Design Fundamentals of Electrical Machines and Drive Systems

Chapter 2
Design Fundamentals of Electrical

Machines and Drive Systems
Abstract This chapter presents a brief summary of the design fundamentals

including the analysis models and methods for electrical machines and drive sys-
tems, based on our design experiences, particularly for permanent magnet electrical
machine with soft magnetic composite cores. Because of the multi-disciplinary
nature, these design models and methods will be investigated at the disciplinary
level, including electromagnetic, thermal, mechanical, power electronics, and
control algorithm designs. Several design examples will be presented to illustrate
the corresponding design models and methods based on our research findings, such
as the finite element model for design analysis of motors, and the model predictive
control algorithm and its improvement form for the drive systems. These models
and algorithms will be employed in the design optimization of electrical machines
and drive systems in the following chapters.

Keywords Electrical drive systems Electromagnetic design Thermal design

Mechanical design Power electronics design Control algorithms Finite element

model Model predictive control
2.1 Introduction
2.1.1 Framework of Multi-disciplinary Design
Figure 2.1 illustrates a general framework of multi-disciplinary design for electrical

machines and drive systems. As shown, three main components, i.e., motor, power
electronics and controller, have to be investigated when designing such electrical
drive systems [1, 2]. The main design procedure includes the following steps.
Firstly, define the specifications of the electrical machine and drive system
required by a given application, which include the steady state specifications, such
as the rated power, speed range, voltage, current, efficiency, power factor (in case of
AC machines), volume and cost, and dynamic performances, such as the maximum
overshoot, settling time, and stability.
© Springer-Verlag Berlin Heidelberg 2016 25

G. Lei et al., Multidisciplinary Design Optimization Methods
for Electrical Machines and Drive Systems, Power Systems,
DOI 10.1007/978-3-662-49271-0_2
26 2 Design Fundamentals of Electrical Machines and Drive Systems
Fig. 2.1 Multi-disciplinary design framework of electrical machines and drive systems
Secondly, select a type of the motor, power electronic converter, and control
algorithm from possible options. The motor options include permanent magnet
(PM) motors, induction machines, synchronous machines, DC machines, and swit-
ched reluctance machines. For servo drives, stepping motors and other types of servo
motors can be considered. In this step, different motor topologies have to be inves-
tigated as well. The power electronic converter options mainly include the different
topologies of AC/DC, DC/DC, and DC/AC converters. The controller design mainly
investigates the control strategies and algorithms, such as field oriented control
(FOC), direct torque control (DTC), and model predictive control (MPC).
Thirdly, based on the selected motor type, converter circuit, and control scheme,
various disciplinary-level analyses should be conducted to evaluate the performance
of the drive system. For example, the motor design analysis consists of mainly the
electromagnetic, thermal and mechanical analyses (the shaded boxes in the figure).
Coupled-field analyses may be required in the design process, such as
electromagnetic-thermal and electromagnetic-mechanical stress analyses.
In summary, the design of electrical machines and drive systems mainly consists
of the analyses of five coupled disciplines or domains: electromagnetic, thermal,
mechanical, power electronics, and controller designs. The following sections will
present the popular design analysis models and methods for each discipline.
2.1.2 Power Losses and Efficiency
Power losses and efficiency are two main issues in the design analysis of electrical
machines and drive systems. The power losses are mainly composed of the copper
loss, core loss, mechanical loss, and stray loss.
2.1 Introduction 27
(1) The copper loss or Ohmic loss: PCu ¼ I 2 R is the power dissipated in stator and
rotor windings due to the resistance of copper wire, where I is the winding
current and R the winding resistance. Normally the DC resistance is used in
the calculation. However, it should be noted that the winding resistance
depends on the operating conditions, i.e., temperature and frequency (due to
the skin effects). In case of the brushes and slip rings/commutator, the effect of
contact resistance is often accounted for by assuming a voltage drop of 2 V.
(2) The core loss is the power dissipated in a magnetic core due to the variation of
magnetic field. This occurs in the stator and/or rotor iron core of an electrical
machine subject to AC excitations. Practically, it can be measured by
open-circuit or no-load tests. When the magnetic material is under an alter-
nating sinusoidal flux excitation, the alternating core loss can be calculated by
Pa ¼ Cha fBh þ Cea ðfBÞ2 þ Caa ðfBÞ1:5 ð2:1Þ
where f is the excitation frequency, B the magnitude of sinusoidal magnetic

flux density, and Cha, Cea, Caa, and h are the alternating core loss coefficients.
In case of rotating electrical machines, the rotational core losses have to be
considered. Figure 2.2 plots the average core losses with alternating flux
density from 2 to 2,000 Hz and circular rotating flux density vectors from 5 to
1,000 Hz of a cubic soft magnetic composite (SMC) SOMALYTM 500 sample
[3]. These are the standard core loss data used to identify the core loss model
parameters. The circularly rotational core loss can be calculated by
Pr ¼ Phr þ Cer ðfBÞ2 þ Car ðfBÞ1:5 ð2:2Þ
where
" #
Phr 1=s 1=ð2 sÞ
¼ a1
f ða2 þ 1=sÞ2 þ a23 ½a2 þ 1=ð2 sÞ2 þ a23
and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
B
s¼1 1 ½1=ða22 þ a23 Þ
Bs
Bs is the saturation flux density, and Cer, Car, a1, a2 and a3 are the rotational
core loss coefficients.
When the material is under a two dimensional elliptically rotating B excitation,
the core loss can be computed by
Per ¼ RB Pr þ ð1 RB Þ2 Pa ð2:3Þ
Fig. 2.2 Average core losses

under a alternating and
b circular rotating magnetic
fluxes [3]
where RB ¼ Bmin =Bmaj is the axis ratio, Bmin and Bmaj are the magnitudes of
the minor and major axes of the ellipse, respectively, and Pr and Pa the
corresponding rotational and alternating core losses when B = Bmaj. More
details about the rotational core losses can be found in [3–9].
(3) The mechanical losses are the power losses caused by the friction (brushes,
slip rings/commutator, shaft and bearing), damping, windage, and cooling fan.
It can be approximately determined by no-load test. In design, empirical data
are used.
(4) The stray loss is the power loss caused by stray factors that are hard to
determine separately, such as the non-uniform current distribution in con-
ductors and additional core loss due to distorted magnetic flux distribution for
various reasons. Because it is usually difficult to determine accurately the stray
loss, estimations based on experimental tests and empirical judgment are
2.1 Introduction 29
acceptable. For most types of machines, this can be assumed to be 1 % of the

output power.
In most electrical machines, the stator and/or rotor cores subject to varying
magnetic fluxes are made of laminated silicon steels, which have low core loss, and
hence the major power loss is the copper loss. Depending on the type of machine,
the copper loss normally accounts for 80–90 % of the total loss.
Based on the above analysis, the efficiency of a machine can be calculated by
Pout Pin ðPCu þ PCore þ PMec þ PStray Þ

g¼ ¼ ð2:4Þ
Pin Pin
Typical values of full load efficiency for rotating machines are:

• 50 % or less for fractional horse power motors (a few W to a few hundreds of
W),
• 75–85 % for electrical machines of 1 kW to a few tens of kW,
• 85–95 % for electrical machines of 100 kW to 1 MW, and
• 95–98 % or above for electrical machines of 1 MW to a few hundreds of MW
(e.g. 98 % for 100 MVA turbo generator).
2.2 Electromagnetic Design
Since electrical machines are electromagnetic devices for transforming electrical

power at one voltage to another (transformers) or converting electric power into
mechanical power or vice versa (motors or generators) by the principle of elec-
tromagnetic induction, electromagnetic design is a fundamental design stage of
electrical machines and drive systems, and is usually based on the following three
kinds of analysis models: the analytical model, magnetic circuit model, and finite
element model (FEM) [10–20].
2.2.1 Analytical Model
Analytical model is generally used to calculate the performance indicators of

electrical machines, such as the output power, torque, and cogging torque. For
example, the power and sizing equations are the powerful ways to guide the design
of PM motors [10, 11]. By utilizing the current density in the sizing equation, some
basic internal relationships can be found among the main dimensions to maximize
the torque density.
Assuming the flux linkage of stator winding in a PM motor is sinusoidal, and
ignoring the winding resistance, the input power Pin can be expressed by
ZT ZT
m m 2p 2p m
Pin ¼ eðtÞiðtÞdt ¼ Em sin t Im sin t dt ¼ Em Im ð2:5Þ
T T T T 2
0 0
where m is the number of phases, Em the peak value of back electromotive force
(EMF), Im the peak value of phase current, and T the electrical time period. The
output torque can be calculated by
Pout m
Tout ¼ ¼ g pkp Ksf As Jm ð2:6Þ
xr 2
where η is the efficiency, p the number of pole pairs, λp the peak value of PM flux
linkage, ωr the mechanical rotary speed, Ksf the slot fill factor, As the slot area, and
Jm the peak of current density. For different kinds of PM motors, λp and As are
related differently to their dimensions [21–23].
2.2.2 Magnetic Circuit Model
The magnetic circuit model acts as a uniform principle in descriptive magneto-

statics, and as an approximate computational aid in electrical machine design. The
model uses the conception of magnetic reluctance to establish an equivalent circuit
for approximate analysis of static magnetic field in electrical machines [24]. To
illustrate this model, a PM transverse flux machine (TFM) designed by SMC
material is investigated.
The SMC material is a relatively new soft magnetic material that has many
advantages over the conventional silicon steel sheets. The main advantages of SMC
material are the magnetic and mechanical isotropy and low cost, high productivity,
and high quality manufacturing capability of complex electromagnetic components
by the matured powder metallurgical molding technology, which will enable low
cost high productivity commercial manufacturing of SMC motors for a great variety
of electrical appliances [24–32].
In our previous work, a 3D flux PM TFM with SMC stator core was developed.
Figure 2.3 shows a photo of the PM-SMC TFM prototype. This machine was
initially designed to deliver an output power of 640 W at 1800 rev/min. It has 20
poles in the external PM rotor, i.e., 120 PMs in the rotor and 60 SMC teeth in the
stator. The stator core is made of SMC SOMALOYTM 500. The operating fre-
quency of this motor is 300 Hz at 1800 rev/min. Table 2.1 tabulates the main design
dimensions for this TFM [24, 25].
In order to briefly predict the performance of this TFM, a sketchy magnetic circuit
model as shown in Fig. 2.4 can be used. Figure 2.4 illustrates the main flux circuit
model and flux path of this TFM, where the resistances represent the magnetic
reluctances and the current source (PhiPM) stands for the magneto-motive forces
(mmfs) of PMs, and thus Rry represents the magnetic reluctance of rotor, Rm the
2.2 Electromagnetic Design 31
(a)
(b)
Fig. 2.3 Photo of the PM-SMC TFM prototype, a PM rotor, and b 3 stack SMC stator
Table 2.1 Main design Par. Description Unit Value

dimensions of PM-SMC TFM
– Number of phases – 3
– Number of poles – 20
– Number of stator teeth – 60
– Number of magnets – 120
– Stator outer radius mm 40
– Effective stator axial length mm 93
x1 PM circumferential angle degree 12
x2 PM width mm 9
x3 SMC tooth circumferential width mm 9
x4 SMC tooth axial width mm 8
x5 SMC tooth radial height mm 10.5
x6 Number of turns – 125
x7 Diameter of copper wire mm 1.25
x8 Air gap length mm 1.0
magnetic reluctance of PM, Rg the magnetic reluctance of the air gap, Rst1 the
magnetic reluctance of the stator teeth, Rst2 and Rsy stand for the magnetic reluctance
of the stator yoke. By analyzing this model, the main magnetic flux can be calculated.
Meanwhile, the magnetic flux leakage is a serious problem in this TFM, thus it
should be considered in the magnetic circuit model. Several flux leakage models
can be constructed for this TFM. Figure 2.5 illustrates the main flux leakage model.
In this model, the adjacent PM in the one side of the machine is modeled, where
Rry1 represents the magnetic reluctance of rotor, Rg1 and Rg2 represent the magnetic
reluctance of the air gap, Rs1 stands for the magnetic reluctance of the stator.
With the computed flux linkage, the resultant magnetic flux density in the air gap
and the flux per turn of coil can be estimated. After calculation, the obtained flux
per turn of this PM-SMC TFM is 0.32 mWb, which is higher than the calculated
result (0.28 mWb) by using the FEM [24].
This model can be also used to evaluate the performance of the motor. Based on
the calculated magnetic flux of the motor, the flux linkage per phase equals the
(a)
(b)
Fig. 2.4 Main flux circuit and flux path of the PM-SMC TFM, a magnetic circuit model, b flux
path in 2D plane
(a)
(b)
Fig. 2.5 Flux leakage circuit and path of the PM-SMC TFM, a magnetic circuit model, b leakage
path in 2D plane
number of coil turns multiplied by the magnetic flux of each coil turn, and it can be
computed as
kPM ¼ kl Ncoil pUgap ð2:7Þ
where λPM is the PM flux linkage per phase, kl the leakage coefficient, Ncoil the
number of turns of the phase winding, p the number of pole pairs, and Φgap is flux
per coil turn. The back EMF can be expressed as
Em ¼ xe kPM ¼ pxm kPM ð2:8Þ
where ωe = pωm is the electrical angular frequency, and ωm the mechanical angular
speed. The electromagnetic torque Tem can be expressed as
pffiffiffi
Pem 2
Tem ¼ ¼ mpkPM Im ð2:9Þ
xm 2
After the calculation, the no-load back EMF is 53.26 V at the rated speed of
1800 rev/min. According to (2.9), the electromagnetic torque is 4.66 Nm at the
rated current of 5.5 A (RMS value). Compared to the electromagnetic torque
obtained from FEM, i.e., 4.08 Nm, the relative error is about 0.58/4.08 = 14.21 %.
2.2.3 Finite Element Model
FEM is a widely used analysis model for field analysis in electrical machines as
well as other electromagnetic devices. The theory of FEM can be found in many
books and research papers. The PM-SMC TFM investigated above will be
employed as an example to show the application of FEM for designing electrical
machines.
When analyzing the magnetic field distribution, we used field analysis software
package ANSYS, and taking advantage of the periodical symmetry, we only need
to analyze one pole-pair region of the machine, as shown in Fig. 2.6a. At the two
radial boundary planes, the magnetic scalar potential obeys the periodical boundary
conditions:
um ðr; Dh; zÞ ¼ um ðr; Dh; zÞ ð2:10Þ
Fig. 2.6 a One pole pitch of FEM solution region for one phase (stack), and b magnetic field
distribution under no-load
Table 2.2 Key PM-SMC Parameter Unit Calculated Measured

TFM parameters
Motor back EMF Vs 0.247 0.244
constant
Phase resistance Ω 0.310 0.305
Phase inductance mH 6.68 6.53
Maximal cogging torque Nm 0.339 0.320
where Dh ¼ 18 mechanical is the angle of one pole pitch. The origin of the
cylindrical coordinate is located at the center of the stack.
Figure 2.6b illustrates the magnetic field distribution under no-load. Based on
the FEM analysis, the calculated key motor parameters for this machine are listed in
Table 2.2. The measured parameters are also listed in the table to show the effec-
tiveness of the FEM method. As shown, the measured motor back EMF constant is
0.244 Vs, 1 % lower than the calculated value of 0.247 Vs. The calculated phase
resistance and inductance, and maximal cogging torque are 0.310 Ω, 6.68 mH and
0.339 Nm, respectively, which are very close to the measured values (0.305 Ω,
6.53 mH and 0.320 Nm). In summary, the estimated parameters calculated by the
FEM-based method are well aligned with the experimental results. Therefore, FEM
is better than magnetic circuit model, and it is reliable to be used for optimization of
the electromagnetic design of electrical machines.
Moreover, the output performance parameters, such as output power, torque and
efficiency, can be estimated with the calculated electromagnetic parameters men-
tioned above. In the estimation, the control method is assumed to maintain that the
d-axis component of current equals zero. Figure 2.7 shows the per phase equivalent
electric circuit of this motor under the assumed control method.
Based on this per phase equivalent electrical circuit, the main relationships of the
motor can be predicted by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Vin ¼ ðEa þ Ia Ra Þ2 þ ðxe La Ia Þ2 ð2:11Þ
Pin ¼ 3Vin Ia cos u ð2:12Þ
Pout ¼ Pin Pcore Pcopper Pmech ð2:13Þ
Pout
Tout ¼ ð2:14Þ
xr
where Vin is the input voltage, Ea the back EMF, Ia the armature current, ωe the
electric angular frequency, La the inductance, Ra the resistance, φ the angle between
Vin and Ea, Pin the input power, Pout the output power, Pcore the core loss, Pcopper
the copper loss, Pmech the mechanical loss, Tout the output torque, and ωr the
mechanical angular speed.
In motor with SMC cores, unlike the conventional motors made of silicon sheet
steels, the core loss can be a major part among all power losses, and the mechanical
Fig. 2.7 Per phase equivalent

electric circuit and phasor
diagrams of the motor
loss is generally considered as 1–1.5 % of the output power. In general, the core
loss prediction in the TFM should be calculated by using the FEM based on the
multi-frequency core loss characteristic of the material. More comparison results
can be seen in [24, 25].
2.3 Thermal Design
2.3.1 Thermal Limits in Electrical Machines
The rating of an electrical machine gives its working capability under the specified
electrical and environmental conditions. Major factors that determine the ratings are
thermal and mechanical considerations. To obtain an economic utilization of the
materials and safe operation of the motor, it is necessary to predict with reasonable
accuracy the temperature rise of the internal parts, especially in the coils and
magnets.
The temperature rise resulted from the power losses in an electrical machine
plays a key role in rating the power capacity of the machine, i.e., the amount of
power it can convert without being burnt for a specified length of life time. The life
expectancy of a large industrial electrical machine ranges from 10 to 50 years or
more. In an aircraft or electronic equipment, it can be of the order of a few thousand
hours, whereas in a military application, e.g. missile, it can be only a few minutes.
Table 2.3 Classification of electrical insulation materials

Class Maximum temperature rise Materials
(°C)
O 90 Paper, cotton, silk
A 105 Cellulose, phenolic resins
B 130 Mica, glass, asbestos with organic binder
F 155 Same as above with suitable binder
H 180 Mica, glass, asbestos with silicone binder, silicone
resin, Teflon
The operating temperature of a machine is closely associated with its life

expectancy because deterioration of the insulation is a function of both time and
temperature. Such deterioration is a chemical phenomenon involving slow oxida-
tion and brittle hardening, leading to loss of mechanical durability and dielectric
strength. In many cases the deterioration rate is such that the life of the insulation
can be expressed as
Life Time ¼ AeB=T ð2:15Þ
where A and B are constants and T is the absolute temperature. Roughly, it says that
for each 10 °C temperature rise exceeding the maximum allowable temperature rise,
the life time of insulation is halved.
Insulation materials used in electrical machines are classified by the maximum
allowable temperature rise that can be safely withstood. Table 2.3 lists the classi-
fication of electrical insulation materials by the IEC (International Electrotechnical
Commission).
Generally, there are two kinds of analysis models for thermal analysis in elec-
trical machines, namely the thermal network model and the FEM [14, 15, 20, 25].
The following sections will present examples for the two methods.
2.3.2 Thermal Network Model
Two design examples will be illustrated to show the usage of thermal network
model for the thermal analysis of PM-SMC motors. The first one is a TFM, and the
second is a high speed claw pole motor.
A. Transverse flux machine
In this study, the temperature rise was calculated by using a hybrid thermal network
model with distributed heat sources, as shown in Fig. 2.8.
For high computation accuracy, every part, e.g. the air gap, is divided into two or
more segments. The thermal resistances to heat conduction in the following sections
are calculated: rotor yoke (Rry), magnets (Rm), glue between magnets and rotor yoke
2.3 Thermal Design 37
Fig. 2.8 Thermal network model of the TFM prototype
(Rmg), air gap (Rag), stator yoke (RFe1), stator side discs (RFe2), stator teeth (RFe3),
varnished copper wire (Rcu), and insulations (RI1, RI2, RI3) between the winding and
the stator yoke, the stator wall disc, and the air gap, respectively. In addition, the
thermal resistances of the stator shaft (Rss), the aluminum end plates (Ral), and the
stationary air (Rsa) between the side stator discs and the end plates are calculated
separately [25].
The equivalent thermal resistances to the heat convection of the following
sections are calculated: that between the stator tooth surface and the inner air in the
air gap (RFeA), that between the winding and the inner air (RWA), that between the
magnet and the inner air (RmA), that between the rotor yoke and the inner air
(RryA1), and that between the rotor yoke and the outer air (RryA2).
The heat sources include the stator winding copper losses (Pcu), the stator and
rotor core losses (PFes, PFer), and the mechanical losses due to windage and friction
(Pmec). The improved method for core loss calculation can obtain the loss distri-
bution, which is a great advantage for thermal calculation by the hybrid thermal
model.
The temperature rises in the middle of several parts are calculated as 64.9 °C in
the stator winding, 78.6 °C in the stator core, 59.3 °C in the air gap, 36.1 °C in the
magnets, and 25.3 °C in the rotor yoke outer surface. The experimentally measured
results are 66 °C in the stator winding and 27 °C in the rotor yoke, and it can be
seen that the maximum relative error between the calculated and measured results is
only 3 %. Thus, it is reliable to use the thermal network method for design of this
TFM.
Fig. 2.9 Prototype of a high speed claw pole motor
B. High speed claw pole motor

In our previous work, a high speed claw pole motor as shown in Fig. 2.9 with an
SMC stator core was developed [3, 8, 9]. The major motor parameters are tabulated
in Table 2.4, and the structure (one pole pitch of one stack) is shown in Fig. 2.10.
Figure 2.11 illustrates the topology of one of the three stacks. The stator consists
of the claw poles, the yoke, and the phase winding. The rotor is simply made of a
ring PM magnetized in four poles and mounted on the rotor core. The three stator
stacks are shifted for 120o (electrical) apart from each other.
Table 2.4 Main dimensions Parameter Unit Value

and design parameters
Number of phases – 3
Rated power W 2000
Rated frequency Hz 666.7
Rated speed rev/min 20,000
Number of poles – 4
Stator outer diameter mm 78
Rotor outer diameter mm 29
Rotor inner diameter mm 18
Airgap length mm 1
Axial length mm 48
Stator core material – SOMALOYTM 500
PM material – NdFeB
Fig. 2.10 Magnetically

relevant parts of one stack of
three-phase claw pole motor
Fig. 2.11 Structure of a

high-speed claw pole motor
(one pole pitch of one stack)
In general, the geometrical complexity of an electrical machine requires a large

thermal network if a high resolution of temperature distribution is required. Instead
of using a whole model, the geometrical symmetries of the machine can be used to
reduce the size of the model. The distributed thermal properties have been lumped
together to form a small thermal network, representing the whole machine. For the
calculation of temperature distribution in the SMC motor, a thermal resistance
network, as shown in Fig. 2.12, is used. It has ten nodes (the outer air, frame, yoke,
and so on). Each node represents a specific part or region of the machine, and the
thermal resistances (Rn, n = 1,…,16) between the nodes include complex processes,
Fig. 2.12 Thermal network

of one stack of a high-speed
claw pole motor
such as the 2D and 3D heat flow, convection, internal heat generation, and varia-
tions in material properties. To account for the three dimensional heat flows at a
node, the thermal structure shown in Fig. 2.13 can be employed.
As shown in Fig. 2.13, the thermal resistances of an element are built in three
directions, and the heat source if any can be placed at the center point. In this
model, the thermal conduction equation can be expressed as
Tb Ta Tc Ta Td Ta Te Ta Tf Ta Tg Ta
þ þ þ þ þ þ qa
Rab Rac Rad Rae Raf Rag
0
@ðTa Ta Þ
¼ Ca ð2:16Þ
@t
where Ta, Tb, Tc, Td, Te, Tf, and Tg are the temperatures at nodes a, b, c, d, e, f, and
g, Rab, Rac, Rad, Rae, Raf, and Rag the thermal resistance between nodes a-b, a-c, a-d,
0
a-e, a-f, and a-g, respectively, qa is the heat source, Ca the heat specific, and Ta the
Fig. 2.13 Nodal thermal

structure for 3D heat flow
temperature of node a at the next time instant. The thermal resistance in Fig. 2.13
can be calculated by
DX
Rab ¼ Rac ¼ ð2:17Þ
2kx DYDZ
DY
Rad ¼ Rae ¼ ð2:18Þ
2ky DXDZ
DZ
Raf ¼ Rae ¼ ð2:19Þ
2kz DXDY
where λx, λy and λz are the thermal conductivities in the x, y and z directions,
respectively [8].
The calculation results at no load are 324.6 K in the frame, 326.3 K in the yoke,
330.8 K in the winding, 337.7 K in the claw poles, 334.4 K in the air gap, 331 K in
the magnets, and 324.7 K in the bearing.
2.3.3 Finite Element Model
In the thermal network, the core loss at each node cannot be obtained easily from
the magnetic field calculation. In most cases, the average value is used. Since the
core loss distribution is quite different in different positions of the stator core, 3D
FEM is used to analyze the temperature distribution in this section. Two design
examples investigated in the previous section will be illustrated to show the usage
of FEM for the thermal analysis of PM-SMC motors.
Fig. 2.14 Temperature distribution in the PM-SMC TFM obtained by 3D FEM
Figure 2.14 illustrates the temperature distribution of the PM-SMC TFM based
on FEM. As shown, the average temperature rises in the winding is 62.5 °C, which
is close to the measured value 65 °C.
Figure 2.15 depicts the distributions of core loss and temperature at full load in
the SMC core of the high speed claw pole motor. The temperature is measured by
an infrared temperature probe. At 20,000 rev/min and no load, the frame temper-
ature is 331.4 K and the stator yoke temperature is 333.5 K, respectively. The
Fig. 2.15 a Distributions of core loss, and b temperature in SMC core of the high speed claw pole
motor
measured temperatures are slightly higher than the FEM results, because the actual
loss is greater than the calculation. The FEM method is more accurate than the
thermal network method because there are only ten nodes in the network. The
advantage of the thermal network is the calculation speed, which is much faster than
the FEM method [8, 9].
2.4 Mechanical Design
Mechanical design is another important issue in the design analysis of electrical

machines, especially for high speed motors. Generally, the following three aspects
should be investigated for the mechanical design analysis:
(1) Mechanical structures and materials,
(2) Field of stress and material strength (including elastic and plastic deforma-
tions), and
(3) Modal analysis for vibration and noise.
The first and the second aspects are often noncritical and can be readily satisfied
through empirical design, whereas the third one requires special attention for most
situations, especially those operated at high frequencies. The modal analysis is
generally used to calculate the resonance frequency of the motor in operation.
Enough distance between this frequency and the electromagnetic frequency should
be designed for motors to avoid resonance. The modal analysis is generally con-
ducted by using the FEM. The two design examples used in the previous section
will be employed as follows.
The PM-SMC TFM is operated at 300 Hz, which is relatively high compared
with the conventional motors operated at the power frequency of either 50 or 60 Hz.
From experience, we only need to do the first-order modal analysis and compare the
frequency with the electromagnetic frequency [20]. Figure 2.16 illustrates the
first-order modal analysis for this motor of the electromagnetic optimized design. It
can be seen that the resonance frequency of the optimal motor is about 4,435 Hz,
which is much higher than the electromagnetic frequency of 300 Hz.
Regarding the high speed claw pole motor, because it is operated at high speed,
it is essential to carry out a modal analysis to find and adjust the resonant points, so
that in practical operation, these frequencies can be avoided. Figure 2.17 shows the
vibration patterns and the corresponding resonant frequencies of the rotor structure.
These frequencies are well above the operating frequency and therefore have almost
no influence to the practical operation. Through the analysis and adjustment, it was
found that the bearing stiffness, the shaft length, the shaft diameter and the position
of bearing have significant influence on the rotor natural frequency [3].
Fig. 2.16 Illustration of first order modal analysis for PM-SMC TFM
Fig. 2.17 Vibration patterns at a 4,102 Hz (Y axis), b 4,102 Hz (X axis), c 9,562 Hz (Z axis), and
d 10,321 Hz (Y axis)
2.5 Power Electronics Design 45
Fig. 2.18 A three-phase

inverter
2.5 Power Electronics Design
The design of power electronics for electrical machines and drive systems is also an
important and complex stage. Among many aspects in power electronics, the
converter/inverter and switching scheme are two main concerns in the design of
electrical machines and drive systems.
The converter/inverter is an important component to drive an electrical machine.
An inverter, for example, is an electronic apparatus that can convert a DC voltage to
an AC voltage of specified waveform, frequency, magnitude, and phase angle.
Among many different topologies, the three phase bridge power circuit as shown in
Fig. 2.18 has become favorite and standard for use in the control systems of
electrical machines. Many different topologies can be obtained from this structure
for different applications. For example, two extra switches can be added to establish
two bridges for the fault tolerant control scheme [33, 34].
For controlling the waveform, frequency, magnitude, and phase angle of the AC
voltage, many switching schemes can be used, such as square wave and sine wave
pulse width modulations (PWMs) and space vector modulation (SVM), as well as
hard and soft switching.
2.6 Control Algorithms Design
Control algorithms play an important role in the determination of dynamic and

state-state performances of electrical drive systems. Various control algorithms
have been developed and employed successfully in commercial drive systems, such
as the six-step control, FOC, DTC and MPC [35–39].
While FOC is commonly used in various high performance electrical drive

system, the merits of DTC are simple in structure (thus low cost), fast dynamic
response, and strong robustness against motor parameter variation [40–42]. The
major advantages that affect the commercial application of the conventional DTC
are large torque and flux ripples, variable switching frequency, and excessive
acoustic noises.
To overcome these problems, many methods have been proposed in the litera-
ture. One of them is to apply the technique of SVM to DTC, known as the
SVM-DTC. In the conventional DTC, the switching table only includes a limited
number of voltage vectors with fixed amplitudes and positions. The implementation
of SVM enables the generation of an arbitrary voltage vector with any amplitude
and position [43–48]. In this way, the SVM-DTC can generate the torque and flux
more accurately to eliminate the ripples. Another merit of SVM-DTC is that the
sampling frequency required is constant and lower than that of the conventional
DTC.
Recently, the MPC has attracted increasing attention in industry and academic
communities [49–54]. In the SVM-DTC, the power converter with modulation can
be considered as a gain in controller design. In the predictive control methods, the
discrete nature of power converters is taken into account by considering the con-
verter and the motor from a systemic viewpoint. There are various different versions
of predictive control algorithms, differing in the principle of vector selection,
number of the applied vectors and predictive horizon.
The conventional DTC and MPC are similar in that they both select only one
voltage vector in each sampling period. This can result in overregulation, leading to
large torque and flux ripples and acoustic noise.
As all the design examples used in this book are permanent magnet synchronous
machines (PMSMs), several control algorithms will be presented with details for
PMSMs in the following sections. Numerical and experimental examples will be
presented for some of them.
2.6.1 Six-Step Control
The six-step control method was oriented to drive brushless DC (BLDC) motors
with trapezoidal back EMF waveforms. In many applications, however, the
trapezoidal excitation is also used to drive PMSMs with sinusoidal back EMF
waveforms because the trapezoidal excitation or six-step method based drive is
robust and low cost [35].
In the six-step control scheme, the stationary reference frame is always used to
model the PMSM. The phase variables are used to express the machine equations as
they can account for the real waveforms of the back EMF and phase current.
Assuming that the resistances of three phase stator windings are equal, the three
phase voltage equations of the motor can be written as
2.6 Control Algorithms Design 47
2 3 2 32 3 82 32 39 2 3
d <4 aa
va Rs 0 0 ia L Lba Lca ia = ea
4 vb 5 ¼ 4 0 Rs 0 54 ib 5 þ Lba Lbb Lcb 54 ib 5 þ 4 eb 5
dt : ;
vc 0 0 Rs ic Lca Lcb Lcc ic ec
ð2:20Þ
where va, vb, and vc are the phase voltages, ia, ib, and ic the 2phase currents, e3a, eb,
Laa Lba Lca
and ec the phase back EMF, Rs is the phase resistance, and 4 Lba Lbb Lcb 5 the
Lca Lcb Lcc
inductance matrix, including both the self-and mutual-inductances.
Assuming further that the reluctance is independent of the rotor position, one can
obtain

La ¼ Lb ¼ Lc ¼ Ls
ð2:21Þ
Lab ¼ Lca ¼ Lbc ¼ M
As ia + ib + ic = 0 for a symmetric three phase system, the voltage equation can

be simplified as
2 3 2 32 3 82 32 39 2 3
va Rs 0 0 ia
d < LM 0 0 ia = ea
4 vb 5 ¼ 4 0 Rs 0 54 ib 5 þ 4 0 LM 0 54 ib 5 þ 4 eb 5
dt : ;
vc 0 0 Rs ic 0 0 LM ic ec
ð2:22Þ
Assuming linear system, the machine model in state space form can be expressed
as
2 3 2 382 3 2 32 3 2 39
i 1=ðL M Þ 0 0 < va Rs 0 0 ia ea =
d 4 a5 4 5 4 vb 5 4 0
ib ¼ 0 1=ðL M Þ 0 Rs 0 54 i b 5 4 eb 5
dt : ;
ic 0 0 1=ðL M Þ vc 0 0 Rs ic ec
ð2:23Þ
The generated electromagnetic torque is given by
Te ¼ ðea ia þ eb ib þ ec ic Þ=xm ð2:24Þ
where ωm is the mechanical angular speed of the rotor.

The mechanical equation of the machine is
dxm
Te ¼ J þ Fxm þ TL ð2:25Þ
dt
where J is the inertia of the machine rotating parts, F the friction coefficient, and TL
the load torque on the rotor shaft.
DC voltage source or
rectified from AC power
iDC
-
ω ref + Speed + Current PWM & Inverter
Controller Controller Commutation
ωr -
Speed θ Hall
Sensor PMSM
Calculation
Fig. 2.19 Block diagram of PMSM six-step drive system
Figure 2.19 shows the block diagram of six-step drive scheme. The drive system
is operated with the feedback information of rotor position, which is obtained at
fixed points, typically every 60 electrical degrees for commutation of the phase
currents.
The 120° conduction mode is applied to drive the PMSM. The voltage may be
applied to the motor every 120° (electrical), with a current limit to hold the phase
currents within the motor’s capabilities. Because the phase currents are excited in
synchronism with the back EMF, a constant torque is generated. A simulation
model is built in MATLAB/SIMULINK as shown in Fig. 2.20.
Fig. 2.20 Simulation block diagram of six-step controlled PMSM drive system
As shown in Fig. 2.20, the rotor position information comes from the Hall effect
sensors, which are integrated in the machine model in MATLAB/SIMULINK. The
resolution of the feedback signals is only 60° (electrical). Since most applications
require a stable speed, a speed feedback loop is employed. The rotor speed infor-
mation can be deduced from the low resolution Hall signals, which is marked as
Speed Calculation in Fig. 2.20. Typically, the average speed in one 60° section is
used as the speed feedback.
However, by using the average speed, there is always a lag when the motor
speed is not constant in accelerating or other dynamic state. To overcome this, the
rotor position can be expressed in Taylor’s series as the following:
ð2Þ
ð1Þ h2k
hðtÞ ¼ hk ðtÞ þ h1k ðt tk Þ þ ðt t k Þ2 þ ð2:26Þ
2!
ð1Þ
p=3
where tk is the last commutation time, h1k ¼ tk tk1
the average speed of last section,
ð1Þ ð1Þ
ð2Þ h h
and h2k ¼ 1ktk tk1
1ðk1Þ
the average acceleration of last section.
As shown above, with the higher order calculation, more accurate speed and
position information can be deduced, whereas the computing cost rises. As a
compromise, in some situations, the following equations are used to estimate the
rotor position and speed:
( ð2Þ
ð1Þ h2k
hðtÞ ¼ hk ðtÞ þ h1k ðt tk Þ þ 2! ðt tk Þ2 ð2:27Þ
ð1Þ ð2Þ
xðtÞ ¼ h1k ðtÞ þ h2k ðt tk Þ
2.6.2 Field Oriented Control
For a PMSM under sinusoidal excitations, the original voltage equations can be
expressed in the stationary reference frame as the following
2 3 2 3 2 3
va ia ka
4 vb 5 ¼ Rs 4 ib 5 þ d 4 kb 5 ð2:28Þ
dt
vc ic kc
where λa, λb, and λc are the flux linkages of phases a, b, and c, respectively.
Equation (2.28) represents a system of differential equations with time varying
(periodic) coefficients. For sinusoidally distributed windings, a Park-Clark trans-
formation can be used to transform the above equations to a system of differential
equations with constant coefficients, represented in a d-q coordinate frame attached
to the rotor. The reference frames are shown in Fig. 2.21.
The Park-Clark orthogonal transformation can be expressed in the matrix form
as
Fig. 2.21 Stationary and

rotating reference frames
2 3 rffiffiffi2 32 3
rd cos h cos h 23 p cos h þ 23 p ra
4 rq 5 ¼ 24 sin h sin h 23 p sin h þ 23 p 54 rb 5 ð2:29Þ
3 p1ffiffi p1ffiffi p1ffiffi
r0 2 2 2
rc
where θ is defined as the angle between two reference frames.

The subscripts d, q, and 0 in (2.29) represent some fictitious windings attached to
the rotor. The variables σd, σq, σ0, σa, σb, and σc may represent voltages, currents, or
flux linkages. As a result, the transformed set of electrical equations describing the
behavior of PMSM in the d-q rotating frame become
8
< vd ¼ Rs id þ dtd kd kq dh
dt
v ¼ Rs iq þ dtd kq þ kd dh ð2:30Þ
: q dt
v0 ¼ Rs i0 þ dtd k0
where vd, vq, and v0 are the phase voltages, id, iq, and i0 the phase currents, and λd,
λq, and λ0 the phase flux linkages.
For the linear PMSM model, the magnetic saturation saliency is not considered.
The flux linkages of the d- and q-axes can be further expressed as

kd ¼ Ld id þ km
ð2:31Þ
kq ¼ Lq iq
where Ld and Lq are the constant d- and q-axes inductances, respectively, and λm is
the flux linkage generated by the rotor PMs.
On the other hand, the voltage equation of the 0 axis in (2.30) is usually ignored
by assuming well-balanced three-phase windings for the controller design.
Therefore, the electrical voltage equations in the rotor reference frame can be
rewritten as
(
vd ¼ Rs id þ Ld didtd Lq iq dh
di
dt
ð2:32Þ
vq ¼ Rs iq þ Lq dtq þ ðLd id þ km Þ dh
dt
The torque expression after the application of the transformation becomes
3
Te ¼ p kd iq kq id
2 ð2:33Þ
3
¼ p km iq þ Ld Lq id iq
2
where p is the number of pole pairs.

By this transformation, the flux and torque control of the PMSM are decoupled.
The q-axis current, in the FOC method, is regulated to produce sufficient torque
while the d-axis current is controlled to modify the air-gap flux linkage. For normal
operation, the d-axis current is set to zero to achieve the maximum
torque-to-ampere ratio, and for the flux weakening control, the d-axis current is
modified to weaken the air-gap flux.
The reference speed value is the main input for the drive system, and the
electromagnetic torque and rotor speed are the output. Two feedback loops, current
or torque loop and speed loop, are added to provide desired performance. The
output of the speed controller will be the reference value for the q-axis current while
the d-axis current is set to zero. Both of the d- and q-axes currents are controlled to
generate the torque and achieve the maximum efficiency drive. Figure 2.22 shows
the implementation diagram of the typical FOC scheme, where the traditional PWM
method is applied for the variable speed drive by the vectorial variable voltage and
variable frequency control strategy.
ω ref + + vd va
PI PI va
ωr - - 1/ Park
vß
1/ Clark vb
PWM VSI
+ vq Transform Transform vc
PI
id_ref -
θ
id ia ia
iq Park iß Clarke
Transform ib
Transform
Position and Encoder
Speed PMSM
Estimation
Fig. 2.22 Block diagram of FOC scheme for PMSM drive

Fig. 2.23 Simulation block diagram of typical FOC based PMSM drive system
Similar to the six-step method, a simulation model of the FOC scheme based
PMSM drive is built in MATLAB/SIMULINK. The sinusoidal back EMF machine
model is selected from the SimPowerSystem tool box, in which the current sensors
and rotor position sensor are integrated. The Park and Clark transformations are
synthesized as one ‘abc_to_dq’ block to transfer the variables between the sta-
tionary and rotating reference frames, as shown in Fig. 2.23. Two discrete PI
controllers are used for the speed and current feedback loops.
The traditional triangulation PWM generation technique is applied. A triangular
carrier wave sampling signal is compared directly with a sinusoidal modulating
wave to determine the switching instants, and therefore the resultant pulse widths.
2.6.3 Direct Torque Control
In the DTC strategy, the flux linkage and torque are calculated in the two-phase
stator reference frame, i.e., the α-β frame, which is transformed from the
three-phase a-b-c reference frame by using the Clark transformation. The Clark
transformation can be expressed in the matrix form as

rffiffiffi" #2 3
1
p12ffiffi ra
ra 2 1 pffiffi2 4 rb 5
¼ ð2:34Þ
rb 3 0 3
23
2 rc
After the measured phase voltages and currents are transformed to the α-β frame,
the flux linkage components of the α- and β-axes can be calculated as
R
ka ¼ R ðva Rs ia Þdt
ð2:35Þ
kb ¼ vb Rs ib dt
λ ref + eλ Hysteresis
- controller
Switch
Inverter
table
ω ref + + eT Hysteresis
Controller
- - controller λ
ωr s
vdc
Flux & Torque ia
calculation ib
Encoder
1/s PMSM
Fig. 2.24 Block diagram of typical DTC scheme based PMSM drive
The torque observer can then be designed as
3 pm
Te ¼ ka ib kb ia ð2:36Þ
2 2
Figure 2.24 shows the block diagram of a typical DTC scheme for PMSM drive.
Two hysteresis controllers are applied to the flux linkage and torque control loops.
The calculated flux linkage is also sent to the switching table to identify the current
flux vector position.
From (2.35), the stator flux linkage is
Z
ks ¼ ðvs Rs is Þdt ð2:37Þ
where vs and is are the stator voltage and current spatial vectors, respectively.
In the case of a PMSM, λs always varies even when the zero voltage vectors are
applied because of the rotating rotor magnets, and thus, zero voltage vectors are not
used for DTC driven PMSM. λs should always be in motion with respect to the rotor
flux.
According to (2.36), the electromagnetic torque can be controlled effectively by
controlling the amplitude and rotating speed of λs. For counter-clockwise operation,
if the actual torque is smaller than the reference, the voltage vectors that keep λs
rotating in the same direction are selected. The angle increases as fast as it can, and
the actual torque increases as well. Once the actual torque is greater than the
reference, the voltage vectors that keep λs rotating in the reverse direction are
selected instead of the zero voltage vectors. The angle decreases, so does the torque.
By selecting the voltage vectors in this way, λs will rotate all the time in the
Table 2.5 Switching table of typical DTC scheme for PMSM drive
Δeλ ΔeT θ
θ1 θ2 θ3 θ4 θ5 θ6
1 1 V2(110) V3(010) V4(011) V5(001) V6(101) V1(100)
0 V6(101) V1(100) V2(110) V3(010) V4(011) V5(001)
0 1 V3(010) V4(011) V5(001) V6(101) V1(100) V2(110)
0 V5(001) V6(101) V1(100) V2(110) V3(010) V4(011)
Fig. 2.25 Voltage vectors

and spatial sector definition
direction determined by the output of the hysteresis controller for the torque. The
switching table for controlling both the amplitude and rotating direction is shown in
Table 2.5, in which the inverter voltage vector and spatial sector definitions are
illustrated in Fig. 2.25.
Figure 2.26 shows the simulation model built based on the typical DTC scheme.
The inverter switching status and DC bus voltage are utilized to calculate the stator
voltage. The stator flux linkage is obtained in the observer. The traditional two-level
hysteresis controllers are applied and the switching table is designed based on
Table 2.5.
2.6.4 Model Predictive Control
The principle of MPC was introduced for industrial control applications in the
1970s after the publication of this strategy in the 1960s. The MPC requires great
computational effort and it has been formerly limited to slowly varying systems,
such as chemical processes. With the availability of inexpensive high computing
Fig. 2.26 Simulation block diagram of typical DTC based PMSM drive system
power microcomputers and modern digital control techniques, MPC is able to be

applied to electrical drive systems [36, 55, 56].
Different from the employment of hysteresis comparators and the switching table
in conventional DTC, the principle of vector selection in MPC is based on eval-
uating a defined cost function. The selected voltage vector from the conventional
switching table in DTC may not necessarily be the best one for the purposes of
torque and flux ripple reduction. Since there are limited discrete voltage vectors in
the two-level inverter-fed PMSM drives, it is possible to evaluate the effects of each
voltage vector and select the one minimizing the cost function.
The key technology of MPC lies in the definition of the cost function, which is
related to the control objectives. The greatest concerns of PMSM drive applications
are the torque and stator flux, and thus, the cost function is defined in such a way
that both the torque and stator flux at the end of control period are as close as
possible to the reference values. In this book, the cost function is defined as

min: G ¼ jTe Tek þ 1 j þ k1 jws j jwsk þ 1 j
ð2:38Þ
s.t. uks 2 fV0 ; V1 ; . . .; V7 g
where Te and ws are the reference torque and flux, Tek þ 1 and wsk þ 1 the predicted
values of torque and flux, respectively, and k1 is the weighting factor. Because the
physical natures of electromagnetic torque and stator flux are different, the
weighting factor k1 is introduced to unify these terms. In this work, k1 is selected to
be Tn =wn , where Tn and wn are the rated values of torque and stator flux, respec-
tively. It should be noted that when a null vector is selected, the specific state (V0 or
V7) will be determined based on the principle of minimal switching commutations,
which is related to the switching states of the previous voltage vector.
The voltage equations in the d-q reference frame are as follows:
did
ud ¼ Rs id þ Ld xLq iq ð2:39Þ
dt
diq
u q ¼ R s i q þ Lq þ xLd id þ xwf ð2:40Þ
dt
Given the voltage and current values at sampling instant k, the predicted current,
torque and flux at instant k + 1 can be expressed as follows:
1
ikd þ 1 ¼ ikd þ Rs ikd þ xk Lq ikq þ ukd Ts ð2:41Þ
Ld
1 k k
ikq þ 1 ¼ ikq þ x Ld id Rs ikq þ ukq xk wf Ts ð2:42Þ
Lq

wks þ 1 ¼ Ld ikd þ 1 þ wf þ jLq ikq þ 1 ð2:43Þ
3
Tek þ 1 ¼ pwks þ 1 iks þ 1 ð2:44Þ
2
where ikd þ 1 and ikq þ 1 are the predicted values of stator current for the sampling
instant k + 1, Ts is the sampling period, Tek þ 1 and wks þ 1 are the predicted values of
torque and flux, respectively, which are also the main concerns for the cost function
in the following MPC control scheme [1, 36, 49].
The block diagram of MPC is shown in Fig. 2.27. The inputs of the system are
the reference and estimated values of torque and flux. By evaluating the effects of
each voltage vector when applied to the machine, the voltage vector which mini-
mizes the difference between the reference and predicted values is first selected, and
then it is generated by the inverter.
Fig. 2.27 Block diagram of MPC drive system in MATLAB/SIMULINK

2.6.4.1 One-Step Delay Compensation
The cost function in (2.38) assumes that all calculations and judgments are
implemented at the kth instant and the selected vector will be applied immediately.
However, in practical digital implementation, this assumption is not true and the
applied voltage vector is not applied until the (k + 1)th instant.
In other words, for the duration between the kth and (k + 1)th instants, the
applied rotor voltage vector uks has been decided by the value in the (k-1)th instant
and the evolutions of ws and Te for this duration are uncontrollable. What is left to
be decided is actually the stator voltage vector uks þ 1 , which is applied at the
beginning of the (k + 1)th instant. To eliminate this one step delay, the variables of
wks þ 2 and Tek þ 2 should be used rather than wks þ 1 and Tek þ 1 for the evaluation of the
cost function in (2.38). This fact is clearly illustrated in Fig. 2.28, where x indicates
the state variables of a dynamic system and u is the input to be decided. For PMSM,
x represents torque or stator flux value.
To eliminate the one-step delay in digital implementation, the cost function in
(2.38) should be changed to (2.45) as shown below

min: G ¼ jTe Tek þ 2 j þ k1 jws j jwks þ 2 j
ð2:45Þ
s.t. uks 2 fV0 ; V1 ; . . .; V7 g
Obtaining wks þ 2 and Tek þ 2 in (2.45) requires a two-step prediction. To obtain the
best voltage vector minimizing the cost function in (2.45), each possible configu-
ration for uke þ 1 will be evaluated to obtain the value at the (k + 2)th instant.
2.6.4.2 Linear Multiple Horizon Prediction
A linear multiple horizon prediction formula is introduced in this section. This

formula incorporates two formulas. The first one is the same as in (2.38). The linear
Fig. 2.28 One-step delay in digital control systems

multiple horizon prediction formula, which is multiplied by a factor A, considers the

errors in the (k + N)th instant (N > 1). Different from the model-based predictions
for wsk þ 1 and Tek þ 1 , the stator flux and torque at the (k + N)th instant are predicted
from the value at the kth and (k + 1)th instants using linear extrapolations, which are
expressed as
Tek þ N ¼ Tek þ ðN 1ÞðTek þ 1 Tek Þ ð2:46Þ

jwk þ N j ¼ jwk j þ ðN 1Þjwk þ 1 j jwk j
s s s s ð2:47Þ
The expression of the proposed cost function is

min: G ¼ jTe Tek þ 1 j þ k1 jws j jwsk þ 1 j

þ A jTe Tek þ N j þ k1 jws j jwks þ N j s:t: uks 2 fV0 ; V1 ; . . .; V7 g
ð2:48Þ
2.6.5 Numerical and Experimental Comparisons of DTC

and MPC
2.6.5.1 Numerical Simulation
In this section, the simulation tests of DTC and MPC are carried out by using
Matlab/Simulink. The parameters of the motor are listed in Table 2.6. The sampling
frequency of both methods is set to 5 kHz. The values of control parameters are
k1 ¼ 25:4; A ¼ 0:1; and N ¼ 10 [36].
This simulation test combines start-up, steady-state and external load tests. The
motor starts up from 0 s with several reference speeds (500 rev/min, 1000 rev/min,
1500 rev/min and 2000 rev/min). After reaching the reference speed, the motor
maintains the speed for at least 0.2 s and an external load is applied at 0.3 s.
Figures 2.29, 2.30, 2.31 and 2.32 show the combined load test for four control
strategies for one reference speed, 1000 rev/min. From top to bottom, the curves are
the stator current, stator flux, torque, motor speed, and switching frequency,
respectively. The test results for other speed situations can be found in [36].
Table 2.6 Motor parameters Number of pole pairs p 3

Permanent magnet flux wf 0.1057 Wb
Stator resistance Rs 1.8 Ω
d- and q-axis inductance Ld, Lq 15 mH
Rated torque TN 4.5 Nm
DC bus voltage Vdc 200 V
Fig. 2.29 Combined load test Classic DTC

for DTC 10
ia/A
0
-10
0.2
Phi/Wb
0.15
0.1
0.05
6
4
Te/Nm
2
0
2000
Speed/rpm
1500
1000
500
0
4000
fs/Hz
2000
0
0 0.1 0.2 0.3 0.4 0.5
t/s
By comparing Fig. 2.30 with Fig. 2.29, it is shown that the torque and flux
ripples of MPC are lower than that of DTC. In Fig. 2.31, MPC with one-step delay
compensation (indicated as MPC + comp) presents torque and flux ripples even
lower than MPC along with an increase in switching frequency. Figure 2.32
illustrates the responses by using cost function (2.48), where factor A is included in
the simulation. As shown, the introduction of linear multiple horizon prediction
(factor A, and indicated as MPC + A) can greatly reduce the switching frequency
only with a quite limited degradation of torque and flux ripples. As shown, all these
methods present similar dynamic performance and the motor can reach the refer-
ence speed rapidly. When the load was applied, the motor speed returned to its
original value in a very short time period.
The recorded data from 0.1 to 0.3 s are picked to calculate the torque and flux
ripples (obtained by standard deviations). The torque and flux ripples of these
control methods are summarized in Table 2.7. A segment (three periods) of the
stator current of phase A is used to calculate the total harmonic distortion
(THD) and current harmonic spectrum.
As shown, MPC can achieve lower torque ripple than that of DTC as proven.
However, MPC’s characteristic in flux ripple reduction is quite unstable. With the
help of one-step delay compensation, the steady-state performance of MPC is
Fig. 2.30 Combined load test MPC

for MPC 10
ia/A
0
-10
0.2
Phi/Wb
0.15
0.1
0.05
6
4
Te/Nm 2
0
2000
Speed/rpm
1500
1000
500
0
4000
fs/Hz
2000
0
0 0.1 0.2 0.3 0.4 0.5
t/s
improved significantly. It should be noticed that the switching frequency has

increased by almost two times in most tests when one-step delay is compensated.
The introduction of linear multiple horizon prediction can effectively reduce the
switching frequency and flux ripple. However, its ability on torque ripple reduction
is quite insignificant.
2.6.5.2 Experimental Testing
In addition to the simulation study, the control methods mentioned above are
further experimentally tested on a two-level inverter-fed PMSM motor drive. The
experimental setup is illustrated in Fig. 2.33. A dSPACE DS1104 PPC/DSP control
board is employed to implement the real-time algorithm coding using C language.
A three phase intelligent power module equipped with an insulated-gate bipolar
transistor (IGBT) is used as an inverter. The gating pulses are generated in the
DS1104 board and then sent to the inverter. The load is applied using a pro-
grammable dynamo-meter controller DSP6000 (Fig. 2.34). A 2500-pulse incre-
mental encoder is equipped to obtain the rotor speed of PMSM. All experimental
results are recorded by the ControlDesk interfaced with DS1104 and PC at 5 kHz
sampling frequency [36].
Fig. 2.31 Combined load test MPC + comp

for MPC with one-step delay 10
compensation
ia/A
0
-10
0.2
Phi/Wb
0.15
0.1
0.05
6
4
Te/Nm 2
0
2000
Speed/rpm
1500
1000
500
0
4000
fs/Hz
2000
0
0 0.1 0.2 0.3 0.4 0.5
t/s
The steady-state responses at 1000 rpm are presented in this section. From top to
bottom, the curves shown are torque, stator flux and switching frequency,
respectively.
Figures 2.35 and 2.36 show the measured steady-state performance at 1000 rpm.
It is seen that the implementation of MPC can reduce the torque ripple, but does not
reduce the flux ripple. When the one-step delay is compensated, a significant
decrease of torque and flux ripples can be found as well as an obvious increase of
switching frequency. When the linear multiple horizon prediction is added to MPC,
it can be seen that the torque and flux ripples are slightly decreased along with a
limited reduction of the switching frequency.
Table 2.8 lists the torque and flux ripples of these control methods in experiment.
As shown, similar conclusions can be obtained as those from Table 2.7. According
to the analysis above, it can be concluded that:
(1) MPC can achieve lower torque ripple than that of DTC whilst
maintaining/reducing the switching frequency as proven in both simulation
and experimental tests. However, MPC’s ability in flux ripple reduction is
insignificant and even unstable.
Fig. 2.32 Combined load test MPC + A

for MPC with linear multiple 10
horizon prediction
ia/A
0
-10
0.2
Phi/Wb
0.15
0.1
0.05
6
4
Te/Nm 2
0
2000
Speed/rpm
1500
1000
500
0
4000
fs/Hz
2000
0
0 0.1 0.2 0.3 0.4 0.5
t/s
Table 2.7 Steady-state Method THD fav wrip ðWbÞ Trip ðNmÞ
response (simulation) (%) (kHz)
DTC 28.83 1.5972 0.0155 0.6869
MPC 18.55 1.5692 0.0138 0.4952
MPC + comp 8.17 1.5812 0.0059 0.2253
MPC + A 15.52 0.6640 0.0090 0.5102
(2) When one-step delay is compensated, the steady-state performance of MPC in

terms of torque and flux ripples reduction is significantly improved. It should
be noticed that the performance improvement also comes with a remarkable
switching frequency increase (two times or more).
(3) By introducing linear multiple horizon prediction to MPCs, a significant
switching frequency reduction can be found as well as an obvious decrease in
flux ripple. However, it comes with heavy penalty of torque ripple increasing,
especially at low motor speed.
Fig. 2.33 Experimental setup of testing system: a overview of the testing platform and b front
view of the PMSM and inverter control board
2.6.6 Improved MPC with Duty Ratio Optimization
There are many improvements for these control algorithms. One of them known as
MPC with duty ratio optimization will be selected for the control of PM-SMC
TFM. As a general algorithm, the theory and test results will be presented in this
section.
Fig. 2.34 Dynamo-meter controller DSP6000
(a) Classic DTC (b) MPC

2 2
1 1
Te/Nm
Te/Nm
0 0
-1 -1
-2 -2
0.16 0.16
• s/Wb
• s/Wb
0.10 0.10
0.04 0.04
4000 4000
fs/Hz
fs/Hz
2000 2000
0 0
0 1 2 0 1 2
t/s t/s
Fig. 2.35 Steady-state response for: a DTC, b MPC
In the conventional MPC, the selected voltage vector works during the whole
sampling period. In many cases, it is not necessary to work for the entire period to
meet the performance requirement of torque and flux. This is one of the main
reasons for the torque and flux ripples. By introducing a null vector to each sam-
pling period, the effects of voltage on torque can be adjusted to be more moderate,
in order to diminish the ripples of torque and flux.
(a) MPC + comp (b) MPC + A

2 2
1 1
Te/Nm
Te/Nm
0 0
-1 -1
-2 -2
0.16 0.16
• s/Wb
• s/Wb
0.10 0.10
0.04 0.04
4000 4000
fs/Hz
fs/Hz
2000 2000
0 0
0 1 2 0 1 2
t/s t/s
Fig. 2.36 Steady-state response for: a MPC with one-step delay compensation, and b MPC with
linear multiple horizon prediction
Table 2.8 Steady-state Method fav (kHz) wrip ðWbÞ Trip ðNmÞ
response at 1000 rpm
(experimental) DTC 1.2129 0.0167 0.8446
MPC 1.1393 0.0173 0.6394
MPC + comp 2.7335 0.0056 0.2310
MPC + A 0.6045 0.0136 0.4460
Actually, the torque can be changed by adjusting the amplitude and time
duration of us . The amplitude is decided by the DC bus voltage and is usually fixed,
while the time duration of us can be varied from zero to the whole period, which is
equivalent to changing the voltage vector length. The null vector only decreases the
torque, while appropriate non-zero vectors can increase the torque, and it is possible
to employ both null and nonzero vectors during one cycle to reduce the torque
ripple. The appropriate non-zero vectors are also referred as ‘active vector’. The key
issue is how to determine the time duration of the two vectors, or the duty ratio of
the active vector.
The expression of duty ratio for MPC is shown as follows

T Tek þ 1 ws wek þ 1
d ¼ e þ

;
ð2:49Þ
C T C w
Fig. 2.37 Diagram of an improved MPC with duty ratio optimization in MATLAB/SIMULINK
where d is the duty ratio of the active voltage vector, and CT and Cψ are two
positive parameters. The idea of this method is that the larger the difference between
the reference and predicted torque values, the larger is the duty ratio value [36]. On
the other hand, the lower the CT and Cψ values, the quicker is the dynamic response
(e.g. take less time to reach the given speed), but the poorer will be the steady-state
response (e.g. higher torque and flux ripples). Higher values of CT and Cψ could
lead to better steady-state responses, but slower dynamic responses. Therefore, the
determination of these values is a compromise between the steady-state and
dynamic performances. Extensive simulation and experimental results have proven
that the PM flux value and half-rated torque value for CT and Cψ can provide a good
compromise between the steady state performance and the dynamic response. The
block diagram of the proposed improved MPC is shown in Fig. 2.37.
2.6.7 Numerical and Experimental Comparisons of DTC

and MPC with Duty Ratio Optimization
2.6.7.1 Numerical Simulation
The parameters of the motor and control system simulated in this section are listed
in Table 2.9. Similar to the previous test example, this simulation test combines the
start-up, steady-state and external load tests. The motor starts up from 0 s with
several reference speeds (500 rev/min, 1000 rev/min, 1500 rev/min and 2000
rev/min). After reaching the reference speed, the motor maintains the speed for at
least 0.2 s and an external load is applied at 0.3 s. Figure 2.38 shows the combined
load test for one reference speed, 1000 rev/min. From top to bottom, the curves are
Table 2.9 Motor and control Parameter Symbol Value

system parameters
Number of pole pairs p 3
Permanent magnet flux wf 0.1057 Wb
Stator resistance Rs 1.8 Ω
d-axis and q-axis inductance Ld, Lq 15 mH
DC bus voltage Vdc 200 V
Inertia J 0.002 kg m2
Torque constant gain CT 2
Flux constant gain Cw 0.1
Sampling frequency fsp 5 kHz
Fig. 2.38 Combined load test MPC + Duty

for MPC with duty ratio 10
optimization at 1000 rev/min
ia/A
-10
0.2
Phi/Wb
0.15
0.1
0.05
6
4
Te/Nm
2
0
2000
Speed/rpm
1500
1000
500
0
1
Duty
0.5
6000
fs/Hz
4000
2000
0
0 0.1 0.2 0.3 0.4 0.5
t/s
the stator current, stator flux, torque, motor speed, and switching frequency,
respectively. The test results for other speed situations can be found in [36].
It can be found that the proposed MPC scheme present very low torque and flux
ripples and excellent dynamic response. The proposed MPC scheme also presents
very low stator current THDs and narrow harmonic spectrums with the dominant
harmonics of around 5 kHz.
2.6.7.2 Experimental Test
The experimental tests are performed on the same testing platform introduced in the
last section. Figure 2.39 shows the steady-state responses at 1000 rpm for three
control strategies, namely (a) DTC, (b) MPC, and (c) MPC with duty ratio
optimization.
(a) Classic DTC (b) MPC

2 2
1 1
Te/Nm
Te/Nm
0 0
-1 -1
-2 -2
0.16 0.16
Phis/Wb
Phis/Wb
0.10 0.10
0.04 0.04
4000 4000
fs/Hz
fs/Hz
2000 2000
0 0
0 1 2 0 1 2
t/s t/s
MPC Duty
(c) 2
1
Te/Nm
0
-1
-2
0.16
Phis/Wb
0.10
0.04
4000
fs/Hz
2000
0
1
duty
0.5
0
0 1 2
t/s
Fig. 2.39 Steady-state response at 1000 rpm for: a DTC, b MPC and c MPC with duty ratio
optimization
It can be seen that in MPC with duty ratio optimization, the torque and flux
ripples are reduced significantly compared to other methods. The duty ratio
increases along with the increase in motor speed.
According to the analysis above, it can be concluded that:
(1) MPC with duty ratio optimization can achieve a better performance than DTC
and original MPC in terms of torque and flux ripples reduction;
(2) Under the same system sampling frequency (5 kHz), the switching frequency
of the improved method is much higher than other methods; and
(3) In DTC and MPC, the switching frequency slightly decreases along with the
increase of motor speed. However, the switching frequency is almost stable in
the proposed method.
More experimental results including different speed, dynamic response and data
analysis can be found in [36].
2.7 Summary
This chapter presents the multi-disciplinary design analysis models and methods for
electrical machines and drive systems. All the models and methods are discussed in
terms of the three major parts of electrical drive systems, namely electrical
machines, power electronic converters and controllers. Electromagnetic, thermal
and mechanical analyses based on different models, e.g. FEM, have been investi-
gated for the design of electrical machines with several prototypes developed in our
research center. Various kinds of popular control algorithms have been described
for the controller design. Several examples investigated in our previous work have
been presented to show the effectiveness of the proposed models and analysis
methods.
References
1. Lei G, Wang TS, Guo YG, Zhu JG, Wang SH (2014) System level design optimization
methods for electrical drive systems: deterministic approach. IEEE Trans Ind Electron 61
(12):6591–6602
2. Lei G, Wang TS, Zhu JG, Guo YG, Wang SH (2015) System level design optimization
method for electrical drive system: robust approach. IEEE Trans Ind Electron 62(8):4702–
4713
3. Zhu JG, Guo YG, Lin ZW, Li YJ, Huang YK (2011) Development of PM transverse flux
motors with soft magnetic composite cores. IEEE Trans Magn 47(10):4376–4383
4. Zhu JG, Ramsden VS (1998) Improved formulations for rotational core losses in rotating
electrical machines. IEEE Trans Magn 34(4):2234–2242
5. Guo YG, Zhu JG, Lu HY, Lin ZW, Li YJ (2012) Core loss calculation for soft magnetic
composite electrical machines. IEEE Trans Magn 48(11):3112–3115
6. Guo YG, Zhu JG, Lu HY, Li YJ, Jin JX (2014) Core loss computation in a permanent magnet
transverse flux motor with rotating fluxes. IEEE Trans Magn 50(11). Article#: 6301004
7. Guo YG, Zhu JG, Zhong JJ, Wu W (2003) Core losses in claw pole permanent magnet
machines with soft magnetic composite stators. IEEE Trans Magn 39(5):3199–3201
8. Huang YK, Zhu JG, Guo YG, Lin ZW, Hu Q (2007) Design and analysis of a high speed claw
pole motor with soft magnetic composite core. IEEE Trans Magn 43(6):2492–2494
9. Huang YK, Zhu JG et al (2009) Thermal analysis of high-speed SMC motor based on thermal
network and 3-D FEA with rotational core loss included. IEEE Trans Magn 45(10):4680–4683
10. Pfister P-D, Perriard Y (2010) Very-high-speed slotless permanent-magnet motors: analytical
modeling, optimization, design, and torque measurement methods. IEEE Trans Ind Electron
57(1):296–303
11. Komeza K, Dems M (2012) Finite-element and analytical calculations of no-load core losses
in energy-saving induction motors. IEEE Trans Ind Electron 59(7):2934–2946
12. Wang SH, Meng XJ, Guo NN, Li HB, Qiu J, Zhu JG et al (2009) Multilevel optimization for
surface mounted PM machine incorporating with FEM. IEEE Trans Magn 45(10):4700–4703
13. Barcaro M, Bianchi N, Magnussen F (2012) Permanent-magnet optimization in
permanent-magnet-assisted synchronous reluctance motor for a wide constant-power speed
range. IEEE Trans Ind Electron 59(6):2495–2502
14. Vese I, Marignetti F, Radulescu MM (2010) Multiphysics approach to numerical modeling of
a permanent-magnet tubular linear motor. IEEE Trans Ind Electron 57(1):320–326
15. Bornschlegell AS, Pelle J, Harmand S, Fasquelle A, Corriou J-P (2013) Thermal optimization
of a high-power salient-pole electrical machine. IEEE Trans Ind Electron 60(5):1734–1746
16. Lee D-H, Pham TH, Ahn J-W (2013) Design and operation characteristics of four-two pole
high-speed SRM for torque ripple reduction. IEEE Trans Ind Electron 60(9):3637–3643
17. Flieller D, Nguyen NK, Wira P, Sturtzer G, Abdeslam DO, Merckle J (2014) A self-learning
solution for torque ripple reduction for nonsinusoidal permanent-magnet motor drives based
on artificial neural networks. IEEE Trans Ind Electron 61(2):655–666
18. Hasanien HM, Abd-Rabou AS, Sakr SM (2010) Design optimization of transverse flux linear
motor for weight reduction and performance improvement using response surface
methodology and genetic algorithms. IEEE Trans Energy Convers 25(3):598–605
19. Hasanien HM (2011) Particle swarm design optimization of transverse flux linear motor for
weight reduction and improvement of thrust force. IEEE Trans Ind Electron 58(9):4048–4056
20. Lei G, Liu CC, Guo YG, Zhu JG (2015) Multidisciplinary design analysis for PM motors with
soft magnetic composite cores. IEEE Trans Magn 51(11). Article 8109704
21. Hua W, Cheng M, Zhu ZQ, Howe D (2006) Design of flux-switching permanent magnet
machine considering the limitation of inverter and flux-weakening capability. In: Proceedings
of 41st IAS annual meeting-industry applications conference, vol 5, pp 2403–2410
22. Liu CC, Zhu JG, Wang YH, Lei G, Guo YG, Liu XY (2014) A low-cost permanent magnet
synchronous motor with SMC and ferrite PM. In: Proceedings of 17th international conference
on electrical machines and systems (ICEMS), pp 397–400
23. Fei W, Luk PCK, Shen JX, Wang Y, Jin M (2012) A novel permanent-magnet flux switching
machine with an outer-rotor configuration for in-wheel light traction applications. IEEE Trans
Ind Appl 48(5):1496–1506
24. Guo YG (2003) Development of low cost high performance permanent magnet motors using
new soft magnetic composite materials, UTS thesis (PhD)
25. Guo YG, Zhu JG, Watterson PA, Wei Wu (2006) Development of a PM transverse flux motor
with soft magnetic composite core. IEEE Trans Energy Conver 21(2):426–434
26. Guo YG, Zhu JG, Watterson PA, Wei Wu (2003) Comparative study of 3-D flux electrical
machines with soft magnetic composite cores. IEEE Trans Ind Appl 39(6):1696–1703
27. Guo YG, Zhu JG, Dorrell D (2009) Design and analysis of a claw pole PM motor with molded
SMC core. IEEE Trans Magn 45(10):582–4585
28. Lei G, Shao KR, Guo YG, Zhu JG (2012) Multi-objective sequential optimization method for
the design of industrial electromagnetic devices. IEEE Trans Magn 48(11):4538–4541
References 71
29. Lei G, Guo YG, Zhu JG et al (2012) System level six sigma robust optimization of a drive
system with PM transverse flux machine. IEEE Trans Magn 48(2):923–926
30. Lei G, Zhu JG, Guo YG, Hu JF, Xu W, Shao KR (2013) Robust design optimization of
PM-SMC motors for Six Sigma quality manufacturing. IEEE Trans Magn 49(7):3953–3956
31. Lei G, Zhu JG, Guo YG, Shao KR, Xu W (2014) Multiobjective sequential design
optimization of PM-SMC motors for six sigma quality manufacturing. IEEE Trans Magn, 50
(2). Article 7017704
32. Liu CC, Zhu JG, Wang YH, Guo YG, Lei G, Liu XY (2015) Development of a low-cost
double rotor axial flux motor with soft magnetic composite and ferrite permanent magnet
materials. J Appl Phys, 117(17). Article # 17B507
33. Teng QF, Zhu JG, Wang TS, Lei G (2012) Fault tolerant direct torque control of three-phase
permanent magnet synchronous motors. WSEAS Trans Syst 8(11):465–476
34. Teng QF, Bai J, Zhu JG, Sun Y (2013) Fault tolerant model predictive control of three-phase
permanent magnet synchronous motors. WSEAS Trans Syst 12(8):385–397
35. Wang Y (2011) Investigation of rotor position detection schemes for PMSM drives based on
analytical machine model incorporating nonlinear saliencies, UTS thesis (PhD)
36. Wang TS (2013) Model predictive torque control of PMSM with duty ratio optimization for
torque ripple reduction, UTS thesis (Master degree)
37. Kim SY, Lee W, Rho MS, Park SY (2010) Effective dead-time compensation using a simple
vectorial disturbance estimator in PMSM drives. IEEE Trans Ind Electron 57(5):1609–1614
38. Lee J, Hong J, Nam K, Ortega R, Praly L, Astolfi A (2010) Sensorless control of
surface-mount permanent-magnet synchronous motors based on a nonlinear observer. IEEE
Trans Power Electron 25(2):290–297
39. Genduso F, Miceli R, Rando C, Galluzzo GR (2010) Back EMF sensorless-control algorithm
for high-dynamic performance PMSM. IEEE Trans Ind Electron 57(6):2092–2100
40. Takahashi I, Noguchi T (1986) A new quick-response and high-efficiency control strategy of
an induction motor. IEEE Trans Ind Appl 22(5):820–827
41. Depenbrock M (1988) Direct self-control (DSC) of inverter-fed induction machine. IEEE
Trans Power Electron 3(4):420–429
42. Buja GS, Kazmierkowski MP (2004) Direct torque control of PWM inverter-fed AC motors-A
survey. IEEE Trans Ind Electron 51(4):744–757
43. Lai YS, Chen JH (2001) A new approach to direct torque control of induction motor drives for
constant inverter switching frequency and torque ripple reduction. IEEE Trans Energy
Convers 16(3):220–227
44. Lascu C, Trzynadlowski A (2004) A sensorless hybrid DTC drive for high-volume low-cost
applications. IEEE Trans Ind Electron 51(5):1048–1055
45. Zhang Y, Zhu J, Xu W, Hu J, Dorrell DG, Zhao Z (2010) Speed sensorless stator flux oriented
control of three-level inverter-fed induction motor drive based on fuzzy logic and sliding mode
control. In: Proceedings of 36th IEEE IECON, pp 2926–293
46. Zhang Y, Zhu J (2011) Direct torque control of permanent magnet synchronous motor with
reduced torque ripple and commutation frequency. IEEE Trans Power Electron 26(1):235–248
47. Zhang Y, Zhu J (2011) A novel duty cycle control strategy to reduce both torque and flux
ripples for DTC of permanent magnet synchronous motor drives with switching frequency
reduction. IEEE Trans Power Electron 26(10):3055–3067
48. Zhang Y, Zhu J, Xu W, Guo Y (2011) A simple method to reduce torque ripple in direct
torque-controlled permanent-magnet synchronous motor by using vectors with variable
amplitude and angle. IEEE Trans Ind Electron 58(7):2848–2859
49. Wang TS, Zhu JG, Zhang YC (2011) Model predictive torque control for PMSM with duty
ratio optimization. In Proceedings of 2011 international conference on electrical machines and
systems (ICEMS), pp 1–5, 20–23 August 2011
50. Miranda H, Cortes P, Yuz J, Rodriguez J (2009) Predictive torque control of induction
machines based on state-space models. IEEE Trans Ind Electron 56(6):1916–1924
51. Geyer T, Papafotiou G, Morari M (2009) Model predictive direct torque control—Part I:
Concept, algorithm, and analysis. IEEE Trans Ind Electron 56(6):1894–1905
52. Kouro S, Cortes P, Vargas R, Ammann U, Rodriguez J (2009) Model predictive control—a
simple and powerful method to control power converters. IEEE Trans Ind Electron 56
(6):1826–1838
53. Morel F, Retif J-M, Lin-Shi X, Valentin C (2008) Permanent magnet synchronous machine
hybrid torque control. IEEE Trans Ind Electron 55(2):501–511
54. Drobnic K, Nemec M, Nedeljkovic D, Ambrozic V (2009) Predictive direct control applied to
AC drives and active power filter. IEEE Trans Ind Electron 56(6):1884–1893
55. Zhang Y, Xie W (2014) Low complexity model predictive control-single vector-based
approach. IEEE Trans Power Electron 29(10):5532–5541
56. Zhang Y, Qu C (2015) Model predictive direct power control of PWM rectifiers under
unbalanced network conditions. IEEE Trans Ind Electron 62(7):4011–4022
http://www.springer.com/978-3-662-49269-7

Design Fundamentals of Electrical Machines and Drive Systems

Uploaded by

Copyright:

Available Formats

Design Fundamentals of Electrical Machines and Drive Systems

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 Fundamentals of Electrical Machines and Drive Systems

Uploaded by

Copyright:

Available Formats

Chapter 2

Design Fundamentals of Electrical

Abstract This chapter presents a brief summary of the design fundamentals

2.1.1 Framework of Multi-disciplinary Design

Figure 2.1 illustrates a general framework of multi-disciplinary design for electrical

© Springer-Verlag Berlin Heidelberg 2016 25

2.1.2 Power Losses and Efﬁciency

Pa ¼ Cha fBh þ Cea ðfBÞ2 þ Caa ðfBÞ1:5 ð2:1Þ

where f is the excitation frequency, B the magnitude of sinusoidal magnetic

Pr ¼ Phr þ Cer ðfBÞ2 þ Car ðfBÞ1:5 ð2:2Þ

Fig. 2.2 Average core losses

acceptable. For most types of machines, this can be assumed to be 1 % of the

Pout Pin ðPCu þ PCore þ PMec þ PStray Þ

Typical values of full load efﬁciency for rotating machines are:

2.2 Electromagnetic Design

Since electrical machines are electromagnetic devices for transforming electrical

2.2.1 Analytical Model

Analytical model is generally used to calculate the performance indicators of

2.2.2 Magnetic Circuit Model

The magnetic circuit model acts as a uniform principle in descriptive magneto-

Table 2.1 Main design Par. Description Unit Value

kPM ¼ kl Ncoil pUgap ð2:7Þ

Em ¼ xe kPM ¼ pxm kPM ð2:8Þ

2.2.3 Finite Element Model

um ðr; Dh; zÞ ¼ um ðr; Dh; zÞ ð2:10Þ

Table 2.2 Key PM-SMC Parameter Unit Calculated Measured

Pin ¼ 3Vin Ia cos u ð2:12Þ

Pout ¼ Pin Pcore Pcopper Pmech ð2:13Þ

Fig. 2.7 Per phase equivalent

2.3 Thermal Design

2.3.1 Thermal Limits in Electrical Machines

Table 2.3 Classiﬁcation of electrical insulation materials

The operating temperature of a machine is closely associated with its life

Life Time ¼ AeB=T ð2:15Þ

2.3.2 Thermal Network Model

Fig. 2.8 Thermal network model of the TFM prototype

Fig. 2.9 Prototype of a high speed claw pole motor

B. High speed claw pole motor

Table 2.4 Main dimensions Parameter Unit Value

Fig. 2.10 Magnetically

Fig. 2.11 Structure of a

In general, the geometrical complexity of an electrical machine requires a large

Fig. 2.12 Thermal network

Fig. 2.13 Nodal thermal

2.3.3 Finite Element Model

Fig. 2.14 Temperature distribution in the PM-SMC TFM obtained by 3D FEM

2.4 Mechanical Design

Mechanical design is another important issue in the design analysis of electrical

Fig. 2.18 A three-phase

2.5 Power Electronics Design

2.6 Control Algorithms Design

Control algorithms play an important role in the determination of dynamic and

While FOC is commonly used in various high performance electrical drive

2.6.1 Six-Step Control

As ia + ib + ic = 0 for a symmetric three phase system, the voltage equation can

The generated electromagnetic torque is given by

Te ¼ ðea ia þ eb ib þ ec ic Þ=xm ð2:24Þ

where ωm is the mechanical angular speed of the rotor.

Fig. 2.19 Block diagram of PMSM six-step drive system