Design Optimization With Flux-Weakening of

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.2017.2726962, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Design optimization with flux-weakening of

High-Speed PMSM for electrical vehicle
considering the driving cycle
L. Dang, N. Bernard, N. Bracikowski, G. Berthiau
Abstract—In this paper, the design optimization of a non- kec eddy currents specific loss coefficient
salient high-speed Permanent Magnet Synchronous Ma- kf slot filling factor
chine (PMSM) for electric vehicle applications is presented. kh hysteresis specific loss coefficient
It will be shown how, with a new approach, it is possible kt slot opening to tooth opening ratio
to find a deterministic solution to solve the sizing of the kw winding coefficient
machine from a given driving cycle. The optimal geometry
and the optimal control strategy over the cycle minimizing kL coefficient for correcting the active length
both the energy losses and the volume of the machine L active length (m)
will be calculated. At first, the 1-D analytical model used p number of pole pairs
is presented and validated for the most significant point R outer stator radius (m)
of the driving cycle using a finite element method. Then, Rs inner stator radius (m)
the design methodology and the results through a specific
application are detailed. Particularly, it will be shown how Rw outer winding radius (m)
the flux weakening, directly given by the design process Rr outer rotor radius (m)
via the optimization of the control strategy, allows reducing rs reduced inner stator radius
both the energy losses and the constraints on the power rw reduced outer winding radius
converter. At last, in order to validate the solution consid- he air-gap thickness (m)
ering the whole cycle while keeping a reduced computation
time, a reluctance network model of the PMSM is used. This hpm permanent magnet thickness (m)
model validate the energy losses and the flux densities in Ns number of turns per phase
the steel parts over the cycle. The study will be done con- ΩL load mechanical angular velocity (rad/s)
sidering the Urban Dynamometer Driving Schedule (UDDS). g gear ratio (g = ΩM /ΩL )
ΩM machine mechanical angular velocity (rad/s)
Index Terms—High-speed PMSM, design optimization, M Vehicle mass (kg)
driving cycle, electric vehicle, flux weakening, reluctance cx Air drag coefficient
network. f Rolling resistance coefficient
S Vehicle frontal area (m2 )
I. N OMENCLATURE
Bf m magnitude of radial magnet flux density (T ) II. I NTRODUCTION
FM M stator magnetomotive force (At) The electrical vehicle is an interesting and efficient solution
Tem torque (N.m) for personal and urban transportation due to the fuel energy
Pc copper losses (W ) resources exhaustion and environmental concern [1]. Owing to
Pmg iron losses (W ) the limited capacity of batteries, electric motors for automotive
ψ torque angle (rad) applications are usually designed to maximize the energy
heq heat transfer coefficient (W/m2 .K) efficiency with a reduced volume [2]. On the other hand,
Sth heat transfer surface (m2 ) traction devices for transport vehicles are required to work
θ temperature (◦ C) on highly variable torque and speed which can be represented
σ electric conductivity (S.m−1 ) by standard driving cycles such as the Urban Dynamometer
µ0 permeability of the air (H.m−1 ) Driving Schedule (UDDS) or the New European Driving Cycle
µr relative permeability of iron (NEDC) [3]. These cycles are frequently composed of several
kad additional loss coefficient hundred working points. Thus, the optimal design process
based on a numerical calculation is one of the main difficulties
because it requires quite a lot of computation time, particu-
L. Dang, N. Bernard, N. Bracikowski and G. Berthiau are with the
Laboratoire Institut de Recherche en Energie Electrique de Nantes- larly when the profiles of the optimal control parameters are
Atlantique, Boulevard de l’Université, Saint Nazaire 44600, France searched in addition to the optimal geometry. In the literature,
(e-mail: thi-nhat-linh.dang@etu.univ-nantes.fr; nicolas.bernard@univ- we can find many references dealing with this problem. In [4]
nantes.fr; nicolas.bracikowski@univ-nantes.fr; gerard.berthiau@univ-
nantes.fr). [5] [6] [7], in order to solve it and evaluate properly the energy
consumption of the machine over the cycle, the whole cycle
0278-0046 (c) 2017 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.2017.2726962, IEEE
is reduced to some points in which the energy consumption of

the motor is the highest. The torque-speed plane can also be
divided into several uniform areas replaced by their barycenter
[8] [9]. In the optimization procedure, the performance of
the motor is evaluated from these points taking into account
equivalent weighted coefficients. These coefficients are the
ratio between the number of points in each area and the total
number of points on the cycle. In [10] [11] [12] [13], the
studies deal with the way to evaluate the energy losses at all
operating points. These methods are based on the combination
of the finite element analysis and an analytical model. More
generally, the authors propose analytical formulae, easier to
use in an optimization work, to determine the copper and iron
losses at every working point. Fig. 1. Speed and torque profiles of the vehicle.
Furthermore, the optimal sizing is generally carried out con-
sidering a given control strategy [11] [14] [15], not necessary TABLE I
the optimal one. This strategy consists in a flux-weakening V EHICLE DATA
mode above the base speed only. Below this base speed, the
torque angle (angle between the stator current and back EMF) Symbol Quantity Value
is set to a constant and maximize the torque per Ampere value; M Vehicle mass 700kg
and above the base speed, the torque angle is changed in order cx Air drag coefficient 0.3
to satisfy the voltage limitation due to the power converter, via f Rolling resistance coefficient 0.01
the maximization of the torque per Volt for example [16] [17] S Frontal area 1.5m2
[18]. In that case, the control strategy leads to minimize, for a
given torque, only the copper losses during operation whereas
iron losses, particularly in high-speed machines, can represent TABLE II
S PECIFIC DATA OF MOTOR TORQUE - SPEED PROFILES WITH g = 35
a significant part of the electric losses [19] [20] [21].
In our study, we will solve these problems considering
Symbol Quantity Value
the case of a high-speed Permanent Magnet Synchronous
Machines (PMSM) thanks to their high energy efficiency Pmax Maximal Power 14.6kW
and power density [22] [23]. In [24], we have proposed a Temmax Maximal Torque 8.7N m
solution, already applied to the case of electric vehicles, where Temef f Effective Torque 3.5N m
the design optimization was performed analytically. However, Temav Average Torque 2.5N m
in this previous work, the armature reaction was assumed Nmax Maximal Speed 31377rpm
negligible. In that case, the iron losses, independent of the Nef f Effective Speed 13545rpm
stator current, lead to an optimal torque angle ψ is zero Nav Average Speed 10834rpm
whatever the operating point. In this paper, we have extended
our research adding the armature reaction in the process to
propose a complete and realistic design optimization. From profiles of the PMSM can be deduced when the gear box g is
the driving cycle, the optimal geometry, optimal flux densities given. The main characteristic of the torque-speed profile of
and optimal control parameters, minimizing both the losses the motor are shown in Tab. II with g = 35.
and the volume of the machine, will be calculated in a way
remaining deterministic. III. ANALYTICAL MODEL
This paper is organized as follows. The section III presents We assume the linearity of the magnetic behavior agree-
the analytical model on which the design method will be ment with naturally low flux densities observed in high-speed
based. In the section IV, the design methodology and the PMSM [26]. The study is limited to the first harmonic and
resulting optimal design of the surface mounted PMSM will be the losses in the magnets due to slot effects are neglected
discussed. At last, a reluctance network model (RN), presented [27]. The relative permeability of the permanent magnets is
in Section V, will be used in Section VI to test the optimal assumed equal to unity. Based on these hypotheses, a 1D
machine with its optimal control strategy. analytical model is established and detailed in [28]. We will
From the data given in the Tab. I (considering an urban only recall here the main points that will serve to present the
vehicle) and the speed profile (UDDS driving cycle) shown in design optimization presented in the next section.
the Fig. 1 (left), the required torque on the wheel is calculated
from the classical formula below (assuming no incline) [25]:
A. Analytical model
dv(t) 1
Fm (t) = M (t) + cx Sv 2 (t) + 9.81M f (1) The electromagnetic torque Tem (t), the copper losses Pc (t),
dt 2 the iron losses in the yoke PY mg (t) and in the teeth PT mg (t)
which is plotted in the same figure (right). The torque-speed are expressed as functions of the following variables : the
thickness (he + hpm ), the active length L and the pole

pair number p (cf. Fig. 2):
12
α = √ rs pLR (9)
2
144 kL p2 L
β = σ 2 r 2 − r 2 R2
(10)
π kf kw w s
√
6 2 µ0
u = (11)
π he + hpm
(1 + rw )rs2
γ1 = π LR2 (12)
p(1 − rw )
2
π (rw − rs2 )rs
γ2 = LR2 (13)
kt rw
Fig. 2. Geometric data of the machine. 3) Constraints: We consider here a constant thermal mode. In
this case, the maximum temperature rise ∆θmax depends
on the average losses. It can be written as [30]:
effective magnetomotive force FM M (t), the torque angle ψ(t)
(angle between the stator current and the back EMF) and the
magnitude of the magnet flux density in the air-gap Bf m . < Ptot (t) >
∆θmax = (14)
heq Sth
1) Electromagnetic torque: Assuming no saliency and no
slotting effect, the air-gap is constant and the machine has In this paper, the effect of mechanical losses will not be
no reluctance torque. There is only the synchronous torque discussed. Note that they can be taken into account in
which is expressed as follows: our method and lead to an over sizing of the machine as
discussed in [30]. Mechanical constraints are imposed by
Tem (t) = αBf m FM M (t)cos(ψ(t)) (2) limitations due to both the maximum stress and the first
natural frequency of the rotor (with the maximum operating
where FM M (t) is, during the cycle, proportional to the
speed assumed to be lower than the first critical speed)
number turns Ns and the profile of the effective current
[31]. The ratio L/Rr and the radius of the rotor Rr must
Is (t) :
therefore satisfy:
FM M (t) = Ns Is (t) (3)
L/Rr ≈ L/Rs ≤ 5 (15)
2) Loss calculation: Here, we consider the copper losses Vpmax
Rr ≤ (16)
Pc (t), the iron losses in the stator yoke PY mg (t) and the Ωmax
iron losses in the stator teeth PT mg (t) :
where Vpmax and Ωmax are respectively the maximum
Ptot (t) = Pc (t) + PY mg (t) + PT mg (t) (4) rotor peripheral speed and the maximum angulaire speed
of the machine.
The copper losses are proportional to the square of the
effective magnetomotive force such as:
B. 1D model validation
2
Pc (t) = βFM M (t) (5)
The 1D analytical model usually gives good results for
And the iron losses are the sum of hysteresis and eddy surface mounted PMSM. Fig. 3, 4 and 5 compare the results
current losses [29] as: between this model and a 2D finite element model (FEM).
For this comparison, we have considered a reference machine
PY mg (t) = γ1 kad [kec p2 Ω2M (t) + kh pΩM (t)]Brm (t)2 (6) whose design and parameters are shown in Fig. 6 and in
PT mg (t) = γ2 kad [kec p2 Ω2M (t) + kh pΩM (t)]Brm (t)2 (7) Tab. IV (case with ψopt (t)). The effective magnetomotive
force and the torque angle used with this reference machine
with Brm , the resulting air-gap flux density that can be are respectively FM M ef f = 384At and ψ = 41◦ . Fig. 3
written as follows: and 5 present the results with different variations around the
reference geometry. They show that the analytical model (AM)
2
Brm (t) = Bf2m + u2 FM
2
M (t) is sufficiently robust with respect to the geometry and control
−2uFM M (t)Bf m sin(ψ(t)) (8) variables. It can be noted that the main error on the torque is
due to the saturation effect. However, in the case of high-speed
The terms α, β, u, γ1 and γ2 are functions of the main machines, with naturally low flux densities, this constraint is
geometric parameters R, rs , rw , the magnetic air-gap not a problem.
the speed profile and the vehicle characteristics are given. We

will show that it is possible to solve this problem using a
deterministic way since the three following steps:
- Step 1: The control parameters FM M (t) and ψ(t) minimiz-
ing the total losses at a given working point is expressed
as a function of the torque, the speed and the remaining
optimization parameters.
- Step 2: The optimal magnitude of the radial airgap flux
density Bf m is obtained by minimizing the average total
losses considering the torque profile Tem (t) and the speed
profile Ω(t). Using the results from the first step, it is
possible to express this optimal flux density Bf m according
to the geometrical parameters R, rs , rw and the pole pair
Fig. 3. Torque for various FM M source, ψ and rs . number p.
- Step 3: The optimal parameters rs , rw and p are calculated to
minimize the average losses from step 2. The external radius
R is also deduced considering both the thermal constraint
and the volume minimization.
1) Step 1: From (2), (5), (6), (7) and (8), the total electrical
losses, Ptot (t) can be written as a function of the optimal
parameter ψ(t) as follows:
γ0 (t)β
Ptot (t) = B2
β + γ0 (t)u2 f m
T 2 (t) 1
+(β + γ0 (t)u2 ) em2
α Bf2m
1
+(β + γ0 (t)u2 ) 2
Fig. 4. Average flux densities in stator tooth for various FM M source, ψ Bf m
and rw . 2
Tem (t) γ0 (t)u 2
× tan(ψ(t)) − B (17)
α β + γ0 (t)u2 f m
IV. DESIGN METHODOLOGY
where γ0 (t) = (γ1 + γ2 )kad [kec p2 Ω2M (t) + kh pΩM (t)].
In this part, we will firstly show how the model presented
Then it can be demonstrated that the optimal value of ψ(t)
in the previous section can be used to solve our optimization
which minimizes Ptot (t) at a given working point can be
problem. In a second stage, this design methodology will be
expressed as:
applied considering a UDDS driving cycle, with two different
control strategies: with ψ(t) = 0 and with ψopt (t). !
αγ0 (t)u Bf2m
ψopt (t) = tan−1 (18)
A. Design methodology β + γ0 (t)u2 Tem (t)
The goal here is to minimize the energy losses over the And the optimal effective magnetomotive force FM M opt (t)
driving cycle as well as the volume of the machine when can be obtained at a given torque:
s 2 2
Tem (t) γ0 (t)uBf m
FM M opt (t) = + (19)
αBf m β + γ0 (t)u2
2) Step 2: Once the motor works with the ψopt (t) and
FM M opt (t), the average total losses over the cycle <
Ptot (t) > can be deduced and given by:
Z
2 1 γ0 (t)β
< Ptot (t) >= Bf m dt
T β + γ0 (t)u2
T 2 (t)
Z
1 1
+ 2 (β + γ0 (t)u2 ) em2 dt (20)
Bf m T α
This equation has a form f = x2 A + (1/x2 )B with two
Fig. 5. Average flux densities in stator yoke for various FM M source, ψ positive constants A and B.√It can be demonstrated that
and rw . f is minimal and equal to 2 AB when x2 A = (1/x2 )B
or x = (B/A)1/4 . As a consequence, the optimal value of

Bf m , noted Bf mopt , which minimize < Ptot (t) > can be
expressed as follows:
 4
1
1 T 2 (t)
γ0 (t)u2 ) em
R
T (β + α2 dt 
Bf mopt =  γ0 (t)β
(21)
1
R
T β+γ0 (t)u2 dt
3) Step 3: Finally, by substituting Bf mopt obtained previously

in (21), we can express < Ptot (t) > as follows:
s Z
1 T 2 (t)
< Ptot (t) >= 2 × (β + γ0 (t)u2 ) em2 dt
T α Fig. 6. Optimal design with ψopt (t), ls = 2.5mm, ht1 = 1mm, ht2 =
s Z 1mm .
1 γ0 (t)β
× dt (22)
T β + γ0 (t)u2
In the specific case where the armature reaction is negli-
gible (u = 0), (22) can be solved analytically. It appears
then that the sizing data of the cycle are the effective (rms)
torque Temef f , the average speed Ωav and the effective
speed Ωef f . The volume and the optimal variables (rw , rs ,
and p) can be expressed analytically [24]. In the general
case (u 6= 0), the use of a numerical algorithm is necessary.
However, the computation time remains reduced compared
to the classical method when all the points of the cycle are
considered without reduction of the model or discretization
of the model. The minimization of both the losses and
the volume leads to a Pareto-optimal front in which all Fig. 7. Time evolution of iron losses of the motor optimal with ψopt (t)
and with ψ(t) = 0.
the optimal machines have the same parameters: p, rs , and
rw . Among all the solutions, only one satisfy the thermal
constraint allowing to find Ropt .
ψ(t) = 0 (59.5W atts for ψopt (t) against 67.8W atts for the
ψ(t) equal to 0).
B. Optimal Designed Motor Fig. 8 shows the map of the voltage-current operating points
In this part, we will apply the design methodology presented of these optimal motors. We can see that the flux-weakening
above, from step 1 to step 3, considering the two following leads to a lower magnitude of supply voltage reducing the
different cases: constraint on the power converter (220V with optimal ψopt (t)
against 325V with ψ(t) = 0). Here, the number of turns
- In the first case, the torque angle ψ(t) is not optimzed
of the winding, not optimized in this work, is set to 45.
and set to zero for all working points of the driving cycle.
The optimization of this parameter will be made in a next
This control strategy is usually chosen for the motor in the
paper in which the power converter and its loss model will be
constant torque area.
considered.
- In the second case, the torque angle is optimized. The motor
then works in flux-weakening regime at all working point
of the cycle.
For these two case, the constant data used in the optimiza-
tion process are given in the Tab. III. After step 1 and 2,
a numerical optimization algorithm is used to minimize the
formula (22). Its gives the optimal values of the motor as
presented in the Tab. IV.
We can see that the motor optimized with ψopt (t) is smaller
(7%) than the motor optimized with ψ = 0. It is due to the
fact that, with ψopt (t), the motor works permanently in the
flux-weakening regime that produces lower flux densities and
lower iron losses. The variations of iron losses with time of
the optimal motors, for the two cases, are plotted in Fig. 7.
The motor with ψopt (t) consumes less than 12% of average Fig. 8. Current-Voltage plane of the optimal motor with ψopt (t) (left) and
losses < Ptot (t) > compared to the motor controlled with with ψ(t) = 0 (right).
TABLE III The proposed method is applied here to a non-salient ma-

C ONSTANT DATA [30] chine. Note however that, for a salient machine, the different
steps of the optimization remain valid with the appropriate
Quantity Value changes in the analytical model.
h 10W/m2 K
∆θmax 70◦ C V. R ELUCTANCE N ETWORK M ODEL
kad 3 (0.3mm Si.Fe)
kec 6.5 × 10−3 (0.3mm Si.Fe) In the following sections, the optimal solution will be
kh 15 (0.3mm Si.Fe) validated considering the driving cycle. For this purpose, the
kf 0.3 performance of the designed machine needs to be evaluated
kt 0.5 considering all the operating points (1370 points for the
kL 1.2 UDDS). Such a validation cannot be easily performed using
kw 0.9 a 2D FEM because of an important amount of the calculation
time. We have therefore developed a reluctance network (RN)
model of the motor. This approach indeed allows an acceptable
TABLE IV calculation time while keeping a good accuracy [33] [34] [35]
O PTIMAL GEOMETRY DATA AND O PTIMAL PERFORMANCES
[36].
Quantity ψ(t) = 0 ψopt (t)
popt 1 1
A. Reluctance network
rsopt 0.3 0.43 The iron parts of the machine can be split into elements
rwopt 0.65 0.8 in a simple way via the insight of flux paths. Using linear
R 72mm 67mm (−7%) magnetic material, these elements can be modeled by constant
Bf mopt 0.46T 0.39T reluctances. Figs. 9 and 10 show the RNs of the stator and
< Pc (t) > 29.8W 34.5W rotor. In these figures, <to , <yo , <le and <r are respectively
< Pmg (t) > 38W 25W the reluctance of stator tooth, stator yoke, stator tooth flux
< Ptot (t) > 67.8W 59.5W (−12%) leakage and rotor yoke. These values can be determined as
follows (cf. Fig. 2):
TABLE V l
1 ht1 1 ht2 ln( lt2 )
MOTOR DATA <to = + t1
µ0 µr Llt1 µ0 µr L(lt2 − lt1 )

lt2 l
Symbol Quantity Value 2 ht3 ln( lt3 ) 2 ht3 ln( lt3
t4
)
+ + (23)
µ0 µr L(lt2 − lt3 ) µ0 µr L(lt3 − lt4 )
q Number of phases 3
Qs Number of stator slots 18
1 lys
<yo = (24)
Ns Number of stator winding turns 45 µ0 µr Lhyo
ha Air gap length 2.5mm 1 ls
<le = (25)
hpm Permanent magnet length 2.5mm µ0 Lht1
ls Opening slot width 2.5mm 1 lyr
ht1 Tooth tip thickness 1mm
<r = (26)
µ0 µr Lhyr
ht2 Tooth isthmus thickness 1mm
F M MP M and F M MSi are respectively magnetomo-
tive forces of permanent magnets and armature currents.
F M MP M is defined as:
Note that owing to the assumptions (no losses in the rotor),
the optimization doesn’t depend on geometric parameters of F M MP M = hpm Hc (27)
the rotor which are not present in (2)-(13). As a consequence,
the choice of the magnets, their dimensions and the airgap where Hc is the coercive field of the permanent magnets.
length can be done separately. hpm and ha have been chosen The distribution of F M MSi is calculated using the method
equal to 2.5mm. The value of the gear box ratio g has presented in [37].
been chosen here equal to 35. As shown in [30] [32], g can The configuration of the RN in the air-gap is shown in the
be optimized considering mechanical losses. It will not be Fig. 11. In this figure, the rotor node N Rj is connected to the
discussed in this paper. The choice of the winding and the node N Si on the stator through an air-gap reluctance. For the
choice of its parameters (number of slots, slot opening,...) studied motor, 18 × 18 air-gap reluctances are used (18 slots).
can be done in a final step in order to minimize harmonics These air-gap reluctances depend on the position of the rotor
for example. In this paper, this work is not discussed and the θr . Some values of an air-gap reluctance for various positions
winding is chosen with the parameters summarized in the Tab. of the rotor can be evaluated using the Tooth Contour Method
V. (TCM) which is based on a finite element calculation [38].
Fig. 9. Stator reluctance network. Fig. 12. Flux density at no load.
Fig. 10. Rotor reluctance network.

Fig. 13. Flux density at full load: FM M = 708(At), ψ = 18◦ .
Each reluctance can be then approximated by the following

analytical formula [24]: VI. VALIDATION OVER THE DRIVING CYCLE
<max − <min
<(θr ) = <max − (28) In this section, the performance of the motor optimized with
Pm r 2mβm
1 + 1 λθm ψopt (t) in section IV is evaluated at each point of the driving
cycle using the developed RN model. The assumption on the
where <max , <min , βm , γm and m can be chosen to best fit
linearity of the magnetic material is verified and the average
the simulation data given by the TCM.
value of energy losses calculated over the cycle is compared
with analytical results.
B. RN Validation by FEM In Fig. 14, the magnitude of the flux densities calculated
Simulations with 2D finite elements were used to validate with the RN, in the yoke and in the teeth, are plotted. As
the RN model. The flux densities in different parts of the motor it can be shown, the magnetic flux density during the cycle
obtained by the two models are compared at no-load and full is always lower than the saturation level which is 1.3T for
load regimes. Figs. 12 and 13 plot the flux densities in the air M 270 − 35A iron.
gap, rotor teeth and stator yoke. It shows that the developed Figs. 15 and 16 present the iron losses in the yoke and the
RN model offers a good accordance with FE. At no-load and teeth, determined by the analytical model and the reluctance
full load, the difference at the maximal value of flux density
is about 5%.
Fig. 11. Reluctance network in the airgap. Fig. 14. Maximum flux density in yokes and teeth over the cycle.
TABLE VI to validate the optimal design at all points of the cycle, a

I RON LOSSES fast and reliable reluctance network model was developed.
This model allows an acceptable computation time (1.6h for
AM RN Variation (%) 1370 operating points of the UDDS cycle in the presented
< Pyoke (t) > 19.3 16.3 15% case) while a calculation by FEM would lead to a prohibitive
< Pteeth (t) > 5.6 4.6 18% time. The maximal flux density in different parts of the
< Pmg (t) > 24.9 20.9 16% motor calculated by the RN model at all points of the cycle
confirms our assumption on the linearity of ferromagnetic parts
of the motor. In this paper, the main geometric parameters
(p, L, R, rs and rw ) are optimized. For the others, such as
the rotor parameters and the winding distribution, a work can
be done in a final step to choose/optimize these parameters.
From the airgap flux density created by the magnet (Bf m ), it is
possible to calculate the size and shape of the magnet that give
the optimal value Bf mopt and minimize the volume of magnets
or/and minimize the space harmonics. From the effective and
maximum magnetomotive force (FM M opt ), it is possible to
optimize the number of slots/pole/phase and the numbers of
turns that minimize the space harmonics and the constraint on
the power converters. The methodology presented in section
Fig. 15. Iron losses in the yoke of the motor over the cycle. IV can be applied for any cycle and application. It should be
noted also that it would remain valid in the case of a salient
machine.
network. Their average values over the cycle are summarized
in Tab. VI. R EFERENCES
The total losses, (< Pc (t) + Pmg (t) >), calculated by the
[1] C. C. Chan, “The state of the art of electric, hybrid, and fuel cell
RN is 55.4W against 59.5W for the analytical model which vehicles,” Proc. IEEE, vol. 95, no. 4, pp. 704–718, Apr. 2007.
gives 6.7% of difference. [2] W. Q. Chu, Z. Q. Zhu, J. Zhang, X. Liu, D. A. Stone, and M. P. Foster,
“Investigation on operational envelops and efficiency maps of electrically
excited machines for electrical vehicle applications,” IEEE Transactions
VII. C ONCLUSION on Magnetics, vol. 51, no. 4, pp. 1–10, Apr. 2015.
[3] T.J.Barlows, S.Latham, I.S.McCrane, and P.G.Boutler, A reference book
A novel design methodology of PMSM is presented in of driving cycles for use in the measurement of road vehicle emissions.
order to solve in a deterministic an optimization problem Wokingham, Berkshire, UK: Transportation Research Laboratory (TRL),
that takes into consideration both torque and speed profiles. 2009.
[4] P. Lazari, J. Wang, and L. Chen, “A computationally efficient design
This optimization is based on a real and complete sizing technique for electric vehicle traction machines,” International Confer-
model of the machine. It is shown that contrary to the ence on Electrical Machines (ICEM), pp. 2596–2602, 2012.
conventional control strategy with a flux weakening above the [5] L. Chen, J. Wang, P. Lombard, P. Lazari, and V. Leconte, “Design
optimisation of permanent magnet assisted synchronous reluctance ma-
base speed only, for high speed motors, working in a flux- chines for electric vehicle applications,” in International Conference on
weakening regime whatever the working point has several Electrical Machines (ICEM), pp. 2647–2653, 2012.
advantages. Firstly, seen by the motor side, it leads to a [6] L. Chen, J. Wang, P. Lazari, and X. Chen, “Optimizations of a permanent
magnet machine targeting different driving cycles for electric vehi-
smaller volume and lower level of losses. Secondly, seen by cles,” in IEEE International Electric Machines and Drives Conference
the power converter side, a lower supply voltage is required (IEMDC), pp. 855–862, 2013.
that reduces the constraint on the power converter (220V with [7] P. Caillard, F. Gillon, M. Hecquet, S.-A. Randi, and N. Janiaud, “An
optimization methodology to pre design an electric vehicle powertrain,”
optimal ψopt (t) against 325V with ψ(t) set to zero). In order in Vehicle Power and Propulsion Conference (VPPC), 2014.
[8] E. Carraro, M. Morandin, and N. Bianchi, “Optimization of a Traction
PMASR Motor According To a Given Driving Cycle,” IEEE Trans-
portation Electrification Conference and Expo (ITEC), vol. 2, pp. 1–6,
2014.
[9] G. Krebs, E. De Cecco, and C. Marchand, “Design approach of an axial
flux motor for electrical powertrain vehicle,” in International Conference
on Electrical Machines (ICEM), pp. 2812–2817, 2012.
[10] V. Ruuskanen, J. Nerg, J. Pyrhönen, S. Ruotsalainen, and R. Kennel,
“Drive Cycle Analysis of a Permanent-Magnet Traction Motor Based
on Magnetostatic Finite-Element Analysis,” IEEE Transactions on Ve-
hicular Technology, vol. 64, no. 3, pp. 1249–1254, 2015.
[11] P. H. Nguyen, E. Hoang, M. Gabsi, L. Kobylanski, and D. Condamin,
“Permanent magnet synchronous machines: Performances during driving
cycles for a hybrid electric vehicle application,” IEEE International
Symposium on Industrial Electronics (ISIE), pp. 1432–1438, 2010.
[12] P. H. Mellor, R. Wrobel, and D. Holliday, “A computationally efficient
iron loss model for brushless AC machines that caters for rated flux and
field weakened operation,” IEEE International Electric Machines and
Fig. 16. Iron losses in the teeth of the motor over the cycle. Drives Conference (IEMDC), pp. 490–494, 2009.
[13] V. Ruuskanen, J. Nerg, M. Rilla, and J. Pyrhönen, “Iron Loss Analysis of [33] G. Y. Sizov, A. Sayed-Ahmed, C. C. Yeh, and N. A. O. Demerdash,
the Permanent-Magnet Synchronous Machine Based on Finite-Element “Analysis and diagnostics of adjacent and nonadjacent broken-rotor-
Analysis Over the Electrical Vehicle Drive Cycle,” IEEE Transactions bar faults in squirrel-cage induction machines,” IEEE Transactions on
on Industrial Electronics, vol. 63, no. 7, pp. 4129–4136, 2016. Industrial Electronics, vol. 56, no. 11, pp. 4627–4641, Nov. 2009.
[14] F. Parasiliti, M. Villani, S. Lucidi, and F. Rinaldi, “Finite-element-based [34] N. Bracikowski, M. Hecquet, P. Brochet, and S. V. Shirinskii, “Multi-
multiobjective design optimization procedure of interior permanent physics modeling of a permanent magnet synchronous machine by using
magnet synchronous motors for wide constant-power region operation,” lumped models,” IEEE Transactions on Industrial Electronics, vol. 59,
IEEE Transactions on Industrial Electronics, vol. 59, no. 6, pp. 2503– no. 6, pp. 2426–2437, Jun. 2012.
2514, Jun. 2012. [35] M. F. Hsieh and Y. C. Hsu, “A generalized magnetic circuit modeling
[15] L. R. Huang, Z. Q. Zhu, and W. Q. Chu, “Optimization of electrically approach for design of surface permanent-magnet machines,” IEEE
excited synchronous machine for electrical vehicle applications,” in Transactions on Industrial Electronics, vol. 59, no. 2, pp. 779–792, Feb.
8th IET International Conference on Power Electronics, Machines and 2012.
Drives (PEMD 2016), pp. 1–6, Apr. 2016. [36] G. Liu, S. Jiang, W. Zhao, and Q. Chen, “Modular reluctance network
[16] W. L. Soong and T. J. E. Miller, “Theoretical limitations to the field- simulation of a linear permanent magnet vernier machine using new
weakening performance of the five classes of brushless synchronous mesh generation methods,” IEEE Transactions on Industrial Electronics,
ac motor drive,” in 1993 Sixth International Conference on Electrical vol. PP, no. 99, pp. 1–1, 2017.
Machines and Drives, pp. 127–132, Sep. 1993. [37] H. Roisse, M. Hecquet, and P. Brochet, “Simulations of Synchronous
[17] S. Morimoto, M. Sanada, and Y. Takeda, “Wide-speed operation of Machines Using a Electric-Magnetic Coupled Network Model,” IEEE
interior permanent magnet synchronous motors with high-performance Transactions on Magnetics, vol. 34, no. 5, 1998.
current regulator,” IEEE Transactions on Industry Applications, vol. 30, [38] E. Ilhan, J. Paulides, L. Encica, and E. Lomonova, “Tooth Contour
no. 4, pp. 920–926, Jul. 1994. Method Implementation for the Flux-Switching PM Machines,” XIX
[18] B. Cheng and T. R. Tesch, “Torque feedforward control technique International Conference on Electrical Machines (ICEM), pp. 1–6, 2010.
for permanent-magnet synchronous motors,” IEEE Transactions on
Industrial Electronics, vol. 57, no. 3, pp. 969–974, Mar. 2010.
Linh Dang was born in 1987 in Vietnam. She
[19] S. Morimoto, Y. Tong, Y. Takeda, and T. Hirasa, “Loss minimization
received the Engineering degree in electrical
control of permanent magnet synchronous motor drives,” IEEE Trans-
engineering from the Institut Polytechnique de
actions on Industrial Electronics, vol. 41, no. 5, pp. 511–517, Oct. 1994.
Hanoi, Hanoi, Vietnam, in 2010, and the M.Sc.
[20] A. Fonseca, C. Chillet, E. Atienza, A. L. Bui-Van, and J. Bigeon, “New
degree in electrical engineering from Ecole Poly-
modelling methodology for different pm motors for electric and hybrid
technique de Nantes, Nantes, France, in 2014.
vehicles,” in IEMDC 2001. IEEE International Electric Machines and
She is currently working toward her doctoral the-
Drives Conference, pp. 173–178, 2001.
sis at the Institut de Recherche en Energie Elec-
[21] A. Frias, A. Kedous-Lebouc, C. Chillet, L. Albert, L. Calegari, and
trique de Nantes-Atlantique Laboratory, Saint-
O. Messal, “Loss minimization of an electrical vehicle machine con-
Nazaire, France, on the design methodology
sidering its control and iron losses,” IEEE Transactions on Magnetics,
of high speed permanent magnet synchronous
vol. 52, no. 5, pp. 1–4, May. 2016.
machine for electric vehicles taking into account the driving cycle.
[22] C. Zwyssig, S. D. Round, and J. W. Kolar, “An ultrahigh-speed,
low power electrical drive system,” IEEE Transactions on Industrial
Electronics, vol. 55, no. 2, pp. 577–585, Feb. 2008. Nicolas Bernard received the M.S. degree in
[23] K. T. Chau, C. C. Chan, and C. Liu, “Overview of permanent-magnet electrical engineering from the University of
brushless drives for electric and hybrid electric vehicles,” IEEE Trans- Poitiers, France, in 1995, followed by the Agre-
actions on Industrial Electronics, vol. 55, no. 6, pp. 2246–2257, Jun. gation Instructors Credential in 1997 and the
2008. Ph.D. degree from Ecole Normale Superieure
[24] L. Dang, N. Bernard, N. Bracikowski, and G. Berthiau, “Analytical de Cachan, Cachan, France, in 2001. Since
model and reluctance network for high-speed pmsm design optimization 2004, he has been an Assistant Professor with
application to electric vehicles,” in 2016 XXII International Conference the Institut de Recherche en Energie Electrique
on Electrical Machines (ICEM), pp. 1359–1365, Sep. 2016. de Nantes-Atlantique Laboratory, Saint-Nazaire,
[25] M. Ehsani, Y. Gao, S. E. Gay, and A. Emadi, Modern Electric, Hybrid France. His research interests include flywheel
Electric, and Fuel Cell Vehicles. Boca Raton, FL: CRC Press, 2005. energy storage; design, modeling, and optimiza-
[26] R. Missoum, N. Bernard, M. E. Zaim, and J. Bonnefous, “Optimiza- tion of new topologies for electromagnetic actuators; high-speeds ma-
tion of high speed surface mounted permanent magnet synchronous chines.
machines,” in 2007 International Aegean Conference on Electrical
Machines and Power Electronics, pp. 446–451, Sep. 2007. Nicolas Bracikowski received the B.Sc. and
[27] F. Martin, M. E.-h. Zaı̈m, A. Tounzi, and N. Bernard, “Improved M.Sc. degrees in electrical engineering from
Analytical Determination of Eddy Current Losses in Surface Mounted the University of Artois (France). He received
Permanent Magnets of Synchronous Machine,” IEEE Transactions on his Ph.D. degree from Ecole Centrale de Lille
Magnetics, vol. 50, no. 6, 2014. (France), in 2012. Since 2013, he is an Asso-
[28] N. Bernard, R. Missoum, L. Dang, N. Bekka, H. B. Ahmed, and ciate Professor at University of Nantes (France),
M. E. H. Zaim, “Design methodology for high-speed permanent magnet where he carries out research in IREENA Lab-
synchronous machines,” IEEE Transactions on Energy Conversion, oratory (Institut de Recherche en Energie Elec-
vol. 31, no. 2, pp. 477–485, 2016. trique de Nantes Atlantique). His area research
[29] Z. Gmyrek, A. Boglietti, and A. Cavagnino, “Estimation of iron losses include: optimal design of electrical machine,
in induction motors: Calculation method, results, and analysis,” IEEE lumped models and vibro-acoustic modeling.
Transactions on Industrial Electronics, vol. 57, no. 1, pp. 161–171, Jan.
2010. Gérard Berthiau is Doctor of Ecole Centrale
[30] N. Bernard, L. Dang, J. C. Olivier, N. Bracikowski, G. Wasselynck, and Paris. He worked during 15 years at the Atomic
G. Berthiau, “Design Optimization of High-Speed Pmsm for Electric Energy Commission in Saclay before he joined
Vehicles,” in Vehicle Power and Propulsion Conference (VPPC), 2015. the University of Nantes in 2001 as Full-
[31] A. Borisavljevic, H. Polinder, and J. A. Ferreira, “On the speed lim- Professor. Today, he manages the team of Elec-
its of permanent-magnet machines,” IEEE Transactions on Industrial tromagnetic Device Modeling of the laboratory
Electronics, vol. 57, no. 1, pp. 220–227, Jan. 2010. IREENA. His research domains are Electromag-
[32] N. Bernard, F. Martin, and M. E.-h. Zaim, “Design Methodology of netic Field Computation, Electromagnetic Non
a Permanent Magnet Synchronous Machine for a Screwdriver Appli- Destructive Evaluation and Global Optimization
cation,” IEEE Transactions on Energy Conversion, vol. 27, no. 3, pp. Heuristics.
624–633, 2012.

Design Optimization With Flux-Weakening of

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Design Optimization With Flux-Weakening of

Uploaded by

Copyright:

Available Formats

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

Design optimization with flux-weakening of

is reduced to some points in which the energy consumption of

thickness (he + hpm ), the active length L and the pole

the speed profile and the vehicle characteristics are given. We

or x = (B/A)1/4 . As a consequence, the optimal value of

3) Step 3: Finally, by substituting Bf mopt obtained previously

TABLE III The proposed method is applied here to a non-salient ma-

Quantity ψ(t) = 0 ψopt (t)

µ0 µr Llt1 µ0 µr L(lt2 − lt1 )

Fig. 9. Stator reluctance network. Fig. 12. Flux density at no load.

Fig. 10. Rotor reluctance network.

Each reluctance can be then approximated by the following

TABLE VI to validate the optimal design at all points of the cycle, a

You might also like