Applied Mathematical Modelling 93 (2021) 792–810

Contents lists available at ScienceDirect

Applied Mathematical Modelling

journal homepage: www.elsevier.com/locate/apm

Analysis of the rub-impact forces between a controlled

nonlinear rotating shaft system and the electromagnet pole
legs
N.A. Saeed a,∗, Emad Mahrous Awwad b,c, Mohammed A. EL-meligy d,
Emad Abouel Nasr e,f
a
Department of Physics and Engineering Mathematics, Faculty of Electronic Engineering, Menoufia University, Menouf, 32952, Egypt
b
Electrical Engineering Department, College of Engineering, King Saud University, PO Box 800, Riyadh 11421, Saudi Arabia
c
Industrial Electronics and Control Engineering Dept., Faculty of Electronic Engineering, Menoufia University, Menouf, 32952, Egypt
d
Advanced Manufacturing Institute, King Saud University, P.O. Box 800, Riyadh11421, Saudi Arabia
e
Industrial Engineering Department, College of Engineering, King Saud University, P.O. Box 800, Riyadh 11421, Saudi Arabia
f
Faculty of Engineering, Mechanical Engineering Department, Helwan University, Cairo 11732, Egypt

a r t i c l e i n f o a b s t r a c t

Article history: This article aimed to investigate the dynamical behaviours of a controlled asymmetric ro-
Received 13 August 2020 tating shaft system when the rub-impact forces between the rotor and stator occur. A
Revised 3 January 2021
nonlinear position-velocity controller is proposed to control the system’s lateral vibrations.
Accepted 4 January 2021
The suggested control algorithm is integrated into the shaft system via four electromag-
Available online 12 January 2021
netic poles that are fixed on a fixed frame and act as active actuators. The dynamics of the
Keywords: asymmetric shaft, the electromagnetic coupling between the magnetic poles and the rotat-
Asymmetric rotating shaft ing shaft, and the rub-impact forces between the rotating shaft and the stator are included
Rub-impact forces in the studied model. The derived dynamical model is firstly analyzed as a continuous non-
Full annular rub linear dynamical system utilizing the asymptotic analysis while neglecting the rub-impact
Quasiperiodic partial rub forces. Different response curves are plotted to predict the conditions under which the ro-
Poincaré map tating shaft may suffer from the rub-impact forces. Secondly, the whole system model is
Frequency spectrum analyzed numerically as a discontinuous nonlinear dynamical system utilizing the bifur-
cation diagram, Poincaré map, and frequency spectrum. The main obtained results illus-
trated that the rub-impact forces occurrence between the uncontrolled rotating shaft and
the stator induces unbounded oscillation that can destruct the rotating shaft. However, the
suggested control algorithm has improved the vibratory behaviours of the considered sys-
tem via minimizing the oscillation amplitudes and preventing the rub-impact forces for a
wide range of the disk eccentricities. Moreover, it was found that the rotating shaft can
whirl in a full annular rub mode with bounded oscillations if the proposed controller fails
to prevent the rub-impact forces occurrence due to either the large disk eccentricity or the
large asymmetric stiffness coefficients.
© 2021 Elsevier Inc. All rights reserved.

∗
Corresponding author.
E-mail addresses: [email protected]fia.edu.eg, [email protected] (N.A. Saeed).

https://doi.org/10.1016/j.apm.2021.01.008
0307-904X/© 2021 Elsevier Inc. All rights reserved.
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

1. Introduction

Nonlinear lateral vibrations are a common feature of the different rotating machinery. The main causes of the rotating
machinery vibrations are the rotating shaft eccentricity, the shaft cracks, the shaft asymmetry, the wear of the bearing, mis-
alignment when the rotating shafts are improperly assembled…etc. The asymmetry of the rotating shafts may be unintended
and appears as a result of production defects. However, sometimes the asymmetrical rotating shafts are required for some
engineering applications as in the case of the two-pole generators. Along the rotating shafts of the two-pole generators,
there exist two slots in two opposite directions on which electric coils are installed. As a result of these slots, the restoring
forces of the rotating shaft can be asymmetric in any two perpendicular directions. Therefore, asymmetrical rotating shaft
systems may have oscillatory characteristics that are not similar to the symmetrical ones. Accordingly, the nonlinear dynam-
ics of the asymmetric rotating machinery has received many investigations. Ardayfio and Frohrib [1] studied the oscillatory
behaviours of an asymmetrical rotating shaft system that is supported by symmetric bearings. The authors proved that the
stiffness coefficients of the bearings may be the main cause of system instability. Iwatsubo et al. [2] studied both the forced
and free vibrations of an asymmetric rotor system supported by asymmetric bearings. The authors found that the system
may lose its stability as a result of the shaft asymmetry or the bearing asymmetry independently. Park [3] investigated the
nonlinear oscillations of a two-pole rotor generator, while Hsieh et al. [4] studied the lateral and torsional oscillations of
an asymmetric rotor-bearings system that is modeled as a continuous Timoshenko beam. The authors in [3,4] concluded
that the system may vibrate with large oscillation amplitudes when the generator spinning speed is close or equal to the
critical speed. Shahgholi and Khadem [5] explored the lateral vibrations of an asymmetrical rotating shaft system that has a
stretching nonlinearity and is subject to both the primary and parametric excitations. Saeed [6] studied the lateral vibrations
of an asymmetric vertically supported nonlinear rotating shaft system. The system is modeled as a two-degree-of-freedom
Jeffcott rotor having both linear and nonlinear asymmetric stiffness coefficients. The author concluded that the asymmetric
rotating shaft may exhibit forward and backward whirling motions when the rotor spinning speed is close to the critical
speed. In addition, he found that the asymmetric system can respond with large oscillation amplitudes even if the disk
imbalance is very small. Saeed [7] investigated the lateral oscillations and the corresponding whirling motions of an asym-
metric horizontally supported rotating shaft system, where the shaft mass has been included in the studied model. Effects
of both the linear and nonlinear asymmetric stiffness coefficients on the system lateral vibration are explored. The author
concluded that system vibration amplitude is a monotonic increasing function of both the linear and nonlinear asymmetric
stiffness coefficients. In addition, he found that increasing the asymmetric linear stiffness coefficient to a specific limit can
eliminate the system’s backward whirling motions. The oscillatory characteristics of asymmetric cracked rotating shafts that
are modeled as a Jeffcott rotor system are investigated in [8,9], while in [10,11], the lateral vibrations and the correspond-
ing whirling motions of both the horizontally and vertically supported cracked Jeffcott rotor system have been explored at
different resonance cases.
The rub-impact forces between rotating shafts and their casing have been recognized as the main cause of the rotating
machinery failure. Therefore, many researchers have investigated the oscillatory behaviours of the rotating machinery when
the rub-impact forces are considered. Patel et al. [12] studied the interactions between the lateral-torsional vibrations and
the rub-impact forces of a nonlinear rotor system. The authors reported that the dynamic frictional coefficient, the mass
ratio, and the contact damping have complex effects on the system motion. Hu et al. [13] investigated the nonlinear dynam-
ics of an asymmetric two-disk rotor when the rub-impact forces and oil-film instability are included in the system model.
The numerical and experimental results showed that the oil-whirl decreases while whip motion keeps unchanged when the
disk eccentricity or the contact stiffness increases. Wang et al. [14] introduced theoretical and experimental investigations
for the dynamical behaviours of the turbofan rotor system when subjected to a sudden loss of a blade. The authors con-
cluded that the sudden imbalance may cause rub-impact forces that result in a full annular rub or quasiperiodic partial rub,
depending on the spinning-speed. Yang et al. studied the dynamical behaviours of a nonlinear rotor system when subjected
to both rub-impact and axial load in [15], while in [16], they investigated the effects of the disk imbalance, the rub, and
pedestal looseness on the nonlinear rotor system. Hong et al. [17] studied the modal characteristics of the rotor system
when subjected to rub-impact forces. Man et al. [18–20] studied different types of rub-impact forces in the rotor system.
Tai et al. [21] investigated the steady-state motion and stability of a simple lumped parameter rotor model when subjected
to rub-impact forces. Li et al. [22] explored the dynamical behaviors of a rotor-blade system when subjected to a normal
rubbing force. Hou et al. [23] explored the rub-impact forces of a nonlinear rotor system numerically when excited by a
maneuvering load as in the case of aircraft motors.
As the rotating machine vibrations are unavoidable phenomena during their operation, many researchers directed their
efforts towards designing different passive absorbers [24–26], and active controllers [27–30] to get rid of these undesirable
oscillations. Bab et al. [24] utilized the nonlinear energy sink as a passive absorber to mitigate the nonlinear lateral vibra-
tions and eliminate the corresponding rub-impact forces in a gas turbine rotor system. Tehrani et al. [25,26] applied both
the tuned mass damper and the nonlinear energy sink as passive absorbers to mitigate the nonlinear vibrations of a bladed
rotor system considering the rub-impact forces. Ishida and Inoue [27] studied the nonlinear lateral vibrations of a verti-
cally supported Jeffcott rotor system. The authors applied a linear active vibration absorber via four electromagnetic poles.
Saeed et al. [28,29] applied both the linear proportional-derivative controller and time-delay linear proportional-derivative
controller to eliminate the lateral vibrations of a horizontally supported Jeffcott rotor system. The authors integrated their
suggested controllers into the rotor system using electromagnetic poles. Saeed et al. [30] studied the nonlinear vibration

793
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

control of an asymmetric vertically supported rotor system. They applied linear and nonlinear position-velocity controllers
simultaneously, where the control currents in U and V-directions are suggested such that iu = γ1 u + γ2 u˙ + γ3 u3 + γ4 u˙ 3 and
iv = γ1 v + γ2 v˙ + γ3 v3 + γ4 v˙ 3 . The proposed controller is integrated into the system using four electromagnetic poles. The
authors investigated the derived model as a continuous nonlinear dynamical system with neglecting the rub-impact forces
between the rotating shaft and the stator. They concluded that the best controller for such systems is a combination of a
linear position-velocity and nonlinear velocity controller (i.e. the best control method is to design the control currents to
be iu = γ1 u + γ2 u˙ + γ4 u˙ 3 and iv = γ1 v + γ2 v˙ + γ4 v˙ 3 , with neglecting the cubic position term). Moreover, it is reported that
the system may suffer from rub-impact forces when the disk eccentricity exceeds a critical value. The authors predicted
the occurrence of rub-impact forces between the rotor and stator depending on the system oscillation amplitudes without
including these forces in the studied model.
Within this article, the dynamical behaviours of the same model studied in [30] are investigated when the rub-impact
forces between the rotating shaft and the poles legs are included. The optimal control method reported in [30] has been
applied to mitigate the considered system lateral vibrations. The whole system mathematical model is derived taking into
account the asymmetric rotor dynamics, the nonlinear electromagnetic coupling between the rotating shaft and the electro-
magnetic poles, and the rub-impact forces between the rotor and stator. As the derived mathematical model is a discon-
tinuous strong nonlinear dynamical system due to including the rub-impact forces, it has been investigated in two basic
steps. Firstly, the system is analyzed as a continuous nonlinear system while neglecting the rub-impact forces utilizing the
asymptotic analysis as in Ref. [30]. Based on the obtained solution, the conditions under which the system may suffer from
rub-impact forces between the rotating shaft and the stator are reported. In the second step of the analysis, the whole sys-
tem model (i.e. including the rub-impact forces) is investigated numerically as a discontinuous nonlinear dynamical system
by means of bifurcation diagram, whirling-orbit, and frequency-spectrum. The main obtained results illustrated that the oc-
currence of rub-impact forces between the uncontrolled rotating shaft and the poles legs induce an unbounded oscillation
that can destruct the rotating shaft system. However, the proposed controller can improve the vibratory behaviours of the
considered system via mitigating the oscillation amplitudes and preventing the occurrence of the rub-impact forces. It is
also found that the rotating shaft system can whirl in a full annular rub mode with bounded oscillations if the applied
controller fails to prevent the rub-impact forces due to either the large disk eccentricity or the large asymmetric stiffness
coefficients.

2. Mathematical model

The nonlinear asymmetrical rotating shaft system can be modeled as a vertically suspended Jeffcott rotor system con-
sisting of a rotating disk and asymmetric massless elastic shaft. Accordingly, the equations governing the system lateral
vibrations in X− and Y− directions can be given as follows [30,31]:
3
mẍ + cX x˙ + kL x + kN (x3 + xy2 ) = mδυ 2 cos(υ t + α ) + kL x cos(2υ t ) + kL y sin(2υ t ) + kN x3 cos(2υ t )
4
1 1
+ kN (y3 + 3x2 y ) sin(2υ t ) + kN (3xy2 − x3 ) cos(4υ t )
2 4
1
+ kN (y − 3x y ) sin(4υ t ) + FXM + FXC ,
3 2
(1.1)
4

3
mÿ + cY y˙ + kL y + kN (y3 + x2 y ) = mδυ 2 sin(υ t + α ) + kL x sin(2υ t ) − kL y cos(2υ t ) + kN y3 cos(2υ t )
4
1 1
+ kN (x3 + 3xy2 ) sin(2υ t ) + kN (3x2 y − y3 ) cos(4υ t )
2 4
1
+ kN (3xy − x ) sin(4υ t ) + FY M + FYC .
2 3
(1.2)
4
where FXM , FYM , are two electromagnetic forces that generated via four-poles active magnetic bearings system to control the
shaft lateral vibrations in X and Y directions as shown in Fig. 1. FXC and FYC are the horizontal and vertical components of
the rub-impact forces that occur between the rotor and stator when the oscillation amplitude exceeds the air-gap size s0 as
illustrated in Fig. 1a. The suggested control algorithm is integrated into the rotor system via two pairs of electromagnetic
poles located at two perpendicular directions (i.e. in X and Y directions) and have been fixed on the stator as shown in
Fig. 1. According to Fig. 1a, the system is designed in such a way that pole-1 and pole-3 are responsible for the system
oscillations in X direction via applying the control force FXM , while pole-2 and pole-4 control the lateral vibrations of the
system in Y direction via imposing the control force FYM . The magnetic forces FXM , and FYM can be written as follows [32]:
 2  2 
I0 − iX I0 + iX
FXM = F (x, iX ) = β − (2.1)
s0 − x s0 + x
 2  2 
I0 − iY I0 + iY
FY M = F (y, iY ) = β − (2.2)
s0 − y s0 + y

794
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

Fig. 1. (a) non-deflected rotor, and (b) deflected rotor with both lateral impact and rub force.

where β is the electromagnetic force constant, I0 is a constant current known as bias current that is fed through the four
electromagnetic poles. iX and iY are the control currents in X and Y directions, respectively. s0 is the static air-gap size
between the rotating disk and the stator as shown in Fig. 1a.
Expanding Eqs. (2) in terms of (iX iY , x, y,) up to the third-order approximation using Mclaurin series about the equilib-
rium point (iX = iY = x= y= 0) yields
4β  3 
FXM = −s0 I0 iX + s20 I02 x + 2I02 x3 − 3s0 I0 iX x2 + s20 i2X x (3.1)
s50
4β  3 
FY M = −s0 I0 iY + s20 I02 y + 2I02 y3 − 3s0 I0 iY y2 + s20 iY2 y (3.2)
s50
Because the shaft asymmetry induces multi-parametric excitations as given in Eqs. (1) [7], it has been suggested that
a combination of both linear position-velocity and cubic-velocity feedback current controllers is introduced to mitigate the
lateral vibrations of the considered system as reported in [30]. Therefore, the control currents in X and Y directions are
proposed as follows:
iX = η1 x + η2 x˙ + η3 x˙ 3 , iY = η1 y + η2 y˙ + η3 y˙ 3 (4)
where η1 and η2 are the linear position and linear velocity gains, while η3 is the nonlinear velocity. The rotating shaft is
constrained within a stationary housing (i.e. within stator) with clearance  s0 as shown in Fig. 1a. So, A contact between the
rotating shaft and the stator can occur when the radial deflection (r = x2 + y2 ) of the rotor geometric center (G) exceeds
the clearance (i.e. when r ≥ s0 ). When the contact between the rotor and stator occurs, an impact restoring force (FN ) and a
tangential frictional force (FT ) appear at the rotor-stator interface as shown in Fig. 1b. The impact force is proportional to the
interference between the rotating shaft and stator (i.e. FN ∝(r− s0 ), r− s0 ≥ 0) by the linear impact stiffness coefficient (kc ),
while the tangent force is proportional to the normal force (i.e. FT ∝FN ) by the dynamic frictional coefficient (μd ). Therefore,
it is possible to express both the impact and tangential forces as follows:
FN = kc (r − s0 ) U (r − s0 ), (5.1)

FT = μd FN = μd kc (r − s0 ) U (r − s0 ). (5.2)
where U is the unit step function. It is assumed that the shaft rotates in the counterclockwise direction. Therefore, the
frictional force direction will be constant in the negative tangential direction. From the geometry of Fig. 1b, we can obtain
the components FXC and FYC of FN and FT in X and Y directions as follows:
kc
FXC = FT sin(υ t ) − FN cos(υ t ) = (r − s0 )(μd y − x ) U (r − s0 ), (6.1)
r
kc
FYC = −FT cos(υ t ) − FN sin(υ t ) = − (r − s0 )(μd x + y ) U (r − s0 ). (6.2)
r
795
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

where cos(υ t ) = xr , sin( υ t ) = yr . Substituting Eqs. (4) into Eqs. (3), and then inserting the resulting equations and

Eqs. (5) into Eqs. (1), with introducing the dimensionless variables τ = ωt, u = x
s0 , v = y
s0 , ρ= r
s0 = u2 + v2 , we get the
following dimensionless equations of motions:
3
ü + μu u˙ + u + λ ( u3 + uv2 ) = f 2
cos( τ + α ) + α1 u cos(2 τ ) + α1 v sin(2 τ ) + α2 u3 cos(2 τ )
4
1 1 1
+ α2 (v3 + 3u2 v ) sin(2 τ ) + λ(3uv2 − u3 ) cos(4 τ ) + λ(v3
2 4 4

− 3u2 v ) sin(4 τ ) − γ1 u + γ2 u˙ + γ3 u3 + γ4 u2 u˙ + γ5 uu˙ 2 + γ6 u˙ 3
 α3
+ γ7 u2 u˙ 3 + γ8 uu˙ 4 +γ9 uu˙ 6 + (ρ − 1 )(μd v − u ) U (ρ − 1 ), (7.1)
ρ

3
v̈ + μv v˙ + v + λ(v3 + u2 v ) = f 2
sin( τ + α ) + α1 u sin(2 τ ) − α1 v cos(2 τ ) + α2 v3 cos(2 τ )
4
1 1 1
+ α2 (u3 + 3uv2 ) sin(2 τ ) + λ(3u2 v − v3 ) cos(4 τ ) + λ(3uv2
2 4 4
− u3 ) sin(4 τ ) − (γ1 v + γ2 v˙ + γ3 v3 + γ4 v2 v˙ + γ5 vv˙ 2 + γ6 v˙ 3
 α3
+ γ7 v2 v˙ 3 + γ8 vv˙ 4 +γ9 vv˙ 6 − (ρ − 1 )(μd u + v ) U (ρ − 1 ). (7.2)
ρ
4β I02 s20 kN kL s20 kN
() ≡ (˙), ω = f = sδ , υ, μ = μv = λ= α1 = , α2 = α3 =
kL cX cY kc p
where d
dτ
= m, = ω u mω , mω , , , ,
kL 1
=
ms30 0 kL kL kL

ω s0 ω 3 s3
I0 η1 , d1 η2 , d2 = I0 0 η3 , γ1 = ( p1 − 1 ), γ2 = d1 , γ3 = (3 p1 − p21 − 2 ), γ4 = (3d1 − 2 p1 d1 ), γ5 = −d12 , γ6 = d2 , γ7 =
s0
= I0
(3d2 − 2 p1 d2 ), γ8 = −2d1 d2 , γ9 = −d22 .
The system equations of motion are a discontinuous nonlinear dynamical system. Therefore, there is no closed-form solu-
tion for Eq. (7). It is important to illustrate that the system lateral vibrations x(t) in X− direction, and y(t) in Y− direction are
transformed to the normalized displacements u(τ ) = xs(t ) and v(τ ) = ys(t ) . This means that as long as |u(τ )| < 1 and |v(τ )| <
0 0
1, the rotating shaft oscillates without any contact with the stator (i.e. α 3 = 0). Accordingly, the system equations of motion
are analyzed in this work as follows: Firstly, the rub-impact forces between the rotating shaft and the stator are neglected
via setting α 3 = 0. In this case, the resulting continuous nonlinear system is analyzed by applying the multiple time scales
perturbation method as given in the appendix to determine the conditions under which the rub-impact forces between the
rotating shaft and the stator can occur. Secondly, the rub-impact forces are included and the resulting discontinuous nonlin-
ear dynamical system is investigated numerically, using bifurcation diagram, Poincaré map, and frequency-spectrum when
the contact between the rotating shaft and the stator is assured.

List of symbols
u, u˙ ,ü Displacement, velocity, and acceleration of the rotating shaft system in X direction.
v, v˙ ,v̈ Displacement, velocity, and acceleration of the rotating shaft 
system in Y direction
ρ Radial displacement of the rotating shaft system, where ρ = u2 + v2 .
μu , μ v Linear damping parameters of the rotating shaft in X and Y direction, respectively.
λ Cubic nonlinearity stiffness parameter.
f Rotating shaft eccentricity.
Rotating shaft spinning-speed.
α1 Linear asymmetric stiffness coefficient.
α2 Nonlinear asymmetric stiffness coefficient.
α3 Impact stiffness coefficient.
μd Frictional coefficient between the rotating shaft and the stator.
p1 Linear proportional control gain.
d1 , d2 Linear and nonlinear derivative control gains, respectively.
σ Detuning parameter that characterizes the closeness of the shaft spinning-speed to the system natural frequencies
(σ = − 1).
au , av Steady-state oscillation amplitudes in X and Y directions, respectively.
θ 1, θ 2 Steady-state phase angels corresponding to the vibrational amplitudes in X and Y directions, respectively.

3. Response curves and bifurcation diagrams

The autonomous nonlinear differential equations that are given in the appendix by Eq. (A2) govern the oscillation
amplitudes and the associated phase angles of the system equations of motion (i.e. Eq. (7)) when α 3 = 0. So, setting

796
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

Fig. 2. Asymmetric rotating shaft system spinning-speed response-curves when neglecting the rub-impact forces at f= 0.025, α 3 = 0.0 and μd = 0.0: (a)
steady-state oscillation amplitude au in X− direction, and (b) steady-state oscillation amplitude av in Y− direction.

Table 1
System parameters.

Physical system parameters Dimensionless system parameters

Disk radius R= 0.15m μu 0.015

Disk thickness d= 0.015m μv 0.015
Disk-mass m= 8kg λ 0.05
Disk eccentricity δ= 1.25 × 10−4 m α1 0.1
Linear damping coefficients cX =cY = 7.35Nm−1 s α2 0.03
Linear stiffness coefficient kL = 3 × 104 Nm−1 α3 5.0
Asymmetric Linear stiffness coefficient kL = 3 × 103 Nm−1 μd 0.2
Nonlinear stiffness coefficient kN = 2.4 × 108 Nm−3 p1 1.0
Asymmetric nonlinear stiffness coefficient kN = 7.2 × 106 Nm−3 d1 0.09
Impact stiffness coefficient kc = 5kL = 15 × 104 Nm−1 d2 0.09
Air-gap size s0 = 5 × 10−3 m α 0.0
The magnetic-pole cross-sectional area Aa cos ϕ = 7.44 × 10−4 m2 f 0.025
Coil turn-numbers n= 1000  1 +σ
Bias current I 0 =2 A
Magnetic permeability μ0 = 4π× 10−7 NA−2
Magnetic force constant β = 14 μ0 n2 Aa cos φ = 2.34 × 10−4
linear position control gain η1 = 400
linear velocity control gain η2 = 0.6
Nonlinear velocity control gain η3 = 6.27

d
dτ
au = ddτ av = ddτ θ1 = ddτ θ2 = 0 into Eq. (A2) results in a system of nonlinear algebraic equations that govern the steady-
state oscillation amplitudes (au & av ) and the associated phase angles (θ1 & θ2 ) of the considered system when the rub-
impact forces are neglected. By solving these algebraic equations numerically in terms of the disk eccentricity (f) or the
detuning parameter (σ ), one can obtain the system spinning-speed or the eccentricity response-curves. In addition, the sta-
bility of the obtained solution can be investigated via checking the eigenvalues of the Jacobian-matrix of the right-hand side
of Eq. (A2). By plotting the steady-state oscillation amplitudes (au & av ) against one of the system parameters (σ or f), one
can predict the whirling motion radius. It is important to remember that Eq. (A2) govern the system oscillation as long as
|u(τ ) = au cos( τ − θ 1 )| < 1 and |v(τ ) = av cos( τ − θ 2 )| < 1. This means that the obtained response curves (i.e. Figs. 2,
7, 8a, 8b… etc.) simulate accurately the oscillation amplitudes as long as au < 1 and av < 1. But, when au ≥ 1 or av ≥ 1
or both au ≥ 1 and av ≥ 1, the rub-impact forces between the rotating shaft and the stator occur, which necessitates the
investigation of the whole system model (i.e. Eq. (7) when α 3 = 0) numerically using the bifurcation diagram.
Coming back to the system parameters that are defined next to Eq. (7), with fixing the main system parameters (m,
kL , kN , s0 , I0 , ω) constant, we can deduce that the parameters α 1 = kL /kL and α2 = s20 kN /kL represent the linear and non-
linear asymmetric stiffness coefficients, respectively. α 3 = kc /kL characterizes the impact stiffness coefficient. p1 = s0 η1 /I0
denotes the linear position control gain. d1 = ωs0 η2 /I0 and d2 = ω3 s30 η3 /I0 refer to the linear and nonlinear velocity con-
trol gains, respectively. The system dynamical behaviours are explored in this work by adopting the physical parameters
and the corresponding dimensionless parameters given in Table 1 as in [30,34]. The following subsections are organized

797
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

Fig. 3. Bifurcation diagram of uncontrolled asymmetric rotating shaft system according to Fig. 2 at f= 0.025, α 3 = 5.0, and μd = 0.2 (the steady-state
whirling radius ρ versus the detuning parameter σ ).

as follows. Section 3.1 is intended to investigate the dynamical behaviours of the asymmetric system before control, while
Section 3.2 is devoted to explore the effect of the different controller parameters on the system oscillatory behaviors. In
addition, Section 3.3 discusses the system dynamics at the optimum controller parameters.

3.1. The system dynamics before control

It is worth to remember that the detuning parameter σ characterizes the closeness of the shaft spinning-speed to the
system normalized natural frequency as defined in Eq. (A3). Accordingly, we can utilize σ as a bifurcation control parameter
to represent the spinning-speed response curve of the rotor system. The spinning-speed response-curve of the asymmet-
ric system before control (i.e. γ j = 0, j= 1, 2, …9) is shown in Fig. 2 when the rub-impact between the rotating shaft
and the stator is neglected. Fig. 2 illustrates the steady-state lateral vibration amplitudes (au and av ) versus the detuning
parameter σ . It is clear from Fig. 2a and b that the asymmetric rotating shaft system can oscillate with one of four vibration
modes (that are the single periodic solution, bi-stable periodic solution, tri-stable periodic solution, and quadri-stable pe-
riodic solution), depending on the shaft spinning-speed and its initial conditions. However, if we suppose that the rotating
shaft system is constrained within the stator housing, and that the controller is turned off or abrupt failure has occurred
for it, the system motion will not be the case as illustrated in Fig. 2, where for oscillation amplitudes larger than unity
(i.e. when au ≥ 1 or av ≥ 1 or both au ≥ 1 and av ≥ 1), the rub-impact forces between the rotating shaft and the stator
will occur. Accordingly, by examining Fig. 2, it is expected that the system will be subject to rub-impact forces when the
shaft spinning-speed = 1 + σ , σ ∈ [ − 0.048, 0.058]. Fig. 3 shows the system bifurcation diagram according to the
spinning-speed response curve given in Fig. 2, where Poincaré map of the whirling radius (ρ ) is plotted against σ . The im-
pact stiffness coefficient (α 3 ) has been taken to be five times the system linear stiffness coefficient (i.e. α 3 = 5.0), while
the dynamic frictional coefficient is μd = 0.2. Comparing Fig. 3 with Fig. 2, it is clear that there is a significant agree-
ment between the spinning-speed response-curve and the bifurcation diagram, where the rub-impact forces occur when σ
∈ ( − 0.0485, 0.0585) as reported in Fig. 2. Moreover, Fig. 3 shows that the uncontrolled system oscillates with one of three
rub-impact modes (that are full annular rub, quasiperiodic partial rub, and unstable motion), depending on the shaft spin-
ning speed. According to Fig. 3, the temporal vibration of the whirling radius ρ (τ ), the corresponding steady-state whirling-
orbit, and the frequency-spectrum at σ = −0.1, 0.0, 0.04 and 0.055 are illustrated in Figs. 4, 5, and 6, respectively. It is clear
from Fig. 4 that the system oscillates in safe mode without rub-impact occurrence at σ = −0.1, while Fig. 5 confirms the
quasiperiodic partial rub oscillations at σ = 0.0. In addition, Fig. 6 demonstrates that the uncontrolled system may lose its
stability at specific spinning speeds (i.e. when = 1 + σ , σ = 0.04 or 0.055).

3.2. Sensitivity investigations

The influences of the control parameters on the dynamical behaviours of the rotating shaft system are investigated within
this section. The conditions under which the rub-impact forces can occur are explored. The oscillatory behaviours of the
controlled system with and without the occurrence of the rub-impact forces are analyzed. The effect of both the linear
and nonlinear-velocity feedback control gains on the system spinning-speed response-curves when p1 = 1.0 is compared in
Fig. 7 for two different values of the disk eccentricity f. Fig. 7a and b compares the effect of both the linear and nonlinear-
velocity feedback controllers on the system spinning-speed response-curve when the eccentricity f= 0.01, while Fig. 7c and

798
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

Fig. 4. uncontrolled asymmetric rotating shaft system at f= 0.025, α 3 = 5.0, μd = 0.2, and σ = −0.1 according to Fig. 3:(a)The instantaneous normalized
whirling radius, (b) Steady-state whirling orbit, and (c) frequency-spectrum.

Fig. 5. uncontrolled asymmetric rotating shaft system at f= 0.025, α 3 = 5.0, μd = 0.2, and σ = 0.0 according to Fig. 3:(a)The instantaneous whirling
radius, (b) Steady-state whirling orbit, and (c) frequency-spectrum.

Fig. 6. uncontrolled asymmetric rotating shaft system at f= 0.025, α 3 = 5.0, and μd = 0.2 according to Fig. 3:(a) Instantaneous whirling radius when
σ = 0.04, (b,) Instantaneous whirling radius when σ = 0.055.

Fig. 7. The effect of both the linear-velocity and nonlinear-velocity feedback control gains on the system spinning-speed response-curves when p1 = 0.1:
(a, b) the disk eccentricity f= 0.01, (c, d) the disk eccentricity f= 0.05.

799
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

Fig. 8. The system spinning-speed response-curves and the corresponding bifurcation diagram at f= 0.1, p1 = 0.1, d1 = 0.09 and d2 = 0.0: (a, b) The
spinning-speed response-curves, (c) The corresponding bifurcation diagram.

d illustrates the same comparison when f= 0.05. In general, Fig. 7 confirms that the linear-velocity feedback controller has
higher efficiency than the nonlinear one at small disk eccentricity (i.e. when f= 0.01), while the nonlinear-velocity feedback
controller exhibits higher efficiency than the linear one at the large disk eccentricity (i.e. when f= 0.05) as reported in [30].
Fig. 8a and b shows the system spinning-speed response curve and the corresponding bifurcation diagram at p1 = 1.0,
d1 = 0.09 and d2 = 0.0 at the large disk eccentricity f= 0.1, while Fig. 9a and b illustrates the asymmetric system spinning-
speed response-curves and the corresponding bifurcation diagram at p1 = 1.0, d1 = 0.0 and d2 = 0.09. It is clear from Fig. 8a
and b that the maximum steady-state oscillation amplitude is approximately twice the maximum oscillation amplitudes that
are illustrated in Fig. 9a and b. This confirms that the nonlinear-velocity feedback control has higher efficiency in suppressing
the system lateral vibrations than the linear-velocity one. In addition, Fig. 8a and b shows that the system lateral vibration
amplitudes are higher than unity (i.e. au ≥ 1 and av ≥ 1) when the shaft spinning speed = 1 + σ , σ ∈ [ − 0.057, 0.0 72].
This implies that the controlled system may be subjected to rub-impact forces between the rotating shaft and the magnetic
poles legs when the shaft spinning-speed belongs to the interval = 1 + σ , σ ∈ [ − 0.057, 0.072]. Therefore, according
to Fig. 8a and b, the system bifurcation diagram is obtained as in Fig. 8c in such a way that considers the impact stiffness
coefficient α 5 = 5.0 and the dynamical frictional coefficient μd = 0.2. The figure demonstrates that the system may suffer
from a rub-impact occurrence when σ ∈ [ − 0.057, 0.072], where the system will oscillate by one of three whirling modes,
namely full annular rub, quasiperiodic partial rub, and period-3 partial rub. Comparing Fig. 8a, b with Fig. 8c, it is clear that
there is excellent agreement between the spinning-speed response-curves and the bifurcation diagram, where the three
figures show that the system may be subject to rub-impact forces when σ ∈ [ − 0.057, 0.072].

800
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

Fig. 9. The system spinning-speed response-curves and the corresponding bifurcation diagram at f= 0.1, p1 = 0.1, d2 = 0.09 and d1 = 0.0: (a, b) The
spinning-speed response-curves, (c) The corresponding bifurcation diagram.

Fig. 9c illustrates the system bifurcation diagram according to Fig. 9a, and b. The figure shows that the asymmetric system
will suffer from rub-impact forces occurrence when σ ∈ [ − 0.068, 0.078], where the system can oscillate by one of three
oscillation modes, namely full annular rub, quasiperiodic partial rub, and period-3 partial rub. In general, Fig. 9 confirms the
excellent correspondence between the obtained spinning-speed response-curves in Fig. 9a, b, and the numerical bifurcation
diagram in Fig. 9c, where the three figures confirm the occurrence of the rub-impact forces when σ ∈ [ − 0.068, 0.078].
Accordingly, Figs. 2, 8a, b, 9a, and b confirm the possibility of accurate determination of the conditions under which the rub-
impact forces can occur for the discontinuous nonlinear system given by Eq. (7) via solving the corresponding continuous
system. However, obtaining the bifurcation diagram is important to enable us to explore the nature of the system motion
when the rub-impact occurrence is assured.
The effects of both the linear and nonlinear-velocity feedback controller on the system eccentricity response-curves
when p1 = 1.0 are compared in Fig. 10. It is clear from the figure that the linear-velocity feedback controller has higher
efficiency than the nonlinear one as long as the disk eccentricity f < 0.05, while for the large disk eccentricity (i.e.f > 0.05),
the nonlinear-velocity feedback controller exhibits vibration suppression efficiency higher than the linear-velocity one. Also,
Fig. 10 shows that the controlled system may be subject to rub-impact forces as long as the disk eccentricity is larger than
a critical value (i.e. if f > fcrit = 0.0262) in the case of the linear-velocity feedback controller (i.e. at p1 = 1.0, d1 = 0.09,
d2 = 0.0). However, the system may suffer from the rub-impact forces when the nonlinear-velocity feedback controller is
applied (i.e. at p1 = 1.0, d1 = 0.0, d2 = 0.09) regardless of the disk eccentricity magnitude.

801
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

Fig. 10. The effect of both the linear and nonlinear velocity feedback controllers on the system eccentricity response-curves at.

Fig. 11. The controlled system bifurcation diagram according to Fig. 10:(a) the linear-position and linear-velocity feedback controller is applied, and (b) the
linear-position and nonlinear-velocity feedback controller is applied.

According to Fig. 10, the system bifurcation diagrams are obtained assuming α 3 = 5.0 and μd = 0.2 as in Fig. 11a
(when p1 = 1.0, d1 = 0.09, d2 = 0.0) and Fig. 11b (when p1 = 1.0, d2 = 0.09, d1 = 0.0). Fig. 11a confirms that the system
can perform a periodic oscillatory motion without any physical contact between the rotating shaft and the stator as long
as the disk eccentricity f < fcrit = 0.0262, while for f > fcrit the system exhibits rub-impact forces that will result in the
system oscillation with a quasiperiodic partial rub mode. Fig. 11b illustrates that the system will be subject to rub-impact
forces regardless of the disk eccentricity magnitudes, where the rotating shaft can exhibit one of three oscillation modes
(that are quasiperiodic partial rub, period-3 motion, and full annular rub) depending on the disk eccentricity magnitude. By
comparing the eccentricity response-curves given in Fig. 10 with the bifurcation diagrams given in Fig. 11, we can notice the
excellent correspondence between the analytic and numerical results.
Based on Fig. 11, the whirling-orbits and the corresponding frequency-spectrum of the considered system are illustrated
in Figs. 12 and 13 at four different values of the disk eccentricity. Fig. 12 simulates the system whirling behaviours according
to Fig. 11a (i.e. when the linear-position and linear-velocity feedback controller is applied), while Fig. 13 shows the system
whirling motion according to Fig. 11b (i.e. when the linear-position and nonlinear-velocity feedback controller is activated).
It is clear that there is a significant agreement between Figs. 11a and 12, where the rotating shaft system can per-
form periodic whirling motion without any physical contact with the stator when f= 0.02 as in Fig. 12a and b, while at
the disk eccentricities f= 0.3, 0.7, and 0.9, the system exhibits quasiperiodic motion as in Fig. 12c–h. By examining the
frequency-spectrum given in Fig. 12b, d, f, and h, we can notice from Fig. 12b that the frequency components of the sys-
tem motion consist of a single angular frequency 1 that is equal to the excitation frequency (i.e. = 1 = 1). This
means that the only reason for the system oscillation before the rub-impact occurrence is the eccentricity excitation forces
(i.e. f 2 cos( τ + α ) and f 2 sin( τ + α )). However, Fig. 12d, f, and h show that the frequency components of the sys-
tem motion consist of two incommensurate angular frequencies 1 and 2 when the rub-impact forces occur. The first

802
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

Fig. 12. The system steady-state whirling-orbit and the corresponding frequency-spectrum according to Fig. 11a (i.e. when σ = 0, p1 = 1.0, d1 = 0.09,
d2 = 0.0, α 3 = 5.0, and μd = 0.2) at different values of the disk eccentricity: (a, b) f= 0.02, (c, d)) f= 0.3, (e,f)) f= 0.7, and (g, h)) f= 0.9.

frequency component ( 1 = 1) arises due to the eccentricity excitation force, while the second frequency component ( 2 )
appears due to the rub-impact forces which has an angular frequency depending on the excitation force (i.e. at f= 0.3
then 2 = 1.534, at f= 0.7 then 2 = 1.467, at f= 0.9 then 2 = 1.429). Moreover, Fig. 12d, f, and h illustrate that increasing
the disk eccentricity decreases the peak amplitude of the angular frequency 1 and increases the peak amplitude of the
angular frequency 2 . This means that the increase of the disk eccentricity increases the impact force magnitude. One of
the other important notes from Fig. 12d, f, and h is that the increase of the disk eccentricity decreases the frequency of the
impact force.
Fig. 13 illustrates the system whirling-orbits and the corresponding frequency-spectrum at four different values of the
disk eccentricity according to Fig. 11b (i.e. when the linear-position and nonlinear-velocity feedback controller is applied).
Fig. 13a, c, and e shows that the system exhibits quasiperiodic partial rub motions, while Fig. 13g illustrates that the system
performs period-1 motion. It is clear from Fig. 13b, d, f, and h that the system motion consists of two frequency components
( 1 and 2 ). The first component ( 1 ) is exciting due to the disk eccentricity f, while the second component ( 2 ) is exciting
because of the impact force. By comparing Fig. 13b, d, f, and h, we can deduce that increasing the disk eccentricity increases
the angular frequency of the impact force ( 2 ) from 2 = 1.726 at f= 0.02 to 2 = 2.0 at f= 0.9. Accordingly, the system
responds with a periodic or aperiodic motion depending on the ratio between the angular frequencies 1 and 2 , whereas
long as 1 and 2 are incommensurate, the system will exhibit aperiodic motion as in Fig. 13b, d, and f. However, the
system can perform periodic motion when 1 and 2 are commensurate as in Fig. 13h.
The performance of both the linear and nonlinear-velocity feedback controllers in suppressing the considered system
lateral vibrations is compared in Figs. 14 and 15 at different values of the parametric forces α 1 and α 2 . Fig. 14 compares the
efficiency of the linear and nonlinear-velocity controllers in mitigating the system lateral vibrations at two different values
of the linear asymmetric stiffness coefficient α 1 . Figs. 14a and b compare the controllers’ efficiency at α 1 = 0.1, while Fig. 14c
and d present the same comparison at α 1 = 0.2. The efficiency of the linear and nonlinear-velocity feedback controllers in
mitigating the system lateral vibrations at two different values of α 2 is compared in Fig. 15. Fig. 15a and b compares the
controllers’ efficiency at α 2 = 0.05, while Fig. 15c and d shows the same comparison at α 2 = 0.1. In general, Figs. 14 and
15 confirm that the nonlinear-velocity feedback controller has higher efficiency than the linear-velocity one in mitigating
the nonlinear vibrations of the parametrically excited systems.

3.3. The controlled system dynamics

Based on the above discussion, combining both the linear and nonlinear-velocity feedback controllers in one controller
will improve the control efficiency against the different excitation forces. The spinning-speed response-curves of the asym-

803
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

Fig. 13. The system steady-state whirling-orbit and the corresponding frequency-spectrum according to Fig. 11b (i.e. when σ = 0, p1 = 1.0, d1 = 0.0,
d2 = 0.09, α 3 = 5.0, and μd = 0.2) at different values of the disk eccentricity: (a, b) f= 0.02, (c, d)) f= 0.3, (e,f)) f= 0.7, and (g, h)) f= 0.9.

Fig. 14. The effect of both the linear-velocity and nonlinear-velocity feedback control gains on the system spinning-speed response-curves at p1 = 0.1: (a,
b) when the linear asymmetric stiffness coefficient α 1 = 0.1, (c, d) when the linear asymmetric stiffness coefficient α 1 = 0.2.

Fig. 15. The effect of both the linear-velocity and nonlinear-velocity feedback control gains on the system spinning-speed response-curves at p1 = 0.1: (a,
b) when the nonlinear asymmetric stiffness coefficient α 2 = 0.05, (c, d) when the nonlinear asymmetric stiffness coefficient α 2 = 0.1.

804
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

Fig. 16. The controlled system spinning-speed response-curves when p1 = 1.0, and d1 = d2 = 0.09 at four different values of the disk eccentricity: (a)
steady-state lateral vibration amplitude au in X− direction, and (b) steady-state lateral vibration amplitude av in Y− direction.

Fig. 17. The controlled system eccentricity response-curves at p1 = 1.0, and d1 = d2 = 0.09 when σ = 0: (a) steady-state lateral vibration amplitude au in
X− direction, (b) steady-state lateral vibration amplitude av in Y− direction, and (c)bifurcation diagram at α 3 = 5.0, μd = 0.2.

metric rotating shaft system when both the linear and nonlinear-velocity feedback controllers are applied simultaneously
(i.e. when p1 = 1.0, and d1 = d2 = 0.09) are illustrated in Fig. 16 at four different levels of the disk eccentricity. Comparing
Fig. 16 with Fig. 2, it is clear that the nonlinear behaviours of the considered system such as the coexistence of multi-stable
solutions and jump phenomenon as in Fig. 2 have been eliminated, and that the system responds as a linear one. In addition,
Fig. 16 demonstrates that the system oscillation amplitudes became less sensitive to the disk eccentricity. Moreover, the fig-
ure shows that the controlled system may suffer from the rub-impact forces at the large disk eccentricity f= 0.15. According
to Fig. 16, the controlled system eccentricity response-curves and the corresponding bifurcation diagram when σ = 0 are
depicted in Fig. 17. It is clear from Fig. 17a and b that the system can oscillate in safe mode (i.e. without rub-impact oc-
currence) as long as the disk eccentricity f < fcrit = 0.112; otherwise, the system may be subject to rub-impact forces. To
confirm the obtained results in Fig. 17a and b, the system bifurcation diagram is illustrated in Fig. 17c. The Poincaré map
of the steady-state whirling radius is plotted against the eccentricity f at p1 = 1.0, d1 = d2 = 0.09, α 3 = 5.0, μd = 0.2
and σ = 0. Comparing Figs.17a and 17b with Fig. 17c, it is clear that there is a significant correspondence between the
approximate solution and numerical results, where the figures show that the critical disk eccentricity is fcrit = 0.112. More-
over, Fig. 17c shows that the controlled system can execute periodic whirling motion with whirling radius ρ > 1 when the
disk eccentricity f > fcrit . This means that the controlled system will exhibit full annular rub motion as long as the disk
eccentricity f > fcrit . According to Fig. 17c, the whirling-orbit and the corresponding frequency-spectrum of the controlled
system at two different values of the disk eccentricity are illustrated in Fig. 18. It is clear from Fig. 18a and b that the
system performs periodic oscillation without rub-impact forces occurrence, while Fig. 18c and d demonstrate the periodic
oscillation of the system in a full annular rub mode in the absence of the impact force. The frequency-spectrum in Fig. 18d

805
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

Fig. 18. The controlled system whirling-orbit and the corresponding frequency-spectrum according to Fig. 17c at p1 = 1.0, d1 = d2 = 0.09, α 3 = 5.0,
μd = 0.2 and σ = 0 at four different values of the disk eccentricity: (a, b) f= 0.1, and (c, d) f= 0.7.

Fig. 19. The controlled system α 1 response-curves when p1 = 1.0, and d1 = d2 = 0.09 at σ = 0: (a) steady-state lateral vibration amplitude au , (b)
steady-state lateral vibration amplitude av , and (c) bifurcation diagram when α 3 = 5.0, μd = 0.2.

confirms the disappearance of the impact force between the rotating shaft and the stator, where only the frequency com-
ponent 1 that is excited due to the disk eccentricity appeared.
In Fig. 19, α 1 is utilized as a bifurcation control parameter to determine the critical value of the asymmetric linear
stiffness coefficient at which the system starts to suffer from the rub-impact forces. It is clear from Fig. 19a and b that the
applied controller can prevent the rub-impact occurrence as long as α 1 < α 1crit = 0.1777, but for α 1 > 0.1777, the system may
be subjected to rub-impact forces when the shaft spinning speed is close to the system’s natural frequency. To investigate
the nature of the system motion when the rub-impact is assured (i.e. when α 1 > α 1crit = 0.1777), the system bifurcation
diagram is illustrated in Fig. 19c. It is clear from the figure that there is a good agreement with Fig. 19a and b. In addition,
Fig. 19c shows that the system will oscillate in a full annular rub mode when α 1 > α 1crit . Fig. 20 investigates the effect of
increasing the nonlinear asymmetric stiffness coefficient α 2 on the dynamical behaviours of the controlled system. Fig. 20a
and b confirms that the system may suffer from a rub-impact when α 2 > 0.2408, which is confirmed in Fig. 20c. Based
on Figs. 17c, 19c, and 20c, we can conclude that the controlled system will oscillate in a full annular rub mode with the
disappearance of the impact force between the rotating shaft and the stator when the applied controller fails to prevent the
contact between the shaft and the stator.

4. Actual control currents in ampere

The actual lateral vibrations (x = s0 u & y= s0 v) in meter and the corresponding actual control currents iX and iY in
ampere are simulated in Fig. 21 according to Fig. 16 (i.e. p1 = 1.0, d1 = d2 = 0.09) when σ = 0.0 at three different val-
ues of the disk eccentricity (i.e. f= 0.025, 0.05, and 0.1). The control currents (iX and iY ) are obtained using Eq. (4), where
iX = η1 (s0 u ) + η2 (s0 ωu˙ ) + η3 (s0 ωu˙ )3 and iY = η1 (s0 v ) + η2 (s0 ωv˙ ) + η3 (s0 ωv˙ )3 . It is clear from Fig. 21 that the increasing
of the disk eccentricity increases the system lateral vibrations, which in turn increases the control currents. In addition, the
maximum control current amplitude does not exceed two amperes at the disk eccentricity f= 0.1. Accordingly, the maxi-
mum total current in each pole (i.e. I0 + iX or I0 + iY ) does not exceed four amperes, which confirms the feasibility and

806
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

Fig. 20. The controlled system α 2 response-curves when p1 = 1.0, and d1 = d2 = 0.09 at σ = 0: (a) steady-state lateral vibration amplitude au , (b)
steady-state lateral vibration amplitude av , and (c) bifurcation diagram when α 3 = 5.0, μd = 0.2.

Fig. 21. The actual system lateral vibrations and the corresponding control currents according to Fig. 16 when σ = 0.0 at three different values of the shaft
eccentricity: (a, b) the system lateral vibrations in X and Y direction in meter, and (c, d) the corresponding control currents iX and iY in ampere.

applicability of the applied controller in suppressing the system vibrations with small control currents even in the case of
the large disk eccentricity.

5. Conclusions

The dynamical behaviours of the controlled asymmetric rotating shaft system investigated in [30] have been analyzed
numerically within this work when including the rub-impact forces between the rotor and stator. The whole system math-
ematical model is derived, taking into account the asymmetric rotor dynamics, the electromagnetic coupling, and the rub-
impact forces between the rotating shaft and the poles legs. As the obtained mathematical model is a discontinuous strong

807
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

nonlinear dynamical system, the analyses are introduced in two stages. In the first stage, the rub-impact forces between the
rotor and stator are neglected and the corresponding continuous system is investigated applying perturbation analysis. In
the second stage of analysis, the whole system model including the rub-impact forces is investigated numerically as a dis-
continuous strong nonlinear dynamical system by means of the bifurcation diagram, Poincaré map, and frequency spectrum.
It is found that there is an excellent agreement between all obtained response curves and the corresponding bifurcations
diagrams, where the conditions under which the system may suffer from the rub-impact forces are accurately determined
using response curves and then verified numerically using the bifurcation diagrams. Based on the presented study, the fol-
lowing important remarks can be concluded.

1. The rotating shaft system may suffer from rub-impact forces when the linear position-velocity feedback controller is
only activated as long as the shaft eccentricity is greater than 0.0262, where the system will oscillate in a quasiperi-
odic partial rub mode.
2. The rotating shaft system will be subject to rub-impact forces regardless of shaft eccentricity magnitude when only
the linear-position nonlinear-velocity feedback controller is activated, where the system will exhibit one of three os-
cillation modes, namely quasiperiodic partial rub, period-3 motion, and full annular rub.
3. When the proposed control strategy (that is a combination of both linear position-velocity and cubic-velocity feedback
controllers) is activated, the rotating shaft system can operate safely without rub-impact forces occurrence as long as
the shaft eccentricity is lower than 0.112. Moreover, if the controller fails to prevent the rub-impact forces at the shaft
eccentricity greater than 0.112, the system will oscillate periodically in a full annular rub mode.
4. The system can exhibit a periodic motion only if the impact force has disappeared as in Figs. 17c, 19c, and 20c (i.e.
the system operates in a full annular rub mode) or when the angular frequencies of the impact force and the shaft
spinning speed are commensurate as in Fig. 13g.
5. The asymmetric rotating shaft system may lose its stability at a specific operational speed and responds with un-
bounded oscillations if the applied controller abruptly fails as shown in Fig. 6.

Declaration of Competing Interest

The authors declared no potential conflicts of interest concerning the research, authorships, and/or publication of this
article.

Acknowledgments

The authors are grateful to the Researchers Supporting Project number (RSP-2020/164), King Saud University, Riyadh,
Saudi Arabia.

Appendix

The equations of motions (i.e. Eq. (7)) are analyzed as a continuous nonlinear system after neglecting the rub-impact
forces between the rotating shaft and the stator via setting α 3 = 0. By applying the multiple time scale perturbation method
following the same solution procedure as in [30,33], we can get the periodic solution of Eq. (7) as:

u(τ ) = au cos( τ − θ1 ) , (A1.1)

v(τ ) = av cos( τ − θ2 ). (A1.2)

where au and av denote the oscillation amplitudes of the system in X and Y− directions, respectively, while θ 1 and θ 2
are the corresponding phase angles. The amplitude-phase modulating equations that govern the system amplitudes and the
corresponding phase angle are given as follows:

d 1 1 1 1 1 3
a u = − μu a u + α1 au + α2 a3u sin(2θ1 ) − λa3u sin(4θ1 ) − α1 av + α2 a2u av
dτ 2 4 8 64 4 32
3 3 3
+ α2 a3v cos(θ1 + θ2 ) + λau a2v sin(2θ1 + 2θ2 ) − λau a2v sin(2θ1 − 2θ2 )
32 64 32
3 3 1
+ λa2 av cos(3θ1 + θ2 ) − α2 a2u av cos(3θ1 − θ2 ) − λa3v cos(θ1 + 3θ2 )
64 u 32 64
1 1 1 1 1
+ α2 a3v cos(θ1 − 3θ2 ) + f 2 sin(θ1 + α ) − γ2 au − (γ4 + 3γ6 )a3u − γ7 a5u , (A2.1)
32 2 2 8 16

808
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

d 1 1 1 1 1 3
a v = − μv a v − α1 av − α2 a3v sin(2θ2 ) − λa3v sin(4 θ2 ) − α1 au + α2 au a2v
dτ 2 4 8 64 4 32
3 3 3
+ α2 a3u cos(θ2 + θ1 ) + λa2u av sin(2θ2 + 2θ1 ) − λa2u av sin(2θ2 − 2θ1 )
32 64 32
3 3 1
− λau a2v cos(3θ2 + θ1 ) − α2 au a2v cos(3θ2 − θ1 ) + λa3u cos(θ2 + 3θ1 )
64 32 64
1 1 2 1 1 1
+ α2 a3u cos(θ2 − 3θ1 ) − f cos(θ2 + α ) − γ2 av − (γ4 + 3γ6 )a3v − γ7 a5v , (A2.2)
32 2 2 8 16

d 9 3 1 1  1 1
θ1 = σ − λa2u − λa2v + α1 + α2 a2u cos(2θ1 ) − λa2u cos(4θ1 ) + α1 av
dτ 32 16 4 4 64 4a1
3 9 3 3
+ α2 a3v + α2 au av sin(θ1 + θ2 ) + λa2v cos(2θ1 + 2θ2 ) − λa2v cos(2θ1 − 2θ2 )
32au 32 64 32
3 3 1
− λau av sin(3θ1 + θ2 ) + α2 au av sin(3θ1 − θ2 ) + λa3 sin(θ1 + 3θ2 )
64 32 64au v
1 1 1 1 1 5
− α2 a3v sin(θ1 − 3θ2 ) + 2
f cos(θ1 + α ) − γ1 − (3γ3 + γ5 )a2u − γ8 a4u − γ9 a6u , (A2.3)
32a1 2au 2 8 16 128

d 9 3 1 1 1 1
θ2 = σ − λa2v − ρ a2u − α1 − α2 a2v cos(2θ2 ) − λa2v cos(4 θ2 ) + α1 au
dτ 32 16 4 4 64 4av
9 3 3 3
+ α2 au av + α2 a3u sin(θ2 + θ1 ) + λa2u cos(2θ2 + 2θ1 ) − λa2u cos(2θ2 − 2θ1 )
32 32av 64 32
3 3 1
+ λau av sin(3θ2 + θ1 ) + α2 au av sin(3θ2 − θ1 ) − λa3 sin(θ2 + 3θ1 )
64 32 64av 1
1 1 1 1 1 5
− α2 a3u sin(θ2 − 3θ1 ) + 2
f sin(θ1 + α ) − γ1 − (3γ3 + γ5 )a2v − γ8 a4v − γ9 a6v . (A2.4)
32av 2av 2 8 16 128
where,

=1+σ (A3)

References

[1] D. Ardayfio, D.A. Frohrib, Instabilities of an asymmetric rotor with asymmetric shaft mounted on symmetric elastic supports, J. Eng. Ind. 98 (4) (1976)
1161–1165.
[2] T. Iwatsubo, Y. Tsujiuchi, T. Inouev, Vibration of asymmetric rotor supported by oil film bearings, Arch. Appl. Mech. 56 (1) (1986) 1–15.
[3] J. Park, Diagnosis of excessive vibration signals of two- pole generator rotors in balancing, KSME Int. J. 14 (6) (20 0 0) 590–596.
[4] S. Hsieh, J. Chen, A. Lee, A modified transfer matrix method for the coupled lateral and torsional vibrations of asymmetric rotor-bearing systems, J.
Sound Vib. 312 (4–5) (2008) 563–571.
[5] M. Shahgholi, S.E. Khadem, Primary and parametric resonances of asymmetrical rotating shafts with stretching nonlinearity, Mech. Mach. Theory 51
(2012) 131–144.
[6] N.A. Saeed, On vibration behavior and motion bifurcation of a nonlinear asymmetric rotating shaft, Arch. Appl. Mech. 89 (2019) 1899–1921.
[7] N.A. Saeed, On the steady-state forward and backward whirling motion of asymmetric nonlinear rotor system, Eur. J. Mech. A Solids 80 (2020) 103878,
doi:10.1016/j.euromechsol.2019.103878.
[8] Q. Han, F. Chu, The effect of transverse crack upon parametric instability of a rotor-bearing system with an asymmetric disk, Commun. Nonlinear Sci.
Numer. Simul. 17 (12) (2012) 5189–5200.
[9] Q. Han, F. Chu, Parametric instability of a Jeffcott rotor with rotationally asymmetric inertia and transverse crack, Nonlinear Dyn. 73 (1–2) (2013)
827–842.
[10] N.A. Saeed, M. Eissa, Bifurcations of periodic motion of a horizontally supported nonlinear Jeffcott rotor system having transversely cracked shaft, Int.
J. Non Linear Mech. (101) (2018) 113–130.
[11] N.A. Saeed, M. Eissa, Bifurcation analysis of a transversely cracked nonlinear Jeffcott rotor system at different resonance cases, Int. J. Acoust. Vib. 24
(2) (2019) 284–302.
[12] T.H. Patel, M.J. Zuo, X. Zhao, Nonlinear lateral-torsional coupled motion of a rotor contacting a viscoelastically suspended stator, Nonlinear Dyn. 69
(2012) 325–339.
[13] A. Hu, L. Hou, L. Xiang, Dynamic simulation and experimental study of an asymmetric double-disk rotor-bearing system with rub-impact and oil-film
instability, Nonlinear Dyn. 84 (2016) 641–659.
[14] C. Wang, D. Zhang, Y. Ma, Z. Liang, J. Hong, Theoretical and experimental investigation on the sudden unbalance and rub-impact in rotor system
caused by blade off, Mech. Syst. Signal Process. 76-77 (2016) 111–135.
[15] Y. Yang, Y. Xu, Y. Yang, D. Cao, Dynamics characteristics of a rotor-casing system subjected to axial load and radial rub, Int. J. Non Linear Mech. 99
(2018) 59–68.
[16] Y. Yang, Y. Yang, D. Cao, G. Chen, Y. Jin, Response evaluation of imbalance-rub-pedestal looseness coupling fault on a geometrically nonlinear rotor
system, Mech. Syst. Signal Process. 118 (2019) 423–442.
[17] J. Hong, P. Yu, D. Zhang, Y. Ma, Nonlinear dynamic analysis using the complex nonlinear modes for a rotor system with an additional constraint due
to rub-impact, Mech. Syst. Signal Process. 116 (2019) 443–461.
[18] H. Man, C. Shi, Q. Han, B. Wen, Fixed-point rubbing fault characteristic analysis of a rotor system based on contact theory, Mech. Syst. Signal Process.
38 (2013) 137–153.
[19] H. Man, Z. Wu, X. Tai, B. Wen, Dynamic characteristics analysis of a rotor system with two types of limiters, Int. J. Mech. Sci. 88 (2014) 192–201.

809
N.A. Saeed, E.M. Awwad, M.A. EL-meligy et al. Applied Mathematical Modelling 93 (2021) 792–810

[20] H. Ma, Q. Zhao, X. Zhao, Q. Han, B. Wen, Dynamic characteristics analysis of a rotor–stator system under different rubbing forms, Appl. Math. Model.
39 (2015) 2392–2408.
[21] X. Tai, H. Ma, F. Liu, Y. Liu, B. Wen, Stability and steady-state response analysis of a single rub-impact rotor system, Arch. Appl. Mech. 85 (2015)
133–148.
[22] B. Li, H. Ma, X. Yu, J. Zeng, X. Guo, B. Wen, Nonlinear vibration and dynamic stability analysis of rotor-blade system with nonlinear supports, Arch.
Appl. Mech. 89 (2019) 1375–1402.
[23] L. Hou, Y. Chen, Q. Cao, Nonlinear vibration phenomenon of an aircraft rub-impact rotor system due to hovering flight, Commun. Nonlinear Sci. Numer.
Simul. 19 (2014) 286–297.
[24] S. Bab, M. Najafi, J.F. Sola, A. Abbasi, Annihilation of non-stationary vibration of a gas turbine rotor system under rub-impact effect using a nonlinear
absorber, Mech. Mach. Theory 139 (2019) 379–406.
[25] G.G. Tehrani, M. Dardel, Vibration mitigation of a flexible bladed rotor dynamic system with passive dynamic absorbers, Commun. Nonlinear Sci.
Numer. Simul. (69) (2019) 1–30.
[26] G.G. Tehrani, M. Dardel, M.H. Pashaei, Passive vibration absorbers for vibration reduction in the multi-bladed rotor with rotor and stator contact, Acta
Mech. (231) (2020) 597–623.
[27] Y. Ishida, T. Inoue, Vibration Suppression of nonlinear rotor systems using a dynamic damper, J. Vib. Control 13 (8) (2007) 1127–1143.
[28] N.A. Saeed, M. Kamel, Nonlinear PD-controller to suppress the nonlinear oscillations of horizontally supported Jeffcott-rotor system, Int. J. Non Linear
Mech. 87 (2016) 109–124.
[29] N.A. Saeed, W.A. El-Ganaini, Time-delayed control to suppress the nonlinear vibrations of a horizontally suspended Jeffcott-rotor system, Appl. Math.
Model. 44 (2017) 523–539.
[30] N.A. Saeed, E.M. Awwad, M.A. El-Meligy, E. Abouel-Nasr, Sensitivity analysis and vibration control of asymmetric nonlinear rotating shaft system
utilizing 4-pole AMBs as an actuator, Eur. J. Mech. A Solids 86 (2021) 104145, doi:10.1016/j.euromechsol.2020.104145.
[31] Y. Ishida, T. Yamamoto, Linear and Nonlinear Rotordynamics, Wiley-VCH, 2012.
[32] H. Bleuler, M. Cole, P. Keogh, R. Larsonneur, E. Maslen, R. Nordmann, Y. Okada, G. Schweitzer, A. Traxler, Magnetic Bearings Theory, Design, and
Application to Rotating Machinery, Springer, Berlin Heidelberg, 2009.
[33] A. Nayfeh, D. Mook, Nonlinear Oscillations, Wiley, New York, 1995.
[34] N.A. Saeed, M. Kamel, Active magnetic bearing-based tuned controller to suppress lateral vibrations of a nonlinear Jeffcott rotor system, Nonlinear
Dyn. 90 (1) (2017) 457–478.

810