Table of Contents

DESIGN, MANUFACTURE, AND TESTING OF A HIGH SPEED 10MW

PERMANENT MAGNET MOTOR AND DISCUSSION OF POTENTIAL APPLICATIONS

by
James S. Smith
Director, Technology Integration
DRS Power and Control Technologies, Inc.
Fitchburg, Massachusetts
and
Andrew P. Watson
Senior Design Engineer
Elliott Company
Jeannette, Pennsylvania

ABSTRACT
James S. Smith is Director, Motor Center The paper will present the design, manufacturing, and testing of
of Excellence, for DRS Power and a high speed 10 MW permanent magnet motor and discussion of
Controls Technology, Inc., in Fitchburg, potential applications. The paper will cover from concept through
Massachusetts. His career has been design and development to the conclusion of prototype testing. As
centered on development and integration part of the concept phase, potential applications will be briefly
of technologies to improve power plant discussed that were used to justify the project. These applications
system performance in nuclear powered include replacing variable speed gas turbines, steam turbines, and
submarines and industrial (petrochemical) wound field, synchronous, and induction motors, as well as fixed
applications. Mr. Smith’s experience spans speed electric motor/gear drivers. The mechanical and electrical
the design, construction, test, and delivery of challenges for high speed, high power permanent magnet motors
all nuclear propulsion plant equipment in will be presented along with the engineering developments that
delivered submarines from the early Los Angeles and Ohio Class met these challenges. The paper will also touch on the various
ships to the present Virginia Class, the LSV2 experimental manufacturing and assembly hurdles needed to be overcome to
submarine electric drive system, and the development and bring the design to fruition. The paper will then present the
installation of the largest electric motors and solid state, variable prototype test data to show the design targets have been met or
frequency drive systems (three systems rated at 60 MW each per exceeded. Finally, a discussion on the extension of the prototype
facility) in the liquefied natural gas industry. Mr. Smith named as design to much higher power at lower speeds and to higher speeds
sole or co-inventor, has been awarded over 24 United States Patents. at the same or lower powers will be presented.
Mr. Smith holds a B.S. degree (Physics) from Rensselaer
Polytechnic Institute.
INTRODUCTION
Gas and steam turbines are routinely selected to drive industrial
Andrew P. Watson is Senior Design compression and other process equipment, and have established
Engineer for axial and single-stage well-known performance standards in terms of efficiency,
centrifugal compressors at Elliott availability, emissions, and power density. Interest in the
Company, in Jeannette, Pennsylvania. He is application of electric motors to these services has increased in
responsible for mechanical design, recent years, as documented in papers presented in past Texas
aerodynamic performance, stress analysis, A&M Turbomachinery Symposia. A key performance shortfall in
and rotordynamics of these product lines. the implementation of electric motors in such applications is the
In addition, he has been assigned special requirement to drive typical gas compression loads through a
projects related to multistage compressors speed increasing gear, as traditional motor designs have been
and steam turbines, including seal design, limitedin rotating speed to 3600 rpm in applications where the
rotordynamics, balancing, AIGV design, power is provided at 60 Hz and 3000 rpm in areas served by 50 Hz
and design and implementation of magnetic bearings. Mr. Watson’s distribution systems. Typical industrial compression equipment
current project is the design of a permanent magnetic electric operates at speeds significantly higher, necessitating the use of a
motor, and he holds a U.S. Patent related to permanent magnet speed increasing gear.
motor design. He has been with Elliott Company since 1981 and Permanent magnet (PM) motors cannot be powered directly
previously had eight years’ experience in turbomachinery design “across the line” using a fixed frequency power source such as the
with other companies. local power grid. Specific controls are needed to match the power
Mr. Watson has a B.S.M.E. degree (1973) from Pennsylvania being supplied to the motor, the motor’s response, and the load to
State University and an M.S.M.E. degree (1981) from Rensselaer ensure stable operation. These controls are readily implemented in
Polytechnic Institute. He has authored several technical papers, is variable frequency drives (VFDs). The required use of VFDs
a member of ASME, and a registered Professional Engineer in the imposes a cost premium; however, it also eliminates the input
State of Pennsylvania. power frequency limitation of 50 Hz or 60 Hz. The effect of
19
 Table of Contents
20 PROCEEDINGS OF THE THIRTY-FIFTH TURBOMACHINERY SYMPOSIUM • 2006

enabling increased input power frequency is to enable a wider (RTDs) to completely map the core and end turn temperature
range of electromagnetic configurations. Some of these new profile while also recording air flow. The high magnetic flux levels
choices can be exploited to enable improved rotordynamic per- and frequency, combined with high velocity air flow made the air
formance and to increase the direct drive speed range of PM flow measurement goal impractical. As a result, only the detailed
motors. core temperature profile was available to validate the stator and
PM motors can also be designed to provide a wide operating rotor thermal performance predictions. Figure 1 shows the
speed and load range with extremely high motor efficiency, predicted rotor temperature profile, which ranges from a high of
typically in excess of 90 percent from less than 50 percent speed 150ºC (302ºF) at the center to nearly 70ºC (158ºF) at the ends. A
driving a typical pump or compressor load to a peak range in similar temperature range was predicted for the stator, with the high
excess of 97 percent for high speed applications. For the purpose approaching 170ºC (338ºF) the center. These values were within the
of this paper, high speed is defined at rotor peripheral speeds in rotor magnet temperature limit and the stator insulation system
excess of 180 meters per second (m/s). material design limits, so the design was approved for manufacture.
The focus of this paper is a flexible permanent magnet machine
design, capable of achieving operating rotor peripheral speeds of
up to 250 m/s that is axially scaleable from a low power rating of
approximately 4 MW to a high of approximately 32 MW. The
design was prototyped and tested at a design power rating of 10
MW, running at 6225 rpm. Basic features of the initial design
objectives were to use a totally enclosed, water to air cooling
(TEWAC) system typical of the petrochemical/industrial motor
applications, a Class H rated insulation system with a Class F
design temperature rise limit, and voltage range, depending on
rating, from 4160 volts to 5500 volts maximum. The actual
performance achieved was a power rating in excess of 11 MW.

STATOR DESIGN AND DEVELOPMENT

The basic electromagnetic design was highly dependent on the
structural and rotordynamics analyses performed on the rotor to
determine the lowest practical pole count that could be achieved at
the 250 m/s peripheral speed. This value was determined to be eight
poles, which, at 6225 rpm, corresponds to a fundamental operating
frequency of 415 Hz. A basic departure from higher power rated Figure 1. Predicted Rotor Temperature Profile from Thermal
machinery, which has historically been limited to 60 Hz or lower at Model.
power levels of greater than several MW, this fundamental
frequency increase imposed specific material and manufacturing ROTOR DESIGN AND DEVELOPMENT
method challenges. Chief among them were the selection of litz The biggest distinction between induction or synchronous
wire for the stator core conductors, and relatively thin stator electric motors and a permanent magnet motor is the rotor design.
laminations that were specially processed to ensure no burrs The stators for all these machines are basically the same in design,
remained from the punching process. The litz wire was needed to material selections, and manufacturing. The PM machine rotor
reduce the losses from skin effects due to the higher fundamental carries permanent magnets while all other motor types use
operating frequency, and the thinner lamination steel, and processing, electricity running in loops on the rotor to develop the rotor
were selected to reduce eddy current losses in the stator core. magnetic field. This key distinction is one of the features of a PM
The bulk of the electromagnetic design was conducted using machine that can be exploited to enable motor configurations that
traditional, empirically validated electric motor design methods. are not practical using other motor technologies.
These methods were supplemented with detailed electromagnetic Until now permanent magnet motor applications have been
finite element analyses (EM FEA) where necessary, and, to generally limited to low power (<500 kW, most under 100 kW),
develop a soon to be empirically validated EM FEA model of the high speed service. Recently additional industrial application in
basic geometry that could be used for other rotor diameter and low speed, high torque services have been proposed. Most of
axial length versions of this basic design. these existing PM motors use surface mounted permanent magnets,
Loss predictions were made using several numerical methods, meaning the magnets are placed on the rotor, which is ferromag-
which were calibrated using data from other machines and specific netic material (steel) with their direction of magnetization aligned
supplemental test data where needed. A thermal model of the stator radially on the shaft. This is one of two design approaches used to
geometry and materials was built using internally developed attach magnets to rotors. The surface mounted magnets mostly use
modeling tools. Computational fluid dynamics (CFD) analysis an adhesive to attach the magnets to the surface and a high strength
methods were used, employing the same geometry from the fiber wrap around the outside of the magnets to keep them in place.
thermal model, to predict cooling air distribution in the machine, This restraining membrane is typically either a high strength nickel
which was then coupled to the thermal model to predict the stator based alloy (low conductivity and very thin to reduce eddy current
and rotor temperatures under steady-state and transients. heating) or a composite material (carbon fiber, for example). The
A fundamental difference from other electric motor types, a PM second method to attach magnets to a rotor is called an embedded
motor can be designed to have essentially negligible rotor losses, design. Here the magnets are held to the shaft by some mechanical
eliminating it as a heat source requiring a specific means for arrangement. Prior methods of embedding magnets typically relied
cooling. Consequently, in the subject design, no specific rotor on the pole pieces, which limits speed and can restrict the magnet
cooling provisions, other than a fixed rate of flushing the air from circuit so that the full benefit of the magnets is not realized.
the air gap, were found necessary. The goal of the project described in this paper was to develop a
To validate these computer modeling efforts, and reduce the risk design that addresses the limitations of prior art and significantly
of their application to extended versions of the prototype machine, extends the speed-power capability of permanent magnet motors.
it was originally intended to install flow measuring devices in The project would encompass a line of motors that extend from 5
strategic locations as well as 114 resistance temperature detectors MW to 120 MW and 3000 rpm to 15,000 rpm. The first frame size
 Table of Contents
DESIGN, MANUFACTURE, AND TESTING OF A HIGH SPEED 10 MW 21
PERMANENT MAGNET MOTOR AND DISCUSSION OF POTENTIAL APPLICATIONS

to be developed, with a mid-sized rotor diameter of 27 inches, is The laminations are made up of thin sheets of the selected material
designated Cypress and covers up to 32 MW and 6200 rpm. The and joined together by bonding or other means to form the required
prototype selected to prove-out the Cypress frame size design, and size lamination. Laminating is required in electrical devices to
the basic electromagnetic, structural, mechanical, and thermal minimize the heat generating effects of eddy currents. Typical
approach to be used for the other larger and smaller frame sizes, electric motor laminations are made from silicon-iron, which give
were chosen to be a 10 MW, 6200 rpm, machine. The Cypress excellent magnetic properties. The mechanical strength, however,
frame is currently being extended to 7000 rpm in Phase II of is too low to be used for the rotating laminations on this project. An
product development. The following presents the design, develop- iron-cobalt lamination material was chosen because it has high
ment, and manufacture of the prototype rotor goals. strength along with the required electromagnetic properties.
Because of the inherent rotor heating and structural challenges
of a surface mounted magnet design, it was decided to pursue an • Magnet carriers—Because of the inherent low strength of the
magnet material, a special holder is needed to support the magnet
embedded magnet design. The design was also required to
and transfer the magnet’s load to the shaft. The required material
overcome the shortcomings of prior embedded designs. This
not only needs to be nonmagnetic but also have high electrical
dictated no bolted pieces holding the magnets, no nonmagnetic
resistivity to minimize eddy current heating of the parts. The
parts in the magnetic circuit, and no magnetic materials outside the
magnet holders also need to have high strength to weight ratio to
magnetic circuit that could short circuit the flux path. Also parts
outside the magnetic circuit needed to have low electrical conduc- support the magnets while minimizing adding additional load to
tivity to minimize eddy current heating. the shaft. The selection was a titanium alloy that fit these criteria.
The major steps in the development of this high speed, high • Magnets—Rare earth magnets are the magnets of choice for
power rotor design were: permanent magnet motors. The two rare earth magnet materials
available are samarium cobalt and neodymium iron boron. For this
• Select speed, maximum power, and dimensions for initial frame application neodymium iron boron (NdFeB, or NIB) was chosen
size design due to its better magnetic and mechanical properties. NIB magnets
• Select materials and test to optimize properties offer the largest energy product of any magnetic material available
below 200ºC (392ºF). Energy product is the field intensity (H) of
• Preliminary design to perform mechanical and EM FEAs and the magnet times the flux density (B) of the magnet. H is measured
conduct mechanical tests to validate predictions
in oersteds and is analogous to voltage in an electric circuit. B is
• Finalize prototype design and release for build measured in gauss and is analogous to current in an electric circuit.
The product of the B*H is given in megagaussoersteds (MGOe).
• Manufacture and test prototype unit The magnets are made by compressing and sintering a powdered
• Optimize final design and release for production mixture Nd, Fe, and B. The downside of compacting and sintering
rare earth powders is that the magnets are mechanically weak,
Initial Size and Speed Selection
especially in tension. This challenges the design of the magnet
The initial rotor size was based on applications currently using carriers to support the magnets in such a way as to ensure the
midsize gas turbines. These type applications have speeds up to magnet stresses are mainly in compression, with as little bending
6200 rpm and power requirements up to 32 MW. The initial elec- or tension as possible.
tromagnetic design goal was to attain a given shear stress in the air
gap. Criteria also required a minimum number of poles to Design
minimize stator losses. The same electromagnetic analysis deter- The design goals for the first frame size to be developed were
mined the preliminary magnet size. After several mechanical determined by looking at the most common driver power and speed
design iterations, the stage length and diameter of the rotor were for the market this PM motor is intended. Based on that study the
chosen to meet these criteria. The mechanical analysis included following goals were established for the rotor design:
building and exercising mechanical FEA and rotordynamics
models. The preliminary FEA showed the initial frame size rotor • Maximum continuous speed: 6200 rpm
diameter limits the minimum number of poles to eight. • Minimum per stage power of 1.7 MW
Materials • Maximum number of stages per motor to be at least 17
Having sized the rotor, the next step was to specify the materials • Meet at least the following criteria:
for each part that meet the electromagnetic and mechanical • 20 percent overspeed capability (NEMA MG 1, 1993)
requirements for the design. Because of the nature of the applica- • Minimum of 5000 starts/stops (API 546, 1997)
tion the materials used had to have specific mechanical and
electromagnetic properties. The material requirements and selec- Initial Design
tions for each part of the rotor are as follows: The initial eight pole design is shown in Figure 2. The initial
design consists of a vertically split titanium ring that captures the
• Shaft—Because the main shaft body carried the load of the magnets beneath the bridge connecting the two rings and it
rotating parts as well as being the means of transmitting torque, it
had to have unusually high tensile and yield strengths, supple- straddles the laminations. The first choice to make in designing the
mented by unusually good fracture mechanics features, such as motor was the number of poles. Since the number of poles is
elongation and brittle fracture resistance. Because it was in close directly proportional to frequency for a given speed, the goal was
proximity to the magnets, it also had to be nonmagnetic to prevent to restrict the number of poles to limit the operating frequency of
short-circuiting the magnetic flux path. These restrictions limited the motor to under 500 Hz. Preliminary FEA work showed for the
materials to some combination of nickel, cobalt, titanium, rotor diameter, carrier/magnet, and lamination sizes, and speed
chromium, and other nonferrous materials. To limit costs it was chosen for the design a minimum of eight poles was required.
decided to use shrunk-on AISI 4340 steel for the stub shafts since Eight poles running at 6200 rpm require a frequency of 413 Hz.
those areas did not need to be nonmagnetic. Figure 3 shows the increase in normalized stress as the number of
poles is decreased. The initial magnet size was a first pass estimate
• Laminations—Laminations are materials of high magnetic per- by the electromagnetic designer. To expedite the project, project
meability that are placed in a magnet circuit to ìurgeî the majority leaders decided at the beginning to do the electromagnetic and
of the flux from the magnets to follow the design magnetic circuit. rotor mechanical design work concurrently as much as possible.
 Table of Contents
22 PROCEEDINGS OF THE THIRTY-FIFTH TURBOMACHINERY SYMPOSIUM • 2006

The decision meant the rotor mechanical design would forge ahead Final Design
based on the initial magnet size estimate. The plan for the mechan- The FEA work supported by the spin test showed the initial
ical design was to complete detail development work including design was marginal for overspeed capability. At the same time the
FEA and build a single-stage rotor to perform straingauged spin rotor spin test was being performed, the electromagnetic FEA was
tests including spinning to destruction. The design was completed, completed and showed the magnets had to be considerably larger to
the single-stage rotor built, and the mechanical tests completed. A attain the goal of 1.7 MW/stage. The rotor design was changed,
picture of the single-stage spin rotor is shown in Figure 4. moved the highest stresses from the ring to the shaft, and enabled
the design to work with the larger magnets and have a comfortable
margin of safety. The individual carrier design is shown in Figure 5.

Figure 2. Exploded View of Initial Design.

Figure 5. Final Desgn Exploded View.

Prototype Rotor Manufacture

Having finalized the electromagnetic and mechanical aspects of
the rotor design, work began on building a 10 MW prototype motor
to test. The most unique, and therefore significant, challenge
imposed in building the prototype rotor was handling the magnets
during the rotor assembly. The magnets were large and powerful
requiring care and safety at every point in the process.
Special magnet handling techniques, fixtures, and procedures
had to be devised. It was decided early that the magnetic flux would
be contained at every step in the process of rotor assembly after the
magnets were initially magnetized. This required the magnet
carriers along with necessary clamps be sent to the magnet vendor.
The assembly fixture was designed to capture the magnets
before the clamps were removed, and steel shunts were installed to
reduce the magnitude of the flux field as the magnets approached
and were installed into the slots in the shaft. The laminations were
slowly lowered into place between the magnets using a hydraulic
Figure 3. Normalized Shaft Stress Versus Number of Poles. cylinder to prevent impact damage with the magnets. The complete
ring assembly was then pushed axially onto the shaft using a
specially designed sizing fixture. A picture of the prototype
assembly is shown in Figure 6.

Figure 4. Single-Stage Spin Rotor. Figure 6. Prototype Rotor Assembly.

 Table of Contents
DESIGN, MANUFACTURE, AND TESTING OF A HIGH SPEED 10 MW 23
PERMANENT MAGNET MOTOR AND DISCUSSION OF POTENTIAL APPLICATIONS

TESTING
Mechanical
The first mechanical test of the project was a straingauged spin
test of the initial design. As discussed above, it was found before
the test that the design would need to change. As with any FEA, the
quality of the results is directly proportional to the accuracy of the
model. Since the PM rotor model was an assembly of different
materials, it was decided to go ahead with the spin test to correlate
the FEA work with the straingauge data. The results of the tests
showed an excellent correlation between FEA and straingauge data
as shown by two typical gauges plotted against the FEA prediction
in Figure 7.

Figure 9. Prototype Rotor Vibration Probe Data.

Initial phases of the motor test program involved essentially
routine electrical tests to establish the machine’s initial
conditions and compare them to the design predictions. A test
spinning the motor to operating speed and recording the terminal
voltage readings was the first, most significant test. The design
voltage of 4160 VAC was slightly low, as the readings at typical
room temperature were approximately 4100 volts. This, along
with resistance and inductance readings confirmed the basic
electromagnetic design had been reproduced in real materials.
The program budgets prevented a long term loaded test program
at 10 MW using a conventional load, such as a water brake or
compressor load. Consequently the prototype machine was wound
as two independent three-phase machines, with the second three-
phase winding offset to allow the machine to be run as a six-phase
motor or generator, or allow the machine to be run as a 5 MW
motor driving a 5 MW generator in the same electromagnetic
Figure 7. Results of Single-Stage Spin Test. envelope. A relatively small, under 500 hp, machine was built to
test this concept and verify we could fully load the machine. Figure
The rotordynamic analysis correlated very well with the actual 10 shows a schematic of the loaded testing set up. Loaded testing
results of the six-stage prototype motor test. Figure 8 shows the was initiated and proved the machine met the design rating of 10
predicted first unbalance response and Figure 9 the actual rotor MW with significant margins, with the highest internal stator tem-
vibration results. The actual vibration level was much lower than perature reading at 152ºC (305.6ºF) at a load of 11.1 MW. This
the API requirement of a maximum of 1.4 mils at 6200 rpm. This stator temperature corresponds to a predicted maximum rotor
was a result of the rigorous balance criteria applied to the rotor. temperature of approximately 130ºC (266ºF). Both of these
Because the prototype design did not allow for center plane temperatures are well within the thermal limits of the design.
balancing, the shaft was balanced at speed prior to assembly. This
allowed for both static and dynamic corrections at the end and
midplanes. The carrier/magnet assemblies and laminations were
moment weighed. A special computer program was developed to
pair up carrier and lamination assemblies and locate them on the
shaft to give the minimum residual static and dynamic unbalances.
The complete assembly was then balanced at speed with minor
corrections on the end planes. A center plane balance will be
incorporated into the production version of the motor.

Figure 10. Prototype Test Schemaatic for Loaded Testing.

Once the proof of concept testing was complete, a number of

special tests to validate some of the claimed benefits of PM
machines were conducted. One such claim, that PM machines are
not limited by thermal considerations in restarting when hot, was
addressed by simulating a casualty where the cooling fans are
Figure 8. Rotordynamic Analysis of Prototype Rotor. disabled, the machine is allowed to operate until the hot spot
 Table of Contents
24 PROCEEDINGS OF THE THIRTY-FIFTH TURBOMACHINERY SYMPOSIUM • 2006

temperature increases to 105 percent of its starting point, and then of the machine once the transient was initiated by shutting the fans
the machine is tripped, allowed to coast to a stop under load, left off and leaving the load applied. In Figure 13 the fans are once
for a few minutes without the fans running, then the fans are turned again turned off with the motor operating at rated load and speed,
back on, and moments later the machine is restarted under load. and then the drive is tripped. The motor comes to a stop without
This entire set of transients was recorded using the 114 RTDs re-energizing the fan motors. After several minutes with motor
embedded in the stator, and the results are provided in Figures 11 stopped the fans are turned back on, and then, several minutes later,
through 13. the motor is restarted under load and ramped up to rated speed and
load in 120 seconds. It is operated at rated speed and load until
thermal equilibrium is approached.
The highest temperature reached in the machine, 158.6ºC
(317.48ºF), occurs, as expected, when the machine is run at rated
load and speed without the fans for several minutes (this occurs
twice, and the second time the period of operating without the fans
was much shorter so the highest temperature reached was
somewhat less). In no other period of the transient performance
testing did the temperatures even rise above the steady-state
thermal equilibrium values. These data show the PM machine is
essentially unaffected by a momentary loss of cooling, and can be
restarted immediately after a shutdown without any extra thermal
stress or imposing a waiting period for the machine to cool off.

CONCLUSION
Figure 11. RTD Recording Showing the Highest Stator Temperatures
at Thermal Equilibrium. The overall test results, summarized in Table 1, showed the
design met or exceeded all of its design mechanical and electrical
performance requirements with substantial margins. While no
specific acoustic goals were imposed on the design, the airborne
acoustic performance has been shown to be unacceptable “as-is.”
This aspect of the design is being evaluated and corrective actions
are being implemented for the prototype, which will be tested in
June of 2006, and all future designs will be based on the Phase II
improved rotor configuration that reduces the source of the noise
as well as increases the speed rating from 6225 rpm to 7000 rpm.

Table 1. Summary of Test Results.

Figure 12. RTD Temperature Measurement Recordings During

Extreme Operational Transient with Cooling Fans Shut Off and
Motor Operating at Rated Load and Speed (Maximum
Temperature Reached was 158.6ºC [317.48ºF]).

Based on the testing to date the benefits of PM machines, which

range from improved efficiency, reduced size and weight, greater
reliability, and available to improved safety and greater adaptability
to driven equipment, appear to be significantly closer to becoming
commercially available in the power and speed ranges needed for
industrial and petrochemical processing.

REFERENCES
API Standard 546, 1997, “Brushless Synchronous Machines—500
Figure 13. RTD Temperature Measurement Recordings kVA and Larger,” American Petroleum Institute, Washington,
D.C.
During Simulated Emergency Shut Down with Cooling
Fans Shut Off While Operating at Rated Load and Speed NEMA Standard Publication MG-1, 1993, “Motors and
Generators,” National Electrical Manufacturers Association,
Followed by a Restart of the Cooling Fans and the Motor.
Washington, D.C.
The data from the RTDs embedded in the hottest part of the
stator, along with some distributed in cooler areas, are presented
as Figure 11, as the machine reached and then stayed at thermal
equilibrium for a reasonable period of time. Data from the same
RTDs are shown in Figure 12, which displays the thermal response