IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Abstract—This paper presents innovative research towards the
development of a 45 kW high speed switched reluctance drive as
an alternative starter-generator for future aero-engines. To
perform such a function the machine had to be designed with a
very wide constant power-speed range. During engine-
start/motoring mode, a peak torque demand of 54 Nm at 8 krpm
was met, whilst in generating mode, 19.2-32 krpm, the machine
was designed to deliver a constant power of 45 kW. The key
enabling feature of the design lies in the novel rotor structure
developed so as to allow for such a wide speed range. The results
presented, are those measured during the initial testing phase
and validate the system design and performance in the low-speed
region with the machine operated in starting-mode. The
measured machine power density is at 9.8 kW/ltr, whilst the
global system efficiency is at 82%.
Index Terms— Electric Machines, High speed generators,
Reluctance Generators.
I. I
NTRODUCTION
IR travel has increased in popularity and scope, a trend
which is likely to continue for years to come [1]. Such
growth however, is taking place in a very demanding
environment for both operators and component manufacturers,
who struggle to lower operational/running costs and
environmental impact, whilst at the same time increasing
aircraft safety and reliability.
To meet such demands, a lot of effort is being put into the
development of a More Electric Aircraft, encompassing a
More Electric Engine [1, 2]. Both concepts express the need
for placing electricity as the unitary vector of energy, over
hydraulic and pneumatic methods, providing advantages in
terms of: relative ease of power routing within the airframe, a
reduction in both working-fluid volume and piping
Manuscript received December 14, 2015; revised May 16, 2016 and
August 3, 2016; accepted August 22, 2016.
J. Borg-Bartolo was with the Power Electronics and Control Group
at the University of Nottingham during the course of this work (e-mail:
james.borg-bartolo@maxonmotor.com Phone: 0041 764274225.
M. Degano, J. Espina and C Gerada are with Power Electronics
Machines and Control group (PEMC) at the University of Nottingham,
University Park, Nottingham, NG72RD (e-mails:
Marco.Degano@nottingham.ac.uk; Jordi.Espina@nottingham.ac.uk;
Chistopher.Gerada@nottingham.ac.uk ).
Fig.1. Typical Torque-speed requirements considered for the design of
a modern starter-generator
infrastructure, potentially lighter aircraft and an increase in
global system efficiency and reliability through lower
component count and spool speed decoupling.
Such a situation is thus bringing about an increased need for
larger, more reliable, on-board generation systems [2].
In response to this, and following from earlier work
presented in [3, 4] this paper details the work done in
designing, building and testing of a new switched reluctance
based starter/generator (S/G) system for a modern regional jet
aero-engine; capable of satisfying the wide constant power
speed range detailed in Fig.1.
Such a function requires the machine to initially act as a
motor, spinning the aero-engine through the starting sequence,
till light-off speed (maximum torque condition). It then
continues to assist the engine till idle speed is reached,
following which the machine acts as a generator, ref. Fig.1.
The machine presented, was designed to deliver its
maximum torque of 53.8 Nm at 8 krpm, and will be required
to supply a 45 kW load at shaft speeds up to 32 krpm.
An identical physical envelope to that imposed on current
state-of-the art S/G machines was adopted such that the
maximum permissible stator outer diameter was limited to
200 mm, and a shared oil system was considered, with an
outer-jacket stator-cooling topology. The inlet oil temperature
was specified at 120˚C.
Such an envelope was found to be in close agreement with
what was previously reported in [5-7] for similar SR-based
drives. However, in all of the mentioned cases the starting-
power demand was much lower than that required during
generating mode, resulting in a reported constant-power speed
Design and Initial Testing of a High Speed
45 kW Switched Reluctance Drive for
Aerospace Application
James Borg Bartolo
,
Marco Degano
Member
,
IEEE
,
Jordi Espina and Chris
Gerada
Member
,
IEEE
A
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
range-ratio (CPSRR) which varied from 1:2 up to 1:3.25 [8].
The drive detailed in this work was required to achieve a
CPSRR of 1:4, in discontinuous-conduction mode and thus
presents a novelty to the field of SR based starter/generators
rated at this power level.
Following this introduction, the present document is divided
into three main parts. The first, section II, details the
governing principles, and salient analysis performed in
designing the novel switched-reluctance machine (SRM)
capable of achieving the stated wide constant power speed
range. The second part, in Sections III, IV and V, presents the
built power electronic converter; adopted control strategy and
test setup for initial machine testing. The third part of the
paper, in section VI, presents the measured results obtained
from initial testing of the assembled drive. Finally a number of
conclusions are presented in section VII.
II. T
HE
S
WITCHED
R
ELUCTANCE
M
ACHINE
The SRM is a doubly salient, singly excited machine. In its
classical form, the machine presents itself with a ‘simple’
construction having a single element rotor and concentrated
windings. These attributes coupled with diagnostic techniques
such as those presented in [9], renders such a topology a
favourite for high speed, harsh environment applications. The
studied variant of this machine is of such classical type.
A number of stator and rotor pole number combinations
were initially investigated. It was found however that the 6/4
pole combination provided the best compromise between iron
losses, self-starting capability and torque ripple [5-7].
The presented design makes use of a novel, rotor pole
chamfer, ref. section II D, which allows for the demanded
wide constant power speed range to be achieved.
A.
Basic operation and principles
Due to its sequential commutation, the electromechanical
energy conversion in such machines occurs in discrete cycles,
or strokes, through the interaction of one (or more), stator and
rotor pole pairs.
The energy supplied to a model system formed by a set of
current-carrying phase coils whose flux couples with, and
induces a resultant force within a rotating element, can be
summarized in terms of measurable (or easily derivable)
quantities via conservation principles as:
(1)
Where: i is the current flowing into the system, ψ is the flux
linkage, w
s
is the total stored energy, T
e
is the developed
torque, and ϑ is the general rotor displacement.
In order to facilitate the initial treatise of the system, it is
very useful to consider the flux linkage setup in a phase coil
system as being a function of the currents flowing within it
rather than vice versa, such that (1) can be restated in terms of
the incremental change in co-energy W
f
, for the single phase
system by:
(2)
Due to the cyclic nature of the energy conversion process
undertaken, and considering steady state conditions, it can be
shown that no net change in the energy level of the system
occurs over one cycle, thus one can restate (2) in its integral
form by:
(3)
Where: C
i
is the path in the coordinate space i-. This
relationship thus states that the developed torque during one
cycle can be evaluated by taking the integral of the flux
linkage locus in a diagnostic plane, -i along the contour
specified in the coordinate space i-, which in practice is none
other than the time-domain current waveform imposed on the
winding; thus highlighting the main advantage of using the co-
energy principle in correlating measurable electrical terminal
quantities to measureable mechanical effects.
Using standard differential theory and the co-energy
expression given in (2) it can be shown that the torque
developed under one pole can be stated in terms of the
experienced rate-of-change of energy with respect to rotor
positions as:
(4)
Where:
m
refers to the change in angular position during
the imposed phase-conduction period
cond
, and
is
the energy converted during such cycle, evaluated as the area
enclosed by the flux-linkage locus in the -i domain (one such
locus was derived numerically in Fig.5).
By considering (4), the area enclosed by the flux-linkage
locus in the -i domain must therefore relate to the average
developed torque per stroke such that:
!
"
#
$
%
&
'
(5)
Where: M is the number of strokes per mechanical cycle.
Thus, knowing the value of the required average torque the
machine should develop, the energy per cycle can be
estimated. This, together with using a linear inductance model
and a reluctance network approach allows for values flux
linkage to be obtained at various rotor positions and hence an
initial machine geometry to be estimated, as described in [3].
B.
Electromagnetic design considerations
To meet the wide constant power-speed requirements set
out in Fig.1, the two extreme power nodes were considered as
design points.
The first, defined as the base-speed, was chosen at 8krpm
(peak torque node), with initial machine sizing carried out at
this point for a peak phase current of 600 A. Such an exercise
was carried out during the earlier design stages, using an
analytic method developed by the author, based on (4), (5) and
work published in [10-14]. In summary, the proposed stator
model adopted a linear approximation to the magnetic circuit,
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
assuming that only the stator poles are in heavy saturation
[11]. Thus initial sizing relationships for the stator tooth arc
angle s, stator back iron b
sy
, stator tooth width T
sb
and bore
Diameter D were derived. The rotor geometry-estimation-
model was similarly based on a reluctance network approach
[3]. A set of curves for the estimated flux linkage variation
within the stator tooth, for a given MMF level, as a function of
machine air-gap l
g
and rotor chamfer angle
cm
were derived
[3]. Following this initial sizing exercise, an iterative
optimization technique relying on finite element simulations
was later used to tune the obtained dimensions and arrive to a
final design for the required machine fulfilling the necessary
base speed ratings.
The second design point considered was the high speed
constant power node at 45 kW, 32 krpm. The operation of the
base-speed-optimized machine at such high speed was
verified, via analytical methods based on [15] and the required
geometry optimization was carried out as detailed in [3] to
meet requirements.
The final geometry for the optimized machine is given in
Fig.2. It has a 120 mm bore diameter, with an air gap of
0.5 mm and a stack length of 150.6 mm; its outer diameter is
192 mm, and occupies a volume of 4.592 ltrs.
An overlapping current excitation was assumed throughout
the design process. This, together with the implemented
circumferentially-alternating pole distribution, allowed for a
short flux-path during commutation to the neighbouring pole.
Both stator and rotor employed a cobalt-iron laminate
material at 0.1 mm, post-process annealed at different
conditions to reach optimal electromagnetic and mechanical
properties respectively.
C.
Stator design
The main feature in the stator design was the inclusion of
the stator-pole tips, which allowed for a reduction in tooth
width whilst ensuring that the minimum pole arc at the bore
diameter s, required for self-starting, is maintained. Whilst
not being novel, this feature contributed towards reaching the
required operating torque-speed envelope. By solely
considering the low speed performance as a first iterate, a
value of 0.49 was chosen for the pole envelope at the tooth-
base “S
penv
”. However, during the high speed optimization
process this value was reduced to 0.44 which allowed for a
3.4% increase in the high speed performance of the machine.
Due to the incurred reduction in tooth-surface area, an earlier
saturation of the machine was brought about, such that at low
speed mode, a 2.8% reduction in torque was noted. As
discussed by the author in [4] notwithstanding this reduction,
the design still met the low speed requirement with a healthy
margin as detailed in the coming sections.
D.
The novel rotor structure
The terminal voltage equation of the SRM can be stated in
terms of the flux linkage variation as:
(
)
*
*
(6)
Fig.2. Design of SRM having a very wide constant power speed range
Fig.3. Ideal position inductance-variation for the linear machine with
rotor pole chamfer
Where: v
t
is the terminal voltage, r
s
is the phase resistance,
and (i,
m
) is the flux linkage variation as a function of
current and rotor position
m
. It can be noted that one of the
major hindrances to developing the required torque at high
speed values, for a fixed converter voltage, is the induced
pseudo-Back EMF term (third term RHS) which reduces the
available voltage ceiling and thus blocking the required
current from being present in the winding.
By introducing a pole chamfer angle
cm
, ref. Fig.2, a handle
on such a term could thus be introduced.
1)
Rotor Electromagnetic Design
During pole alignment the inductance increases in a
proportional way to the stator pole-overlap, such that the
maximum change-of-inductance is only limited by the
magnetic state of the stator pole arc and its duration is dictated
by s.
By introducing a pole chamfer, the rotor crown arc angle
rc
could be decoupled from r such that a pre-alignment period
could be defined by
cm
, and expressed as:
+
(7)
where:
rc
is the rotor crown angle and r is the
conventional rotor pole arc angle.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Fig.4. Plot showing the flux-linkage for the chamfered and non-
chamfered designs. Current increased by 25A at each
,
rp
Fig.5. Estimated flux linkage loci at 32 krpm for different chamfer
angles,-single voltage pulse operation, 270V; duration: 24º elect
The change-of-inductance period is thus ‘elongated’ and
two different alignment periods, with two different gradients
are introduced, ref. Fig.3. During the first period, the poles
start overlapping but the increased reluctance due to the
chamfer limits its positional-rate-of change. During the second
period, starting at
1cm
, the rotor crown starts to overlap in a
similar fashion to the original non-chamfered designs. Thus, a
similar positional-rate-of-change of inductance is experienced,
with duration approximately equal to
s
. A number of effects
need to be noted.
The first is a positive effect in that the introduction of such
a chamfer widens the rotor slot at the air-gap, thus decreasing
the unaligned inductance when compared to a similar non-
chamfer design, ref. Fig.4, thus enabling higher torque
production in the high-speed region as per Fig.5.
The second effect is however detrimental in the low-speed,
high-torque region, in that the introduced chamfer increases
the magnetic impedance when compared to the non-chamfered
designs thus increasing stator pole leakage flux. Due to this
increased reluctance, a slight reduction in the total flux
handled by the circuit could be observed for identical current
loading conditions. Such an effect lowers the positional rate-
of-change of inductance even during the main-alignment
period (which should ideally have the same gradient as the
non-chamfered design). Thus, if not carefully catered for, the
introduction of such a chamfer might bring about a reduction
in the Torque/Ampere ratio of the machine.
Fig.6. Plot showing the flux-linkage for the chamfered and non-
chamfered designs at maximum current.
Fig.7.
Plot showing mutual flux linkage with Phase 3 for the chamfered
and non-chamfered designs
Both these effects can be observed in Fig.4 and Fig.6 which
portray the self-flux linkage variation obtained via a 2D Finite
Element (FE) profiling analysis for a sub-optimal design.
During such ‘quasi-static’ evaluation the rotor was moved in
discrete steps through one electrical revolution (90º
mechanical) under constant current, single coil excitation. In
this case, forty fixed current steps were considered until a
phase current of 1000 A was reached. The FE suite used was
Magnet™ from Infolytica
©
. The model employed 530,000
nodes and used a non-linear BH characteristic for the
ferromagnetic materials considered, (Vacoflux48 for the stator
and Vacodur S+ for the rotor).
The required initial magnetisation curves were provided by
the material supplier Vacuumschmelze GmbH.
A third important effect of introducing such a chamfer is the
effective control on the mutual flux contribution. This is of
particular importance during current overlap where the mutual
flux contribution holds the turning-off phase into saturation
even though the current through it starts decreasing. This
brings about a reduction in the aligned unsaturated inductance
Lau and therefore a further reduction in the torque-developing
potential of the phase. The reduction of such mutual flux
contribution is shown for the same design in Fig.7.
The combined effect of the mentioned features resulted in a
machine capable of operation at and above the given torque-
speed envelope. The final design had stator and effective rotor
pole arc angles at 30º and 41.75º respectively together with a
pole chamfer angle of 8/32.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Fig.8.
Comparison of simulated performance between proposed and
conventional geometry machines
An initial Torque-speed characteristic comparison to a
standard machine geometry (having conventional stator and
rotor teeth and identical dimensions to the new design) was
performed using a Simulink
®
based SRM block model with
results reported in Fig.8. At low speeds, (up to 8000 rpm), the
current was limited to 600A (nominal) via a hysteresis
controller set at ±10% nominal current.
2)
Rotor Structural evaluation
Further to the introduction of the described chamfer, two
other features were added to the rotor profile. The first,
described in [4] saw the introduction of a cavity within the
rotor pole structure to increase its magnetic utilization as
shown in Fig.2. With the correct sizing this feature reduced
the rotor weight by 3.5% with minimal torque reduction
(0.57%), at base speed.
The second feature allowed for the inclusion of rotor-slot
inserts to render the whole structure more aerodynamically
efficient. This required a minor modification in rotor-slot
geometry to allow for their retention at such high speeds. The
chosen material was a high temperature thermo-setting
polymer, (µ
r
=1), called Meldin7001, rated for continuous use
at temperatures above 300˚C.
An optimization of the rotor geometry to reduce stress
concentration issues and ensure its safe operation at the
operating speed of 32krpm was thus carried out.
Due to the predominantly radial nature of the experienced
forces, a 2D plane-stress model was setup within ANSYS
®
Workbench R15, an off-the shelf, FE suite. The model used
185873 nodes with 60249 elements of type Plane82 for the
lamination geometry. A linear elastic behaviour was assumed
for all materials, whilst allowing for large displacements,
geometric non-linearities and thermally induced stresses. The
thermal and centrifugal conditions were imposed on the
complete body or bodies.
The rotor surface temperature was estimated at 300ºC using
the thermal model similar to that described in [16] as detailed
in [4] and correlations given in [17-19]. The input iron and
copper losses are reported in Table I. Results for the
undergone structural analysis in the ‘Hot rotor’ condition at
nominal speed (32krpm) are reported in Fig.9.
Using the estimated stress distribution, the final laminate
material was chosen to be VACODURS+ with a minimum
proof stress of 750 MPa.
Fig.9. Equivalent Von Mises Stress distribution in the finalized rotor
geometry at the peak estimated temperatures including inserts
At the imposed rotor temperatures, and rated speed, the
peak deformation of the rotor pole at the air-gap was estimated
at 0.24 mm whilst the Meldin deformed by a maximum of
0.7 mm. Such a large deformation was however allowed since
the inserts’ outer edge was purposely located at a smaller
diameter than the air gap.
E.
Simulated Drive performance summary
In order to better quantify the machine performance
together with the incurred iron and switching losses,
particularly for the low speed high torque region in which
current control is provided via a hysteresis controller, a co-
simulation approach was adopted based on a Matlab
®
Simulink
®
platform.
The required asymmetric H-bridge driving the SRM was
modelled using a PLECS™ library for non-ideal Power-
Electronic (PE) devices and controlled via a hysteretic current
controller in Simulink
®
. This control environment was linked
via a dedicated Simulink
®
block provided by the FE suite
developer, Infolytica
©
. The PE losses included both
conduction and switching loss terms, both of which were
estimated using an off-the-self PLECS library model for
IGBTs. Briefly the conduction losses were based on an
equivalent IGBT model using a series connection of a DC
voltage source, U
ceo
, ( representing the device on-state-zero-
current collector-emitter voltage), and a collector-emitter on-
state resistance , r
ce
, values of which were read-off the
respective datasheet and inputted to the model. The switching
loss energies for both IGBT and diode (mainly reverse
recovery energy during IGBT turn-on) were estimated using
the datasheet reported energy loss, computed collector current
and collector-emitter voltage. A fixed time interval had to be
implemented for the FE-solving block to function correctly.
The maximum simulation time-step was set to 1 ns.
Using this method, the simulated performance for the final
drive design at three salient power-speed nodes was evaluated,
proving the wide constant power speed range capability of the
designed machine, ref. Table I. Note a worst-case switching-
loss condition was assumed by imposing a switching strategy
in which both top and bottom devices (ref. Fig.10) were
turned-off simultaneously.