Documente Academic
Documente Profesional
Documente Cultură
Executive Summary.......................................................... 3
Acknowledgements......................................................... 23
EXEcUtIVE SUMMARY
Electric drives are found in all areas of transportation for electromechanical energy converters to replace
and industrial automation and cover power ranges from traditional hydraulic systems and, more generally,
fractional to thousands of horsepower and beyond. mechanical systems. This trend will continue and even
Design of components of electric drives systems or the accelerate due to significant benefits of weight and
even more challenging task of designing sub-systems cost reduction, increased reliability of electrical systems,
or systems belonging to the electric drive applications and convenient control and automation via electric and
area is of great interest to many electrical engineers. electronic means.
The main reason for this interest is the current trend
ELEctRIc DRIVES DESIGN
cHALLENGES
The trend to utilize more electric drives significantly The needed attention to detail requires accurate design
impacts the requirements users have for simulation and simulation of many components along the energy
tools. flow channel. Logical consequences of this requirement
are both the need of the component-level simulation
• component, sub-system and system-level simulation tool to communicate with the system-level simulation
are all desirable since this global approach allows tool and the consideration of thermal and stress con-
both concentration on component performance but sequences of electromagnetic fields at the component
also makes possible the end-to-end simulation of level.
the final product;
• the physical nature of the application may require Ansoft provides a unique set of integrated tools that
the consideration of thermal and/or stress conse makes the global design and simulation of sophisticated
quences of electromagnetic fields; electric drive systems and components inherently
• integration in the simulation environment of thorough, accurate and intuitive for engineers.
productivity features such as parametrics sweeps,
optimization, statistical analysis and distributed
solving (on computer network) provides additional
benefits to users.
Fig 3. Ansoft’s Electromechanical (EM) Integrated Solution
Maxwell® is software for the simulation and ePhysics™ provides the ability to perform thermal and
analysis of high-performance electromagnetic and stress analysis and couple the results to Maxwell to
electromechanical components. The software allows determine electrical, power dissipation and mechanical
users to study static, frequency-domain, and time- integrity constraints.
varying electromagnetic fields in complex structures.
The ability to import high-fidelity component models
Simplorer® is multi-domain, system simulation software from Maxwell into Simplorer ensures accuracy at the
for the design of high-performance electromechanical system-level simulation. Additionally, the automatic
systems. The software includes a wide range of data link between Maxwell and ePhysics enables for
modeling techniques and statistical analysis capability, the simulation of thermal and stress effects of the
adheres to IEEE modeling standards, and allows components.
engineers to study design performance of electrical,
mechatronic, power-electronic, and electromechanical
systems.
ELEctRIc DRIVES
cOMPONENtS DESIGN
In many situations, particularly for medium voltage and be studied and optimized. In a radial magnetic field
high short-circuit current applications, major contact topology, an azimuthal electromagnetic force
manufacturers of vacuum-operating switching devices acting on the vacuum arc forces the arc to move over
(Siemens, ABB, etc.) use contacts with specially shaped the surface of the contact at a high speed of around 100
contact geometries to achieve superior performance m/s, with favorable effects on the aging of the surfaces
required by the application. Thus, over 100,000 and capable of improving the reliable current
switching cycles can be achieved while being able to interrupting ability.
ensure reliable large short-circuit current (tens of kA)
interruptions of over 100 times during the lifetime of Axial magnetic fields (two- or four-pole axial magnetic
the vacuum interrupter. Both radial and axial magnetic field in the contact gap) configurations can be explored
field topologies are used in such high-performance with available FEM simulation tools capable of
vacuum interrupters (Fig. 4). predicting the distribution of current and magnetic field
at arbitrary locations inside the device and the phase
This superior performance was in part due to FEM shift between currents and corresponding magnetic
simulation of different contact geometries using fields with important consequences on the dynamic of
particular distribution of conductive and magnetic the electric arc.
material properties. As a result, a favorable distribution
and dynamics of the electric arc between contacts can
Similar concerns regarding reliable switching and capa- (circuit breaker) is very different. Here, too, the concern
bility to interrupt high short-circuit currents face design- is to use various techniques, many times a combination
ers of low-voltage switching devices. For this area of of techniques aimed in essence at extracting as much
switching applications, the configuration of the device energy as possible from the burning electric arc.
The model in Fig. 6 exhibits an arc divider/extinguisher post-processing in such an application: Current density
chamber as well as an integrated coil, both with differ- and magnetic field magnitudes are presented as well
ent, distinct roles in the process of reliably interrupting as a parametric study of the force calculation acting on
the current (normal current or short-circuit current). For the electric arc (simulated as a cylindrical, conducting
such devices, the distribution of magnetic field and con- plasma channel).
duction current is important. Fig. 7 shows examples of
Fig. 7b Lorentz force distribution on the arc and force vs. position, parametric study
In Fig. 7b, the plot represents the variation of the force ers. To extend the simulation capability into the inves-
acting on the electric arc plasma as the respective elec- tigation of the thermal and stress consequences of the
tric arc is “attracted” inside the ferromagnetic walls of currents flowing in the device, Maxwell is coupled with
the extinguisher. The parametric study was performed ePhysics. The nature of the coupling is dynamic in the
by using Maxwell’s distributed processing capability by sense that once the communication channel has been
solving different variations on parallel computers. established between the two desktops, needed data
flows automatically and allows the adaptive meshing
However, in this application of equal interest are the technology to be used in both. For this reason, since the
thermal and stress consequences of electromagnetic meshes in the two-coupled model may be different, an
fields. Indeed, for both normal operation regime and adequate mapping is used and applicable for the power
fault (short-circuit) regime, the thermal and stress loss density as well as for the force distribution, respec-
behavior of the device is of intense interest for design- tively, depending on the application.
Fig. 8 The principle of Maxwell and ePhysics coupling for steady-state electromagnetics
The coupling between the Maxwell and ePhysics desk- such that the full automatic nature of the coupling can
tops is much more general than the situation presented be used even in complex parametric sweeps.
in Fig. 8 and covers all types of electromagnetic solu-
tions from static ones to transient ones. Also note that Using the technology sketched above, Figs. 8a and 8b
the dynamic datalink technology used for the coupling present the temperature distribution and deformation
allows parameter mapping between coupled designs respectively of the parts of interest.
Fig. 8a Temperature distribution, thermal Fig. 8b. Thermal deformation of the current path,
steady-state simulation deformation is amplified for visualization using a
convenient scale
Inductors are used in many electric drives applications. ic analysis phase, but thermal and/or stress effects also
Inductors may be used as part of filters to provide cur- may be investigated. Shunts, used mainly for DC current
rent smoothing effects or di/dt limitations as needed by measurements, are modeled in a similar way. Typical
the application. Their design involves an electromagnet- examples of inductor geometries are presented in Fig. 9.
Fig. 9 Different inductor geometries; left: quarter model, right: high current, full model
Inductor modeling (as the modeling of many other Maxwell, inductance and resistance of components can
electromagnetic devices) is possible from a dual per- be exported in a convenient format for sub-system/sys-
spective: a component analysis using a field solution tem-level analysis in Simplorer. Fig.10 shows how such
(Maxwell with or without coupling with ePhysics) but equivalent models with field effect can be exported
also from the perspective of exporting the electromag- from Maxwell for use in a sub-system or system-level
netic essence of the device as an equivalent circuit with simulation together with other components.
field effects. Thus, using the ECE feature available in
Using the Maxwell-to-ePhysics datalink, the power-loss then used to obtain a sequence of static stress solutions
density distribution is mapped from the eddy-current using the respective ePhysics stress solver. The results of
solution to the thermal transient model. The respective this chained example are shown in Fig. 11.
temperature distributions at user-selected times are
Fig. 11 Eddy current -> thermal transient -> stress daisy-chained simulation
Note that in the above example, each of the steady/ In many drives applications (particularly in large power),
static solvers is using its own adaptive meshing the role of the power transformer can be very complex.
solution (as set by the user) while the thermal transient Let us consider the sophisticated medium-voltage
simulation is performed with fixed mesh (final mesh in Harmony drive from Siemens AG, Fig. 12.
the eddy-current solution) and its own adaptive time-
stepping algorithm.
To achieve the needed medium voltage, the above Maxwell is used to produce a very large inductance
solution uses a power transformer with many (9 matrix. In an 18 secondary winding transformer, there
– 18) phase shifted secondary windings and a series are at least 37 coils, giving a matrix with 666 coupling
connection of power cells to reach the input/output factors. This very large equivalent circuit (the inductance
power quality parameters and the desired level matrix) is then exported to Simplorer for further
of voltage to be applied to the motor. There is no analysis of the system. Other transient simulations are
additional reactance between transformer and cell, performed in Maxwell using the harmonic spectrum
so the commutating inductance of the diode bridge predicted from the Simplorer simulation (see paragraph
is entirely due to the leakage inductance of the D for more info).
transformer. This creates an interesting design problem,
as the transformer losses are greatly affected by the The thermal simulation of such large units (1,100 KVA
harmonic content of the secondary windings. in this case) is also a challenge. Fig. 13 shows a typical
For the electromagnetic design of the transformer, situation with forced air cooling used.
10
Fig. 13. 1,100 KVA power transformer unit with six forced convection channels per phase
Thus, the thermal modeling of this application makes channels in the model. Examples of measured
use of sophisticated forced convection boundary temperatures and the respective match with simulation
conditions in ePhysics that are based on customer- results are presented in Fig. 14.
supplied measurements of air velocities in all convection
C
C
B
B
A
A
Temp Rise A B C
Fig. 13. 1,100 KVA power transformer unit with six forced convection channels per phase
A good match was also reported for the primary and Medium-voltage transformers pose other design
secondary winding steady-state temperature simulation challenges as well. For example, the model shown in
with forced/natural convection and radiation boundary Fig. 15 is analyzed from an electrostatic field simulation
conditions (as applicable, depending on location). perspective.
11
Fig. 15 Example of bushing application for medium/high-voltage transformer
For this application, the requirement of simulating with simulated. Using specific geometry modeling and
high accuracy the maximum level of electric field in applying mesh operations in Maxwell, the subsequent
critical areas leads to the need to model real curvatures model that uses also automatic adaptive meshing is
in the respective regions and perform adequate capable of rendering the high electric field magnitudes
meshing such that the magnitude of the electric fields in the targeted regions (Fig. 16).
in the respective regions is adequately and realistically
Fig. 16 Shows main area of interest and electric field solution superimposed on local mesh (detail)
Bus-bar systems are used in most high-power electric of amps, sometimes even more. Fig. 17 presents an
drives applications. For such applications, the bus bars extreme application with a system of bus bars and
are capable of carrying typically hundreds to thousands cables capable of carrying over 100,000 A (RMS).
Fig. 17. 150,000 Amps (peak) system used in steel smelting plant
12
For such very large industrial size applications, many large shield systems and of the bus-bar system itself.
aspects are of interest, from the level of fields in the However, for most drives, the usual type of bus bar used
neighboring regions, adjacent biological aspects, to transfer the power is contained inside a cabinet, such
shielding and thermal and stress simulation of such as shown in Fig. 18.
Fig. 18 Medium voltage electric drive cabinet (from Siemens AG) Bus-bar capacity: 2,000 A
For this bus-bar application, the requirement is two- simulation together with other components of interest)
fold: of interest are the R, L and C of the different and the thermal and thermal stress simulation as well.
sections of the bus bar (to be used in a system-level
The equivalent circuit parameters are extracted using of the simulation, temperature distribution and the
Q3D Extractor®. For the thermal and stress simulation, corresponding thermal stress under rated current flow
we use Maxwell and ePhysics. Fig. 20 presents the result conditions.
13
PCBs containing components, ICs, power electronics As an example, let us consider Fig. 21, which presents
devices, IGBTs, FETs and SCRs, are eligible mainly for a Maxwell (eddy current) design coupled with ePhysics
thermal (steady-state or transient) simulations. In most thermal steady state. The respective simulation
cases, a stand-alone ePhysics simulation is used for these concentrates on the weak point of the design, the hot
specific applications, but, of course, the combination spot in the connecting wire, which gets to dangerous
with Maxwell also is useable in the same applications. levels of temperature.
For the thermal analysis of PCBs in most cases, unneeded stress solver modules. There are situations, however,
detail (from a thermal perspective) is not included in where the geometry complexity is manageable, and the
the models. PCBs with many layers of copper and FR4 problem can be solved using its full complexity. Such is
are modeled, for example, as thermally anisotropic the case of the model shown in Fig. 22, exhibiting a PCB
structures. ePhysics has the capability to use anisotropy with a few layers of copper traces and ICs.
settings for the material properties in both thermal and
Stress modeling also proves necessary in some PCB of the multi-layer PCB and over 200 applied points of
applications, such as is the case presented below, where pressure. A study of placement of the anchor points is
a “bed of pins” model simulation is shown. The model then performed such that deformation of the PCB is
uses an anisotropic setting for the material property kept within specified limits.
14
Electric machines are usually difficult to simulate. There 2D/3D FEM analysis using field solvers. The advantage
is the question of the modeling approach to use; an of using the Maxwell desktop is that different levels
analytical approach (less accurate in general but very of abstraction (and accuracy) are possible for the
fast) vs. a finite element approach. Both methods are simulation while making the data transfer between the
possible in the Maxwell desktop, which integrates available tools quite easy, as shown in Fig. 24.
analytical modeling (with RMxprt™) and accurate
3,735,782 DOFs
65
Matrix solution time
60
55
50
[min]
Series1
45
40
35
30
0 1 2 3 4 5
No of Processors
Fig. 25 Large magnetostatic, high-accuracy model (over 3.5 million degrees of freedom)
of a stepper motor using multiple processing
15
The brush-type DC motor model presented in Fig. 26 brush and commutator bar a function of position, such
makes use of commutating elements embedded in that the respective effect is considered automatically
the library of elements available for use in transient during the simulation. Mechanical elements connected
applications. The simulation uses the capability of to the moving part, such as load force/torque, inertia,
making the contact resistance at the interface between and damping, can be specified during set-up.
Some of the results of the simulation are shown in Fig. EMF, torque, core loss, etc.); others can be specified
26. Note that a whole range of results are available as additional quantities to be calculated by the field
for post-processing, some of which are calculated calculator.
automatically (winding current, flux linkage, back
For electric machines, the thermal and mechanical The synchronous, permanent magnet electric motor
stress consequences are two other distinct areas that analyzed from a thermal performance perspective has
can now be investigated using the available couplings a max of 600 KW output and is used in a completely
between Maxwell and ePhysics. The following example new gearless drive system for future high-speed trains.
is the result of cooperation with Siemens AG (Large The configuration in which the thermal performance is
Drives, Nuremberg, Germany) who provided the model, analyzed is presented in Figs. 27 and 28.
material properties and thermal test measurement data
for comparisons.
16
Fig. 27. PM motor in a direct drive configuration
Cooling Jacket
with Spiral Channel
Inlet
Stator Winding
PM Rotor
17
For electric machines, the thermal and mechanical The synchronous, permanent magnet electric motor
stress consequences are two other distinct areas that analyzed from a thermal performance perspective has
can now be investigated using the available couplings a max of 600 KW output and is used in a completely
between Maxwell and ePhysics. The following example new gearless drive system for future high-speed trains.
is the result of cooperation with Siemens AG (Large The configuration in which the thermal performance is
Drives, Nuremberg, Germany) who provided the model, analyzed is presented in Figs. 27 and 28.
material properties and thermal test measurement data
for comparisons.
Other modeling features used in the thermal Simulation of the above-described device, performed
simulation were the anisotropic distribution of in ePhysics, showed a good match with experimental
thermal conductivity of the laminations for the rotor data for all regions of the motor. As an example, Fig.
and stator, two power levels and two inlet cooling 30 presents temperature distribution in the rotor and
fluid temperatures taken into account, convective permanent magnets, useful information due to the
and radiation boundary conditions used for external possible impact on the performance of the motor.
surfaces, and temperature-dependent cooling fluid
properties.
The match of permanent magnets temperature combination of electromagnetic and centrifugal body
distribution at three locations (the two ends and forces distributions is critical in the design process.
middle) was within 5% of the respective measured Indeed, due to the embedded PM topology used for the
values. The same good agreement was reported for rotor in this application, for these motors, the thickness
the windings and stator locations, showing a clear of the bridge must be thick enough to withstand the
anisotropic behavior of the motor mainly due to the stress due to the combined forces and yet adequate
gradual heating of the cooling fluid mixture (50%-50% also from an electromagnetic perspective to allow
water and ethylene glycol). easy saturation of the material. These two possibly
conflicting requirements must be balanced in such a
In some PM electric motors (such as in embedded way that both constraints are satisfied. In Fig. 31, the
PM rotor configurations) operating at high speeds, stress and deformation distribution for such a rotor
analyzing the stress and deformation due to a operating at 10,000 rpm is shown.
18
Fig. 31 Deformation (left) and von Mises stress (right) distributions in PM rotor
under combined electromagnetic and centrifugal force distributions
ePhysics also can be used to determine the stress an example of such an application that provides the
distribution for a given placement of the heavy displacement distribution of the frame only when
components inside the drive cabinet. Fig. 32 presents moved by an overhead crane in the assembly plant.
19
ELEctRIc DRIVES SUB-
SYStEM/SYStEM-LEVEL
DESIGN
The usage of system-level simulation for electric drives ease of use and needed accuracy of the system-level
simulation cannot be overestimated. The current analysis of electric drives. The following paragraphs
level of complexity in this area combined with the provide a few examples along the above lines.
increasing need for accuracy in the simulation makes
the availability of system components with field effect The example presented below shows the usefulness
more than a trendy technology; it clearly defines a of having a field solver and a system-level simulator
viable market. Thus, the combination of Maxwell working together to simulate the over-voltages
and Simplorer, which allows the creation and use of occurring in a low-voltage drive for a PWM driven AC
equivalent components with field effect, becomes a very motor. The model chosen for the electric cable is a
useful tool in the hands of the system design engineer. low loss model, which does not take into account the
Adding the possibility of using the dynamic coupling distributed (transversal) conductance of the cable. It
between Maxwell and Simplorer allows users to get a does take into account the distributed self and mutual
full picture of the multi-dimensionality of the coupling inductance capacitance of the cable. Resistance may also
between the two desktops. be included in the equivalent model. Using a number
of such “cells,” a model with distributed parameters is
For drives applications, it becomes increasingly included in Simplorer as shown in Fig. 33 (only 5 “cells”
important to model in an efficient but also accurate way shown) in such a way that the actual length of the cable
the complexity of many applications in this area. The is reflected in the model. The motor model used in this
equivalent models of transformers, inductors, bus-bar application corresponds to a 10 HP, 460 V induction
systems and motors—to mention just a few, all available motor.
from the Maxwell desktop—provide the convenience,
Fig. 33 Example of a building block in the Simplorer application for the study of the
consequences of the reflected waves
The combined effects of the PWM of the applied the impedance mismatch with the induction motor
voltage, the distributed nature of the cable model, model, are the reason for the over-voltage predicted by
which exhibits a certain characteristic impedance and the Simplorer simulation. (Fig. 34)
Fig. 35 Inductance matrix (self and mutual inductances) of the power transformer
The voltage amplitude and phase shifts must be application and not incur excessive losses. The overall
closely controlled to achieve best-input harmonic drive model is presented in Figs. 36 and 37 and contains
results. The task here is to obtain a transformer that the transformer equivalent circuit (inductance matrix)
has suitable properties to function properly in this models of the IGBTs and the control logic.
Fig. 36 System-level model of the drive, sub-sheet containing power transformer and rectifying bridge
AcKNOWLEDGEMENtS
Special thanks go to Dr. Leon Voss, Richard Osman,
PE, and Johannes Germishuizen, of Siemens AG, who
provided, for many applications presented here, models
and supporting material property and measurement
data used in putting together this document.
Copyright © 2007 Siemens and Ansoft Corporation TPS105-0207