SPEEDAM 2010
International Symposium on Power Electronics,
Electrical Drives, Automation and Motion
Investigating the Air Flow Rate of Self-
Ventilated Traction Motors by Means of
Computational Fluid Dynamics
B. Streibl* and H. Neudorfer*
* Traktionssysteme Austria GmbH, Brown-Boveri-Straße 1, 2351 Wiener Neudorf, Austria
Abstract--A self-ventilated traction motor was method is the fact that it is rather time consuming when
investigated by means of Computational Fluid Dynamics compared with other design tools. This can partly be
(CFD) with special emphasis on the prediction of its cooling overcome by parametrically building up reasonable
air flow rate. This investigation was suggested by the fact simplified motor geometries which are then subjected to
that the motor had to be designed in a non usual way as far an automatic meshing procedure and the subsequent CFD
as air flow is concerned and the subsequent thermal design
of the motor had to be based on a well predicted flow rate.
calculation (see for instance [7]).
The cooling flow through the investigated motor is driven by In this paper we present the results and the conclusions
an axis-mounted radial-blade centrifugal fan. Two variants drawn from a CFD analysis which was done in order to
of the motor differing in its fan wheel were analyzed by predict the flow rate of a self-ventilated traction motor
means of CFD. One of the two fan variants employs a two- and to fix its ventilation concept. The design of the
fans-in-one-wheel configuration. The results together with investigated motor had to meet certain requirements
analytical fluid-mechanical considerations helped to judge arising from possible water exposures of the machine and
between these variants under the premise of maximum flow as a consequence the air flow path had to be altered
rate. compared to our motors of similar power. As no network
of flow resistances was available which captured this
Index Terms--Computational Fluid Dynamics, cooling air
flow, thermal management, traction motors
design and also no measurements on geometrically
similar machines were at hand we decided to analyze this
motor by means of CFD in order to predict its air flow
rate. The CFD calculations were based on two full scale
I. INTRODUCTION
models of the motor which differed in the axis-mounted
fan. The results regarding the flow rate of both fan
The thermal management of traction motors is of designs were supported by analytical fluid-mechanical
crucial importance in order to reach high power-volume considerations.
densities (see for instance [1]). For the thermal
calculation procedure to be efficient it is necessary to II. COMPUTATIONAL SETUP
have on the one hand calculation networks which
represent the thermal capacities of the machine as well as
the path along which heat is being transferred. Especially Industrial CFD packages like ANSYS® FLUENT or
in the case of traction motors simulating the heat flow in ANSYS® CFX numerically solve a set of equations
the time domain is mandatory as the motor design must consisting of the so called Reynolds Averaged Navier-
be based on prescribed operational cycles [2]. Stokes equations (RANS), the energy-, continuity- and
On the other hand networks of the flow resistances in state equation and at least one equation to model the
the machine are obligatory for a fast and proper effects of turbulence. The need for using a turbulence
prediction of the air flow rate of a self-ventilated model arises from the time- or ensemble-averaging of the
machine. Navier-Stokes equations leading to the above mentioned
In the development of these networks Computational RANS equations. This averaging is done in order to avoid
Fluid Dynamics (CFD) together with flow and heat the numerical resolution of a large range of scales. So
transfer measurements is a highly valuable tool as it called two-equation models are widely used which add
allows to conduct virtual experiments whereby the two additional transport equations for turbulence related
influence of various, predominantly geometrical quantities to the above set of equations. These transport
parameters on heat transfer as well as cooling flow can be equations cannot be derived from first principles but
investigated [3][4]. Without the use of numerical methods result from heuristic arguments concerning the nature of
costly experiments would be needed. Important research turbulence. Among the variants of these two-equation
work on the application of the method to the field of models the k- and the k- SST model are often
electrical machinery was done in [5]. An example for the employed.
application of the method to a salient pole synchronous For our calculations we used ANSYS® CFX and the k-
generator can be found in [6]. The major drawback of the SST model. The cooling air flow was assumed to be
incompressible which is justified by the value of pressure
978-1-4244-4987-3/10/$25.00 ©2010 IEEE
736
rise in a self-ventilated traction motor. For coping with and which is convected by the rotor air stream enters the
rotation a multiple-frame of reference approach was fan. Now it is well known from basic fan theory that pre-
chosen (the frozen-rotor model of CFX). Hereby the swirl which enters a fan lowers its theoretical total
computational region is divided into two types of pressure rise. This fact is mathematically expressed by
domains. In the first type of domain the above mentioned Euler's turbine equation:
set of equations is solved in a rotating coordinate frame
whereas in the second type a stationary coordinate frame pth = (u 2 c2u − u1c1u ) (1)
is used. At each computational node of the common
border of the domains the fluid mechanical quantities are
Hereby pth denotes the ideal (theoretical) total pressure
handed over. With this approach a mild rotor-stator
rise which equals the shaft power of the fan, the fluid
interaction can be handled with the lowest computational
density (which we assume to be nearly constant), u1 and
effort compared to other rotor-stator interaction models.
u2 the circumferential velocity of the fan at the radial
For details see [8].
position of the inlet-area (index 1) and outlet-area (index
The air flow rate of the machine was determined in a
2) of the fan. The velocity u2 is also denoted as tip-speed.
three-step procedure. In the first computational step a
c1u and c2u denote the components of the fluid velocity in
certain flow rate was prescribed in the CFD calculation
circumferential direction at the radial position of the inlet
and the resulting difference between the fixed static
and outlet area respectively.1
pressure level at the air-outlet and the floating total
pressure level at the air-inlet was monitored. In a second
step this was repeated with another value of the flow rate. Air outlet
As there is – within limits and also for our case - a Stator air stream
quadratic dependency of pressure on flow rate in a fan-
duct system one is able to determine in a third step the
flow rate from the two preceding CFD calculations by
Rotor air stream
quadratic interpolation. The (self-adjusting) flow rate of
the machine is characterized by zero difference between
static pressure at the air-outlet and total pressure at the Air inlet
air-inlet. Furthermore it follows from dimensional
analysis that the flow rate linearly depends on the
Single fan wheel
rotational speed of the machine. Therefore it is sufficient
to conduct the above mentioned CFD calculations at a
single rotational speed of the machine in order to fully
know the flow rate characteristic of the machine.
Fig. 1. CFD-results for the distribution of static pressure in an axial
III. VENTILATION CONCEPT cross section of the motor which is equipped with a single fan.
At an early stage of the design process the ventilation
Two-in-one fan wheel
concept of the motor which was to be simulated with
CFD was fixed in as far as the cooling air was supposed
to be driven by suction in a parallel manner through axial
channels in rotor and stator by an axis-mounted radial-
Stator channel
blade centrifugal fan (see Fig. 1). The number and
diameter of these channels was determined in accordance
with the calculation of the magnetic flux through rotor
Rotor channel
and stator as well as mechanical considerations. The
design of these cooling channels is therefore an example
for the multi-physical approach which is necessary in the End winding
construction and development of electrical machines.
With this basic concept fixed the following two design
options were analyzed in the further course of the
investigation. First one could think of discharging the air
streams of the two parallel paths through rotor and stator Fig. 2. CFD-results for the distribution of static pressure in an axial
by a single fan after reuniting them in the space between cross section of the motor which is equipped with a two-in-one fan.
end-winding, rotor front end and fan entrance (see Fig.
1). The second option which we investigated is to use a
two-in-one fan wheel with two separate chambers (see 1
Equation (1) in the above stated form is valid under the assumption
Fig. 2) in order to separately discharge the rotor- and the of constant velocities at the inlet- and outlet-area. CFD allows to
stator air stream. In the case of a single fan so called pre- calculate the integral from of (1) which is valid in general. For details
see [5].
swirl which is generated by the revolution of the rotor
737
It follows from (1) that a two-in-one fan which the flow rate in the machine would adjust itself to a lower
discharges the swirling rotor air stream through a radially level than the prescribed one. Its exact value was
inward chamber and the stator air stream through a determined by the procedure outlined in chapter II.
radially outward chamber should be advantageous in When employing the two-in-one fan variant the
terms of pressure rise. This presupposes that the air gap cooling flow through the machine is split up into two
between both air streams is kept as small as possible and distinct streams along two parallel paths – denoted as
that therefore the radially inward chamber of the fan stator path with fan #1 and rotor path with fan #2 in Fig.
reaches as near as possible to the rotor. In this case the 4 and Fig. 5 (see also Fig. 1 and Fig. 2). Both streams
radially outward fan chamber is kept free from pre-swirl. reunite at the outlet of the fan. It can be seen in Fig. 4 and
Furthermore the rotor and the radially inward chamber Fig. 5 that the two-in-one fan variant is advantageous as
together form a fluid mechanical system which can be far as the pressure balance is concerned. Fan #1 (in the
regarded as a fan and which utilizes the rotor in this stator path) exhibits an increased ideal pressure rise when
sense. compared with the single fan variant. As stated above this
It also follows from (1) that an increase of the fan can be attributed to the absence of pre-swirl and the
diameter and – and therefore a greater tip-speed u2 – higher tip-speed u2, which accounts according to our CFD
improves the ideal pressure rise of a fan. As the primary results for 61% and 39% of the increase in ideal pressure
aim of our investigation was to calculate and maximize rise respectively. As a negative consequence of the
the flow rate of the motor we decided to increase the absence of pre-swirl higher shock losses in the fan occur.
outer blade diameter of the two-in-one fan variant by Nevertheless the sum of all losses which include the
6.6% when compared with the single fan. Such an losses arising from the discharge through-flow of the
increase has to be traded off against the accompanying stator (stator loss) is exceeded by the ideal pressure rise
increase in sound-pressure level of the machine which is of the fan.
linked to the tip-speed u2.
900
IV. RESULTS 800
700
Total Pressure (Pa)
600 Exhaust loss
The results of the CFD calculations support the above
500
considerations. The comparison of both fan variants Fan #1 loss
400
(single and two-in-one) was based on calculations at one 300 Stator loss
single (prescribed) flow rate. In Fig. 3 to Fig. 5 the 200 Fan #1 ideal pressure rise
various pressure losses and pressure gains in the machine 100
are shown for both variants. One encounters the 0
following pressure losses in the single fan variant: Pressure Losses Pressure Gains
exhaust loss, fan loss and system loss. The exhaust loss Two-in-one fan variant - stator path
equals the total pressure of the cooling flow at the outlet
Fig. 4. Pressure losses and pressure gains in the stator path of the two-
of the fan. All total pressures are measured relative to the
in-one fan variant at a prescribed flow rate.
ambient static pressure, which does not influence the
results as long one can regard the flow in the motor as
900
incompressible. The fan losses cover friction losses and 800
shock losses in the fan. Leakage losses were not 700
Total Pressure (Pa)
accounted for in the simulations. 600 Rotor ideal pressure rise
500 Exhaust loss
900 400 Fan #2 loss
800 300 Rotor loss
700 200 Fan #2 ideal pressure rise
Total Pressure (Pa)
600 Exhaust loss 100
500 0
Fan loss
400 Pressure Losses Pressure Gains
300 System loss
Two-in-one fan variant - rotor path
200 Fan ideal pressure rise
100 Fig. 5. Pressure losses and pressure gains in the rotor path of the two-
0
in-one fan variant at a prescribed flow rate.
Pressure Losses Pressure Gains
Single fan variant
The pressure gains in the rotor path of the two-in-one
fan variant (see Fig. 5) consist of the ideal pressure rise of
Fig. 3. Pressure losses and pressure gains obtained from the CFD fan #2 as well as the ideal pressure rise of the rotor. As
simulation of the single fan variant at a prescribed flow rate.
can be seen in Fig. 5 this rotor pressure rise is exceeded
by the rotor losses and therefore the rotor acts as a flow
It can be seen in Fig. 3 that for the prescribed flow rate resistance. Nevertheless – and what must not be ignored –
in the simulation the pressure losses are not wholly there is a pressure contribution of the rotor. But in
balanced by the ideal pressure rise of the fan. Therefore contrary to the single fan the swirl generated by the rotor
738
is immediately taken over by the subsequent fan #2 [2] Kral, C.; Haumer, A.; Plainer, M.: Simulation of a thermal
thereby keeping fan #1 completely free from pre-swirl. It model of a surface cooled squirrel cage induction machine
is - besides the higher tip-speed u2 - this co-operation by means of the SimpleFlow-library Proc. of the 4th
between rotor and fan #2 which results in a positive International Modelica Conference, Hamburg, 2005
[3] Mugglestone, J.; Pickering, S.; Lampard, D.: Effect of
pressure balance along both paths for the prescribed flow geometric changes on the flow and heat transfer in the end
rate when compared to the single fan variant. region of a TEFC induction motor Ninth International
Conference on Electrical Machines and Drives, 1999, 40-
V. CONCLUSION 44
[4] Hettegger, M.; Streibl, B.; Biro, O., Neudorfer, H.:
Identifying the heat transfer on the end-windings of an
Based on analytical considerations and the here
electrical machine by measurements and simulations
presented CFD calculations one can state that it is ICEM International Conference on Electrical Machines,
advantageous to use a two-in-one fan in a self-ventilated Rome, 2010
traction motor with axial cooling channels through rotor [5] Farnleitner, E. Computational Fluid Dynamics Analysis for
and stator instead of a single fan. This conclusion is Rotating Electrical Machinery PhD thesis, University of
drawn under the premise of maximum flow rate in the Leoben, Austria, 1999
motor and furthermore presupposes that suction is the [6] Lang, H.; Kral, C.; Haumer, A.; Haigis, M.; Schulz, R.
principal mechanism to drive the cooling flow through Investigation of the Thermal Behavior of a Salient Pole
the motor. Mechanical aspects (moment of inertia, Synchronous Machine ICEM International Conference on
Electrical Machines, Crete, 2006
weight, stresses, etc.) of both fan variants as well as
[7] Streibl, B.; Schlemmer, E., Schönauer J., Ramsauer, F.:
commercial aspects were not taken into account in this The Implementation of Computational Fluid Dynamics and
investigation. Electromagnetic Finite Element Methods as Standard
Design Tools for Large Synchronous Generators ICEM
REFERENCES International Conference on Electrical Machines, Crete,
2006
[1] Neudorfer, H.: Thermische Untersuchung und Berechnung [8] ANSYS®, Release 11.0, Help System. ANSYS, Inc.
eines flüssigkeitsgekühlten Traktionsmotors mit
Getriebeölwellenkühlung, PhD thesis, Technical
University of Vienna, 1998
739
