Ocean Engineering xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Analysis of the asymmetric behavior of propeller–rudder system of twin

screw ships by CFD
⁎
Roberto Muscaria, , Giulio Dubbiosoa, Michele Vivianib, Andrea Di Mascioc
a
CNR-INSEAN, via del Vallerano 139, 00128 Roma, Italy
b
DINAEL, via Montallegro 1, 00128 Genova, Italy
c
CNR-IAC, via dei Taurini 139, 00128 Roma, Italy

A R T I C L E I N F O A BS T RAC T

Keywords: The interference between the hull, propeller and rudder remarkably affects the control and maneuvering
Computational fluid dynamics capabilities of marine vehicles. In case of twin screw/twin rudder ships, the asymmetric evolution of the wake
Overlapping grids past the hull causes the asymmetric functioning of the propeller–rudder system. Systematic investigations on
Ship maneuvering this aspect for twin screw ships are limited. Available experimental data carried out on simplified hull–
Propeller–rudder interaction
propeller–rudder system and captive model tests do not allow to completely understand the fluid mechanism at
Flow straightening effect
the basis of the hydrodynamic interaction that should be taken into account in simplified maneuvering
mathematical models for preliminary predictions. In this paper the hull–propeller–rudder interactions
phenomena for a twin screw/twin rudder model are investigated by URANS simulations, with a particular
focus on the asymmetry of the propeller–rudder system. To this aim, captive model tests consisting of pure
rudder and coupled drift–yaw motions corresponding to the steady phases of turning circle maneuvers at
different rudder angles (δ = 15° ÷ 35°) are performed at the speed correspondent to Fr=0.265. Moreover, a free
running maneuvering simulation is also performed to gain more insight on the transient phase of the maneuver.
An identity rudder lift methodology is applied to synthesize the hull–propeller–rudder interactions by means of
a flow straightening coefficient; the analysis highlights that these effects are weak and invariant with respect to
the rudder angle on the windward shaft, whereas on the leeward side these effects are extremely sensitive to the
evolution of the hull and propeller wake.

1. Introduction literature is poor of dedicated studies, in particular for twin screw

configurations, where the propeller–rudder system inevitably operate
Marine rudders operate in a complex flow region that, during in a completely different flow field during a maneuver: the leeward side
maneuvers, is characterized by massive separation and by swirled and is highly perturbed by the wake of the hull and the presence of strong
accelerated flow induced by the action of the propeller. The correct vortices detached from the appendages, whereas the windward side is
prediction of these complex interferences is necessary to estimate the characterized by non negligible cross flow that leads the system to be
control force exerted by the rudder and, consequently, to predict the impinged by a flow at high incidence. Obviously, the resultant asym-
control and maneuvering capabilities of a ship. Due to the complexity metric loading of the propellers amplifies the asymmetric inflow
of the problem, traditional approaches based on semi–empirical condition determined by the hull. Although the crucial aspects of the
mathematical models are still fashionable in the early design stage propeller–rudder systems have been thoroughly documented in many
due to their low computational demand. In their modern physical– experimental works carried out for isolated and basic ship configura-
based structure, the modeling effort is devoted to describe the tions (Molland and Turnock, 2006) or by systematic analysis through
propeller–rudder system by traditional actuator disk and airfoil constrained model tests (Hirano, 1980; Inoue et al., 1981a, 1981b), a
theories, properly coupled, by means of interference factors, with a reliable generalized model is not yet available, and ad hoc tuning of the
separate description of the hull in terms of hydrodynamic coefficients. parameters is still required to minimize the discrepancies between
The drawback of these tools lies in the lack of empirical regressions that simulations and experimental data (either in model or full scale). The
provide reliable estimation of these parameters to a broader set of ship data provided in Molland and Turnock (2006) and relative simulations
geometries and propeller–rudder configurations. In this regard, the performed in Badoe et al. (2015) are an excellent starting point for

⁎
Corresponding author.
E-mail address: [email protected] (R. Muscari).

http://dx.doi.org/10.1016/j.oceaneng.2017.07.056
Received 15 February 2017; Received in revised form 31 May 2017; Accepted 27 July 2017
0029-8018/ © 2017 Elsevier Ltd. All rights reserved.

Please cite this article as: Muscari, R., Ocean Engineering (2017), http://dx.doi.org/10.1016/j.oceaneng.2017.07.056
R. Muscari et al. Ocean Engineering xxx (xxxx) xxx–xxx

Nomenclature D Propeller diameter

FX − R, FY − R,
List of symbols MZ − R Axial force, lateral force and yaw moment of rudder
G Center of gravity of the model
α Rudder geometric angle of incidence J Advance coefficient of the propeller
αeff Rudder effective angle of incidence KT Propeller thrust coefficient, KT = T /(ρN2D 4 )
βH Geometric drift of the hull lR Rudder arm (with respect to G )
βR Geometric drift of the rudder L pp Model length between perpendiculars
Δu Increment of the flow speed in the slipstream (momentum N Propeller rate of revolutions
theory) t Non–dimensional time
δ Nominal deflection of the rudder u′, v′, r′ Velocities of the model in its frame of reference
γ Flow straightening coefficient V0 Approach speed of the model
A0 Rudder area out from the propeller slipstream Vw Mean velocity in the wake
Aapp Plane area of all appendages Vrace Flow speed in the propeller wake at rudder location
Arace Rudder area covered by the propeller slipstream VTOT Total velocity at the rudder
Arud Plane area of the rudder x′, y′, z′ Model frame of reference
Askeg Plane area of the skeg X , Y , Z Inertial frame of reference
CP Center of pressure of the rudder

model development; nevertheless, the latter data and derived regres- propeller–rudder interaction during steady and unsteady maneuvers
sions are specific to the adopted hull shape and should be used is affected by the propeller loading condition and the attitude of the
carefully for different hull shapes. In addition, due to the particular model. Moreover, the drawbacks of simplified propeller–rudder mod-
facility (wind tunnel in the case of Molland and Turnock (2006)), the els to account for these effects are critically discussed.
data are not reliable for the actual incidence typical of propeller–
rudder system during a realistic maneuver that, in case of the tightest 2. CFD solver
turning rate, may be as high as 30° (twice the value considered in the
experiments and related numerical computations). Moreover, the The CFD solver χnavis is a general purpose simulation code
analysis of the interference effect is often indicative and must be developed at CNR–INSEAN. The code solves the Navier-Stokes
carefully considered due to the strong sensitivity to the local geometric Equations for unsteady high Reynolds number free surface–flows
characteristic of the stern and propeller–rudder system. around complex geometries. The main features of the solver are only
The present work aims to gain a deeper insight into the propeller– briefly recalled here; the interested reader is referred to (Di Mascio
rudder system of a fast twin screw / twin rudder configuration and to et al., 2014, 2007; Dubbioso et al., 2013; Muscari et al., 2013, 2011) for
bridge the gap with the open issues raised by the experimental and details. The solver is based on a cell-centered finite volume formula-
numerical analysis described in Coraddu et al. (2013), Dubbioso tion. The spatial discretization of the convective terms can be done by
(2015). In particular, Coraddu et al. (2013) showed by means of free second, third or fourth order approximation; in the present simula-
running tests a marked asymmetric overloading of the propellers, with tions, a third order upwind biased scheme was adopted. The diffusive
an increase of shaft torque up to 100% of the value in the straight terms are discretized with second order centered scheme, whereas time
course. In order to better investigate the effects of the asymmetric derivatives are approximated by a second order implicit formula (three
propeller behavior on the rudder performance and, consequently, the points backward). The solver being fully implicit, the divergence–free
maneuverability behavior of twin screws/twin rudders configurations, solution at each time step is computed iteratively by a pseudo-time
a further series of experiments was carried out for the same model, integration; convergence to steady state in pseudo–time is accelerated
with the rudder instrumented by a torque meter (Dubbioso, 2015). In by an Euler implicit scheme with approximate factorization, local
this work, the measurement of rudder torque showed an asymmetric pseudo–time step and multi-grid iteration. Turbulent stresses are
behavior of the internal and external rudder; however, the rudder taken into account by the Boussinesq hypothesis; several turbulence
torque alone is not sufficient to synthesize the rudder performance models (both algebraic and differential) are implemented. In the
because of its dependency on both the side force and lever arm with present computations, the one–equation model originally introduced
respect to the rudder stock. A preliminary analysis by means of CFD by Spalart and Allmaras (1994) has been used. Complex geometries
simulations on the isolated propeller–rudder system and a simplified and multiple bodies in relative motion are handled by a dynamical
maneuvering mathematical model, showed that this behavior could be overlapping grid approach (Di Mascio et al., 2006). High performance
ascribed to an asymmetric flow straightening. In particular, the leeward computing is achieved by an efficient shared and distributed memory
rudder experiences a flow with an incidence close to the nominal parallelization.
rudder angle δ, whereas on the windward side the straightening of the
flow was negligible.
3. Test case, numerical set–up and computational mesh
In order to analyze the asymmetric behavior of the rudders,
numerical simulations were carried out by CFD on the fully appended
The simulations were carried out for the twin screw model considered
model, for both captive and free running maneuvers. The constrained
in Ortolani et al. (2015a, 2015b), Dubbioso et al. (2017), whose principal
tests consisted of pure rudder simulations at different propeller loading
geometric characteristics are listed in Table 1. The geometry of the model
and circular motion tests with the same kinematic conditions (drift
is represented in Fig. 1. The model is characterized by a pram stern and
angle and yaw rate) of the steady phase of the turning circle maneuvers
equipped with a medium size skeg that guarantee good course keeping
at three different rudder angles (δ = 15°, 25°, 35°). In addition, a free
qualities (the plane area of the skeg, Askeg, is almost equal to the sum of
running maneuver was performed at the highest rudder angle
rudder areas). The rudders are outward misaligned with respect to the
(δ = 35°), to investigate the complexity of the propeller–rudder inter-
propeller axis and are inclined by about 10° with respect to the vertical
action during the transient phase.
plane. Moreover, the propeller shafts are inclined longitudinally by about
The analysis highlights that the asymmetric behavior of the
6° with respect to the horizontal plane.

2
R. Muscari et al. Ocean Engineering xxx (xxxx) xxx–xxx

Table 1
Geometric characteristics of the ship and propeller model. B is the breadth of the model,
• pure rudder tests: the model is set at zero drift angle βH and only the
rudder is deflected. The simulations are performed at propeller
T is the draught.
loadings KT / J 2 similar to those experienced by the external shaft
MODEL DATA during the steady turning maneuvers. In these tests the reference
velocity is reduced up to 0.70 Fr and the propeller loads are derived
L pp /B 7.5 from the open water curves at these speeds.
B/T
CB
3.25
0.5
• steady turning tests: the computational set–up is the same as
adopted in Dubbioso et al. (2017) with the rudder included; the
Arud /(L pp T ) 0.0157
Askeg /(L pp T ) 0.031
attitude and kinematic of the model is prescribed according to the
Aapp /(L pp T ) 0.15 experiments (see Table 2 in Dubbioso et al. (2017)). In the second
column of Table 3 the propeller loading refers to the internal
propeller, whereas the value for the external propeller is written in
PROPELLER DATA
brackets.
Number of blades
Pitch ratio
5
1.35
• free running maneuver: the maneuvering response of the model is
computed by the coupled Navier–Stokes and rigid body motion
Expanded area ratio 0.79
equations. The model is first accelerated by a fictitious force to about
Hub ratio 0.250
the approach velocity with the propellers switched–off. Then, the
propellers are activated and the model stabilizes to the self propul-
sion condition at the rectilinear condition; finally, the rudders are
activated and the maneuver starts.

The forces and moments are computed in the inertial reference

frame and are converted to a frame fixed at the center of gravity G of
the ship, where the x′ axis is directed forward, the y′ axis is directed to
port–side and the z′ axis upwards (the two frames coincide at the start
of the simulation); the rotations (specifically, yaw and rudder) follow
the right hand rule (see Fig. 2). All the maneuvers were carried out with
negative rudder angle δ. For the circular motion tests and the free
running maneuver δ < 0 corresponds to positive yaw rate and the
model turns to portside; consequently, the port and starboard rudder
are located in the internal (leeward) and external (windward) side. The
rudder characteristics by the pure rudder tests are specified for the
starboard rudder (windward/external); therefore, the values for δ > 0
correspond to the port rudder (leeward/internal) with opposite sign. If
not otherwise specified, in case of the prescribed and free turning
simulations the rudder angles are reported in tables and figures in
absolute value. The forces on the hull, propellers and rudders are
Fig. 1. Geometry of the model.
identified with subscripts H, P and R, respectively Fig. 3.
The computational mesh is the same as considered in Dubbioso
et al. (2017), with the inclusion of the all–movable rudders. The
domain has been discretized by 186 body–fitted patched and over-
lapped blocks, for a total of about 12 M cells. Grid distribution is such 3.1. Propeller modeling
that the thickness of the first cell on the wall is always below one in wall
unit, and at least 30 cells are within the boundary layer thickness In all numerical simulations the propeller is modeled by a body
(y+ = O (1), with Δ/ L pp = O (20/ Re), Δ being the thickness of the cell force approach. The propeller thrust and torque were set equal to the
adjacent to the wall, Lpp the model length between perpendiculars). measured values, KT and KQ , and were distributed over the actuator
Details of the cells distribution are listed in Table 2, where the “rudder disk as additional momentum sources in the Navier–Stokes equations
background” is made of two refinement blocks surrounding the rudders following two different approaches. In case of the constrained man-
and is used to provide a smooth transition from the very fine mesh euvers (pure rudder tests and coupled drift–yaw motions), the
around the rudders and the coarse background mesh. The topology and momentum sources fax and fθ were distributed according to the radial,
mesh details of the domains around the rudder and the surrounding axial–symmetric law provided by Hough (1966).
refinement block are shown in Figs. 3a–b. On the contrary, in case of the free running maneuver, the body
In the following, all the quantities are made non dimensional using forces are distributed consistently to the local angle of incidence
the length between perpendiculars (L ref = L pp ), the approach speed at experienced by the blade sections (Dubbioso et al., 2017). In particular,
model scale (Vref = V0 ), and the density of water ( ρref = 1000 kg/m3) as a weighting function W (r , θ ) is introduced:
reference quantities. In particular, the time is made non dimensional
by tref = L ref / Vref . Table 2
Details of grid cells distribution.
In the simulations, carried out at nominal speeds corresponding to
Froude number not exceeding Fr = 0.26 (Re=1.60·107), the free surface DOMAIN CELLS PERCENT
is not resolved, because experiments highlighted that at this speed both
the free surface effects and motions in the transverse plane were HULL 2.36 M 14%
APPENDAGES 8.60 M 61%
negligible. All the numerical simulations were unsteady and the time
RUDDERS 2.2 M 12%
step was set to Δt ⋍0.001. RUDDER BACKGROUND 1.8 M 10%
The different numerical tests are summarized in Table 3 and BACKGROUND 0.5 M 3%
consisted of: TOTAL 15.0 M 100%

3
R. Muscari et al. Ocean Engineering xxx (xxxx) xxx–xxx

Table 3
Captive model tests.

TEST KT /J 2 δ [deg] βH [deg] Yaw rate

PURE RUDDER 0, 0.2, 0.4, 0.75 0°, 20° − 35° 0.0 0.0
PURE TURNING 0.200 (0.367) 15° 7.5° 0.236
PURE TURNING 0.215 (0.518) 25° 12° 0.333
PURE TURNING 0.238 (0.769) 35° 13.5° 0.363

Fig. 4. Modeling of the inflow to the rudder.

4. Results

The numerical results aim to describe the asymmetric behavior of

the rudders consequent to the different interaction with the propeller
and the hull. The results are analyzed and discussed referring to
established physical models of ship maneuvering, given their reliability
to synthesize the key aspects of the phenomenon; at the same time, this
approach allows to emphasize the drawbacks of the simplified model-
Fig. 2. Reference system (view from the bottom).
ing of hull–propeller–rudder interaction in case of twin screw ship.
α′(r , θ ) The side force developed by the rudder depends on the inflow
W (r , θ ) = velocity VTOT and the incidence angle αeff :
∑Disk α′(r , θ ) (1)
∂CN 2
FY − R = 0.5ρArud αeff VTOT
where α′(r , θ ) is the incidence angle experienced by the generic blade ∂α (3)
section at (r, θ); the term ∑Disk α′(r , θ ) at the denominator normalize where Arud is the plane area and ∂CN
is the side force coefficient (in the
the weights, so that they sum up to one; the axial and circumferential ∂α
ship reference frame) of the rudder.
body forces at each element of the disk are then computed as
In simplified models these contributions are separately evaluated as
described in Figs. 4 and 5:
fax (r , θ ) = KT W (r , θ ) fθ (r , θ ) = KQ W (r , θ ) (2)

This approach yield a more realistic approximation of the correct • the effective inflow to the rudder (Fig. 4) is estimated from the flow
speed inside and outside the propeller race, weighted by the
propeller loading condition and propeller–rudder interference, and
percentage of the rudder plane area covered by these two parts.
hence of the model dynamic response.
The accelerated flow is estimated by the classical actuator disk

Fig. 3. Topology and mesh details of the rudder and rudder refinement block.

4
R. Muscari et al. Ocean Engineering xxx (xxxx) xxx–xxx

theory, that provides a correction Δu that depends on the propeller ward shift of the center of pressure due to the onset of flow separation
loading factor KT / J 2 , and by a further correction factor KP that at the leading edge. On the contrary, for δ < 0°, the reduction of the
depends on the longitudinal distance from the propeller (Gutshe rudder torque is smoother and the passage of CP past the rudder stock
correction, (Molland and Turnock, 2006)): is slightly delayed. The different behavior of the rudders for positive
and negative δ is described in Fig. 7 for the propeller loading condition
Arace Vrace + A0 Vw
VTOT = KT / J 2 = 0.4 in terms of the pressure field: up to flow separation,
Arud (4)
pressure distribution looks similar for positive and negative deflections
(up to about δ = ± 20°). For larger deflection angles, the flow is
completely separated for positive deflection (e.g. δ = 35°, with a shift
⎛ ⎛ 8KT ⎞⎞
Vrace = Vw ⎜1 + KP ⎜ 1 + − 1⎟ ⎟ of the pressure center past the rudder stock), whereas only partial
⎝ ⎝ πJ 2
⎠⎠ (5) separation happens for negative deflection (e.g. δ = −35°).
It is also interesting to observe that the rudder torque is slightly
Usually, the accelerated flow is further corrected by means of semi–
affected by the propeller loading for negative δ.
empirical factors, which reduce the theoretical value (Martelli et al.,
The results of the steady turning maneuvers are reported in Tables 5–6
2014); in other cases, further corrections are used to take into
for the internal and external rudder, respectively, in terms of the effective
account the turbulent mixing in terms of both the diameter of the
incidence αeff (obtained by the identity lift approach), geometric drift βR and
slipstream and the velocity.
γ (from Eq. (6)).
• the effective incidence angle (Fig. 4), determined by the interaction
In the tables, the values of αeff and γ for the un-propelled case are
with the propeller and the hull, is evaluated by a flow straightening
also reported in parentheses, in order to gain a deeper insight of the
coefficient γ:
effect of the propeller during a coupled drift–yaw motion. It is
αeff = δ − γβR (6) interesting to observe that the drift angle experienced by the rudder
v ′ + lR r ′ is almost twice the value for the hull due to rotation (βyaw). The details
where βR = is the geometric drift angle. It has to be noticed
u′ of the flow and pressure field on the rudders are shown in Figs. 8 and 9
that the geometric drift at the rudder can be alternatively evaluated for the limiting cases at δ = 15° and 35°. In Fig. 11, the rudder side
considering, in the denominator, the additional velocity by the force and the flow straightening coefficient obtained for the three
propeller; obviously, the choice should be consistent with the different maneuvers are visualized for the propelled and unpropelled
procedure developed to obtain γ from experiments or numerical configurations.
tests. From fluid dynamic point of view, the external rudder experiences
the most interesting behavior as a consequence of a complex interac-
The analysis is focused on the flow straightening coefficient since tion with the propeller race. The rudder is not completely immersed in
the accelerating effect of the propeller is based on theoretical grounds. the propeller race, the latter being deflected towards the leeward side
The flow straightening coefficient γ is evaluated following a side force by the cross flow. In particular, at δ = 15°, the vortex tube accelerates
identity approach, explained schematically in Fig. 5: the effective the flow of the face of the rudder (confirmed by the region of negative
incidence angle experienced by the rudder during a maneuvering
pressure at the face of the rudder in Fig. 9b), counteracting the
condition is identified by comparing the rudder side force with the pressure jump on the suction side. As a result, the effective angle of
characteristic curve of the rudder at the corresponding propeller
attack of the rudder is very small (see values reported in Table 6 and
loading coefficient. The characteristic curves of the rudder are obtained Figs. 8b–e–f) and only the port rudder provides the desired control
by pure rudder simulations and, therefore, take implicitly into account
force.
the effect of the wake of the hull in straight ahead condition; moreover, At the tightest maneuver, the propeller race is more deflected (the
the propeller functioning is also related to the rectilinear motion. The
cross flow is stronger). As a consequence of the external location with
rudder characteristics curves were computed for a limited set of respect to the propeller axis, the external rudder experiences a “clean”
propeller loading spanning the propeller regime observed experimen- flow and γ is very close to unity (the geometric drift angle is not altered
tally (see Table 3) in order to reduce the number of computations; the by hydrodynamic interference). The flow is not stalled, as emphasized
lift curves at the generic KT / J 2 are obtained from the computed ones by by the velocity vectors.
linear interpolation. On the contrary, the internal rudder always takes benefit of the flow
accelerated by the propeller; at the tightest maneuver, this interaction
4.1. Captive model tests causes the flow past the suction side of the rudder to stall. The stall is

The results of the pure rudder tests are shown in Fig. 6 for the
resistance FX − R , side force FY − R , torque MZ − R (evaluated about the
rudder stock) and the center of pressure xCP (provided in percentage of
the mean chord of the rudder).
In general, the forces and moment increase with the propeller
loading because the inflow velocity VTOT increases (see Eq. (3)); rudder
stall is triggered around δ = 20°, confirming the preliminary computa-
tions obtained for the open–water propeller–rudder configuration
(Dubbioso, 2015). The increase of the propeller loading does not affect
the stall angle and slightly alters the post–stall regime for negative δ,
probably because the rudder rotates towards a region more affected by
the hull wake (in fact, the trailing edge moves inward with respect to
the propeller axis). This asymmetric behavior is clear for the zero
loading curve and can be evidenced by the slightly retarded stall angle
for δ < 0°. The different regime of the rudder for positive and negative
deflections is further highlighted by the rudder torque MZ − R and the
center of pressure CP, see Figs. 6c and d. For positive δ, MZ − R drops
abruptly to negative values after δ = 20° as consequence of the after- Fig. 5. Identity lift approach.

5
R. Muscari et al. Ocean Engineering xxx (xxxx) xxx–xxx

Fig. 6. Rudder loads.

Table 4 distributed over the whole span of the rudder, with a smaller influence
Lateral force and torque of the rudder - Circular motion tests. on the inflow (γ increases). The starboard rudder experiences values of
γ that are very close to unity, the geometric drift of the rudder being
δ KT /J 2 FY − R (w/prop) FY − R (wo/prop)
almost equal to the effective one (see the left column of Fig. 10).
15° INTERNAL 0.200 −1.07E-005 1.94E-005 In case of the propelled cases, γ shows remarkable variations on the
25° INTERNAL 0.215 −2.57E-005 −1.48E-005 internal side. Comparison of the windward and leeward side on the
35° INTERNAL 0.238 −8.54E-005 6.40E-007 right column of Fig. 10 confirms that the propeller wash alters
15° EXTERNAL 0.367 5.73E-005 −3.88E-005
completely the rudder inflow along the span. The straightening of the
25° EXTERNAL 0.518 1.18E-004 −3.25E-005
35° EXTERNAL 0.769 1.25E-004 −4.03E-005
propeller is clear for δ = 15° and δ = 25°; at the highest angle, the
resultant force developed by the rudder is affected by the increase of
the counterbalancing contribution from the lower portion of the rudder
not related to propeller loading (because the propeller thrust only (evidenced by the large suction at the trailing edge on the face of the
slightly increases with respect to the straight ahead condition) but rudder, see Fig. 7c).
rather to the effective inflow angle. On the external shaft, γ is slightly affected by the propeller;
The effects of the flow field are quantitatively described by the trend specifically, with the increase of the yaw rate the effect of the slipstream
of the flow straightening coefficients for the starboard and port rudder. gradually reduces from the top of the rudder (right column of Fig. 10).
For the unpropelled case configuration, the internal (port) rudder It is to be observed the change of sign of the side force at δ = 15°, i.e.
experiences the strongest interaction as a consequence of the evolution the rudder provides a stabilizing, although small, side force.
of the wake; the strongest effect of the wake on the inflow to the rudder In general, the computations with the constrained model confirmed
is at δ = 25°, because the vortex structures generated at the bilge and the experimental finding and supported the assumption about the
the skeg are more aligned with the rudder, as described by the axial asymmetrical flow straightening coefficients (Dubbioso, 2015), i.e. that
velocity contours on the left half of Fig. 10b. At δ = 35° (left half of γ well synthesizes the superposition of various effects, due to both hull
Fig. 10c), the disturbance is further deflected towards the lee side and wake deformation during the coupled yaw–drift motion and the

6
R. Muscari et al. Ocean Engineering xxx (xxxx) xxx–xxx

Fig. 7. Pressure distribution; KT /J 2 = 0.4 .

Table 5 4.2. Free running maneuver

Evaluation of the flow straightening coefficient - internal rudder.
The free running maneuver for δ = 35° (maneuver to port side, see
δ KT /J 2 αeff (w/prop) αeff (w/prop) βyaw βR [deg] γ (γH )
[deg] comments in Section 3) is carried out to further describe the propeller–
[deg] [deg]
rudder interaction and the variation of γ during the transient phase.
15° 0.200 14.32 5.25 7.08 14.35 0.047 This analysis is motivated by the fact that, usually, in simplified
(0.6788) maneuvering modeling, the flow straightening coefficient is assumed
25° 0.215 13.91 11.76 11.03 22.18 0.499
constant or it is implemented from experiments consisting in oblique
(0.5969)
35° 0.238 8.53 12.54 13.45 25.61 1.033 towing or circular motion tests (for example by MMG maneuvering
(0.8769) models (Khanfir et al., 2011)); therefore, this may lead to incorrect
estimation of control forces and consequently of ship response.
The flow straightening coefficient is analyzed at each time step
Table 6 accordingly to the identity of side force procedure described in previous
Evaluation of the flow straightening coefficient - external rudder.
section Fig. 11.
δ αeff (w/prop) αeff (w/prop) βyaw βR [deg] γ (γH ) Comparison of CFD results with experiments is reported in Fig. 12
KT /J 2
[deg] [deg] [deg] in terms of trajectory, speed drop and yaw rate. The macroscopic
parameters are summarized in Table 7, where a satisfactory agreement
15° 0.367 −1.77 −1.90 7.08 14.35 0.921 can be observed, the average error being less than 5%.
(0.912)
In Fig. 13 the separate contribution of the fully appended hull,
25° 0.518 4.24 3.66 11.03 22.18 0.936
(0.962) propellers (total contribution) and rudders are reported in terms of
35° 0.769 9.41 10.75 13.45 25.61 0.998 side force and yaw moment. It is worth noticing that the propellers acts
(0.946) to counteract the turn and in the steady turning phase the overall
contribution from the propeller amount to about 45% and 50% of the
rudders side force and moment, respectively. The side force developed
variation of propeller operating condition. These effects are not directly
by the propellers (30% and 14% of the thrust developed during the
taken into account in system based models, their complex nature being
rectilinear approach phase) is in satisfactory agreement with the
difficult to generalize even for similar ships, due to the sensitivity to
experimental values (see Dubbioso et al. (2017), Tables 6 and 7) and
both local and global geometric details of the stern. The introduction of
this supports the reliability of the prescribed body force model. In
γ is an effective mean to partially account for these effects and,
general, the maneuvering loads developed by the hull and propeller
therefore, its physical meaning is more than a geometric indicator of
show a trend similar to that of the yaw rate (Fig. 12b).
the straightening of the flow to the rudder.
On the other hand, the rudder forces rise quickly up to t = 1

7
R. Muscari et al. Ocean Engineering xxx (xxxx) xxx–xxx

Fig. 8. Details of the flow on the rudders, longitudinal velocity component; horizontal plane through propeller center.

achieving almost the same peak, because the response of the hull is circulatory component (i.e., added mass) is relevant with respect to the
slow (the model showed good course keeping qualities) and, therefore, lift. After the peak, the asymmetric evolution between the rudders
the reduction of the rudder incidence due to βR is delayed (see Eq. (6)). becomes evident: on the internal rudder, the side force FYINT − R drops
The forces quickly rise also because the contribution of the non– faster than FYEXT
− R , because the rudder inflow could be more affected by

8
R. Muscari et al. Ocean Engineering xxx (xxxx) xxx–xxx

Fig. 9. Pressure distributions during steady turning.

the interaction with the wake of the hull. On the contrary, on the δ − αeff
γ=
external shaft the inflow is less perturbed by the wake of the hull and βR (7)
the variation is smoother.
The asymmetric behavior of the rudders during the stabilized phase Therefore, δ and βR being the same for both rudders, this behavior it
of turn is less evident with respect to the steady phase simulation. FYINT closely connected to slower or faster variation of αeff (i.e., the rudder
−R
is in good agreement with the previous simulation whereas FYEXT force). Time evolution for these terms is represented in Fig. 15. On the
− R is
overestimated. This is due to the fact that the drift angle of the model external shaft (see Fig. 15b), the faster increase of αeff is associated to
establishes around a value smaller than the prescribed one (i.e. 12° the fact that the propeller loading Kt / J 2 increases monotonically and
compared to 13.5°) and, consequently, the rudder might experience a the deflection of the propeller race is weak (Fig. 15c). On the leeward
positive interaction with the propeller race. side (Fig. 15a), the inflow velocity to the rudder is slower because of
The analysis of the asymmetric behavior of γ by the identity lift both the inward deflection of the hull wake and of the reduction of the
procedure for inspecting the transients is approximate and the results propeller load: consequently, δ − αeff increases faster than βR .
have to be considered qualitative: in fact, the value of αeff does not Approximately after the complete rudder rotation (t ∼ 1.1), the
properly account for unsteady effects related to the rotation of the variation of FYEXT
− R gradually reduces because of the motion stabilization

rudder superposed to the variation of inflow caused by the motion and drop of the inflow speed caused by the deflection of the propeller
(alteration of the wake past the hull and propeller operating regime), race; consequently, γ increases for the external rudder. On the internal
because the rudder force used to this purpose results from pure rudder side, γ shows a peaked behavior that corresponds to the abrupt drop of
tests (in steady configuration). − R (see Fig. 13) related to flow separation.
FYINT
The outcome of the quasi–steady analysis is synthesized in Fig. 14 The terms involved in the evaluation of γ during the initial transient
in terms of γ and αeff . It has to be observed that during the stabilized (identified by the superscript UNST ) are reported in Tables 8–9 at rudder
turn, γ achieves approximately the same values obtained by the circular angles equal to the steady analysis. It is worth noticing the deeply
motion tests at δ = 35° (Tables 5–6). The asymmetry of the flow different dependence of γ on δ between internal and external rudder,
straightening experienced by the rudders is amplified during the determined by the different interaction with the propeller wake. Note
transient phase, as highlighted by the zoomed view for the interval also that βUNST
R is considerably smaller than the steady value βSTR, the
0 < t < 3, reported for the sake of clarity in Fig. 15. It is to be noted that gap being reduced when δ increases, because of the faster stabilization
the evolution of γ for the two rudders during the early stages of of the turn.
deflection is opposite to what observed during the stabilized phase.
This fact has to be related to the different direction of rudder forces. In 4.3. Considerations on simplified propeller–rudder modeling in
fact, from Eq. (6) we have system based maneuvering models

The asymmetric behavior of the propeller–rudder system is an

9
R. Muscari et al. Ocean Engineering xxx (xxxx) xxx–xxx

Fig. 10. Inflow to the rudders; contours of axial velocity. Left: no propellers; right: with propellers.

important characteristic of maneuvering twin screw hulls. Indeed, a fixed incidence angle, the side force generated by the rudder always
these effects should be included in simplified maneuvering models to increases with the propeller loading, because of the implicit assumption
achieve useful indications during the preliminary design of the control that the propeller race remains straight. Moreover, the regressions for γ
devices. proposed in literature, based only on hull/stern macroscopic charac-
The core of propeller–rudder interaction, synthesized by (4)–(5), teristics, are reliable for single screw ships (Inoue et al., 1981a, 1981b),
highlights some drawbacks of the simplified theories. In particular, for while they fail for most twin screws/twin rudder ships, because they do

Fig. 11. Rudder performance; steady turning simulations.

10
R. Muscari et al. Ocean Engineering xxx (xxxx) xxx–xxx

Fig. 12. Trajectory and kinematic response of the model.

Table 7 ad hoc experiments. The major limit of this procedure is that very often
Macroscopic parameters of δ = 35° maneuver. V is the velocity of the model at steady quasi–steady states are usually considered, and, as shown in the
phase. previous sections, this may be misleading, in particular during
EXP CFD err %
transient phases.
Moreover, in simplified models the propeller is modeled through
Advance 3.55 3.197 −9.94 the open water characteristics or by simplified models (i.e., Gutshe
Transfer 1.6 1.627 1.68 (Cassella et al., 1989)). More accurate and/or validated propeller side
Tactical diameter 4.09 3.993 −2.37
Final diameter 4.11 4.012 −2.38
force models (for example, the method of Ribner (1943) is standar-
V /V0 0.75 0.75 0.10 dized for flight vehicle maneuvering predictions) should be included in
Yaw rate 0.352 0.36 2.27 physical models. The improvement of the physical consistency of the
Drift angle [deg] 13.57 12.05 8.88 propeller–rudder interaction should be able to reduce, for operational
ranges not affected by speed, the uncertainty and the difficulties related
to the identification of a generic model for the hull; namely, the
not take into account the different flows around the internal and
characterization of the hull by a single set of hydrodynamic coefficients
external propeller–rudder system.
reliable to accurately describe the maneuvering characteristic of the
In particular, the flow straightening coefficient used in simplified
ship for a broader set of rudder angles and maneuvers typology.
theories synthesizes different hydrodynamic effects by a constant,
averaged reduction of the incident flow to the rudders; hence, it cannot
properly represent the correct inflow along the span of the rudder that 5. Conclusions
is strongly dependent on the geometric configuration of the propeller–
rudder system, the arrangement of the stern appendages, and the The asymmetric performance of the propeller–rudder system of a
different overloading of the propeller. Moreover, for twin screws/twin twin screw model, with particular emphasis on the rudder hydrody-
rudder ships, the use the same γ for both internal and external rudder namics, were analyzed by CFD simulations in order to assess the
represents a further simplification of the flow physic modeling. For reliability of simplified physical models used for ship maneuvering
these cases, the observed asymmetry of propeller performances and simulations. The fully appended model were considered in the numer-
flow straightening is usually included by means of heuristic models or ical simulations for three kind of typical maneuvers: pure rudder,

Fig. 13. Decomposition of the loads acting on the hull, propeller and rudders.

11
R. Muscari et al. Ocean Engineering xxx (xxxx) xxx–xxx

Fig. 14. Transient flow straightening coefficient analysis.

Fig. 15. Contributions to the variation of γ (Eq. (7)).

12
R. Muscari et al. Ocean Engineering xxx (xxxx) xxx–xxx

Table 8 maneuvering models and computational results.

Steady identity lift approach applied to transient maneuvers: internal rudder.
Acknowledgments
δ KT /J 2 FY − R αeff [deg] βST
R [deg] βUNST
R [deg] γ UNST

15° 0.1848 −1.994E-4 12.49 14.35 4.7 0.53 This research activity is funded by the Flagship Project RITMARE -
25° 0.1748 −2.618E-4 17.87 22.17 10.76 0.66 The Italian Research for the Sea – coordinated by the Italian National
35° 0.1775 −3.135E-4 20 25.61 18.06 0.83 Research Council and funded by the Italian Ministry of Education,
University and Research within the National Research Program 2011–
2013. The authors would like to acknowledge Mr. Fabio Carta for his
Table 9
valuable technical contribution and his support for adjustments and
Steady identity lift approach applied to transient maneuvers: external rudder.
refinements to CAD models.
δ KT /J 2 FY − R αeff [deg] βST
R [deg] βUNST
R [deg] γ UNST
References
15° 0.2212 −1.216E-4 10.92 14.35 4.7 0.868
25° 0.2546 −2.185E-4 18.21 22.17 10.76 0.631
Badoe, C., Philips, A., Turnock, S., 2015. Influence of drift angle on the computation of
35° 0.3043 −3.066E-4 35 25.61 18.06 0
hull-propeller-rudder interaction. Ocean Eng. 103, 64–77. http://dx.doi.org/
10.1016/j.oceaneng.2015.04.059.
Cassella, P., Mandarino, M., Scamardella, A., 1989. Systematic tests with b-wageningen
steady turning and free running model tests. The inspection of the screw propellers in non-axial flow: presentation and analysis of the experimental
results. In: Practical Design of Ships and Floating Structures (PRADS).
complete flow field developed around the hull clarified the key
Coraddu, A., Dubbioso, G., Mauro, S., Viviani, M., 2013. Analysis of twin screw ships'
mechanisms that affect the rudder performance in steady and unsteady asymmetric propeller behaviour by means of free running model tests. Ocean Eng.
conditions. 68, 47–64. http://dx.doi.org/10.1016/j.oceaneng.2013.04.013.
It was found that, for the particular geometrical configuration that Di Mascio, A., Broglia, R., Muscari, R., 2007. On the application of the one-phase level set
method for naval hydrodynamic flows. Comput. Fluids 36, 868–886.
characterizes the adopted model (outward misalignment of the rudder Di Mascio, A., Muscari, R., Broglia, R., 2006. An Overlapping Grids Approach for Moving
with respect to the propeller), the differences between the internal and Bodies Problems. In: Proc. of 16th Int. Offshore and Polar Engineering Conference,
external rudder have to be ascribed to the kinematic state of the model San Francisco, California (USA).
Di Mascio, A., Muscari, R., Dubbioso, G., 2014. On the wake dynamics of a propeller
and the propeller performance. During steady turning maneuvers, the wake operating in drift. J. Fluid Mech. 754, 263–307. http://dx.doi.org/10.1017/
external rudder resulted less efficient than the internal one, because the jfm.2014.390.
deflection of the propeller slipstream induced a detrimental effect in terms Dubbioso, G., Muscari, R., Mascio, A.D., 2013. Analysis of the performances of a marine
propeller operating in oblique flow. Comput. Fluids 75, 86–102. http://dx.doi.org/
of inflow velocity and equivalent incidence angle. The internal rudder was 10.1016/j.compfluid.2013.01.017.
always covered by the propeller slipstream; however, the principal benefit Dubbioso, G., Muscari, R., Ortolani, F., Di Mascio, A., 2017. Analysis of propeller bearing
from this interaction is indirectly related to the straightening of the flow, loads by CFD. Part I: quasi-steady maneuvers. Ocean Eng. 130, 241–259.
Dubbioso, G., Ortolani, F., Mauro, S., Martelli, M., Nataletti, M., Villa, D., Viviani, M.,
since the internal propeller did not experience a relevant overloading.
2015. Experimental and numerical investigation of asymmetrical behaviour of
Moreover, the rudder hydrodynamics is affected by a strong interaction rudder-propeller for twin screw ships. In: Proc. of MARSIM, Newcastle, UK.
with the wake of the hull, that counteracts the regularization of the flow Hirano, M., 1980. A practical calculation method of ship maneuvering motion at design
stage. J. Soc. Nav. Archit. Jpn. 147, 68–80.
imparted by the propeller with the increase of drift angle and yaw rate.
Hough, Ordway, 1966. On the generalized actuator disk. Technical Report No. 433976.
During the transient phases (characterized by moderate drift angles and Therm Advanced Research, Inc.,
yaw rate), the performance of the external rudder is superior with respect Inoue, S., Hirano, M., Kijima, K., 1981a. Hydrodynamic derivatives of ship maneuvering.
to the internal one. In fact, the propeller is overloaded and its slipstream is Int. Shipbuild. Prog., 28.
Inoue, S., Hirano, M., Kijima, K., Takashina, J., 1981b. A practical calculation method for
not deflected because the transverse motion of the hull is still slow; ship maneuvering motion. Int. Shipbuild. Prog., 28.
conversely, on the internal side, the control force is reduced because of the Khanfir, S., Hasegawa, K., Nagarajan, V., Shouji, K., Lee, S., 2011. Manoeuvering
less efficient propeller and the disturbance caused by the wake evolution. characteristics of twin-rudder systems: rudder hull interaction effect on the
manoeuvrability of twin-rudder ships. J. Mar. Sci. Technol. 16, 472–492.
On the leeward side, the simultaneous interaction of the propeller race Martelli, M., Viviani, M., Altosole, M., Figari, M., Vignolo, S., 2014. Numerical modelling
and hull wake causes the abrupt reduction of the loads induced by flow of propulsion, control and ship motions in 6 degrees of freedom. Journal of
separation; conversely, the external rudder encounters a less perturbed Engineering for the Maritime Environment, Proc. IMech, Part M 228, 373–397.
Molland, A., Turnock, P., 2006. Marine Rudders and Control Surfaces first ed..
flow and the variation of the force is smoother with respect to the opposite Butterworth-Heinemann, New York.
side. Muscari, R., Di Mascio, A., Verzicco, R., 2013. Modelling of vortex dynamics in the wake
The inclusion of the dependence of rudder performance on kinematic of a marine propeller. Comput. Fluids 73, 65–79.
Muscari, R., Felli, M., Di Mascio, A., 2011. Analysis of the flow past a fully appended hull
parameters and propeller loading and slipstream is of paramount
with propellers by computational and experimental fluid dynamics. J. Fluids Eng.
importance to improve the prediction capability of the simplified math- 133 (061104), 1–16.
ematical models and, therefore, to develop more robust tools to be used Ortolani, F., Mauro, S., Dubbioso, G., 2015a. Investigation of the radial bearing force
developed during actual ship operations part 1: straight ahead sailing and turning
during the earlier ship design phases. In particular, the present result
maneuvers. Ocean Eng. 94, 67–87. http://dx.doi.org/10.1016/
emphasized that the lateral gap between rudder and propeller can amplify j.oceaneng.2014.11.032.
the asymmetric behavior during maneuvers, with detriment of rudder Ortolani, F., Mauro, S., Dubbioso, G., 2015b. Investigation of the radial bearing force
efficiency and possible increase of vibratory loads. Most of experimental developed during actual ship operations part 2: unsteady maneuvers. Ocean Eng.
106, 424–445. http://dx.doi.org/10.1016/j.oceaneng.2015.06.058.
data found in the literature (i.e., (Molland and Turnock, 2006)) investi- Ribner, H., 1943. Propellers in Yaw. National Advisory Committee for Aeronautics
gated systematically the effect of longitudinal and lateral distance on (NACA), (Technical Report No. 3L09).
rudder performance only for the zero drift angle; their extension to non Spalart, P.R., Allmaras, S.R., 1994. A one-equation turbulence model for aerodynamic
flows. La Rech. Aérospatiale 1, 5–21.
zero drift angle can improve data set used to validate simplified

13