Alexandria Engineering Journal (2018) xxx, xxx–xxx

H O S T E D BY
Alexandria University

Alexandria Engineering Journal

www.elsevier.com/locate/aej
www.sciencedirect.com

ORIGINAL ARTICLE

Numerical prediction of sheet cavitation on marine

propellers using CFD simulation with transition-
sensitive turbulence model
Mohamed M. Helal a,*, Tamer M. Ahmed b, Adel A. Banawan b, Mohamed A. Kotb c

a
VSE Corporation, Alexandria, Egypt
b
Dept. of Naval Architecture and Marine Engineering, Faculty of Engineering, Alexandria University, Alexandria, Egypt
c
Arab Academy for Science, Technology and Maritime Transport, Alexandria, Egypt

Received 3 January 2018; revised 18 February 2018; accepted 12 March 2018

KEYWORDS Abstract One of the big challenges, yet to be addressed, in the numerical simulation of cavitating
CFD simulations; flow on marine propellers is; the existence of laminar and turbulence transition flows over the pro-
Marine propellers; peller’s blades. The majority of previous studies employed turbulence models that were only appro-
Cavitation; priate for fully turbulent flows. These models mostly caused high discrepancies between numerical
Multi-phase flow; predictions and experimental measurements especially at low rotational speeds where, Reynolds
Turbulence models; number decreases and laminar and transient flows exist. The present paper proposes a complete
Transition-sensitive models and detailed procedure for the CFD simulation of cavitating flow on marine propellers using the
‘K-Kl-x’ transition-sensitive model. Results are obtained using ‘ANSYS FLUENT 16’. The pro-
peller under consideration is the ‘INSEAN E779A’ propeller model. The fully turbulent standard
‘k-e’ model is also adopted for comparison. Obtained results, based on both turbulence models,
are validated by comparison with experimental data available in the literature. Predictions based
on the ‘K-Kl-x’ transition-sensitive model are found to be in better agreement with experiments
at lower rotational speeds i.e. at low Reynolds numbers.
Ó 2018 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V. This is an
open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction defined cavitation as: ‘‘the evaporation of a liquid when its

pressure decreases below its saturation pressure”. Applying
Cavitation is a widespread phenomenon that might exist in this principle to marine propellers means that; when a pro-
various marine applications. Prediction of cavitation, espe- peller runs in water at high rotational speeds, the water’s local
cially for marine propellers, is very important to avoid –or pos- pressure decreases. This decrease in pressure is proportional to
sibly eliminate– its adverse effects. Franc and Michel [1] the local flow velocity squared. If the pressure of water drops
below the vapor formation pressure corresponding to the
* Corresponding author. flow’s temperature, bubbles of vapor (cavities) form. Subse-
E-mail address: [email protected] (M.M. Helal). quently, these cavities collapse with an explosive manner
Peer review under responsibility of Faculty of Engineering, Alexandria resulting in severe pitting to the blade surface. If the cavitation
University. extent is sufficiently large, adverse effects will take place such
https://doi.org/10.1016/j.aej.2018.03.008
1110-0168 Ó 2018 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Please cite this article in press as: M.M. Helal et al., Numerical prediction of sheet cavitation on marine propellers using CFD simulation with transition-sensitive
turbulence model, Alexandria Eng. J. (2018), https://doi.org/10.1016/j.aej.2018.03.008
2 M.M. Helal et al.

Nomenclature

cp pressure coefficient g efficiency

D diameter (m) l viscosity (Pa s)
J advance ratio t kinematic fluid viscosity (m2 s1)
k turbulent kinetic energy per unit mass (m2/s2) q density (kg/m3)
K turbulent kinetic energy (J) r cavitation number
Kl laminar kinetic energy (J) swall shear stress at the wall (N/m2)
KQ torque coefficient x inverse turbulent time scale (s1)
KT thrust coefficient
L length (m) Subscribts
n rotational speed (rps) m mixture
p local pressure (Pa) L liquid
p1 free stream pressure (Pa) V vapor
pv vapor pressure (Pa) Charact. characteristic
Q torque (N m)
r radius (position) of any blade section (m) Abbreviations
R radius of the propeller (m) CFD computational fluid dynamics
Re Reynolds number INSEAN Istituto nazionale per studi ed esperienze Di
T thrust force (N) architettura navale (Research Institute active
U incident velocity (m/s) in the field of naval architecture and marine
V total velocity (m/s) engineering within the frame of the National
VA advance velocity (m/s) Research Council of Italy).
y+ non-dimensional normal distance from the wall PROPCAV propeller cavitation research
y normal distance from the wall (m) BEM boundary element method
RANS Reynolds-averaged Navier-Stokes equation
Greek symbols MRF multiple reference frame
a vapor volume fraction
e dissipation rate (W)

as; a decrease in produced thrust, an increase in required tor- merged super-cavitating propellers along with experimental
que, damage to the propeller’s material (erosion), strong vibra- measurements. Viscous flow methods for predicting cavitating
tions and, noise. Using photography–via– experiments to flows around two-dimensional foils started getting in use since
capture cavitation patterns and extents over marine propellers’ the 1990s.
blades is a very complex task. Experiments are, in general, Numerical cavitation methods based on Reynolds Aver-
expensive to setup and time consuming to conduct. aged Navier-Stokes (RANS) models have been developed to
Numerical predictions of cavitating flows over marine pro- deal with viscous flows and, are currently still an active topic
pellers’ blades using CFD simulations have become a good of research. The first results obtained through RANS simula-
alternative to experiments and subsequently, have become a tion of cavitating marine propellers were obtained by Mak-
trending research topic. Researchers nowadays may harvest soud [10] and watanabe et al. [11]. Salvatore et al. [12]
the great advancement of computer performance that has carried out computational predictions using the ‘INSEAN’
made these simulations possible; refer to Blazek [2] as well as propeller flow code. In addition, experiments were carried
Versteeg and Malalasekera [3]. out to validate the predicted cavitation under uniform flow.
Previously, flow about propellers had been predicted using Zhu and Fang [13] investigated the performance characteristics
the lifting-line theory with, a vortex line representing the pro- of propellers under cavitation using viscous multiphase flow
peller’s blade and helicoidal vortices representing the pro- methods based on Navier-Stokes and bubble dynamics equa-
peller’s wake. Numerical models developed swiftly starting tions. Pereira et al. [14] ran an experimental and theoretical
from the 1960s. Salvatore et al. [4] utilized the perturbation study of a cavitating propeller in uniform inflow. Advanced
methods to demonstrate the lifting-line theory. The lifting sur- imaging techniques were utilised to obtain quantitative data
face model was developed later on. Dang [5] and Vaz [6] pro- on the extent of the cavity. Pereira and Sequeira [15] developed
vided the possibility for boundary element methods (BEM), a turbulent vorticity confinement strategy for RANS based
also referred to as the boundary integral or panel methods, industrial propeller flow simulations. The aim of the study
to be considered for simulating the flow about two- was to improve the prediction of tip vortices.
dimensional geometrically complex bodies. The application In general and from the presented brief literature, it is
of BEM was first utilized for partially cavitating flows of apparent that many researchers have significantly contributed
two-dimensional foils by Uhlman [7] and then by Lee et al. to the prediction of propellers’ cavitating flow using CFD sim-
[8]. Kinnas and Young [9] introduced the ‘PROPCAV’ method ulations to address various aspects of the problem. Among
employing BEM for the numerical prediction of fully sub- several challenges, not yet fully investigated is; the existence

Please cite this article in press as: M.M. Helal et al., Numerical prediction of sheet cavitation on marine propellers using CFD simulation with transition-sensitive
turbulence model, Alexandria Eng. J. (2018), https://doi.org/10.1016/j.aej.2018.03.008
Numerical prediction of sheet cavitation on marine propellers 3

of laminar and turbulence transition flows over the propellers’ @ ! ! ! !

ðq V m Þ þ r  ðq V m V m Þ ¼ rp þ r  ½lm ðr V m
blades especially, at low rotational speeds i.e. at low Reynolds @t
!
numbers. It is necessary to take into account the transient flow þ rVT Þ þ q !
m g þF m ð2Þ
problem by employing transition-sensitive models and, to
avoid the assumption that the flow is fully turbulent over the Eq. (2) represents the multi-phase RANS equation where,
whole propeller. In this paper, the detailed procedure of a the left hand side represents changes in the mean momentum
CFD simulation is proposed for solving the three- of a fluid element due to unsteadiness in the mean flow and
dimensional, viscous, cavitating flow of a marine propeller. is attributed to convection by the mean flow. These changes
Subsequently, the cavitation patterns, extents and, the pro- are balanced by the mean body force. As known from conven-
peller’s performance characteristics under cavitation are tional RANS equations for single-phase flow and for the
predicted. apparent stress to close, additional modeling is required. This
Additionally, the paper investigates the influence of the has led to the development of many different turbulence mod-
existence of transition flow on the predictions. This is achieved els. A turbulence model is a tool for specifying the Reynolds
by using two different turbulence models, a transition-sensitive stresses hence, closing the mean-flow equations.
model and, a fully turbulent model, namely, the ‘K-Kl-x’ tran- For multi-phase flows, the density constitution of each
sition model [16,17] and the standard ‘k-e’ model [17], phase in a fluid cell is represented by a scalar volume fraction
respectively. as follows:
The CFD software ‘ANSYS FLUENT 160 is used to per- qm ¼ aqv þ ð1  aÞqL ð3Þ
form the simulations where, both turbulence models are imple-
mented. The investigation is carried out for the ‘INSEAN In a similar manner
E779A’ propeller model. Results from the two turbulence lm ¼ alv þ ð1  aÞlL ð4Þ
models are assessed against experimental data available in
the literature [14,18]. Finally and to solve the system of the governing equations,
turbulence closures are used. As clarified earlier, the transition-
sensitive ‘K-Kl-x’ and the standard ‘k-e’ turbulence models are
2. Theoretical analysis and numerical methods
employed in this paper for this purpose. The ‘k-e’ model [17]
deals with mechanisms that influence the turbulent kinetic
2.1. Performance characteristics of marine propellers
energy. It is a two-equation model which means; it solves
two additional partial differential equations for k and e in
The performance of propellers is traditionally represented in order to specify the Reynolds stress. The ‘k-e’ model is
terms of; the thrust coefficient KT, the torque coefficient KQ employed in the simulation presented here within as it con-
and the efficiency g, together with, their variation with the verges relatively easily and, it has been reported suitable for
advance coefficient J. These parameters are introduced as simulating various turbulent flows. On the other hand, the
follows: ‘K-Kl-x’ transition model [16,17] is used to predict the bound-
T Q J KT VA ary layer development and transition onset. This model is
KT ¼ ; KQ ¼ ; g¼ ; J¼ employed in the simulation presented here within as it
qn2 D4 qn2 D5 2p KQ nD
addresses –effectively- the transition of the boundary layer
from laminar to turbulent regime specially, in rotating flows.
2.2. Cavitation calculations The model is a three-equation, eddy-viscosity type of model
which includes transport equations for K, Kl and x. One
The onset of the process of cavitation is usually termed ‘cavi- unknown variable is yet to be identified in Eqs. (3) and (4),
tation inception’. The cavitation number r and the pressure namely, the vapor volume fraction, a, which is obtained using
coefficient cp may be defined as given below: the mass transfer (cavitation) model.
p  pv p  p1
r¼11 2
; cp ¼ 1 2.4. Mass transfer models
2
qðnDÞ 2
qðnDÞ2
The traditional cavitation criterion based upon r and cp Mass transfer models idealise the interphase mass transfer rate
can be given as: If r 6 cp , cavitation occurs. due to cavitation. Several mass transfer models have been pro-
posed by researchers. In the current study, the ‘full cavitation
2.3. Multi-phase RANS method model’ which was developed by Singhal et al. [19] and is avail-
able in ‘ANSYS FLUENT 160 , is employed.
The cavitating flow is a mixture of two phases (vapor and liq-
uid) that are simulated as one phase. Both phases share the 3. Geometric modeling
same velocity and pressure fields. Assuming that the mixture
is homogeneous, the multi-phase flow can be solved with con- The propeller selected as a test case for this paper is the
ventional RANS equations after some modifications. The con- ‘INSEAN E779A’ propeller model which is a four bladed,
tinuity equation of the mixture flow is given as: fixed-pitch, right-handed propeller. Geometric modeling of
@ ! this propeller is obtained using the ‘HYDROCOMP PROP-
ðq Þ þ r  ðqm V m Þ ¼ 0 ð1Þ
@t m CAD’ software as shown in Fig. 1. All geometrical data of this
propeller is available in the literature [20] as summarized in
And the momentum equation can be expressed as:
Table 1.

Please cite this article in press as: M.M. Helal et al., Numerical prediction of sheet cavitation on marine propellers using CFD simulation with transition-sensitive
turbulence model, Alexandria Eng. J. (2018), https://doi.org/10.1016/j.aej.2018.03.008
4 M.M. Helal et al.

5. Grid generation

To generate an unstructured grid with hybrid cells, ‘ANSYS

FLUENT 160 is used for grid generation. As concluded by
Morgut and Nobile [21], this type of meshing is preferred since;
hybrid meshes give higher accuracy for numerical prediction of
propulsive performance. In addition, hybrid meshes need less
effort in generation as compared to hexa-structured meshing.
The mesh is generated taking into account that; the size of
cells near the blade wall is small whereas, size increases
towards the outer boundary. Fig. 3 shows size of the grid cells
over the outer surface of the domain while Fig. 4 clarifies the
size of hybrid cells inside the domain. After convergence, the
total number of cells generated over the entire grid is
238,387. To obtain a suitable resolution of the viscous sub-
layer, the wall normal resolution over all surfaces of the blades
is within a range of: y+ < 0.5 where, y+ is defined as:
qffiffiffiffi
Fig. 1 The ‘INSEAN E779A’ propeller model developed in sw
y
q
‘PROPCAD’. yþ ¼
t

Table 1 Geometric details of the ‘INSEAN E779A’ Propeller 6. Boundary conditions

model.
E779a Propeller model The numerical simulation is performed over a range of
Propeller Diameter 227.27 mm advance coefficients and cavitation numbers equivalent to
Number of blades 4 the range at which the propeller was experimentally tested
Pitch ratio 1.1 [14,18].
Skew angle at blade tip 4°480 (positive) Values assigned to the advance coefficient and cavitation
Rake (nominal) 4° 350 (forward) number are as follows:
Expanded area ratio 0.689
Hub diameter (at prop. Ref. line) 45.53 mm  J = 0.9, r = 4.455
Hub length 68.30 mm  J = 0.83, r = 2.063
 J = 0.77, r = 1.783
 J = 0.71, r = 1.763

4. Computational domain To obtain these values, various propeller rotational speeds

of (1500 rpm, 1800 rpm, 2400 rpm and 3000 rpm) are used.
For the simulation under consideration, the computational The advance velocity values prescribed at the domain inlet
domain is taken to be a cylinder as shown in Fig. 2. While are calculated and are as listed in Table 2. The free stream
the inlet is placed at a distance of 4D upstream from the pro- pressure p1 used in each run is calculated based on the
peller plane, the outlet is placed at a distance of 5D down- corresponding value of r. In this study, it is assumed that
stream from the propeller plane. In the radial direction, the Poutlet = p1. All boundary conditions imposed are shown in
domain radius is extended up to a distance of 2.5D from the
axis of the hub.

Fig. 3 Grid over the domain surface developed in ‘ANSYS-

Fig. 2 Design of the computational domain. FLUENT’.

Please cite this article in press as: M.M. Helal et al., Numerical prediction of sheet cavitation on marine propellers using CFD simulation with transition-sensitive
turbulence model, Alexandria Eng. J. (2018), https://doi.org/10.1016/j.aej.2018.03.008
Numerical prediction of sheet cavitation on marine propellers 5

 For unsteady simulations, a second order implicit scheme is

applied for time derivatives. The selected time step at each
run corresponds to a rotation of 1°.

The simulation is carried out by nine processors (Intel Ò

Core(TM)2 Duo CPU E8400@ 3 GHZ). Table 3 shows details
of the simulation settings.

8. Frame of reference

The frame of reference of the ‘INSEAN E779A’ propeller

model adopted in this study is shown in Fig. 6 where, the ‘x-
axis’ points downstream, the ‘y-axis’ is directed upwards
and, the ‘z-axis’ follows the right hand rule in the ‘x-y’ plane.
Viewing form aft., the direction of rotation is clockwise as the
Fig. 4 Longitudinal section of the computational domain propeller is a right-handed one.
showing the size of cells.
9. Reynolds number distribution over the blades

Table 2 Flow parameters of each run. Reynolds number varies along each blade with variations in
n (rps) VA (m/s) J r
the rotational speed and at every radius ratio r/R. Subse-
quently, and along the blade, the flow could change from lam-
25 5.1 0.9 4.455 inar to transient to fully turbulent as Reynolds number values
30 5.65 0.83 2.063
increase. Reynolds number Re is defined as:
40 6.99 0.77 1.783
50 8 0.71 1.763 Ucharact Lcharact
Re ¼
tfluid
For a propeller blade, Reynolds number may be estimated
Fig. 5. The continuum is chosen to be fluid with properties of based on several characteristic dimensions, refer to literature
water assigned to it. The propeller rotation is simulated using [22] and [23]. In this study, it has been opted for a Reynolds
the Multiple Reference Frame (MRF) approach [17]. For the number that is based on (Lcharct = r) and characteristic veloc-
stationary region (the outer far field wall), the governing equa- ity (Ucharct = 2npr) at each radius ratio, i.e. at each r/R. The
tions are solved in a fixed frame of reference while for all rotat- variation of Re with n for various blade sections (r/R = 0.3,
ing regions (blade, hub and fluid zone), the governing 0.5, 0.7 and 1) is presented in Table 4 and Fig. 7. According
equations are solved in a rotating frame of reference. to De Witt [24], the threshold for the turbulent region is at
A rotating frame of reference is assigned to the fluid at the Re = 0.5  106. It can be noted from Table 4 and Fig. 7 that
selected absolute speeds given earlier in Table 2. A relative the flow becomes fully turbulent as the rotational speed
rotational speed of zero –with respect to adjacent cell zones- increases and beyond a radius ratio of r/R = 0.5.
is prescribed to the walls of the propeller’s blade and hub. From that, the standard ‘k-e’ turbulence model is expected
The far field boundary is treated as an inviscid wall to which, to be more accurate for the higher rotational speeds at (r/R ˃
an absolute rotational speed of zero is prescribed. The four 0.5) since the flow is fully turbulent. On the other hand and for
blades of the propeller are set at a regular angular interval of the range of radius ratios (r/R < 0.5), transient flow takes
90° hence, the modeling of one angular sector of 90° and place and laminar flow exists specifically as the rotational
one blade will suffice in solving the entire flow domain as speed decreases i.e. at low Re. Over this range, the ‘K-Kl-x’
shown previously in Fig. 5. The fact that other blades are pre- transition model is expected to be well suited.
sent is taken into account by imposing periodic boundary con-
ditions on the two sides of the blade. On these periodic 10. Results and discussion
boundaries, rotational periodicity is ensured.
10.1. Prediction of cavitation patterns and extents
7. Solution and solver settings
In this section, predictions of the pressure distribution, based
The commercially available CFD software, ‘ANSYS FLU- on the ‘k-kl-x’ transition-sensitive turbulence model, is pre-
ENT 160 is adopted to solve the three- dimensional viscous, sented. This distribution plays a significant role in predicting
unsteady, cavitating problem. The equations are solved as the cavitation inception. Later on, cavitation patterns -at the
follows: selected conditions in Table 2 will be shown based on both tur-
bulence models. As mentioned earlier, the relevant experimen-
 The pressure and velocity components are solved using cou- tal data of the cavitating flow for this propeller model can be
pled technique. found in the literature, Pereira et al. [14] and Salvatore et al.
 The second order ‘QUICK’ scheme is applied for convec- [18]. It covers the following (for various advance coefficients
tion terms in all transport equations. and cavitation numbers):

Please cite this article in press as: M.M. Helal et al., Numerical prediction of sheet cavitation on marine propellers using CFD simulation with transition-sensitive
turbulence model, Alexandria Eng. J. (2018), https://doi.org/10.1016/j.aej.2018.03.008
6 M.M. Helal et al.

Fig. 5 Boundary Conditions: (a) 3D view. (b) Front view.

Table 3 Solver settings.

Solver settings
Pressure link SIMPLE
Discretization scheme for convective fluxes and QUICK
turbulence parameters
Turbulence models 1. Standard (k-
Є)
2. Transition
(K-Kl-x)
Near wall treatment Standard wall
functions
Phases 1. Water
2. Water vapor
Solver Unsteady
Absolute Vapor pressure (KPa) 2.337
CPU time/processor/run 1–28 h (k-Є)
2–30.2 h (K-Kl-
x)
Fig. 6 Frame of reference.

Please cite this article in press as: M.M. Helal et al., Numerical prediction of sheet cavitation on marine propellers using CFD simulation with transition-sensitive
turbulence model, Alexandria Eng. J. (2018), https://doi.org/10.1016/j.aej.2018.03.008
Numerical prediction of sheet cavitation on marine propellers 7

Table 4 Variation of Reynolds number (Re) with rotational

speed (n) at various blade sections.
r/R n (rps) Re 106
1 25 2
30 2.4
40 3.2
50 4
0.7 25 0.99
30 1.1
40 1.5
50 1.98
0.5 25 0.5
30 0.6
40 0.8
50 1
0.3 25 0.16
30 0.19
40 0.25
50 0.3

Fig. 8 Pressure distribution on the face and back sides of a blade

section. r/R = 0.7, n = 50 rps and H = 135°.

Fig. 7 Variation of Reynolds number (Re) with rotational speed

(n) at various blade sections.

 Photographs of cavitation patterns.

 Measurements of cavitation extents.
 Measurements of performance characteristics.

At the end of this section, predictions of the values of cav-

itation extents, based on both turbulence models, will be vali- Fig. 9 Pressure distribution on the face and back sides of a blade
dated by comparison with corresponding experimental section. r/R = 0.7, n = 50 rps and H = 180°.
measurements.
Figs. 8–10 show variations of the negative pressure coeffi-
cient (cp) with the position ratio (x/c) over the face and back to H = 90°. Moreover, it is clearly observed that cavitation
sides of a blade section at (r/R = 0.7) and (n = 50rps) for inception is more likely to occur over the back of the blade sec-
(H = 135°), (H = 180°) and (H = 225°), respectively. By tion (suction side) and close to the leading edge.
comparing the three figures it can be deduced that; the peak An important issue that has been investigated in the past is;
value of the negative pressure coefficient (cp) increases with how to define the cavitation area [18], with two possible
a decrease in the hydrostatic pressure, i.e. a higher risk of cav- answers; either as an iso-surface of (cp = r) or based on
itation. Subsequently, the probability of cavitation inception the iso-surface of vapor volume fraction (a = const). This
decreases as the blade rotates from H = 90° to H = 270° point is addressed in this study where, the cavitation area
and then increases again as the blade rotates from H = 270° extent over the ‘E779A’ propeller blade is obtained based on

Please cite this article in press as: M.M. Helal et al., Numerical prediction of sheet cavitation on marine propellers using CFD simulation with transition-sensitive
turbulence model, Alexandria Eng. J. (2018), https://doi.org/10.1016/j.aej.2018.03.008
8 M.M. Helal et al.

This indicates that the cavitation criterion of (cp = r) is

not reliable in defining cavitation.
On the other hand, a variation of the vapor volume fraction
a over the suction side of the propeller blade at the selected
parameters is obtained based on both turbulence models. By
comparing results predicted at various values of a with exper-
iments, it is found that a value of (a = 0.5) led to the best cav-
itation predictions in terms of location, shape and extent.
Subsequently, it is decided that cavitation predictions will be
carried out based on the criterion of iso-suface of vapour vol-
ume fraction at (a = 0.5).
Fig. 12 shows a comparison of predictions of cavitation
patterns over the suction side of a blade, based on both turbu-
lence models and at (a = 0.5), against corresponding patterns
from photographs obtained experimentally. Comparisons
reveal that the predicted cavitating regions, at the four operat-
ing conditions and based on both turbulence models, are in
good agreement with those in the corresponding photographs
[14,18]. Both predictions and photographs exhibit quite similar
cavitation areas on the blade surface in terms of location,
shape and extent. This demonstrates the correctness of the pro-
posed simulation procedure as well as the validity of selection
of the iso-surface of vapor volume fraction criterion and its
assigned value.
Fig. 10 Pressure distribution on the face and back sides of a
In addition, the cavitation regions are all found to be close
blade section. r/R = 0.7, n = 50 rps and H = 225°.
to the leading edge of the suction side of the blade which
emphasizes the previously drawn observations from Figs. 8–
both cavitation criteria; (a = const) and (cp = r) and, is 10.
assessed against photographs obtained experimentally to judge An overall comparison of predictions and experiments
which of these criteria is more accurate in defining the cavita- shown in Fig. 12 reveals that; the ‘K-Kl-x’ transition-
tion area. sensitive turbulence model is a better suited choice for high val-
Fig. 11 shows the predicted distribution of the pressure ues of the advance ratio (J) while the ‘k-e’ fully turbulent stan-
coefficient over the suction side of the propeller blade, based dard model is a better suited choice for low values of (J). This
on the ‘K-Kl-x’ transition model, at (r = 1.76, J = 0.71, is as expected since in this study, flow parameters were selected
n = 50 rps and H = 90°). It is clear that large discrepancies such that; a higher advance ratio means a lower rotational
exist between the numerically predicted and the experimentally speed hence, less possibility for full turbulence to occur. Gen-
obtained cavitation areas, at (cp = 1.76) [14,18]. In particu- erally speaking and when compared to experiments, the extents
lar, the predicted and experimental cavitation areas don’t exhi- of cavitation are over predicted -using both turbulence models-
bit the same extent, shape or location on the blade surface. and at all advance ratios.

Fig. 11 Cavitation area over the blade suction side at (J = 0.71, n = 50 rps, r = 1.76 and H = 90°): (a) Experimental photograph of
the cavitation area (b) Predictions based on the ‘k-kl-x’ transition model and defined by (cp = r = 1.76).

Please cite this article in press as: M.M. Helal et al., Numerical prediction of sheet cavitation on marine propellers using CFD simulation with transition-sensitive
turbulence model, Alexandria Eng. J. (2018), https://doi.org/10.1016/j.aej.2018.03.008
Numerical prediction of sheet cavitation on marine propellers 9

Exp. J = 0.71, = 1.763 k- . J = 0.71, = 1.763 K-Kl- . J = 0.71, = 1.763

Exp. J = 0.77, = 1.783 k- . J = 0.77, = 1.783 K-Kl- .. J = 0.77, = 1.783

Exp. J = 0.83, = 2.063 k- . J = 0.83, = 2.063 K-Kl- .. J = 0.83, = 2.063

Exp. J = 0.9, = 4.455 k- . J = 0.9, = 4.455 K-Kl- .. J = 0.9, = 4.455

Fig. 12 Comparison of cavitation patterns, over the suction side of the blade, between experiments and predictions. Results are based on
the ‘k-e’ and the ‘k-kl-x‘ turbulence models for the same operational conditions and at a = 0.5.

CFD results  Exp: results

Table 5 Relative error of the cavitation extent based on the R: Errorð%Þ ¼  100
Exp: results
‘k-e’ and ‘k-kl-x’ turbulence models.
J r Relative error (%)
10.2. Prediction of performance characteristics under cavitation
k-kl-x k-e conditions
0.9 4.455 +10 +22.5
0.83 2.063 +11.6 +20 In this section, thrust and torque are obtained from numerical
0.77 1.783 +17.5 +14.2 simulations, based on both turbulence models and, at various
0.71 1.763 +24.7 +9.4
rotational speeds of the propeller as shown earlier in Table 2.
Subsequently, the non-dimensional coefficients of performance
(KT and KQ) are calculated for each value of J.
Table 5 gives a detailed account of the relative error per- A comparison of the numerically estimated coefficients,
centage of the predicted extent of cavitation i.e. predicted area based on both turbulence models, against experimental mea-
of cavitation. This relative error may be calculated as: surements is carried out and plotted in Figs. 13 and 14 for

Please cite this article in press as: M.M. Helal et al., Numerical prediction of sheet cavitation on marine propellers using CFD simulation with transition-sensitive
turbulence model, Alexandria Eng. J. (2018), https://doi.org/10.1016/j.aej.2018.03.008
10 M.M. Helal et al.

0.28 reversed. This agrees well with conclusions drawn previously

Exp.data
0.26 from Fig. 12 and Table 5.
k-kl-ω
0.24 k-ε
0.22 11. Conclusions
KT 0.2
0.18 This paper proposes a complete and detailed CFD procedure
0.16 for simulating the viscous, three-dimensional cavitating flow
0.14 about marine propellers. Subsequently, the cavitation patterns,
extents and the performance characteristics are predicted. This
0.12
procedure is developed taking into account the possibility of
0.1
existence of laminar and transient flow regimes over the pro-
0.7 0.75 0.8 0.85 0.9 0.95
peller. The ‘K-Kl-x’ transition-sensitive turbulence model is
J
selected for this purpose. The fully turbulent standard ‘k-e’
Fig. 13 Numerical/Experimental results of the thrust coefficient model is also employed for comparison. The investigation is
(KT). carried out for the ‘INSEAN E779A’ propeller model using
‘ANSYS FLUENT 16’. The multiphase RANSE technique
based on Multiple Reference Frame approach and full cavita-
tion model is applied. The cavitation patterns and extents as
0.48
well as the performance characteristics are predicted with
0.46
k-kl- results being validated by comparison with experimental data
0.44
k- available in the literature.
0.42 For the propeller under consideration, Reynolds number
0.4 distribution against various blade radius ratios demonstrates
Q
0.38 the existence of laminar and transient flow regimes over the
0.36 blade with decreasing rotational speeds. In this paper, the
0.34 advance speed was varied -in a way- throughout all calcula-
0.32 tions to ensure an advance ratio that varies inversely with
0.3 the rotational speed.
0.7 0.75 0.8 0.85 0.9 0.95 One of the important issues investigated in this paper is
J evaluating the two commonly used criteria for defining the
cavitation area. For the propeller under investigation and by
Fig. 14 Numerical/Experimental results of the torque coefficient
comparison with available experiments, it is revealed that;
(KQ). the iso-surface of vapor volume fraction (a = const.) criterion
is much more reliable in defining the cavitation area than the
KT and KQ, respectively. Detailed values of the coefficients and iso-surface of (cp = r) criterion. Subsequently, all predic-
of the relative error percentage are also shown in Table 6. tions obtained in this paper are based on the former.
It may be concluded from Figs. 13 and 14 and Table 6 that; Predictions of cavitation patterns, based on both turbu-
the threshold for the occurrence of the effect of transient flow – lence models, are in good agreement with corresponding exper-
for this propeller and associated flow conditions- is at an iments. Both predictions and experiments exhibit similar
advance ratio of approx. J = 0.80. The agreement between cavitation areas on the blade surface in terms of location,
predicted propeller characteristics, based on both turbulence shape and extent. Moreover, predictions have shown that cav-
models, and experiments is remarkably good for both, the itation inception is more likely to occur over the back of the
KT and KQ coefficients. blade section (suction side) and close to the leading edge. This
While –for both turbulence models- values of KT are agrees well with what is documented in the literature. As for
slightly under-predicted at all advance ratios, values of KQ the suitability of the two selected turbulence models, conclu-
are slightly over-predicted; Table 6. At higher advance ratios sions drawn agree with what is expected that is; while the
(J), the ‘K-Kl-x’ transition-sensitive turbulence gives a slight ‘K-Kl-x’ transition-sensitive turbulence model is more suitable
advantage over the ‘k-e’ fully turbulent model when compared for low rotational speeds, the fully turbulent standard ‘k-e’
to experiments while at lower advance ratios, the advantage is model is more suitable for high ones.

Table 6 Numerical/Experimental results of the thrust (KT) and torque (KQ) coefficients.
J r R. error (%) for KT R. error (%) for KQ
k-kl-x k-e k-kl-x k-e
0.9 4.455 2.4 8.7 +1.5 +3.8
0.83 2.063 3.9 6.4 +2.2 +2.6
0.77 1.783 5.7 3.3 +2.8 +1.9
0.71 1.763 7.6 2.9 +3.2 +1.7

Please cite this article in press as: M.M. Helal et al., Numerical prediction of sheet cavitation on marine propellers using CFD simulation with transition-sensitive
turbulence model, Alexandria Eng. J. (2018), https://doi.org/10.1016/j.aej.2018.03.008
Numerical prediction of sheet cavitation on marine propellers 11

Obtained results of the cavitating performance characteris- Symposium on Cavitation (CAV2001); 2001 June; California,
tics (KT, KQ) of the propeller, based on both turbulence mod- USA, 2001.
els, are compared with corresponding experimental results. [10] M. Maksoud, Numerical and experimental study of cavitation
Agreement between predictions and experiments is quite behaviour of a propeller, STG Sprechtag Kavitation; Hamburg,
Germany, 2003.
remarkable for all characteristics. Once again, while the ‘K-
[11] T. Watanabe, Kawamura, Takekoshi, Maeda, S.H. Rhee,
Kl-x’ transition-sensitive turbulence model gives an advantage Simulation of steady and unsteady cavitation on a marine
over the ‘k-e’ fully turbulent model at low rotational speeds, propeller using a RANS – CFD Code, in: Fifth International
the advantage is reversed at high rotational speeds. Symposium on Cavitation (CAV 2003); 2003 November; Osaka,
Overall, the proposed CFD simulation procedure based on Japan, 2003.
the selected transition-sensitive turbulence model has proved [12] F. Salvatore, L. Greco, D. Calcagni, Computational analysis of
to be an efficient and reliable tool in predicting the cavitation marine propeller performance and cavitation by using an
patterns and extents as well as the performance characteristics inviscid-flow BEM model, in: Second International
–under cavitation- of the marine propeller under considera- Symposium on Marine Propulsors; 2011. September;
tion. This is particularly apparent at low Reynolds numbers Hamburg, Germany, 2011.
[13] Z. Zhu, S. Fang, Numerical investigation of cavitation
i.e. low rotational speeds which correspond to high advance
performance of ship propellers, J. Hydrodynam. 24 (3) (2012)
ratios in this case study. 347–353.
The CPU time/processor/run required for simulations, [14] F. Pereira, F. Di Felice, F. Salvatore, Measurement and
based on the transition sensitive turbulence model, is very close modeling of propeller cavitation in uniform inflow, J. Fluids
to that required for simulations based on the standard fully Eng. 126 (4) (2004) 671–679.
turbulent model. This concludes that the processing time [15] F. Pereira, A. Sequeira, Propeller-flow predictions using
may be eliminated as a factor when evaluating the feasibility turbulent vorticity-confinement, in: European Conference on
of which turbulence model to opt for. The sole factor in decid- Computational Fluid Dynamics (ECCOMAS CFD); Lisbon,
ing on the appropriate model should only be the possible exis- Portugal, 2010.
tence of laminar and transient flow regimes over sections of the [16] D.K. Walter, D. Cokljat, A three-equation eddy-viscosity model
for Reynolds-averaged Navier-Stokes simulations of transitional
blade.
flow, J. Fluids Eng. 130 (12) (2008) 121401.
[17] ANSYS Inc., ANSYS Fluent Theory guide, Release 15.0, 2013.
References [18] F. Salvatore, H. Streckwall, T. Terwisga, Propeller cavitation
modelling by CFD-results from the VIRTUE 2008, in:
[1] J.P. Franc, J.M. Michel, Fundamentals of Cavitation, Kluwer Proceedings of the 1st International Symposium on Cavitation
Academic Press, Dordrecht, 2004. (SMP ’09); 2009 June; Trondheim, Norway, 2009, 362–371.
[2] J. Blazek, Computational Fluid Dynamics: Principles and [19] A.K. Singhal, M.M. Athavale, H. Li, Y. Jiang, Mathematical
Applications, second ed., Elsevier, New York, 2005. basis and validation of the full cavitation model, J. Fluids Eng.
[3] H.K. Versteeg, W. Malalasekera, An Introduction to 124 (3) (2002) 617–624.
Computational Fluid Dynamics: the Finite Volume Method, [20] F. Pereira, F. Di Felice, F. Salvatore, Numerical investigation of
second ed., Pearson Education Limited, London, 2007. the cavitation pattern on a marine propeller: validation vs.
[4] F. Salvatore, C. Testa, S. Ianniello, F. Pereira, Theoretical experiments, in: Proceedings of the 23rd ITTC; Rome, Italy.
modeling of unsteady cavitation and induced noise (INSEAN, Group discussion A.2, accuracy of CFD predictions, 2002.
Italian Ship Model Basin, Rome, Italy), in: Sixth International [21] M. Morgut, E. Nobile, Comparison of hexa-structured and
Symposium on Cavitation (CAV2006); Wageningen, hybrid-unstructured meshing approaches for numerical
Netherlands, 2006. prediction of the flow around marine propellers, in: First
[5] J. Dang, Numerical Simulation of Unsteady Partial Cavity International Symposium on Marine Propulsors; 2009 June;
Flows [doctoral dissertation], Delft University, Delft, Trondheim, Norway, 2009.
Netherlands, 2001. [22] Guilmineau, Deng, Leroyer, et al. Influence of the turbulence
[6] G. Vaz, Modeling of Sheet Cavitation on Hydrofoils and closures for the wake prediction of a marine propeller. In:
Marine Propellers Using Boundary Element Methods [doctoral Fourth International Symposium on Marine Propulsors
dissertation], Universidad Técnica de Lisboa, Lisbon, Portugal, (smp’15), Texas, USA, June 2015.
2005. [23] Da-Qing, Validation of RANS Predictions of open water
[7] J.S. Uhlman, The surface singularity method applied to partially performance of highly skewed propeller with experiments, in:
cavitating hydrofoils, J. Ship Res. 31 (2) (1987) 107–124. Conference of Global Chinese Scholars on Hydrodynamics,
[8] C.S. Lee, Kim, J.T. Lee, A potential based panel method for the Shanghai, China, 2006.
analysis of two-dimensional super or partially cavitating [24] D.P. De Witt, F.P. Incropera, Fundamentals of Heat and Mass
hydrofoil, J. Ship Res. 36 (2) (1992). Transfer, third ed., Willy, New York, 1990.
[9] S.A. Kinnas, Young, Numerical modelling of super cavitating
and surface-piercing propeller flows, in: Fourth International

Please cite this article in press as: M.M. Helal et al., Numerical prediction of sheet cavitation on marine propellers using CFD simulation with transition-sensitive
turbulence model, Alexandria Eng. J. (2018), https://doi.org/10.1016/j.aej.2018.03.008