Overview of the Presentation

Introduction to Magnetic Bearings

Introduction
Lecture presented in Quality Improvement
Program (QIP08) at Indian Institute of Design of Active Magnetic Bearings
Technology Guwahati
Control Engineering of Magnetic Bearings

Control of Rotor by using Magnetic Bearings

Jagu Srinivasa Rao, (Research Scholar)
Department of Mechanical Engineering Conclusions
Indian Institute of Technology Guwahati
December, 2008

Radial Magnetic Bearing

Introduction
An active magnetic bearing (AMB) system supports Eight-Pole Radial Magnetic-Bearing
a rotating shaft, without any physical contact by
suspending the rotor in the air, with an electrically
controlled (or/and permanent magnet) magnetic
force.
It is a mechatronic product which involves different
fields of engineering such as Mechanical, Electrical,
Control Systems, and Computer Science etc.

Test Apparatus for rotor control

Thrust Magnetic Bearing Introduction to Active Magnetic Bearings

Rotor Disc
Left AMB
Coil Winding
Upper AMB
Rotor shaft

Rotor shaft

Left AMB Lower AMB

Horizontal shaft Vertical shaft Typical Actuator Controller unit of an AMB

1
 Introduction to Active Magnetic Bearings
Advantages of Magnetic Bearings
Magnetic Bearings are free of contact and can be utilized in
vacuum techniques, clean and sterile rooms, transportation of
Electro magnet
aggressive media or pure media
Power
f Highest speeds are possible even till the ultimate strength of

Rotor
Amplifier
the rotor
Sensor
Absence of lubrication seals allows the larger and stiffer
Controller rotor shafts

Absence of mechanical wear results in lower maintenance

costs and longer life of the system
Working principle of magnetic bearing
Adaptable stiffness can be used in vibration isolation,
passing critical speeds, robust to external disturbances

Classification of Magnetic Bearings Applications of Magnetic Bearings

According to Load supported
Axial or Thrust Turbo molecular pumps Test rig for high speed tires
control action
Active Radial or Journal
Conical Blood pumps Magnarails and maglev systems
Passive
Hybrid Molecular beam choppers Gears, Chains, Conveyors, etc
Magnetic effect
Forcing action Electro magnetic
Electro dynamic Epitaxy centrifuges Energy Storage Flywheels
Repulsive
Attractive
Application Contact free linear guides High precision position stages
Sensing action Precision flotors
Linear motors Variable speed spindles Active magnetic dampers
Sensor sensing
Self sensing Levitated rotors
Bearingless motors Pipeline compressor Smart Aero Engines
Contactless Geartrains
Elastic rotor control Turbo machines

Fields of Applications of Magnetic Bearings

Semiconductor Industry Maglev Transportation

Electromagnetism
Bio-medical Engineering Precision Engineering

Vacuum Technology Energy Storage Electromagnetic field

Lorenz force
Structural Isolation Aero Space

Rotor Dynamics Turbo Machines

2
 Electromagnetism Electromagnetism
When a charged particle is
at rest it wont emit When the particle is in
electromagnetic waves accelerated motion then
rather it is surrounded by the magnetic field will be
electrostatic field oscillating.

3d electrostatic field
surrounding a
When the charged particle is charged particle
in uniform motion (i.e. the In electromagnetic
motion with uniform velocity waves both the electric
in a direction) the and magnetic fields are
electrostatic field is oscillating and harmonic.
associated with
magnetostatic field.
Magnetostatic field

Feed back loop of electromagnetism

Feed back loop of electromagnetism
The electric and magnetic The electric and magnetic
fields are generated by fields produce forces on The electric and magnetic
electric charges electric charges fields are generated by
electric charges
Charges generate electric Electric force which is
fields generated by the electric field
Movement of charges and is in same direction as
generate magnetic fields electric field
The electric The electric and
The electric and magnetic magnetic force which is charges move in magnetic fields
generated by the magnetic space when they interact only with
fields interact only with each field and is perpendicular both are acted upon by each other
other to magnetic field and to field forces
Changing electric field acts like velocity of charge
a current, generating vortex of
magnetic field The electric charges move in
The electric and magnetic
Changing magnetic field space fields produce forces on
induces (negative) vortex of The electric charges move in electric charges
electric field space when they are acted
upon by field forces

The four fundamental forces The four fundamental forces

Strong nuclear force Weak nuclear force Gravitational force Electromagnetic force
which holds atomic which causes Which causes the Which is caused by
nuclei together certain forms of masses to attract electromagnetic fields on
radioactive decay each other electrically charged
particles

3
 The four fundamental forces Force between two electrically charged particles

r
All the other forces are derived from Coulomb force (Static)
these four fundamental forces q1q2
fc = r
4 0 r 3 q1 fc q2
Electro-magnetic force is one of these
four fundamental forces
Lorenz force (Dynamic)
q q r q q vr
fl = 1 2 3 + v 1 2 2 3
4 0 r 4 0 c r

Electric and magnetic components Comparison Electric and magnetic

of Lorenz force components of Lorenz force
If q1=q then v B v2 1
2 23
E c 10
F = q (E + v B )
q2r q2 v r Three conclusions:
E= Electric flux; B= Magnetic flux;
4 0 r 3 4 0 c 2 r 3 Magnetic component of Lorenz force is at least smaller by a factor of 1023!
But we dont face the effect of electric field in conductors because protons
1 and electrons are equal in number and generate equal and opposite electric fields
r= r; = Lorenz factor; canceling each other
2
1 (v / c) Protons have no motion with reference to conductor and there wont be
magnetic component from them. Thus the magnetic component observed is
the relativistic effect of electrons only
0 = 8.854 1012 C 2 / J-m Electric permeability of vacuum; When the conductor is moving with reference to another frame both the
protons and electrons will move with the same velocity thus the relativistic
effects due to the velocity of conductor will be cancelled out
1
= 0 = 410 -7 N/A 2 Magnetic permeability of vacuum;
0c 2

Effective Lorenz force in macro calculations Relations between E and B

For macro calculations Lorenz force is Gauss Law for linear q 1

reduced to the form materials E =
0 E ds =
S
0
V
qdv

Gauss Law for

Bw
B magnetism B = 0 B ds = 0
F = q ( v B) Faradays law of
S

B
v magnetic induction E =
t E dl = t B ds
L S

Amperes law and

F E E
Maxwell's
extension
B = 0 J + 0


t L B dl = 0 S J + 0 t ds
Lorenz force acts perpendicular to both velocity
of charged particle and magnetic flux
These relations are called simplified Maxwell's relations who formulated
the original relations from previous works

4
 Specifications

Mechanical design

magnetic bearing systems

Design of magnetic

Design methodology of
no
Performance O.K?

actuator yes
Magnetic actuator design

no
Performance O.K?
Bearing magnet
yes
Magnetic circuit Control system design
Coil
Simulation

Experimentation

Performance O.K?
no
yes
End

Areas involved in the design of magnetic Magneto mechanical systems

bearing systems
Magnetic bearing
system design
According to the known technology till
now, magnetic bearings can be classified
Mechanical Magnetic actuator
design design
Control system
design for their design according to the purpose
Modal
frequencies
Bearing magnet
design
Coil design Power
amplifier
Controller Sensor
design
of the levitated object as
design
design
Admissible coil Position
Topology temperature Stiffness sensing
Losses

Damping Velocity
Load Number of Bandwidth sensing
estimation turns
Stability
Current
Winding sensing
Magnetic scheme
circuit Balancing
design Flux
Coil head sensing

Self sensing

Magneto mechanical systems Magneto mechanical systems

Linear motors
Precision flotors (precision stages,
(Contactless sliders,
isolation bases, isolation springs)
maglev trains and
Levitation force
conveyors)
Propulsion force
Levitation force
Propulsion force

Levitation force Propulsion force

A magnetic Precision Stage

Principle of a linear motor

5
 Magneto mechanical systems Magneto mechanical systems

Levitated rotors Bearingless motors

(gas turbines, (canned pumps,
energy storage compact pumps, blood
flywheels, high pumps, spindle
speed spindles, drives,
balancing and semiconductor
vibration control process)
of rotors)
Radial load
Radial load
Thrust load Thrust load
Rotor levitated by Radial and Torque Bearingless Motor
Axial Active Magnetic Bearings

Magneto mechanical systems Linear systems from rotary systems

Contactless Gears and

Couplers
Regulated torque
transmission

Macro Geometry of Thrust Magnetic Bearing

Design of a thrust magnetic bearing

Back-wall
Inner wall

Coil Outer wall

Space for shaft

Space for coil

Figure 1: Parts of Thrust Magnetic Bearing

6
 Optimal design Magnetic circuit
Optimal design is carried out Actuator Rg
in two steps
Modeling the magnetic Ra Ni Ra
circuit Coil Ni
Determines the accuracy of
Rl Rg
achieving the objective Rl
Equivalent electric (dc)
circuit representation
gap Levitated object
Optimization of the
Magnetic circuit
parameters
Determines the efficiency of l fp l fp
R= =
the achieving the objective A 0 r A

Magnetic circuit analogy Ideal magnetic circuit model

with electric circuit la

L
H dl = J nda
S
( Ampere's law )

2 H g lg + H a la + H s ls = ni H
Magnetic circuit Electric circuit J
B = H or H = B / lg
Electro Motive
Magneto Motive B B
Force (EMF) or 2 Bg lg + 0 a la + s ls = 0 ni
Force (MMF) a s ls
Voltage (V)
B B
Magnetic Flux ( ) Electric Current (i) if 0 a la + s ls is neglected
a s

Reluctance (R) Resistance (R) 0 ni

Bg =
2lg

Flux density is used to find the force exerted

Extension of the ideal model Force by using flux density

la
B B
if K a is added for 0 a la + s ls
a s
as core loss factor and K i is added
as coil loss factor, then
lg B= =
Ni 0
A 2(lg x )
The model reduces to i0 + i B0 + B
ls B02
2 K a Bg lg = 0 Ki ni F = Ag i0 i B0 B
20

0 K i ni
Bg = 2 Ag
2 K alg F = B0 B
0

Differential actuator

7
 Linear range of flux density Terminology used in
magnetic circuit
Quantity Symbol Formula Units Magnitude
Permeability 1 Vs/Am
0
Magnetic flux density, T

Bsat 0c2 4 10 7
of vacuum
Permeability 0 r Vs/Am 0.026
max Bsat Reluctance R
l fp
=
l fp Vs/A 7.95e5 for air
A 0 r wl
3.97e4 for Fe
Magneto ni ni A- 1600
motive force turns
min Bsat Linear Range Ni 0wl
Magnetic flux Ni
2Rg
=
2(g x)
Wb 0.0010

Flux density Ni 0 T 10.05

B =
A 2( g x )
Magnetic force, N Magnetic flux Wb- 0.1005
N
linkage turns
Hysteresis is assumed to be negligible
while setting the linear range

Different quantities used in

magnetic circuit Design vector for optimal design
Quantity Symbol Formula Units Magnitude
Current i
=
i A/m2 16e4
J Known parameters are
density A wl
Magnetic = n 2 0 wl H=Wb/A Gap
L
inductance i 2 ( lg x ) Inner radius of the bearing
Nominal n 2 0 wl H 0.0063 Outer radius of the bearing
L0 L x=0 =
inductance 2lg
Free parameters
Magnetic N 804.2
L0i 2 Inner radius of the coil space
force by F
2lg
inductance Outer radius of the coil space
Magnetic N 804.2 Height of the coil space
force by flux F B02
Ag
density 20 Current density supplied
Magnetic Ag N 19.84
force for diff F
2
(B +
2
B2 ) All the other parameters are dependant
0
actuator

Input parameters taken for the design of thrust magnetic bearing Eight pole radial magnetic bearing
Parameter Value Parameter Value
Inner radius of the 25.00mm Specific gravity of the 7.77g/cm 3
bearing stator iron
Operating air gap 4.00mm Specific gravity of the 8.91g/cm 3
copper
Operating load 2025N Specific gravity of 7.5g/cm 3
permanent magnet material
neodymium-iron-baron
Variation in the gap 5% Coil mmf loss factor 1.394
Variation in the load 10% Actuator loss factor 1.072
Saturation flux density 1.00T Flux leakage factor 0.840
Remnant flux density of 1.2T Packing factor 0.85 Eight Pole AMB
bias magnets
Saturation current 4.0A/mm 2 Maximum allowable coil 820mm 3
density volume
Maximum outer radius 120mm Maximum height of bearing 70mm
of bearing

8
 Radial magnetic bearing Three pole radial magnetic bearing

The component of force will be 0 Ag ( K i ni ) 2

at an angle of half of the angle F= cos
4( K a lg ) 2 2
between two poles

Magnetic Circuit for

three pole AMB
Three Pole AMB

Coil design
Winding scheme
Admissible coil temperature is determined by
the choice of insulation type
Number of turns are chosen such that it
generates maximum admissible magneto
motive force at the maximum current supplied
by the power amplifier

Coil

Permanent magnetic bearings Permanent magnetic bearings

Br

Hc Ha BH max
BH

9
 Introduction

Control is the process of bringing a

system into desired path when it is
going away from it
MAGNETIC BEARINGS Earnshaw(1842) had shown that it is
impossible to hover a body in all six
CONTROL degrees of freedom by using permanent
magnets
But it is possible to maintain the body in
equilibrium condition by active control

Types of control systems Types of control systems

Open loop control systems Closed loop control systems
U ( s) G( s) Y (s)
U ( s) G (s) Y (s)
x
H (s)
The control in which the output of the system has
no effect on input is called open loop control If the control maintains a prescribed output and
Open loop control is used when the input is known reference input relation by comparing them and
and there are no external disturbances uses their difference as controlling quantity, it is
An example of open loop control is washing called feedback or closed loop control
machine which works on time basis rather than the Temperature control of a room or a furnace is an
cleanliness of clothes example of closed loop system

Classification of controllers Classification of controllers

According to control action controllers are Two-position or on-off controllers

classified as: The output of the controller y (t ) will be a
Two-position or on-off controllers maximum or minimum according to the state of
Proportional controllers error e(t ) as below:
Integral controllers y (t ) = y0 for e(t ) < 0
Proportional-integral controllers = y1 for e(t ) > 0
Proportional-differential controllers
Proportional-differential-integral controllers y0 and y1 are minimum and maximum values of
output

10
 Classification of controllers Classification of controllers

Proportional controllers: Integral controllers:

The output of the controller y (t ) is proportional to In integral control action, the value of the
the magnitude of the actuating error e(t ) signal as controller output y(t ) is changed at a rate
y (t ) = g p e(t ) proportional to the actuating error signal e(t )
By Laplace transformation dy(t ) t
y (t ) = g i e(t )dt
= g i e(t ) (or) 0
Y (s) dt
= gp
E (s) By Laplace transformation
g p is called proportional gain Y (s) gi
=
E (s) s
g i is called integral gain

Classification of controllers Classification of controllers

Proportional-Integral (PI) controllers: proportional-differential (PD) controllers:

Control action is a combination of both The control action is defined by
proportional and integral action de(t )
y (t ) = g p e(t ) + g pTd
gp t dt
y (t ) = g p e(t ) +
Ti e(t )dt
0 By Laplace transformation
By Laplace transformation Y (s)
= g p (1 + Td s )
Y ( s) 1 E (s)
= g p 1 +
E ( s) Tis

Classification of controllers Control Design

proportional-Integral-differential (PID) Input System Output

controllers:
An over all system
It has the advantages of all three actions. So this is
the most common type of industrial controllers
Mathematical form of PID action is U(s) G(s) Y(s) Y ( s ) = G ( s)U ( s )
t
gp de(t ) Transfer-function representation of a system

Ti 0
y (t ) = g p e(t ) + e(t )dt + g pTd
dt
x(t )
By Laplace transformation u(t) y(t) y (t ) = Ax(t ) + Bu(t )
x(t )
Y (s) 1
= g p 1 + + Td s State-space representation of a system
E ( s) Ti s

11
 Control Design Methods of design and
analysis of controllers

Input System Output

Methods of design and analysis
An over all system

Known Known unknown

Studying the behaviour of a system
Transfer-function method State-variable method
Known Unknown known
Studying the characteristics of a system
Transient and Linear-
Frequency
steady state Root locus Pole-placement quadratic
response
Response analysis analysis optimization
analysis
analysis
Unknown Unknown known
(Classical control) (Modern control)
Designing of a control system of required behaviour

Mechanical and electro magnetic

Transfer-function State-space method stiffness
method

Classical control method
Modern control method

Used for single input single output
Used for multi input multi output fs x
(SISO) systems (MIMO) systems can be used for SISO Equilibrium
x0 Rotor
also position
mg

It is useful for linear and simple
It is useful for nonlinear and Mechanical spring stiffness
systems only complex systems also.
mechanical spring

Frequency domain method
Time domain method

Analysis consists of single higher
Analysis consists of system of n first
order differential equation order differential equations.
fm xm

Steady state and transient response
Pole-placement method and Linear- Operating
analysis, Root locus analysis and quadratic optimization are the main x0 position
frequency response analysis are the methods of design and analysis. mg
main methods of design and analysis magnetic displacement stiffness
Magnetic spring

Magnetic Bearing Control

im Instantaneous
Operating i0 current
current Equilibrium and Operating points
For a mechanical spring there will be an
equilibrium point where the force resisted by the
fm spring is equal to the force applied on the spring
Operating
x0 x0 position
For electro magnets there will be a quantity of
mg current corresponding to position of the object and
Magnetic spring
force applied. At this point the gravity force and
magnetic force will be equal. A slight movement
form this point will cause indefinite movement of
the body. This point is called operating point

12
 Linearization at operating point
mg i0 x0 Linearized formula around the operating point will be

f ( x, i ) = k x x + ki i

where
x is the displacement from the
operating position
i is the deviation of current
from operating current
Linearization of current Linearization of displacement

i = im i0 x = x0 xm k x is displacement stiffness
f = ki i f = kx x ki is current stiffness
im is the instantaneous current xm is the instantaneous position
f is instantaneous force

Magnetic Bearing Control

Linearized equation is suitable for most of the

applications of magnetic bearings f = k x x + ki i

It is not valid in three occasions By Newton's law

When x = x0 the rotor touches the bearing magnet
f = mx Active magnetic
When there are strong currents such that magnetic bearing system
saturation of the material occurs Combing above two equations we get k c

When i = i0 or very small currents there wont be

levitation of the rotor because of very small mx k x x = kii f x
m
Rotor
magnetic forces.
spring mass damper
system

If controlling current i is zero then If we supply controlling current i such that

mx k x x = 0 k + kx c
i ( x) = x + x
ki ki
And the response grows exponentially thus
the rotor may fall down or touch the magnet then it becomes mx + cx + kx = 0
And the response is imitated to a spring mass damper
system by the magnetic bearing system
k c

f x
m
Rotor

spring mass damper system

Response of magnetic
bearing without control Response of magnetic
bearing with control

13
 PD controller model

x k +k c
( x) = x + with
The model is iPD-controller x proportional
ki ki r Controller Amplifier
and differential feed back
e ic i f x x x
P=
k + kx
D=
c + ic = Pe + De i = ic ki + 1/ m
ki ki

In design of controller we choose the stiffness kx

and damping to ensure the system come to y

y=x
steady state in optimum time.
The optimal stiffness suggested is Sensor
k =systems
The range of damping ratio for better kx
Block diagram of PD controller with
suggested is 0.1 to 1 current control

r Controller fload
Amplifier
ic = Pe + De ic i f x
e x x
+ 1 i = ic ki + 1/ m
Ti Control of rotors by using
+ edt

kx magnetic bearings
y
y=x

Sensor

Block diagram of PID controller with

current control

Differences between mechanical and

magnetic bearing models
Topics to be covered
Mechanical bearing model Magnetic bearing model

Stiffness is very high Stiffness is very low thus

Rigid rotor model thus the vibration of the the rotor can rotate freely
rotor will be transmitted about the principal axes of
Flexible rotor model to foundation inertia which results in a
vibration isolation system.

Damping is directly As the rotor is free in the

observed due to air there is no coulomb
hydrodynamic effects damping acting on the
system. The control law
will have damping term.

14
 Rigid rotor model Rigid rotor model
If the variable vector is chosen as
T T
x = [ x1 x2 x3 x4 ] = [ x y]
Angular velocity vector can be expressed as

x 0  cos t +  sin t

= y 0 =  sin t +  cos t
z 0

y y y y
y'

Rotor mechanical bearing system z'
O O
z z z z

Angular velocity of shaft

z'

x x x x
d d x'
Infinitesimal rotation about x axis  =  = Motion about x- axis Motion about y- axis
Infinitesimal rotation about y axis dt dt

Rigid rotor model Rigid rotor model

Kinetic energy is expressed as Equations (1) can be expressed in matrix form by rearranging
1 1
T = m ( x 2 + y 2 + z 2 ) + ( J x 0x20 + J y 0 y20 + J z 0 z20 ) x + ( G + C ) x = F
M
2 2
Equations of equilibrium can be obtained as by using Lagranges principle
M is the inertia matrix (M = M T )
G is the gyroscopic matrix(G = -GT )
d T T C is the damping matrix (C = C T )
+ = Fi
dt xi xi
F can be expressed as

Fi is the generalized force corresponding to i variableth

F = (K + N )x

K is conservative force matrix (K = K T )

N is non-conservative force matrix(N = - N T )

Rigid rotor model

Conservative forces include Rigid rotor model

forces due to stiffness
Non-conservative or circulatory forces From Eq. (2) and (3) we get
include
x + ( G + C ) x + ( K + N ) x = 0
M
Internal or structural damping
Steam or gas whirl in turbines
Seal effects If the non-conservative and gyroscopic
Process forces such as in grinding forces neglected, we have
Unbalance, etc
x + Cx + Kx = 0
M
Damping include
Coulomb damping due to hydrodynamic effects

15
 Natural modes Natural modes

The solution of the equations (5) gives

four modes, for there are four degrees
of freedom considered Forward whirl Backward whirl

Translation mode Rotation mode

Forward nutation Backward nutation

Magnetic bearing model Weight considerations

In a magnetic bearing if we neglect the

conservative, non-conservative, and
damping effects, we will have
mg
x + Gx = F
M
For small rotations gyroscopic effects

can be neglected and the equations in x
and y directions can be decoupled
x=F
M f g = mg = ki i0 fg =
mg
= ki i0
cos

Imbalance considerations Magnetic bearing model

It can be written as
mx = f k + f c f g + f

where
e

f k = kx x
m is the imbalance mass f c = ki i
 2 cos ( t + )
f = me f g = mg = ki i0
e is the eccentricity of
imbalance mass  2 cos ( t + )
f = me
is the angular position
of imbalance mass

16
 Magnetic bearing model Rigid rotor with magnetic bearing
Three steps involved:
Formulation with respect to centre of gravity
It will be Transformation with respect to the bearing
coordinates
mx = k x x + ki ( i i0 ) + me
 2 cos ( t + ) Transformation with respect to the sensor
coordinates

i at any instant will be Bearing Centre of gravity

x
 2 cos ( t + )
mx kx x me
i = i0 +
ki O
z

Sensor

With respect to centre of gravity

Why with respect to sensor
In slow role x and y directions can be decoupled
coordinates
p mx = f
Sensors cannot be A x B
arranged directly in the I  = p
y
magnetic actuator. f
x O
z
In matrix form as
This requires certain
gap between the y x=f
M
f ax fbx
magnet and the sensor. a
where
b
The displacements with m 0 x
respect to sensor c d M = ,x =

0 I y
coordinates will be
transformed to bearing f
f =
coordinates p

With respect to bearing coordinates

With respect to bearing coordinates
Forces are transformed as
Displacement vector can be
p transformed as
A x B
f = f ax + f bx
f p = af ax + bf bx p x = TB x B
O
x z
A x B x
f = Tf fB x=

y
f
f ax fbx 1 f ax x O
1 z
1 b a
a b Tf = ,f = TB =
a b B f bx xa y xb b a 1 1
c d
a b x
xB = a
x
b
c d

17
 State feed back
With respect to sensor coordinates State space form with respect to sensor coordinates
As = Ts AbTs1
x x s = As x s + Bsu Bs = Bb
xS = c
p xd The control vector is found by using control
A B
f
x
law
1 c x u = Fx s
O
TS = x=
x z
1 d

We do not know the velocity components
y
xc fax fbx xd x S = TS x = TS TB x B directly from sensors. So a state observer is
a b required to find the velocities
xS
xs = =T x
 S s b
c d
x x s is the full state vector
xs = Cx S
TS TB 0 x S is the vector from the sensor
Ts =
0 TS TB

State feed back

Model at high speeds
The whole closed loop system can be shown as
block diagram At high speeds the gyroscopic effects cannot
be neglected, thus the model becomes
x s = As x s + Bsu u x s xs xS
Mx + Gx = F

Bs +
C
x S = Cx s
The displacements in x and y directions no
u = Fx s As longer decoupled, so four forces and four
displacements should be taken into
x s = ( As Bs F ) x s d
dt consideration simultaneously.
F
As Bs F decides the closed loop
The same procedure is to be followed as for
dynamics of the system
the slow rotation

Model at high speeds Conclusions on rigid rotor model

There is an optimal design for each speed

m 0 0 0 0 0 0 0
0 I 0 0 0 1
0 0
The optimal design at higher speed may not
M = y
G= be stable at lower speeds, for the gyroscopic
0 0 m 0 0 0 0 0
effects are reduced.
0 0 0 Ix 0 1 0 0
The optimal design at zero speed may not be the
optimal at higher speeds
fx xa
p
x
The gyroscopic effects will not destabilize the system
f = y xB = b which is stable at lower speeds.
fy y Further more the design at lower speeds is decoupled
a
px yb and easier to design. Decentralized designs for lower
speeds can be implemented

18
 Conclusions on rigid rotor model
Flexible rotor model
Thus for stability considerations and other
advantages systems are designed for Rigid rotor can be Flexible rotor has
lower speeds and with decentralization defined by two infinite degrees of
points freedom. One
uxa = Fxa xa cannot define
uniquely by some of
uxb = Fxb xb
the points
u ya = Fya xa
Decentralized control mode scheme
u yb = Fyb xb

Flexible rotor model

Flexible rotor model
By substituting we get
Equation motion of
d 4Y ( z ) d 2 q (t )
an Euler-Bernoulli 4
EI dz dt 2
z beam is given by = = 2
m Y ( z) q (t )
4 y 2 y
L EI +m 2 =0 By rewriting we get
z 4 t
The variable d 4Y ( z )
4Y ( z ) = 0, =
2
separable form is dz 4 EI / m

y ( z , t ) = Y ( z )q (t )
d 2 q (t )
+ 2 q (t ) = 0
dt 2

2
Flexible rotor model The mode shapes or
modal functions
z depend on the end
L By applying initial
conditions
conditions and solving
we get the natural
frequencies
3
0
z
By substituting the
Eigen values in (29) we
1 get the Eigen functions Flexible rotor modes
z or model functions Y ( z )

Rigid rotor modes

19
 Actuator sensor location
Actuator sensor location
Sensor should not be set at nodes We can conclude that the sensor can be
set at a place where we can get
information from each mode under
consideration

Sensor and actuator should not lie on

opposite sides of a node
actuator

sensor

Modal reduction (mathematical representation)

Modal reduction
Mathematical model of the
While designing a flexible rotor system, we full system x = Ax + Bu
y = Cx
can not consider all the modes of the system
for they are infinite x M AM AMR x M BM
Thus we consider first n number of modes Divided system x = A +
AR x R BR
u
R RM
corresponding to first n natural frequencies x
and neglect the remaining modes y = [ CM CR ] M
xR
If we study the effect of the reduced modes
we can find the number of modes which we x M = AM x M + BM u
can consider without destabilizing the system Reduced system y = CM x

Modal reduction
Modal reduction Modeled modes
AM
The reduced modes
x M AM AMR x M BM u xM y
give three kinds of x = A +
AR x R BR
u BM + CM +
R RM
effects on the system
x
called spillovers y = [ CM CR ] M
xR ARM Interconnection AMR
Control spillover (By the spillover
input) xR
Interconnection spillover
BR + CR

Control Observation
(By the parameters of the Input System Output spillover
AR spillover
system)
Unmodeled modes
Observation spillover (on Control Interconnection Observation
spillover spillover spillover
the estimated output)
Block diagram of effect of model reduction

20
 Conclusion on flexible rotor control Conclusions

Modal reduction is studied to consider Magnetic bearings advantages and

the number modes to be taken into applications have been discussed
consideration for having stable control Electromagnetism and Control system
Mechanical design is studied for finding technologies have been introduced
the sensor actuator locations Design of thrust and radial magnetic
bearings have been studied
Control of a rotor by rigid rotor and
flexible rotor models have been studied

Further References

Schweitzer, G., Bleuler, H. and Traxler, A., 2003, Active

Magnetic Bearings: Basics, Properties and Applications of Active
Magnetic Bearings, Authors Working Group, www.mcgs.ch Anton, V.L. , 2000, Analysis and initial synthesis of a
reprint. novel linear actuator with active magnetic suspension, 0-7803-
8486-5/04/$20.00 2004 IEEE
Chiba, A., Fukao, T., Ichikawa, O., Oshima, M., Takemoto,
M. and Dorrell, D.G., 2005, Magnetic Bearings & Bearingless
Drives, Newnes, Elsevier. Chee, K.L., 1999, A Piezo-on-Slider Type Linear
Ultrasonic Motor for theApplication of Positioning Stages,
Maslen, E., 2000, Magnetic Bearings, University of Proceedingsof the 1999IEEE/ASME.
Virginia.

Groom N.J. and Bloodgood, V.D. Jr., 2000, A Shyh-Leh, C., 2002, Optimal Design of a Three-Pole
Comparison of Analytical and Experimental Data for a Magnetic Active Magnetic Bearing, IEEE TRANSACTIONS ON MAGNETICS,
Actuator, NASA-2000-tm210328. VOL. 38, NO. 5.

Bloodgood, V.D. Jr., Groom, N.J. and Britcher, C.P., 2000,

Further development of an optimal design approach applied to
axial magnetic bearings, NASA-2000-7ismb-vdb.

21