You are on page 1of 10

Ocean Engineering 42 (2012) 61–70

Contents lists available at SciVerse ScienceDirect

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

Effect of a propeller duct on AUV maneuverability


E.A. de Barros n, J.L.D. Dantas
~ Paulo. Av. Prof. Mello Moraes, 2231, 05508-900 SP, Brazil
Department of Mechatronics Engineering and Mechanical Systems, University of Sao

a r t i c l e i n f o abstract

Article history: A number of autonomous underwater vehicles, AUV, are equipped with commercial ducted propellers,
Received 6 January 2011 most of them produced originally for the remote operated vehicle, ROV, industry. However, AUVs and
Accepted 2 January 2012 ROVs are supposed to work quite differently since the ROV operates in almost the bollard pull
Editor-in-Chief: A.I. Incecik
condition, while the AUV works at larger cruising speeds. Moreover, they can have an influence in the
Available online 25 January 2012
maneuverability of AUV due to the lift the duct generates in the most distant place of the vehicle’s
Keywords: center of mass. In this work, it is proposed the modeling of the hydrodynamic forces and moment on a
AUV duct propeller according to a numerical (CFD) simulation, and analytical and semi-empirical, ASE,
Ducted propeller approaches. Predicted values are compared to experimental results produced in a towing tank. Results
Maneuverability
confirm the advantages of the symbiosis between CFD and ASE methods for modeling the influence of
CFD
the propeller duct in the AUV maneuverability.
Hydrodynamic coefficients
Prediction method & 2012 Elsevier Ltd. All rights reserved.

1. Introduction results that were validated experimentally for some duct profiles
investigated by Morgan and Caster (1965).
A number of autonomous underwater vehicles, AUVs, are Considering the numerical simulation, there are a number of
equipped with commercial ducted propellers, most of them pro- works on the application of inviscid flow modeling to the duct
duced originally for the operation in remote operated vehicles, ROVs. propeller analysis. The works of Lewis and Ryan (1972), Gibson
However, AUVs and ROVs are supposed to work quite differently and Lewis (1973) were based on the placement of singularities for
since the ROV operates in almost the bollard pull condition, while the modeling the duct and the actuator disk, which represents the
AUV works at larger cruising speeds. propeller. Falca~ o de Campos (1983) introduced numerical meth-
Ducts can help the AUV to accelerate from zero to cruising ods for accounting shear flow effects. Surface panel methods have
speed. At cruising speed, however, the duct drag effect may cancel been developed as well for modeling the duct and the propeller
or even surpass the additional thrust it produces. Moreover, the blades (Kerwin et al., 1987; Baltazar and Falca~ o de Campos, 2009).
duct also contributes to increase the damping effect during Vortex lattice methods were applied for analyzing the unsteady
manoeuvres. Considering its common location far behind the cavitating performance of ducted propellers subject to non-
vehicle center of gravity, the duct effect on the vehicle maneuver- axisymmetric inflows (Kinnas et al., 2005).
ability can be quite significant. Methods based on the Reynolds Average Navier Stokes, RANS,
There is a lack of works on the evaluation of the duct effect on formulation have been applied also to the duct propellers analysis
maneuverability for this kind of vehicle. Van Gusteren and van in order to treat explicitly the viscous effects. Hoekstra (2006)
Gusteren (1972) has proposed an analytical estimation method has investigated the duct profiles 19A and 37 for the zero angle
for predicting the effect of the ducted propeller on the steering of of attack case, modeling the propeller as an actuator disk, and
ships. Minsaas et al. (1973) have analyzed experimentally the obtained good agreement with the experimental results.
effect of duct propellers on de maneuvering of tankers. Consider- A number of authors have applied numerical tools to predict
ing the combination of duct and a body of revolution, Falca~ o de added masses, static forces and moments, and the wake field
Campos (1983), based on numerical flow simulation tools, has of submarines and other underwater vehicles. In the work of
analyzed the effect of the duct on the flow at the stern of the hull, Vaz et al. (2010), two different CFD (computational fluid dynamic)
however, only the zero angle of attack case was considered. solvers were used for testing a number of turbulence models
Lift, drag and moment coefficients for the duct alone case have related to static RANS, applying the American Society of Mechan-
been computed by analytical formulas derived from theoretical ical Engineers numerical verification and validation methodology,
the ‘‘ASME V&V’’. Unsteady RANS and the Large Eddy Simulation
formulation were tested as well. The flow simulations were
n
Corresponding author.Tel.: þ 55 11 3091 5761; fax: þ3091 5471. related to two configurations of submarines (with and without
E-mail address: eabarros@usp.br (E.A. de Barros). the sail and control surfaces).

0029-8018/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.oceaneng.2012.01.014
62 E.A. de Barros, J.L.D. Dantas / Ocean Engineering 42 (2012) 61–70

Recently, CFD methods were used to predict the stability This work proposes the combination of CFD and ASE approaches
derivatives of AUVs. For example, in the works of Tyagi and Sen applied to a different problem: the prediction of static efforts acting
(2006) and Jagadeesh et al. (2009) the forces and moments acting upon the AUV bare hull combined with a propeller duct. The main
upon the AUVs hulls, when they are inclined to the flow (i.e. have motivation of this study is to achieve a method of modeling the
an angle of attack), are compared with those obtained in experi- effects of the duct on the AUV for manoeuvre simulation and control
mental tests in the towing tank. system design. During investigations on the modeling and control of
Phillips et al. (2007) applied CFD methods for predicting the AUV dynamics, it is common to refer to the hydrodynamic deriva-
dynamic stability derivatives of the Autosub AUV. The turning tives in the equations of motion. Methods for predicting those
manoeuvre was simulated, and the computed efforts were com- parameters may include analytical and semi-empirical, ASE, formula
pared to those ones obtained during towing tank tests with a based on the vehicle geometry and mass distribution. In de Barros
planar motion mechanism (Kimber and Scrimshaw, 1994). et al. (2008a), the duct effect on those derivatives was included by
Tang et al. (2009) also used CFD methods in the prediction of the addition of the ASE formula proposed by Morgan and Caster
hydrodynamics coefficients of AUVs in block-type shapes. The (1965), without considering the interaction between hull and duct.
validation was done by comparison of the vehicle trajectory Moreover, the formula adopted had not been tested for a typical
obtained in simulations and manoeuvres performed in the field. accelerating duct such as the 19A.
In addition to the identification of hydrodynamic coefficients, In this work, a CFD technique is used for simulating the flow
the CFD tools are also used to optimize the shape of AUVs around a propeller duct at different angles of attack. The numerical
(Yamamoto, 2007; Inoue et al., 2010), performance of AUV predictions are validated according to experimental results for the
propellers (Husaini et al., 2008), and to examine the efforts upon duct alone case. In a next phase, the interaction between the duct
AUV hull during the docking process (Wu, 2010). and the bare hull wake is investigated. The vehicle considered
The economy of computational resources and expertise in is an AUV, which was investigated previously (de Barros et al.,
numerical models make the analytical and semi-empirical meth- 2008a, b). Predictions based on CFD, ASE and on the combination
ods, ASE, still an interesting alternative to the CFD approaches. of both approaches are compared to those ones produced during
The calculation of hydrodynamic derivatives yield approximate experiments in a towing tank.
results that can be used to predict maneuverability characteris-
tics, select hydroplanes, and investigate control strategies at an
early stage of vehicle design. ASE methods are directly focused on 2. Numerical model
the estimation of parameters such as added mass and inertias
(acceleration related coefficients), linear and non-linear damping The solver adopted in CFD simulations is the Ansys Fluent 12.1,
coefficients (related to velocities), and control action related using the ‘‘k-o shear stress transport’’ turbulence model (Menter,
parameters. Advanced approaches to AUV design may also 1994), due to the good results obtained in former investigations
involve combined plant/controller optimization, where prediction (de Barros et al., 2008a). Moreover, in Phillips et al. (2010), the
methods of hydrodynamic derivatives play an important role k-o shear stress transport was the 2-equation model, which
(Silvestre et al., 1998). provided the best prediction of efforts and vortice distribution
The ASE approach to derivatives estimation provides an around a submarine vehicle. A slightly better result was produced
analytical formulation based on physical concepts that can help by the Reynolds Stress Method. However, its computational cost
in the interpretation of experimental and CFD results as well as in is much higher, and the method convergence is more unstable
defining uncertainty intervals for hydrodynamic parameters in when compared to the ‘‘k-o’’ model.
order to help the design of robust control systems. Considering the meshing generation, grids around the duct
On the other hand, some elements of the numerical simula- alone and for the hull-duct combination are similar, both
tions can be used to improve the ASE predictions. For instance, having the same kind of elements and are distributed using the
the flow visualization at the stern of an AUV bare hull was used O-topology. The grid space was decomposed into three regions:
for improving the normal force prediction in de Barros et al. a boundary layer region, the space closely around the vehicle
(2008a). In the work of Jeans et al. (2010) the CFD simulations (Fig. 1) or duct, and the outer region. All three regions included
were applied for modeling the hydrodynamic impulse, which was only hexagonal elements, the size of which was increased from
used for estimating the normal force distribution along axisym- the body surface to the outer region according to an exponential
metric hulls at incidence. law. The thickness of the closest elements to the vehicle were

Fig. 1. Mesh generated closed to the bare-hull and duct, using hexaedral elemets with O-topology.
E.A. de Barros, J.L.D. Dantas / Ocean Engineering 42 (2012) 61–70 63

defined according to the boundary layer size of similar axisym- perturbations are observed. For the case of hull-duct combination
metric bodies (ESDU, 1978). The outer region is a rectangular the propeller boss was represented in the grid. Moreover, a conical
volume for the hull-duct combination, whose length is 15 times body, connecting the hull base and the propeller hub, was also
the vehicle length (26.0 m), the height is about 85 times the included.
vehicle diameter (20.0 m) and the width is about 25times the The analysis of the mesh refinement influence on the com-
vehicle diameter (6.0 m). In the simulations considering the duct puted results is based on the ‘‘ASME V&V’’, as applied by Ec- a et al.
alone, taking the chord length as unit, the outer region has 100 (2010). The same procedure is used to define the uncertainty level
units length (7.5 m), a height of 80 units (6.0 m) and a 20 units of the CFD calculation, which is represented by the corresponding
width (1.5 m), resulting in a mesh with 1.5 million elements. bar for each computed point in Figs. 3–5, 8, and 12. During the
Since the hull and duct are axisymmetric bodies, and subjected application of such methodology, five different grids have been
only to a variation in the angle of attack, a grid including just half tested for the flow simulation around each body (see Tables 1–4).
body was considered, in order to save computational resources. The grid index (hi) is used to express the degree of mesh
In the near-wall region (i.e. close to the body or the duct refinement in each case. It is a geometric parameter adopted for
external surface), the chosen turbulence model use the so called defining the typical cell size of a grid (Ec- a et al., 2010). The grid
Enchanced Wall Function, that requires a minimum spacing index is defined in this work as the first cell thickness, i.e. the
between grid elements in order to correct model the viscous distance of the closest grid node to the body surface. The grid
sublayer in the boundary layer. The size of these elements are index ratio is then calculated by the ratio between the grid index
given in function of the non-dimensional wall distance y þ , for each case, to that one corresponding to the first case.
representative of the local Reynolds number, and must be as
close to 1 as possible (Fluent, 2005). This parameter is defined as
Table 1
ryun
yþ ¼ ð1Þ Bare-hull mesh properties.
m
pffiffiffiffiffiffiffiffiffiffiffi Total Number First cell yþ Mean Grid Index
where, y is the distance to the body surface; un ¼ tw =r, is the of elements Thickness (hi) (mm) value ratio (hi/h1)
friction velocity; tw is the shear stress at the body surface; r is the
density of the fluid, m is the local dynamic viscosity of the fluid. Case 1 2008108 0.050 0.87 1
pffiffiffi
In the space closely around the bodies, the grid refinement Case 2 1807332 0.071 1.25 2
depends on how intensively the flow is disturbed. This means that Case 3 1650662 0.010 1.75 2
pffiffiffi
Case 4 1468441 0.141 2.50 2 2
the size of the elements needs to be decreased where flow separa-
Case 5 1385595 0.200 3.50 4
tion and vortices occur, compared to other areas where minor

Fig. 2. Lift and moment coefficient slope in function of the duct chord-diameter ratio, predicted by the analytical and semi-empirical formula presented in Morgan and
Caster (1965).

Fig. 3. Lift force (a) and pitch moment (b) coefficient on the Clark-Y duct with t/c¼ 0.117, and Dd/cd ¼ 1.5, calculated in the duct leading edge.
64 E.A. de Barros, J.L.D. Dantas / Ocean Engineering 42 (2012) 61–70

Table 2
Bare-hull with duct mesh properties.

Total number First cell yþ mean Grid index


of elements thickness (hi) (mm) value ratio (hi/h1)

Case 1 2306423 0.040 0.68 1


pffiffiffi
Case 2 2220223 0.056 0.95 2
Case 3 2111227 0.080 1.34 2
pffiffiffi
Case 4 2015283 0.113 1.98 2 2
Case 5 1929083 0.160 2.80 4

Table 3
19A Duct mesh properties.

Total number First cell yþ mean Grid index


of elements thickness (hi) (mm) value ratio (hi/h1)

Case 1 1459968 0.015 0.63 1.00


Case 2 1436640 0.020 0.85 1.33
Case 3 1404960 0.025 1.07 1.67
Case 4 1383840 0.030 1.23 2.00
Case 5 1297440 0.050 2.00 3.33 Fig. 4. Hydrodynamic center position as a function of the angle of attack for the
Clark Y duct, with t/c¼ 0.117, and Dd/cd ¼ 1.5.

Table 4 of attack range the CFD predictions clearly follow the experi-
Clark Y duct mesh properties. mental values closer than the original analytical prediction.
Therefore, it seems that the adopted turbulence model could
Total number First cell yþ mean Grid index
not correctly identify the duct stall. It is believed that the complex
of elements thickness (hi) (mm) value ratio (hi/h1)
flow at the stall condition may exhibit pressure gradients large
Case 1 1470544 0.020 0.45 1.00 enough so that the isotropic turbulent viscosity (Boussisneq
Case 2 1308414 0.035 0.78 1.75 hypothesis), which is implicit in the ‘‘k-o’’ model, can no longer
Case 3 1145768 0.050 1.10 2.50 be assumed. Using a more complex CFD model, that do not
Case 4 1016808 0.065 1.42 3.25
use the RANS (Reynolds Averaged Navier-Stokes) approach, like
Case 5 920234 0.080 1.77 4.00
the transients ‘‘Large Eddy Simulation’’ or the ‘‘Detatched Eddy
Simulation’’, could be further tried, at a much larger computa-
tional cost, in order to improve the predictions at the stall region.
3. Preliminary tests However, considering the estimation of stability derivatives,
this phase of the investigation was focused on the agreement
This study started with the application of the flow numerical between experiment and calculation in the range of 0 to 151.
simulation on duct profiles that have been used in experiments Taking into account the agreement between predicted and
reported in the literature. After comparing the measured results experimental results up to this phase, it was decided to keep
on lift and moment coefficients with those ones predicted by CFD, the ‘‘k-o’’ model in the next tests with the 19A duct, which was
it was possible to proceed with more confidence on the numerical combined later to the AUV bare hull for tank tests.
approach. The 19A duct profile is more usual in marine applications
The flow around the Clark-Y duct was simulated, adopting the nowadays than the others presented in the previous section. The
value 0.117 for the thickness to chord ratio. This was the profile analytical and semi-empirical predictions (see Fig. 2) produced by
chosen by Morgan and Caster (1965) for validating the analytical Morgan and Caster (1965) and validated using the previous
and semi-empirical approach developed for estimating lift and profiles were also applied to this case. The results were compared
pitch moment coefficients (see Fig. 2). This profile was tested to the ones generated by CFD. The duct dimensions were defined
experimentally by Fletcher (1957) in a wind tunnel for a number in order to fit it to the AUV stern: the diameter is 150 mm and the
of different diameter/chord ratios. duct chord is 75 mm.
In Fig. 3(a), the experimental results reported by Fletcher Results showing the ASE and CFD predictions are presented in
(1957) are included, as well as those generated by CFD, adopting Fig. 5a and b. For the normal force and pitch moment coefficients,
air as the fluid. Additionally, results obtained in CFD using water, it is observed a dead band in the range of 0 to 31 of angle of attack.
and the analytical prediction proposed by Morgan and Caster This can be explained by the significant slope of the external side
(1965) are also presented. of the duct, which results in a large incidence angle at small
Except for the angle of attack range where stall occurs (i.e., angles of attack, producing the stall effect (Fig. 6a). As the angle of
between 15 and 20 degrees), a fair agreement between experi- attack is increased, the flow lines tend to be aligned with the duct
ment and CFD based estimations can be observed. Even better profile at the lower part, producing lift in that region (Fig. 6b).
results can be observed in the prediction of the moment coeffi- Deducing the dead band effect from the ASE calculations, it is
cient, as indicated in Fig. 3(b). possible to observe a good agreement between ASE and CFD normal
The center of pressure location is better predicted by the CFD force predictions for all the angle of attack range considered. In the
estimations when compared to the analytical formula derived by case of the moment coefficient, the agreement is restricted to the
Morgan and Caster (1965), as can be observed in Fig. 4. There is linear angle of attack range. This may be explained by the viscous
still a discrepancy between experiment and the results derived by drag on the upper side, which is responsible for the change in the
both approaches in the stall region. However, in the higher angle total pressure center location.
E.A. de Barros, J.L.D. Dantas / Ocean Engineering 42 (2012) 61–70 65

Fig. 5. Comparison of the normal force (a) and pitch moment (b) coefficient estimated by CFD and ASE formulation (Morgan and Caster, 1965) for the 19A duct profile.

Fig. 6. Streamlines of the flow around the duct with 01 (a) and 251 (b) of angle of attack.

4. Duct body combination be taken into account by the prediction of the wake coefficient,
improving the ASE estimation.
The combination of bare hull and the 19A propeller duct is The dead band in the normal force, as predicted by the CFD
tested in the CFD simulations and experimental trials in a towing simulation, is not observed for the duct body combination.
tank. Flow simulations of the duct-body combination indicate an The next phase in the duct body combination study is the
important consequence of the bare hull effect on the incidence at prediction of the efforts based on analytical and semi-empirical
the duct. Flow lines converge following the duct profile shape, models, and the comparison to the experimental results.
and the stall is no longer observed at the low angles of attack The bare hull force and moment were predicted based on a
(Fig. 7). modification of the slender body theory, and related formula
Fig. 8a and b show the comparison between experimental derived by Munk (1923), and Hoak and Finck (1978). Results were
results and the CFD predicted normal force and moment coeffi- experimentally validated (de Barros et al., 2008a). The proposed
cients as function of the angle of attack. The AUV Reynolds formula for the bare hull normal force coefficient, CN(a), is given by
number based on the vehicle length is 1.8  106. The agreement
C N ðaÞ ¼ ðK 2 K 1 ÞC Np ðaÞð1sink aÞ þ C Nv ðaÞ, ð2Þ
between numerical and experimental results is very good, parti-
cularly in the linear range. The coefficients CNp(a) and CNv(a) denote the contribution of
Analytical and semi-empirical predictions applied to the nor- potential and viscous terms, respectively. In the equation above
mal force and moment in the bare hull, and in the body with duct
Sn
are also represented in the figures, for comparison. The bare C Np ðaÞ ¼ sinð2aÞ ð3Þ
Sref
hull prediction was based on the ASE expressions presented in
formulas (2)–(8). The body with duct case was considered by the and
simple addition of the bare hull contribution to that one provided
Sp
by the duct alone, as shown in Fig. 2. In this last case, no C Nv ðaÞ ¼ ZC dn sin2 a ð4Þ
Sref
interference between the flows at the hull and duct was con-
sidered. As expected, the experimental results lie between both where, Sn is the sectional area in a station at a longitudinal
curves. In the sequel, it is shown that the hull-duct interaction can distance xn0 from the nose tip. The coefficients K1 and K2 are,
66 E.A. de Barros, J.L.D. Dantas / Ocean Engineering 42 (2012) 61–70

Fig. 7. Streamlines of the flow in the duct with and without the presence of the bare hull.

Fig. 8. Comparison of normal force (a) and pitch moment (b) coefficients of body-duct combination.

respectively, the longitudinal and transversal apparent mass fac-


tors (Munk, 1923). The factor (1sinka) is related to the boundary
layer thickening, vorticity and cross-flow separation in the bare
hull (de Barros et al., 2008b). For the hull shape considered, k¼1.3
was adopted. The parameter Sp is the planform area at the xy plane
(Fig. 9), and Sref is the reference area (taken as L2 in this work). The
coefficient Z is a correction factor, equal to the ratio of C dn for a
finite length cylinder to that for an infinite cylinder. This parameter
is a function of the fineness ratio L/d (Jorgensen, 1977), where L is
the body length and d is the body maximum diameter.
The distance xn0 is the average value between the hull length and
Fig. 9. Definition of the considered coordinated system.
x0. The parameter x0 is adopted by the Datcom handbook (Hoak and
Finck, 1978), as the axial distance between the nose and the station,
which the flow can no longer be considered as potential. The semi- the flow visualization in Fig. 10a (de Barros et al., 2008a). The
empirical expression for estimating x0 is given by numerical flow simulations also indicate that the presence of the
duct does not provoke a change in the position of such station at the
x0 ¼ 0:378Lþ 0:527x1 ð5Þ
stern (Fig. 10b). In fact, the velocity profile at the stern suffers no
where x1 is the coordinate where the body profile has the most significant change because of the presence of the duct (Fig. 11).
negative slope in the aft direction. The moment coefficient Cm is defined by
The parameter xn0 is adopted as a better approximation than x0 Z xn0
calculated by (5), since it takes into account that a significant portion Cm ¼ cN ðxÞðxm xÞdx ð6Þ
of the flow at the stern can be considered as potential, according to 0
E.A. de Barros, J.L.D. Dantas / Ocean Engineering 42 (2012) 61–70 67

Fig. 10. Comparison of the flow streamlines and the station where the flow can no longer be considered as potential (x0) for the bare hull alone and in the presence of the
duct, for some angles of attack (a).

where xm is the axial distance from the nose tip to the center of rotation, and Vn is the volume between the nose tip and such
rotation and station. The parameter xc is the axial distance from the nose tip to
the centroid of the bare-hull planform area at the xy plane.
dSðxÞ
cN ðxÞ ¼ ðK 2 K 1 Þsinð2aÞ ð1sink aÞ þ 2ZC d sin2 ðaÞrðxÞ ð7Þ The prediction of the normal force for the combination should
dx
take into account the effect of the hull wake at the duct location.
with r(x) denoting the cross section radius at the axial distance x The ratio between the dynamic pressures at the duct and that of
from the nose tip and S(x) is the corresponding cross-sectional area. the undisturbed flow defines the efficiency factor as follows:
From formulas (6) and (7), the final expression for the moment
coefficient results q V2 V 2 ð1oÞ2
 n n n  Zo ¼ ¼ 2 ¼ 1 2 ¼ ð1oÞ2 ð9Þ
V S ðx0 xm Þ Sp xm xc  2 q1 V1 V1
Cm ¼ 3
sinð2aÞð1sink aÞ þ ZC dn 2 sin ðaÞ,
L L L
The wake coefficient, o, was estimated by two different
ð8Þ
approaches. The first one is an ASE method. Jackson (1992)
where Sn is the cross-sectional area at the station distant xn0 from presented a formula for estimating the wake coefficient based
the nose, xm is the axial distance from the nose tip to the center of on the relationship between the bow and stern surface areas in
68 E.A. de Barros, J.L.D. Dantas / Ocean Engineering 42 (2012) 61–70

Fig. 11. Comparison of the velocity profile in the bare hull tail with and without the presence of the duct. The angle of attack is zero.

the case of typical submarine hull geometries. The wake coeffi- The duct normal force is composed by the lift and drag forces.
cient is calculated from the expression Expressing them in the non-dimensional form it follows:
0:01382 D D C NDuct ðaÞ ¼ C LDuct ðaÞcos aC DDuct sin a ð15Þ
ð1oÞ ¼ 0:3674þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ 0:008406 þ 1:6732 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ,
ðL=dÞK o d d ðL=dÞK o where, C LDuct ðaÞ is given by
ð10Þ
dC LDuct
C LDuct ðaÞ ¼ a ð16Þ
where L is the hull length, d is the hull maximum diameter, and D da
is the propeller diameter. The coefficient Ko is proportional to the The lift slope coefficient (dC LDuct =da) is given in Fig. 2a. The drag
wetted surface area coefficients of the forebody and afterbody, as coefficient was adopted as the experimental value presented by
follows: Morgan and Caster (1965) for the forward thrust increase type
K o ¼ 62:4C wsf 3:6C wsa ð11Þ of duct
C d Rd
where, C DDuct ¼ 0:48 ð17Þ
L2
Sf
C wsf ¼ , ð12Þ For composing the moment coefficient, it was assumed the
pdLf algebraic summation of the bare hull contribution (Eq. (9)) to the
and duct contribution calculated as follows:

Sa ðxm xLE Þ dC mDuct


C wsa ¼ ð13Þ C mDuct ðaÞ ¼ C NDuct ðaÞZo þ a ð18Þ
pdLa L da
where, xLE is the axial distance from the nose tip to the duct
where Sf, Lf are the forebody wetted area and length respectively,
leading edge, and dC mDuct =da is the moment slope coefficient,
while Sa, La are the same parameters related to the afterbody.
related to the duct leading edge, as given in Fig. 2b.
The second approach for estimating the wake coefficient is
In Fig. 12a and b the experimental and CFD results are
based on the velocity distribution at the duct station produced by
compared to the curves generated by the ASE method. In those
the numerical flow simulation. Considering the flow around the
figures, the normal force and moment coefficients variation with
bare hull (without the presence of the duct), the velocity field was
the angle of attack are displayed according to different conditions.
estimated by the CFD just at the zero incidence case by a single
They correspond to the bare hull case, the duct alone case, and the
2-D simulation. An average value was calculated from the velocity
duct-hull combination with and without the influence of the hull
distribution at the station relative to the virtual location of the
wake. All the curves belong to the area defined by the bare hull
duct hydrodynamic center (taken at 25% of its chord length from
case, and the curve generated by the algebraic summation of bare
the leading edge).
hull estimative with the duct alone case.
R ro R ro
r VðrÞrdr r i VðrÞrdr
Comparing to the bare hull generated efforts, as expected, the
V ¼ V 1 ð1oÞ ¼ iR ro ¼ 2 r 2 Þ
ð14Þ duct contribution is significant, especially for the pitch moment.
r rdr ð1=2Þðr o i
i
Two different curves are generated for each estimated value
where ri is taken as the propeller boss radius (0.015 m), and r0 is of the wake coefficient. The Jackson method (Eqs. (8) and (9))
the duct external radius (0.0908 m) at the hydrodynamic center produced a wake coefficient equal to 0.22, while the value
position. generated by the analysis of the CFD simulation is 0.36. It can
According to numerical predictions, the velocity distribution at be observed also that the best fit to experimental results, mainly
the body stern is not significantly affected by the presence of the in the moment coefficient, are provided by the ASE using the
duct, as indicated by Fig. 11. The presence of the duct should not wake coefficient value estimated through the CFD simulation.
affect as well the approximation to the station of the body where
the flow ceases to be considered potential (see Fig. 10). Therefore, 5. Conclusions
it is possible to consider the summation of the normal force
generated on the bare hull (Eq. (2)) to the duct contribution This paper presented a comparative study of CFD and ASE
multiplied by the efficiency factor defined in Eq. (9). methods applied to the prediction of normal force and moment
E.A. de Barros, J.L.D. Dantas / Ocean Engineering 42 (2012) 61–70 69

Fig. 12. Normal force (a) and pitch moment (b) coefficient of the bare hull plus duct combination, for values of wake coefficient coming from analytical and numerical
methods.

coefficients of a Myring type hull of an AUV combined to a ESDU Aerodynamics, 1978. The influence of body geometry and flow conditions on
propeller duct. The methods were applied to the body-duct axisymmetric velocity distribution at subcritical Mach numbers. ESDU Inter-
combination. national, London, UK, Data Item 78037.
Falca~ o de Campos, J.A.C., 1983. On the calculation of ducted propeller performance
The CFD calculations have produced very good predictions of in axisymmetric flows. Ph.D. Thesis. Delft University.
the normal force and moment coefficients for the bare-hull and Fletcher, H.S., 1957. Experimental investigation of lift, drag, and pitching moment
duct combination. This makes the information given by the flow of five annular airfoils. National Advisory Committee for Aeronautics TN 4117.
Fluent Inc., 2005. Fluent 6.2 User’s Guide. Fluent Inc.
visualization and pressure distribution very useful during the Hoak, D., Finck, 1978. USAF Stability and Control Datcom. Wright-Paterson Air
selection or tuning of the ASE formulas. Force Base, Ohio.
The accuracy of the estimations when considering the vehicle Hoekstra, M., 2006. A RANS-based analysis tool for ducted propeller systems in
length up to a station in between the base and that one suggested open water condition. International Shipbuilding Progress 53, IOS Press,
Maritime Research Institute Netherlands, Wageningen, The Netherlands.
by the Datcom approach has not been affected by the duct Husaini, M., Samad, Z., Arshad, M.R., 2008. Optimum Design of URRG-AUV
presence at the stern. Propeller Using PVL. 2nd Technical Seminar on Underwater System Technol-
The Morgan Caster formulation for the normal force and ogy: Breaking New Frontiers 2008.
Inoue, T., Suzuki, H., Kitamoto, R., Watanabe, Y., Yoshida, H., 2010. Hull form
moment generated by the duct provided good results for the
design of underwater vehicle applying CFD (Computational Fluid Dynamics).
19A profile. IEEE, 1–5.
The analysis of the duct-hull combination, as in the case of the Jackson, H.A., 1992. Fundamental of Submarine Concept Design. SNAME Transac-
bare hull normal force and moment coefficients (de Barros et al., tions 100, 419–448.
Jagadeesh, P., Murali, K., Idichandy, V.G., 2009. Experimental investigation of hydro-
2008b), shows that combining ASE formula to CFD results can dynamic force coefficients over AUV hull form. Ocean Eng. 36 (1), 113–118.
provide qualitative understanding and more physical meaning to Jeans, T.L., Holloway, A.G.L., Watt, G.D., Gerber, A.G.A., 2010. A force estimation
experimental and CFD results, using quite simple hydrodynamic method for viscous separated flow over slender axisymmetric bodies with
tapered tails. J. Ship Res. 54 (1), 53–67.
concepts. Moreover, the ASE estimation could be improved using
Jorgensen, L.H., 1977. Prediction of static aerodynamic characteristics for slender
a CFD based result based on a single case, without requiring the bodies alone and with lifting surfaces to very high angles of attack. NASA
usual long time and computer resources related to the numerical Technical Report. TR-R-474.
simulation approaches. Kerwin, E.J., Kinnas, S.A., Lee, J.-T., Shih., W.-Z., 1987. A surface panel method for
the hydrodynamic analysis of ducted propellers. SNAME Trans. 95, 93–112.
Next steps will include the propeller modeling inside the duct Kimber, N.I., Scrimshaw, K.H., 1994. Hydrodynamic testing of a 3/4 scale Autosub
under incidence. The inclusion of such effect may change the model. In: Oceanology International 94.
pressure distribution over the vehicle stern so that the moment Kinnas, S.A., Lee, H., Gu, H., Deng, Y., 2005. Prediction of performance and design
coefficient would be affected. However, it is the authors’ belief via optimization of ducted propellers subject to non-axisymmetric inflows.
SNAME Trans., 99–121.
that changes in the normal force and moment coefficients will be Lewis, R.I., Ryan, P.G., 1972. Surface vorticity theory for axisymmetric potential
not so significant when compared to the results presented in flow past annular aerofoils and bodies of revolution with application to ducted
this work. propellers and cowls. J. Mech. Eng. Sci. 14 (4).
Minsaas, K.J., G.M. Jacobsen, H. Okamoto, 1973. The design of large ducted
propellers for optimum efficiency and manoeuvrability. Symposium on Ducted
Propellers. The Royal Institution of Naval Architects. pp. 134–160.
Menter, F.R., 1994. Two-equation eddy-viscosity turbulence models for engineer-
ing applications. AIAA J. Am. Inst. Aeronaut. Astronaut. 32 (8), 1598–1605.
References Morgan, W.B., Caster, E.B., 1965. Prediction of Aerodynamic Characteristics of
Annular Airfoils. Department of TheNavy, Washington, D.C. Report1830.
Baltazar, J., Falca~ o de Campos, J.A.C., 2009. On the modelling of the flow in ducted Munk, M.M., 1923. The Aerodynamic Forces on Airship Hulls. NACA Rep. p. 184.
propellers with a panel method. In: Proceedings of the First Symposium in Phillips, A., Furlong, M., Turnock, S., 2007. The use of computational Fluid dynamics
Marine Propulsors. Norway. to determine the dynamic stability of an autonomous underwater vehicle. In:
de Barros, E.A., Pascoal, A., de Sá, E., 2008a. Investigation of a method for predicting 10th Numerical Towing Tank Symposium (NuTTS’07). Hamburg, Germany, p. 6.
AUV derivatives. Elsevier. Ocean Engineering, pp. 1627–1636. Phillips, A., Furlong, M., Turnock, M., 2010. Influence of turbulence closure models
de Barros, E.A., Dantas, J.L.D., Pascoal, A.M., de Sá, E., 2008b. Investigation of on the vortical flow field around a submarine body undergoing steady drift.
normal force and moment coefficients for an auv at nonlinear angle of attack J. Mar. Sci. Technol. 15 (3), 201–217.
and sideslip range. IEEE J. Oceanic Eng., 538–549. Silvestre, C., Pascoal, A., Kaminer, I., and Healey, A., 1998. Combined plant/
Ec-a, L., Vaz, G., Hoekstra, M., 2010. Code verification, solution verification and controller optimization with application to autonomous underwater vehicles.
validation in RANS solvers. ASME Conference Proceedings, ASME, vol. 2010, no. In: Proceedings of the Control Applied Marine System, Fukuoka, Japan.
49149, p. 597–605. pp. 361–366.
70 E.A. de Barros, J.L.D. Dantas / Ocean Engineering 42 (2012) 61–70

Tang, S., Ura, T., Nakatani, T., Thornton, B., Jiang, T., 2009. Estimation of the International Conference on Ocean, Offshore and Arctic Engineering. Shanghai,
hydrodynamic coefficients of the complex-shaped autonomous underwater China, vol. 6, pp. 621–633.
vehicle tuna-sand. J. Mar. Sci. Technol. 14 (3), 373–386. Wu, L., 2010. Applying dynamic hybrid grids method to simulate auv docking with
Tyagi, A., Sen, D., 2006. Calculation of transverse hydrodynamic coefficients using a tube. In: Information and Automation (ICIA), 2010 IEEE International
computational fluid dynamic approach. Ocean Eng. 33 (5–6), 798–809. Conference. pp. 1363–1366.
Van Gusteren, L.A., van Gusteren, F.F., 1972. The effect of a nozzle on steering Yamamoto, I., 2007. Research and development of past, present and future AUV
characteristics. Int. Shipbuilding Prog., 139–151. technologies. In: Masterclass in AUV Technology for Polar Science: British
Vaz, G., Toxopeus, S., Holmes, S., 2010. Calculation of manoeuvring forces on Library, (469, 4), pp. 99–102.
submarines using two viscous-flow solvers. In: OMAE2010. ASME 29th

You might also like