Study of wake dynamics of a rim-driven thruster
based on numerical methods
Bao Liu, Maarten Vanierschot, and Frank Buysschaert
Abstract—The present work examines the wake dynamics of a
rim-driven thruster (RDT) with Computational Fluid Dynamics
(CFD). The unsteady Reynolds-averaged Navier-Stokes (URANS)
equations were solved in the commercial solver ANSYS Fluent
using an SST k − ω turbulence model. The application of both
the moving reference frame (MRF) and the sliding mesh (SM)
approach for handling the rotational movement of the propeller
were compared with each other. Validation and verification of
the numerical model was performed to ensure the numerical
accuracy. Two representative scenarios were considered, i.e. the
bollard condition (advance coefficient J=0) and a very light
loading condition (J=0.7), respectively. From the results it is
confirmed that compared to the SM method, the MRF method is
not suitable for resolving the unsteady flow details as it only gives
general averaged wake features. By evaluating the simulation
results of the SM approach, the instantaneous wake flow field
under both conditions is presented and analyzed, most notably
the helical vortex structure. It is observed from the results that
the tip, blade shed and hub vortices are present in the wake
flow field and are convected downstream in a highly non-linear
way. Moreover, the shear layer vortices shedding from the duct
displayed a strong interaction with the distorted tip vortices of
the propeller.
Index Terms—Wake dynamics; Rim-driven thruster; Compu-
tational fluid dynamics; Sliding mesh.
I. I NTRODUCTION
OMPLEX flow structures are often observed in marine
C propeller systems when flow passes through the pro-
peller. Vortical structures are generated on the blade surface
and are convected downstream in regular patterns. These
coherent structures can often maintain their structure and shape
for a certain period until instability occurs and they start
to break down in the far field. For the conventional open
propeller, there are three types of blade vortices including the
tip vortex, the thin vortex sheets and the root vortex. Since
there is always a central hub connected to all propeller blades,
a hub vortex also exists. These vortices form the wake vortical
system for a propeller, namely the helical vortex system.
During the evolution of the wake vortex system, vortices shed
from different parts of the propeller system often exhibit strong
self- and mutual induction, causing instabilities and break-
up. The instability mode is related to the helical pitch, the
vortex core size and the vortex strength. Since the dynamic
development of the wave vortex is highly responsible for the
oscillations and noise performance of a propeller system, it is
M. Vanierschot is with the Department of Mechanical Engineer-
ing, Group T Leuven Campus, KU Leuven, Leuven, Belgium e-mail:
maarten.vanierschot@kuleuven.be.
B. Liu and F. Buysschaert are with the KU Leuven.
Manuscript received December 10, 2021; revised December 20, 2021.
considered to be paramount to have a thorough understanding
of the flow mechanisms as they can often provide valuable
insights into improved hydrodynamic performance prediction.
The study of the wake evolution of a propeller can be
performed with experimental methods. [1], [2] Particle Image
Velocimetry (PIV) and Laser Doppler Velocimetry (LDV)
measurements are useful tools in extracting the distribution
of field variables, like velocity. However, in recent years,
numerical simulations are gaining more popularity as they can
provide even more detailed insights into the flow field with less
costs if used properly. [3], [4]
It is often observed that the wake vortices exhibit different
patterns based on propeller loading conditions and geometry
profiles. For open propellers, the development of wake vortical
structures is mainly affected by the mutual interactions of
vortex shedding from adjacent blades. However, for a ducted
propeller, the existence of the duct complicates the composi-
tion and evolutionary mechanism of these wake vortices. Tip
leakage vortices are formed due to the tip clearance between
the blade tip and the duct’s interior surface. At the trailing
edge of the duct, flow separates from the duct and interacts
with the propeller tip vortices, causing different wake evolution
patterns compared to an open propeller.
A rim-driven thruster, which integrates the propeller and
motor, also consists of a duct and propeller structures. [5]
However, unlike a ducted propeller, there is no tip clearance
and therefore no tip leakage vortices are present. The gap
formed by the rim and duct interior surface is liable to
affect the formation of propeller tip vortices. Current studies
on RDTs mainly focus on the structural design and general
hydrodynamic performance prediction. [6]–[9] Hence, it is
meaningful to conduct a study on the vortical structure com-
position and wake evolution for an RDT in order to better
predict its performance.
The objectives of this study are: 1) assess the abilities of the
moving reference frame (MRF) and Sliding Mesh (SM) tech-
niques for propeller rotation problems in transient simulations;
2) study of the wake vortex structures of a rim-driven thruster
and its influence on the hydrodynamic performance. The lay-
out of the study is organized as follows. Section II covers the
theoretic background applied in this numerical study. Section
III sets up the physical model and its treatment including the
mesh discretization and solver settings. Numerical results and
related discussion are presented in Section IV and conclusions
are summarized in Section V. mds
December 10, 2021
 component, which represents the rotational velocity of the
grid, is included to take into account the rotational effects.

Fig. 1: The geometry of the ducted propeller (left) and RDT

(right) used in this study

TABLE I: Discretization error estimation with GCI for J=0.1.

J=0.1
Mesh Total cells Refinement
density (million) ratio GCI(%) Fig. 2: Comparison of instantaneous axial vorticity predicted
Ktp /Kq
Ktp /Kq by MRF (top) and SM (bottom) for the ducted propeller
Coarse 5.02 - 0.257/0.0456 -
Medium 9.79 1.25 0.262/0.0465 4.32/4.39
Fine 19.6 1.26 0.263/0.0461 0.86/1.95

II. T HEORETICAL BACKGROUND

The governing equations for incompressible Newtonian
fluid flows are given in tensor notation as:
∂ui
= 0, (1)
∂xi
    
∂ui ∂ui ∂p ∂ ∂ui ∂uj
ρ + uj =− + µ + +
∂t ∂xj ∂xi ∂xj ∂xj ∂xi
∂  
−ρu′i u′j ,
∂xj
(2)
where ρ is the fluid density, ui (i, j = 1, 2, 3) is the mean
velocity component, t is the flow time, p is the pressure, µ Fig. 3: Comparison of instantaneous wake vortical structures
is the dynamic viscosity, and −ρu′i u′j is the Reynolds stress predicted by MRF (top) and SM (bottom) for the ducted
term. For incompressible Newtonian flows, the Reynolds stress propeller; Q=500 1/s
can be related to the mean strain rate and eddy viscosity as
follows,
 
∂ui ∂uj 2 III. P HYSICAL MODEL AND GRID GENERATION
− ρu′i u′j = µt + − ρkδij , (3)
∂xj ∂xi 3 The physical models used in this study are presented in
where µt is the turbulent viscosity and δij is the Kronecker Figure 1. The computational domain has a cylindrical shape
symbol. with a length of 14D and a radius of 5D, in which D represents
The PISO (Pressure-Implicit with Splitting of Operators) al- the diameter of the propeller. The velocity inlet is located 4D
gorithm is adopted for the pressure and velocity coupling. upstream of the propeller and the pressure outlet is 10D down-
Second-order upwind schemes are used for the discretization stream. The computational domain consists of a static zone and
of momentum and turbulence terms. To simulate the rotational a rotating zone containing the propeller, rim and hub.
movement of the propeller, the Moving Reference Frame In order to balance the computational time and mesh quality,
(MRF) and Sliding Mesh (SM) approaches are employed. the grid employed in this study is constructed in a hybrid
Both methods are apt for the analysis of situations involving form, in which the moving zone is made with an unstructured
domains that are rotating relatively to each other. In the mesh and the static zone with a structured mesh. Several
case of MRF, the governing equations for the flow in the sets of meshes with varied density are generated for a mesh
selected rotating zone are solved in a relative rotating frame sensitivity study to ensure the numerical accuracy. Detailed
by further considering two acceleration terms, i.e. the Coriolis information can be found in a previous study ( [6]). Table I
and Centripetal accelerations, while for the SM technique, the provides a brief view of the results between the different
rotating zone indeed rotates with respect to the static zone. meshes. The Grid Convergence Index (GCI) values, described
Thus there is no reference frame attached to the computational in [10], can reflect the numerical error introduced by the grids.
domain. In the governing equations, an additional velocity It is observed that the GCI of the fine mesh is lower than that
of the medium mesh, which demonstrates that the numerical
uncertainty becomes smaller when the mesh is refined. Based
on the results, the fine mesh is adopted for the transient
simulations.
IV. R ESULTS AND D ISCUSSION
A. Comparison of the MRF and SM approach
To assess the abilities of the MRF and SM approaches
in resolving the unsteady effects induced by the propeller
rotation, the wake field predicted by both methods for the
validation case, the ducted propeller, is compared in the
following discussion.
The instantaneous vortical structures in the wake field of
Fig. 4: Comparison of instantaneous axial vorticity in the wake
field of the RDT at J=0 (top) and J=0.7 (bottom)
the ducted propeller, in which the 3D vortical structures are
visualized using the Q-criterion [11], are shown in Figure 2
and Figure 3. Significant discrepancy can be observed for
the results calculated with the MRF and SM approaches.
The wake vortex system is better illustrated using the SM,
including the tip and root vortices. The trajectories of these
vortices shed from different blades are distinctively formed,
even though the entire helical vortex system is not fully defined
in which more information can be extracted such as break
up during evolution. This is due to the limitation of the
RANS calculations, which can be countered using LES-like
turbulence models. By comparison, the MRF gives even worse
results, in which the flow details are almost smoothed out and
useful insights can hardly be obtained.
B. Results for the RDT
The wake flow field of the RDT using the SM approach is
presented in this section. By setting the same range of vorticity
contours, the vorticity distribution patterns calculated for J=0
and J=0.7 are compared in Figure 4. Both conditions evidently
display a tip vortex region (0.8 < |y/R| < 1.1), a blade shed
vortex region (0.3 < |y/R| < 0.8), and a hub vortex region
(0 < |y/R| < 0.3), in which R represents the radius of the
Fig. 5: Comparison of instantaneous wake vortical structures
at J=0 (top) and J=0.7 (bottom); Q=250 1/s
Fig. 6: Vortical structures inside the duct at J=0 (top) and J=0.7
(bottom); Q=250 1/s
propeller. It is also observed that the vortex strength of J=0 is
significantly larger than that of J=0.7. At the tip vortex region,
a shear layer flow emerges when the viscous fluid passes
the surface of the duct, resulting in the shedding of several
vortical structures from the trailing edge of the duct. In the
tip region, vortical structures are deformed by the interaction
between solid surfaces such as blade and duct. These vortex
distributions are important for the production of the thrust
of the duct. The instantaneous flow field for two working
conditions is presented in Figure 5. The flow separation on
the exterior side of the duct is distinct at J=0.7.
Compared to the ducted propeller, the tip vortex in the RDT
wake exhibits a more ambiguous pattern. This is most likely
caused by the blunt trailing edge of the duct which alters
the distribution of the tip vortex energy. The root vortex is
not affected, and its shape resembles the one observed in a
ducted propeller. At a light loading condition, the tip vortex is
more regularly formed and diffuses gradually when convected
downstream. From Figure 6, it is clearly seen that the vortical
structures near the duct internal surface are more complicated
at J=0. Under heavy loading condition, the vortex deforms
because of a stronger interaction between the propeller and
duct surface and therefore the energy distribution is changed.
The circumferential flow around the duct is intensified and
thus propulsion efficiency is augmented.
Figure 7 provides an overview of the fluctuation of propeller
thrust for one revolution at two considered working condi-
tions. At J=0, the propeller thrust changes periodically within
 Fig. 7: Thrust variation on propeller during one revolution time at J=0 (left) and J=0.7 (right)
an almost constant amplitude. The curve is approximately
sinusoidal. Compared to the one at J=0.7, the variation is
much more substantial, which is probably caused by the more
intensified vortex shedding from the blade because of heavy
loading.
The comparison of propeller and duct thrust predicted by the
MRF and SM is presented in Table II. The thrusts predicted by
both methods are quite close, especially for the light loading
condition. This may suggest that the steady MRF can already
provide reasonable insights into hydrodynamic performance
prediction while the SM has the potential to further resolve
the flow details.
TABLE II: Comparison of forces prediction by MRF and SM
approaches.
J=0 J=0.7
Approaches
Ktp Ktd Ktp Ktd
Moving Reference Frame 0.263 0.115 0.104 -0.042
Sliding Mesh 0.271 0.118 0.103 -0.041
V. C ONCLUSION
The wake dynamics of a propeller is highly important for its
performance prediction as it provides valuable insights into the
formation and evolution of complex vortical structures in the
flow field. The present work investigates the wake topology of
an RDT and its influence on the hydrodynamic performance.
The Moving Reference Frame (MRF) and Sliding Mesh (SM)
techniques, which are often used in the simulation of rotating
machines, are employed and compared. It is confirmed that
the MRF is not suitable to study the transient wake dynamics
because unsteady effects are ignored, leaving little information
about the instantaneous wake structure. Therefore, to account
for the transient effects, the sliding mesh approach is nec-
essary. Therefore, for the simulations of the RDT, the SM
is adopted and two different propeller loading conditions are
considered. The numerical results reveal that the captured
wake topology exhibits significant difference to that of a
ducted propeller. The vortex distributions at different load-
ing conditions display considerable discrepancy. The vortical
structures at the heavy loading condition are obviously more
complex and stronger than at light loading condition. Due to
the existence of the duct, the development of the tip vortex
shedding from the propeller is interfered and distorted.
ACKNOWLEDGMENT
The authors would like to thank the China Scholarship
Council (CSC) for their financial support for the first author
(Grant No. 201806950010). The computational resources and
services used in this work were provided by the VSC (Flemish
Supercomputer Center), funded by the Research Foundation -
Flanders (FWO) and the Flemish Government – department
EWI.
R EFERENCES
[1] M. Felli, F. Di Felice, G. Guj and R. Camussi, Analysis of the propeller
wake evolution by pressure and velocity phase measurements. Experi-
ments in Fluids. 2006, 41, 441-451.
[2] M. Felli, R. Camussi and F. Di Felice, Mechanisms of evolution of the
propeller wake in the transition and far fields. Experiments in Fluids.
2011, 682, 5-53.
[3] J. Gong, C. Y. Guo, D. G. Zhao, T. C. Wu and K. W. Song, A comparative
DES study of wake vortex evolution for ducted and non-ducted propellers.
Ocean Engineering. 2018, 160, 78-93.
[4] Q. Zhang, and R. K. Jaiman, Numerical analysis on the wake dynamics
of a ducted propeller. Ocean Engineering. 2019, 171, 202-224.
[5] X. P. Yan, X. X Liang, W. Ouyang, Z. L Liu, B. Liu and J. F. Lan, A
review of progress and applications of ship shaft-less rim-driven thrusters.
Ocean Engineering. 2017, 144, 142-156.
[6] B. Liu and M. Vanierschot, Numerical study of the hydrodynamic
characteristics comparison between a ducted propeller and a rim-driven
thruster. Applied Sciences-Basel. 2021, 11, 4919.
[7] S. Gaggero, Numerical design of a Rim-driven thruster using a RANS-
based optimization approach. Applied Ocean Research. 2020, 94,
101941.
[8] B. W. Song, Y. J. Wang and W. L. Tian, Open water performance
comparison between hub-type and hubless rim driven thrusters based
on CFD method. Ocean Engineering. 2015, 103, 55-63.
[9] A. J. Dubas, N. W. Bressloff and S. M. Sharkh, Numerical modelling
of rotor–stator interaction in rim driven thrusters. Ocean Engineering.
2015, 106, 281-288.
[10] I. B. Celik, U. Ghia, P. J. Roache and C. J. Freitas, Procedure for
estimation and reporting of uncertainty due to discretization in CFD
applications. Journal of Fludids Engineering. 2008, 130.
[11] J. Jeong and F. Hussain, On the identification of a vortex. Journal of
Fluid Mechanics. 1995, 285, 69-94.