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

net/publication/305234196

Unsteady flow analysis of a two-phase hydraulic coupling

Article · June 2016

DOI: 10.1063/1.4951790

CITATIONS READS

0 13

6 authors, including:

Nahmkeon Hur Minhee Kwak

Sogang University Sogang University
140 PUBLICATIONS   690 CITATIONS    2 PUBLICATIONS   0 CITATIONS   

SEE PROFILE SEE PROFILE

Mohammad Moshfeghi
Sogang University
21 PUBLICATIONS   81 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Blood pump design View project

All content following this page was uploaded by Mohammad Moshfeghi on 17 December 2018.

The user has requested enhancement of the downloaded file.

JID: CAF
ARTICLE IN PRESS [m5G;November 15, 2016;3:45]

Computers and Fluids 0 0 0 (2016) 1–9

Contents lists available at ScienceDirect

Computers and Fluids

journal homepage: www.elsevier.com/locate/compfluid

Flow and performance analyses of a partially-charged water retarder

Nahmkeon Hur a,b,∗, Mohammad Moshfeghi b, Wonju Lee a
a
Department of Mechanical Engineering, Sogang University, Seoul, Korea
b
Multi-Phenomena CFD Engineering Research Center, Seoul, Korea

a r t i c l e i n f o a b s t r a c t

Article history: Two-phase flow inside a water retarder is numerically studied to understand the water retarder per-
Received 7 April 2016 formance and momentum dissipation mechanism for partially and fully-charged conditions at different
Revised 30 September 2016
rotational speeds, and the effects of external cooling flow into the water retarder. In order to accurately
Accepted 15 October 2016
capture the flow interaction between the blades of different rotating speeds, the transient simulations are
Available online xxx
performed using unsteady turbulent flow assumption using a sliding mesh technique.
Keywords: The Realizable k-ε turbulence model and the Volume of Fluid (VOF) method are adopted to simulate
Water retarder two-phase motion and interface of air and water mixture inside the retarder. A momentum index analysis
Partially charged is proposed for investigation of flow momentum and its dissipation. Results of partially-charged cases
CFD demonstrate that as a general trend, air is accumulated at the center of the hydraulic system forming
Multi-phase flow a toroidal shape. At a low charged water volume flow, circulation inside the retarder and, hence, the
Volume of fluid (VOF)
retarding torque are trivial. Furthermore, it is observed that the retarding torque becomes very small for
the charged water values lower than a special value. Finally, the simulations of the fully-charged open
system with external cooling flow (through an inlet and an outlet) reveal that the retarding torque and
also flow circulation between wheels decrease with the increase in flowrate.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction light and heavy vehicles, marine, etc [1]. Therefore, in hydraulic
couplings the second wheel is also connected to a shaft and ro-
Hydraulic retarders are being industrially employed as primary tates as a result of the applied torque. Hence, the rotor and stator
or supplementary brake systems for rotating shafts deceleration. A are called primary and secondary wheels.
hydraulic retarder is basically composed of one rotating wheel (ro- Because of the fundamental similarities between the hydraulic
tor) and one stationary wheel (stator), as shown in Fig. 1(a,b). As couplings and retarders, both have been usually studied numeri-
it is illustrated, the wheels include a number of straight or curved cally and/or experimentally as one category. Ainley et al. [2] mea-
blades, which provide passages (channels) for a fluid circulation in- sured the flow velocity inside a torque converter filled with oil
side the system. The assembly may be fully or partially charged at a constant pump speed at different speed ratios. The outcomes
with oil or water. As the rotor rotates, it centrifugally accelerates revealed a noticeable area of separation inside the hydraulic cou-
the working fluid and shoots it out from the outer part of its chan- pling. Also, mass flowrates were found to decrease as the speed
nels. The flow then enters the stator and streams inside the chan- ratio increased. In addition, they compared the core-to-shell and
nels along casing curvature towards inner radii. Meanwhile, it ap- blade-to-blade torque distributions. In another set of experiments
plies external torque to the stator walls. Since the stator is fixed, it done by Flack et al. [3], the effects of pump speed at a fixed speed
also applies an opposite torque to the fluid, which acts as a resist- ratio and also different oil viscosities have been measured. The re-
ing torque and, in turn, decelerate the rotor. The circulating fluid sults showed that torque converter mass flowrate increased ap-
then leaves the secondary wheel at inner radii and re-enters to proximately linearly with increasing pump rotational speed at a
the primary wheel, and the cycle continues. During this process, fixed speed ratio. However, they reported that the flowrate was not
the mechanical energy of the circulating fluid is dissipated into directly proportional to pump rotational speed. In addition, their
heat. Almost similar working principle is valid for hydraulic cou- measurements showed that mass flowrate decreased as the oil vis-
plings. However, the hydraulic couplings are used for transmitting cosity decreases. Huitenga and Mitra [4] numerically studied the 3-
(converting) torque between separated shafts of power generators, D fluid flow in order to understand the time dependent effects and
to improve the startup behavior of a hydraulic coupling. They mod-
∗
ified the geometry of the primary wheel (pump) and investigated
Corresponding author.
the effects of modifications on the pump and secondary wheel
E-mail address: [email protected] (N. Hur).

http://dx.doi.org/10.1016/j.compfluid.2016.10.033
0045-7930/© 2016 Elsevier Ltd. All rights reserved.

Please cite this article as: N. Hur et al., Flow and performance analyses of a partially-charged water retarder, Computers and Fluids
(2016), http://dx.doi.org/10.1016/j.compfluid.2016.10.033
JID: CAF
ARTICLE IN PRESS [m5G;November 15, 2016;3:45]

2 N. Hur et al. / Computers and Fluids 000 (2016) 1–9

Fig. 1. Geometry and size of the baseline water retarder (a) stator and (b) rotor; (c) Convention for non-dimensional radial position.

(turbine) [5]. Habsieger and Flack [6] studied flow characteristics cosity of oil and water on torque transmission using two different
at the interface of the primary and secondary wheels of a scaled charged conditions. Their conclusion showed that the transmitted
model of a truck torque converter at seven different speed ra- torque decreases as viscosity of fluid increases. Hur et al. [13] nu-
tios. The laser velocimetry was adopted for their measurement and merically investigated flow physics and streamline patterns inside
flow velocity at the interfacial plane was investigated. The results a water drive at three speed ratios and three charged fluid condi-
showed that as the speed ratio increases, the high velocities move tions. They reported the negative effects of unwet surface effects
to the pressure-shell corner and then to both the core-suction and at cases at high speed ratios. Liu et al. [14] investigated internal
the pressure shell corners at the primary wheel exit. Also they ob- torque converters using RNG k-ε , DES and LES turbulence models
served that there exist a reverse flow at the turbine inlet at low and compared flow field with experimental data. Their conclusions
speed ratios. Schweitzer and Gandham [7] validated CFD results of revealed that although the results of DES and LES can precisely
a torque converter using k-ε turbulence model with log-law as- predict the flow details inside highly complex geometries; how-
sumption. In another research, Hampel et al. [8] investigated the ever, the RNG k-ε model are also reasonably accurate for the pre-
average fluid distribution in the transversal planes of a hydrody- dictions of velocity magnitudes and flow structures.
namic coupling through gamma tomography measurements at dif- As it can be seen in the literature survey, a majority of re-
ferent speed ratios and charged water volumes. The experimental search projects are limited to special operational conditions of hy-
investigations reconstructed phase distributions together with use- draulic drives. Hence, there is a serious need for a comprehensive
ful information about inter-relation of the coupling design and the study of hydraulic retarders in a wide range of operational con-
local transfer of mass and energy between the wheels. Da silva et ditions. Therefore, the present research aims to investigate three-
al. [9] employed a planar array sensor method and measured flow dimensional two-phase flow numerical analysis of a water retarder
distribution and two-phase flow patterns inside a water coupling with different values of charged water at various rotating speeds
at two different speed ratios for a partially-charged coupling. Their in order to provide a performance map as a fundamental stand-
tests showed that at the zero slip condition a steady flow was es- point for interpretation and analysis of flow details inside a water
tablished and the liquid level was parallel to the rotation axis and retarder.
the phase transition between air and water was very sharp. Luo et
al. [10] investigated the flow inside a partially-charged hydraulic
coupling using RNG k-ε turbulence model and Volume of Fluid
(VOF) model. The working fluid was oil (VOF = 80%) and the ro-
2. Physics and governing equations
tating speed was fixed at 30 0 0 rpm. Their conclusion showed for-
mation of an oil wrap around a toroidal-shape air accumulation
The flow field is solved using the incompressible fluid assump-
at the center. In another research the steady state flow of a fully-
tions for both water and air. The simulations are performed us-
charged water driver at different speed ratios using SST k-ω tur-
ing VOF method. Hence, the same governing equations of a sin-
bulence model has been investigated. The result showed similar
gle phase flow are applied to each cell in the domain, and the
flow field distribution, despite different rotational speeds. Also, it
equations are solved for an equivalent fluid whose physical prop-
was concluded that the pressure distribution on the torus section
erties are calculated as linear functions of the physical properties
became more radially distributed at higher speed ratios [11]. Jain
of its constituent phases and their volume fractions. In addition, all
and Tiwari [12] compared performance of hydraulic coupling for
phases share the same velocity and pressure fields. For each com-
different working fluids in order to investigate the effects of vis-
putational cell, the equations of density (ρ ) and dynamic viscosity

Please cite this article as: N. Hur et al., Flow and performance analyses of a partially-charged water retarder, Computers and Fluids
(2016), http://dx.doi.org/10.1016/j.compfluid.2016.10.033
JID: CAF
ARTICLE IN PRESS [m5G;November 15, 2016;3:45]

N. Hur et al. / Computers and Fluids 000 (2016) 1–9 3

Fig. 2. A sample of initial condition for transient simulations (charged volume ratio
of 75%).
Fig. 4. Mesh resolution used for water retarder simulations.

tion equations are considered as:

∂
(u ) = 0 (3)
∂ xi i
∂ ui ∂   ∂p
ρ + ρ u j u i − τi j = − + Si (4)
∂t ∂xj ∂ xi
where ui is the fluid velocity in ith direction, p pressure, τ ij stress
tensor and Si momentum source. In addition, the conservation
equation of the transport of volume fractions is:
d
    αi Dρi 
αi dV + αi (v − vg )da = sαi − dV (5)
dt V S V ρi Dt
where sα i is source of the ith phase and Dρ i /Dt the Lagrangian
derivative of the ith phase density.
Furthermore, the Realizable k-ε turbulence model is adopted to
simulate turbulent behavior of the fluid mixture inside the water
retarder. This model is a recent development of the standard k-ε
Fig. 3. Mesh density sensitivity analysis of hydraulic coupling and the selected turbulence model, which contains a new transport equation for the
mesh.
turbulent dissipation rate, ε . The term Cμ , which is an important
factor for turbulent viscosity (μt ) calculation and is assumed to be
constant (Cμ = 0.09) in the standard k-ε turbulence model is con-
(μ) can be shown as:
sidered as a function of mean flow in the Realizable k-ε turbulence
 model. Accordingly, the Realizable k-ε turbulence model is appro-
ρ= αi ρi (1)
priate for swirling flows and the results can be relied on [15]. The
i
transport equations of k and ε can be presented as:

μ=

αi μi (2) ∂ ∂    μt  ∂ k
ρ
( )k + ρ k u − μ + = Gk + Gb − ρε + Sk
i ∂t ∂xj j
σk ∂ x j
(6)
where i represents the ith phase number and α is the correspond-
ing volume fraction (α i=Vi/V ). The mass and momentum conserva-

Please cite this article as: N. Hur et al., Flow and performance analyses of a partially-charged water retarder, Computers and Fluids
(2016), http://dx.doi.org/10.1016/j.compfluid.2016.10.033
JID: CAF
ARTICLE IN PRESS [m5G;November 15, 2016;3:45]

4 N. Hur et al. / Computers and Fluids 000 (2016) 1–9

Fig. 5. Radial distribution of momentum triangle in a fully-charged water retarder at different rotational speeds.

∂ ∂    μ  ∂ε ε2 energy due to buoyancy; and Sk and Sε the user defined source
(ρε ) + ρε u j − μ + t = ρC1 Sε − ρC2 √ terms for k and ε , respectively.
∂t ∂xj σε ∂ x j k+ vε
ε
+ C1ε C3ε Gb + Sε (7)
k 3. Computational approach

where Gk is generation of turbulence kinetic energy due to the The rotor and stator wheels include 48 and 45 curved blades,
mean velocity gradients; Gb the generation of turbulence kinetic respectively. The blade angle of the baseline model, defined as

Please cite this article as: N. Hur et al., Flow and performance analyses of a partially-charged water retarder, Computers and Fluids
(2016), http://dx.doi.org/10.1016/j.compfluid.2016.10.033
JID: CAF
ARTICLE IN PRESS [m5G;November 15, 2016;3:45]

N. Hur et al. / Computers and Fluids 000 (2016) 1–9 5

the angle between the blade walls and the axis of rotation, is
45°. In addition, there exists a 1 mm gap between these wheels.
The geometry of the rotor and stator wheels of the baseline
model together with the non-dimensional radial position is shown
in Fig. 1.
The study is carried out using the commercial CFD software
STAR-CCM+ v9.02 [15] on the computer cluster of CFD-ERC of So-
gang University, Korea. In order to accurately model the rotation of
the rotor and the interaction between the rotor and stator blades,
a sliding grid technique is used, and the simulations are performed
in unsteady framework. Since we are interested in the quasi-steady
state operation of the hydraulic coupling, the initial condition is
not an important factor. However, for the faster convergence of the
solutions, the air is initially assumed to be trapped in the central
region of the coupling as a toroidal shape as shown in Fig. 2. All
other variables are set zero as an initial condition. The typical so-
lution time for the simulations vary drastically from 30 days for
charged-water volume of 5% to 2 days for fully-charged cases us-
ing 12 cores (Intel Xeon E5-2660 v2, 2.20 GHz CPUs).
Fig. 6. Retarding torque at different operational conditions.
Prior to the main investigation, a mesh density sensitivity anal-
ysis has been carried out using polyhedral cells with three prism
layers on the baseline model. The sensitivity test is conducted
at the charged water condition of α = 90% and rotation speed
of 20 0 0 rpm, and the value of the retarding torque is used for
the assessment of mesh density sensitivity analysis. As it can be
clearly seen in Fig. 3, the cases with number of cells higher than
860,0 0 0 cells can be considered as insensitive ones. Hence, the
mesh resolution associated with this case is chosen for all of later
simulations. The corresponding mesh resolution is also presented
in Fig. 4.

4. Results and discussions

The baseline water retarder is investigated from different as-

pects. First, momentum dissipation mechanism inside the baseline
model at the fully-charged condition at different rotational speeds
is investigated via a concept of a momentum index. Later, the two-
phase flow analyses are conducted with various amount of charged
water inside the water retarder at different rotor speeds. Finally,
Fig. 7. Effects of amount of charged-water on the retarding torque at 20 0 0 rpm.
effects of external mass flowrates, which is practically used for
cooling process, are investigated.
4.2. Effect of amount of charged water in a closed system
4.1. Effects of rotor speed at fully-charged condition
In this part, the retarder is simulated at different amounts of
Owing to importance of understanding fundamental phenom- charged-water from α = 5% to 100%, at different rotational speeds.
ena inside the water retarder and, also, in order to provide a back- Similar to the previous section, the system is assumed to be closed.
bone dataset for further comparison, the baseline model is inves- As it can be seen (Fig. 6), the CFD simulations of fully-charged sys-
tigated as a fully-charged closed system with no inlet and out- tem follow the same trend with the experimental data.
let flow. Here, mutual effects between circulating water and the However, the predicted values are slightly underestimated, es-
wheels are investigated through a concept of flow momentum in- pecially at the lower speeds. Also, the results clearly show that the
dex triangle using local velocity multiplied by density. An impor- retarding torque is directly related to the amount of water inside
tant point about this concept is simultaneous demonstration of the retarder. Also, the curves’ slopes increase with rotational speed.
flow direction and the value of flow momentum at each loca- It can be also deduced that all cases with different amount of
tion. Fig. 5 displays the radial distribution of the flow momen- charged-water generates almost similar amount of retarding torque
tum index at different rotational speeds. As it can be seen, the for the rotational speed slower than 500 rpm.
high-momentum flow regions are mostly at outer radii, and at the Furthermore, the effects of amount of charged water on the re-
center of the hydraulic coupling momentum is almost zero. Also, tarding torque at a specific operational condition of 20 0 0 RMP is
the graphs clearly show that momentum values are relatively very plotted in Fig. 7. From the figure, it can be easily predicted that
small 500 rpm. However, the values increase as the rotor rotates the retarding torque becomes trivial for the charged water values
faster. In addition, the vectors clearly show more vigorous con- lower than approximately α = 40%. In addition, the rate of change
tribution of axial momentum in the outer radii. From the perfor- of torque increases at the higher water volumes.
mance point of view, the dissipation of flow momentum in each To investigate the details of flow inside the hydraulic coupling,
case can be calculated based on the surface integral over the in- phase distribution of different cases are shown in Fig. 8. Evidently,
terfacial plane. In addition, this figures together with the velocity the air is accumulated in the center part of the hydraulic cou-
triangles can be used for blade optimization. pling wrapped by water forming a toroidal shape. As the amount

Please cite this article as: N. Hur et al., Flow and performance analyses of a partially-charged water retarder, Computers and Fluids
(2016), http://dx.doi.org/10.1016/j.compfluid.2016.10.033
JID: CAF
ARTICLE IN PRESS [m5G;November 15, 2016;3:45]

6 N. Hur et al. / Computers and Fluids 000 (2016) 1–9

Fig. 8. Phase distribution inside the water retarder for different charged-water con-
ditions. Fig. 9. Velocity vector plots inside the water retarder for different charged-water
conditions.

of water decreases, it lowers the amount of retarding torque at

the same rotational speed, as previously shown in Fig. 7. More-
over, the velocity vector plots are depicted in Fig. 9, in which it
can be clearly observed that the flow circulation for the cases with
smaller amount of charged water are considerably slower. This is,
in fact, the predominant reason for the reduction in torque values.
In addition, in order to show the distributions of the phases
and velocity magnitude, these quantities are shown in Figs 10 and
11 on a unwrapped cylindrical surface passing through the center
of each section (r/R = 0). The figures provide valuable information
about the differences between distribution patterns as a result of
different charged-water values. It can be clearly seen in Fig. 10 that
the water tends to circulate adjacent to the each wheels wall and,
hence, the inner space becomes empty. Consequently, with less
water inside the system, the fluid velocity decreases (Fig. 11). Also,
as it can be deduced from the Fig. 10, a major part of the retarder
walls do not have contribution in the case with 25% charged water
volume. However, as the water volume increases a larger area of
the blade surfaces become wet and, hence, the retarder operates
Fig. 10. Phase distribution inside the water retarder on an expanded cylindrical sur-
more effectively. face at r/R = 0.0 for different charged- water conditions.
Moreover, velocity components on the planar interface between
the wheels are plotted in Fig. 12(a–c). As a general trend in Fig.
12(a), the axial velocity of the flow decreases with the decrease in which the tangential velocities become zero. For each case, the
the amount of charged water. Also, the radial portion associated change in the direction of the tangential velocity is a direct re-
with the flow into the stator wheel (position of zero axial velocity) sult of stator blade local twist angle. Physically, this issue causes
is at slightly in the outer part of the wheel section (0 < r/R < 0.08). tangential flow recirculation near the interface of two wheels. Fi-
In addition, this radial position trivially moves outwards as the nally, the radial velocity curves (Fig. 12(c)) show that for all of dif-
amount of charged water decreases. Physically, it means that flow ferent charged conditions, flow at the interface moves outwards
at a major part of interface is the decelerated flow coming out of from an approximate radius of r/R =−0.6 towards the outer radius.
the stator wheel. Similar to the axial velocity, the tangential ve- Then it reaches a maximum velocity at a location of approximately
locity (Fig. 12(b)) of cases with the lower amount of water are r/R = 0.7, and decelerates at the outmost radial position. It is note-
smaller. It can be seen that, in spite of the axial velocity com- worthy to mention that from the point of view of hydraulic cou-
ponent, there exists a noticeable difference between the radii at pling performance, the axial and tangential velocity components

Please cite this article as: N. Hur et al., Flow and performance analyses of a partially-charged water retarder, Computers and Fluids
(2016), http://dx.doi.org/10.1016/j.compfluid.2016.10.033
JID: CAF
ARTICLE IN PRESS [m5G;November 15, 2016;3:45]

N. Hur et al. / Computers and Fluids 000 (2016) 1–9 7

Fig. 11. Velocity magnitude inside the water retarder on an expanded cylindrical
surface at r/R = 0.0 for different charged- water conditions.

have more important contribution to the retarding torque genera-

tion phenomenon.
In addition, in order to investigate the details of momentum
dissipation through water circulation in a partially-charged con-
dition, demonstrations of the momentum triangle concept (calcu-
lated as the multiplication of local velocity by local density) for
different charged conditions are plotted in Fig. 13. Since the flow
in a partially-charged case is a two-phase flow, the figures here in-
herently include the effects of local water volume of fraction; and,
hence, they clearly show simultaneous effects of two-phase flow
and the velocity on the momentum dissipation mechanism. From
the figures, as the amount of charged-water volume increases the
flow momentum also increases, which is in complete agreement
with the retarding torque values (Figs. 6 and 7). In addition, the
vectors clearly show that the outer radial positions are the regions
mainly responsible for high-momentum flow circulations. This is
in accordance with Fig. 13(a) in which the axial velocity values are
very small near the center.

4.3. Effects of external flowrate in an open system

In a real application, there exists an external flow circulation

in order to cool down the working fluid. Hence, for this part, it is
assumed that water enters into the fully-charged retarder rotating
at a fixed speed of 20 0 0 rpm at different flowrates from 0 to 500
liter per minute (Lpm). The inlet and outlet are located at the sta-
tor casing as shown in Fig 1(a). For these simulations, the results
of retarding torque as well as the amount of circulating flow inside
the water retarder (Fig. 14) show a decreasing trend as the external
flowrate increases. This external cooling flow causes approximately
the same rate of decrease in the torque transmission and also the
circulating flowrate between the wheels. This is in fact expected to
be the result of the disturbances which are imposed to the velocity
flow field inside the stator by the in/out flows.

5. Concluding remarks

Unsteady two-phase flow analysis of a water retarder is per-

formed using a sliding mesh technique and VOF method on an
appropriately insensitive polyhedral mesh. Different amounts of
Fig. 12. Velocity components on the interface between wheels at 20 0 0 rpm.
charged water, rotating speeds and external flowrates are investi-
gated. The retarding torque is investigated via a vectorial demon-
stration of a momentum index, calculated based on multiplication

Please cite this article as: N. Hur et al., Flow and performance analyses of a partially-charged water retarder, Computers and Fluids
(2016), http://dx.doi.org/10.1016/j.compfluid.2016.10.033
JID: CAF
ARTICLE IN PRESS [m5G;November 15, 2016;3:45]

8 N. Hur et al. / Computers and Fluids 000 (2016) 1–9

Fig. 13. Radial distribution of momentum triangle at different amount of charged water at 20 0 0 rpm.

of local velocity and local density. The momentum triangles obvi- which causes a considerable fall in the value of retarding torque.
ously reveal that a major portion of high-momentum flow streams This is in accordance with the reduction in axial flowrate at the in-
at the outer radii. For partially-charged cases, since the vector terfacial plane. Also, since in a real application, the retarder in con-
demonstrations inherently contain the effects of local density, they nected to and external cooling flow (open system), a fully-charged
provide valuable information about the flow direction and local case at a fixed rotational speed is simulated for different external
flow momentum, simultaneously. The results display that flow cir- cooling flowrates. The results demonstrate very similar decreasing
culation is very small at a low amount of charged water volume,

Please cite this article as: N. Hur et al., Flow and performance analyses of a partially-charged water retarder, Computers and Fluids
(2016), http://dx.doi.org/10.1016/j.compfluid.2016.10.033
JID: CAF
ARTICLE IN PRESS [m5G;November 15, 2016;3:45]

N. Hur et al. / Computers and Fluids 000 (2016) 1–9 9

[2] Ainley SB, Flack RD, Brun K, Rovello TJ. Laser velocimeter measurements in
the pump of an automotive torque converter part Ⅰ- Effect of speed ratio. Int J
Rotat Mach 20 0 0;6(3):167–80.
[3] Flack RD, Ainley SB, Brun K, Whitehead L. Laser velocimeter measurements in
the pump of an automotive torque converter Part Ⅱ- effect of pump speed and
oil viscosity. Int J Rotat Mach 20 0 0;6(3):181–90.
[4] Huitenga H, Mitra NK. Improving startup behavior of fluid couplings through
modification of runner geometry: Part Ⅰ- modification of runner geometry and
its effects on the operation characteristics. J Fluids Eng 20 0 0;122:683–8.
[5] Huitenga H, Mitra NK. Improving startup behavior of fluid couplings through
modification of runner geometry: Part Ⅱ- modification of runner geometry and
its effects on the operation characteristics. J Fluids Eng 20 0 0;122:689–93.
[6] Habsieger A, Flack RD. Flow characteristics at the pump–turbine interface of a
torque converter at extreme speed ratios. Int J Rotat Mach 2003;9:419–26.
[7] Schweitzer J, Gandham J. Computational fluid dynamics in torque converter:
validation and application. Int J Rotat Mach 2003;9:411–18.
[8] Hampel U, Hoppe D, Diele K-H, Fietz J, Holler H, Kernchen R, et al. Application
of gamma tomography to the measurement of fluid distributions in a hydro-
dynamic coupling. J Flow Meas Instrum 2005;16:85–90.
[9] Da Silva MJ, Lu Y, Suhenl T, Schleicher E, Thiele S, Kernchen R, et al. Au-
tonomous planer conductivity array sensor for fast liquid distribution imaging
in a fluid coupling. Sens Actuators A 2008;147:508–15.
[10] Luo L, Feng LH, Liu SH, Chen TG, Fan HG. Numerical simulation of the two-
Fig. 14. Retarding torque and flow circulation between the wheels in an open sys-
phase flows in a hydraulic coupling by solving VOF model. IOP Conf. Ser.:
tem at different external cooling flow rates.
Mater. Sci. Eng. 19–22 Sep., Beijing, China; 2013. available at: http://iopscience.
iop.org/article/10.1088/1757-899X/52/7/072022/pdf (accessed 11 March 2016).
[11] Luo L, Zuo ZG, Liu SH, Fan HG, Zhung WL. Numerical comparisons of the
trend for both circulating flowrate between two wheels and also performance of a hydraulic coupling with different pump rotational speeds.
the transmitted torque. IOP Conf. Ser.: Mater. Sci. Eng. 19–22 Sep., Beijing, China; 2013. available at:
http://iopscience.iop.org/article/10.1088/1757-899X/52/7/072023/pdf (accessed
11 March 2016).
Acknowledgments [12] Jain N, Tiwari A. Comparative study of fluid coupling for oil and water as work-
ing fluid. Int J Eng Res Dev 2013;8:56–61.
[13] Hur N, Kwak M, Lee WJ, Chang C-S, Kang N-W. Unsteady flow analysis of a
This work was supported by the Energy Efficiency & Resources two-phase hydraulic coupling. 13th international conference of numerical anal-
Core Technology Program of the Korea Institute of Energy Tech- ysis and applied mathematics. Rhodes, Greece, Sspt; 2015.
nology Evaluation and Planning (KETEP) granted financial resource [14] Liu C, Liu C, Ma W. RANS, detached Eddy simulation and large Eddy simulation
of internal Torque converters flows: a comparative study. Eng Appl Comput
from the Ministry of Trade, Industry & Energy, Republic of Korea Fluid Mech 2015;1:114–25.
(No. 20142020102630). [15] STAR-CCM+ Ver. 9.06.011 user, CD-adapco, London, UK, (2014).

References

[1] Nunney MJ. Light and heavy vehicle technology. Butterworth-Heinemann;

2007. p. 317.

Please cite this article as: N. Hur et al., Flow and performance analyses of a partially-charged water retarder, Computers and Fluids
(2016), http://dx.doi.org/10.1016/j.compfluid.2016.10.033
 Journal of Mechanical Science and Technology 00 (2010) 0000~0000
www.springerlink.com/content/1738-494x
submitted manuscript under review

Numerical flow analyses of a two-phase hydraulic coupling

N. Hur1,2*, M. Kwak1, M. Moshfeghi2, C.-S. Chang3 and N.-W. Kang3

1
Department of Mechanical Engineering, Sogang University, Seoul, 04107, Korea
2
Multi-phenomena CFD Engineering Research Enter, Sogang University, Seoul, 04107, Korea
3
VS Engineering, Seoul, 06159, Korea

(Manuscript Received 000 0, 2009; Revised 000 0, 2009; Accepted 000 0, 2009) -please leave blank

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Abstract

The purpose of this paper is to investigate flow characteristics in a hydraulic coupling at different charged water conditions and speed
ratios. Hence, simulations have been performed for three-dimensional two-phase flow by using the VOF method. The Realizable k-ε
turbulence model is adopted. In order to resolve the interaction of passing blades of the primary and secondary wheels, the simulations
are conducted in the unsteady framework using a sliding grid technique. The results show that the water-air distribution inside the wheel
is strongly dependent upon both amount of charged water and speed ratio. Generally, air is accumulated in the center of the wheel
forming a toroidal shape wrapped by the circulating water. The results also show that at high speed ratios, the solid-body-like rotation
causes unwet areas on the periphery of the wheels and, hence, considerably decreases the circulating flow rate and also the transmitted
torque. Furthermore, the momentum transfer has been investigated through the concept of a mass flux triangle base on the local velocity
multiplied by the local mixture density instead of the velocity triangle commonly used in a single phase turbomachine analysis. Also, the
mass fluxes along the radius of the coupling in the partially charged and fully charged cases are found to be completely different. It is
shown that the flow rate at the interfacial plane and also the transmitted torque are closely related and are strongly dependent upon both
the amount of charged water and speed ratio. Finally, a conceptual categorization together with two comprehensive maps is provided for
the torque transmission and also circulating flow rates. These two maps in turn exhibit valuable engineering information and can be
served as a basis for an optimal design of a hydraulic coupling.
Keywords: CFD; Hydraulic coupling; Two-phase flow; Volume of Fluid (VOF)
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

result of the torque exerted by the rotating fluid. On the other

1. Introduction
hand, in the retarders, the secondary wheel is often stationary,
Hydraulic couplings are being widely used for torque which is intended to slow down the driving shaft connected to
transmission between separated shafts in industrial the primary wheel by resisting against circulation fluid by
applications such as power generators, light and heavy dissipating mechanical energy to heat energy.
vehicles, trains, marine, etc. [1]. Torque transmission is With the increasing trend of implementations of hydraulic
controlled by changing the amount of working fluid inside the couplings within the past two decades, these machines have
coupling. A typical hydraulic coupling is composed of been investigated via both experiments and CFD simulations.
separately rotating primary (pump) and secondary (turbine) Huitenga and Mitra studied 3-D fluid flow analysis in order to
wheels as shown in Fig. 1. The hydraulic couplings are understand the time dependent effects and to improve the
usually categorized into two types, i.e. drive and retarder. The startup behavior of fluid coupling by modifying the geometry
former is used for to drive the secondary wheel with different of the driving part [2-3]. Ainley and his colleagues measured
rotational speeds, while, the latter is used to decelerate the the flow velocity inside a torque converter charged with oil at
primary wheel. In either category, the torque is transmitted different speed ratios (SR) at constant pump speed [4]. The
from the primary wheel to the secondary wheel by circulating mass flows were calculated from the velocity data and were
flow inside the chamber between the blades of these two found to decrease as the speed ratio increased. Habsieger and
wheels. In the case of a hydraulic drive, the rotation of Flack investigated flow characteristics at the interface of the
primary wheel causes the flow circulation inside the passages primary and secondary wheels of a torque converter at various
of the secondary wheel, and consequently, the secondary speed ratios by using the laser velocimetry [5]. The results
wheel rotates at a lower speed then the primary wheel as a showed that there exists a reverse flow at the turbine inlet at
†
This paper was recommended for publication in revised form by Associate Editor
000 000-please leave blank.
*
Corresponding author. Tel.: +82 2 705 8637, Fax.: +82 2 713 8637
E-mail address: [email protected]
© KSME & Springer 2017
0000 G. Bell et al. / Journal of Mechanical Science and Technology 00 (2010) 0000~0000

In addition to the experimental measurements, which are

usually costly, the computer-based simulations also have been
being broadly used. Schweitzer and Gandham successfully
conducted CFD simulations of a 3-D torque converter using
the k-ε turbulence model [10]. Fan et al. numerically analyzed
the temperature and flow fields inside the flow channels of a
high-power hydrodynamic coupling with oil inlet and outlet at
different rotational speeds using the standard k-ε turbulence
model [11]. Luo et al. investigated flow field inside a partially
filled (80%) hydraulic coupling using the RNG k-ε turbulence
model and the Volume of Fluid (VOF) method [12]. Their
conclusion showed formation of oil wrap around a toroidal
shape air accumulation at the center of coupling. In another
(a) research, Luo et al. studied steady state flow inside a fully
charged water driver at various speed ratios using the SST k-ω
turbulence model [13]. Their result show similar flow field
distribution, despite different rotational speeds. They also
concluded that the pressure distribution on the torus section
becomes more radially distributed at higher speed ratios. Hur
et al. numerically investigate water-air two-phase flow pattern
inside a hydraulic coupling in different speed ratios and
charged conditions using the standard k-ε turbulence model
and VOF method together with the sliding mesh techniques
[14,15]. They investigated phase distributions and also torque
ratios inside a coupling at different operating conditions. Liu
et al. studied internal torque converters using the Detached
Eddy Simulation, Large Eddy Simulation and RNG k-ε,
(b) (c)
turbulence models and compared the details of flow field with
the experimental data [16]. They conclude that although the
Fig. 1. (a) Schematic view of a hydraulic coupling; (b) Side view of a
simulated model; (c) Section plot of a simulated model showing results of DES and LES are very accurate for highly complex
convention for the non-dimensional radial position. geometries; however, the RNG k-ε models can also result in
reasonably accurate predictions of velocity magnitudes and
low speed ratios, which is an indication of flow leakage flow structures.
through the core part. In another set of experiments the effects Despite the great values of different experimental methods,
of pump speed at a fixed speed ratio and also different oil practically, they are very expensive and lack to provide a
viscosities have been measured [6]. Their results revealed that complete set of data, especially for the details of flow field
torque converter mass flows increased approximately linearly inside hydraulic coupling. In addition, a majority of previous
with increasing the pump rotational speed at a fixed speed numerical simulations have not been shown to be numerically
ratio. In addition, mass flow rates were seen to decrease due to insensitive, and hence valid. Also their simulations usually are
the oil viscosity decrease and the resulting increase of the confined to limited cases at specific operational conditions.
Reynolds number. Hampel et al. investigated average fluid Thus, there exists a real need for a comprehensive study of
distribution in transversal planes of a hydrodynamic coupling hydraulic couplings at a wide range of operational condition
through gamma tomography at different speed ratios and also using an insensitive mesh resolution. To this end, the present
water volume fractions [7]. They reconstructed fluid fraction research aims at 3-D numerical analyses of performance and
distributions together with a considerable amount of useful flow fields of a hydraulic coupling with different charged
information such as the local transfer of the mass and energy water conditions at various rotating speeds. The Realizable k-ε
between the bladed wheels. Da silva and colleagues measured turbulence model together with the VOF model is applied.
flow distribution and two-phase flow patterns inside a water First, a mesh sensitivity analysis is accomplished while using
coupling at two different speed ratios for an 80% charged a sliding grid technique. The results are presented for different
coupling, in order to illustrate the real condition inside the amount of charged conditions at various speed ratios. The
coupling [8]. Jain and Tiwari experimentally compared the phase distributions, the flow patterns and the transmitted
performances of a hydraulic coupling for different working torque are investigated. In addition, two comprehensive maps
fluids in order to investigate the effects of viscosity on the are provided for the torque transmission and also circulating
transmitted torque at two different amount of charged fluid [9]. flow rates.
 G. Bell et al. / Journal of Mechanical Science and Technology 23 (2009) 1261~1269 1263

2. Governing equations and introduces a well-established length scale distribution near

the solid walls. The Realizable k-ε turbulence equations are:
The VOF method is used to simulate effect of multi-phase
flow inside the hydraulic coupling. This method is known to
    t  k 
be suited for simulating flows of several immiscible fluids on  k     ku j       
numerical grids capable of resolving the interface between the t x j    k  x j  (7)
phases of the mixture. The VOF tracks the interface between  Gk  Gb    Sk
different phases using the equivalent fluid properties whose
physical properties are calculated as functions of the physical
       
      u j      t  
properties of its constituent phases and their volume fractions.
Therefore, the equations of density (ρ) and molecular viscosity t x j      x j  (8)
(μ) can be shown as:
2 
  C1S   C2  C1 C3 Gb  S
k  v k
    i i (1)
i
where Gk represents generation of turbulence kinetic energy
due to the mean velocity gradients; Gb the generation of the
  i i (2) turbulence kinetic energy due to buoyancy; and Sk and Sε the
i user defined source terms for the k and ε, respectively.
Since the blades of the primary and secondary wheels pass
where i represents the ith phase number and α is the related
each other at different rotational speeds, numerical analyses
volume fraction.
are conducted in unsteady situations with a timestep of 25
vi micro seconds. Also, a sliding grid technique, which precisely
i  (3) simulates the relative movement of the rotating parts, is used
V
for modeling the wheels rotations.
The conservation equation for transport of volume fractions The numerical transient simulations are conducted based on
is: SIMPLE algorithm with second-order upwind scheme for the
convection term treatment and first-order fully implicit
  D i 
i dV   i  v  v g   da    Si  i
d method for the time discretization.
dt V
 dV (4)
S V
i Dt 
3. Geometry and numerical simulation
where vg is grid velocity, Sαi represents source or sink of the ith
phase and the Dρi/Dt is the Lagrangian derivative of density of The maximum diameter of hydraulic coupling is
the ith phase. approximately 400 mm. The number of blades in the primary
Then for the incompressible flow, the mass conservation and secondary wheels is 48 and 45, respectively. The blades
equation can be written as: are assumed to be radially straight with no bending, and their
thickness is supposed to be 3 mm. The schematic view of
 three-dimensional geometry and the assembled model of the
 ui   0 (5) flow passages are shown in Fig.1.
xi
The numerical simulations are carried out using commercial
where ui is the velocity in ith direction. Also, the momentum
conservation equation is:

ui 
 
t x j
 u jui  ij    xp  Si (6)
i

where, p is pressure, τij the stress tensor and Si the momentum

source.
Moreover, the Realizable k-ε turbulence model is adopted
to simulate turbulent behavior, since this model is an accurate
version of the standard k-ε turbulence model for separated
flows, swirling and rotating flows and flows with large
streamline curvature. In addition, in order to deal with the y+
effect, the “two layer all y+ wall treatment” option is chosen
so that the flow inside the viscous sublayer can be modeled
appropriately. This option eliminates the need for high mesh Fig. 2. Sample of initial condition (for the 60% charged cases).
resolution near solid walls, improves the convergence rates
0000 G. Bell et al. / Journal of Mechanical Science and Technology 00 (2010) 0000~0000

(a) 2.5×106 (b) 6×106 (c) 10×106 (d) 18×106

cells cells cells cells

Fig. 3. Phase distribution inside a hydraulic coupling for different Fig. 5. Mesh resolution used in the simulations of a hydraulic
number of meshes for 60% charged case at a speed ratio of 0.5 coupling.

Fig. 4. Mesh sensitivity analysis of the present flow simulation of Fig. 6. Time evolution of transmitted torque for fully charged case at
a hydraulic coupling at 60% charged case at a speed ratio of 0.5. a speed ratio of 0.5

software STAR-CCM+ (v9.06, CD-Adapco, 2014) [17]. The requires huge computational resources as described above it
polyhedral mesh has been used for the computational domains, seems to be not practical to accurately resolve the interface
and the domains are assumed as closed systems without inlet and dispersed bubbles and/or droplets near the interface. In
and outlet flow. The walls of the domains and blades have order to see the effect of mesh sizes on the computational
“no-slip” boundary condition. In addition, for the partially results a mesh density sensitivity test has been conducted by
charged cases the air is assumed to be initially at the center of using a 60% charged condition at a speed ratio of 0.5. As
the domain in a toroidal shape as shown in Fig. 2 in order to shown in Fig. 3, with smaller meshes the interface becomes
achieve faster quasi steady state results. The simulations are sharper and additional dispersed bubbles and droplets appear
performed using an Intel Xeon CPU cluster. The converged in the flow field, while the flow structure in a global sense
results were obtained after 8 days of parallel computation per remains unchanged. The effect of mesh sizes on the
case using 40 cores. transmitted torque is shown in Fig. 4, where approximately
Generally, a sufficient number of computational cells is 12% of difference in transmitted torque is seen between the
needed to capture a sharp interface between the phases. Since cases with 2.5 and 18 million computational cells. Also the
the VOF moving mesh simulation adopted in the present study results clear ly demonstrate unchanged volumes for the cases
0000 G. Bell et al. / Journal of Mechanical Science and Technology 00 (2010) 0000~0000

(a) Radial velocity (b) Tangential velocity (c) Axial vecoity

Fig. 7. Velocity components at different speed ratios for the fully charged case.

with number of cells more than 6 million. From these results,

it is believed that the simulations using VOF method with an
adequate number of computational cells can give us a good
qualitative results and reasonably good quantitative results as
well. Throughout the paper, the case with 6 million cells as
shown in Fig. 5 is chosen for discussions of the detailed flow
field, while results with 2.5 million cells are used for
comparison of the performances of the hydraulic coupling.

4. Results and discussions

4.1 Performance analysis at fully charged condition
The hydraulic coupling is simulated first as a fully charged
case at different speed ratios. Since in these cases the system
is single-phase, the results can provide valuable information Fig. 8. Mass flux triangle based on the local mixture density ρ
about sectional distribution of momentum transfer between
the fluid and the hydraulic coupling. Here, the primary wheel for the instantaneous transmitted torque at a speed ratio of 0.5
rotates at a fixed speed of 1780 rpm while the secondary for a fully charged condition during two revolutions of the
wheel rotational speed varies for different cases due to the primary wheel, using 1200 data points for each revolution . As
simulation of different speed ratios from 0.1 to 0.9. shown in Fig 6, the average torque and RMS values are about
While the hydraulic coupling is rotating, relative motion of 5072 and 6.6 (approximately 0.13%).
the blades causes temporal fluctuations in the results of each Furthermore, the details distributions of velocity components
properties. Figure 6 displays a sample of these fluctuations at different speed ratios at the interface plane between the

(a) SR = 0.4 (b) SR = 0.6 (c) SR = 0.8

Fig. 9. Mass flux triangle for fully charged case at different speed ratios.
0000 G. Bell et al. / Journal of Mechanical Science and Technology 00 (2010) 0000~0000

(a) 60% charged case (b) 70% charged case

(c) 80% charged case (d) 90% charged case

Fig. 10. Phase distribution inside a hydraulic coupling with different amount of changed working fluid.

wheels are presented in Fig.7. The radial, tangential, and axial generally increase as the radius increase, although in some
velocity components are monitored and circumferentially sections small reductions are observed. Since the coupling is
averaged at each special radius and plotted with respect to fully charged, the tangential velocity components of the water
non-dimensional radial position (Fig.1(c)). It can be clearly directly correspond to the local tangential rotational speed.
seen that the radial velocities are not vigorous, as compared to Also, it can be observed that at the outer radii, all tangential
the tangential and axial velocities. In Fig. 7(a), components of velocities seem to overlap at certain value, which shows the
radial velocities show very abrupt decreases from the position dominancy of rotating effects of the blades of wheel at this
of r/R=-1.0 to r/R=-0.95, which is because of the different part. Moreover, the results of axial velocity (Fig. 7(c)) show
geometries of primary and secondary wheels at r/R=-1.0. =All that the directions of the velocities are in opposite directions in
values increase after this radial position, and, interestingly, at the inner and outer radii, which is physically equivalent to the
r/R~-0.85, all cases pass through point of zero radial velocity. flow circulation between two wheels. The cases with higher
Further travel towards the outer radii, the radial velocities speed ratios have smaller values, which means the lower mass
become positive until to the location of r/R=0.1. Finally, at the flow transfer between the wheels. It is noteworthy to mention
outmost radius all the cases reach approximately the same that torque transmission between the wheels is a compound
values. In addition, Fig. 7(b) displays that tangential velocities effect of the tangential and axial velocity components.
0000 G. Bell et al. / Journal of Mechanical Science and Technology 00 (2010) 0000~0000

(a) Radial velocity (b) Tangential velocity (c) Axial vecoity

Fig. 11. Velocity components at different speed ratios for the 60% charged case.

(a) SR = 0.4 (b) SR = 0.6 (c) SR = 0.8

Fig. 12. Mass flux triangle for 60% charged condition at different speed ratios.

In addition, in order to investigate the details of momentum velocity components and also radial component of momentum
transfer between the wheels and the circulating water, mass transfer between the fluid and the hydraulic coupling. Similar
fluxes are calculated via multiplication of the mixture velocity to the previous section, the primary wheel rotates at a fixed
and its density. Figure 8 is the convection for the mass flux speed of 1780 rpm and different speed ratios are simulated by
triangle. Also, the sectional vector demonstration of a mass changing the rotating speed of the secondary wheel.
flux concept via multiplication of the mixture velocity and its Figure 10 shows the phase distributions inside the hydraulic
density are presented in Fig.9 for different speed ratios. As the coupling for different amount of charged water (60%-90%).
vectors clearly show, the water at the outer radii transfers As shown, the speed ratio has a noticeable impact on phase
higher amount of momentum, than the center of each section. distribution. At SR=0.4 and SR=0.6, the air is accumulated in
Also, it can be seen that values of the momentum decrease by the center of the wheel forming a toroidal shape, and water is
increasing the speed ratios. Finally, since the energy transfer circulating around this air torus in all cases. Also it can be
inside a hydraulic coupling is directly related to this concept, seen that center of this air torus moves towards the secondary
these figures together with velocity plots can be practically wheel, as the charged water volume increases. However, there
used for blade angle optimization. are remarkable differences in the phase distributions when the
speed ratio is 0.8. For such a case, since the two wheels rotate
4.2 Performance analysis at partially charged condition at closer speeds, the effects of relative movement of the
wheels become weaker and solid-body-like rotation occur.
In addition to the fully charged conditions, the hydraulic
This, in turn, causes the fluid to only circulate in the outer
coupling is simulated for different partially charged conditions
portion of the coupling, and as it will be shown later it reduces
at various speed ratios by using two-phase flow simulations.
the water circulation and transmitted torque. In addition, as it
The results are analyzed in terms of the phase distribution,
is shown, at SR=0.8, the fluid accumulates in outer portion of
0000 G. Bell et al. / Journal of Mechanical Science and Technology 00 (2010) 0000~0000

SR 0.4 SR 0.6 SR 0.8

(a) 60% charged case

SR 0.4 SR 0.6 SR 0.8

(b) 80% charged case

SR 0.4 SR 0.6 SR 0.8

(c) 100% charged case

Fig. 13. Mass flux at different speed ratios and charged conditions.

wheels while the air is in the inner portion, except for charged plots of tangential velocity values reveal that the empty space
water volume 90%. This also causes some ineffective unwet inside the wheels trivially increases the tangential velocity
areas appear on the wheels blades. Overall, it can be predicted values at r/R=-1.0 and r/R=+1.0. Also, it can be seen that the
that the transmitted torque decreases as the amount of charged distinguish trend of the fully charged condition is disappeared,
water decrease. especially at the two ends of the radial positions.
Furthermore, in order to investigate the velocity field for a Moreover, the momentum transfer between the fluid and
partially charged case, the velocity components of the case the wheels are demonstrated in Fig. 12 for a partially charged
with water volume fraction of 60% are presented in Fig. 11. condition of 60% through the concept of mass flux. The local
As it can be observed, the trends and details are completely density is calculated using Eq. 1 and the vectors are calculated
different from the fully charged cases. The centrifugal force by multiplication of the local velocity and the local density.
together with the empty space inside the wheels causes the The figures clearly demonstrate the radial positions of the
maximum radial velocities to become higher than those in the hydraulic wheel through which the momentum is transferred
fully charged cases. Also, it can be seen that the radial and more effectively. In other words, these plots are measure for
axial velocities at r/R<0.0 for high speed ratios are very small, the distribution of energy transfer between the water and the
which is in consistence with the solid-body rotation. Finally, wheels. The results show that for the cases with SR=0.4
0000 G. Bell et al. / Journal of Mechanical Science and Technology 00 (2010) 0000~0000

(a) Fully charged (b) Partially charged (SR<< 1.0) (c) Partially charged (SR ≈ 1.0)

Fig. 14. Schematics of phase distribution with different simulating conditions.

Fig. 15. Circulating water flow rate at interface at different volume Fig. 16. Transmitted torque for different amounts of changed water and
fractions and different speed ratios (All results are from simulations speed ratios (All results are from simulations with 2.5 million
with 2.5 million computational cells). computational cells).

more energy is transfer at the outer radii as the radial position charged at high speed ratio. Schematics summaries of these
moves away from r/R=0.0. Very similar phenomenon happens patterns are shown in Fig. 14. This categorization, together
at SR=0.6. However, at SR=0.8, the situation is completely with the flow angles (presented in Figs. 9 and 12) are certainly
different. In a complete agreement with what has been seen in useful information needed for every hydraulic coupling design.
Fig. 6, the water is accumulated in the outer radii. Therefore, Moreover, since the torque is transmitted via flow passing
the whole circulation occurs in the upper half of the cross through the interface of the wheels, the circulating water flow
section, and one can see the considerable unwet area at the rate at the interface between primary and secondary wheels
periphery of the wheels. The vectors demonstration of SR=0.8 are investigated in Fig. 15. As it is demonstrated, generally,
also confirm the trivial energy transfer between the hydraulic the circulating flow rate decreases when the amount of
coupling and the water. In such cases, the flow circulation is charged water is reduces. Also, for each charged condition,
very small, and, hence the transmitted torque is trivial. It is there exists an abrupt fall in the amount of circulating flow.
also important to mention that almost no energy is transferred Overall, the plots show that for a case with a lower amount of
in the region of 0.0<r/R<0.5 in cases with SR=0.4 and 0.6 for charged water this phenomenon occurs at the lower speed
these partially charged conditions. ratios.
Figure 13 describes the axial mass flux at different speed Furthermore, the torque values, as shown in Fig. 16, show
ratios and charged working fluid volumes with phase that the wheel in fully charged condition has the highest
distributions. As it shows, the axial mass flux has highest torque transmitting ability. For such a case, the transmitted
value at outer radii, which corresponds to more fluid torque smoothly decreases. It can also be seen that when the
circulation inside the wheels. In fully charged cases at speed speed ratio is close to 1.0, the wheel rotation is not influential,
ratios of 0.4 and 0.6, flow circulates in whole part of wheels. and the transmitted torque is small. Also, at a constant speed
However, in speed ratio 0.8 and partially charged cases, flow ratio, the transmitted torque decreases when the amount of
circulates only upper part of the plane. charged water decreases. Similar to the circulating flow rate,
Based on the observations, the flow patterns inside the there is an abrupt decrease for the partially charged cases.
wheel show three different and distinguished types for fully Interestingly, the speed ratios of fall are very similar to those
charged, partially charged at low speed ratio and partially observed in Fig.15.
0000 G. Bell et al. / Journal of Mechanical Science and Technology 00 (2010) 0000~0000

Finally, by cross-referring Figs. 15-16 to Fig. 13 one can Nomenclature

easily conclude the crucial effects of speed ratio on the phase
ρi : Density of ith phase
distribution and existence of unwet surfaces, and, accordingly,
αi : Volume fraction of ith phase
on the torque transmission. For example, as it is observed at
μi : Molecular viscosity of ith phase
SR=0.8, all charged water values lower than 90% show that
vg : Grid velocity
the air is accumulated near to the hydraulic casing as a result
Sαi : Source or sink of the ith phase
of solid-body-like rotation. Hence, the water recirculation is
ui : Velocity component in ith direction
severely interfered, and, consequently, the torque is reduced.
p : Pressure
Also, for the cases which are trivially charged, the rate of flow
τij : stress tensor
circulation becomes low and, therefore, the transmitted torque
Gk : Turbulence kinetic energy generation due to
decreases.
the mean velocity gradient
Gb : Turbulence kinetic energy generation due to
5. Concluding Remarks
to buoyancy
This paper investigates flow characteristics of a hydraulic Sk : Source term for k
coupling at different charged water conditions and speed ratios. Sε : Source term for ε
Hence, simulations are performed in a 3-D domain. In order to
resolve the interaction of passing blades of the primary and References
secondary wheels, the simulations are conducted in transient
unsteady framework using a sliding grid technique. [1] M.J. Nunney, Light and heavy Vehicle Technology,
Results show that water-air distribution inside the wheel is Butterworth-Heinemann.,Oxford, UK, (2007) 317.
strongly dependent upon both amount of charged water and [2] H. Huitenga and N.K. Mitra, Improving startup behavior of
speed ratio. Generally, air is accumulated in center of the fluid couplings through modification of runner geometry:
wheel forming a toroidal shape wrapped by the circulating Part I- Modification of runner geometry and its effects on the
water. The results also show that at high speed ratios, the Operation Characteristics, J.Fluids Eng, 122 (2000) 683-688.
solid-body-like rotation causes unwet areas on the periphery [3] H. Huitenga and N.K. Mitra, Improving startup behavior of
of the wheels and, hence, considerably decreases the fluid couplings through modification of runner geometry:
circulating flow rate and also the transmitted torque. Part II- Modification of runner geometry and its effects on
Furthermore, the momentum transfer has been investigated the operation characteristics, J. Fluids Eng, 122 (2000) 689-
through the concept of mass flux based on the local velocity 693.
multiplied by the local density. The mass fluxes along the [4] S.B. Ainley, R.D Flack, K. Brun and T.J Rovello, Laser
radius of the coupling in the partially charged and fully Velocimeter Measurements in the Pump of an Automotive
charged cases are found to be completely different. Also, Torque Converter Part I- Effect of Speed Ratio,
details of flow angle for each radial section is calculated, International Journal of Rotating Machinery, 6 (3) (2000)
which is a critical parameter needed to be considered for an 167-180.
optimum blade design, especially the relative angle between [5] A. Habsieger and R.D Flack, Flow characteristics at the
primary and secondary wheels. Besides, it is shown that the pump–turbine interface of a torque converter at extreme
flow rate at the interfacial plane and also the transmitted speed ratios”, International Journal of Rotating Machinery,
torque are closely related and are strongly dependent upon 9 (2003) 419-426.
both the amount of charged water and speed ratio. Finally, two [6] R.D. Flack, S.B. Ainley, K Brun. and L. Whitehead, Laser
comprehensive maps are provided for the torque transmission velocimeter measurements in the pump of an automotive
and also circulating flow rates. These two maps, in turn, torque converter Part II- effect of pump speed and oil
exhibit valuable engineering information and can be served as viscosity, International Journal of Rotating Machinary, 6 (3)
a basis for an optimal design of a hydraulic coupling. (2000) 181-190.
[7] U. Hampel, D. Hoppe, K.-H. Diele, J. Fietz, H. Holler, R.
Kernchen, H.-M. Prasser and C. Zipper, Application of
Acknowledgment
gamma tomography to the measurement of fluid
This work was financially supported by the Energy distributions in a hydrodynamic coupling, Flow
Efficiency & Resources Core Technology Program of the Measurement and Instrumentation, 16 (2005) 85-90.
Korea Institute of Energy Technology Evaluation and [8] M.J. Da Silva, y. Lu, T. Suhenl, E. Schleicher, S. Thiele, R.
Planning (KETEP) through a grant from the Ministry of Trade, Kernchen, K.-H. Diele and U. Hample, Autonomous planer
Industry & Energy, Republic of Korea (No. 20142020102630). conductivity array sensor for fast liquid distribution imaging
Also, a part of this paper was an invited contribution for “The in a fluid coupling, Sensors and Actuators AL Physical, 147
10th International Symposium on Numerical Analysis of Fluid (2008), 508-515.
Flow and Heat Transfer” – Numerical Fluids 2015, Rhodes, [9] N. Jain and A. Tiwari, Comparative study of fluid coupling
Greece. for oil and water as working fluid, International Journal of
 G. Bell et al. / Journal of Mechanical Science and Technology 23 (2009) 1261~1269 1263

Engineering Research and Development, 8 (2013) 56-61. Nahmkeon Hur is a Professor in

[10] J. Schweitzer and J. Gandham, Computational fluid Department of Mechanical Engineering
dynamics in torque converter: validation and application, of Sogang University in Seoul, Korea
International Journal of Rotating Machinery, 9 (2003)411- and a director of Multi-phenomena
CFD Engineering Research Center
418.
(ERC) funded by National Research
[11] L. Fan, W. Ma and W. Cai, Thermal-hydraulic analysis of Foundation of Korea. He received his
hydrodynamic coupling, Advances in information Sciences PhD from Stevens Institute of Technology in 1988. His
and Service Sciences, 4 (2012) 393-399. research interests include multi-physics and multi-dynamics
[12] Y. Luo, Z.G. Zuo, H.G. Fan, and W.L. Zhung, Numerical CFD and their industrial applications.
simulation of the two-phase flows in a hydraulic coupling by
solving VOF model, IOP Conf. Ser.: Mater. Sci. Eng., Minhee Kwak received her B.S. and
Beijing, China, (2013). M.S. in Mechanical Engineering from
Sogang University, Seoul, Korea in
[13] Y. Luo, L.H. Feng, S.H. Liu, T.J Chen, and H.G. Fan,
2015 and 2017, respectively. Her study
Numerical comparisons of the performance of a hydraulic interests at CFD-ERC laboratory at
coupling with different pump rotational speeds, IOP Conf. Sogang University include multi-phase
Ser.: Mater. Sci. Eng., Beijing, China, (2013). CFD analysis and applications.
[14] N. Hur, W.J. Lee, C.-S. Chang, and N.-W. Kang, “Two-
Phase flow analysis of a hydraulic coupling”, ASME-JSME-
Mohammad Moshfeghi received his
KSME Joint Fluids Eng. Conf., Seoul, Korea, (2015).
B.S. and M.S. in Aerospace
[15] N. Hur, M. Moshfeghi and W.J. Lee, “Flow and Engineering from Tehran Polytechnic
performance analyses of partially-charged water retarder”, in 2002 and 2005. He also received his
Computer and fluids, http://dx.doi.org/10.1016/ j.compfluid. PhD about flow control over wind
2016.10.033. turbine blades from Xi’an Jiaotong
[16] C. Liu, C. Liu and W. Ma, RANS, detached eddy simulation University, Xi’an, China in 2013. His
field of interest include CFD
and large eddy simulation of internal torque converters
simulations of multi-phase flows, turbo-machineries and wind
flows: A comparative study, Engineering Applications of turbine aerodynamics.
computational Fluid Mechanics,1 (2015) 114-125.
[17] STAR-CCM+ Ver. 9.06.011 User Guide, CD-adapco, Choon-Soo Chang is a Managing
London, UK, (2014). Director of the VS ENG in Seoul,
Korea. He received his M.S. degree
in production engineering from
KAIST, Korea. As a principal
engineer, he is responsible for the
development of a variable speed
hydraulic coupling.
 Proceedings of the International Mechanical Engineering Congress & Exposition
2015 IMECE
November 13-19, 2015, Houston, Texas, USA

IMECE2015-53349

UNSTEADY TWO-PHASE FLOW ANALYSIS ON WATER RETARDER PERFORMANCE

USING SLIDING MESH TECHNIQUE AND VOF METHOD

Wonju Lee Nahmkeon Hur *

Department of Mechanical Engineering, Department of Mechanical Engineering,
Sogang University, Seoul, Korea Sogang University, Seoul, Korea
[email protected]

ABSTRACT retarder and slows down. There are numerous reports and
technical investigations about the mechanism of hydraulic
Hydraulic retarders are used as auxiliary brake system in torque converters and retarders. Song et al. [1] developed the
heavy vehicles and high speed trains. A hydraulic retarder is torque converter design optimization system and studied the
composed of two parts, a rotor and a stator. When the system is design parameter and also blade profile optimization. Fuente et
activated, the working fluid is injected into the wheel and al. [2] investigated the unsteady flow at the pump exit and
circulates between the rotor and stator vanes using the resisting turbine inlet has been analyzed through instantaneous flow
torque of the stator to slow down the vehicle. The purpose of fields. As Yamaguchi and Tanaka [3] expressed, the effect of
this research is to investigate a water retarder system and the fluid inertia is enormous for vehicle launch situations, and the
details of flow characteristics of the water, and to investigate mass flow rate inside the torque converter is decreased due to
the device performance as well. The water retarder is basically fluid inertia. Talukder and Huynh [4] simulated performance of
composed of a rotor and a stator. In the present research, the a torque converter employing a power-law fluid by using a
rotor rotating speed is fixed at 2000 rpm. Since the performance CFD software package. Adibi-Asl et al. [5] developed a
characteristic of the water retarder is dependent upon the water mathematical torque converter model and studied the effects of
volume ratio, different volume ratios have been investigated. In the model parameters on the torque converter performance
this paper water retarder simulations are carried out using CFD (efficiency and capacity factor). Schweitzer and Gandham [6]
using sliding mesh technique. To capture the unsteady effects, simulate torque converter with CFD based on validation study.
the cases have been solved as transient simulations using In the present research, a water retarder is simulated under
standard k-ε turbulence model. The simulations have been different operational condition at constant rotating speed, in
solved as two phase flow, water and air. The results are order to estimate the mechanical parameters, using VOF
compared for different water volume ratios. The result show technique by the means of computational fluid mechanics.
that the air particles are accumulated in the center of the wheels
forming a tube shape (doughnut shape) and water particles are
at the outside, wrapping the air particles. In addition, torque
values are sensitively dependent upon water volume fraction.

INTRODUCTION

Hydraulic retarders operating mechanism is similar with

hydraulic torque converters, which is composed of two wheels,
Figure 1. Hydraulic torque converter/ retarder assembly
one rotor and one stator, as shown in Fig 1. While the purpose
of a torque converter is to transport the force to operate any
other mechanical device, a water retarder transports the force to
2. PHYICS AND GOVERNING EQUATIONS
assist the braking system of a vehicle during the decelerations
of the vehicle. When the breaking system is activated, the water Similar to any other CFD simulation, the basic
is injected from the water tank and flows into the water conservation equations have to be satisfied for the two-phase
retarder. Therefore, the system faces a resisting force from

Copyright © 2015 by ASME

flow. For a single phase flow, the mass conservation equation gap between the rotor and stator in one mm and the blade angle
can be written by: is 60 degrees. The geometry is generated in ANSA software and

𝜕𝜌 𝜕
+ �𝜌𝑢𝑗 � = 0 (1)
𝜕𝑡 𝜕𝑥𝑖

The momentum conservation equation is:

𝜕𝜌𝑢𝑖 𝜕 𝜕𝑝
+ �𝜌𝑢𝑗 𝑢𝑗 − 𝜏𝑖𝑗 � = − + 𝑆𝑖 (2)
𝜕𝑡 𝜕𝑥𝑖 𝜕𝑥𝑖
where the quantity Si is the momentum of source.
Since the model contains water and air, the Volume of
Fluid (VOF) method has to be selected. It is suited to model
flows of several immiscible fluids and is capable of resolving
the interface between the phases of the mixture (free surface
modelling). The model assumes that all phases present in a
control volume share velocity, pressure, and temperature fields. Figure 2. Schematic of the model geometry
Hence, the same governing equations of the mass, momentum,
and energy transport of a single phase flow is applied to a VOF
problem. imported to the STAR CCM+ Ver. 9.04 for mesh generation and
In addition, the equations are solved for an equivalent processing.
fluid whose physical properties are calculated as functions of The model is meshed using polyhedral mesh. For the high
the physical properties of its constituent phases and their accuracy of result mesh sensitivity test has been done prior to
volume fractions (αi ). Therefore, the equations of density, the main simulations as show in Fig. 3.As it is clear the case
molecular viscosity and the specific heat transfer can be shown with 860,000 cells can be considered as an insensitive case.
as:

𝜌 = � 𝛼𝑖 𝜌𝑖
𝑖
(3)

𝜇 = � 𝛼𝑖 𝜇 𝑖
𝑖
(4)

𝐶𝑝 𝑖 𝜌𝑖
𝐶𝑝 = � 𝛼𝑖
𝜌 (5)
𝑖

In addition, the conservation equation of the transport of

volume fractions is:
Figure 3. Mesh sensitivity analysis

𝑑 𝛼𝑖 𝐷𝜌𝑖
� 𝛼𝑖 𝑑𝑉 + � 𝛼𝑖 (𝒗 − 𝒗g )𝑑𝒂 = �(𝑠𝛼𝑖 − )𝑑𝑉
𝑑𝑡 𝜌𝑖 𝐷𝑡 (6)
𝑉 𝑆 𝑉

Similar to the single phase, the 𝑠𝛼𝑖 is the source of the i th

phase.

3 GEOMETRY OF MODEL AND MESH

The schematic layout of the rotor-stator assembly of the

simulated water retarder is shown in Fig. 2. In the model, the
Figure 4. Rotor-stator assembly and mesh arrangement

Copyright © 2015 by ASME

 5% 10% 25% 50% 75% 90% 95% 100%

Figure 5. Cases with different volume of fraction values after one second

Owing to the importance of the simulation of flow near the

wall, three prism layers have been located in the vicinity of the
walls. Also, because flow is more complex near the wheel, a
higher mesh resolution is used around the wheel. The mesh
arrangement in the stator-rotor assembly is shown in Fig. 4.

4. RESULTS AND DISCUSSIONS

To simulate the retarder movement, unsteady simulations

have been carried out using sliding mesh technique. Realizable
k-ε turbulence model is selected and the VOF method is
assumed. Since cases with different water volume fractions
have been simulated, the flow inside the wheel is a mixture of
water and air. In order to set a proper initial condition, the
initial conditions are defined by using by toroidal coordinate Figure 6. Torque related to different water volume fraction
system.
During the water retarder operation, typically, the air
particles are accumulated in the center of the wheels, forming a 5. CONCLUDING REMARKS
toroidal (doughnut shape), and water particles are in the outside
wrapping the air particles. Figure 5 presents the result of the In this paper, a water retarder has been simulated by three
two-phase separation for different water volume fractions after dimensional CFD simulation using VOF method with different
the period of one second. water fraction values. The realizable k-ε turbulence model is
Furthermore, an important performance factor in water used and for the high accuracy moving sliding mesh technic has
retarders is the torque value. Hence, the dependency of the been adopted. The results show that during the operation of the
torque on the water fraction is plotted in in Fig. 6. As the results system the air particles are accumulated in the center forming a
show, the slope of change in the torque value increases toroidal doughnut-shape and water particles wrap the doughnut.
gradually from the water volume fraction of 5% to 75%. In addition, the dependency of the torque values upon the water
However, for the higher water volume fractions, the slope of volume fraction is investigated. The results demonstrate that as
the curve increases dramatically and increases approximately the slope of the increase of the torque dramatically increase for
400% . the volume ration of higher than 75%.

Copyright © 2015 by ASME

ACKNOWLEDGMENTS [3] Yamaguchi, T. and Tanaka, K., 2012, “Torque converter
transient characteristics prediction using computational
This work was supported by the National Research fluid dynamics”, IOP Conf.Series:Earth and Environmental
Foundation of Korea (NRF) grant funded by the Korea Science 15.
Government (MSIP): Multi-phenomena CFD Research Center [4] Talukder, S.A. and Huynh, B.P., 2011, “Effects of Number
(ERC) in Sogang University (No. 2009-0093128). of stator Blades on the Performance of a Torque Converter”,
Proceedings of the ASME 2011 International Mechanical
Engineering Congress and Exposition, Denver, Colorado,
REFERENCES USA, 11-17 November
[5] H. Adibi Asl, N. Lashgarian Azad, J. McPhee, 2011, “Math-
[1] Song, K. Y., Kim, K. S., Park, J. I., and Oh, J. S., 2008, based Modeling and Parametric Sensitivity Analysis of
“Development of the Integrated Process for Torque Torque Converter Performance Characteristics, SAE Paper”,
Converter Design and Anlysis”, SAE International. 2008- 2011-01-0732, SAE Word Congress, Detroit, USA, April
01-0785. 2011.
[2] Fuente, P De la., Stoff, H., Volgmann, W., and Wozniak, M., [6] Schweitzer, J. and Gandham, J., 2003, “Computational
2011, “Numerical Analysis into the Effects of the Unsteady Fluid Dynamics in Torque Converters:Validation and
Flow in an Automotive Hydrodynamic Torque Converter”, Application”, International Journal of Rotating Machinery,
Proceedings of the World Congress on Engineering 2011, 9: 411–418, 2003, ISSN: 1023-621X print DOI:
Vol. 3. 10.1080/10236210390241646.

Copyright © 2015 by ASME

View publication stats