2004
for the compressible Navier–Stokes equations. AIAA J 22. Maki K, Doctors L, Rhee S, Wilson W, Beck R, Troesch A
16:393–402 (2007) Resistance predictions for a high-speed sealift trimaran.

123
J Mar Sci Technol (2011) 16:254–269 269

In: Proceedings of 9th international conference on naval ship 34. Stern F, Carrica P, Kandasamy M, Gorski J, O’Dea J, Hughes M,
hydrodynamics, Ann Arbor, Michigan, USA, 2007 Miller R, Hendrix D, Kring D, Milewski W, Hoffman R, Cary C
23. Merkle CL, Athavale M (1987) Time-accurate unsteady incom- (2006) Computational hydrodynamic tools for high-speed trans-
pressible flow algorithm based on artificially compressibility. ports. In: Transaction SNAME, vol 114
AIAA paper, pp 87–1137 35. Stern F, Wilson RV, Coleman HW, Paterson E (2001) Compre-
24. Miller R, Gorski J, Xing T, Carrica P, Stern F (2006) Resistance hensive approach to verification and validation of cfd simula-
prediction of high speed mono and multihull ships with and tions—Part 1: methodology and procedures. J Fluid Eng
without propulsors using URANS. In: Proceedings of 26th sym- 123:793–802
posium on naval hydrodynamics, Rome, Italy, 2006 36. Sussman M, Smekerda P, Osher SJ (1994) A level set approach
25. Millward A (1992) The effect of hull separation and restricted for computing solutions to incompressible two-phase flow.
water depth on catamaran resistance. Trans RINA 134:341–349 J Comput Phys 114:146–159
26. Molland A, Wellicome J, Couser P (1996) Resistance experi- 37. Tarafder S, Suzuki K (2008) Wave-making resistance of a cata-
ments on a systematic series of high speed catamaran forms: maran hull in shallow water using potential-based panel method.
variation of length-displacement ratio and breadth-draught ratio. J Ship Res 52(1):16–29
Trans Royal Inst Naval Archit 138:59–71 38. Van Leer B (1979) Towards the ultimate conservative difference
27. Molland AF, Wilson PA, Taunton DJ, Chandraprabha S, Ghani scheme V. A second-order sequel to Godunov’s method. J Com-
PA (2004) Resistance and wash wave measurements on a series put Phys 32:101–136
of high speed displacement monohull and catamaran forms in 39. Van’t Veer R (1998a) Experimental results of motions and
shallow water. Int J Marit Eng (Trans RINA) 146(A2):19–38 structural loads on the 372 catamaran model in head and oblique
28. Moraes HB, Vasconcellos JM, Latorre RG (2004) Wave resis- waves. TU Delft report, N.1130
tance for high speed catamarans. Ocean Eng 31:2253–2282 40. Van’t Veer R (1998b) Experimental results of motions, hydro-
29. Muscari R, Di Mascio A (2005) Simulation of the flow around dynamic coefficients and wave loads on the 372 Catamaran
complex hull geometries by an overlapping grid approach. In: model. TU Delft report, N.1129
Proceedings of 5th Osaka Colloquium, Osaka, Japan 41. Wilson RV, Stern F, Coleman HW, Paterson E (2001) Compre-
30. Osher S, Sethian JA (1988) Fronts propagating with curvature- hensive approach to verification and validation of CFD simula-
dependant speed: algorithms based on Hamilton–Jacobi formu- tions—Part 2: application for RANS simulation of a cargo/
lations. J Comput Phys 79:12–40 container ship. J Fluid Eng 123:803–810
31. Pinto A, Broglia R, Campana E, Di Mascio A, Peri D (2006) 42. Zlatev Z, Milanov E, Chotukova V, Sakamoto N, Stern F (2009)
RANS analysis of the interference effect in catamarans. In: Combined model-scale EFD-CFD investigation of the maneu-
Proceedings of 9th NuTTS, Le Croisic, France vering characteristics of a high speed catamaran. In: Proceedings
32. Souto-Iglesias A, Zamora-Rodrı́guez R, Fernández-Gutiérrez D, of the 10th International Conference on fast and sea transporta-
Pérez-Rojas L (2007) Analysis of the wave system of a catamaran tion, FAST 2009, Athens, Greece, October 2009
for CFD validation. Exp Fluids 42:321–333
33. Spalart PR, Allmaras SR (1994) A one-equation turbulence
model for aerodynamic flows. La Recherche Aérospatiale 1:5–21

123

View publication stats