See discussions, stats, and author profiles for this publication at: [Link]

net/publication/280387221

Detailed CFD Modeling of Engine Cooling Fan Systems Airflow

Conference Paper · March 2003

DOI: 10.4271/2003-01-0615

CITATIONS READS

6 2,228

3 authors, including:

M. Henner Stéphane Moreau

Valeo Université de Sherbrooke
64 PUBLICATIONS 584 CITATIONS 494 PUBLICATIONS 7,990 CITATIONS

SEE PROFILE SEE PROFILE

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

SmartAnswer View project

UAV acoustics View project

All content following this page was uploaded by Stéphane Moreau on 26 July 2015.

The user has requested enhancement of the downloaded file.

 2003-01-0615

Detailed CFD Modeling of Engine Cooling Fan Systems

Airflow
M. Henner, A. Levasseur and S. Moreau
Valeo Motors and Actuators

Copyright © 2003 Society of Automotive Engineers, Inc.

ABSTRACT and the secondary flow moving around the motor.

Therefore to achieve the optimal cooling of a given
Requirement for more compact and more efficient fan electrical motor, fan systems designers nowadays need a
systems to improve the automotive thermal management better and a more detailed understanding of the flow
has urged Valeo to develop a complete automatic CFD phenomena under the fan hub and around the electrical
procedure to numerically simulate the fan system. The motor. This will also improve the fan system efficiency.
present study has focused on the description and the
validation of the flexible grid template that is meant to Since 1996 [2-5], Valeo has developed a complete fan
simulate the internal flow field within the fan system. This system design strategy using 3D Navier-Stokes
template closely fits in the previously developed rotor- computation fluid dynamics (CFD) intensively. It not only
stator parametric grid. The initial small rib reference helped Valeo improving its products but also reducing the
design has been meshed and four different large rib cases cost and the delay of the product development for a given
with increasing geometrical difficulties have helped testing client request. Today, CFD is used daily in our R&D group
this new capability. The simulations have been performed and has already lead to significant, constant and
on a medium size grid that has been shown to capture predictable improvement of the design of fan and stator
most of the important flow features. The ribs act as blades. It has also led to optimized cooling passages
centrifugal pumps that drain airflow through the electrical within its state-of-the-art compact electrical motors. The
motor. Larger ribs are shown to increase the flow rate overall CFD approach has already been presented in [3-5]
through the motor by up to 25% with a marginal increase and relies on accurate low diffusion simulations on
with swept ribs compared to radial ones. multiblock structured grids with CFX-TASCflow™ from
AEA Technology. As stressed in [5], the basic design
INTRODUCTION level has evolved from a basic parametric 3D in-duct
model where the fan blade is set in an ideal cylindrical
Automotive thermal management by the cooling system in duct to a complete model which can include the tip
tighter and hotter underhood environments requires more clearance, a recirculation area behind the hub and some
and more sophisticated fan systems, producing both a stator vanes behind the fan blade. In the present work,
high flow rate and a high pressure rise in a minimum this topology is further extended to include the detailed
space. Higher pressure rise often means higher torque hub structure with its ribs, the exact outer boundary of the
and higher electrical motor loading and internal heating. electrical motor and the inner ring of the support. Thus a
Moreover, better underhood airflow management suggests new specific grid template has been developed to study
using puller fan systems, which create a more stringent the fan system internal airflow. This new template yields
environment for the electrical motor. Efficient cooling of the various so-called “under-fan-hub”configurations. After
the electrical motor then becomes a necessity. This is going through the design background which involves both
even more critical with Valeo latest compact motor experimental and numerical investigations, the overall
technology, which brings the highest power density of model and grid specifications of the under-fan-hub airflow
electrically driven fan systems on the market. are presented. The parametric features are particularly
emphasized. The aim of this template is to quickly
In [1], the detailed CFD methodology put together to estimate the effect of various parameters on the airflow
improve the internal flow structure of the electrical motor going through the motors and the impact on the fan
was presented. This will improve the pressure-flow rate system efficiency. The first results with various large rib
characteristics of a given motor. Yet the exact position of designs are then achieved in the present investigation.
the actual operating conditions on this curve is a complex They are compared to the reference calculation with small
combination of the main flow going through the axial fan ribs and temperature measurements in climate cells
1
DESIGN AND EXPERIMENTAL BACKGROUND GLOBAL RIB DESIGN – The ribs within the fan hub act as
two centrifugal pumps in series. As a preliminary guideline
INITIAL CFD STUDIES – A preliminary CFD investigation the quasi-1D approach developed in [6] was applied. The
was first undertaken to understand the general features of flow within the rib passage is assumed uniform. Loss
the internal flow of engine cooling fan systems. The functions are accounted for to yield the final flow
simplified model did not account for the fan blades and characteristics of the centrifugal pump (figure 1). The
replaced the external flow through the blade by the operating point is found by the intersection with the
theoretical exit speed triangles. Similarly the airflow internal flow characteristics of the electrical motor. The
through the electrical motor was simplified to the main latter is obtained experimentally on the test rig shown in
internal airflow passage. The geometrical parameters were figure 2. From this initial design a range of inlet and outlet
the hub size, the external motor ring size and the rib flow angles were determined. These angles then provided
height. The rotational velocity was also varied. It was the necessary inputs for the CAD concept of the fan hub
found that the ring and the hub radii should be similar and with CATIA (figure 3). At this stage only circular mean
that a larger hub diameter and a larger rotational velocity lines had been assumed. Four different prototypes were
increased the airflow through the motor. The presence of then built from this CAD study: one has radial ribs and is
ribs was also found to be crucial for motor cooling. This termed “radial”, three have swept ribs and are called 160-
led us to a first overall topology to focus on for the next 90, 160-110 (shown in figure 4) and 160-130 respectively.
detailed study. Having investigated the overall
configuration it was time to consider the influence of the
rib geometry and the effect of the coupling between the
external flow through the blades and the internal flow
under the fan hub.

11TA45 Restriction Curve

0.14

y = 0.0048x2 + 0.002x + 0.0015

0.12 y = 0.0038x2 + 0.0031x + 0.0017

rise

0.1
11TA45Mot#1
11TA45Mot#2
Pressure
dp in inH2O

0.08 Polynomial (11TA45Mot#1)

Polynomial (11TA45Mot#2)

0.06

0.04

0.02

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
flow cfm
Flow rate
Figure 3. CATIA design of fan hub with swept ribs

Figure 1. Typical rib flow characteristics

Figure 4. Fan prototype 160-110 with swept ribs

Figure 2. Test rig for electrical motors flow characteristics

2
EXPERIMENTAL INVESTIGATIONS – The previously built
fan prototypes were then assembled with the same
support and several instrumented compact electrical
motors (shown in figure 5). Each resulting fan system was
then put on the same set of heat exchangers and tested
in a climate cell at an ambient temperature of 100°C.
Temperatures were measured with thermocouples at
several key locations within the electrical motor: the
brushes, the bushing on the back plate, the bearing on
the front plate and the diode on the plastic brush holder.
Strong temperature differences were found between the
reference small ribs fan and the above four prototypes: the
large ribs decreased the local temperatures by 5 to 10
degrees. Yet little temperature difference was found
between the radial and the curved ribs. The latter result
was a further incentive for a more detailed CFD analysis of
the internal flow under the fan hub.

Figure 6. Overall model topology

FAN SYSTEM MODEL– The fan is assumed symmetric

and the support is assumed to have a number of stator
vanes that are multiple of the rotor blades. The fan system
model then involves only one rotor blade section and
some multiple (here two) stator vane passages. The grid
template for the external airflow is the stage configuration
shown in figure 7 presented and validated in details in [5].
It comes from the extensive experience acquired on
several blade families that has guided the technical
choices for the grid generation, such as the overall
topology, the node distribution and density. The end result
is a high quality grid well suited to our very compact
turbomachine stage. In the present study, the stator
blades have not been considered to limit the final grid
Figure 5. Instrumented electrical motor
size.
NEW UNDER-FAN-HUB DEVELOPMENTS

OVERALL MODEL – The overall model consists of four

main components: the fan model which provides the
detailed external airflow; the electrical motor model which
retains only the necessary aerodynamic characteristics,
the under-fan-hub model which exhibits the detailed
internal airflow and the support model which accounts for
the structural blockage and possible flow leakage. The
overall topology is outlined in figure 6. To get high quality
and consistent results in the numerical simulations,
parametric grid templates have been emphasized for all
detailed models to ensure repeatability of grid generation
and consequently comparable quality level of simulations
on similar cases. These templates are written in the ACL
command language of TASCgrid the integrated mesh
generator of CFX-TASCflow from AEA Technology. The
grid of each part is built once for a generic case and can Figure 7. External fan system topology
be duplicated in a few minutes for all variations on the
same topology. The four parts are then assembled to yield Mesh Topology – The resulting grid topology has four
the final grid for the flow simulations. blocks wrapped around the blade, one small inlet block
3
and one large outlet block. The central blocks require the the symmetry and the 40° periodicity, the actual motor
ring, the hub and the motor profiles, and the blade airfoil cooling passages are modeled by an equivalent annular
sections at different radii. All grids used for fan simulation passage at the same mean radius on the motor case with
involve a hybrid H-O-H or H-C-H topology [8]. The O- or C- the same area as sketched in figure 10. The porous
grids are wrapped around the blade to ensure a proper medium is then applied to this annulus. The actual
resolution of the boundary layers around the profile. Four characteristics of this porous medium are set by an
other H-blocks are included in the domain limited by each external subroutine in the solver with a law that best fits
periodic face on the pressure and suction side of the the experimental measurements:
blades. Various additional blocks model the tip clearance ∆p = K ρ (v2/2) ∆L
and the gap between the hub limit and the motor back where K represents the internal resistance of the motor, v
plate. The resulting grid on the “solid” surfaces is shown in the mean speed going through the motor and ∆L the
figure 8. length of the motor case. This model was first validated on
a simple cube.

11TA45 Restriction Curve

0.14

y = 0.0048x2

0.12 y = 0.0038x2 + 0.0031x + 0.00

drop
0.1
11TA45Mot#1
11TA45Mot#2

dp in inH2O
Pressure
0.08 Polynomial (11TA
Polynomial (11TA

0.06

0.04

0.02

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
flow cfm
Flow rate

Figure 9. Compact electrical motor flow characteristics

Figure 8. Surface grid of the blade section

Boundary Conditions – The mass flow rate is set at the

inlet and a constant pressure (atmospheric conditions) is
imposed at the outlet. No slip condition is imposed in the
proper reference frame for all solid walls. At the periodic
boundaries, a periodic general grid interface (GGI) is
imposed to account for the high fan blade stagger angles
and to maintain the grid orthogonality at a reasonable Actual geometry Modeled geometry
level. Other frozen-rotor GGI are set at the interfaces
between the stationary and rotary parts.
Figure 10. Model of electrical motor flow passages
MOTOR MODEL– The internal structure of the electrical
HUB AND RIB MODEL– The parametric hub model is built
motor is not fully accounted for to keep the grid and model
in the same way as the fan system model. The number of
size reasonable. A typical grid size within the motor would
ribs is assumed to be a multiple of the rotor blades.
add about a million nodes and would require simulating
Presently only two rib passages have been considered
the full fan systems with its nine blades [1]. Yet a porous
following the results of a previous design of experiments
media models the effect of the internal structure with the
that had been achieved on this parameter to decrease the
experimental internal flow characteristics as inputs. Figure
motor temperatures with the same experimental set-up as
9 is the latter for the present compact electrical motors.
above. This hub topology is then meant to fit exactly in
This pressure drop is measured between the front and
the previous fan system topology: under the red part (fan
back plates of the motor for a given flow rate. The two
blade root) of figure 7.
experimental curves in figure 9 give an idea of the
dispersion due to measurements and process. To keep
4
Mesh Topology – The grid topology requires the rib, the mesh sizes in the order of 600,000 nodes. Similar grid
hub and the motor profiles. Similarly to the above fan studies as in [5] have been done. A fine grid would require
blade section, the hub parametric grids involve a simple H- at least a million nodes. Accounting for the internal
grid topology. The node distribution in the H-grids wrapped structure of the electrical motor would then require at least
around the ribs are fine enough to ensure a proper 10 millions nodes without the stator vanes!
resolution of the boundary layers around their profile. The
node distribution is also set to insure a smooth transition
with the fan system grid (figure 8). The resulting grid is
shown in figure 11. Again it should be stressed that a new
grid can be quickly created for any change of sweep,
stagger and profile of the ribs.

Boundary Conditions –No slip condition is imposed in the

proper reference frame for all solid walls. A one-to-one
periodic interface is imposed at the periodic boundaries. A
stationary frozen-rotor interface is set between the
electrical motor cooling passage and the hub grid. Finally
a GGI is set at the interface between the outlet of the hub
grid and the fan system grid.

Figure 12. Complete under-fan-hub topology

PHYSICAL MODEL – To account for the usual

incompressible, highly rotational and three-dimensional
environment of fan systems, the flow field has been
assumed to be fully turbulent and has been modeled by
the 3D RANS Navier-Stokes equations with a two-
equations k -ε turbulence model. The resulting set of
conservative equations has been solved with CFX-
TASCflow from AEA Technology on multi-block
structured grids using its Multiple-Frame-of-Reference
capability. Details of the numerical second-order schemes
used for this type of computation can be found in [3].

Figure 11. Internal hub topology and mesh At the solid boundaries wall laws have been applied to
reduce the number of nodes close to the walls. All
SUPPORT MODEL AND ASSEMBLY – The final part of computations have been performed at a rotational speed
the under-fan-hub model is the support. It involves the of 2500 rpm and a volumetric flow rate of 2500 m3/h.
central ring of the support with the actual clearance with
the motor case. It fits exactly in the previous fan system RESULTS
topology: under the blue part (stator feet) of figure 7. The
grid is a simple H-grid topology whose node density is Figures 13 and 14 show the induced flow topology in the
meant to match the node density of the fan system and hub and around the motor case for the large and straight
hub templates as closely as possible. No-slip condition is rib case. As already seen in the initial simplified CFD
imposed on the ring. One-to-one periodic boundary model, the ribs act as centrifugal pumps that drain air
conditions are imposed on the periodic faces. GGI from the back of the motor into the hub. Figure 13 shows
interfaces are set at the fluid boundaries with the other the streamlines colored by speed within the hub with the
templates. motor and the support removed from the plot for sake of
clarity. It emphasizes that the large ribs induce a
For the final assembly, all the above parts are located in significant flow field (large speeds) and therefore a
the same reference frame with their proper relative position significant mass flow rate into the electrical motor.
to yield the topology shown in figure 12. The Moreover significant momentum fluxes are injected in the
corresponding partial grids are attached to yield the final hub boundary layer at the blade root as conjectured in [8]
simulation grid. The same grid quality has been targeted and measured by LDV. A comparison with the initial small
for the different cases according to previous test rig ribs show similar patterns but with a much smaller flow
simulations with medium resolution [4-5]. This has led to rate. Figure 14 is a velocity vector plot in a meridional cut
5
through the electrical motor case, the fan hub and the Rib Radial 160-90 160-110 160-130
support ring between two ribs. It first points out that the Qmotor 12.35 12.93 12.7 12.21
inlet flow rate splits between a main flow rate that enters (m3/h)
the back of the motor and a secondary leakage flow Gain % base 4.70 2.83 -1.13
between the support ring and the motor case. The latter Table 1: Volumetric flow rates in the electrical motor
goes back all the way to the front plate of the motor. The
outlet flow is all concentrated between the ring and hub When comparing the mean flow velocity through the
end. Figure 14 also shows that several secondary flow motor, all four large rib designs lie around 5 m/s whereas
patterns happen in the hub: a large flow recirculation in the initial small rib design lie 25% lower at about 4 m/s.
the hub corner and another one in front of the support ring. The corresponding aerodynamic operating points of the
electrical motor are shown in figure 15. These results
closely correlate with the temperature measurements in
climate cells described above. The flow field in the portion
of the rib lying along the hub edge can explain the
marginal increase in flow rate observed with the swept
ribs. This second centrifugal pump currently operates off-
design in a quasi-stalled mode. On the contrary, the inlet
angle of the first centrifugal pump is improved over the
radial design yielding the marginal efficiency improvement.
Dp dans le moteur

120.00

100.00

80.00
Optimized ribs (+25%)
Dp (Pa)

60.00

Initial small ribs

40.00

Essais
Curve fit
Figure 13. Streamlines colored by speed with the support
20.00 Test data
valeurs essais
removed
Simulations
Obtenu calcu K98l

0.00
0 1 2 3 4 5 6 7 8 9
V (m/s)

Hub
Figure 15. Electrical motor operating conditions

The final investigation has been the influence of the rib

Q design and particularly the large rib case on the axial fan
Motor performances. Table 2 summarizes the results on the fan
pressure rise and torque. No significant difference is
Q exit found.

Rib Radial 160-90 160-110 160-130

∆Pfan (Pa) 216 215.7 215.8 215
Support Motor
Gain % base -0.14 -0.09 -0.46
Q support
Torque 1.3385 1.3379 1.3352 1.3348
(Nm)

Figure 14. Velocity vectors in a cut between the ribs Gain % base -0.04 -0.25 -0.28
Table 2: Fan pressure rise and torque
The next step has been to simulate and compare the
different rib designs. Similar flow patterns as figure 14 are CONCLUSION
obtained in all cases. Table 1 shows that quantitatively
the flow rates achieved in the motor with the four large The extensive use of CFD for fan design has been
designs are very similar with a marginal increase of 5% for extended to the full fan systems with both the external
the 160-90 configuration. flow field through the fan blades and the internal flow field
under the fan hub through the electrical motor. Previous

6
work had involved automatic grid templates for three kinds 4. “On the use of CFD in the Automotive Engine Cooling
of topologies, respectively the in-duct configuration, the Fan System Design,” E. Coggiola, B. Dessale, S.
virtual test rig and the rotor-stator configuration. The latter Moreau, and R. Broberg, AIAA 98-0772 paper, Reno,
included some new relevant local effects as the detailed January 1998.
tip clearance labyrinth and the electrical motor presence. 5. “Latest Improvements of CFD Models of Engine
Based on this upgraded topology, a new automatic grid Cooling Axial Fan Systems,” M. Henner, S. Kessaci
template for the under-fan-hub configuration has been and S. Moreau, SAE-02HX-36 paper, Detroit,
developed which exactly fits in the previous template. It is February 2002.
designed to cover a wide range of rib design with high 6. “Modélisation du transfert de chaleur au sein des
sweep and potentially various profiles and stacking. machines électriques tournantes. Dimensionnement
et Optimisation de leur système de refroidissement,”
Four different large rib designs with increasing sweep and C. Vasilescu, PhD dissertation, Université Paris VI,
therefore geometrical difficulty were used to test this new 2002.
grid capability. They involved the same fan used in 7. “Unsteady Rotor-Stator Interactions in Automotive
previous studies. No stator blades have been included to Engine Cooling Fan Systems,” M. Henner, M.
limit the model size. All four geometry was meshed Stanciu, S. Moreau, S. Aubert and P. Ferrand,
successfully with an equivalent grid quality of about Proceedings of the ISUAAAT 2000 Conference,
600000 nodes designed to capture the main relevant flow Ecully, September 2000.
features. 8. “Evaluating CFD Models of Axial Fans by
Comparisons with Phase-Averaged Experimental
A second step was to develop a simple realistic electrical Data,” D. Neal, S. Moreau, M. Henner and J. Foss,
model that bore most of its aerodynamic characteristics. VTMS5-01-89 paper, Nashville, May 2001.
A best fit of the available measured pressure drop was
then supplied to the flow solver as a porous medium
external subroutine. The flow simulations were then easy
and fast enough to provide a detailed investigation of the
four different rib shapes and to compare then to the initial
small rib design. Gains of about 25% in flow rates with
respect to the reference fan system were estimated for
each large rib case at a single operating condition. That
confirmed the trend observed in earlier temperature
measurements made in climate cells in key locations
within the electrical motor. The numerical comparison of
various fan system internal configurations is now effective
at an early stage of the product design process.

Future work will include more elaborate rib design that will
address the second pump stall issue. Additional flow
rates will also be run to assess the sensitivity of the
internal flow rates to the external flow rates. Finally results
from [8] will be revisited to study the impact of the ribs on
the wake topology of this nine-blade fan mounted in the
MSU test rig.

REFERENCES

1. “Thermal Analysis of Electric Motors in Engine

Cooling Fan Systems,” T. Hong, M. Rakotovao, M.
Henner, S. Moreau and J. Savage, SAE-2001-01-
10174 paper, Detroit, February 2001.
2. “Improvement of Fan Design using CFD,” S. Moreau
and E. Bennett, SAE-970934 paper, Detroit, February
1997.
3. “CFD based Design for Automotive Engine Cooling
Fan Systems,” E. Coggiola, B. Dessale, S. Moreau,
R. Broberg and F. Bakir, SAE-980427 paper, Detroit,
February 1998.

View publication stats