This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2017.2681975, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

Direct Liquid Cooling Method verified with

an Axial-Flux Permanent-Magnet Traction
Machine Prototype
P. Lindh, Member, IEEE, I. Petrov, J. Pyrhönen, Member, IEEE, M. Satrústegui, M. Martinez-
Iturralde, A. Jaatinen-Värri and A. Grönman

n Rotational speed
Abstract—Efficient cooling is needed e.g. in traction Ns Number of stator coil turns in series per phase
motors which face regularly high torque peaks and p Number of pole pairs
generate high stator Joule losses. This paper studies the q Slots per pole and phase
feasibility of the direct liquid cooling in the thermal Qs Stator slots
management of a low-power low-voltage permanent- Rs Phase resistance
magnet machine. A tooth-coil axial-flux (AF) permanent- stan Tangential stress
magnet (PM) double-stator-single-rotor test machine was T Torque
first equipped with indirect liquid cooling using water zQ Number conductors in the slot
cooling jackets and then with direct winding cooling. The
winding material used is a hybrid conductor comprising a I. INTRODUCTION
stainless steel coolant conduit tightly wrapped with
stranded Litz wire. The performance of the motor is
examined at various power levels using oil or water as the
cooling fluid. The results confirm that the proposed direct
H igh torque, high efficiency, compact size, and high
overload capability motors are required by original
equipment manufacturers, especially in the case of vehicle
cooling method is practical also in small machines, and traction [1-3]. The motor loss balance during acceleration of a
furthermore, it offers significant improvements in the vehicle shows high copper losses, which are dominating when
machine thermal management, especially, in cases where the current density is high during the periods of acceleration or
stator Joule losses dominate.
climbing a hill. In traditional cooling methods, the copper
Index Terms— electric machines, permanent magnet mass heat capacity is used to store the excess Joule losses
motors, rotating machines, temperature dependence, during high torque periods. In direct liquid cooling, the
liquid cooling. cooling performance is much more efficient. Lately, some
other interesting cooling methods have been introduced: for
instance, the use of a thermally conductive potting material
NOMENCLATURE [4], heat paths made of thermally well-conducting copper bars
a Number of stator parallel branches [5–7], or aluminium [6] have been suggested.
D se Stator external diameter In some cases axial-flux machines are also manufactured as
D si Stator inner diameter inner stator machines, which can be relatively complex and
h sy Stator yoke thickness employ indirect cooling [8]. In general, totally-enclosed
h PM Permanent magnet height
J Current density
double-stator-single-rotor axial-flux machines can have
kFe Stator lamination space factor difficulties in the indirect heat removal and, therefore, more
kw5 Working harmonic winding factor efficient methods are needed to manage their losses. Liquid-
lPM Permanent magnet length immersion has also been applied to intensify the cooling of
ltot Stack total radial length high-torque machines [9-11].
M Mass Test setups implementing separate cooling channels inside
mPM Magnet mass
m rb Rotor bridge mass
traction machine slots have also been reported [12–14].
Applying winding’s direct liquid cooling in a low-voltage
axial-flux motor was in the authors’ interest because earlier
Manuscript received February 2, 2017; revised Month 2, 2017;
accepted Month 2, 2017. prototypes have demonstrated difficulties, especially in the
P. M. Lindh, I. Petrov, A. Jaatinen-Värri, A. Grönman, and J. J. cooling of axial-flux machines in the power range interesting
Pyrhönen are with Lappeenranta University of Technology, LUT for heavy machinery traction applications. This direct cooling
School of Energy Systems, P.O. Box 53851, Lappeenranta, Finland.
(email:[email protected] or [email protected]). allows applying a much higher torque-producing tangential
M. Martinez-Iturralde and M. Satrústegui are with Electrical stress than any indirect cooling.
Machines and Electrical Circuits at Tecnun, the School of Engineering Instead of using water, certain low-viscosity oils can also
of the University of Navarra, P° de Manuel Lardizabal, 13. 20018
Donostia, San Sebastian, Spain. (email: [email protected]).
serve as coolant fluids. The benefit of oil is that it normally

0278-0046 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2017.2681975, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

does not form ions, and therefore, there is no need to take in [4, 7].
actions to maintain the low conductivity of the fluid. In the
A. Prototype Machine
case of water, impurities and ions are easily formed in the
direct cooling system, and therefore, there may be a need to A tooth-coil AFPM prototype machine with 12 stator slots,
monitor the conductivity of the fluid during operation. The ten poles, and 60–150 kW target power at 1500 min-1 was
problems of oil are its lower heat capacity (in the range of 2 constructed to gain experience of the electrical and, especially,
kJ/(kgK)) and much higher viscosity than with water. In the the thermal performance of the machine. The machine slots
case of oil immersion, the cooling oil works both as a coolant are wider than normally (slot width 40 mm), resulting in not
and as a lubricant for the gearbox [9, 15]. Direct liquid cooling very efficient traditional cooling of the stator winding.
has mainly been used in large synchronous machines [16–18]. However, direct liquid cooling is suitable for large slots,
However, in this study the authors aimed to verify that it can because it removes the heat all along the slot. The rotor is also
be successfully applied also in smaller-power machines. developed to work together with the open-slot tooth-coil
Another objective of this work is to compare the indirect and stator. With such a construction the rotor is burdened by high
direct liquid cooling in the same low-voltage machine and to amount of high amplitude harmonics and therefore the rotor
gain a better understanding of the direct liquid cooling conductivity should be very low to prevent eddy currents to
approach. The coolants studied with the same stator were flow in the materials. The rotor structure of the test machine is
polyalphaolefin oil (PAO) and distilled water. illustrated in Fig. 1.
In the direct cooling of this prototype machine, the extra
costs can be the costs of a cooling circuit if there is not
available any cooling in the application. In turn the cost of
cooling circuit depends on its complexity and required
characteristics. The production process of the studied winding
method can be automated, as well as the winding routine of
inserting (punching ready-made coils) the winding in the
stator slots. The volume of utilized copper is smaller than in
original machine. The performance of the direct cooling
method is much better due the temperatures are much lower
than with forced air cooling or indirect cooling of e.g. stator Figure 1. Rotor structure of the 12/10 axial-flux machine. From left to the
rotor centre: glass fibre rotor body, ten pole shoes made of laminated steel to
frame. Therefore the lifetime of new direct cooled motor avoid high permeance and current linkage harmonic losses in the magnets and
would be longer than the traditional ones. to collect flux to the stator teeth, ten segmented magnets.

The rotor body is made of electrically nonconductive

II. DESIGNS OF THE PROOF-OF -CONCEPT MACHINES
impregnated glass fibre in order to have a light-weight rotor
To verify the proposed cooling method, an, originally, and to minimize the iron losses in the rotor. The details of the
60 kW permanent-magnet, axial-flux, double-stator, single- rotor structure are given in [4]. The permanent magnets used
rotor PM machine with tooth-coil windings cooled indirectly in the rotor have a remanence of 1.15 T, a coercive field
with bearing shield water jackets was refitted with directly strength of 870 kA/m, and a relative permeability of 1.05 (at
water cooled tooth-coil windings. The tooth-coil stator with 120 °C). The main dimensions and parameters of the machine
totally open slots offered easy winding work with the directly are listed in Table I.
liquid cooled Litz material. Prefabricated coils could be used.
TABLE I
Tooth coil machines are widely used in different vehicles as DIMENSIONS AND P ARAMETERS OF THE 12-SLOT 10-POLE P ROOF-OF-
traction motors as they offer suitable synchronous inductance CONCEPT MACHINE
for field weakening [19, 20]. Parameter Value
The proof-of-concept windings were manufactured using, in Design tangential stress value stan [kPa] indirect
34/57
principle, standard motor Litz wire, with the exception that the cooling/direct cooling
centremost conductor bunch was replaced by a stainless steel Stack total radial length ltot [mm] 70
cooling-fluid tube. At the motor terminals, the copper Stator inner diameter Dsi [mm] 250
conductors and the cooling fluid conduits were separated from Stator external diameter Dse[mm] 390
Stator yoke thickness hsy [mm] 70
each other, and the cooling conduits were connected by
Number of stator slots in one stator Qs 12
electrically isolated tubes to a heat exchanger to transmit the Permanent magnet height hPM [mm] 2×15
motor losses to a secondary cooling circuit with no extra Average permanent magnet width hPM [mm] 70
purity requirements for the coolant. Appropriate Permanent magnet length l PM [mm] 70
instrumentation and a data acquisition system were used in the Winding type: fractional-slot concentrated non-
0.4
proof-of-concept system. overlapping, slots per phase and pole q = Qs/mp
As mentioned the electrical machine was originally built to Working harmonic winding factor kw5 0.933
evaluate an indirect liquid cooling approach with water jackets Number of pole pairs p 5
Winding connection Star
on both end shields. The previous results have been published

0278-0046 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2017.2681975, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

Prototype 1: In the first prototype machine, the stator 2) Cooling circuit

winding was constructed with regular round enamelled copper Fig. 4 illustrates the construction of one stator. There are
wires, and both stators had a liquid jacket on the end shield. twelve tooth coils in each stator. Two tooth coils adjacent to
This is called CASE A. Later on, for prototype 2, the original each other and belonging to the same phase are connected as
winding was removed and replaced with direct liquid cooling pairs in order to have fewer cooling liquid connections and an
windings, and the cooling jackets were removed from the end acceptably short path for the cooling circuit. The stator
shields, CASEs B and C. Stator 1 has indirect water cooling lamination was constructed from M400-50A electrical steel.
with water jackets covering a 0.1 m2 surface area per end
shield. For better heat removal capacity, also three copper
thermal conductors were embedded in the frame and the yoke
in each stator tooth and arranged in contact with the cooling
fluid at the other end. This method and measurement results
are exhaustively explained in [6]. Fig. 2 shows that there are
12 regular stranded copper conductors around each stator
tooth. The end windings are as short as possible because of the
tooth coils, shown in Fig. 4.
Slot wedge Figure 4. One stator of the proof-of-concept machine and its coils.
Teeth
There are two circuits in parallel as shown in Fig. 4: the
electrical circuit with Litz wires and the cooling circuit with
Slot heigth 48 mm

Copper stainless steel tubes for the cooling liquid circulation. The
Area
32 mm2
cooling loops contain a primary loop travelling through the
conductors and a heat-exchanger primary, and a secondary
loop operating as a heat sink. Direct liquid cooling needs a
Average Slot width 40 mm galvanic isolation of the electric circuit and the cooling circuit,
tooth width and therefore, the cooling fluid itself must be nonconducting.
44 mm
The easiest way to achieve this is to use suitable oil for the
Figure 2. Winding conductors inside one slot of prototype 1 (CASE A). 12 heat transfer on the primary side. The liquid-tightness of the
conductors around one tooth. The total cross-sectional area of each copper machine has been ensured by manufacturing the phase-group
conductor is 32 mm2. The net slot copper space factor under the slot key is coils of single 6 m lengths of the Litz wire. Therefore, all the
48%
cooling-conduit connections take place outside of the stator,
Prototype 2: and in this POC case, outside of the whole machine. Standard
1) Directly liquid cooled windings, CASEs B and C plumbing connections are used between the stainless steel
The original copper windings of the two-stator-one-rotor tubes and the nylon tubes providing the galvanic insulation. A
axial-flux permanent-magnet machine were removed and risk of this kind of a mechanical solution can be insufficient
replaced with a new directly liquid cooled winding made of durability of the liquid circuit connections, and thereby, leak
Litz wires, Fig. 3. There are 188 strands of 0.5 mm diameter of oil or water. However, during three-month measurements,
no leaking occurred. In traditional direct liquid cooling
resulting in an effective copper space factor of about 61% in
solutions of high-voltage high-power turbogenerators, the
the conductor. The inner and the outer diameter of the
whole machine interior is usually pressurized with hydrogen
stainless steel water tubes are 3 mm and 4 mm, respectively.
so that if there is a leak in the liquid cooling system, the leak
The net slot copper space factor under the slot key is 37%. A will take place inwards from the pressurized hydrogen towards
higher factor could have been reached with a more successful the cooling circuit. In a case of a small machine, such an
selection of the Litz conductor dimensions. arrangement is practically impossible, and therefore, the
Slot wedge cooling circuit connections must be reliable enough and
Teeth Litz Tube for the liquid suitably outside the active area of the machine. If there is a
leak, it can be noticed from the leaking fluid without severe
harm for the machine itself.
Slot heigth 48 mm

Copper Using oil instead of water somewhat brings pros and cons
Area
37 mm2 with it. Typically, oils have a clearly higher viscosity and
lower heat capacity than water. The oil viscosity varies more
as a function of temperature leading to higher friction and
Average Slot width 40 mm pumping losses. Also water cooling has its pros and cons. It
tooth width
44 mm
has a low viscosity and a high heat transfer capacity, around
4.2 kJ/(kgK). However, water has three main drawbacks.
Figure 3. Stranded and transposed winding conductors showing also the Firstly, water should be fairly ion free also in low-voltage
coolant conduits. There are eight conductors wound around each tooth. The applications, secondly, water may freeze, which necessitates
height of one conductor is approx. 7.6 mm and the width 8 mm. The cross-
sectional area of the conductor copper is 37 mm2. using some anti-freeze additive in the primary cooling circuit

0278-0046 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2017.2681975, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

and thirdly hot water causes erosion in tubing materials, achieve the same power is higher. This is, naturally, not an
especially at high flow speeds. In these measurements, pure ideal case, but the availability of the Litz wire did not allow
water was used. The conductivity of water was regularly exactly similar conditions. In the indirectly cooled machine,
measured to monitor a possible line-to line or earth fault the windings were constructed with 12 stranded turns around
current flow in the cooling water. For the water used in the each tooth, and the phase resistance Rs was 0.0265 Ω. The
study, the conductivity was approx. 135 µS/cm. Using this hybrid tooth-coil windings were constructed with only eight
water as such would result in an about 1.5 MW isolation Litz stranded turns around each tooth, and their phase
resistance over an 8 mm diameter, 1 m long nylon tube. This resistance was measured to be 0.0165 Ω. At this temperature,
value is low and therefore better deionized water was acquired the resistance of the stainless steel tube is about 0.5 Ω.
with which a conductivity value of 3 µS/cm was achieved, Therefore, the resistance of copper is about 0.017 Ω.
resulting in an about 70 MW isolation resistance per water The losses were investigated by 2D numerical analysis of
connection. In practice, both winding ends have isolation of the double-stator single-rotor axial-flux PMSM; the calculated
this kind, which is certainly more than sufficient for a low- efficiencies are plotted in Fig. 5. Only the stator winding Joule
voltage machine. The water conductivity was monitored losses and the iron losses are taken into account (no friction
weekly, and no significant changes were observed. losses or extra losses are included).
3) Losses
From the electrical point of view, also the risk of short-
circuits and eddy current losses in the cooling tubes was
investigated. The main electrical values of the machine are
shown in Table II, and the electromagnetic design results are
described in more detail in [21]. The ratio stator Joule losses is
4.2 and the ratio of total losses is 1.5 between Prototypes 1
and 2. These figures indicate how dramatically the Joule
losses’ evacuation is improved in the case of direct liquid
cooling compared to the indirect one in this case. However,
also the total losses can be evacuated efficiently in this case
with the stator winding direct cooling.
TABLE II
Figure 5. Efficiencies of the motor under study (with Litz wires and cooling
PARAMETERS AND COMPUTED 3D MODEL VALUES FOR THE M ACHINES
tubes) according to the FEA. The stator winding Joule losses and the iron
UNDER STUDY [21,22]
losses are taken into account (no friction losses or extra losses).
Indirect cooling Direct cooling
Case A Cases B and C Possible eddy currents in the cooling tubes were
Power, kW 60 100
Rated frequency [Hz] 125 125 investigated by modelling them as hollow solid conductors
Number conductors in the slot zQ 24 16 and applying a finite element analysis as shown in Fig 6. Only
Number of stator parallel branches
1 1 the cooling tubes closest to the air gap are influenced by eddy
a currents in the tube walls. The worst case (highest current
Number of stator coil turns per
phase Ns (both stators are included)
96 64 density at the rated load) is shown in Fig. 6.
Total winding length per coil pair
including 0.2 m connection lengths 9 6
at both ends [m]
Proportion of useful copper length
3.36/37.4 2.24/37.3
per coil pair [m/%]
Total end winding length per coil
5.64/62.6 3.76/62.6
pair [m/%]
Phase resistance, ohm 0.027 0.017
Rated line-to-line voltage [Vrms] 400 272
Induced phase voltage, [Vrms] 216 147
Rated torque, [Nm] 380 639
Rated current, [A] 104 260
PCu, [W] 814 3450
PFe,s, [W] 945 945
PFe,r + PFe,PM, [W] 480 480
Padd, W] 1200 1200
Add. mechanical losses, [W] 1600 1600 Figure 6. Current density in four stainless steel conductors located topmost in
Ptotal, [W] 5039 7675 a slot. There are 16 tubes in a slot in total. The eddy-current losses remain low
Synchronous inductance, per unit 0.24 0.4 in practice.
Stator resistance, per unit 0.012 0.028
The eddy current losses of the eight cooling tubes are
presented in Fig. 7, indicating that the amplitudes are
Applying a stator of the same size and volume with Litz reasonably low. The total eddy current losses in all 12 slots
wires and cooling tubes inside resulted in fewer winding turns and their 192 solid stainless steel tubes is 50 W. During the
than in CASE A applying round enamelled wires. Therefore, measurement there were no indications of excessive eddy-
the induced voltage is lower, and the current to be used to current losses in the tubes, either. There is also some torque-

0278-0046 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2017.2681975, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

producing current flowing in the stainless steel tube. However, In addition, a periodic behavior in the circumferential axis is
its resistance is so high compared with the Litz strands in assumed, where only one pole and one slot are modelled. For
parallel with the tube that only about 3% of the torque- the physical setup, fluids (air, oil or water) have been
producing current may flow in the tube. considered incompressible and have been modelled with the
properties given in table III, which also shows the
2
thermophysical properties of different solids constituting the
machine.
Edyy current losses (W)

TABLE III
THERMOPHYSICAL P ROPERTIES OF THE MACHINES STUDIED

1 Thermal
Density Spec. heat Viscosity
conductivity
(kg/m3) (J/(kgK)) (kg/(sm))
(W/(Km))
Water 998.2 4182 0.6 0.001003
Oil 884 1910 0.144 0.4862
Air 1.225 1006 0.0242 1.8E-5
0
Fiber glass 2580 1024 0.43 -
0 0.008
Time (s) Steel 8030 502.5 16.27 -
Copper 8978 381 387.6 -
Figure 7. Eddy-current-induced losses in the cooling tubes. Each line
represents the loss in one of the conductors as a function of time as the Lamination 7700 450 4.43/39/39 -
machine is rotating and the flux density in the slot is varying. Magnet 7475 578 9 -
Aluminium 2719 871 202.4 -
III. THERMAL ANALYSIS
Thermal analysis is performed through CFD simulations, The behaviour of fluids is obtained using Navier Stokes
using the commercial software Ansys Fluent v.16.2. This equations, which relate the speed and pressure of the fluids
study summarizes three different cooling CASEs: and combined with the initial and boundary conditions
constitute the mathematical model for the fluid flow. In
A. Normal stator windings. Indirect cooling with water
turbulent flows, as in this case in the air gap, it is a really
jackets covering a 0.1 m2 surface area per shield plus
difficult task to solve the Navier-Stokes equations, so an
three copper thermal conductors embedded in each stator
approach is used to get an average value. This approach is
tooth at one end and in contact with the cooling fluid at
called the Reynolds Averaged Navier Stokes (RANS).
the other end.
However, new unknown variables called Reynolds stresses
B. Direct liquid cooling (described in this paper) that uses
appear in the RANS model. To get these values, the standard
tooth-coil windings formed from a hybrid Litz cable
k-ε model (which is considered one of the most robust
combining an electrical conductor with a coaxial stainless
turbulence models) has been used. Furthermore, the turbulence
cooling conduit. Coolant polyalphaolefin oil.
model has been combined with standard wall functions to
C. Similar direct liquid cooling that uses distilled water as a
describe the turbulence near the walls. The interaction
coolant.
between the rotor (rotating at 1500 min-1) and the stationary
This section presents two 3D CFD models of the direct parts was modelled using a multiple reference frame available
cooling systems using polyalphaolefin-oil (case B) and in the software employed.
distilled water (case C). Therefore, the main characteristics of As boundary conditions, the parameters exposed in table IV
the CFD models are presented first and later the results have been employed. In this table, losses are considered for a
obtained are shown. First, the geometry is simplified in both load point of 380 Nm and 1500 min-1 corresponding to about
cases in order to reduce the number of elements in the model 60 kW output. The convergence criteria for these simulations
and minimize the computational time. Consequently, an axial is obtained when a reduction of the scaled residuals in three
symmetry is considered, modelling only one stator and half of orders of magnitude is reached and the temperatures through
the rotor. In this sense, Fig. 8 shows the cutaway view where
iterations in the coils and the magnets remained constant.
the symmetry plane is placed and also the assumed
periodicity. TABLE IV
BOUNDARY CONDITIONS STABLISHED FOR BOTH SIMULATIONS
CASE B CASE C
Rotor Fe losses (W) 666 666
Rotor magnet losses (W) 400 400
Stator Joule losses (W) 1500 1500
Stator Fe losses (W) 625 625
Rotational speed (rpm) 1500 1500
Ambient temperature of air (ºC) 24.3 24.3
Inlet temperature of the coolant (ºC) 31.5 18.6
Figure. 8. Place of symmetry and the simulated coolant tubes in the 3D model. Coolant flow rate (l/min) 4.6 5.9

0278-0046 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2017.2681975, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

Finally, table V shows the mean temperatures of the active IV. MEASUREMENT ANALYSIS
parts for both cases. The results really reveal the effectiveness The previous results of the axial-flux machine
of the direct stator cooling. This machine, unfortunately measurements with indirect cooling are repeated here for the
suffers from unintentional extra losses in the rotor which
readers’ convenience. The three scenarios A, B, and C defined
limits the performance of the machine via hot PM
in Section III are compared. A single load point was selected
temperatures. However, the cool stator in case of water
cooling also helps keeping the rotor magnets at acceptable for comparison of the cooling methods A, B, and C. Each
temperatures. machine was warmed until the temperatures stabilized. The
output–input powers were measured for the efficiency
TABLE V analysis. The liquid flow speed, pressures, and several
MEAN TEMPERATURES OF THE ACTIVE PARTS (ºC)
CASE B CASE C
temperatures were recorded. The input values were obtained
Stator lamination 69.9 32.9 with a Yogokawa PZ4000 power analyzer, and the output
Rotor lamination 151.7 115.3 values were measured using a Magtrol torque transducer
Stator coils 68.8 31.4 placed on the shaft. The measurement results at 380 Nm and
Magnets 153.8 117.4 1500 min-1 were selected for comparison because the weakest
cooling method A could only sustain this load without a risk
In addition, temperatures in the stator laminations for both of burning the stator insulation. The temperature rise in the
cases are presented in Fig. 9 and the rotor temperatures are windings recorded for each of these three cooling approaches
depicted in Fig. 10.
with a 380 Nm load (100% load for the indirectly water-
cooled machine A and 60% load for the directly liquid-cooled
machine C) (B being in the middle) and 1500 rpm speed are
illustrated in Fig. 12.

150
A, Indirect water cooling
B, Direct Liquid cooling - Oil
C, Direct Liquid cooling - Water
Temperature rise (C)

100

Figure 9. Temperatures in the stator laminations at steady state 60 kW

operation at 1500 min-1 for direct oil cooling (on the left) and direct water 50
cooling (on the right) in ºC.

0
0 50 100 150 200 250
Time (min)
Figure 12. Winding temperature rise recorded for the indirect water cooling
approach A and the new direct liquid cooling methods B (PAO) and C
(Water).

For the cooling method A, a maximum temperature rise of

110 °C was recorded at the end windings. For B, the proposed
Figure 10. Rotor temperatures with oil cooling (on the left) and water cooling direct liquid cooling method with PAO oil as coolant, the
(on the right) in ºC. temperatures stabilized at approximately 86°C, with a
temperature rise of 60 °C. There was more than a 50°C
Temperature distribution in the windings from 3D CFD temperature rise difference between the indirect and direct
analyses are presented for direct water cooling in Fig. 11 a) liquid cooling approaches, thus validating the predicted
and direct oil cooling in Fig. 11 b). superiority of the latter. Further, in case C (direct liquid
cooling with water), the maximum temperature of the winding
surface was only 57 °C with a temperature rise of only 31 °C.
The temperature rise difference between the oil and water
coolants was 29 °C at the selected load point.
Table VI presents the measurement results taken from the
testing of the cooling CASEs A, B, and C. Because traditional
windings were replaced with a hybrid tooth-coil winding for
the versions B and C, there were substantial differences in the
stator windings for the indirect and direct liquid cooling tests.
With direct oil cooling, the maximum temperature is 86 °C,
Figure 11. Temperature distribution (K) in the windings for a) direct water and with direct water cooling only 57 °C. The flow rates in the
cooling and b) for direct oil cooling from 3D CFD analysis. primary and secondary cooling circuits were 6 l/min and 8
l/min, respectively.

0278-0046 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2017.2681975, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

TABLE VI cooling could provide acceptable (below 155 °C) end winding
MEASUREMENT RESULTS OF THE PROOF-OF-CONCEPT MACHINE USING temperatures even with 500 A, with a current density of J = 14
COOLING METHODS B AND C (LOAD 380 NM, SPEED 1500 RPM)
A/mm2.
CASE CASE CASE
Parameter
A B C
Shaft power, kW 59.1 59.2 59.2
Input power, kW 62.9 62.7 62.7
Current, A 100.6 147.3 147.3
Copper losses, kW 0.8 1.0 1.0
Total losses, kW 3.8 3.5 3.5
Efficiency 94.1 94.4 94.4
Maximum temperature in windings, °C 137 86 57
Maximum temperature rise in windings, °C 110 57 28
Average temperature of end windings, °C 137 70 40
Average temperature inside of slot, °C 127 66 36

The water flow speed could still be increased, resulting in

still slightly lower temperatures. When the cooling flows were
increased to 9 l/min (primary circuit) and to 30 l/min
(secondary circuit), the temperatures decreased further by Figure 13. CASE B, coil temperature rise, oil flow, oil input temperature,
approx. 3 °C. measured losses, and estimated amount of heat losses removed by the coolant.
At speed 300 min-1.
The measured values are compared with the previous CFD
results. The results are gathered in Table VII. With cooling B TABLE VIII
CASE S B, C, SPEED 300 MIN -1, TORQUE 378 NM, AND SHAFT POWER 11.9 KW.
and cooling C, the computed results are close to the measured
Cooling fluid PAO Oil, Water,
ones. All the measurement sensors are located on the surfaces CASE B CASE C
of the coils, and thus, the mean value in each case was C1*Primary cooling flow, l/min 2.6 6
compared with the surface-averaged value obtained from the Liquid inlet temperature C1, °C 23.6 15
CFD. Liquid outlet temperature C1, °C 28.5 17.2
C2**Secondary cooling flow, l/min 8 8
TABLE VII
Current, A 147.8 149.5
COMPARISON BETWEEN THE CFD S IMULATION AND EXPERIMENTAL
Total losses, kW 1.2 1.3
TEMPERATURES (°C). THE EXPERIMENTAL TEMPERATURE IS THE AVERAGE
Losses removed by the liquid, kW 0.3 0.96
TEMPERATURE OF 12 TEMPERATURE S ENSORS LOCATED IN BOTH STATORS.
Maximum temperature in the windings, °C 43.2 27.3
Experimental *C1, Main cooling circuit through the coils
CFD Simulation
measurements **C2, Secondary cooling circuit with tap water
CASE B 74 69
CASE C 37 39 With oil cooling, the flow rate is approx. 6 l/min at the
maximum. The oil flow circulating in the tubes caused a slight
Direct cooling analysis pressure difference. The upper oil pressure difference was
The flow speed of oil varies with a certain pressure difference close to 6 bar with oil. It is the maximum pressure that the oil
because of the oil viscosity behaviour as a function of the fluid pressure sensor used in the study can sustain, and therefore,
temperature. The heat tests take a long time until the oil this oil could not be driven with a higher flow rate. The oil
temperatures stabilize to a value that changes less than one used was Chemlube 221 PAO-based compressor lubricant.
degrees Celsius per 30 minutes. The direct oil cooling However, with water, the pressure difference is much lower,
performance was investigated at 300 min-1 speed. and higher flows could be tested. Fig. 14 presents the highest
A set of measurements were performed with a constant measured temperatures of the prototype machine windings as
speed with different loads. The obtained results are depicted in a function of current at the 300 min-1 speed.
Fig. 13, and the measurement values are shown in Table VIII. 50

The results show for instance that at 150 A current the

machine losses were 1.2 kW, and the oil removes approx. 0.3 40

kW of these losses. The PAO oil flow was only 2.6 l/min, and
the maximum measured temperature was 40 °C. At the highest 30
Temperature (oC)

measured current value of 325 A, the machine losses were 5.2

Oil circuit
kW, and in this case, the proportion of the direct PAO oil 20
cooling was approx. 3.8 kW. The maximum winding Water circuit
temperature exceeded a value of 80 °C at 5.2 kW losses. 325 10
A (J = 9.1 A/mm2) was the maximum inverter supply, and
therefore, the machine could not be driven at higher currents 0
in the stabilized condition. However, based on the 0 50 100 150 200 250 300

measurement results, the windings could most probably Current (A)

sustain much larger currents and current densities. This liquid Figure 14. Maximum temperatures (Celsius) as a function of current at 300
min-1 speed, CASEs B and C.

0278-0046 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2017.2681975, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

The measurements clearly indicate the temperature-rise density and should result in about 110 kPa tangential stress. At
difference with water and oil as a coolant as can be seen in the rated operation point of 100 kW the stress value was 57
Fig. 14. The efficiencies of the axial-flux machine were kPa. The thermal CFD simulations are accurate according to
measured with three speeds: 600, 900, and 1500 min-1, and the verifications of measured temperatures from a real machine.
results are gathered in Fig. 15. The temperature measurement results confirm that the direct
cooling method is effective in small traction machines, and
improved thermal management.
The AFPMSM used in this analysis offers only moderately
high tangential stress values at the rated operation. This is
partly because the effective proportion of the copper winding
length is only 37 % in this case despite the tooth-coil design.
The design stress value for direct liquid cooling was 57 kPa
with J = 7 A/mm2, but obviously, the water-cooled version
could, at least in low-speed operation, offer about 110 kPa
continuous tangential stress. Such values would be appreciated
in traction applications. The problem with the present design
Figure 15. Efficiencies of the axial-flux machine from measurements for three is that it needs a high voltage reserve to provide such high
speeds 600, 900, and 1500 min-1, CASE B. values. Therefore, they are available only below the rated
The efficiencies of the original motor utilising round speed. Also this AFPMSM has quite a low useful winding
enamelled copper wires and the new motor utilising Litz-wires proportion as the end windings are still long despite the tooth-
were actually quite similar. In the indirectly cooled machine, coil design. Naturally, the winding material only allows a
the windings were constructed with 12 conductors around certain minimum bending radius, and the machine has now
each tooth, and the phase resistance was 0.0265 Ω. In the been designed based on such a value. A radial flux machine
directly cooled Litz windings, windings were constructed with with a longer stack might be capable of offering higher
only 8 conductors around each tooth, and their phase tangential stress values with direct water cooling. This cooling
resistance was measured to be 0.0165 Ω. All measured method seems to provide an opportunity to achieve the most
efficiencies of the original motor (case A) and the new directly compact machine designs in cases with high Joule losses,
cooled motor case C are shown in Fig. 16. The directly cooled especially in traction applications.
motor C has higher efficiencies at 0.5 p.u. load area, but
suffers of higher copper losses as the torque increases above REFERENCES
0.8 p.u. as can be seen in the results depicted in Fig. 16. [1] W. Hua, H. Zhang, M. Cheng, J. Meng, and C. Hou, “An Outer-Rotor
96
Flux-Switching Permanent-Magnet-Machine With Wedge-Shaped
Magnets for In-Wheel Light Traction,” IEEE Transactions on Industrial
94 Electronics, Vol. 64, n. 1, pp. 69–80, 2017
[2] S-U. Chung, S-H. Moon, D-J. Kim, and J-M. Kim, “Development of a
92 20-Pole–24-Slot SPMSM With Consequent Pole Rotor for In-Wheel
90 Direct Drive,” IEEE Transactions on Industrial Electronics, Vol. 63,
Indirect liquid cooling, A n. 1, pp. 302–309, 2016
Efficiency (%)

88 [3] X. Liu, H. Chen, J. Zhao, and A. Belahcen, “Research on the

Direct water cooling, C Performances and Parameters of Interior PMSM Used for Electric
86 Vehicles”, IEEE Transactions on Industrial Electronics, Vol. 63, n. 6,
84 pp. 3533–3545, 2016
[4] M. Polikarpova, P. Lindh, C. Gerada, M. Rilla, V. Naumanen, and J.
82 Pyrhönen, “Thermal Effects of Stator Potting in an Axial-Flux
Permanent Magnet Synchronous Generator,” Applied Thermal
80
0.0 0.2 0.4 0.6 0.8 1.0 1.2 Engineering, Elsevier, pp. 1359–4311, 2014

Power (p.u.) [5] Galea, M., Gerada, C., Raminosoa, T, and Wheeler, P. “A Thermal
Improvement Technique for the Phase Windings of Electrical
Figure 16. Measured efficiencies of the original motor case A and directly
Machines,” IEEE Transactions on Industry Applications, Vol. 48, n. 1,
liquid cooled motor case C at 1500 min-1.
2012, pp. 79–87.
[6] J. Pyrhönen, P. Lindh, M. Polikarpova, E. Kurvinen, and V. Naumanen,
V. CONCLUSION ”Heat-Transfer Improvements in an Axial-Flux Permanent-Magnet
Synchronous Machine,” Applied Thermal Engineering, Elsevier, pp. 1–
Direct water cooling was applied in a low-voltage 7, 2015
permanent magnet axial-flux machine. The cooling [7] M. Polikarpova, P. Ponomarev, P. Lindh, I. Petrov, W. Jara, V.
effectiveness is far superior to indirect water cooling Naumanen, J. A. Tapia, and J. Pyrhönen, “Hybrid Cooling Method of
Axial-Flux Permanent Magnet Machines for a Vehicle Application,”
effectiveness. The original AFPMSM tolerated about 4 A/mm2 IEEE Transactions on Industrial Electronics, Vol. 62, n. 12, pp. 7382–
current density with about 34 kPa tangential stress at the rated 7390, 2015
operating point (CASE A). The tests with direct water cooling [8] R. Camilleri, D. A. Howey, and M. D. McCulloch, “Predicting the
Temperature and Flow Distribution in a Direct Oil-Cooled Electrical
(CASE C) indicate that with these dimensions, even 14 Machine With Segmented Stator”, IEEE Transactions on Industrial
A/mm2 is feasible. This is 3.5 times the original current Electronics, Vol. 63, n. 1, pp. 82–91, 2015

0278-0046 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2017.2681975, IEEE
Transactions on Industrial Electronics

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

[9] B. Zhang; T. Seidler; R. Dierken; M. Doppelbauer, “Development of a Juha Pyrhönen (M’06), was born in 1957 in
Yokeless and Segmented Armature Axial Flux Machine,” IEEE Kuusankoski, Finland, received the Doctor of
Transactions on Industrial Electronics, Vol. 63, n. 4, pp. 2062–2071, Science (D.Sc.) degree in Electrical
2016 Engineering from Lappeenranta University of
[10] J. Montonen, S. Sinkko, P. Lindh, J. Pyrhönen, “Design of a traction Technology (LUT), Finland in 1991. He
motor with two-step gearbox for high-torque applications,” in 2014 became an Associate Professor of Electrical
International Conference on Electrical Machines (ICEM), 2014, pp. Engineering at LUT in 1993 and a Professor of
1069–1075. Electrical Machines and Drives in 1997. He is
[11] P. Ponomarev, R. Åman, H. Handroos, P. Immonen, Juha Pyrhönen, and engaged in research and development of
L. Laurila, "High power density integrated electro-hydraulic energy electric motors and electric drives. His current
converter for heavy hybrid off-highway working vehicles,” IET interests include different synchronous
Electrical Systems in Transportation, Vol. 4, n. 4, pp. 114 – 121, 2014 machines and drives, induction motors and drives and solid-rotor high-
[12] F. Caricchi, F. Crescimbini, A. Di Napoli, “Prototype of Innovative speed induction machines and drives.
Wheel Direct Drive with Water-cooled Axial-flux PM Motor for Electric
Vehicle Applications,” in Applied Power Electronics Conference and Marco Satrústegui (San Sebastián, Spain,
Exposition, APEC, pp. 764–770, 1996. 1988) is a PhD student in the Transport and
[13] M. Schiefer, and M. Doppelbauer, “Indirect Slot Cooling for High- Energy Division at CEIT-IK4 Technology
Power-Density Machines with Concentrated Winding,” IEMDC, 2015, Center. He received his degree in Industrial
pp. 1820–1825. Engineering (2012) from the University of
[14] Z. Ping, L. Ranran, P. Thelin, E. Nordlund, and C. Sadarangani, Navarra in San Sebastián, Spain.
“Research on the cooling system of a 4QT proof-of-concept machine His main research interests are in the field of
used for HEV,” IEEE Transactions on Energy Conversion, vol. 23, no. electrical machine design, and specifically in
1, pp. 61–67, 2008. thermal design.
[15] S. Nategh, Z. Huang, A. Krings, O. Wallmark, and M. Leksell, “Thermal
modeling of directly cooled electric machines using lumped parameter
and limited CFD analysis,” IEEE Transactions on Energy Conversion,
vol. 28, no. 4, pp. 979-990,2013.
Miguel Martinez-Iturralde is a researcher in
[16] M. Polikarpova, S. Semken and J. Pyrhönen, “Reliability analysis of a
the Transport and Energy Division at CEIT-IK4
direct-liquid cooling system of direct drive permanent magnet
Technology Center. He received his degree in
synchronous generator,” Reliability and Maintainability Symposium
Industrial Engineering (2001) and his Ph.D.
(RAMS), 2013.
degree (2005) from the University of Navarra.
[17] G. K. Ridley, “The UK's first direct-water-cooled pumped-storage
He has been a lecturer of Electrical Machines
generator-motor,” IEE Electronics and Power, pp. 678–681, Oct. 1982
and Electrical Circuits at Tecnun, the School of
[18] R. F. Gray, L. W. Montgomery, R. J. Nelson, J. R. Pipkin, S. Joki-
Engineering of the University of Navarra, in San
Korpela, and F. Caguiat, “Designing the Cooling Systems for the
Sebastian, Spain since 2006.
World’s Most Powerful Turbogenerator –Olkiluoto Unit 3,” IEEE 2006
His main research interests and expertise are in
[19] Gurakuq Dajaku; Harald Hofmann; Fatmir Hetemi; Xhevat Dajaku; Wei
the field of the design and control of electrical
Xie; Dieter Gerling, "Comparison of Two Different IPM Traction
machines for transport applications.
Machines With Concentrated Winding,” IEEE Transactions on
Industrial Electronics, Vol. 63, n. 7, pp. 4137–4149, 2016
Ahti Jaatinen-Värri received his M.Sc. in
[20] Xiangdong Liu; Hao Chen; Jing Zhao; Anouar Belahcen, “Research on
energy technology from Lappenranta Univeristy
the Performances and Parameters of Interior PMSM Used for Electric
of Tehcnology in 2005, and his doctoral degree
Vehicles,” IEEE Transactions on Industrial Electronics, Vol. 63, n. 6,
from LUT in 2009. He currently works as an
pp. 3533–3545, 2016
associate professor in the Laboratory of Fluid
[21] W. Jara, P. Lindh, J. A. Tapia, I. Petrov, A. K. Repo, and J. Pyrhonen,
Dynamic at LUT. He also acts as a head of
"Rotor Eddy Current Losses Reduction in an Axial Flux Permanent
Energy Technology degree programme. His
Magnet Machine," in IEEE Transactions on Industrial Electronics ,
main academic interests are compressor and
pp.1–9, 2016.
turbine aerodynamics, small gas turbines and
[22] P. Lindh, I. Petrov, R. Semken, M. Niemelä, J. Pyrhönen, L.
educational development.
Aarniovuori, T. Vaimann, A. Kallaste, “Direct Liquid Cooling in Low-
Power Electrical Machines–Proof-of-Concept,” IEEE Transactions on
Aki Grönman, born in 1980 in Pori, Finland,
Energy Conversion, Vol. PP, n. 99, 2016
received his Doctor of Science (D.Sc.) degree
from Lappeenranta University of Technology
(LUT), Finland in 2010. He is currently working
Pia Lindh (M’04) born in Helsinki in 1969,
as an associate professor at the School of
received her M.Sc. degree in energy technology in
Energy Systems at LUT, where he is engaged
1998 and her D.Sc. degree in electrical
in teaching and research of turbomachinery,
engineering (Technology) in 2004 from
applied fluid dynamics and renewable energy.
Lappeenranta University of Technology (LUT),
His current research interests include combining
Lappeenranta, Finland. She is currently serving as
fluid dynamics into different research areas with
an associate professor at the Department of
special interest in renewable energy production.
Electrical Engineering in LUT Energy,
Lappeenranta, where she is engaged in teaching
and research of electric motors and electric drives.
.

Ilya Petrov received the D.Sc. degree from

Lappeenranta University of Technology (LUT),
Lappeenranta, Finland, in 2015. He is currently a
Fellow Researcher in the Department of Electrical
Engineering, LUT.
His current research interests include the design
and optimization of permanent-magnet machines.

0278-0046 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.