Magnetic Bearing For Bently Nevada RK4 Rotor Kits
Magnetic Bearing For Bently Nevada RK4 Rotor Kits
Magnetic Bearing For Bently Nevada RK4 Rotor Kits
Since this project is a result of a class assignment, it has been graded and accepted as
fulfillment of the course requirements. Acceptance does not imply technical accuracy or
reliability. Any use of information in this report is done at the risk of the user. These risks may
include catastrophic failure of the device or infringement of patent or copyright laws. California
Polytechnic State University at San Luis Obispo and its staff cannot be held liable for any use or
misuse of the project.
Senior Project Final Design Report:
Magnetic Bearing for Bently Nevada RK4 Rotor Kits
Design By:
Sean Fowler
[email protected]
Garrett Olson
[email protected]
1
Table of Contents
Page
Safety Warning……………………………………………………………….. 4
1. Introduction………………………………………………………………….. 5
2. Objective……………………………………………………………………… 5-6
3. Background…………………………………………………………………… 6-7
4. Design Requirements….…………………..………………………………… 7-8
5. Management Plan……………………………………………………………. 9
6. Design Analysis………………………………………………………………. 10-30
6.1. Static and Dynamic Load Capacity
6.2. Actuator Generated Magnetic Force
6.3. FEA Analysis
6.4. Heat Transfer
6.5. Physical Constraints
6.6. Physical Parameter Optimization
6.7. Design Results
7. AMB Design………………………………………………………………… 30-37
7.1. Stator and Rotor Mass Model
7.2. Stator Case Model
7.3. Control System Model
8. ADAMS and MATLAB Co-Simulation……………………………………. 37-47
8.1. ADAMS Model Setup
8.2. Constraints and Applied Motion
8.3. Force and Stiffness Interactions
8.4. State Variables
8.5. MATLAB Modeling
8.6. Simulation Procedure and Results
9. Design Hazard Identification……………………………………………… 48
10. Bill of Materials…………………..…………………………………………. 48-49
11. Electronics Assembly……………………………………………………….. 49-52
11.1. Controller and Data Acquisition Board
11.2. Power Amplifier and Power Supply
11.3. Proximity Sensor and Amplifier
12. Manufacturing and Construction…………………………………………… 53-60
12.1. Laminated Stator and Rotor Mass
12.2. Stator Case
12.3. Magnetic Coils
12.4. RK4 Rotor System
12.5. Power Electronics Enclosure
13. Wire Diagrams……………………………………………………………….. 60-64
13.1. System Schematic
2
13.2. Servo Drives to Power Supply
13.3. Servo Drives to NI FPGA Controller
13.4. Servo Drives to AMB
14. Servo Drive Operation……………………………………………………… 64-71
14.1. General Inspection
14.2. Tuning Procedure
14.3. Block Diagram
14.4. Response Results
15. LabVIEW Real-time Simulation………………………………………….. 72-74
16. Testing and System Identification……………………………................... 75-77
16.1. Test Results
16.2. Observations
17. Recommendations…………………………………………………………. 78
18. References………………………………………………………………….. 79-80
19. Appendices
19.1. Appendix A- QFD
19.2. Appendix B- Codes and Standards
19.3. Appendix C- Two Plane Rotor Model
19.4. Appendix D- Stator/Rotor/Coil Design Calculations
19.5. Appendix E- Servo Drive (Amplifier) Model
19.6. Appendix F- Control Model
19.7. Appendix G – ADAMS Model
19.8. Appendix H – FMEA
19.9. Appendix I – Polar Laser Laminations (Stator and Rotor) Receipt
19.10. Appendix J – NI cRIO-9063 Receipt and Recommended Reading
19.11. Appendix K – Part Drawings
19.12. Appendix L – Concept Sketch
19.13. Appendix M – Gantt Chart
3
ATTENTION READER:
Since you are reading this, you are likely the inheritor of the hardware associated with
this senior project. As such, it is paramount that you understand the following for your own
safety:
1. This magnetic bearing system is not a trivial piece of equipment. It uses high voltage,
high current electronics, which have the potential to expose the user to deadly electric
shock if the device is misused.
2. Never operate the device without first reading and understanding the full operating
procedures of this report.
3. All electrical components and wires must be physically avoided during operation. When
adjusting electrical components outside of the lockable black box, all input signals must
be turned off and the power supply plug disconnected. After this is done, energy must
also be allowed to leave the system; there is about a 30 second delay that must be
waited between when the input signal is set to zero and the output control current in
the actuator coils also falls to zero. This current should always be monitored during use.
4. The lockable black box should remain locked at all times to prevent accidental exposure
to electric shock and remind users of the danger.
5. When opening the lockable black box, the system should always be powered down and
unplugged with the one exception being when adjusting the servo drive potentiometers
or switches. A minimum of one hour must be waited after disconnecting the power
supply and turning off input signals before opening the box. This is to allow the large
blue capacitor to fully discharge. (4RC = time for full capacitor discharge. A 30 kOhm
bleed resistor across the power supply rectifier = R. The capacitance is 0.024 F.
4 x 30,000 x 0.024 = 2,880 sec = 48 mins for full discharge. Rounded up to one hour for
safety.) When adjusting servo drive potentiometers and switches, great care must be
taken so as not to expose oneself to electric shock by bumping any other components in
the box.
6. The rotor kit motor is capable of rotating the shaft at 10,000 RPM. At this speed, debris
or untightened screws can result in serious user injury, particularly to unprotected eyes.
7. While safety measures have been put into place, we do not pretend to have created a
foolproof device by any means. As such, great care, forethought, and critical awareness
are vital to operating the device safely and effectively.
These are the most crucial safety precautions that must be taken when operating the device.
Please consider this section as a stern warning of the unique dangers associated with this
project.
4
1. Introduction
Active Magnetic Bearings (AMB) are contact free bearings that support loads by
magnetic levitation. This is accomplished by generating magnetic forces with electric current
through a series of electromagnets surrounding a suspended rotor mass. Along with a set of
electromagnets, an AMB assembly also consists of power amplifiers for each electromagnet, a
controller, and proximity sensors. The proximity sensors provide rotor position feedback to the
control system, which modulates power to the amplifiers [4].
Magnetic bearings are used in several industrial applications today, including but not
limited to compressors, turbines, pumps, motors, and generators. Advantages of their utility
include very low to no friction, longer life than bearings with lubrication (no wear), high
operation speeds, environmental friendliness (oil and contamination free), and the ability to
accommodate irregularities in the mass distribution automatically. Disadvantages include
complexity, little to no damping, difficulty to control, and high cost. Magnetic bearings are
finding increasing use as the technology progresses and components becomes less expensive.
(a) (b)
Figure 1.1. (a) Industrial full-size radial magnetic bearing. (b) SKF AMB assembly in an industrial
water pump.
2. Objective
For the purpose of this project, an AMB was to be sized and retrofitted to a Bently
Nevada RK4 rotor kit used in the vibrations and mechanical controls labs. These kits are used
for undergraduate and graduate courses as teaching and research tools in the areas of rotor-
dynamics and active machine control. The production of this AMB is the first step in a
conceived series of projects that will look to continue to improve the efficiency, size, and
performance of this technology. A number of companies and research labs also stand to
5
benefit from such a product, as SKF is currently the only company which sells magnetic bearings
compatible with the rotor kits.
The specific goal of this senior project was to manufacture one magnetic bearing and
integrate it with the necessary electronics hardware so that it would be capable of sending and
receiving control data to and from MATLAB. The system was to be designed for future use by
students as a teaching and research aid. This complements the Cal Poly Engineering
Department’s creed of “Learn by Doing” and will provide students with the opportunity to learn
unique skills in the fields of mechatronics, rotor dynamics, and controls.
3. Background
There have been a few universities and independent parties that have successfully
designed and manufactured a small scale AMB, however the technology is still relatively new
with much room for improvement. University of Patras, Greece [1] designed an AMB for the
same model rotor kit (the RK4) as used at Cal Poly (Figure 3.1). The bearing controller utilizes
the PID control method, implemented using MATLAB software. This method provides good
robustness and stability as long as the operating point is inside the linear performance range of
the AMB. PID control requires linearization, therefore the non-linear equations of motion of
the rotor were linearized with coupled x-y equations of motion. PID gains were derived
through numerous simulations and the resulting control resulted in satisfactory damping and
stiffness of the overall system. The rotor itself was modeled as rigid with one mass balance
plane located near the AMB housing.
Figure 3.1. AMB designed by the Machine Design Laboratory of the Department of Mechanical
Engineering and Aeronautics, University of Patras, Greece.
An independent group from the Republic of Korea [3] provides a successful roadmap for
the analysis and design phases of the AMB. The “probable flux paths method” is introduced,
6
offering fundamental insight for electromagnet development. This general method assumes: 1)
linear behavior between magnetic flux and current, 2) average magnetic flux passes through
the centroid of the cross-section, and 3) varying cross-section sizes are treated as separate
pieces in series or parallel. Analysis begins with determining the mass of the levitating object or
the effective mass of the rotor. The electromagnet core and coils are then sized to achieve the
attractive force required to lift the shaft and provide feedback. The next steps include
electronics hardware and control system design.
Schweitzer [2] provides a complete roadmap of the AMB design process, providing
extensive information on rotor system modeling, necessary hardware components, actuator
design, potential losses in the AMB, and active control techniques. Rotor system modeling
consists of accurately determining the characteristic equation of the rotor system and
predicting the vibratory response induced by a disturbance force (mass imbalance during
rotation). Load capacity is determined from the maximum static and dynamic loads
experienced by the rotor and is then used to size the iron core of the AMB (magnetic force
ranges). Apart from the iron core, additional components with guidelines for sizing are
introduced. Parasitic losses from eddy currents, hysteresis, and induced heating effects are
considered when determining geometry and selecting materials. Active control techniques vary
in complexity and include the PID method, H-infinity, and mu-synthesis. The proper method is
chosen based on the degrees-of-freedom of the AMB and how many inputs/outputs are being
controlled simultaneously.
Due to an AMB’s complexity and precision, proper safety guidelines must be established
and followed for the design and implementation of an AMB. Failures can occur mechanically,
electronically, and in the software. Schweitzer [2] offers insight in regards to safety practice
and proper procedures. Standardized procedures as described in the ISO 9000 series are highly
recommended. A company or establishment following these procedures can be recognized as a
certified institution with a defined quality level. Furthermore, the ISO 14839 1-4 series should
be followed as it lays out design guidelines pertaining to mechanical vibrations in rotating
machinery equipped with AMB’s.
4. Design Requirements
Discussions with the sponsor, Dr. Wu, resulted in a list of design requirements for
meeting the goal and scope of the project. The size and function of the RK4 rotor kit was
already established, providing constraints that had to be considered while designing the AMB.
With the inboard end of the rotor fixed axially by the motor, a 2 degree-of-freedom AMB was
the logical choice. This constrained the control methodology to SISO (single-input/single-
output) as opposed to MIMO (multiple-input/multiple-output). MIMO is more commonly
implemented for multiple AMB systems with more degrees-of-freedom. The design
7
requirements were divided into customer requirements with corresponding engineering
requirements (Table 4.1).
1. Has to work with existing rotor -One magnetic bearing to support one shaft end
kits -Can maintain controlled levitation up to max speed
that the vibes lab rotor kits run at (0-3,000 rpm
typically, Max 10,000 rpm)
-Stator outer diameter cannot exceed 12 cm
5. Safe for use by students -No touchable component should exceed 100°F
(below the 111°F threshold for 1st degree burns) [6]
-Appropriately sized guards shall prevent the user
from touching the AMB during operation
-Minimize voltages wherever possible
-Adhere to electrical safety guidelines presented in
class
8
5. Management Plan
Due to the small size of the project group, each of the members were intimately
involved with each aspect of the project. However, some of the tasks were distributed
according to the unique strengths of each member. Sean led the manufacturing effort due to
his prior project experience producing various machined parts. He saw to the final design and
implementation of the AMB stator, rotor mass, and casing as well as the touchdown bearing.
Sean was also responsible for the coils windings and their installation on the AMB stator.
Garrett led the FEA electromagnetics and rotordynamics analyses used for determining the
design specifications of the AMB. He performed the controls analyses in MATLAB and ADAMS.
Finally, Garrett saw to the design and implementation of the power electronics that included
the wiring and enclosure. A third and very important addition to the group, Cameron Naugle
(who was a graduate student working for Dr. Wu), saw to the selection and implementation of
the NI cRIO-9063 FPGA controller and LabVIEW real-time control program that took the place of
the originally conceived Simulink compatible controller mentioned in the customer
requirements.
The senior project design process flow chart for the year is included below in Figure 5.1
to provide an outline of when various tasks were performed. We worked with a slightly
accelerated schedule, in that our detailed design work for the bearing assembly was performed
in the fall. Design work for the electronics systems and controls was the primary focus during
winter quarter.
The first step in designing the AMB was to determine the magnetic force that the
bearing had to produce to levitate the rotor. This involved an analysis of a two-plane mass
rotor system suspended between two bushings. A theoretical model in MATLAB was developed
and tuned to match the experimental results. The rotor model consisted of two similar masses
suspended on a shaft at different lengths from each end of the rotor at L1 and L3 (Figure 6.1a).
Experiments were also conducted using ADRE 408 for data-acquisition. The rotor was
attached to an inboard motor with a maximum operable speed of 10,000 rpm. The rotor was
operated at speeds ranging from 245 rpm (slow roll) to 7,000 rpm with a ramp rate of 5.24
rad/sec2.
(a) (b)
Figure 6.1. (a) Case 1: Rigid shaft, flexible bearings. (b) Case 2: Rigid bearings, flexible shaft.
A theoretical model was developed to validate the experimental results utilizing the
Superposition Method [Notes: Rotordynamics, Wu] where the shaft stiffness matrix was
determined. Spring stiffness values for two cases, 1) a rigid shaft with flexible bearings and 2) a
flexible shaft with rigid bearings (Figure 6.1) were determined and combined in parallel to
produce an equivalent system stiffness matrix. The equations used to develop these matrices
are as follows:
10
2) Flexible Shaft with Rigid Bearings
𝐿1 2 (𝐿2 + 𝐿3 )2 𝐿1 𝐿3 (𝐿2 − 𝐿1 2 − 𝐿3 2 )
𝐶𝑠 = 3𝐸𝐼𝐿 6𝐸𝐼𝐿
𝐿1 𝐿3 (𝐿 − 𝐿1 2 − 𝐿3 2 )
2 2
𝐿3 (𝐿1 + 𝐿2 )2
[ 6𝐸𝐼𝐿 3𝐸𝐼𝐿 ]
3) Total Stiffness
1
𝐾=
(𝐶𝑏 + 𝐶𝑠 )
The Ka and Kb values represent the estimated stiffness of the inboard and outboard
bushings, respectively, and were set at 290,000 N/m. Damping was estimated at 0.0003*K.
The masses included in the mass matrix were those of the disks, each having the same value of
0.5 kg. The mass, stiffness, and damping values were implemented in matrix format:
An expanded format of this equation was created in an ODE45, MATLAB function file
that was used to generate the displacement amplitude of each disk over a specified length of
time:
An eccentricity of 3 mils and a phase of 0 rad and π/8 rad were used for disks 1 and 2,
respectively. These values, along with the estimated stiffness of the inboard and outboard
bushings, were tuned until the theoretical and experimental responses matched. Figure 6.2
shows the responses of each disk ranging from 0 to 4,000 rpm and 0 to 10,000 rpm. Both the
experimental and theoretical models indicate a first natural response at 1,100 rpm and have
similar response profiles.
Once validated by the experimental results, the theoretical model simulation was
extended to 10,000 rpm to view potential responses occurring at the operating limit of the RK4
system. It should be noted that experimental data was cut off above 50 mils (peak to peak)
which is the maximum expected displacement during safe rotor use. In this case, the shaft’s
deflection was physically limited by impacting the proximity sensors. The amplitude response
of the theoretical model predicted the extended measurable response of the actual model.
11
Figure 6.2. Top: MATLAB and Experimental results with matching natural frequencies and
amplitude widths. Bottom: Extended MATLAB model between 0-10,000 rpm with amplitude
response approaching 100 mils.
12
The displacement amplitudes of both masses were used to find the maximum dynamic
forces experienced by the rotor. Figure 6.3 shows the dynamic forces experienced by each
bushing with speed ranging from 0-10,000 rpm. Both the static and dynamic forces of each
disk at their respective locations were combined, and the moments experienced by the inboard
and outboard locations were used with the shaft length to determine the loads at each fixed
end.
Figure 6.3. Total force reactions at the inboard and outboard bushings.
These calculated load capacities are a high estimation of what may be experienced
during typical rotor kit use. Appendix C provides the MATLAB code used in the load capacity
analysis and may be tuned for different rotor system setups. The bushing in the place of the
AMB provides a good initial estimate of rotor performance and bearing stiffness values that
may be used to size the AMB. Once built, stiffness in the AMB may be adjusted by adjusting the
bias current of each actuator pair, as described in the following sections.
13
Table 6.1. RK4 system parameters and expected load capacities experienced at the outboard
location.
L1 (m) 0.197
L3 (m) 0.2
Total Length (m) 0.53
Mass of Disk (kg) 0.8 per disk
Rotor Mass (kg) 0.5
Omega (RPM) 300-3000
Static Load Capacity (N) 14.3
Dynamic Load Capacity (N) 62.4
Total Load Capacity (N) 76.7
With the maximum force required for stable operation calculated, the next step was the
mechanical design and sizing of the AMB. The analysis performed follows the models
developed in [3] and Chapter 3 of both [2] and [7]. These begin with the force produced by a
single magnetic actuator, and then develop into models for two magnetic actuators acting in
opposition to each other. This is what is found on 2-axis AMB's. One pair of actuators must be
used to balance the rotor mass for each axis because magnetic actuators can only provide an
attractive force. The total force on the rotor provided by the opposing electromagnets is
linearized to include a current “stiffness” and displacement stiffness. The analysis assumes that
the cross-sectional area of the back iron (Figure 6.4) is equal to that of the poles and rotor iron
such that a uniform cross-sectional area is maintained throughout the entire magnetic circuit.
It is also assumed that the air gap between the stator and rotor is small compared to the stator
and rotor dimensions.
14
c
The magnetic actuator can then be modeled as a magnetic circuit (Figure 6.5) with
magnetic flux (Φ) as the analog for current, total reluctance (Rt) for resistance, and
magnetomotive force (MMF) as voltage, such that
𝑀𝑀𝐹 = Φ𝑅𝑡 .
𝐿
𝑅 = 𝜇𝐴
L is the mean length of the path traveled by the flux and A is the cross-sectional area of the
material. Because the permeability of free space (μ0=4π x 10-7 [N/A2]) is on the order of 1,700
to 6,000 times smaller than the permeability of the silicon steel or “iron” of the stator and
rotor, the reluctance of the stator and rotor can be neglected so that the total reluctance is
given by
2𝑠
𝑅𝑡 = 𝜇 𝐴 .
0
Note that A in this case is technically the cross-sectional area of the air gap through
which the flux passes. Since fringing (where some flux is allowed to spread out and escape in
making the transition across the air gap) is assumed to be negligible, and since the cross-
sectional area of the material is kept constant through the flux path, A remains equal to the
cross-sectional area of the stator pole (A = bc, as shown in Figure 6.6 below). Since
𝑀𝑀𝐹 = 𝑁𝑖,
where N is the number of coil turns of the electromagnet and i is the current flowing through
the coils, the magnetic flux in the circuit can be written as
Ni𝜇0 𝐴
Φ= .
2s
and can be used to calculate the attractive force produced by the magnetic actuator,
𝐵2 𝐴
𝐹= .
𝜇0
16
*Note, this equation has an error in [7]. The 4 in the denominator is given as a 2. The equation presented below is
confirmed in [2] and [3].
In a form that is more useful for designing and which includes a correction factor, ε, to account
for fringing and leakage losses, the force equation becomes*
𝜀𝜇0 𝑁 2 𝑖 2 𝐴
𝐹= .
4s2
To extend this model to a two actuator system, the total force becomes
𝜀𝜇0 𝑁 2 𝑖 2 𝐴 𝑖1 2 𝑖 2
𝐹𝑡𝑜𝑡 = 𝐹1 − 𝐹2 = [𝑠 2 − 𝑠2 2 ] .
4 1 2
The linearization of this equation begins by defining 𝑖0 as the base current or bias
current and 𝑖𝑥 as the perturbation current. The minimum perturbation current is zero while the
maximum perturbation current is equal to the base current such that imax through an actuator is
equal to 2𝑖0 . Since for the force equation above, F1 > F2, likewise i1 > i2, so i1 and i2 may be
defined as
𝑖1 = 𝑖0 + 𝑖𝑥 ,
and
𝑖2 = 𝑖0 − 𝑖𝑥 .
17
Similarly, with 𝑠0 defined as the nominal air gap present when the rotor is centered and 𝑥
defined as the displacement of the rotor from its center position, 𝑠1 and 𝑠2 can be defined as
𝑠1 = 𝑠0 + 𝑥,
and
𝑠2 = 𝑠0 − 𝑥.
Using Taylor series expansion, the force equation can be modified to take the form
𝐹𝑡𝑜𝑡 = 𝐾𝑖 𝑖𝑥 − 𝐾𝑠 𝑥
𝜀𝜇0 𝑁 2 𝐴𝑖0
𝐾𝑖 = cos(𝛼),
s2
𝜀𝜇0 𝑁 2 𝐴𝑖0 2
𝐾𝑠 = − cos(𝛼) .
s3
Here, the term cos(𝛼) has been added to account for the geometry of a two pole
magnetic actuator where the centerlines of the poles are separated by an angle of 𝛼. For an 8
pole magnetic bearing with the poles spaced equally around the rotor, 𝛼 = 22.5°.
Generated magnetic forces and magnetic flux densities were analyzed utilizing the FEA
method in Abaqus. Magnetic forces by each of the 8 poles were induced by an applied current
to each actuator that acted perpendicular to an applied current density field inside solid copper
windings. The expected force for a full scale AMB model was 36 N given the designed geometry
and bias current. By scaling down the stator and rotor mass to 1/7 the thickness, the expected
generated force became 5 N. This was validated through a convergence study that averaged
the body forces of elements contained in each pole region.
For the purpose of this analysis, material properties were considered for M19 silicon
steel (stator and rotor) and air. Copper properties were not included but should be considered
for a future magnetic/thermal coupled simulation. Table 6.2 provides the material properties
used as well as magnetic force parameters.
18
Table 6.2. Material properties for M19 Silicon Steel and air as well as the magnetic force
parameters governing the analysis.
19
Figure 6.7. Mesh density fields of stator, poles, and air gap.
The FEA model consisted of partitioned sections representing the stator, copper
windings, rotor mass, and air. All partitioned sections were mated together to form one solid
piece. Within each main partition, sub-partitions were created to allow for linear quadratic
elements to be used during the mesh process. The geometry of the assembly prohibited the
use of global seeds, therefore sets of edges were created to define the varying seed sizes.
Figure 6.7 shows the congruency of each partitioned section during meshing. The smallest
elements were in the air gap between the poles and rotor mass. Elements at the outer edge of
the rotor mass were made the same size as those of the air gap.
The rotor mass itself was divided into three rung sections with the largest edge bias
being towards the outside. Much of the generated magnetic flux during simulation was
expected towards the outer edge of the rotor, therefore smaller seed sizes were used in this
rung that lined up with nodes of the air gap elements. The seed sizes were made considerably
larger moving towards the center of the rotor in order to reduce the number of elements
during simulation.
Figure 6.8 and Figure 6.9 give snap shots of both the generated magnetic body forces
and magnetic flux. Results in both figures involved a total of 250,000 elements. Initially 1.1e06
elements were used, however this greatly increased the simulation run time. The magnetic
body forces between each pole and rotor mass were averaged, producing results that matched
those of Table 6.2.
20
Figure 6.8. Force produced by the magnetization of each pole was compared to theoretical
values for validation of the finite element model.
Stator/Rotor Interactions
Figure 6.9. Magnetic flux density field with equally applied current density to copper
surrounding all 8 poles of the stator.
21
Magnetic flux leakage occurring at the pole corners and air gaps is also evident in Figure
6.9. A fringing and leakage coefficient of 0.8 [2], [3] was assumed throughout the design
process. For better accuracy, future FEA analyses may be performed that may quantify this
coefficient through magnetic flux averages in various regions between the stator poles and air.
Furthermore, a thermal/electromagnetic analysis could be performed to estimate the effect of
heating from the copper coils on the magnetic permeability of the stator and rotor.
The FEA study provided a proof of concept and led to our final decision of going with
M19 steel. For this study, 250 coils per actuator were used, however our final design came to
include 300 coils. Due to time constraints, no further analyses were performed, however it is
encouraged to make use of this existing approach moving forward to better quantity system
characteristics for more precise applications.
Heat transfer was also considered throughout the iterative process. Two methods of
heat transfer analysis were included in the MATLAB based design analysis (see Appendix D).
The first took an elementary physics approach and calculated the amount of time required to
increase the temperature of the coils and immediately adjacent core material by 30°F
(presumably from a room temperature of approximately 70°F to the design limit of 100°F). This
analysis assumed, based on the limited convection and conduction coefficients suggested by
[8], that all of the resistive power losses dissipated by the coils would be stored by the coils and
the core material surrounded by the coils as thermal energy. This helped to produce some
intuition as to how rapidly the assembly would heat up.
The second analysis followed more closely to the 1-D heat transfer analysis outlined in
[8]. This method calculates the heat transfer away from the coils assuming a 30°F temperature
difference between the coils and the surrounding air and between the coils and the back iron of
the stator. The calculated heat transfer rate could then be compared to the power dissipated
due to resistive losses.
Both of these methods involved simplified versions of the relatively complex stator
geometry, and can only be taken as ballpark approximations. Analogies to similarly sized
hardware were sought so as to provide better intuition regarding how dissipated electrical
power produces detectable temperature rises. An energy efficient laptop computer, for
example, dissipates 15-25 Watts into a confined space with only modest temperature rises.
This is achieved in part due to a small cooling fan. This analogy sheds some doubt on the
results of the second heat transfer analysis, which indicated that a 30°F temperature difference
between the coils and their open surroundings would only result in the transfer of 5.4 Watts.
Intuitively, given the open nature of the bearing assembly and the large heat sink provided by
the stator, the 5.7 Watts calculated to be dissipated by the coils would likely be transferred to
the surroundings with only a minimal rise in overall AMB temperature. This analysis did
22
demonstrate the utility of minimizing current through the coils, which was taken into
consideration when designing other parameters. Fortunately, this analysis does tell us that the
system ought to be easily cooled using a small computer fan providing forced convection over
the stator.
With the design force equation developed, work then began on determining the
geometry and design parameters of the magnetic bearing. A number of constraints were
determined based on the geometry of the RK4 rotor kit and recommendations made by various
research groups and authorities on magnetic bearing design. In the following analysis, a safety
factor of 3 was used for the design force as recommended in [3]. The outer diameter of the
stator was also constrained to a maximum of 12 cm so that it could be made to fit within the V-
shaped channel of the RK4 rotor kit.
According to [2], the most significant AMB power losses are due to the resistance of the
copper coils. Since wire resistance is inversely proportional to wire cross-sectional area, using
the largest possible wire diameter for the coils would minimize resistive losses. However, the
wire diameter also have to be small enough to allow the coils to be easily wound around the
stator poles and to allow the necessary number of coils to be wrapped around a given pole
without making the poles so long as to exceed the 12 cm limit on the stator outside diameter.
In [8], AWG 18 copper wire (1.02 mm diameter [9]) was used for a similarly sized AMB. By
inspection, this appeared to be the largest wire diameter that would appropriately meet the
described constraints, so it was selected as an upper limit for the analysis. Once the electrical
hardware design began, it became clear that limiting the current provided to the actuators
would be desirable in order to be able to use more available, lower cost power electronics
components. Between AWG 18, 20, and 22, gauge 22 wire was selected. Reducing the wire
diameter allows more windings to be added to each pole, which allows the same design force
to be applied using less current. The tradeoff is that this also reduces the force slew rate (see
Figure 9), which theoretically lowers the maximum operating speed. Gauge 22 wire should still
allow a maximum controllable rotor speed of 10,000 RPM. The reduction in current also meant
that the design adhered to another rule of thumb presented in [8], which is to keep the current
density within the copper coils below 6 [A/mm2] to avoid overheating.
This parameter is also relevant to the selection of the bias current. [7] recommends
beginning by setting the base current equal to the current level that would produce a flux
density equal to one half Bknee, or the value on the stator and core material's B-H curve which
marks the upper limit of the linear range (Figure 6.10 below). Grade M19 silicon steel was
chosen as the rotor and stator material based on manufacturer recommendations [10]: "M19 is
probably the most common grade for motion control products, as it offers nearly the lowest
core loss in this class of material, with only a small cost impact, particularly in low to medium
23
production quantities." The silicon in "silicon steel" is alloyed with the steel to increase the
material's resistance. This reduces losses due to eddy currents forming in the material, which
are induced by the oscillating magnetic field. M19 silicon steel is commonly used in electric
motor laminations. It is also a readily available stock material that most lamination
manufacturers keep on hand, meaning its use should aid in limiting lead times during the
manufacturing stage of the project. From data provided by [11], Bknee for M19 silicon steel is 1
Tesla. This was used as an initial rule of thumb for determining 𝑖0 such that
𝑠0 𝐵𝑘𝑛𝑒𝑒
𝑖0 = .
𝜇0 𝑁
As the analysis progressed, Bknee was replaced by scalable parameter "Bmax", or the
maximum flux density produced by the design. During the design process, this served as a
method to reduce 𝑖0 and therefore minimize resistive losses without drastically changing the
format of the MATLAB analysis (see Appendix D). Bknee was then treated as a constraint on Bmax
such that
Bmax ≤ Bknee .
Here, 𝑣𝑚𝑎𝑥 is the maximum voltage capable of being provided by the amplifier connected to
each actuator. Multiplying this value by Ki gives the force slew rate,
𝑑𝐹 2𝜀𝑖0 𝑣𝑚𝑎𝑥
𝑚𝑎𝑥 | 𝑑𝑡 | = .
𝑠0
The force output of one of the opposing actuators for an axis has a sinusoidal nature in order
keep the rotor balanced at its center position [7]. This force output as a function of time can
therefore be modeled to take the form
𝑓 = 𝑀𝑠𝑖𝑛(𝜔𝑡)
𝑑𝑓
| 𝑑𝑡 | = |𝑀𝑐𝑜𝑠(𝜔𝑡)𝜔|.
𝑑𝑓 𝑑𝐹
is limited by 𝑚𝑎𝑥 | 𝑑𝑡 |, meaning that the magnitude |𝑀|of the output force is limited such
𝑑𝑡
that,
2𝜀𝑖 𝑣
|𝑀| = 0 𝑚𝑎𝑥 .
𝑠 𝜔 0
Therefore, the bias current and voltage output of the amplifier must be high enough so
that the magnetic actuators can provide the design force throughout the entire range of
operating speeds, 𝜔. This is best verified through a log-log plot of the max load capacity vs.
frequency. The intersection of the calculated design force and maximum force as a function of
frequency can then be used to determine whether the system designed is capable of operating
at the speed desired. Such a plot is presented by [7] and included below. During our
conversation with Dr. Zhu from Solar Turbines, he estimated that about 80V, would be required
for our application. As he described, this is because the inductance of the system should be
considered the real driver for determining voltage. The load capacity plot, Figure 6.12, has
been included for a max voltage of 72 Volts, which would theoretically provide a maximum
25
slew-rate limited operating speed of over 10,000 RPM. The real limit on speed then becomes
the frequency at which the controller can run the PID control.
Figure 6.11. Log-log plot of the load capacity vs. frequency from [7].
Figure 6.12. Log-log plot of the load capacity vs. frequency using design values. Note that the
maximum force output is still possible up until the max speed of 10,000 rpm (1,047 rad/sec).
26
6.6. Physical Parameter Optimization
Once as many meaningful constraints were placed within the model as possible, several
key variables were modified iteratively to come to a working design. These variables are
presented in the following table. While more sophisticated techniques for optimizing the design
parameters are available and described in [8], such methods were not possible given the time
constraints on the project. These methods are also best utilized with more accurately
characterized heat transfer phenomena.
Table 6.3. Design parameters and their effect on the AMB system.
Design Parameter Description and Effect of Modification
Number of coils per actuator Increasing N increases actuator force output and resistive
(N) losses by increasing the total coil length.
Bmax ≤ Bknee . Decreasing Bmax decreases bias current and
Maximum flux density (Bmax)
resistive losses.
Increasing Vmax increases the force slew rate and therefore
Max amplifier voltage (Vmax) allows the system to run at higher rotational speeds.
Increasing Vmax also increases the risk of electric shock.
Decreasing s0 decreases the number of coils and the amount
of current required to achieve the design force. s0 is limited
Air gap (s0) by achievable machining tolerances. Based on the
capabilities of the Cal Poly machine shops and precedents
set by [3] and [8], an air gap of 0.5 mm was selected.
Increasing ppc increases the stator outer diameter. 2mm
Pole length past the coil (ppc) was deemed appropriate for preventing the coils from
slipping off the end of the pole.
Increasing nl minimizes the stator outer diameter, but can
Number of coil layers (nl) cause interference between the coils of adjacent poles.
Increasing nl also reduces losses due to leakage.
The pole width also affects the rotor and back iron
thickness (Figure 5). Increasing c increases the actuator
output force but increases stator and rotor outer diameter
Pole width (c) and increases the chance of interference between coils of
adjacent poles. A rule of thumb provided by [2] is to adjust
c to make 50% of the stator inner circumference consist of
pole material.
Increasing b increases the actuator output force but also
means that the rotor and stator laminations must be joined
Stator, rotor axial thickness (b)
more precisely to ensure proper alignment within the
limited air gap. b must also be limited so that when the
27
middle of the shaft experiences a deflection, that deflection
does not cause the inboard end of the rotor to hit the inside
of the stator before the shaft contacts the touchdown
bearing. This was approximated using similar triangles.
Increasing di increases the stator outer diameter but also
allows c and therefore the actuator output force to be
increased without causing adjacent coil interference. 23
mm was selected for di because it struck a balance between
Rotor laminations inner increasing the stator outer diameter and allowing enough
diameter (di) room for adjacent coils. It also allows for the use of the
same connectors that join the Bently Nevada RK4 shaft to
the Bently Nevada journal bearing rotor. Use of this
existing hardware would allow the AMB rotor location on
the shaft to be easily adjustable.
The table below includes selected and calculated parameter values that characterize the
design chosen. Geometric parameters are included and correspond to the figure below,
modified from [2].
Table 6.4. Design results from Appendix D analysis.
Parameter Description (MATLAB variable) Value
Coil Parameters
Number of coils per actuator (N) 300 coils
Magnet wire diameter (dcc) 0.668 mm (AWG 22)
Number of coil layers on one pole (nl) 4 layers
Maximum current density 6.66 A/mm2
Geometric Parameters
Air gap between stator and rotor (s0) 0.5 mm
28
Length of stator pole past the coil (ppc) 2 mm
Stator thickness (b) 35 mm
Pole leg width (c) 8.5 mm
Inner diameter of rotor laminations (di) 23 mm
Coil height (hc) 25.5 mm
Laminated rotor outer diameter (dr) 40 mm
Stator inner diameter (d) 41 mm
Stator outer diameter (da) 11.3 cm < 12 cm
29
The Appendix D MATLAB analyses produced the functional parameters of the AMB’s
mechanical system. A depiction of the resulting stator and rotor is displayed below.
7. AMB Design
With the stator and rotor dimensions determined, the assembly for mounting the
stator, sensors, and touchdown bearing could then be designed. The design illustrated below
makes use of the existing rotor kit bearing mount for the touchdown bearing and the sensor
mount for the proximity sensors. The rotor kit uses lubricated bronze bushings, mounted in the
bearing mount displayed below, to support the shaft. The touchdown bearing for the AMB
made use of the same mount. This bushing was created by drilling out the inner diameter of a
10 mm rotor kit bushing with a lettered Z drill bit, thereby increasing the inner diameter to 10.5
mm. This provided a clearance of 0.25 mm (0.25 mm < 0.5 mm air gap) between the shaft and
the bushing, which was thought to be able to prevent the rotor of the AMB from contacting the
inner diameter of the stator during startup and during loss of control. This is critical to
preventing damage to the rotor and stator. Utilizing the existing mounting hardware for the
touchdown bearing and sensors also allowed for a simple stator mounting case design.
30
Touchdown bushing mount
Trantorque
Coils
RK4 shaft
Figure 6.3. Left: Stator CAD without rotor. Right: Stator CAD with rotor.
31
7.2. Stator Case Model
The designed stator mounting case consists of three pieces. The primary load bearing
support piece is depicted on in Figure 7.4. The necessary complexity and tight tolerances
required that it be CNC machined. 6061-T6 aluminum was chosen for the material due to its
low cost, easiness to machine, and paramagnetic nature. The method for mounting the case to
the V-shaped channel of the rotor kit mimicked the existing bushing and sensor mounts. The
surfaces below the right slot in the figure align with two 45° chamfers on the edge of the rotor
kit channel so that, when screwed onto the channel, the face of the part aligns itself
perpendicularly to the shaft. The stator was designed to rest on two flat edges rather than one
curved edge in order to allow for easier shimming of the case should the case or stator not fall
within tolerance. Tolerances much better than 0.001" or 0.025mm are difficult to achieve with
the CNC mills in Mustang 60.
Figure 7.4. Stator case with stator (Left) and without stator (Right).
The rib with the semicircular cutout provides stiffness to the case while serving as an
alignment surface when installing the stator and clamping it in place with the case cover. Case
deflection was raised as a concern due to the small air gap between the rotor outer diameter
and stator inner diameter. Deflection in the case could change the stator's position relative to
the rotor and compromise the air gap. This concern was addressed with a simple analysis
where the case "base" (Figures 7.5-7.6) and "cover plate" were modeled together as a straight
beam with an equivalent area moment of inertia. The case's dimensions were simplified and
flattened as shown in the figures below to determine an equivalent length for the straight
beam of 150mm.
32
Figure 7.5. The combined force from the stator's weight and the max magnetic force acts
normally at the stator/case interface. Note that while the cover plate has been excluded here
for clarity, its area moment of inertia was included in the beam deflection equation.
240 N
F
Figure 7.6. The case geometry was simplified and then flattened into a straight beam. The
forces were then resolved into a single downward force of 240 N, which is equal to the
combined force from the weight of the stator and the max generated magnetic force.
−𝐹𝑙3
Using the appropriate equation from Shigley's Machine Design text [12], 𝛿 = 48𝐸𝐼 , the
maximum deflection, 𝛿, was calculated to be 0.012mm which constitutes 2.4% of the nominal
0.5mm air gap. While this uses the max design load with a safety factor of 3, this load is not
very realistic as the safety factor of 3 was chosen more as a way of accounting for losses when
calculating the bias current than due to uncertainty regarding the force required to levitate the
rotor. A more realistic max load is about 80 N. This produced a deflection of 0.004mm or 0.8%
of the air gap. This was expected to have little impact on the control of the system, and so the
case design was considered acceptable.
The case’s cover plate holds the stator in place axially while the “roof” secures it
vertically. With the bolts that hold down the roof and attach the case to the rotor table not
33
tightened down, there was designed to be a 0.5mm gap between the interface of the base and
roof of the case. This was to ensure contact between the top of the stator and inside of the
roof. These two parts were also CNC machined from 6061-T6.
Figure 7.7. 3 part stator case with and without stator and rotor.
The control schematics by Schweitzer [2] and the independent Republic of Korea group
[3] were followed in developing the closed-loop system for one electromagnetic actuator.
There are a total of four actuators surrounding the rotor: two opposing actuators in the y-axis
and two opposing actuators in the x-axis. Each actuator is supplied a current by a power
amplifier to induce a magnetic force on the suspended rotor. The power amplifier supplies a
bias current (1.2 Amps) that is set at startup. The current will increase or decrease according to
the rotor’s position relative to the shaft’s center position inside the AMB. A constant tug-of-
war results as each actuator attracts the rotor to the center position while the rotor’s rotational
speed is increased. Two proximity sensors, one for each axis, measure the rotor’s position and
provide feedback to the system’s controller. Deviations from the center position are corrected
by PID control. The corrected signal is then fed back into the amplifier, providing an up-to-date
current magnitude.
34
Figure 7.8. Control block diagram of one actuator with voltage input and displacement
feedback.
The power amplifier output signal (input to AMB) would ideally follow the input signal
to the power amplifier, however this is not the case. The inductance of the AMB coil will resist
any sudden changes in current, therefore fast current changes can only be achieved by a
suitably high internal amplifier voltage [2]. The coil current now becomes part of the system
dynamics, making it important to model the electrical properties of the bearing magnet and
power amplifier (inductance, reluctance, etc.). In this case, the power amplifier must be
modeled as voltage-to-voltage and not voltage-to-current. Figure 7.9 provides a general model
of a trans-conductance operational amplifier (voltage controlled) that will be incorporated into
the AMB system.
Figure 7.9. Voltage controlled power amplifier with output current and actuator load (L + R).
Current from the power amplifier will be supplied to the AMB actuator. Stiffness
coefficients Ki and Ks due to the input current and position of the rotor, respectively, are seen
35
in Figure 7.10. The mass used to model the AMB is the effective mass of the rotor at the
bearing location. The motion induced voltage coefficient Ku is fed back to the power amplifier.
A disturbance representing forces caused by mass imbalance of the rotor system is input into
the model. The greater the disturbance, the larger the current perturbation needed to bring
the rotor back to center.
Figure 7.10. Top: Control block diagram of voltage controlled operational amplifier. Bottom:
Control block diagram of AMB with current/force input and force disturbance from rotor
imbalance.
The following equation is used to describe the total voltage output of the amplifier that
includes the system dynamics [2]:
𝑑𝑖 𝑑𝑥
𝑢 = 𝑅𝑖 + 𝐿 + 𝑘𝑢
𝑑𝑡 𝑑𝑡
36
The coefficient ku describes the dynamic stiffness induced by the rotor mass within the
AMB while R and L are the resistance and inductance of the actuator copper wiring,
respectively. The largest contributor to voltage potential across each actuator is the coil
inductance, L. Schweitzer [2] and Yoon [7] estimate this value with the following equation:
𝜇𝑜 ∗ 𝐴 ∗ 𝑁 2
𝐿=
2 ∗ 𝑠𝑜
This equation produced a coil inductance of 35.3 mH for our system. With a maximum
slew rate of 2140 A/sec, the corresponding inductance requires 75.6 V. Based on these two
results, 80 V provided a safe estimate of the maximum voltage required for each actuator. Dr.
Lei Zhu, senior electrical engineer for Solar Turbines Inc.’s magnetic bearing team, estimated
the required voltage to be as much as 100 V due to the expected inductance and unanticipated
iron losses in our actuators. As a general estimate, Zhu uses the following equations for
required voltage capacity:
𝑉 = 𝐿 ∗ 𝑤1 ∗ 𝑖1 + 𝐿 ∗ 𝑤2 ∗ 𝑖2 + 𝐿 ∗ 𝑤4 ∗ 𝑖4
The “w’s” represent the first, second, and fourth harmonics of the operating speed, in
rad/sec. The “i’s” represent the bias current divided into thirds over each harmonic. In this
case, with a maximum bias current of 1.4 amps, i1 = 1 Amp, i2 = 0.2 Amps, and i4 = 0.2 Amps.
With an expected maximum operating speed of 10,000 rpm, this equation produces 81.4 Volts.
To avoid the issues associated with higher voltage, we reduced the maximum operating
speed to 6,000 rpm. This produced an expected voltage capacity of 48.8 Volts. Anticipating
additional eddy current and hysteresis losses at higher speeds, a safe design voltage capacity
was set to 60-80 Volts.
Disk 1 Bushing 1
37
AMB Stator
Rotor Disk 2
Mass
Bushing Bushing 2
Disk 1
Disk 2
Figure 8.1. Left: AMB and bushing rotor system. Right: Two bushing system.
A co-simulation between ADAMS and MATLAB was developed for the first system
containing the AMB. Relative displacement, velocity, and acceleration state variables were
defined inside the AMB that linked the stator to the center of mass of the “Rotor Mass”
element. These state variables characterized the AMB plant used in a MATLAB/Simulink AMB
system block diagram. Simulation results from MATLAB were imported back into ADAMS and
viewed as an animation during post-processing where the model could be further tuned and
adjusted.
SOLIDWORKS models of the steel rotor balancing disks and a shaft were created to
match the dimensions of the RK4 rotor kit components. Both disks were identical with an OD =
0.075 m, ID = 0.01 m, and Thickness = 0.025 m. They also contained 12 cut holes with Diameter
= 0.005 m used for adding balancing weights. The extruded steel shaft has a Length = 0.56 m
and OD = 0.01 m.
Figure 8.2. Left: Disk 1 and 2 used in AMB system, OD = .075m, Thickness = 0.025m. Right:
Steel shaft used in AMB system, length 0.56 m.
The AMB assembly from SOLIDWORKS was directly imported into ADAMS. Once in
ADAMS, the assembly was translated to line up along the global Z axis. Before any simulations
38
could be performed, a few system parameters were defined that include material definitions,
constraints, applied motion, force and stiffness interactions, and state variables. Figure 8.3
shows the locations of the AMB/bushing and suspended disks.
.05 m .135 m .15 m .1 m .125 m
Figure 8.3. Dimensioned ADAMS assembly starting from the left: Bushing, Disk 1, Disk 2, and
AMB with Rotor Mass.
All pieces of the AMB/bushing and two bushing assembly are made of steel with Young’s
Modulus E = 2.07e5 N/mm2, density ρ= 7.8e-6 kg/mm3, and Poisson’s Ratio ν= 0.29. The
material properties become very important when converting the steel shaft from a rigid to
flexible body.
The steel shaft had 152 nodes and 36 modes of vibration after being modeled as a
flexible body. The length of the shaft was aligned with the global Z axis and the center of its
diameter was set to 0 m in the global X and Y coordinates. Bearings in each model were locked
to the ground and the flexible shaft and had zero translational or rotational degrees of
freedom. The bushings in each model were given a lateral stiffness of 2e05 N/m with 0 N/m
axial stiffness and negligible damping. Each disk’s center of mass was offset from its geometric
center to induce an eccentricity. Disk 1’s eccentricity in global coordinates is 0 m (X direction)
and 0 m (Y direction). Disk 2’s eccentricity in global coordinates is 4e-4 m (X direction) and 0 m
(Y direction). The center of mass of each disk was locked to the flexible shaft and has zero
translational and rotational degrees of freedom.
A rotational joint was added to both models that connected the inboard of the flexible
shaft to the ground. A rotational motion with initial velocity of 4 rev/sec (240 rpm) and
39
acceleration of 35 rev/sec/sec was applied for a time interval 5 seconds, ramping the shaft to a
final velocity of 27 rev/sec (1620 rpm).
Applied current to the AMB stator induces an attractive force and a negative stiffness on
the levitated rotor mass. Both effects tend to pull the levitated mass to each actuator during
operation. Figure 8.4 shows the force and stiffness interactions acting on the center of mass of
the levitated rotor mass in the global X and Y directions. Four points attached to ground were
placed on the inner diameter of the stator along the X and Y axes and halfway between the
poles of each actuator. The other end of the force and stiffness interactions on the laminated
rotor mass were then anchored to these points.
The stiffness is a constant term defined by the bias current, nominal air gap, and
number of coils on each actuator. With a bias current of 1.2 Amps, the stiffness was 3.08e5
N/m. The applied forces in the X and Y directions vary with time during the simulation and
depend on the location of the rotor center mass with respect to the 4 inner stator grounded
points.
Figure 8.4. Left: AMB profile with springs setup in ± X and Y directions (-Ks). Right: Springs with
Force function defined in the ± X and Y directions.
40
Relative displacement, velocity, and acceleration between the center of mass of the
rotor and grounded stator points were defined. These variables were used in a
MATLAB/Simulink block diagram to accurately describe the motion of the suspended rotor
system during simulation. Figure 8.5 shows where the exported ADAMS plant is implemented
in the AMB system transfer function.
Figure 8.5. AMB system block diagram for 1 actuator with highlighted block of the imported
ADAMS rotor plant.
The force and stiffness calculations were conducted in a separate MATLAB function file.
Table 8.1 provides a list of AMB parameters, calculated stiffness terms, and amplifier circuit
parameters (servo drive parameters). The generated magnetic force (F_mag) varies during the
simulation as the rotor mass’ distance varies from each actuator. The Ks term is implemented
in the ADAMS model before the system plant is exported into Simulink. The Ku term was
neglected for this initial analysis but may be implemented if the amplifier dynamics are
considered in future analyses. The Ki term is a current stiffness that becomes the generated
magnetic force of each actuator when multiplied by the applied current.
41
Figure 8.6. Block diagram of Servo Drive implemented in the MATLAB/ADAMS co-simulation.
Included is the Gamp block with corresponding parameters for each of the circuit components.
Table 8.1. AMB and Servo Drive (Amplifier) parameters implemented in MATLAB function file.
Bias Current (Amps) 1.1671
# Coils (Per Actuator) 300
g_o (m) 5.00E-04
mu_o (Perm_Air) 1.25664E-06
Eps (Leakage + Fringing) 0.8
A_act (m2) 2.98E-04
Gap Factor 1.05E+00
A_gap (m2) 3.12E-04
F_mag (N) 3.67E+01
Ki (N/Amp) 1.32E+02
Ks (N/m) 3.08E+05
Ku (V/m/sec) 1.32E+02
Once all the state variables were fully defined, the control plant developed in ADAMS
was exported into MATLAB/Simulink (Figure 8.7). This became a Control_Plant_##.m file that
had to be run in MATLAB. The next step was to run “adams_sys” in the command prompt for
the .m file to become a Simulink block diagram (Figure 8.7). This new block diagram was then
copied and inserted into a developed AMB system block diagram that was then run from a
42
separate .m file containing the complete system parameters. A complete AMB system block
diagram can be viewed in Appendix G.
Figure 8.7. Left: Exported plant variables from ADAMS model to MATLAB/Simulink models.
Right: ADAMS control plant Simulink block diagram.
Displacement in the X and Y directions for Disks 1 and 2 as well as force vectors inside of
the AMB Stator and Bushing 2 were captured in Figures 8.8-8.11. Both simulations ran the
same length of time with identical eccentricity, velocity, and acceleration parameters. The two
suspended disks were also locked at the same locations.
The bushing/bushing model shows an increase in amplitude for Disk 1 as the rotor
approaches the first natural frequency. Interestingly, the AMB/bushing model shows an almost
constant displacement at Disk 1. Displacement values for Disk 2 in the bushing/bushing model
remain constant at a lower amplitude than a constant amplitude displacement in the
AMB/bushing model. Both Disks 1 and 2 hold the same displacement in the AMB/Bushing
model.
As can be expected, the reaction forces experienced by the bearings increases as the
rotational speed increases with time. The bushing/bushing model shows a maximum force = 27
N (Figure 8.10) after 5 seconds. This value supports previous analyses (two-plane superposition
model) used to predict expected loads while sizing the AMB actuators. The forces captured by
the AMB Stator in the X and Y directions are the applied forces to each actuator. The forces
remain close to constant magnitude (153-155 N) however their oscillating frequency increases
with the shaft’s rotation speed.
43
Time (sec)
Time (sec)
Figure 8.8. Top: Disk 1 X and Y displacement in Two Bushing Model. Bottom: Disk 1 X and Y
displacement in AMB and Bushing Model.
44
Time (sec)
Time (sec)
Figure 8.9. Top: Disk 2 X and Y displacement in AMB and Bushing Model. Bottom: Disk 2 X and
Y displacement in Two Bushing Model.
45
Time (sec)
Figure 8.10. X and Y force components in Bushing 2 of the Two Bushing Model.
Time (sec)
46
Time (sec)
Time (sec)
Time (sec)
Figure 8.11. Force components from Top, Bottom, Left, and Right Actuators respectively (top to
bottom).
The co-simulation results provide the user some extremely useful insight into modeling
a system when designing a controller. This model can be adjusted for different rotor system
parameters (location of the disks, eccentricities, length of rotor, etc.). Additionally, poles and
zeros created when adding filtering can be added to the system in either ADAMS or MATLAB.
Filters can remove outside disturbances (noise) and affect time delay after discretizing the
system for digital control.
47
9. Design Hazard Identification
Several potential hazards will exist during the testing and operation of the AMB
assembly and must be addressed with precaution and care. Hazards include 1) electrical wiring,
2) mechanical parts in motion at high speeds, and 3) heat dissipation by the iron core. The
upper limit of voltage consumption by the system is estimated to be 60-80V. Proper wiring
(placement and covering) of all wires will be critical for use by the user as well as limiting any
parasitic losses during operation. The typical rotor operating speed will be between 300-3000
rpm. Anytime a rotor is spinning there will be the chance of metal fragmentation from contact
interference or the assembly not being properly fastened and secured. During operation the
iron core and casing will increase in temperature due to generated heat by the copper coils on
each actuator. An operating time limit with corresponding running speeds may have to be
defined to maintain lower levels of heat generation.
A detailed Failure Modes and Effects Analysis (FMEA) is provided in Appendix H. This
explains all of the foreseen potential failures, their resulting outcome, and design constraints
implemented for prevention. Severity and occurrence scales have also been provided to show
the likelihood of these events as well as how great the consequences.
48
Trantorque couplers 2 No purchase $0
Sheet metal shims NA No purchase $0
Magnetic/Mechanical Subtotal: $2,227
Electronics Hardware
Total Line
Part(s) Quantity Unit Price
Price
GE Proximity sensors 2 No purchase $0
GE Sensor amplifiers 2 No purchase $0
National Instruments Real-Time Control Board
1 $1,246 $1,246
(cRIO-9063)
Advanced Motion Control 12A8 Servo Drives 4 $206.25 $825
Advanced Motion Control PS16L72 Power
1 $650 $650
Supply
AWG 18 Shielded electrical wire (gray) 1 $60 per 50 ft reel $60
PC fan 2 No purchase $0
Hall effect sensor and ±15V power supply 1 No purchase $0
Electrical enclosure/box 1 $150 $150
Electronics Subtotal: $2,781
Total Cost: $5,008
In selecting a controller and data acquisition system, it was decided to combine both of
these functions into one compact solution for convenience and feasibility. National
Instruments gave us the best option in meeting our needs by providing a high sampling rate
49
(>10kHz), programmable logic that interfaces with a computer real-time simulation (LabVIEW),
and an adequate number of input and output terminals (8 and 4, respectively). The selected
unit is the NI cRIO-9063 (Figure 11.1) which combines a dual-core processor, reconfigurable
FPGA (Field Programmable Gate Array), and four slots for input/output modules all in one
chassis. The system is well suited for advanced embedded control and monitoring applications,
thus perfect for operating an AMB. LabVIEW can be used to create, debug, and deploy logic to
the onboard FPGA. Table 11.1 provides the specifications for the chassis setup. Two NI 9215
modules were used to meet the requirement of 8 input channels. For more information
regarding these units, please consult the National Instruments website [18,19,20].
Figure 11.1. Left: NI 9063 FPGA Controller. Right: NI 9215 Simultaneous ± 10V Analog Input and
NI 9263 ± 10V Analog Output.
Table 11.1. Specifications of NI instrumentation for control and analog input/output voltage
signals.
NI 9063 FPGA Controller NI 9215 Analog Input NI 9263 Analog Output
Architecture ARM Cotex-A9 Signal Levels ± 10V Signal Levels ± 10V
Speed 667 MHz Channels 4 Differential Channels 4
Cores 2 Sample Rate 100kS/s/ch Sample Rate 100kS/s/ch
Flash Reboot
100,000 cycles Simultaneous Yes Simultaneous Yes
Endurance
Nonvolatile Memory 512 MB Resolution 16-Bit Resolution 16-Bit
Volatile Memory 256 MB Connectivity BNC Connectivity Spring-Terminal
Accuracy 5 ppm
Voltage Input Range 9-30 VDC
Max Power Input 18 Watts
50
11.2. Power Amplifiers and Power Supply
Voltage signals ranging from ±3.3V to ± 10V are sent from the controller into the power
amplifier. The amplifier’s parameters are sized according to a desired output, in this case 1-2
Amps. The chosen amplifier and power supply are shown in Figure 11.2. The performance
characteristics of the servo drives are shown in Table 11.2. More information regarding the
specifications of these units can be found on the Advanced Motion Controls website [12,13].
Figure 11.2. (Left) 12A8 analog servo drive. (Right) PS16L72 power supply.
Table 11.2. Advanced Motion Controls servo drive and power supply specifications
AMC 12A8 Analog Servo Drive AMC PS30A Power Supply
DC Output
Peak Current 12 A 72 VDC
Voltage
Max Continuous Output
6A 11 A
Current Current
Output
Supply Voltage 20-80 VDC 800 W
Power
Command
± 10 V
Voltage
Switching
36 kHz
Frequency
An important element to the performance of the overall AMB system is the type of
proximity sensor used to measure the placement of the rotor during operation. This sensor
must remain contact free during operation and provide accurate position feedback to the
controller. Typical sensors used in industry for this application are eddy current sensors. Eddy
51
current sensors work by creating a rapidly oscillating current in a sensing coil at the end of the
probe. This creates an alternating magnetic field that induces eddy currents in the target
material being measured. The induced magnetic field in the opposing target material will then
resist the field being generated by the probe, thus providing a measurable voltage potential.
Important parameters to consider when selecting a sensor include: 1) measuring range,
2) linearity, 3) sensitivity, 4) resolution, and 5) frequency range. Measuring range is the range
of distances that can be measured for which the relationship between distance and sensor
output voltage remains linear. Linearity shows to what extent the measured quantity will
deviate from a linear relationship between the measured quantity (mm) and output signal (V).
Sensitivity indicates the ratio of the output signal over the quantity to be measured (increment
size, V/mm). Resolution indicates how much useful signal from the output can be distinguished
from noise produced by the sensor amplifier system. Finally, frequency range indicates the
bandwidth of the linear frequency response that has a cut-off frequency of -3 dB. The chosen
proximity sensor assembly is Bently’s 3300 XL 8mm. Assembly and specifications are shown in
Figure 11.3.
Figure 11.3. Bently 3300 XL 8mm sensor probe and amplifier assembly with specifications
Bently’s product is common in industry today for monitoring rotating machinery and is
currently being used in the vibrations lab. Dr. Lei Zhu recommended a probe sensitivity of
around 5 mV for a magnetic bearing application. Bently’s probe sensitivity is slightly worse at
7.84 mV, but this is still very good for the size and scope of our system. More product
information can be found Bently’s website [21].
52
12. Manufacturing and Construction
12.1. Laminated Stator and Rotor Mass
Since Cal Poly does not possess the expertise to perform effective bonding of the
laminations, these services were sourced externally. Quotes from several vendors were
received. The least expensive quote was $1,650, from Polaris Laser Laminations based in
Chicago (the most expensive being over $13,000 from Wingard Hydroblanking in Baltimore).
Since Polaris Laser Laminations (PLL) also claimed the shortest lead time of about 4 weeks from
purchase completion to shipment, it was chosen as the manufacturer for the rotor and stator.
Gauge 29 (0.014") sheets of M19 silicon steel were laser cut and bonded together to make the
parts. Each layer of glue between the laminations is on the order of 0.0002" thick. Gauge 29 is
a typical gauge found in motor laminations and should be suitable for the speeds we expect to
operate at.
The received parts were within the ± 0.001" requested and no post-cutting operations
were necessary. Quotes for lamination thicknesses between 0.025” and 0.014” were requested
from PLL. There was a marginal increase in cost for thinner laminations, but thinner
laminations reduce losses due to eddy current formation in the material. This allows for higher
operating speeds to be attained. In accordance to advice given by Dr. Zhu of Solar Turbines,
the laminated rotor's axial length was slightly increased past the axial length of the stator to
account for fringing, or the spreading of the magnetic field within the air gap as the flux
transitions from the stator steel to the rotor steel.
Figure 12.1. Laminated rotor and stator. Rotor withdrawn from operating position for image.
53
Appropriately sized Tran-torque connectors from the Bently fluid film bearing kit were
used to connect the laminated rotor to the existing 10 mm rotor shaft. At the time of their
incorporation into the design, it was not realized that their steel construction would introduce
undesirable flux leakage from the laminated rotor into the Tran-torques. These were retained
in the final project as it was determined that their ease of use and ability to keep the laminated
rotor attached to the shaft with limited eccentricity was more valuable than the losses they
produced. Changing the shaft/laminated rotor coupling to a paramagnetic material like
aluminum would be a prudent future upgrade for when attempting to attain higher operating
speeds.
The stator case was CNC milled by Alec Bialek (Cal Poly shop tech) from 6061-T6
aluminum. Aluminum was used to prevent magnetic flux from leaking into the case since
aluminum has a much lower magnetic permeability than steel. The first “base” of the case had
to be scrapped due to a mistake when dimensioning the CAD file. It was assumed that the 10
mm rotor kit shaft rests centered between the threaded holes on either side of the V-channel.
As was revealed, there is actually a slight asymmetry and the shaft is biased to one side. This
was too big of an error to correct through shimming.
Producing the eight magnetic coils for the stator poles presented a challenge. The coils
had to be wound tightly and with an accurate number of windings (150 for each pole) for the
best performance. It was initially hypothesized that the lathe in Mustang 60 could be used to
wind the coils by attaching a spool of magnet wire to the auto-feed and running the machine at
low RPM. A fixture for holding the coil sleeve could be held in the chuck. While this would
likely have produced tight, precise windings, accurately counting the number of windings would
have been challenging. However, a quick YouTube search yielded several examples of
individuals easily producing tight windings by hand with a cordless drill with speed modulation.
55
Positioning
Pulley
ME 405
Microcontroller
Board
This simple setup was combined with a rotary encoder attached to the spinning coil
sleeve mold. This mold consisted of a rectangular piece of aluminum of the same cross section
as a stator pole with a 6-32 through tap at the rotational axis. This allowed a spindle to be
threaded into one end for the drill chuck to hold and a quarter inch shaft to be attached to the
other end. This shaft was connected to the rotary encoder via a shaft coupler to account for
the misalignment in the system. The encoder was held by a bracket mounted to the
mechatronics lab workbench. The drill was held inverted in the vice of the workbench. The
magnet wire reel was mounted to the back of the workbench via a pair of brackets, a wooden
dowel, and a friction plate to keep tension on the wire. A pulley with an adjustable position
was mounted at the front of the workbench near the encoder bracket. Springs on either side of
the pulley shaft provided a consistent starting position but also adjustability so that the user
could guide each subsequent winding down the length of the mold. The encoder was wired to
an ME 405 (Mechatronics) microcontroller. An existing ME 405 lab assignment was then
modified to output the revolution count from the encoder on a computer monitor so that the
user could keep track of the number of completed windings.
56
Figure 12.4. Left: Coil sleeve being wound with magnetic wire. Right: Magnetic coils fastened to
the laminated stator.
Free prototype coil sleeves had been requested early in winter quarter from the
Innovation Sandbox. These were intended to provide additional insulation between the stator
and the coils and also allow the coils to be transferred from their mold to the poles without
falling apart. The Sandbox’s filament deposition printer was claimed to have the precision
necessary to produce these simple parts, and, since they were printed from ABS plastic with a
melting temperature of about 220°F, heat deformation was not an issue. A mechanical failure
in the printer meant that those prototypes were never printed. Several wall thicknesses of the
sleeve design shown below were originally requested, but time constraints and high printing
demand meant that different thicknesses could not be experimented with. The first print was
the final print. Fortunately, the 0.8 mm wall thickness requested proved satisfactory. What
was not anticipated was the difficulty of preventing the wires from slipping off of the sleeve,
particularly when winding over previous layers of wire. This difficulty produced considerable
delays. Modifying the ends of the sleeve to be more like a spool would be prudent for future
iterations. Delrin stops were added to the winding mold to help overcome this design flaw. It
may have been prudent to seek out prototype coil manufacturers in order to reduce the time it
took to produce the coils, but this would also have introduced its own challenges with no
guarantee of less difficulty. Nevertheless, if future coils are to be made, it would be worth
investigating.
57
Figure 12.5. SolidWorks model of an ABS plastic coil sleeve, around which the magnet wires for
one pole were wound.
The AMB was mounted at the outboard end of the RK4 rotor kit assembly. Figure 12.6
illustrates the complete setup when running a two-plane mass experiment. Keeping with the
existing rotor kit configuration, the rotor shaft’s inboard end was connected to the drive motor
via a shaft coupler and supported by a bushing. Rotor disks can be moved along the shaft to
adjust the weight distribution and affect the displacement amplitudes. Additional balancing
weights can be added to each rotor disk to reduce vibrations from mass eccentricities. A three-
sided acrylic cover from Tap Plastics was mounted around the AMB to protect the user from
metal fragments and exposure to electrical wires while providing a minimally obstructed view
during operation.
The touchdown bearing and eddy current probes were mounted on the kit as close as
possible to the AMB on opposite sides. The eddy current probes measure the shaft position
next to the laminated rotor, rather than the laminated rotor itself. This was for convenient
sensor mounting and because eddy current probes typically need factory recalibration if they
are used with a different material than originally intended. The probes are offset from each
other axially to prevent signal interference between the sensors. It is imperative that the
touchdown bearing is as close as possible to the AMB in the event that control is lost by the
actuators. The touchdown bushing was designed to “catch” the spinning rotor, preventing any
damage to the stator poles by the rotor mass. To make this bushing, one of the brass, Bently kit
bushings was drilled out to allow the shaft to be displaced from its center position by about half
of the air gap between the stator and rotor before contacting the bushing. Using a simple,
similar triangles approximation, it was assumed that the laminated rotor would not be able to
contact the inside of the stator so long as the touchdown bearing block was positioned directly
against the base of the stator case. Upon experimentation, it was found that this assumption
58
was insufficient and did not adequately take into account the flexibility of the shaft. The
actuators are strong enough to pull the rotor mass against the inside of the stator. Once this
has happened, the magnetic force is so strong that it is difficult to free the rotor from the stator
without powering down the AMB system. You will notice that the touchdown bushing is held in
place by aluminum sheet metal shims. This was to provide more rigidity to this bushing and
help prevent the rotor from being pulled against the stator. The other bushings use O-rings as
an interface between the brass bushing and the aluminum bushing holder. This helps deal with
shaft misalignment and provides damping, but this also allows the bushing to move within the
housing, which is undesirable for the touchdown bushing.
The small, controller connection box is a junction between the actuators and the servo
drives supplying the current. This is attached to the RK4 base with Velcro and can be moved
along the length of the rotor with the AMB for different configurations. The wire connection
design from the actuator coils to the connection box is intended to allow a Hall-effect current
probe to be easily inserted or removed from any of the four actuator circuits. Individual female
connectors were used and surrounded by heat shrink to provide isolation from accidental
arcing in the case of a voltage spike. Connections on the inside of the connection box are
similarly insulated from each other with hot glue covering the solder points.
Rotor Disks
Inboard Bushing
AMB
Motor
r Controller Connection
Figure 12.6. Left: RK4 rotor kit assembly with mounted AMB. Right: Profile of AMB mounted
between a touchdown bearing and control sensors.
The servo drives and power supply are housed inside of an electronic enclosure set next
to the RK4 rotor assembly. It has holes in the front and back for the servo drive/controller
59
wires and power wire, respectively. A vent and two fans were installed on the ends to allow for
transient cross-flow of air if the system is operated over extended periods of time. Figure 12.7
shows the electronic enclosure next to the RK4 system (Left) as well as the inside with the servo
drives mounted and wired to the power supply (Right). Due to the high voltage and close
proximity of the electronic enclosure to the user, a lock is engaged to keep the lid of the
enclosure closed during operation.
Figure 12.7. Left: Electronic enclosure with wire connections to the AMB and NI FPGA
Controller. Right: Inside of electronic enclosure with the four servo drives mounted to the
power supply.
A schematic of the AMB core and electrical system is provided in Figure 13.1. There is
an X and Y axis probe receiving displacement signals that travel to their respective
amplifiers. These sensor amplifiers are being supplied -17.5/-26 VDC. The signal is then
amplified and sent to the National Instruments RIO control board. This board uses a field
programmable gate array (FPGA) architecture. A control program was written in LabVIEW on a
PC and then loaded onto the FPGA. During operation, the control is performed entirely on the
FPGA hardware. No processing power from the PC is used for control, but the FPGA does send
select data points back to the visual interface on the PC so that the user can monitor its
operation.
60
Figure 13.1. General schematic of the overall mechanical and electronic assembly with flow of
current signals. (Note, the colors used in this schematic do not reflect the wiring colors of the
final assembly).
The four servo drives used to drive the AMB actuators were mounted and wired to the
power supply according to the arrangement in Figure 13.2. As shown on the capacitor, Red and
Black wires correspond with ‘+’ and ‘-’, or high and low voltage, respectively. Wire connections
from each of the servo drives correspond with the pin call-outs in the Advanced Motion
Controls product manual [12]. To conserve wire, the servo drives were grouped in pairs with
the ‘+’ and ‘-’ leads being shared with Gage 18 two conductor wire. The shielding of each wire
was grounded from one end to the power supply base. A single conductor wire connected to
the power supply base is connected to the RK4 rotor kit base. This configuration of the ground
wires is to ensure that everything in the system has the same common ground.
Included in Figure 13.2 is the location and call-out of the servo drive potentiometers
[12]. Each one can be adjusted with a small flathead screwdriver when the power is off. For
61
the purpose of this application, potentiometers 1-3 are adjusted to meet performance
specifications.
Figure 13.2. Wire diagram of servo drives to power supply with additional connections from the
servo drives to the AMB Connection box and NI FPGA Controller.
62
13.3. Servo Drives to NI FPGA Controller
The wire connections between the servo drives and NI FPGA Controller are shown in
Figure 13.3. The “current monitor” connections require a BNC adaptor to inject the +, - signals
coming from the servo drives. The “controller to servo drive” wires do not require a special
adaptor and were directly connected to the controller via spring-terminal. All of the shielded
wires were ground to the controller chassis (Figure 13.3).
63
Figure 13.4. Color coded wire configuration of AMB
1) Shielded cables must be used for all interconnect cables to the drive and the shield of the
cable must be grounded at the closest ground point with the least amount of resistance.
2) The motor chassis, controller chassis, power supply chassis, and analog servo drive chassis
must be grounded to the closest ground point with the least amount of resistance
64
Figure 14.1. System grounding [12].
3) The drive must be mounted in such a manner that the connectors and exposed printed
circuit board are not accessible to be touched by personnel when the product is in
operation.
4) A Fair Rite model 0443167251 round suppression core must be fitted to the low level signal
interconnect cables to prevent pickup from external RF fields.
5) Run separate power supply leads to each drive directly from the power supply filter
capacitor. Never “daisy-chain” any power or DC common connections, use a “stator”-
connection instead.
6) Power Supply- connect the transformer-isolated DC supply high voltage to the DC Power
Input terminal, and the DC supply ground to the power ground terminal.
65
14.2. Block Diagram
POT 2 (Current Limit) Current Loop, DIP Switch 2 (ON/OFF)
66
Potentiometers
DIP Switches
Figure 14.4. Adjustable potentiometers and DIP switches on 12A8 servo drive.
7) Use a function generator to produce the input command. Hook it up to the input reference
of the drive and set it to the input of a 50-100 Hz square wave. Set the square wave
amplitude so the drive outputs a current step similar to what will be expected when the
drive is in operation. Use Channel 1 of oscilloscope to view square wave input.
8) Connect wire of CURRENT MONITOR pin to Channel 2 of oscilloscope to view the output of
the drive. Note: Best option is using a current probe clamped to the MOTOR PHASE A.
Signals from the CURRENT MONITOR pin are unfiltered and may be difficult to view. Also,
this pin may be isolated from the drive power ground. If this is the case, the oscilloscope
must have isolated channels to avoid large ground currents.
9) View initial response of the current loop on oscilloscope. Switch any of the DIP switches
within current loop to adjust response.
67
Figure 14.6. Current loop responses with Switch 2 ON (Left) and Switch 2 OFF (Right) [16]. Note:
These are example responses provided by the manufacturer.
10) Additional current loop tuning with through PCB through-hole components may be
necessary in the case of the following:
a) Motor rapidly overheats even at low current.
b) Drive rapidly overheats even at low current.
c) Vibration sound comes from the drive or motor.
d) The motor (load) has a high inductance (>10mH).
e) The motor has a low inductance (near minimum rating of the drive).
f) Slow system response times.
g) Excessive torque ripple (N/A).
h) Difficulty tuning position or velocity loops (N/A).
i) Electrical noise problems.
j) High power supply voltage (power supply voltage is significantly higher than the motor
voltage rating or near the drive high voltage rating).
k) Low power supply voltage (power supply voltage is near the low voltage rating of the
drive).
11) An Additional resistor may be added at the R30* location, effecting the gain of the current
loop. An additional capacitor may be added at the C10* location, effecting the current loop
integrator (dampened response).
68
R30*/C10*
Through-Holes
Figure 14.7. Inside of Analog Servo Drives with through-hole component call-out for additional
tuning.
14.4. Results
Square wave input tests comprised the first major test of the system. These were
conducted with only one servo drive plugged into the connection box at a time. This one servo
drive was paired with each of the four actuators, one at a time, to produce the oscilloscope
images depicted below. These results demonstrate that there are clear differences between
the responses of the different coils. These tests were repeated for each servo drive, verifying
that the differences were indeed within the coils and not the drives. Each drive produced the
same result for each coils. During the tests, only one actuator was powered at a time. The
laminated rotor mass was inserted within the stator in its normal operating position for the
tests. Two signals are displayed on the oscilloscope images. The first is a clean, square wave
input signal provided to the servo drive powering the coil of interest. The second is the current
signal recorded by a Hall-effect current sensor in series with powered coil. This sensor displays
a voltage signal of 400 mV per 1 Amp measured. This voltage is the signal displayed on the
scope, and is superimposed over the square wave input. The apparent high frequency noise
present in this signal is actually a predictable, high frequency (over 40 kHz) chirp that is believed
to come from the MOSFET switching frequency of the servo drives. Attempts at adjusting the
shape of the signal using the “REF IN GAIN” and “CURR LIMIT” potentiometers did not
significantly impact the shape of the current output signal.
69
Later it was found that the system’s current output was almost perfectly able to follow a
sine wave input with a DC offset (such that the current in the coils would never switch
direction). Since this test was much more representative of the actual operating conditions of
the system, this was a promising result, and it was decided to move forward with PID control
testing.
It was also found that the servo drives were not able to respond to input voltages below
about 0.4 V. Below this input voltage, no current was produced in the coils. Once the input
voltage was increased above this level, a current signal gradually became visible that matched
the input signal. For accurate control, it is important that the input signal provided to the servo
drives never drops below this offset voltage after startup.
Figure 14.8. Hall effect current sensor attachment to the AMB/Servo Drive connection box.
70
(A) (B)
(C) (D)
Figure 14.9. Actuator responses for colored wire configurations: (A) Red/White, (B)
Green/Brown, (C) Blue/Purple, and (D) Gray/Orange.
71
15. LabVIEW Real-Time Simulation
Figure 15.1. LabVIEW project tree. The core of the program is boxed in red. The visual
interface tab is located directly below this box (RT.vi).
72
Figure 15.2. Visual interface of the LabVIEW PID control. From this window, inputs to the
controller can be monitored through the “waveform graph” and the control voltages being
applied to the servo drives can be viewed with the four slider bars. Controller gains (Ki, Kp, Kd,
filter coefficient), voltage biases (which translate to bias currents), and voltage setpoints
(representing the position probes’ voltage outputs when the shaft is centered) can all be
adjusted from this pane in real time, while the AMB is running.
73
While loop
For taking snapshots of the
Timing loop input data to be displayed
in the VI waveform graph.
Buffer
Hardware Inputs Hardware Outputs
74
VI Outputs
Figure 15.3. LabVIEW block diagram for the PID control program. Note that the four hardware inputs
labeled “Mod1” and “Mod2” are currently unused. For more details, contact Cameron Naugle who
was a graduate student working for Dr. Wu at the time of the project.
16. Testing and System Identification
16.1. Test Results
Upon connecting the wires of the pole pairs making up each coil actuator, continuity
checks were conducted with a direct voltage input from a power supply, demonstrating that
these electromagnets were indeed capable of attracting the rotor mass. There did not appear
to be any interruptions in the magnetic coils and each coil was measured to have a resistance of
about 1.3 Ohms.
Once the device was assembled, the eddy current proximity sensors were positioned to
each display about -7 Volts with the shaft at its center position. -5 Volts is reportedly the
recommended sensor zero position for the Bently kits, but -7 Volts was chosen so that at the
maximum and minimum displacements of the rotor allowed by the stator, the sensors would
still be within their linear range. These sensor center voltages were then input as the “blue
setpoint” and “red setpoint” in the LabVIEW control visual interface. These color designations
refer to the colored tape identifying the proximity probes.
When testing the PID control setup, both active feedback and levitation configurations
were attempted. For active feedback, two support bushings were used at either end of rotor
shaft, and the AMB assembly was placed between them. For levitation, the outboard support
bushing was removed, with the AMB attempting to replace it. Full levitation was never
attained, but should be possible. There was simply not enough time available to work through
the procedure of achieving it.
However, active feedback was successfully achieved. As shown in the figures below, the
two eddy current probe position signals were first displayed with the shaft rotating at 1,000
RPM and the AMB off. The AMB was then powered on to resist the oscillations of the shaft
with proportional control only (KI and KD set to zero). When this was done, the AMB
significantly reduced the amplitude of the shaft oscillations. An abundance of high frequency
noise was also produced when turning on the AMB, possibly due to the servo drive switching
frequency interfering with the eddy current sensors. Even with the noise, the dominant eddy
current probe signal is still clearly visible and clearly attenuated from the previous readings with
the AMB off. These oscilloscope plots were recorded with the same scaling and at the same
1,000 RPM rotational speed.
75
Figure 16.1. Position probe outputs at 1,000 RPM with AMB powered off.
Figure 16.2. Position probe outputs at 1,000 RPM with AMB applying active feedback.
76
16.2. Observations
During testing, a variety of observations were made. Regarding the control settings,
only proportional gain was used in the PID loops. A proportional gain of 0.1 was used at
startup. The device was found to be very sensitive to changes to the proportional gain and bias
voltages. To accommodate this, the proportional gain and bias values were incremented very
slowly, with all biases starting at 0 V upon startup. Very small step increments of about 0.01 V
are recommended for adjusting each of these values when ramping up to full operation. In
some instances, sparks appearing to originate from the actuator coils were observed when the
servo drive power supply switch was flicked on with a control signal already being applied. It is
hypothesized that these sparks may be the result of arcing between the stator and sections of
magnet wire where the enamel may have been scratched off.
While experimenting, the proportional gain was never allowed to exceed 1. While the
system was running at a stable operation providing active feedback, the derivative gain was
experimented with. It was allowed to reach about 1 before it produced too much
instability. Whenever integral gain was turned on, the system rapidly became unstable. During
active feedback testing, the laminated rotor mass occasionally stuck to the inside of the stator
when control values were being experimented with. This demonstrated that the touchdown
bearing designed was not sufficient. However, when this occurred, the touchdown bearing did
allow adjustments to the control values to free the rotor mass rather than having to power
down the whole system.
77
17. Recommendations
A number of recommendations for future inheritors of this project have been
stated within other appropriate sections of the report. This section will focus upon additional
recommendations not previously mentioned.
First, it is important to note that full testing of the AMB did not begin until the week
prior to the project expo. For this reason, much remains to be explored regarding the
capabilities of the device. The surface has truly only been scratched.
After a literature review and full familiarization with this report and the existing
hardware, a variety of potential avenues exist for further characterization and improvement of
the system. To begin, a more robust controls analysis and transfer function characterization of
system components is in order. This will allow for more educated controller tuning. This may
include a more advanced rotordynamics analysis using transfer matrices rather than assuming a
lumped mass model. Modal analysis of the rotor shaft would also be valuable and could be
incorporated into the controls to account for the sensors being located away from the center of
the laminated rotor mass. It may also be worth considering different configurations of the eddy
current sensors. For example, with a new mounting method, it could be possible to place the
sensors on opposite sides (axially) of the rotor mass. A variety of alternatives exist.
As previously mentioned, a redesign of the touchdown bushing would be a valuable
investment for improving the safety and longevity of the device. Such a redesign might involve
a technique for holding the rotor shaft and gradually releasing it (once the system has been
fully powered up) for attempts at levitation. Further start up procedures relating to both the
controls and hardware should also explored, particularly in the case of attempting
levitation. Part of this may include modifying the LabVIEW visual interface to produce a more
gradual ramp up step size for the various control values.
Jigger Jumonville, the Cal Poly turbomachinery professor, would be a valuable resource
for further insights and possibilities for the project. In hindsight, it would have been extremely
useful to have contacted him earlier in the project.
It may be useful to build more Hall-effect current probes for simultaneously measuring
the current in multiple coils. Additional probes can be found in the mechatronics closet
(accessible through rooms 116 and 118 in building 192). .
It should also be noted that while improvements to the device are certainly possible
through subsequent senior projects, the large amount of front end research required to
understand magnetic bearings may make improvements much more conducive to a graduate
level project. This is especially true considering the utility that higher level coursework would
serve in attempting to address many of the recommendations made.
78
18. References
1. Farmakopoulos M., Papadopoulos C., Nikolakopoulos P., and A. Tzes (2013). A Control
Model of Active Magnetic Bearings. University of Patras, Greece.
2. H. Bleuler, M. Cole, P. Keogh, R. Larsonneur, E. Maslen, R. Nordmann, Y. Okada, G.
Schweitzer, A. Traxler, 2009 Magnetic Bearings:2. Theory, Design, and Application to
Rotating Machinery. Springer-Verlag Berlin Heidelberg. 535 p. ISBN 978-3-642-00496-
4.
3. Hwang Hun Jeong, So Nam Yun and Joo Ho Yang (2012). Control of Magnetic Bearing
System, Performance Evaluation of Bearings, Dr. Rakesh Sehgal (Ed.), ISBN: 978-953-51-
0786-6, InTech, DOI: 10.5772/51185. Available from:
http://www.intechopen.com/books/performance-evaluation-of-bearings/control-of-
magnetic-bearing-system. Ch.8
4. Wikipedia (Magnetic Bearing).
5. Ungar, Eugene, and Kenneth Stroud. A New Approach to Defining Human Touch
Temperature Standards. Houston: NASA, n.d. Print.
6. Yoon, Se Young, Zongli Yin, and Paul E. Allaire. Control of Surge in Centrifugal
Compressors by Active Magnetic Bearings. London: Springer, 2013. Print.
7. Lijesh, K.P., and Harish Harani. "Optimization of Eight Pole Radial Active Magnetic
Bearing." Journal of Tribology (2015): n. pag. Print.
8. "Wire Gauges - Current Ratings." The Engineering Toolbox. N.p., n.d. Web. 15 Nov. 2015.
<http://www.engineeringtoolbox.com/wire-gauges-d_419.html>.
9. "Silicon Steel (Electrical Steel)." Proto Laminations Inc. N.p., n.d. Web. 15 Nov. 2015.
<http://www.protolam.com/page7.html>.
10. Lamination Steels Third Edition. N.p.: Electric Motor Education and Research
Foundation, 2007. Print.
11. Budynas, Richard, and Keith Nisbett. Shigley's Mechanical Engineering Design (PDF).
10th ed. N.p.: n.p., n.d. Print.
12. Advanced Motion Controls. “12A8 Analog Servo Drive.” Downloadable Resources-Data
Sheet. <http://www.a-m-c.com/support/downloadable_resources.html>
13. Advanced Motion Controls. “PS16 Series Unregulated Power Supplies.” Downloadable
Resources-Data Sheet. <http://www.a-m-
c.com/support/downloadable_resources.html>
14. Advanced Motion Controls. “Analog Panel Mount Servo Drives- Installation of Analog
Panel Mount Servo Drives.” Downloadable Resources. <http://www.a-m-
c.com/support/downloadable_resources.html>
15. Advanced Motion Controls. “AN-011- Analog Drive Configuration and Setup.”
Downloadable Resources. <http://www.a-m-
c.com/support/downloadable_resources.html>
79
16. Advanced Motion Controls. “AN-015- Current Loop Tuning Procedure.” Downloadable
Resources. <http://www.a-m-c.com/support/downloadable_resources.html>
17. Advanced Motion Controls. “AN-016- Analog Drive Current Limits.” Downloadable
Resources. <http://www.a-m-c.com/support/downloadable_resources.html>
18. National Instruments. “NI cRIO-9063.”
<http://sine.ni.com/nips/cds/view/p/lang/en/nid/213090>
19. National Instruments. “NI 9215.”
<http://sine.ni.com/nips/cds/view/p/lang/en/nid/208793>
20. National Instruments. “NI 9263.”
<http://sine.ni.com/nips/cds/view/p/lang/en/nid/208806>
21. GE Bently Nevada. “Bently Nevada 3300 XL Series Proximitor System.”
<https://www.gemeasurement.com/sensors-probes-transducers/proximity-
probes/bently-nevada-3300-xl-series-proximitor-system>
80
19. Appendices
Appendix A
81
Appendix B
82
Appendix C
data2 = csvread('IRPM.csv');
% data3 = csvread('OHor.csv');
data4 = csvread('OVer.csv');
data5 = csvread('IRPM.csv');
% data6 = csvread('IHor.csv');
data7 = csvread('IVer.csv');
Rotor
System Parameters
83
M_r = 0.5; %[kg] AMB rotor mass
ro_st = 7850; %[kg/m3] density of steel
Ds = .01; %[m]
A = pi/4*(Ds^2); %[m2] cross-section of main rotor
g = 9.81; %[m/sec2]
LD = .015; %[m] length of disk
R = sqrt(M_D/(pi*LD*ro_st)); %[m] radius of disk
M_shaft = A*ro_st*L;
M_static = M_D + M_r + M_shaft; % total mass of rotor
E = 200e+09; %N/m^2 modulus of elasticity of steel
I = pi/4*(Ds/2)^4; %m^4, area moment of inertia
% Mass Matrix
M = [.8+.1*(L1+.5*L2)/L 0;...
0 .8+.1*(L3+.5*L2)/L]; %kgram, mass matrix
% Case 1. Rigid shaft, flexible bearings. Solved from geometry and FBD of case
C_b = 1/L^2.*[(L1^2)/Kb+((L2+L3)^2)/Ka, L1*(L1+L2)/Kb+(L2+L3)*L3/Ka;...
L1*(L1+L2)/Kb+(L2+L3)*L3/Ka, ((L1+L2)^2)/Kb+(L3^2)/Ka]; % m/N
v =
1.0e+05 *
84
0.1403
2.2253
wn =
1.0e+03 *
1.1310 4.5047
Test
alpha = 3000/60/60*2*pi; %Rad/s^2
w0 = 0/60*2*pi; %Rad/s
Phi = [0, pi/8]; %rad, [(Phase angle at disk 1), (Phase angle at disk 2)]
a = [3*(.0000254), 3*(.0000254)]; %[m], eccentricity of Disk 1 and Disk 2 [.75*(.0000254),
.75*(.0000254)]
tspan = [0:1/600:200]; %1:1/535:200
y0 = [0 0 0 0]';
options = [];
[t,y] = ode45(@TwoPlaneODE,tspan,y0,options,K,D,M,alpha,w0,Phi,a);
% Stop = length(omega);
% for i = 1:Stop
% F1x(i,1) = M(1,1)*e1*omega(i,1).^2*cos(omega(i,1)*t(i,1) + phi1);
% F2x(i,1) = M(2,2)*e2*omega(i,1).^2*cos(omega(i,1)*t(i,1) + phi2);
% F1y(i,1) = M(1,1)*e1*omega(i,1).^2*sin(omega(i,1)*t(i,1) + phi1) - M_D*g;
% F2y(i,1) = M(2,2)*e2*omega(i,1).^2*sin(omega(i,1)*t(i,1) + phi2) - M_D*g;
% Ftotx(i,1) = F1x(i,1) + F2x(i,1);
% Ftoty(i,1) = F1y(i,1) + F2y(i,1);
% F3x(i,1) = (F2x(i,1)*(L2+L1) + F1x(i,1)*L1)/L;
% F3y(i,1) = (F2y(i,1)*(L2+L1) + F1y(i,1)*L1)/L;
% end
W_RPM = (alpha*t+w0)*60/2/pi;
[Amp1, loc1] = findpeaks(y(:,1));
[NegAmp1, Negloc1] = findpeaks(-y(:,1));
if length(Amp1) > length(NegAmp1)
Amp1(length(NegAmp1)+1:length(Amp1)) = [];
loc1(length(Negloc1)+1:length(loc1)) = [];
else
NegAmp1(length(Amp1)+1:length(NegAmp1)) = [];
Negloc1(length(loc1)+1:length(Negloc1)) = [];
end
Amp1 = Amp1+NegAmp1;
85
if length(Negloc1)>length(loc1)
w1 = W_RPM(Negloc1);
else
w1 = W_RPM(loc1);
end
[Amp2, loc2] = findpeaks(y(:,3));
[NegAmp2, Negloc2] = findpeaks(-y(:,3));
if length(Amp2) > length(NegAmp2)
Amp2(length(NegAmp2)+1:length(Amp2)) = [];
loc2(length(Negloc2)+1:length(loc2)) = [];
else
NegAmp2(length(Amp2)+1:length(NegAmp2)) = [];
Negloc2(length(loc2)+1:length(Negloc2)) = [];
end
Amp2 = Amp2+NegAmp2;
w2 = W_RPM(loc2);
% Magnetic Bearing
A_c = 2.975e-04; %cross section area of actuator
A_g = A_c*1.05; %cross section area of air gap
eps = 0.8; % leakage and fringing correction factor for radial bearing
Bmax = 0.88;
Vmax = 72;
N = 300; %number of coil turns 250
mu_o = 4*pi*10^(-7); %air permeability
g_o = .0005; %[m] nominal air gap
I_B = g_o*Bmax/(mu_o*N); %[Amps] bias current 1.4
Fmag = (eps*mu_o*N^2*I_B^2*A_c)/(4*g_o^2) %[N] generate magnetic force with I_B
Fstatic = 4*Fmag %[N] max generated magnetic force with 2*I_B
ki = (eps*mu_o*A_g*(N^2)*(I_B))/(g_o^2); %[N/m] current stiffness
86
K = 0.25*mu_o*N^2*A ; % Schweitzer Pg. 80
ki_A = 4*eps*K*I_B.*cosd(alpha)/(g_o)^2; % Schweiter Pg. 80
ks = (eps*mu_o*A_g*(N^2)*(I_B.^2))/(g_o^3); %[N/m] position stiffness
ku = ki; % Schweitzer Pg. 48, motion induced voltage stiffness
Lc1 = ki.^2/abs(ks); % Schweitzer Pg. 48
Lc2 = (mu_o*(N^2)*A_g)/(2*g_o); % Yoon Pg. 67, Control of Surge, also Schweitzer Pg. 77
MAXdidt = 2*g_o*Vmax/(mu_o*N^2*A_c); % max slew rate, neglecting iron reluctance and coil
resistance (eq. 3.39 in Yoon text).
MAXdFdt = 2*eps*I_B*Vmax/g_o; % Max time rate of change of the magnetic force that can be applied
to the rotor (eq. 3.40 in Yoon text)
VLc1 = MAXdidt*Lc1;
VLc2 = MAXdidt*Lc2;
wi1 = 6000*(2*pi/60); %[rad/sec]
wi2 = 2*wi1; %[rad/sec]
wi4 = 4*wi1; %[rad/sec]
i1 = 1; %[Amps]
i2 = .2; %[Amps]
i4 = .2; %[Amps]
Vtot = Lc2*(wi1*i1 + wi2*i2 + wi4*i4)
% DS = cell2dataset(F2Tot);
F1Totmax =
80.0837
F2Totmax =
75.3884
Fmag =
36.6668
Fstatic =
146.6670
Vtot =
87
48.8350
Plots
figure('Name','Inboard and Outboard Amplitude Response')
plot(w1,AMP1,'o b',w2,AMP2,'o r',data2(1,:),data4(1,:),'-k',data5(1,:),data7(1,:),'-g')
ylabel('Height [mils]')
xlabel('Omega [RPM]')
axis([300,10000,0,100])
legend('Sim Inboard','Sim Outboard','Exp Inboard','Exp Outboard','Location','east')
88
Published with MATLAB® R2015a
89
Appendix D
MagForceCalc_AWG22.m
Force, Geometry, and Electromagnetic Properties Calculations ............................................................ 90
Design Variables ...................................................................................................................................... 90
Geometry Design Variables..................................................................................................................... 91
Calculated Geometric Variables.............................................................................................................. 91
Constants ................................................................................................................................................ 92
Heat Transfer Constants ......................................................................................................................... 92
Calculated Values .................................................................................................................................... 93
Copper Wire Losses................................................................................................................................. 93
Heat Transfer Calculations ...................................................................................................................... 94
Plots ........................................................................................................................................................ 95
Parallel Coils Study .................................................................................................................................. 96
clc
clear all
Design Variables
N = 300; % number of coils
Npp = N/2; % number of coils per pole
Bmax = .88; % [Tesla] = [kg/A*s^2] flux value corresponding to the knee of the material's BH
curve
% Bknee = 1 [Tesla]
% http://dspace.mit.edu/bitstream/handle/1721.1/66486/6-061-spring-
2007/contents/assignments/steel_data.pdf
Vmax = 72; % [Volts] this is dependent on the amplifier chosen
90
Geometry Design Variables
s0 = 0.5*10^-3; % [m] nominal air gap
dcc = 0.668*10^-3; % [m] Gauge # 22. copper wire + enamel diameter
% http://www.engineeringtoolbox.com/wire-gauges-d_419.html
nl = 4; % number of layers of copper wire on one pole
ppc = 2*10^-3; % [m] stator pole length past the coil
% Dimensions below correspond to Fig. 3.18 (page 105 of Schweitzer pdf, 91
% of book)
b = 35*10^-3; % [m] core thickness
c = 8.5*10^-3; % [m] leg width
di = 23*10^-3; % [m] shaft diameter (rotor diameter that the rotor laminations are mounted over)
91
Laminated rotor section diameter [cm]
4
da =
0.1121
Fraction of stator inner circumference that is pole tip rather than air:
0.5279
Constants
% Mshaft = 0.33; % [kg] mass of shaft only
% Mdisk = 0.8; % [kg] mass of one disk
% Mlam = 0.4; % [kg] mass of laminated rotor section
% Mtot = Mshaft + 2*Mdisk + Mlam; % [kg] total rotor mass
Fworst = 76.73; % [N} From experiments, Cameron Naugle's code (Dr. Wu's grad student)
eps = 0.8; % leakage and fringing correction factor for radial bearing
% included in Ch.3 of 'Control of Surge in Centrifugal
% Compressors by Active Magnetic Bearings' book
u0 = 4*pi*10^(-7); % [N/A^2] permeability of free space
alpha = deg2rad(22.5); % [rad] for radial bearing with 4 pole pairs
92
Calculated Values
i0 = s0*Bmax/(u0*N); % [A] bias current (ib = i0) = one half current needed to reach Bknee
B = u0*N*i0/(2*s0); % [T] nominal flux density in core
ix = i0; % [A] perturbation current. ix must be < i0. Here it is set to i0 for
% worst case.
x = 0.358*10^-3; % [m] rotor displacement (set to nominal air gap for worst case)
MAXdidt = 2*s0*Vmax/(u0*N^2*A); % max slew rate, neglecting iron reluctance and coil resistance
(eq. 3.39 in Yoon text).
MAXdFdt = 2*eps*i0*Vmax/s0; % Max time rate of change of the magnetic force that can be applied
to the rotor (eq. 3.40 in Yoon text)
FgenMAX =
232.5225
[N]
Safety factor on max load, neglecting slew rate complications:
SF =
3.0304
93
disp(PLc1*8)
% Check for current overload based on recommended max current density for
% copper wire of 6 A/mm^2
bias_current_density = i0/((dcc*1000)^2*pi*0.25);
max_current_density = (ix+i0)/((dcc*1000)^2*pi*0.25)
if bias_current_density > 6
disp('WARNING: the bias current density is above recommended')
end
if max_current_density > 6
disp('WARNING: the max current density is above recommended')
end
max_current_density =
6.6605
qcu = m_cu*cp_cu*dT;
qss = m_pole*cp_ss*dT;
t = (qcu + qss)/PLc1; % [s] minimum time to reach 100F
disp('Time [min] for a pole to reach 100F')
disp(t/60)
94
h_conv = 10; % [W/m^2*K]
q_conv = 8*(h_conv*hc*b*(T3-T1) + h_conv*hc*c*(T3-T1)); % [W]
q_tot = q_cond+q_conv
q_tot =
5.4127
Plots
omega = [0:25:1050]; % [rad/s] shaft rotation frequency from 0 to about 10,000 rpm
dFdtBound = MAXdFdt./omega; % bound on Fmax due to slew rate and max time rate of change of
magnetic force
FgenMAX = FgenMAX.*ones(length(omega));
95
Parallel Coils Study
disp('PARALLEL COILS STUDY:')
i0pcs = 1; % [A]
Npcs = 300; % windings
Npppcs = Npcs/2; % windings for 1 pole
FgenMAXpcs = kipcs*i0pcs-kspcs*x ;
SFpcs = FgenMAXpcs/Fworst;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
npw = 3; % number of parallel windings on a pole
Npcs2 = Npcs/npw; % windings
Npppcs2 = Npcs2/2; % number of windings in one parallel winding on one pole
i0pcs2 = i0pcs/npw; % current through one parallel winding
96
Kpcs2 = eps*0.25*u0*Npcs2^2*A ; % Schweitzer equation (3.26)
kipcs2 = 4*Kpcs2*i0pcs2*cos(alpha)/(s0)^2; % Schweitzer equation (3.28)
kspcs2 = -4*Kpcs2*i0pcs2^2*cos(alpha)/(s0)^3; % Schweitzer equation (3.29)
FgenMAXpcs2 = npw*(kipcs2*i0pcs2-kspcs2*x) ;
SFpcs2 = FgenMAXpcs2/Fworst;
% Result: Using parallel windings makes too large of an impact on the force
% output. It does significantly reduce losses though. The more effective
% approach is to just increase the number of coils while reducing current.
97
Appendix E
AMB Amplifier
Wipe the Slate ......................................................................................................................................... 98
Ampflifier Circuit Parameters ................................................................................................................. 98
Open Current Loop ................................................................................................................................. 99
Closed Current Loop ............................................................................................................................... 99
Gamp ....................................................................................................................................................... 99
Korean Amplifier ................................................................................................................................... 100
Plots ...................................................................................................................................................... 100
98
L = 35.3e-03; %[H] with N=150, L=8-10 mH
R = 0.641; %[Ohm]
RC = 1; %[Ohm] 1, 10000
RD = 10000; %[Ohm]
Gamp
C2 = RB*CB*RD*L;
C1 = RD*L + RD*RB*CB*R;
C0 = RD*R;
D2 = RA*L*CB*RD;
D1 = CB*(RA*RB*RC+RA*RD*(R+RC)-RB*RD^2);
D0 = RA*RC - RD^2;
99
Korean Amplifier
A1 = 3000;
A0 = 210000;
B2 = 1;
B1 = 3600;
B0 = 210000;
timeSim = 3; %[sec]
% Tp = .001; %peak time
% Ts = .001; %settling time
% pos = 5; %percentage overshoot
% z = -log(pos/100)/sqrt(pi^2+(log(pos/100)^2)); %zeta, damping ratio
% w_lp = pi/(Tp*sqrt(1-z^2)); %[rad/sec] low-pass frequency with Tp
% w_lp = 4/(Ts*z); %[rad/sec] low-pass frequency with Ts
% sim('AMP')
% open_system('AMP')
Plots
figure('Name','Amp Response') plot(tout(:,1),simout(:,1),'k',tout(:,1),simout2(:,1),'b',[0,5],[-1,-1],'r')
ylabel('Response') xlabel('Time [sec]') axis([0.96,1.2,-1.6,0]) legend('Korean Response','Servo
Drive','Location','northeast')
% figure('Name','Amp Response')
% plot(tout(:,1),simout(:,1),'k',tout(:,1),simout3(:,1),'b')
% % ,[0,5],[-1,-1],'r')
% ylabel('Response')
% xlabel('Time [sec]')
% axis([0.96,1.2,-10,0])
% legend('Korean Response','Servo Drive','Location','northeast')
100
Appendix F
AMB Controls
Wipe the Slate ....................................................................................................................................... 101
Amplifier................................................................................................................................................ 101
Ampflifier Circuit Parameters ............................................................................................................... 102
Open Current Loop ............................................................................................................................... 102
Closed Current Loop ............................................................................................................................. 102
Gamp ..................................................................................................................................................... 103
Korean Amplifier ................................................................................................................................... 103
AMB....................................................................................................................................................... 103
Feedback Proximity Sensor ................................................................................................................... 105
Delay ..................................................................................................................................................... 105
PID Control ............................................................................................................................................ 106
Op Amp Tests ........................................................................................................................................ 106
Control Loop ......................................................................................................................................... 107
MATLAB................................................................................................................................................. 107
Simulink ................................................................................................................................................. 107
Plots ...................................................................................................................................................... 107
Amplifier
R = 6.2; %[ohms] L = 9.2*10^-3; %[mH] Ri = 10000; %[ohms] Rs = 2; %[ohms] Rf = Ri*Rs; %[ohms] Rd =
2.75*10^5; %[ohms] Cf = 3.55*10^-9; %[muf]
101
Ampflifier Circuit Parameters
R1 = 50000;
R2 = 20000;
R4 = 5000;
R5 = 10000;
R6 = 500000;
R7 = 500000;
C7 = .05e-06;
Rpot3 = 1; %6000
Rpot2 = 1;
RA = 10000; %[Ohm]
Ra = 20000; %[Ohm] max without Rb
Rb = 10; %[Ohm] 500,000 max
RB = Ra*Rb/(Ra+Rb); %[Ohm]
% RB = 20000; %[Ohm]
Ca = 0.01e-06; %[F]
Cb = 0.00e-06; %[F] 0.8e-06, 0
CB = Ca + Cb; %[F]
L = 35.3e-03; %[H] with N=150, L=8-10 mH
R = 0.641; %[Ohm]
RC = 1; %[Ohm] 1, 10000
RD = 10000; %[Ohm]
102
numGH_cl = [c1 c0]; %Numerator
denGH_cl = [d2 d1 d0]; %Denominator
Gamp
C2 = RB*CB*RD*L;
C1 = RD*L + RD*RB*CB*R;
C0 = RD*R;
D2 = RA*L*CB*RD;
D1 = CB*(RA*RB*RC+RA*RD*(R+RC)-RB*RD^2);
D0 = RA*RC - RD^2;
Korean Amplifier
A1 = 3000;
A0 = 210000;
B2 = 1;
B1 = 3600;
B0 = 210000;
AMB
m = 0.5; %[kg] modal mass of rotor inside AMB
d1 = .0315;%[m] inner diameter actuator path
d2 = .1045;%[m] outer diameter actuator path
h_act = .002+.0255 + (8.5e-03)/2;%[m] height of actuator pole path
N = 250; %number of coil turns 250
% I_B = 1.4; %[Amps] bias current 1.4
% Imax = 2*I_B; %[Amps] max current
g_o = .0005; %[m] nominal air gap
mu_o = 4*pi*10^(-7); %air permeability
mu = 5000; %permeability of core
d_cu = .037*100; %copper coil diameter
l_cu = (2*10.962)*100; %copper coil length 10.962
% Core
l_c = (pi*d1/8) + (pi*d2/8) + 2*h_act; %length of actuator path
A_c = 2.975e-04; %cross section area of actuator
R_c = l_c/(A_c*mu); %Actuator reluctance
103
% Rotor Mass
% lr = ; %length of rotor mass path
% A_r = ; %cross section area of rotor mass
% R_r = lr/(A_r*mu); %Rotor mass reluctance
% Air gap
% lg = ; %length of air gap [will be changing, approximate at nominal air gap]
A_g = A_c*1.05; %cross section area of air gap
% R_g = lg/(A_g*mu); %Rotor mass reluctance
% Stiffness Coefficients
ki = (eps*mu_o*A_g*(N^2)*(I_B))/(g_o^2); %[N/m] current stiffness
K = 0.25*mu_o*N^2*A ; % Schweitzer Pg. 80
ki_A = 4*eps*K*I_B.*cosd(alpha)/(g_o)^2; % Schweiter Pg. 80
ks = (eps*mu_o*A_g*(N^2)*(I_B.^2))/(g_o^3); %[N/m] position stiffness
ku = ki; % Schweitzer Pg. 48, motion induced voltage stiffness
M_eff = 0.5; %[kg]
FgenMAX = ki.*ix-ks.*x % Schweitzer equation (3.27)
104
% Vc1 = didt*Lc1
% Vc2 = didt*Lc2
Vc1max = MAXdidt*Lc1;
Vc2max = MAXdidt*Lc2;
% VLc3 = MAXdidt*Lc3;
Psat = Bmax^2/(2*mu_o);
Fsat = 2*cosd(alpha)*Psat*A;
% ki_sat = (4*Fsat*I_B)/(Imax);
% Lc3 = (2*g_o*ki_sat)/(cosd(alpha)*I_B); % Schweitzer Pg. 338
FgenMAX =
43.7361
MAXdidt =
3.4238e+03
MAXdFdt =
3.5854e+05
Delay
T_delay = .0001; %[sec] 1-2 times sampling rate of 5-10kHz
s = tf('s');
sys = exp(-T_delay*s);
sysx = pade(sys,3)
sysx =
PID Control
KP = 1; %Proportional
KI = 0; %Integral
KD = 0; %Derivative
% G_PID = tf([KD KP KI],[1])
% A = 223872; %Amp Gain
Vin = 1.5;%Voltage Source
FDis = 15
FDis =
15
Op Amp Tests
time_sim = 5; %[sec]
% Tp = .001; %peak time
% Ts = .001; %settling time
% pos = 5; %percentage overshoot
% z = -log(pos/100)/sqrt(pi^2+(log(pos/100)^2)); %zeta, damping ratio
% % w_lp = pi/(Tp*sqrt(1-z^2)); %[rad/sec] low-pass frequency with Tp
% w_lp = 4/(Ts*z); %[rad/sec] low-pass frequency with Ts
% numGH = [1.8844e07]; %Numerator
% denGH = poly([1 5990.54 1.8844e07]); %Denominator
% GH = tf(numGH,denGH);
% GHzpk = zpk(GH);
106
Control Loop
MATLAB
a1 = 2; a0 = 10000; b2 = .03; b1 = 30000; b0 = 1e08; T_amp = tf([a1 a0],[b2 b1 b0]) G_amp = tf([L*T_amp
+ R*T_amp],[1-T_amp*Rf]) G_AMP = tf([G_amp],[L (R+G_amp*Rf)]) T_AMB = tf([G_PID*G_AMP*ki],[m
ku*ki*G_PID*G_AMP (Gprox*G_PID*G_AMP*ki+ks)])
Simulink
Op-Amp A = L*T_amp; B = R*T_amp; C = 1-T_amp;
% Discrete
% T_Hz = 2*10^6; %[Hz] sampling time
% T_s = inv(T_Hz); %[Hz] sampling time
% T_AMBd = tf([T_amp*ki],[m T_amp*ki*ku/G_amp T_amp*ki*Gprox+ks],'Input Delay',0.3);
% T_AMBdd = c2d(T_AMBd,T_s,'zoh')
Plots
figure('Name','X-direction AMB Force vs. Omega') plot(omega_rpm,F3x) ylabel('Force [N]')
xlabel('Omega [RPM]') % axis([300,3000,0,100]) % legend('Sim Inboard','Sim Outboard','Exp
Inboard','Exp Outboard','Location','east')
% figure('Name','Actuator Response')
% plot(tout(:,1),simout2(:,1),'k')
% ylabel('Displacement [m]')
% xlabel('Time [sec]')
% % axis([0,3,0,.01])
107
% figure('Name','Korean Amp Response')
% plot(tout(:,1),simout(:,1),'k',[0,5],[-1,-1],'b')
% ylabel('Displacement [m]')
% xlabel('Time [sec]')
% axis([0.98,1.2,-1.4,0])
% figure('Name','Amp Response')
% plot(tout(:,1),simout(:,1),'k')
% ylabel('Displacement [m]')
% xlabel('Time [sec]')
% axis([0.98,1.2,-0.95,0])
% figure('Name','Discrete')
% plot(tout(length(simout),1),simout(:,1),'k')
% ylabel('Displacement [m]')
% xlabel('Time [sec]')
% axis([0.98,1.2,0.95,1.05])
108
Appendix G
ADAMS Control
Wipe the Slate ....................................................................................................................................... 109
AMB Parameters ................................................................................................................................... 109
Stiffness Equations ................................................................................................................................ 110
Amplifier................................................................................................................................................ 110
Feedback Proximity Sensor ................................................................................................................... 110
Ampflifier Circuit Parameters ............................................................................................................... 110
Closed Current Loop ............................................................................................................................. 111
Gamp ..................................................................................................................................................... 111
Korean Amplifier ................................................................................................................................... 112
PID Control ............................................................................................................................................ 112
AMB Parameters
M = 0.5; %[kg] modal mass of rotor inside AMB
N = 300; %number of coil turns 250
g_o = .0005; %[m] nominal air gap
x = 0.358*10^-3; % [m] rotor displacement (set to nominal air gap for worst case)
mu_o = 4*pi*10^(-7); %air permeability
% mu = 5000; %permeability of core
Bmax = 1.2;
Bknee = 0.88;
I_B = g_o*Bknee./(mu_o*N);
% I_B = 1.4; %[Amps] bias current
Imax = 2*I_B; %[Amps] max current
B = mu_o*N*I_B/(2*g_o); % [T] nominal flux density in core
b = 35*10^-3; % [m] core thickness
c = 8.5*10^-3; % [m] leg width
alpha = 22.5; % [deg] for radial bearing with 4 pole pairs
109
A_c = c*b; % [m^2] "opposing area of an electromagnet" core pole
A_g = A_c*1.05; %cross section area of air gap
eps = 0.8; % leakage and fringing correction factor for radial bearing
% included in Ch.3 of 'Control of Surge in Centrifugal
% Compressors by Active Magnetic Bearings' book
Stiffness Equations
Ki = (eps*mu_o*A_g*(N^2)*(I_B))/(g_o^2); %[N/A] current stiffness
K = 0.25*mu_o*N^2*A_c ; % Schweitzer Pg. 80
Ki_A = 4*eps*K*I_B.*cosd(alpha)/(g_o)^2; % Schweiter Pg. 80
Ku = Ki; % Schweitzer Pg. 48, motion induced voltage stiffness
Ks = (eps*mu_o*A_g*(N^2)*(I_B.^2))/(g_o^3); %[N/m] position stiffness
Amplifier
R = 0.6410; %copper resistance%[ohms]
Lc1 = Ki_A.^2/abs(Ks); %[mH] Schweitzer Pg. 48
Lc2 = (mu_o*(N^2)*A_g)/(2*g_o); %[mH] Yoon Pg. 67, Control of Surge, also Schweitzer Pg. 77
110
% Ra = 20000; %[Ohm] max without Rb
% Rb = 20000; %[Ohm] 500,000 max
% RB = Ra*Rb/(Ra+Rb); %[Ohm]
RB = 20000; %[Ohm]
Ca = 0.01e-06; %[F]
Cb = 0; %[F] 0.8e-06
CB = Ca + Cb; %[F]
L = 35.3e-03; %[H] with N=150, L=8-10 mH
R = 0.641; %[Ohm]
RC = 10000; %[Ohm] 1
RD = 10000; %[Ohm]
% %% Servo Drive
% c1 = R2*C2*Rf;
% c0 = Rf;
% d2 = R1*L*C2*Rf;
% d1 = C2*(R1*R2*Rs+R1*Rf*(R+Rs));
% d0 = R1*Rs;
% numGH_cl = [c1 c0]; %Numerator
% denGH_cl = [d2 d1 d0]; %Denominator
Gamp
C2 = RB*CB*RD*L;
C1 = RD*L + RD*RB*CB*R;
C0 = RD*R;
D2 = RA*L*CB*RD;
D1 = CB*(RA*RB*RC+RA*RD*(R+RC)-RB*RD^2);
D0 = RA*RC - RD^2;
111
Korean Amplifier
A1 = 3000;
A0 = 210000;
B2 = 1;
B1 = 3600;
B0 = 210000;
% %% Delay
% T_delay = .0001; %[sec] 1-2 times sampling rate of 5-10kHz
% s = tf('s');
% sys = exp(-T_delay*s);
% sysx = pade(sys,3);
PID Control
KP = 1; %Proportional
KI = 0; %Integral
KD = 0; %Derivative
% Discrete
% T_Hz = 2*10^6; %[Hz] sampling time
% T_s = inv(T_Hz); %[Hz] sampling time
% T_AMBd = tf([T_amp*ki],[m T_amp*ki*ku/G_amp T_amp*ki*Gprox+ks],'Input Delay',0.3);
% T_AMBdd = c2d(T_AMBd,T_s,'zoh')
% timeSim = 5; %[sec]
% sim('AMBModel')
% open_system('AMBModel')
%
% %% Plots
% figure('Name','Amp Response')
% plot(tout(:,1),simout1(:,1),'k',[0,5],[1,1],'r')
% ylabel('Displacement [m]')
% xlabel('Time [sec]')
% axis([0.6,1.4,-.1,2])
% figure('Name','Amp Response')
112
% plot(tout(:,1),simout2(:,1),'k')
% % ,[0,5],[1,1],'r')
% ylabel('Displacement [m]')
% xlabel('Time [sec]')
% % axis([0.6,1.4,-.1,1.1])
113
114
115
Appendix H
116
Appendix I
117
Appendix J
National Instruments cRIO-9063 FPGA controller quote.
ALSO SEE the following downloadable documents from National Instruments (also included
on the project Google Drive under DAQ>LabVIEW):
118
Appendix K
Part Drawings
119
NOTES:
*DISCUSS PART WITH DESIGNER
PRIOR TO MACHINING
*ALL DIMS INCLUDED ARE "CRITICAL"
UNLESS OTHERWISE SPECIFIED:
1. ALL DIMS. IN MILLIMETERS
2. TOLERANCES:
X.XX 0.05
ANGLES = 0.5
3. MATERIAL: AL - 6061
4. QUANTITY: 1
84.50 5. STOCK SIZE: 10-12" x 3" x 2.5"
113
25.65 28.11
R56.50
90.0°
2.10
ISO SCALE: 1:2
4x M5 x 0.8 10
Cal Poly Mechanical Engineering PHONE: 858-354-7763 Title: STATOR BASE Drwn. By: SEAN FOWLER
ME 429 - WINTER 2016 [email protected] Date: 3/17/2016 Scale: 1:1 Chkd. By: GARRETT OLSON
NOTES:
*DISCUSS PART WITH DESIGNER
PRIOR TO MACHINING
UNLESS OTHERWISE SPECIFIED:
Critical Surfaces 1. ALL DIMS. IN MILLIMETERS
2. TOLERANCES:
X.XX 0.10
ANGLES = 1
3. MATERIAL: AL - 6061
4. QUANTITY: 1
5. STOCK SIZE: 10-12" x 0.3125" x 3"
4 x 5.2 THRU-ALL
(M5 Clearance Hole. Use #6 drill bit)
Cal Poly Mechanical Engineering PHONE: 858-354-7763 Title: STATOR BASE COVER PLATE Drwn. By: SEAN FOWLER
ME 429 - WINTER 2016 [email protected] Date: 3/17/2016 Scale: 1:1 Chkd. By: GARRETT OLSON
NOTES:
*DISCUSS PART WITH DESIGNER
PRIOR TO MACHINING
UNLESS OTHERWISE SPECIFIED:
1. ALL DIMS. IN MILLIMETERS
2. TOLERANCES:
X.XX 0.10
ANGLES = 1
3. MATERIAL: AL - 6061
4. QUANTITY: 1
5. STOCK SIZE: 10-12" x 3" x 2.5"
Critical Surface
R57.55
Cal Poly Mechanical Engineering PHONE: 858-354-7763 Title: STATOR CASE ROOF Drwn. By: SEAN FOWLER
ME 429 - WINTER 2016 [email protected] Date: 3/17/2016 Scale: 1:1 Chkd. By: GARRETT OLSON
NOTES:
ALL DIMENSIONS IN INCHES
(CONVERTED FROM MILLIMETERS)
UNLESS OTHERWISE SPECIFIED:
1.378 1. TOLERANCES:
X.XXX .003
ANGLES = 1
2. MATERIAL: ARNON-7
3. QUANTITY: 1
1.614 .001
4.449
3.780
8 x 45°
8 x .335
ISOMETRIC VIEW
Cal Poly Mechanical Engineering PHONE: 858-354-7763 Title: Magnetic Bearing Stator Drwn. By: SEAN FOWLER
ME 429 - WINTER 2016 [email protected] Date: 1/16/2016 Scale: 1:1 Chkd. By: GARRETT OLSON
NOTES:
ALL DIMENSIONS IN INCHES
(CONVERTED FROM MILLIMETERS)
UNLESS OTHERWISE SPECIFIED:
1.378 1. TOLERANCES:
X.XXX .003
ANGLES = 1
2. MATERIAL: ARNON-7
3. QUANTITY: 1
.906
1.575 .001
ISOMETRIC VIEW
Cal Poly Mechanical Engineering PHONE: 858-354-7763 Title: Magnetic Bearing Rotor Drwn. By: Sean Fowler
ME 429 - WINTER [email protected] Date: 1/16/2016 Scale: 1:1 Chkd. By: Garrett Olson
Appendix L
Concept Sketch
RK 4 rotor kit base
Stator coils
Cal Poly Mechanical Engineering Lab Section: Assignment # Title: CDR Concept Drwn. By: Sean Fowler
ME 428 -Fall 2015 Dwg. #: Nxt Asb: Date: 2/11/16 Scale:1:2 Chkd. By: ME STAFF
Appendix M
Gantt Chart
ID Task Task Name Duration Start Finish 8, '15 Oct 25, '15 Nov 1, '15 Nov 8, '15 Nov 15, '15 Nov 22, '15 Nov 29, '15 Dec 6, '15 Dec 13, '15 Dec 20, '15 Dec 27, '15 Jan 3, '16 Jan 10, '16 Jan 17, '16 Jan 24, '16 Jan 31, '16 Feb 7, '16 Feb 14, '16 Feb 21, '16 Feb 28, '16 Mar 6, '16 Mar 13, '16 Mar 20, '16 Mar 27, '16 Apr 3, '16 Apr 10, '16 Apr 17, '16 Apr 24, '16 May 1, '16 May 8, '16 May 15, '16 May 22, '16 May 29, '16 Jun 5, '16 Jun 12, '16 Jun 1
Mode W S T F M T S W S T F M T S W S T F M T S W S T F M T S W S T F M T S W S T F M T S W S T F M T S W S T F M T S W S T F M T S W S T F M T S W S T F M T S W S T F M
0 Mag Bearing 168 days? Wed Sat
Project 10/28/15 6/11/16
1 ME 428 33 days Wed 10/28/1Fri 12/11/15
7 Electronics
Hardware
Design
Project: Mag Bearing Project Task Milestone Project Summary Inactive Milestone Manual Task Manual Summary Rollup Start-only External Tasks Deadline Manual Progress
Date: Mon 6/6/16 Split Summary Inactive Task Inactive Summary Duration-only Manual Summary Finish-only External Milestone Progress
Page 1