Structural stability of two-dimensional active symmetric magnetic

bearing systems with 2n-pole legs

∗
Wilfried Takam Sokamte, Fabien Kenmogné, Alexandre Kongne Mando David Yemélé

Research Unit of Mechanics and Modeling of Physics Systems (RU-2MSP),

Dschang School of Science and Technology (DSST),
Faculty of Science, University of Dschang, Po Box 067, Dschang, Cameroon.
∗
Corresponding author; phone:(+237) 699 02 66 48; email: [email protected]

November 20, 2020

Abstract

In this paper, we analyzed the nonlinear dynamics of the rotor Active Magnetic
Bearings (AMBs) system with 2n pairs of electromagnets. The electromagnetic force
with 2n pole legs is obtained by applying the electromagnetic theory. Then a math-
ematical model translating the movement of the rotor-AMB is established on the
basis of Newton’s second law. The equation of the dimensionless movement which
results from the rotor-AMB system is expressed as a nonlinear system with two de-
grees of freedom including the quadratic and cubic nonlinearities whose coefficients
are a function of n. The Hamiltonian associated with the conservative system as
a function of n is established and from this Hamiltonian a state diagram linked to
is also established with the aim of delimiting the domains of stability from criti-
cal parameters of the system. It appears that some domains of the state diagram
disappear when n increases. An amplitude equation system is obtained from the
method of multiple scales. It emerges from this system that the amplitudes of vi-
bration increase when the number of electromagnets increases. Then, the presence
of the jump or bistability phenomenon appears as n increases evidence of a signifi-
cant influence of the number of electromagnet on the dynamics of the system. The
analytical studying of running speed was done under the influence of n, the result
of this study show that the increasing of n reduce the stability in our systems. The
numerical simulations carried out confirm the precision of these results obtained.
Keywords: Rotor-AMB, 2n-pole legs, Hamiltonian, State diagram, amplitudes of vibration,
bistability.

1
1 Introduction
In the areas of engineering, the development of many, as mechanical engineering, electrome-
chanical engineering, have developed many technical so that to perform his equipments. In fact,
for many industries, where the machine use the high speed have known a good increasing manu-
facturing in his productivity. Because of these constraints, rotating machines with conventional
bearings are no longer appropriate since they are the source of a number of problems including
in one hand poor damping due to mechanical contact between rotor and stator and in other
hand low dynamics stability. One of the promising solutions to overcome this limitation has
been the use of Active Magnetic Bearing(AMB) [1], because they offer a number of advantage
namely the lubrication free, lower friction loss, high speed operation, clear environment, long
service life, low maintenance and the potential of active control with high precision [2]. In
view of these advantages and other interesting features AMBs in recent years, has found wide
applications in both military and commercial machinery [3]. However, although the AMBs are
a very promising and appealing technology for machine tools, there are a number of significant
technical challenges that must be overcome to capture their full potential. It seems then natural
to ask whether it is possible to have a tool like a data base which can help with the design
criteria or with the technical characteristics of 2D AMB systems with 2n-pole legs.
Over the past two decades, there were a number of literatures that have been conducted
to study the nonlinear dynamics of magnetic bearing systems. Zhang and Zhan [4] studied
the periodic and chaotic motions of a rotor-AMB system with 8-pole legs and time varying
stiffness. They indicated that the ability of auto-controlling transient state chaos to the steady
state periodic and quasi-periodic motion exist in the system. [5], the authors discussed the
Shilnikov type multi-pulse chaotic dynamics for a rotor-AMB with quadratic and cubic terms.
Li et al. [6]-[7] found a rotor-AMB system exists respectively at 17, 19, 21 and 22 limit
cycles under the different controlling conditions and they investigated the global bifurcations
and chaos for the same systems. After the investigation, they observed that the selection of
singular points has significant influence on the number of limit cycles.
Mohamed and Emad [8] analyzed the nonlinear oscillation of a rigid rotor in two radial
active magnetic bearings and showed that the system undergoes Hopf bifurcation caused by
unstable periodic motion. Using the method of multiple scales, Ji and Hansen [9] analyzed
nonlinear dynamics of rotor-AMB system in the case of primary and internal resonances. They
are found that there exhibits many interesting phenomenon such as bifurcation, jump, sensi-
bility to initial conditions, coexistence of multiple solutions, amplitude-modulated motions and
on. They also indicated that the steady state solutions lose their stability by either saddle-node
bifurcation or bifurcation. Virgin [10], consider the effect of coordinate coupling due to the
geometry of the pole arrangement on dynamic behavior and he has pointed out that multiple
coexisting solutions and fractal boundaries can be obtained. Leung [11] examined the non-
linear oscillation of a rotor-magnetic bearing system for the case of super-harmonic resonance.
They proposed that the controller’s feedback gain and imbalance eccentricity have great in-
fluences on the nonlinear response of the system. Hegazy et al. [12] numerically exploited
the stability and the steady-state response in a time-varying stiffness rotor-AMB system under
combined resonance for various parameters. Inayat-Hussain [13]-[14] discussed the geometric
coupling effect on the bifurcations of a rotor response in active magnetic bearings using numer-
ical method. Furthermore, he presented a bifurcation study for a rigid rotor system supported
by load sharing between magnetic and auxiliary bearings.
There also has been limited research of the rotor-AMB systems focusing on the control
strategies like Flowers et al. [15] designed an integrally augmented state feedback controller
in order to eliminate static offsets which are a important problem in a rotor magnetic bearing

2
system adding to that Maslen and Montie [16] digged into the sliding mode control problem
of magnetic bearings in the case of rotor flexibility and finite amplifier voltages. And the
others works have also been conducted by others researchers whose purpose are to highlight
the conditions for which the AMBs system may experience a stable behavior, like Chinta and
Plazzolo [17] derive equations of motion of a two-DOF mass in a magnetic bearing considering
geometric coupling between the horizontal x and vertical y components of rotor motion. By
means of the numerical integration and the approximate method of trigonometric collocation
and Floquet theory, they found out that in the forced response regime, stable periodic motion
can be obtained. Similarly, Tseng et al. [18] conducted a study on a simply supported beam
controlled without contact manner by the attractive force of magnetic actuators, generated by
the feedback signals from displacement and velocity sensors necessary to stabilize the levitation.
The results showed how feedback gains inuence the behavior of transverse vibrations and must
be selected carefully to improve the stability of the system. Na et al. [19] put forward a new
algorithm for implementing fault-tolerant operation of magnetically suspended, flexible shaft,
rotating machinery.
This paper investigated nonlinear dynamics for a rotor-active magnetic bearing system with
2n pole legs using the method of multiple scales to describe the dynamic of our system [20]-
[21], and also determine the Hamiltonian of our system with 2n pair of electromagnets and
present the different form of potential well which the system can have [22] with n pair of
electromagnets. It demonstrate the influence of the number of pair of electromagnets on the
amplitudes of oscillation of displacement of rotor on the two directions through the response
Frequency-Amplitude and the bifurcation diagrams with parameters control p and l for different
values of n. The system is modeled by two-degree-of-freedom nonlinear ordinary differential
equations with quadratic and cubic nonlinearities which give us the equation which governed
the dynamics AMB system, and parametric excitation.
Thus, several scientific works have been interested in finding how to optimize the stability
of the non-linear dynamic behavior of the system while minimizing or reducing the vibration
amplitudes of the rotor in the electromagnetic bearings. The electromagnetic bearing rotor
system presents in its operation rich in non-linear phenomenon (the phenomenon of jump or
bistability, the phenomenon of chaos, ...) whose only idea of the researchers is to be able to
discover these phenomena and to control them each hence several system designs were made.
The control of these phenomena passes by the creation of mathematical models translating the
non-linear behavior of the system and the control of the coefficients of the model which are the
characteristic elements of the system. Indeed, some works were interested in the geometry of
the bearings, in the gyroscopic effect of the structure, in the electromagnetic impulses through
the electromagnets on the control of the structure, in the ferromagnetic effects,... . This is how
our works are based on the design of electromagnetic bearings with 2n pairs of electromagnets
and on the ability of these electromagnets to be able not only to keep the structure (rotor)
in suspension but also to see how optimized the dynamic stability of our system through the
number electromagnets. But the main questions that aroused our curiosity about this aspect
of the system is:
Can we use the number of electromagnets endlessly? what can be the influence of the
number on: the capacity of the system’s carrying, then the structural stability of the system
and finally the dynamic stability?
In this paper, we have structured the work as follows: after this introductory section, we
will move on to the next section, which is to present a mathematical model governing the non-
linear dynamic behavior of our system with 2n pair of electromagnets. Then, we will make a
static analysis of the stability of the structure, in addition to that, we will make a dynamic

3
study of the behavior of the rotor in the AMB with 2n pair of electromagnet thereafter we will
present, analysis and discuss the numerical results obtained. Finally will follow a conclusion
and remark.

2 Model of AMB system and governing equation

The rotor-AMB system considered is a uniform and symmetric rigid rotor suspended at the
both end by two radial AMB systems as illustrated in Fig.1. Due to this symmetric character,
each of these bearings is constituted of n = 2s identical electromagnets pairs, where s is the
natural arbitrary integer. The particular cases s = 1, 2, 3 have been deeply investigated and
have been recently revisited by Mando et al. [22]. Similarly, the case n = 4 has been recently
considered [?]. In the following, ny taking s an arbitrary integer, we generalise the model of
symmetric electromagnetic pairs. In this model, the equilibrium of the rotor shaft is assured by
the electromagnetic forces. In fact, the case where s = 1 is a 1D configuration and the cases for
which s ≥ 2 represent the 2D configuration [22]. within this configuration, the electromagnetic
force produced by the j th opposed pair of electromagnet can be expressed as follows [?]:
" (n) (n)
#
(n) (n) (n) (ib + Ij )2 (ib − Ij )2 µ0 Aα N 2
fj (Ij , qj ) = (n)
− (n)
cos θ, j = 1, 2, 3, 4, 5, ... (1)
(R0 + qj )2 (R0 − qj )2 4

(n)
with R0 = cr +lf e /2µr , where qj is the radial coordinate of the j electromagnets pair measured
from the stator, µ0 and µr are respectively the permeability of free space, that is the vacuum,
and the relative permeability of the material. N is the number of turns of the coil, A is the
cross-section area of one electromagnet, cr and lf e similarly are the nominal air gap and the
total the length of the magnetic circuit through the iron. The other parameters namely ib and
(n) (n) (n) (n)
Ij = i0 (1 − δ1j ) + ij , stand for the bias and control current, respectively, where ij is the
(n)
feedback component of the control current in the i-direction and i0 the pre-control or more
precisely the static component of the control current which produces a force to balance the
weight of the rotor at rest, while δij is the Kronecker symbol. In fact, in this feedback control
loop of AMB systems, the dynamics of the device components such as sensors, controllers, is
more complex and depend both on the rotor displacement on one hand and velocity on the
other hand. However, when the pre-control current is first supplied, the displacement around
the bearing axis during the working regime of the system becomes small so that the current
(n)
ij can take the following form [24]:
(n) (n)
ij = Kpef f qj + Kdef f q˙j (n) , j = 1, 2, 3, 4, ..., n.. (2)

As showing in Fig 1(a) and taking into account the fact that 2α = π/n, the position of the
(n)
rotor qj can be written as follows:
   
(n) π(j − 1) π(j − 1)
qj = x cos + y sin . (3)
n n
Each electromagnet pair produces an electromagnetic force with the horizontal and vertical
components given by:
n   n  
(n)
X (n) π(j − 1) (n)
X (n) π(j − 1)
Fx = fj cos and Fy = fj sin (4)
j=1
n j=1
n

4
 (n)
where fj are defined by Eq.1. In order to apply the Newton law, we assume that the rotor is
reduced to a point mass with two degree of freedom in the horizontal x and vertical y directions
as indicated in Fig.1. The equation of motion for this rotor can then be written as follows:
ẍ = Fx(n) + meΩ2 cos Ωt,
(5)
ÿ = Fy(n) + meΩ2 sin Ωt + mg,
where m, e, c and Ω, respectively, are the mass, the eccentricity of the rotor, the damping
coefficient and the angular velocity of the rotor. It is appears from Eq.(4) that the expressions
(n)
of the electromagnetic force Fx is highly nonlinear. However, due to the fact that the displace-
(n)
ment qj of the rotor will be small after the action of the precontrol current, this expression
may be approximated by a polynomial function. Thus, within these working conditions of the
bearings, the third order Taylor series can be satisfactory used to approximate this function
as demonstrated in the earlier paper [22]. Thus, by means of this third order Taylor series
development the electromagnetic forces described by Eq.(5), becomes:
Fv(n) = −α¯1 v̇ + α¯2 v̇ 2 v + α¯3 u̇2 v + α¯3 u̇v + α¯5 v − α¯6 v̇v 2
(6)
−α¯7 v̇u2 + α¯8 v 3 + α¯9 v̇u + α¯10 v̇ u̇u − α¯11 u̇uv − α¯12 uv + α¯13 vu2
and
Fu(n) = −β¯0 − β¯1 u̇ − β¯2 u + β¯3 u̇u + β¯4 v̇ 2 u + β¯5 u̇2 u − β¯2 u2 − β¯7 u̇u2
(7)
+β¯8 u3 − β¯9 v̇vu + β¯10 v̇v + β¯11 uu̇v̇ − β¯12 v 2 − β¯13 u̇v 2 + β¯14 v 2 u
where v = x/cr and u = y/cr are the dimensionless displacements. The different coefficients this
expression are given explicitly in ??. From these expressions, the precontrol current obtained
(n) (n)
by setting Fv = 0 and Fu = 0 at the origin can then be established:
(n) mgib π
i0 = tan . (8)
4a 2n
It is obvious that, Eq.(8) depends on the number of the electromagnetic pairs n and more
precisely decreases with increasing value of n as shown by Fig.2. This means that, the carrying
capacity of the bearing increases with the number of electromagnetic pairs. Similarly, the
increasing rate of this carry capacity is more important for small values of  n. For example,

(4) (8) (4) (8) (4)
n = 4 and n = 8, i0 =0, 4142135623 and i0 =0, 1989123673 leading to i0 − i0 /i0 =
 
(8) (16) (8)
0, 5197830650 while for n = 8 and n = 16 one has i0 − i0 /i0 = 0, 504850278.
Similarly, with the dimensionless variables inspired from the work by Mando et al.[22],
Eq.(6) takes the following form:
η̃ 2 2σ̃ (n) σ̃ (n)
    
(n) (n) 2 (n) (n) (n) 2η̃

 v̈ + γ 1 + b23 u + u − uu̇ + η̃v − σ̃ v̇v v̇ + γ −b23 v + uv − v u̇ u̇
3 3 3 3







(n)2 (n) (n) (n)

+ω1 v + c22 v 3 + b12 uv + c21 vu2 = F (n) cos Ω(n) r t





(9)

 (n) (n)
   
η̃ 2σ̃ 2η̃ σ̃

(n)
 (n) (n) 2 (n) 2 (n)
ü + γ 1 − µ u + η̃u − σ̃ uu̇ + v − v v̇ u̇ + γ b23 v + uv − uv̇ v̇





 3 3 3 3




(n)2 (n) (n) (n) (n)
+ω2 u + b11 u2 + c11 u3 + b22 v 2 + b21 v 2 u = F (n) sin Ωr(n) t.


5
where t = (ω0 /α(n) )t̃ is the dimensionless time with ω02 = g/R0 the characteristic frequency.
Similarly, `(n) , p and d are the dimensionless control parameters of the system defined as follows:

2dα(n) `(n) h
 π  
(n) 3π
γ(n) = (n) ; µ = 3 cot − cot ,
` an n 2n 2n

nα(n)2 nα(n)2 h p
    
(n)2 p 2 (n)2 (n)2
i
ω1 = (n) − 1+ 1− ` ; ω2 = (n) − (1 + ` ) = ε(n) ;
(n)2
2` an 2 n 2` an 2

3h2 3h2 d
η̃ = (3 − p); σ̃ (n) = ;
4 4α(n)

α(n)2 h α(n)2 h
 π       
(n) 3π (n) π 3π
b11 = (3 − p) 3 cot − cot , b22 = (3 − p) cot + cot ;
4an 2n 2n 4an 2n 2n

3nh2 α(n)2  2 3nh2 α(n)2 2

    
(n) (n)2
 (n) 8 (n)2
c11 =− p − 6p + 8(1 + ` )=−
; c22 p − 6p + 8 1 + 1 − ` ;
32`(n) an 32`(n) an 3n
π
+ cot 3π



hα(n)2
    
(n) π 3π (n) (n) cot 2n
b12 = (3 − p) cot + cot = 2b22 ; b23 = − π

2n
3π

µ(n) ;
2an 2n 2n 3 cot 2n − cot 2n

2ε(n) an `(n) hp i
α(n)2 = p ; ε(n) = sign − (1 + `(n)2 ) ;
n 2 − (1 + `(n)2 ) 2

(n) (n) (n) (n)

c21 = c12 = b21 = c11 ;

Excepted η̃, all the coefficients of the above equations strongly depends on the number of
electromagnets pairs n. The dimensionless angular frequency Ω(n) of the driving force and
amplitude of the external force F (n) are also n dependent through the relation
 
(n) (n) Ω e
Ωr = α , F (n) = Ω(n)2 . (11)
ω0 cr r

This is also the case of the pre-control current `(n) which is now given by:
π
`(n) = tan `0 and (12)
2n
with `0 = mg/4a is the dimensionless pre-control current for the 1D case.
Equation (9) describes the rotor displacement subjected to electromagnetic forces created
by the electromagnet pairs, and consequently the dynamics of AMB structure in the 2D con-
figuration where 2s is the number of these electromagnet pairs. It obvious that, this equation
is similar to the equation obtained for n = 23 [22], that is, without additional terms. However,
its coefficients ares strongly depend on n and may influence the dynamics of the system as well
as its structural stability. It will be used in the following to study the influence of n on the
properties of this AMBs system.

6
3 Effect of the number of the electromagnetic pairs on the structural
stability of AMB systems
As demonstrated by Mando et al. [22], the stability of the AMBs system is closely connected
to the properties of the Hamiltonian of the conservative part of the system. In this section, we
generalize this analysis to an arbitrary n = 2s pairs of electromagnetic coils. The associated
Hamiltonian of the system is given by:
H (n) (u, v, u̇, v̇) = Ec (u̇, v̇) + Ep(n) (u, v) (13)
with
1 2
u̇ + v̇ 2


Ec (v̇, u̇) =
2 (14)
1 (n)2 1 (n)2 1 (n) 1 (n) 1 (n) (n) 1 (n)
(n)
Ep (v, u) = ω1 v 2 + ω2 u2 + c22 v 4 + c11 u4 + b21 u2 v 2 + b22 uv 2 + b11 u3 .
2 2 4 4 2 3
(n)
where Ec (v̇, u̇) is the kinematic part and Ep (v, u) the potential energy part. The configuration
of this energy allows, at the first step, to get insight on the stability of the system. To this
end, let us determine the fixed points of this system. These fixed points are the static solution
of Eq.(8) and are given by (u = 0, v = 0). Within some conditions, the system will also
exhibit other solutions namely (u1 , 0), (u2 , 0) where u1 and u2 satisfy to the following algebraic
equation:
(n) (n) (n)2
c11 u2 + b11 u + ω2 = 0 (15)
and are given by:
q q
(n) (n) (n) (n)
(n) b11 + ∆u (n) b11 − ∆u
u1 = − (n)
, u2 = − (n)
. (16)
2c11 2c11
with
(n)
n2 α0 h2
 2  2
Bn2
    
Bn Bn
∆(n)
u = (2) 2
−48(1 + ` ) + 2 2
−3 p −6 (2)
− 5 p + 9 2 (1 + ` )
32a2n `(n)2 n2 n2 n
(17)
 2
Bn2 Bn2
  
3 Bn 2
+3p − 2 + 9 p + 12 2 p − 18 2
n2 n n
and  
π π 3π
an = cot( ) and Bn = 3 cot − cot . (18)
2n 2n 2n
(n) (n)
Let us note that, these solutions exist only if ∆u > 0. When ∆u < 0, the system will exhibit
only one fixed point. This will be the case in two different situations:
(n)
• When p ≤ pth1 with pth1 = pth1 (n), ∆u < 0, the system admits (u = 0, v = 0) as
(n)2
equilibrium point which will be stable only if ε(n) and ω1 are all positive. This will be
the case if p > 2(1 + `(n)2 ) on one hand and p > 2 1 + (1 − n2 )`(n)2 on the other hand.
 
This condition is well satisfied here since p is less than pth1 (n). As we will see below, this
single fixed point (0, 0) is unstable for p < pth1 (n).
(n)
• When p > pth1 (n), ∆u will be also negative when the pre-control current `(n) verifies
(n) (n)
`(n) < `2 or `(n) > `1 with

7
and the pre-control current
   2  q 1/2
2
Bn Bn 2
Bn (n)
 2
 n2 − 3 p − 6 n2 − 5 p + 9 n2 + ∆` 

(n)
`1 = −1

 48 

(19)
   2  q 1/2
2 2
 n2 − 3 p − 6 n2 − 5 p + 9 n2 − ∆(n)
 Bn 2 Bn Bn
`


(n)
`2 = −1

 48 


where
( 2
Bn2
 2
Bn2
  
(n) 4 Bn 3
∆` =4 −3 p −6 − 5 p + 54 2
n2 n2 n
(20)
Bn2 Bn4 Bn2 Bn4 Bn2
   
2
+ −510 2 + 36 p − (108 4 − 1116 2 )p + 81 4 − 864 2
n n n n n
Thus, the system will present a one-well configuration for the potential energy or a configuration
with one hill top.

3.1 More than one well potential

The system will also exhibit more than one equilibrium point. That is beside the equilibrium
point (0, 0) which exist for any value of the system parameters, the system may also exhibit the
fixed point (u1 , v = 0), (u2 , v = 0) even (u = 0, v1 ) and (u = 0, v2 ). This situation is possible
when the control parameter p is greater than the threshold pth1 = 1, 90142620349, with ∆u > 0.
(n) (n)
This condition is possible only if the precontrol current satisfy to the constraint `(n) ∈ [`2 , `1 ]
(n) (n)
where `1 and `2 are given by Eq.(19). Let us note also that the system may also admit the
solution (u3 , v+ ) and (u3 , v− ) with
(n) (n) (n)2
c11 u2 + 2b22 u + ω1
v2 = − (n)
, (21)
c22
where u satisfies to the the algebraic equation:
! " # !
(n) (n) (n) (n)2 (n)2 (n)
(n) b 21 3 (n) b 22 (n) (n) 2 (n)2 b ω
21 1 + 2b 22 b22 (n)2
c11 1 − (n) u + b11 − (n) (2b21 + c11 ) u + ω2 − (n)
u− (n) ω1 = 0
c22 c22 c22 c22
(22)
In this case the system will exhibit five equilibrium points. From the previous work on 1D
case and four pairs of electromagnet, the system may exhibit many domain of stability namely
A, B, C, D, E, F, G and H. The results of the state diagram of stability obtained from the
above coefficients of Eq.(9) are shown in Fig.3 for different values of n. It appear from this
state diagram of stability that, domain A remains unchanged for all values of the number of
the electromagnetic pairs n. Thus, the domain, `(n) will be always considered as a quantity of
small magnitude as it was the case of the 1D case. It is also obvious that domain G tends to
(n)2
the disappear for increasing values of n. This is also the case for domains lying between ω1
(n)2 (n)2 (n)2
and ω2 naturalized by the solid curve which disappear since ω1 tends to ω2 for high value

8
of n. Similarly, `(n) tends to `(n) for high values of n as indicated in Fig.??. The above results
suggest that in the practical view point, it is important to use the appropriate state diagram
of stability for selecting the appropriate parameters of the P D controller.
However, in order to make emphasis on the influence of the number of electromagnetic pairs
coils, n, it is important to use a quantity which does not depends implicitly on n. To this
end, let us consider the dimensionless precontrol current for the 1D case as a variable in the
y direction. The results depicted on Fig.3 show that although there is not large qualitative
changes on the shape of this diagram, it appears dilated in the y direction. More precisely,
for n = 4, domain A in which the system exhibits only a one well configuration, covers in the
vertical direction the domain ]0, 0.6[ while for n = 8, it covers ]0, 4] particularly at the abscissa
p = 3. Thus, the domain of stability of the system in the y direction increases with n, indicating
the increases of the carrying capacity of the AMB system with the number of electromagnetic
pairs coils. It appears also that, the area covered by domain F decreases when n increases and
tends to zero for high values of n namely (n > 64). This behavior is also obvious for domain G
in which the system exhibits ... configuration. Similarly the curves describing the behavior of
(n) (n)
`3 and `4 become superimposed. The analytical expression of these quantities are given by:

( )1/2
A2n 2 2
     q 
(n) 1 6 6 A 48 A (n)
`3 =
− 3 − − 4 2 p2 + 6 5 − − 4 2n p − + 36 2n + δ` −1
48 1 − n2 n n n n n n
(2
( )1/2
A2n 2 2
     q 
(n) 1 6 6 A 48 A (n)
`4 = 2

− 3 − − 4 2 p2 + 6 5 − − 4 2n p − + 36 2n − δ` −1
48 1 − n n n n n n n

2
A2 576 A2n 3
   
(n) 6 288 432
δ` = 3 − − 4 2n
p − 36 − 4
+ 2 − 384 − p
n n n n n n2

2544 A2n A4n 2

   
720 1872
+ 36 − + 2 − 2040 − + 864 4 p (24)
n n n n2 n
2
4896 A2n A4n 2 A2n A2n
      
576 3456 48
+ − 2 − 4464 − − 1728 4 p + 3456 1 − + − 36 2
n n n n2 n n n2 n n
0(n) (n)
with An = cot(π/(2n)) + cot(3π/(2n)) and ∆` = n4 δ` .

4 Dynamical behavior of the system: Influence of the number of

electromagnetic pairs of coils.
4.1 Basic formulation
In the working regime, the rotor shaft experiences not only a rotation around its axis but
also another rotation around the bearing axis. This behaviour has been studied by Mando et
al. [22] for the 1D case and for the 2D case with n = 4. In the following, we generalize this
study for arbitrary value of n. Since the dimensionless precontrol current still small for any
value of n, the following magnitude of the quantity `(n) still valid, that is:

9
 e e Ω Ω
`(n) → `(n) , → 3
d → 5/2 d, and → −1/2 . (25)
cr cr ω0 ω0
Thus, the other coefficients will conserve its weight namely:

α(n)2 → α(n)2 , γ 0(n) → 2 γ 0(n) , δω (n)2 → 2 δω (n)2 , µ(n) → µ(n) ,

(n) (n) (n) (n) (26)
σ (n) → 2 σ (n) , b11 → b11 , b12 → b12 , F (n) → 3 F (n) .
Accordingly, the amplitude of vibration in the x and y direction is given as:

(n)
(n) (n)
v1 = A1 (T1 , T2 )ei(ω2 To ) + cc
(n) (n) (n) (27)
u1 = A2 (T1 , T2 )ei(ω2 To ) + cc
with
(n) (n)
a (n) (n) a
= 1 exp(iβ1 )and A2 = 2 exp(iβ2 )
A1 (28)
2 2
where a1 and a2 obeys to the following set of coupled first order differential equations

(n) (n)
˙
(n) F (n) b21 (n)2 (n) a
a1 =− (n)
cos(ϕ1 ) − a a1
(n) 2
sin[2(ϕ2 − ϕ1 )] − 1 γ 0(n)
2ω2 8ω2 2

(n)
(n) F (n) b (n)2 (n) 1 (n) (n)3 (n) (n)2 (n) (n)
a1 ϕ˙1 = (n)
sin(ϕ1 ) + 21(n) a2 a1 cos[2(ϕ2 − ϕ1 )] + (n) (3c11 a1 + 2b21 a2 a1 ) − a1 σ
2ω 2 8ω2 8ω2

˙
(n) F (n)
(n)
c21 (n)2 (n) a
(n) .(2
a2 =− (n)
sin(ϕ2 ) + a a2
(n) 1
sin[2(ϕ2 − ϕ1 )] − 2 γ 0(n)
2ω2 8ω2 2

(n)
(n) F (n) c21 (n)2 (n)
a2 ϕ˙2 =− (n)
cos(ϕ2 ) + a a2 cos[2(ϕ2 − ϕ1 )]+
(n) 1
2ω2 8ω2
1 (n) (n)3 (n) (n)2 (n) (n) (n)
(n)
(3c22 a2 + 2c21 a1 a2 + 4δω (n)2 a2 ) − a2 σ
8ω 2
This set of equations admits a non trivial solution a10 =a20 and ϕ2 − ϕ1 =π2 where [22]
" #
2 (n)2
1 (n) 4(σ − 3γ )
a210 = (n) ∆1 + (n)
+ 4σ (30)
3c11 ∆1
with √
(n)
∆1 = (−72σγ (n)2 − 8σ 3 + 54c11 F (n)2 + 6 3∆(n) )1/3 (31)
in which
(n) (n) (n)2
∆(n) = (16γ (n)6 +32γ (n)4 σ 2 +16σ 4 γ (n)2 −72σc11 γ (n)2 F (n)2 −8c11 σ 3 F (n)2 +27c11 F (n)4 )1/2 (32)
(n)
This expression depends on n through the coefficients of the system namely c11 , γ (n) , F (n) . It
would be very important that to make the analysis on the influence of the n on the frequency
response amplitude.

10
4.2 Influence of n on the frequency response amplitude
From the above analytical investigations, the amplitude response of the system is given by
Eq.(29). The analysis of this expression shows that it strongly depends on the number of the
electromagnetic pair coils. In fact Fig.5 shows that their variation is function of the detuning
frequency σ following some values of n in the different domains of state diagram where our
system can operate like Mandoet al. had demonstrated : domain A1 , F1 , F2 , and H1 . It
appears that firstly, the amplitude a0 of vibration increase when the values of n increasing and
secondly, we notice also that in the Fig.5(a), Fig.5(b), Fig.5(c), Fig.5(d) when the value of n is
greater than 4 we observe the jump phenomenon or bistability in our system. In some figure
like the Fig.5(a), Fig.5(b), and Fig.5(c), the increasing of number of the electromagnet coils
brought our system to have a type of response that called hardening type response. However,
the response of the Fig.5(d) which corresponds to domain H1 is of the softening type response.
For the value of n = 4 the system operate around the resonance frequency that means that
the value of n is greater than 4, the system exhibit different dynamic behavior like jump
phenomenon, bistability and so on. The variation of the amplitudes of our system are born
from the vibrations of the rotor during its operation following its rotation speed. It would be
important to make an analysis of the influence of the number of electromagnets on its angular
speed which will be the subject of the following paragraph.

4.3 Influence of n on the angular frequency of the system

The analysis made above on the influence of the number of electromagnets n on the vibration
amplitudes of the system has shown the different behavior that the system can exhibit in its
operation when the number of electromagnets in the system is varied. In this paragraph, it is a
(n) (1)
question of analyzing the influence of n on the angular speed Ωr /Ωr of the structure when the
parameters p and `0 are varied. Figures (4a), (4b), (4c), and (4d) shows the different domains
of stabilities and instability for certain values of the proportional gain p = 2.5, 3.0, 3.5, 5.0.
We observe in figures (4a1 ), (4b1 ), (4c1 ), (4d1 ), that although in general for different values
of p the stable domains increase when the precontrol current `0 also increases, the angular
speeds decrease when the number of electromagnet n increases. This result therefore indicates
that the more the number of electromagnets increases, the more the angular speed of the
structure decreases as well as that of the precontrol current which refers to the mass of the
rotor. This means that the angular velocity of a heavy structure subjected to a high number
of electromagnet will be small compared to that which would be subjected to a low number of
electromagnet.
To analyze the influence of the number of electromagnets on the rotor by varying p. it would
(n) (1)
therefore be interesting to know beforehand the behavior of the angular speed Ωr /Ωr of the
shaft when p varies. This is how we observe in F ig.5(e), F ig.5(f ), and F ig.5(h), that for a
certain range of low p value, the angular speed decreases until it is canceled. In other words,
for a given mass of the structure in operation, its angular speed becomes low when the bias
current ib is low. The increasing of n in this rang of p makes the angular speed even lower as
we can see in the F ig.5(e1 ), F ig.5(f1 ), and F ig.5(h1 ). It should be noted that the proportional
gain returns the contribution of the bias current ib which is inversely proportional. Beyond this
range of values of p cited above, Fig.5(e), Fig.5(f), and Fig.5(h), shows that when the value of p
(n) (1)
increases, the angular speed Ωr /Ωr also increases up to a certain maximum value. And this
maximum value varies depending on the precontrol current `0 . It would therefore be important
to note that the greater the mass of the structure, the lower its maximum angular speed
(n) (1)
Ωr /Ωr , even when the bias current is large. Moreover, we observe in F ig.5(e1 ), F ig.5(f1 ),

11
and F ig.5(h1 ), that when n increases the maximum of the angular speed decreases whatever
the precontrol current `0 . These mean that although increasing the number of electromagnets
is important to support the load of the rotor, it reduces the rotational movement of the shaft.
It is therefore important to recommend that for a fixed mass of the rotor a small quantity the
number of electromagnet will be sufficient to maintain good stability of the rotor during its
operation.

5 Numerical simulation and discussions

In this section, we will both analyze the behavior of our system according to each domain
of the state diagram (A, B, C, E, F, G and H) of the different values of n (4, 256) through
the bifurcation diagram and then analyzed the influence of n on the displacement of the rotor
in the directions u and v.
To study the behavior of our system function of n pair of electromagnets, we use the numer-
ical method to confirm the analytic predictions. Thus, Eqs.(8) are numerically integrated using
a fourth order RungeKutta algorithm in order to analyze the influence of n on the dynamics of
our system. Results are presented in graphical forms as steady-state amplitude of the system
against time.
Despite the influence of the parameters p and `0 on the dynamic behavior of the system, in
general manner, we observe a reduction in the amplitudes of the displacements in the two direc-
tions u and v when the number of pairs increases. This is how Figures 8-12 of the displacements
u and v on the bifurcation diagrams show that these displacements decrease when n increases.
The same is true of Fig.10 on the time history of the amplitudes of the displacements of which
a clear decrease in the amplitudes is observed when n increases regardless of the values of the
parameters p and `0 .
These observations above allow us to make the following analysis on the influence of n:

5.1 Analysis of parameter p for constant values of `0 on the displacements u and

v as a function of the different values of n
Figures 11-12 show us the different forms of displacements u and v depending on the control
parameter p according to the values of the even number n = 4, and 256 corresponding to certain
domains of the state diagram. Therefore:
For a value of `0 = 0.3.
Figures (6a1 ) and (6b1 ) shows that for a precontrol current fixed at the value `0 = 0.3 under
the influence of n = 4 pairs of electromagnets, for a range of value of the proportional gain
in p between 2 and 4 corresponding to the domain F3 , and A, the displacement profiles in u
and v increase up to maximum values. The maximum value in u corresponds to p = 4 yet in
v it corresponds to p = 3.5 before decreasing. Beyond this range of value p, that is to say p
greater than 4 corresponding to the domain H1 , the profile in u decreases slightly before being
constant however the displacement in v decreases sharply up to a certain value of p = 8 then
becomes constant. When the number of electromagnets increases in other term n greater than
4, the displacement profiles in u and v described above do not change but there is a significant
reduction in the displacement scale up to the order of 10−4 like we see in figures (6c1 ) and (6d1 ).
All these mean that for a fixed load of the rotor the dynamic behavior of our system described
by the displacements in u and v will have displacement profiles as described above according
to each of the domains F3 , and A, even if the p increases.
For a value of `0 = 1.0.

12
 In this case where the precontrol current has increased compared to the case cited above,
the displacement profile in u and in v in figure 7e1 and 7f1 for n = 4 pairs of electromagnets,
show for p belonging to the interval [2.2; 4.0] corresponding to the domains F4 , F1 , F2 , and part
of H1 , increase to a certain maximum value then decrease until canceling out at a critical value
p = 4. Beyond this range of p value which is already in the domain H1 , the two displacement
profiles u and v grow weakly until they become constant when p increases. When increasing
the number of electromagnets (n > 4) in the system, in the same range of p value [2.2; 4.0] the
amplitudes in u and v decrease until canceling out at the value p = 4 then decrease slightly up
to a certain value then becomes as seen in Figures 7g1 and 7h1 . Thus, the increase in the load
of the rotor subjected to the control of several electromagnets describes a displacement profile
in both directions as presented above regardless of the increasing variation of the proportional
gain kp or the progressive reduction of the bias current ib .
Finally for a value of `0 = 1.2.
In the latter case, the displacement profile u and v is similar to that of the preceding case
but the range of values of p differs. For p belonging to [2.6; 5] corresponding to the domain
F1 , F2 , and part of H1 , with n = 4 even electromagnet, the two amplitudes increase up to a
maximum value then decrease until reaching a critical value which cancels out u and v. After
this range of values of p the displacements in u and v start again then become constant when
p increases. When we increase in our system the number of electromagnet pairs, the behaviors
are different in the two amplitude profiles as shown in figures (8i1 ) and (8j1 ). In figure (8i1 ),
in the range of p belonging to [2.5; 5] the amplitudes decrease when p increases. This means
that for a fixed charge in this range of p value the displacement in u will always decrease
when the proportional gain increases for a high number of electromagnets then for values of
p greater than 5 the displacement of the charge in u starts again until it becomes constant
when the proportional gain kp increases or when the bias current ib decreases. For the case of
displacement in v as shown in figure (8j1 ) in the same range of values of p ([2.5; 5]) we observe
an unstable behavior of the load of the rotor for a very high number of electromagnet pairs
even beyond this range where we see a doubling of displacement of v as the proportional gain
increases (f ig.(8l1 )).
Generally speaking, the analyzes made above allows the designer to have not only an idea
on the displacement profiles of the rotor in the two directions u and v and in each domain of the
state diagram when fixing its load and knowing the type of P D controller he uses to design his
rotor-AMB system but these also give him an idea of the type of displacement he can expect
when increasing or decreasing the number of electromagnet pairs .

5.2 Analysis of the parameter `0 for constant values of p on the displacements u

and v as a function of the different values of n
Figures 9-12 show us the different forms of displacements u and v depending on the control
parameter `0 according to the values of the even number n = 4, and 256 corresponding to
certain domains of the state diagram when we fix some to proportional gain p. Therefore:
when p = 2.5.
Figures (9a2 ) and (9b2 ) show that for the value range of `0 [0; 0.5] with n = 4 even elec-
tromagnets the profile of u has a parabolic shape, that is to say a maximum and two values
in which the displacement of u is canceled out. We observe that the same profile with the
displacement in v. Beyond this range of value of the precontrol current, there is a range `0
]0.5; 0.79] for which the displacement in u increases up to a maximum value, however in this
range the displacement in v decreases to a minimum value. After this range mentioned above,
the profile in u decreases until it takes on a constant pace, which is the case with the profile

13
in v, but beyond `0 = 1.0 the profile in v seems to exhibit unstable behavior. This observation
shows that although the role of the bias current is to keep the mass of the rotor in suspension,
it does not guarantee a linear behavior of the load in its displacements in u and in v during
its operation. When the system is subjected to a high number of electromagnet pairs as shown
in figures (9c2 ) and (9d2 ) the u and v profiles appear to be the same and the amplitudes are
considerably reduced compared to the case where n = 4 pairs of electromagnets. Thus, the
number of electromagnets hardly changes the dynamic behavior of our system but reduces the
amplitude of its movements.
Finally, for p=3.0; p=3.5; and p=4.5
We observe in figures (10e2 ), (10f2 ), (11i2 ), (11j2 ) (12m2 ) and (12n2 ) that they present the
same shapes as figures (9a2 ) and (9b2 ) for a system which has 4 pair of electromagnets whose
proportional gain p is fixed by varying the precontrol current `0 . It emerges from these figures
that for a fixed polarization current does not guarantee the linear stability of the system during
its operation when the mass of the rotor varies this due to the effect of inertia although this
polarization current has the role of maintaining in suspension the load whatever the fixed value
of the current. When we increase the number of electromagnets greater than 4 in our system
while keeping the proportional gain fixed, we first observe a reduction in the amplitudes of
considerable displacement as shown in figures (9g2 ), (10h2 ), (11k2 ), (11l2 ), (12o2 ) and (12p2 )
and change in the shape of the curves.

5.3 Analysis of the influence of n on the dynamic of the rotor-AMB system

The dynamic stability of the rotor load is controlled from the electromagnetic forces gen-
erated by the electromagnets through the bias current and the control current which is in the
active electromagnetic bearings. This stability is influenced by the number of electromagnets
that the bearing can have. Mando et al. have established and demonstrated that the state dia-
gram has domains where the dynamics of the system are favorable such as domains A, F1 , F2 ,
and H1 . Thus in the domain A the rotor presents a stable behavior when the bearing consists
of 4 pairs of electromagnets this can be seen in figures (13a) and (12b) where the amplitudes of
the displacements in u and in v remains constant which is verified by theirs frequencies spectra
which each has. When the rotor is subjected this time to a high number of electromagnets, i.e.
256 pairs, the amplitudes of the displacements in u and in v are greatly reduced like we are
seeing in the figures (12c) and (12d) but the behavior of the rotor remains unchanged compared
to the case of n = 4 pairs despite very low frequency spectra doubling in the direction of v. All
these are due to the fact that the domain A has a single potential well in the neighborhood of
(1)
(0, 0) that’s why our rotor operating with only one angular speed Ωr = 0.98.
In the domain F1 , the displacement of the rotor subjected to 4 pairs of electromagnets in
the directions u and v remains stable as shown in figures (13a) and (13b) and operates with
(1)
a low angular speed Ωr = 1.47 unlike the case where our rotor is subjected to 256 pairs of
(1)
electromagnets which operates with an angular speed Ωr = 7.47 and presents a fairly stable
behavior in addition to the low presence of another frequency spectrum as seen in figure (13c)
and (13d) There is also a very large reduction in amplitude. This difference in angular speed
depending on whether the rotor is under 4 pairs or 256 pairs during its operation is due to the
fact that the potential well is very low in this domain F1 and can not be used under a certain
range of angular speed depending on whether it is subjected to 4 pairs or 256 pairs. Thus the
increase of the electromagnets in the system which operates in the domain F1 so much to split
the frequency in the direction u.
Concerning the domain F2 , we observe in figures (14a) and (14b) the same behavior of the

14
rotor under 4 pairs of electromagnets stable as that observed in the domain F1 . Similarly, when
the number of pairs of electromagnets is increased to 256 pairs, figures (14c) and (14d) presents
us with quasi-periodicity with the birth of a spectrum of more frequencies in both directions.
It should be emphasized here that the angular speed of the operation of the rotor when it is
under 4 pairs of electromagnets is different than when it is under 256 pairs those due to the
low potential well in the vicinity of (0, 0). Quasi-periodicity appears in the system when the
number of electromagnet pairs increases during its operation in the domain F2 .
In the case of the H1 domain, the displacement amplitudes in u and in v are stable with
4 pairs of electromagnets as we see in figures (16a) and (16b) where the phase portrait and
frequency spectrum diagram shows that the behavior of the rotor is stable this is confirmed by
the temporal evolution of in the two directions u and v. When increasing the electromagnet pairs
by 256 pairs, one observes firstly on the system a considerable reduction in the displacement
amplitudes in both directions, secondly an increase in frequency spectrum in the direction of v
and perhaps a very slight increase in the direction of v. u as shown in figures (16c) and (16d).In
this case the angular speed of operation of the rotor remains the same both under n = 4 pairs
of electromagnets and on n = 256 pairs because in this area H1 there is only one potential well
which is located around (0, 0) which corresponds to the axis of the bearing.

6 Conclusion
Much work on the non-linear dynamics of the rotor-Active Magnetic Bearings (AMBs)
system has been carried out using even limited numbers of electromagnets such as 2, 4, 8,
finally 16 pairs [25]. No other work has gone beyond these pair numbers. Few studies have
focused on the influence that the number of pairs of electromagnets can have on dynamic
operation and especially on the stability of the system in its operating regime.
Thus in this work, the non-linear behavior of the electromagnetic bearing made up of a rigid
rotor supported by 2n pair of electromagnet was subjected to our study. A little more particular
attention on the influence of the number of electromagnet in general on the nonlinear behavior
of the rotor was considered and also its effects on the geometrical coupling of the displacement
of the rotor in the two directions. Based on the work done by [22]- [23], we highlighted the
Hamiltonian who this time took into account the number of electromagnet. This Hamiltonian
made it possible to determine the conditions of structural stability by integrating the number
n in this condition. All these will allow us to know whatever the number of electromagnets the
condition of stability of this structure under AMB. Our investigations also showed that despite
the contribution of the parameters of the control parameters such as: proportional gain p and
the precontrol current `0 (that the values are taking in [23]) characterizing the P D controller
which have a considerable impact on the control of the stability of the amplitudes of vibration
of the system, we can also use the number of electromagnet as control parameter so that to
control the amplitudes of vibration of the system even see how to control the stability of the
structure in AMB. In addition to this analysis, we highlighted the seven domains of stability
obtained by [22] but strongly influence by the number of pair n because the more this number
n increases the more certain domain of the state diagram disappear.
In addition, from this analysis, the nonlinear rotor vibrations were also examined as a
function of n in the area where the system can exhibit stability. Based on the multiple time scale
method, the 2D behavior amplitude equations for large rotational speeds of the rotor showed
that when we increase the number of pair of electromagnet it appear the jump phenomenon
or bistability in our system. This phenomenon disappear when reduce the number of pairs.
Considering these results, we have also observe through the bifurcation diagram that we plotted

15
that either we have took p or `0 as a control parameters, when we increase the number of
electromagnet coils the amplitudes of both of displacement u and v reduce. we are also noticed
that for the rotors which operate under the the higher number of pair of electromagnet increase
the frequency spectra in both of the direction in domains F1 , F2 , and H1 .
Therefore, from all these results of the work carried out, it is important to note that in the
design of an electromagnetic bearing in addition to taking into account the characteristics of
the controller considered for its good optimization of the stability of the system, it is important
to know that the number of electromagnets has a very large contribution on the load carrying
capacity and a good contribution on the stability of the system during its operation.

16
References
[1] C.R. Knospe ”Active magnetic bearings for machining applications”. Control Engineering
Practice 15: (2007) 307-313.
[2] F. Cong, J. Chen, G. Dong, M. Pecht, Vibration model of rolling element bearings in
a rotor-bearing system for fault diagnosis, Journal of Sound and Vibration 332 (2013)
2081-2097.
[3] S. Lei, A. Palazzolo, Control of flexible rotor systems with AMBs, Journal of Sound and
Vibration 314 (2008) 19-38.
[4] W. Zhang and X. P. Zhan, 2005, Periodic and chaotic motions of a rotor-active magnetic
bearing with quadratic and cubic terms and time-varying stiffness, Nonlinear Dynamics
41 (2008) 331-359.
[5] W. Zhang, J. W. Zu, and F. X. Wang, 2008, Global bifurcations and chaos for a rotor-
active magnetic bearing system with time-varying stiffness, Chaos, Solitons and Fractals
35 (2008) 586-608.
[6] W. Zhang and Jean W. Zu, 2008, Transient and steady nonlinear responses for a rotor-
active magnetic bearings system with time-varying stiffness, Chaos, Solitons and Fractals
38 (2008) 1152-1167.
[7] J. Li, Y. Tian, W. Zhang and S. F. Miao, 2008, Bifurcation of multiple limit cycles for a
rotor- active magnetic bearings system with time-varying stiffness, International Journal
of Bifurcation and Chaos 18 (2008) 755-778.
[8] A.M. Mohamed, F.P. Emad Nonlinear oscillations in magnetic b earing systems, IEEE
Transactions on Automatic Control 38 (1993) 1242-1245.
[9] J. C. Ji and C. H. Hansen, 2001, Nonlinear oscillations of a rotor in active magnetic
bearings, Journal of Sound and Vibration 240 (2001)599-612.
[10] L.N. Virgin, T.F. Walsh, J.D. Knight, Nonlinear behavior of a Magnetic Bearing System,
ASME Journal of Engineering for Gas and Turbines and Power 117 (1995) 582-588.
[11] J. C. Ji, and A.Y. T. Leung, 2003, Non-linear oscillations of a rotor-magnetic bearing
system under superharmonic resonance conditions, International Journal of Non-Linear
Mechanics 38(2001)829-835.
[12] U.H.Hegazy, M.H. Eissa and Y. A. Amer, 2008, A time-varying stiffness rotor active mag-
netic bearings under combined resonance, Journal of Applied Mechanics 75(2008)011011-
1-011011-12.
[13] Jawaid I. Inayat-Hussain, 2009, Geometric coupling effects on the bifurcations of a flexible
rotor response in active magnetic bearings, Chaos, Solitons and Fractals 41(2009)2664-
2671.
[14] Jawaid I.Inayat-Hussain, 2011, Bifurcations in the response of a rigid rotor supported by
load sharing between magnetic and auxiliary bearings, Meccanica 45(2011)1341-1351.
[15] G. T. Flower, G. Szasz, V. S. Trent, and M. E. Greene, 2001,A study of integrally aug-
mented state feedback control for an active magnetic bearing supported rotor system,
Journal of Engineering for Gas Turbines and Power 123(2001)377-381.

17
[16] E. Maslen and D. Montie, 2001, Sliding mode control of magnetic bearings: a hardware
perspective, Transactions of the ASME 123(2001)878-884.
[17] M. Chinta, A.B. Palazzolo, Stability and bifurcation of rotor motion in a Magnetic Bearing,
Journal of Sound and Vibration 214(1998)793-803.
[18] C.Y. Tseng, P.C. Tung, Dynamic of a exible beam with active nonlinear magnetic force,
Journal of Vibration and Acoustics 120(1998)39-46.
[19] U. J. Na, A. B. Palazzolo, and A. Provenza, Test and theory correlation study for a
flexible rotor on fault-tolerant magnetic bearings, Journal of Vibration and Acoustics
124(2002)359-363.
[20] A. H. Nayfeh and D. T. Mook, 1979, Nonlinear Oscillations, New York: Wiley.
[21] A. H. Nayfeh and C. M. Chin, 1995, Nonlinear inter-bifurcation in a parametrically excited
system with widely spaced frequencies, Nonlinear Dynamics 7(1995)195-216.
[22] A.K. Mando, D. Yemele, T.W. Sokamte, A. Fomethe, Structural stability and control at
high speed regime of two-dimensional active magnetic bearing systems. Journal of Sound
and Vibration 455(2019)161-187.
[23] A.K. Mando, D. Yemele, T.W. Sokamte, A. Fomethe, Structural static stability and dy-
namic chaos of active electromagnetic bearing systems: Analytical investigations and nu-
merical simulations. Journal of vibration and Control (2016) 1-20.
[24] W. Zhang, M.H. Yao, X.P. Zhan, Multi-pulse chaotic motions of a magnetic bearing system
with time virring stiffness,Chaos, Soliton. Fractals 27(2006)175186.
[25] R.Q. Wu, W. Zhang , M.H. Yao, Nonlinear dynamics near resonances of a rotor-active
magnetic bearings system with 16-pole legs and time varying stiffness, Mechanical Systems
and Signal Processing100(2018)113134.

18
 Appendix A
n
! n
! n
!
4ahd X
2 4ah3 d2 X
4 4ah3 d2 X
2 2
α¯1 = θ1ni ; α¯2 = θ1ni , α¯3 = θ1ni θ2ni ;
α0 i=1
α02 i=1
α02 i=1

n
! ( n
!)
8a`(n) h2 d X
2 p hp
(n)2
i X
2
α¯4 = θ1ni θ2ni ; α¯5 = 4ah −1+ − (1 + ` ) θ1ni ;
α0 i=1
2 2 i=1

n
! ( n
! )
4ah3 d X
4 3
 2 (n)2
 X
4 (n)2
α¯6 = (3 − p) 1 + θ1ni ; α¯7 = ah p − 6p + 8(1 + ` ) 1+ θ1ni − 8` ;
α0 i=1 i=1

n
! n
! n
!
8a`(n) h2 d X
2 8ah3 d2 X
2 2 8ah3 d X
2 2
α¯8 = θ1ni θ2ni α¯9 = θ1ni θ2ni ; α¯10 = (3 − p) θ1ni θ2ni ;
α0 i=1
α02 i=1
α0 i=1

n
! n
!
X  X
(n) 2 2 3
 2 (n)2 2 2
α¯11 = 8a` h (3 − p) θ1ni θ2ni ; α¯12 = 3ah p − 6p + 8(1 + ` ) θ1ni θ2ni ;
i=1 i=1

n
! n
! n
!
8ah3 d X X hp
i X
α¯13 = (3 − p) 2
θ1ni 2
θ2ni , β¯0 = 4a`(n) θ2ni , β¯1 = 4ah − 1+`(n)2 2
θ2ni
α0 i=1 i=1
2 i=1

n
! n
! n
!
¯ 8ah2 `(n) d X
3 ¯ 4ah3 d2 X
2 2 ¯ 4ah3 d2 X
4
β2 = θ2ni β3 = θ1ni θ2ni β4 = θ2ni
α0 i=1
α02 i=1
α02 i=1

n
! n
!
3
X 4ah d X
β¯5 = 4ah2 `(n) (3 − p) 3
θ2ni ; β¯6 = (3 − p) 4
θ2ni ;
i=1
α0 i=1

n
! n
!
3
X 8ah d X
β¯7 = ah3 p2 − 6p + 8(1 + `(n)2 ) 4
β¯8 = 2 2
 
θ2ni ; (3 − p) θ2ni θ1ni ;
i=1
α0 i=1

n
! n
!
8a`(n) h2 d X 8ah 3 2
d X
β¯9 = 2
θ2ni θ1ni ; β¯10 = (3 − p) 2
θ2ni 2
θ1ni ;
α0 i=1
α02 i=1

n
! n
!
3
X 4ah d X
β¯11 = 4ah2 `(n) (3 − p) 2
θ2ni θ1ni ; β¯12 = (3 − p) 2
θ2ni 2
θ1ni ;
i=1
α0 i=1

n
! n
!
X 4ahd X
β¯13 = 3ah3 p2 − 6p + 8(1 + `(n)2 ) 2 2
β¯14 = 2
 
θ2ni θ1ni ; θ2ni ;
i=1
α0 i=1

Similarly, the expressions of θ1ni and θ2ni are given

   
π(i − 1) π(i − 1)
θ1ni = cos ; θ2ni = sin . (34)
n n

19
Figure Captions
Figure1: Model of AMB system (a) and precontrol current function of n pair of electro-
magnets n (b).

Figure2: State diagram for n = 4 and for different values of the number of pair of electro-
magnets coils: (a) n = 4, (b) n = 16, (c) n = 64, (d) n = 256.

Figure3: The different shapes of the frequency amplitude response as a function of the
detuning for different values of the number of electromagnet coils n=4, 16, 64, 256 following the
proportional p and precontrol current `0 which correspond to different domains of state diagram
A, F1 , F2 ,and H1 , with d = 0.05, h = 1.0 and ecr = 0.03. The different color of the shape
correspond to the numbers of electromagnet coils which follows: for n = 4 blue color, n = 16the
black color, n = 64 red color and to n = 256 violet red color: (a) p = 3.0, `0 = 0.2 (domain
A); (b) p = 2.5, `0 = 1.0 (domain F1 ); (c) p = 2.5, `0 = 1.0 (domain F2 ); (d) p = 5.0, `0 = 2.0
(domain H1 ).
(n) (1)
Figure4: Variation of the critical running speed (Ωr /Ωr ) as a function of the precontrol
current `0 for different values of proportional gain p (Fig5a: p = 2.5, Fig5b: p = 3.0, Fig5c:
p = 3.5, Fig5d: p = 5.0). These shapes shows us the different running speed range of the system
(1)
which corresponds rather to the stable domain or unstable domain. It is good to notice Ωr
is angular speed in 1D. The figures a1 , b1 , c1 , d1 , show us the variation of the critical running
speed as a function of the precontrol current `0 for some values of numbers of electromagnet
coils (n = 4, n = 16, n = 64, n = 256).
(n) (1)
Figure5: Variation of the critical running speed (Ωr /Ωr ) as a function of the propor-
tional gain p for different values of precontrol current `0 (Fig6e: `0 = 0.3, Fig6f: `0 = 1.0,
Fig6h: `0 = 2.5. These shapes shows us the different running speed range of the system which
corresponds rather to the stable domain or unstable domain. The figures e1 , f1 , h1 , show us
the variation of the critical running speed as function of the proportional gain p for some values
of numbers of electromagnet coils (n = 4, n = 16, n = 64, n = 256).

Figure??: Bifurcation Diagram for a value of precontrol current `(1) = 0.3 as function of
parameter p following each case for different directions u and v: (a) and (b) n = 4 pair electro-
magnets, (c) and (d) n = 256 pair electromagnets.

Figure??: Bifurcation Diagram for a value of precontrol current `0 = 1.0 as function of

parameter p following each case for different directions u and v: (a) and (b) n = 4 pair electro-
magnets, (c) and (d) n = 256 pair electromagnets.

Figure??: Bifurcation Diagram for a value of precontrol current `0 = 2.5 as function of

parameter p following each case for different directions u and v: (a) and (b) n = 4 pair electro-
magnets, (c) and (d) n = 256 pair electromagnets.

Figure6: Bifurcation Diagram for a value of proportional gain p = 3.0 as function of

parameter `0 following each case for different directions u and v: (a) and (b) n = 4 pair elec-
tromagnets, (c) and (d) n = 256 pair electromagnets.

20
 Figure7: Bifurcation Diagram for a value of proportional gain p = 2.5 as function of
parameter `0 following each case for different directions u and v: (a) and (b) n = 4 pair elec-
tromagnets, (c) and (d) n = 256 pair electromagnets.

Figure8: Bifurcation Diagram for a value of proportional gain p = 3.5 as function of

parameter `0 following each case for different directions u and v: (a) and (b) n = 4 pair elec-
tromagnets, (c) and (d) n = 256 pair electromagnets.

Figure9: Bifurcation Diagram for a constant parameter of proportional gain p = 4.5 as

function of parameter `0 following each case for different directions u and v: (a) and (b) n = 4
pair electromagnets, (c) and (d) n = 256 pair electromagnets.

Figure??: The time history of the amplitude response and frequency spectra for the two
direction u and v of system in domain A for different value of n (4, 256) with for constant
(1)
values h = 1.0, e/cr = 0.03, p = 3.0, `0 = 0.3, d = 0.05, Ωr = 2.47 ( [23]).

Figure??: The time history of the amplitude response and frequency spectra for the two
direction u and v of system in domain F1 for different value of n (4, 256) with for constant
(1)
values h = 1.0 e/cr = 0.03, p = 5.0, `0 = 2.5, d = 0.05, Ωr = 7.47 for n = 256 ( [23]).
(1)
and Ωr = 1.47 for n = 4
Figure??: The time history of the amplitude response and frequency spectra for the two
direction u and v of system in domain F2 for different value of n (4, 256) with for constant
(1)
values h = 1.0 e/cr = 0.03, p = 2.5, `0 = 1.0, d = 0.05, Ωr = 20.0 ( [23]) for n = 256 and
(1)
Ωr = 18.0 for n = 4 .

Figure10: The time history of the amplitude response and frequency spectra for the two
direction u and v of system in domain H1 for different value of n (4, 256) with for constant
(1)
values h = 1.0, e/cr = 0.03, p = 3.5, `0 = 1.0, d = 0.05, Ωr = 0.98 [23].

21
 (a) (b)

Figure 1:

22
Figure 2:

23
 (a) (b)

(c) (d)

Figure 3:

24
Figure 4:

25
Figure 5:

26
 (a2 ) (b2 )

(c2 ) (d2 )

Figure 6:

27
 (e2 ) (f2 )

(g2 ) (h2 )

Figure 7:

28
 (i2 ) (j2 )

(k2 ) (l2 )

Figure 8:

29
(m2 ) (n2 )

(o2 ) (p2 )

Figure 9:

30
 (a) (b)

(c) (d)

Figure 10:

31