Nonlinear Dynamics 14: 57–87, 1997.

c 1997 Kluwer Academic Publishers. Printed in the Netherlands.

Nonlinear Dynamics of a Rigid Unbalanced Rotor in Journal

Bearings. Part I: Theoretical Analysis

G. ADILETTA, A. R. GUIDO, and C. ROSSI

Dipartimento di Ingegneria Meccanica per l’Energetica, Università degli Studi “Federico II”, via Claudio 21,
80125 Napoli, Italy

(Received: 25 April 1996; accepted: 17 March 1997)

Abstract. The dynamic behaviour of a rigid rotor supported on plain journal bearings was studied, focusing
particular attention on its nonlinear aspects. Under the hypothesis that the motion of the rotor mass center is
plane, the rotor has five Lagrangian co-ordinates which are represented by the co-ordinates of the mass center and
the three angular co-ordinates needed to express the rotor’s rotation with respect to its center of mass. In such
conditions, the system is characterised not only by the nonlinearity of the bearings but also by the nonlinearity due
to the trigonometric functions of the three assigned angular co-ordinates. However, if two angular co-ordinates
have values that are generally quite small because of the small radial clearances in the bearings, the system is de
facto linear in these angular co-ordinates. Moreover, if the third angular co-ordinate is assumed to be cyclic [18],
the number of degrees of freedom in the system is reduced to four and nonlinearity depends solely on the presence
of the journal bearings, whose reactions were predicted with the  -film, short bearing model. After writing the
equations of motion in this way and determining a numerical routine for a Runge–Kutta integration the most
significant aspects of the dynamics of a symmetrical rotor were studied, in the presence of either pure static or pure
couple unbalance and also when both types of unbalance were present. Two categories of rotors, whose motion is
prevailingly a cylindrical whirl or a conical whirl, were put under investigation.
Key words: Rotor dynamics, rigid rotors, chaos, conical whirl, journal bearings.

Nomenclature
ci ; c = radial clearance of i-bearing; radial clearance of bearings for the symmetrical system
C = center of the rotor
Ci = center of the journal of the bearing i, along the geometrical axis of the rotor (i = 1; 2)
d = distance of the two planes that are perpendicular to the  -axis and contain m1 and m2
Di ; D = diameter of i-bearing; diameter of bearings for the symmetrical system
e = dimensional unbalance of the rotor
fi;K = Fi;K =(i Pi ), (i = 1; 2; K = X; Y ): nondimensional components of the fluid film force,
bearing i, along the K -axis
Fi;K = (i = 1; 2; K = X; Y ): dimensional component of the fluid film force, bearing i, along the
K -axis (N)
F ; Fi = resultant force due to the unbalancing masses; force due to the mass mi
g = acceleration of gravity
G = center of mass of the rotor
Iii ; Iij = moment of inertia; product of inertia of the rotor, axes i, j
[I ] = inertia matrix of the rotor, with respect to the frame Gxyz
[I] = inertia matrix of the rotor, with respect to C
li = distance between C1 and C
`1; `2 = l1 ; l2
Li ; L = axial length of bearing i; length of bearings for the symmetrical system
LR = axial length of the rotor
L1 , L2 ; L = different assigned distances between supports; generic distance between supports
m = (! c=g ): nondimensional mass of rotor
2

M = mass of rotor (kg)

m1, m2; mU = unbalancing masses; common value (m1 = m2 = mU )
58 G. Adiletta et al.

fg = fMx My Mz gT : dimensional moment vector of the active and constraining forces about the
mass center, in the Gxyz frame
fMg = fMx My Mz gT : corresponding expression of fM g in the Gxyz frame
fMg = fMx My Mz gT = fMg=(1 P1 c1 ): dimensionless expression of fMg
M; Mi = resultant torque due to unbalancing masses; torque due to the mass mi
n = angular speed of rotor (in r.p.m. = 60!=2 )
P = weight of rotor
Pi = P [1 li =(l1 + l2 )]: part of the weight P supported by the i-bearing
r = radius of the rotor
rU = distance of unbalancing masses from the rotor axis
Ri ; R = radius of i-bearing; radius of bearings for the symmetrical system
[R] = matrix of rotation from Gxyz to Gxyz frame (see Appendix I.A)
t = time
T0 = 2=! : period of driving rotation
u, v = X=c1 , Y=c1 : dimensionless co-ordinates of point C
X, Y , Z = co-ordinates of point C
= 2 P2 =1 P1
= angle given in the plane  in order to fix the positions of m1 and m2 (Figure II.1)
 = Ixy =Izz : dimensionless couple unbalance
" = dimensionless eccentricity (ratio: dimensional ecc./c) of the journal with respect to its bearing
"s = value of " for the symmetrical system, static load and steady-state conditions
 = c2 =c1
i = angular co-ordinate (i = 1; 2; 3)
 = absolute viscosity of the lubricant in both bearings
 = frequency component
0 = !=2: frequency of the driving rotation
 = e=c1 : dimensionless value of unbalance
xx, zz = radii of inertia of the rotor with respect to the x- and z -axis
i ;  = (!=Pi )Ri Li (Li =Di ) (Ri =ci ) , (i = 1; 2): modified Sommerfeld number relative to the
2 2

i-bearing; modified Sommerfeld number for the symmetrical system

 = !t: dimensionless time
! = angular speed of rotor (rad/s)
!i = i-component of the angular speed vector in the Gxyz frame (i = x; y; z)
f!g _ = see Appendix I.B
f!g i
= vector of the angular speed components in the Gxyz frame
f!g 0 = vector with dimensionless components of the angular speed in the Gxyz frame
f!_ g; f! g = f g
derivative of ! ; dimensionless form of !_ f g
[! ] = rotational matrix of the frame Gxyz , given in the same frame
(c! =g )0 value of m = c! 2 =g at threshold of instability
2
=
_,  = derivatives with respect to time
0 , 00 = derivatives with respect to 

1. Introduction

The dynamic behaviour of a machinery rotor is conditional on the nonlinearities that affect the
rotor-supports system. These are mainly represented by the response of lubricated bearings
and squeeze-film dampers, the elastic restoring force acting upon the rotor and the dynamic
actions that arise because of the presence of clearances in the bearings.
The most remarkable effect of the above nonlinearities lies in the fact that as rotor motion
is determined by the unbalance of the rotor itself, it is generally not synchronous with rotation,
which turns out to be periodic – with a period that differs from that of the driving rotation –
nonperiodic or chaotic in given conditions.
In these conditions, the monitoring signals detected at the machine bearings will themselves
be periodic, but not synchronous with the driving rotation, nonperiodic or even chaotic,
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 59

respectively. If the nonlinear nature of the system is ignored, these signals can lead to incorrect
interpretations about their source.
A review that reports a number of investigations regarding the nonlinear effects mentioned
above is given in [15]. Some of the papers referred to deal with simple systems like Jeffcott
rotors. For a rotor of this kind, some examples of nonsynchronous periodic motions that arise
because of nonlinear elastic restoring forces are reported in [17]. In [14] a theoretical study
of the stability and chaotic motions was carried out.
Sub-harmonic and super-harmonic motions that originate because of the strong nonlin-
earity due to radial clearances and rub in the bearings are studied in [3, 5]; under the same
hypotheses, paper [9] reports examples of chaotic behaviour, obtained both theoretically and
in an experimental way.
An investigation of nonlinear effects regarding rotor inclination due to small clearances in
the roller-bearings of the supports is given in [3, 6, 11].
The role of internal damping in the dynamics of rotors that are studied as continuous
systems was examined in [2, 7, 10, 13].
A theoretical analysis of the chaotic behaviour that originates with the presence of fluid
film forces in lubricated bearings that support an unbalanced rigid rotor was made in [12].
Paper [1] investigated the effects of residual unbalance which was not restricted to cylindrical
motions. The theoretical model of the rotor was characterised by means of five independent
co-ordinates, represented by the co-ordinates of the rotor mass center (hypothesis of plane
orbits) and the three Euler angles.
For such rotors, a previous study of the present authors [16], showed theoretical and
experimental results for cylindrical motions and pointed out the possibility of chaotic motions
stemming from the nonlinear response of the bearings. The rotor conditions taken under
examination were set as simple as possible: in particular the rotor was rigid, symmetrical and
such that no conical instability could arise in the presence of static unbalance, thus making
the motion of the rotor a cylindrical whirl in all cases.
In the present paper, the dynamic problem has been set in a more general way and the only
hypothesis from the problem dealt with in [16] that has been kept is rotor rigidity. This was
done in order to allow a relatively straightforward analysis.
In this case, rotor motion is given by the motion of the rotor mass center, which can be
assumed to be plane, and a rotation around this point. Neither of the motions is independent
since the motion of the mass center is determined by forces (the bearing reactions) that are
also functions of the rotations around the mass center.
The system is thus characterised by five degrees of freedom and the co-ordinates that can
express the position of the rotor at any one moment include the pair of mass center co-ordinates
and three suitable angular co-ordinates.
Besides the nonlinearity due to the bearings, the system would be affected by a nonlinearity
that stems from the trigonometric functions of the fixed angular co-ordinates.
Owing to the circumstance that two angular co-ordinates have small values in any case,
because of the small radial clearances in the bearings, it can be assumed that the equations of
motion are linear in these two angular co-ordinates. Moreover, the third angular co-ordinate
is assumed to be cyclic [18], thus restricting the number of degrees of freedom in the system
to four. Hence, under these more general conditions, nonlinearity is entirely represented by
the lubricated bearing response.
After writing the equations of motion according to the previous remarks, a numerical
integration has been carried out, by means of a Runge–Kutta routine.
60 G. Adiletta et al.

With the aim of carrying out an investigation that would highlight the basic aspects of
nonlinear motions of a rigid rotor, two choices have been made: the first consisted of setting
rotor operating conditions that could show up the system’s nonlinearity. The second choice
involved addressing the investigation to two categories of rotors whose motion is prevailingly
a cylindrical whirl or a conical whirl.
With reference to the first indication given above, some suitable values for the quantities
=
 and m ( c!2 =g) have been fixed on the basis of the results given in [16]; as regards
the second search criterion, two numerical values have been assigned, in turn, to the conical
=
parameter  2xx =L: the first less than 1 (0.875) and the second higher than 1 (1.200). The
 parameter represents the ratio given by twice the rotor’s transverse radius of inertia xx to
the distance L between the supports.
In order to set the conical instability threshold in a range with low  -values, it was decided
to give the ratio xx =zz a value suitably greater than 1 (4.26).
Thus, with a set of values given to the triad  , c! 2 =g and xx =zz , and with the values
0.875 and 1.200 fixed in turn for , the instability threshold was sought: in particular, the fact
that cylindrical instability corresponds to the smaller value of  and conical instability to the
higher value, has been checked.
A broad numerical analysis was carried out in order to infer information from the bifurcation
diagrams u ()= ( ) ()f  (u  is a dimensionless co-ordinate of the projection of the mass
center of the rotor onto its axis) obtained with  varying in a suitable interval. In this regard
48 diagrams were collected through numerical integration, fixing in turn:
six unbalance conditions characterised by pure static or couple unbalance. With this aim
three values were selected in turn for the static unbalance  (0.17, 0.25, 0.31) and three
other values for the couple unbalance  ( 2:81
10 3, 3:75
10 3, 4:68
10 3).
the two values for the conical parameter ,
four sets of initial conditions. The first set represents initial displacements and velocities
that are practically null. Two other sets are respectively congruent with cylindrical or
conical motions. The fourth set refers to more generic conditions with both cylindrical
and conical components.
()
The investigation was supported by the harmonic analysis of the u  signal and the
examination of the orbits covered by three assigned points along the rotor axis. One of these
points has been fixed coinciding with the projection of the mass center of the rotor onto its
axis. The other two points are taken on the rotor axis, coinciding with good approximation
with the intersections of the same axis with the bearing middle planes. In particular, this means
that the orbits of the two points are assumed to be plane. In this regard, the hypothesis that the
orbits of all three assigned points are plane, has been justified through analysis.
The results of the scheduled analysis will be discussed in the follow, even though the only
=
data referring to intermediate unbalance conditions ( 0:25 or  = 3:75
10 3) and the
first set of initial conditions have been reported in the paper.

2. The Equations of Motion

A rigid rotor with residual unbalance, fitted with two lubricated bearings is assigned (Figure 1).
Once the projections O and C (Figure 2) of the rotor’s mass center G onto the Z -axis,
common to both bearings, and onto the axis  of rotor have been respectively fixed, three
reference frames can be assigned:
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 61

Figure 1. Drawing of the rotor with its supports.

Figure 2. The co-ordinate systems: OXY Z , fixed frame; Gxyz, nonrotating moving frame; Gxyz , moving frame
fixed to the rotor.

the fixed frame OXY Z , with the Y -axis vertical and the Z axis coinciding with the axis
of bearings;
the nonrotating moving frame Gxyz, which undergoes translation motions while remain-
ing parallel to OXY Z ;
the moving frame Gxyz that is fixed to the rotor.
Of the infinite choices that could be made in order to assign the Gxyz frame, the one in
which the x-axis contains the CG segment and is perpendicular to the plane of unbalancing
torque (Appendix I) was fixed.
In Figure 2, C , C1 and C2 represent the triad of points taken along the axis  of rotor, as
said above, so that their orbits give a further insight into rotor motion.
Because of the clearances in the two bearings, under the hypothesis of plane motion of the
mass center, the rotor exhibits five d.o.f. The XG and YG mass center co-ordinates and the
three angular rotations 1 , 2 , 3 were set as independent co-ordinates. This triad of angles
was set from the 12 available triads so as to represent successive rotations that give the actual
position of the frame Gxyz , beginning from the position of the frame Gxyz.
In general, the angles 1 and 2 are very small because of the small radial clearances in the
bearings, and the equations of motion are linear in 1 and 2 : the two rotations 1 and 2 thus
lose their role of angular co-ordinates and can be assumed to be vectors, whose resultant does
not depend on the order in which rotations follow one another.
62 G. Adiletta et al.

Figure 3. Arrangement of the rotor in the reference frame Gxyz .

In the fixed frame OXY Z the equations of motion can be written:

(
M X G + F1;X + F2;X = 0
M YG + F1;Y + F2;Y P = 0
(1)

with XG , YG : co-ordinates of the mass center of the rotor; Fi;K : component of the reaction of
the i-bearing, along the K -axis (i = 1; 2; K = X; Y ); M , P : mass and weight of the rotor.
Ignoring the effect of misalignment between the bearing axis and the journal axis, the Fi;K
components of the reactions of bearings 1 and 2 (Figure 1) can be expressed as functions of
the co-ordinates of the points C1 and C2 .
In the reference frame Gxyz , the co-ordinates of the points C , C1 and C2 are fxg, fxg1
and fxg2 , respectively, and have the expressions (Figure 3):
8x9 8x 9 8 9 8x 9 8 9
< = >
< 1> = < e= >
< 2> = < e=
fxg = : y ; ; fxg1 = > y1 > = : 0 ; ; fxg2 = > y2 > = : 0 ; ; (2)
z :z ; l11
:z ; l2 2

where e indicates the eccentricity of G.

In the frame Gxyz, the co-ordinates of C , C1 and C2 can be obtained from (2) by means of
[ ] [ ]
the matrix R 1 , inverse of the matrix R (Appendix I), relative to the transformation from
[ ] =[ ]
Gxyz to Gxyz, provided that R 1 R T :
8x9 8 e9 8 > ec3
9
>
< = < = < =
fxg = : y ; = [R] fxg = [R] : 0 ; = > e(12c3 + s3) >
T T
(3)
z 0 : e( 2c3 + 1s3) ;
8x 9 8 e9
>
< i> = < =
fxgi = > yi > = [R]T fxgi = [R]T : 0 ;
: zi ; `i
8 ec3 + 2`i
9
>
< >
=
= > e(12c3 + s3) 1`i > ; i = 1; 2: (4)
: e( 2c3 + 1s3) + `i ;
In the relationship (4) `1 ; = l1 , and `2 = l2 have been set; cos 3 and sin 3 have been
concisely indicated as c3 and s3 , respectively.
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 63

In the fixed reference OXY Z , the points C , C1 , and C2 have co-ordinates:

fX g = fX gG + fxg; (5)
fX gi = fX gG + fxgj ; j = 1; 2: (6)
From (5) the co-ordinates of the mass center in OXY Z can be inferred:
8 e9
< =
fX gG = fX g fxg = fX g [R]T fxg = fX g [R]T : 0 ;
0
8 X + ec3
9
>
< >
=
= > Y + e(1 2c3 + s3 > : (7)
: Z + e(2 c3 1s3) ;
In particular, assuming the absence of a resultant force along the Z direction, the following
relation can be written:
ZG = Z + e(2 c3 1s3) = 0;
from which:
Z = e(2c3 1s3):
The assumption of plane orbits of C , C1 and C2 , given above, can be held assuming that
the term 2 c3 1 s3 is negligible due to the smallness of 1 and 2 , and thus Z 0. =
The acceleration of the mass center can be expressed by differentiating twice Equation (7)
with respect to time:
8 9
>
< X e_32c3 >
=
fX gG = > Y e[(1 2_32 __
21 2 12 12)c3 + (_32 + 2_12_3 + 21_2_3)s3] > :
: 0
>
;
Relation (1) can then be rewritten as a function of the co-ordinates X , Y of C and of the
angular co-ordinates 1 , 2 , 3 :
8  _2
< M X = F1;X + F2;X + Me3 c3 2
>
> M Y = F1;Y + F2;Y + Me[(1 2_3 2_1_2 12 12)c3 (1bis)
: + (_2 + 2_12_3 + 21_2_3)s3] P:
3
_ _  
In good approximation, neglecting those terms that depend on 1 , 2 , 1 , 2 , 1 and 2 , the
equations can be reduced to:
( M X = F + F + Me_2c
1;X 2;X 3 3
_
M Y = F1;Y + F2;Y + Me3 s3 P:
2
(1ter)

The co-ordinates fX gi of C1 and C2 can then be written as a function of the ones given in
point C , making use of relation (6) and expressions (4) and (7):
809 8 > X +  `
9
>
< = < 2 i
=
fX gi = fX gG + fxgi = fX g + [R] : 0 ; = > Y 1`i > ; i = 1; 2:
T
(8)
`i : Z + `i ;
64 G. Adiletta et al.

The fluid film force components Fi;K in bearings 1 and 2 (Figure 1) can be written by
means of relation (9) and their time derivatives.
As regards the motion of rotor rotation around its mass center G, the equations of motion
can be written, in the reference frame Gxyz that is fixed to the rotor itself:
fM g = [I ]f!_ g + [!][I ]f!g (9)
[]
after assuming in the same reference: fM g: vector of the moments of the external forces; I :
[]
matrix of inertia for the rotor; ! : rotational matrix of the frame Gxyz given in the same
frame; and f! g: vector of the components of the angular speed.
8! 9
Because of
8 _ 9
>
< x> = < 1>
> =
f!g = > !y > = [Rv ] > _2 > ; (10)
: !z ; : _ ;
3
it is possible to write
0 8> _ 9 1
< 1>=
f!_ g = ddt B
@[Rv ] > _ CA = [Rv ]fg + [R_ v ]f_g:
: _2 >
(11)
;
3
Some of the matrices concisely written above are explained in Appendix II.
[ ]
Premultiplying relation (9) with the matrix R T , the equations of motion relative to rotor
rotation are obtained in the Gxyz co-ordinate system
fMg = [R]T fM g = [R]T ([I ]f!_ g + [!][I ]f!g): (12)
The moment fMg of the active and constraining forces about the mass center is given by
the sum of the moments fMgi , i = 1; 2, which are due to fluid film reactions in the bearings,
and of the moment fMga of the external active forces
fMg = fMg1 + fMg2 + fMga : (13)
In particular, having respectively indicated with the following symbols:
F1  (F1;X ; F1;Y ; F1;Z )  (F1;x ; F1;y ; F1;z )
F2  (F2;X ; F2;Y ; F2;Z )  (F2;x ; F2;y ; F2;z ) (14)
the fluid film reactions, acting upon C1 and C2 , and their components in the frames GXY Z
and Gxyz, and observing that
fMgi = (Ci G) ^ Fj ;
(Ci G) = (xi; yi ; zi); i = 1; 2;
Fi;X = Fi;x ; Fi;Y = Fi;y; Fi;Z = Fi;z ;
8M 9 8
it can be deduced:
9
>
< x> = > < z1 F1;Y >
=
fMg1 = > My > = > z1F1;X
: Mz ; : x1F1;Y y1F1;X >
;
;
8 M 91 8 9
>
< x> = > < z2 F2;Y >
=
fMg2 = > My > = > z1F2;X :
: Mz ; : x2F2;Y y2F2;X >
(15)
;
2
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 65

Assuming further
8 0 9
< =
fMga = : 0 ; ;
Ma;z
relation (13) can be rewritten, making use of the above expressions:
fMg = fMg1 + fMg2 + fMga
8 (z1F1;Y + z2F2;Y ) 9 8 9
>
< >
= < 0 =
=> zF + z2F2;X > + : 0 ;: (16)
: x1F1;Y + x21F12;X
;Y y1 F1;X y2 F2;X
; Ma;z

Equations (1ter) and (12) represent the rotor’s five equations of motion in the independent
co-ordinates X , Y , 1 , 2 , 3 .
In order to give the equations a dimensionless form, it can be initially assumed that
u = X=c1 ; v = Y=c1 ; w = Z=c1 ;
ui = xi =c1 ; v1 = yi =c1 ; wi = zi =c1 ; i = 1; 2;
 = !t; (17)
where c1 indicates the radial clearance in bearing 1.
On the basis of the last relation, time derivatives can be put in a dimensionless expression:
u_ = !u0; v_ = !v0 ; : : : u = !2u00 ; v = !2v00 ; : : :
u_ i = ! u0i ; v_ i = ! v0i ; : : : ui = ! 2 u00i ; vi = ! 2 v00i ; : : :
_i = !i0 ; i = !2i00: (18)
The reactions Fi;K in the bearings can be written in dimensionless form as follows:

fi;K = Fi;K ; i = 1; 2; K = X; Y;
i Pi
(19)

where
!R  Ri 2  Li 2
i = P i Li
ci Di ; i = 1; 2 (20)
i

is the modified Sommerfeld number relative to the operating conditions of the i-bearing
=
(i 1; 2); in the expression of i the following symbols have been adopted:
 : absolute viscosity of the lubricant in both bearings;
Di ; Ri ; Li : diameter, radius and axial length of the i-bearing;
Pi : P [1 li =(l1 + l2 )]: part of the weight P supported by the i-bearing.
Making use of (17), the vector of the moments can be expressed as
8M 9 8M 9
>
< x> = >
< x> =
> =  1 P1 c1
: Mz > >
: Mz >
My My
; ;
66 G. Adiletta et al.
8 (w1 f1;Y + w2 f2;Y ) 9
>
> >
>
< w1 f1;X +  w2 f2;X
=
= 1 P1c1 > > (21)
>
: u1f1;Y + u2 f2;Y v1f1;X v2 f2;X + Ma;z >
;
Pc 1 1 1

with

= 21PP21 ;  = cc21 : (22)

_
With the further relations that can be given for the vectors f! g and f! g, and the matrices
[!] and [R_ v ] (see Appendix I):
f!g = !f!g; [!] = ![!]; f!_ g = !2f!0 g; [R_ v ] = ![Rv0 ]; (23)

the system of five equations of motion can be written in dimensionless form:

8 1P1 (f + f ) + e c
>
> u00 = Mc
>
> 1!
2 1;X 2;x
c1 3
>
< v00 = 1P1 2 (f1;Y + f2;Y ) + e s3 g 2
Mc1 ! c1 c1 ! 8 9
> I  >
(24)
>
> M 2 !2  I  < Mx >
=
> z
[ R ] T
f! 0 g + [! ] f ! g = > My :
>
: 1 P1 c1 Izz Izz : Mz >
;
The third relation in (24) can be rewritten as follows:
0 8M 9 1
 I  1 B 1 P1c1 > < x>
= I C
f!0g = I @ M2 !2 [R] > [!] I f!gA
: y>
M
zz z ; Mz
zz

and, furthermore, on the basis of (11), (18) and (23):

0 8M 9 1
 I  1 B 1P1 c1 >
< x>= I C
[Rv ]f00g + [Rv0 ]f0g = Izz @ M2 !2 [R] > My > [!] Izz f!gA ; (25)
z : ; Mz

it can finally be written as

8 00 9 0 8M 9 1
>
< 100 > =  I  1 B 1 P1c1 > < x> = I C
>  = [Rv ] 1 I @ M2 !2 [R] > [!] I f!gA
: 2300 > : Myz >
M
; zz z ; zz

8 0 9
>
< 10 >=
[Rv ] [Rv ] > 2 > :
1 0
(26)
: 30 ;
The first two equations in (24) and (26) represent a second order system of five ordinary
differential equations in the independent co-ordinates u, v , 1 , 2 , 3 .
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 67

In order to carry out a numerical integration of the system with a Runge–Kutta method,
the same system has been put in the equivalent form given by a first order system of ten
differential equations. With this aim, the following relations have been adopted:
1 = u; 2 = v; 5 = 1; 6 = 2; 7 = 3;
3 = u0; 4 = v0 ; 8 = 10 ; 9 = 20 ; 10 = 30 :
The ten equations of motion are
8 0 = 
>
> 3
>
1

> 2 = 4
0
>
> 1 P1 (f + f ) + e cos 
>
>  3 =
0
>
> Mc1 !2 1;X 2;X
c1 7

>
> 40 = Mc 1 P1 (f + f ) + e sin  g
< 1!
2 1;Y 2;Y
c1 7
c1 !2
>
> 5 = 8
0 (27)
>
> 60 = 9
>
> 70 = 10
>
>
> 80 = A1
>
>
: 90 ==AA23 ;
0
>
10
where A1 , A2 , and A3 have the following expressions:
8A 9 0 8M 9 1
>
< 1> =  I  1 B 1 P1c1 > < x>
= I C
> A = [Rv ] 1 I @ M2 !2 [R] > [!] I f!gA
: A23 > : Myz >
M
; zz z ; zz

8 9
>
< 8> =
[Rv ] 1[Rv0 ] > 9 > : (28)
: 10 ;
The matrix [Rv ] 1 in relations (26) and (28) is explained in Appendix I.
In particular, the assumption Mz = 0 can be made; thus from the third relation (16) it can
be inferred that
x1F1;Y + x2 F2;Y y1F1;X y2F2;X = Ma;Z : (29)
The parameters determining system dynamics are
 2 P2 c !  L1 2  R1 2 I  z 2
=  P ;  = c ; 1 = P R1L1 D
2
c1 ;  = Ixx = x ;
zz

ci !2  2xx 2 ;  = e ;  = Iyz ; = l1 :
1 1 1 1 1

2ci Ixx ! 2
Pi `2 = g `i c1 Izz l2 (30)

If the system is symmetrical, we have: =  = = 1, and the parameters of the problem

are restricted to the following:
2!
 L 2  R 2
 = P RL D m = c!  = 2Lxx ;
2

c ;
g ;
68 G. Adiletta et al.
 2
 = IIzz = zz ;  = ec ;  = IIyz : (31)
xx xx zz
The last two parameters refer to static and couple unbalance of the rotor, respectively.
With reference to the parameter  it must be stressed that if the frame Gxyz is set with its
x-axis perpendicular to the plane of the unbalancing torque, the only product of inertia that
differs from zero is Iyx (see Appendix II).
= =
In order to perform numerical integration of (27), it has been assumed that 3 !t  ,
with ! = const:, so as to have 30= 1 and 300= 0. Under this constraint, the co-ordinate
becomes a cyclic co-ordinate and thus negligible, hence reducing the d.o.f. of the dynamic
system from five to four. The system of equations (27) is thus replaced by
8 0 = 
>
> 3
>
1
 0 = 4
>
> 2
>
> 1 P1 (f + f ) + e cos 
3 =
>  0
>
< 0 Mc 1!
2 1;X 2;X
c1
 1 P1 e g
> 4 = Mc1 !2 (f1;Y + f2;Y ) + c1 sin c1 !2 (32)
>
> 50 = 8
>
>
> 60 = 9
>
>
: 80 == AA12
0
>
9
with 7 = 3 =  , 10 = 30 = 1.

3. Numerical Integration of the Equations of Motion

Even though the equations of motion (32) have a general aspect, the numerical investigation
has been restricted to symmetrical systems whose dynamic behaviour is determined by the
six parameters given in (31).
The  -film, short bearing model was adopted for the fluid film reactions fi;K . Their
expressions, regardless of misalignment between the bearing axis and the journal axis, have
=
been given in functions of the co-ordinates fX gi (i 1; 2) of the points C1 and C2 , as in [16,
Appendix].
From the analysis [16] already referred to above, it has been inferred that the nonlinear
behaviour of such a system becomes more marked as static eccentricity "s of the rotor increases
and that, under such conditions, for suitable values of the static unbalance , the motion of
the rotor can assume nonperiodic or even chaotic characteristics.
In Figure 4 the static eccentricity "s has been plotted as a function of  together with
=
the zone ABCD of the plane  , m including, with  0:3, points whose co-ordinates are
relative to chaotic conditions of system operation. The quantity m has been expressed as
c!2=g.
In the same Figure 4, the dashed line, indicating the cylindrical instability threshold,
has been analytically obtained with linear approximation regarding the reactions of the two
bearings [19].
It is worth mentioning that the zone ABCD in the above diagram has been fixed by
assigning two values of  and two straight lines with a 45 slope; these lines are the loci of
=
points corresponding to the values 40 and 180 of the parameter m c! 2 =g , respectively.
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 69

Figure 4. Region of the plane in which operating conditions of the rotor have been fixed.

Figure 5. The two distances L1 and L2 fixed for positioning the rotor supports.

After assigning inside this region a point H with co-ordinates

 = 0:1; c!2=g = 7:0; (33)
the parameter  = 2xx =L was given the two values, respectively, less and greater than 1:
1 = 0:857; 2 = 1:200: (34)
Furthermore, a numerical value considerably higher than 1 has been assigned to the para-
meter= xx=zz :
 = xx=zz = 4:25
in order to make the conical instability threshold set in with values of  that are quite small.
It is worth pointing out that, for a cylindrical rotor whose diameter coincides with the
diameters of its journals (see Figure 5), once the ratio xx =zz has been fixed, the ratio of the
length LR to the radius r of the rotor also remains determined and that, once the two values
(34) have been assigned, the ratios of the two distances L1 and L2 between the bearings to the
length LR of the rotor remain fixed.
70 G. Adiletta et al.

Under the assumption that only the distance between the bearings is modified, in accordance
with the two solutions depicted in Figure 5, while both the rotor and the bearings are kept the
same, the instability thresholds in these two cases have been determined. It is worth observing

that under such conditions the values of the parameters  , m and are the same in both cases
and that the values of  and  can also be repeated in the two arrangements.
The two instability thresholds have been determined through numerical integration of the
= =
equations of motion, assuming the rotor to be perfectly balanced (  0) and with  and
 given as above. The values of  and c! 2 =g have been fixed as co-ordinates of points on
()
the straight line (see Figure 4) passing through H and having arctg 2 as its slope. It must be
observed that this corresponds to assuming that the rotor varies its angular speed beyond or
beneath the speed value at P , leaving the values of all the other parameters unchanged, except
for  and c! 2 =g .
Figures 6a and b illustrate the perspective representations of the surfaces given by the suc-
cessive positions assumed by the rotor axis in conditions of instability. These representations
have been obtained by rotating the equidistant planes of the orbits of C1 , C and C2 around two
perpendicular axes. Then the points taken along the three orbits in each instance have been
joined with straight lines. A suitable number of lines referring to instances taken at regular
intervals of time have been collected in each representation.
When the instability is cylindrical, the geometrical locus of the successive positions
assumed by the rotor axis is a cylindrical surface (Figure 6a), but it is not a perfect cylinder.
In the case of conical instability the surface strongly appears to be conical (Figure 6b), even
if not exactly a cone: the point C on the rotor describes a small trajectory around a point
not coinciding with the origin O of the axes. These circumstances depend on the fact that
the rotor’s weight makes the system lose its axial symmetry which could, instead, determine
surfaces that are symmetrical to their geometrical axes.
One remarkable aspect is represented by the fact that under the conditions in Figure 6b,
the orbit of point C is practically circular and, unlike the trajectories of points C1 and C2 , it is
synchronous, its frequency being double that of the frequency characterising the motions of
C1 and C2.
In order to determine either the periodical or the nonperiodical conditions that might affect
rotor motion due to unbalance, a broad numerical analysis was scheduled, with reference to
the two categories of rotors that could be assigned giving , in turn, the values 1 and 2 .
Six unbalance conditions, characterised by pure static or couple unbalance, were fixed.
With this aim three values were selected in turn for the static unbalance  (0.18, 0.25, 0.31)
and three other values for the couple unbalance  ( 2:81
10 3 , 3:75
10 3 , 4:68
10 3 ).
Furthermore, the influence of the initial conditions onto the numerical integration was
taken into account by adopting, in turn, each of the four following sets:

Set 1:
u0 = 0:1
10 5; v0 = 0:1
10 5; u00 = 0; v00 = 0;
1;0 = 0; 2;0 = 0; 3;0 = 0; 10 ;0 = 0; 20 ;0 = 0; 30 ;0 = 1:0:
Set 2:
u0 = 0:5; v0 = 0:5; u00 = 0:5; v00 = 0:5;
1;0 = 0; 2;0 = 0; 3;0 = 0; 10 ;0 = 0; 20 ;0 = 0; 30 ;0 = 1:0:
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 71

Figure 6. Perspective representations of the surfaces given by the successive positions assumed by the rotor axis
in conditions of instability: (a)  = 0:857; (b)  = 1:200.

Set 3:

u0 = 0; v0 = 0; u00 = 0; v00 = 0;
1;0 = 0:5
10 3; 2;0 = 0:5
10 3; 3;0 = 0; 10 ;0 = 0:5
10 3
;

20 ;0 = 0:5
10 3; 30 ;0 = 1:0:

Set 4:

u0 = 0:5; v0 = 0:5; u00 = 0:5; v00 = 0:5;

72 G. Adiletta et al.

1;0 = 0:5
10 3
; 2;0 = 0:5
10 3
; 3;0 = 0; 10 ;0 = 0:5
10 3
;

20 ;0 = 0:5
10 3
; 30 ;0 = 1:0:

In the following the motion of the rotor has been said to be cylindrical when the rotor axis
= =
remains parallel to axis of bearings (1 2 0) and conical in the other conditions (i 6 0; =
=
i 1; 2). In this respect, the four sets of initial conditions can be characterised as follows:
Set 1: practically null initial conditions;
Set 2: initial conditions congruent to cylindrical motions;
Set 3: initial conditions congruent to conical motions;
Set 4: the initial conditions of sets 2 and 3 are combined.
According to the above indications, 48 bifurcation diagrams were obtained, each with one
of the two values of , one of the six unbalance conditions, and one of the four sets of initial
conditions.
In each diagram the abscissa gives the value of  that varies in the interval 0:0295  0:1476,
in steps of 2:95
10 3 . For each value of  , the diagram shows the corresponding values of
()
the dimensionless co-ordinate u  of point C , obtained through numerical integration of the
equations of motion and collected in steps of  = =
k
2, (k 1000  2000). It is worth
pointing out that in the course of the experimental analysis described in Part II of this paper,
=
the above said interval of  values corresponds to the interval n 2000  10000 rev/min of

=
the angular speed of the rotor, and a step n 200 rev/min corresponds to the step fixed for
.
For each of the above said 48 conditions, the orbits of the points C1 , C and C2 were plotted,
fixing further in turn four values of  , suitably assigned on the basis of the related bifurcation
diagram. It is worth to remark that, together with the trajectories in the co-ordinate plane u, v
of the points C1 , C and C2 , respectively, it was useful, particularly for nonperiodic conditions,
[ ( ) ( )] [ ( ) ( )] [ ( ) ( )]
to plot the loci of points u1  ; v1  , u  ; v  , and u2  ; v2  , with  k
2 and k =
taken in a suitable range. In the following, these loci will be indicated concisely as attractors.
[ ( ) ( )] [ ( ) ( )]
In particular, the points u  ; v   1  ; 2  , represent a projection of the Poincaré
section for the given system of independent co-ordinates.
From the data collected through the numerical investigation it can be inferred, among other
things, that:
1. the nonlinear character that affects the system response manifests whatever the unbalance
condition assigned to the rotor, though in a degree that differs from case to case; and
2. the higher the unbalance, the lower is the influence of the initial conditions onto the
motion.
This influence is further reduced when the symmetrical or the antisymmetrical character of the
lowest resonant motion of the rotor-supports system turns out to be favoured by the symmetry
or the antisymmetry of the unbalancing actions.
From the above remarks and with further account of the fact that along experimental trials
it turns out to be practically impossible to fix any set of initial conditions, it follows that it is
suitable to put in comparison only theoretical and experimental data that refer to intermediate
value of the unbalance. Such values of the unbalancing forces, on the one hand, make it
possible to bound the influence of the initial conditions and, on the other hand, are not too
high so as to lose any correspondence with practical cases.
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 73

Figure 7. Bifurcation diagrams with  = 0:857: (a)  = 0,  =

3:75 10 3
; (b)  = 0:31,  = 0.

4. The Results of the Theoretical Investigation

Taking into account the indications given in the above section, the results illustrated in the
following as an example, refer to the intermediate values of the unbalance (pure static, with
=
 0:25, or pure couple, with  = 3:75
10 3) and the set 1 of initial conditions, unless
different values or sets are expressly indicated.
Having said that, the results of the theoretical analysis that was carried out as illustrated in
Section 3, can be concisely reported as follows:
(A) The bifurcation diagrams exhibit considerable similarities from one to another, through
all the conditions that were put under investigation. In fact, when  is given values in the lowest
zone of the fixed interval, all the plots confirm the existence of periodic motions of the rotor,
generally synchronous with rotation. A successive increase of  values determines, whatever
the case, a bifurcation that is followed by nonperiodic behaviour.
With further increase of  , some differences arise from diagram to diagram, depending on
the conditions relative to each case, which make the motion still nonperiodic or periodic again
and synchronous in particular.
As an example of the above said similarities and differences two bifurcation diagrams are
shown in Figure 7a (for  = 1;  0; = = 3:75
10 3 ) and in Figure 7b (for  1 ; =
=
 0:31;  0).=
(B) For low  values, the projection of the Poincaré section is represented in the diagrams
by a single point, making it possible to infer the periodicity of rotor motion. Nevertheless, it
is not sufficient in order to recognize whether the solution has the same period of the driving
rotation, or a sub-multiple of it. This uncertainty has been solved with frequency analysis of
()
u.
()
Figures 8a and b show the RMS amplitude spectra of u  obtained with the parameters
set as in Figures 7a and b, respectively, when  0:044.=
74 G. Adiletta et al.

Figure 8. Frequency analysis (RMS Amplitude Spectrum) of u( ), with  = 0:857 and  = 0:044: (a)  = 0,


 = 3:75 10 3 ; (b)  = 0:25,  = 0.

Figure 9. Orbits of the points C1 , C and C2 obtained with  = 0:857,  = 0:25,  = 0 and  = 0:044.

()
From Figure 8a it can be inferred that u  is harmonic, with a frequency which is twice
()
that of the fundamental, while Figure 8b shows that the u  is synchronous with the driving
rotation, even though a frequency component of order two appears in the spectrum. Under
these last conditions Figure 9 illustrates the trajectories of points C1 , C and C2 , confirming the
synchronous and cylindrical (the orbits of C1 , C and C2 are equal) character of the whirling
motion of the rotor.
This dynamic behaviour is different from that of Figure 8a. An insight into the rotor
motion under the conditions of Figure 8a, is given by Figures 10–12. The frequency spectra
() ()
of u1  and u2  , illustrated in Figure 10, show that both functions have a period equal to
T0. Figure 11 illustrates the orbits of points C1, C and C2 and makes it possible to infer that:
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 75

Figure 10. Frequency analysis (RMS Amplitude Spectrum), with  = 0:857,  = 0,  =

3:75 10 3
and
 = 0:044: (a) relative to u1 ( ); (b) relative to u2 ( ).

Figure 11. Orbits of the points C1 , C and C2 obtained with  = 0:857,  = 0,  =

3:75 10 3
, and  = 0:044.

the rotor motion is practically conical, having C as its vertex; and

the orbits of C1 and C2 are equal and synchronous with rotation, as it is confirmed by
the aspect of the respective attractors, represented in both cases by a single point, whose
position along the trajectories is different from C1 to C2 .
Figure 12a is a zoom and compares the trajectories of C and C1 (or C2 ). The direction
of motion is indicated onto the trajectories: C has a clockwise motion unlike the points C1
and C2 which move in the opposite direction along their respective orbits. Figure 12b shows
a further enlargement of the orbits of C .
This particular precession motion of the rotor, with the point C covering a complete
elliptical orbit in the same interval of time taken by C1 and C2 to describe one half of their
76 G. Adiletta et al.

Figure 12. Solution obtained with  = 0:857,  = 0,  = 3:75 10 3 , and  = 0:044 (see Figure 11): (a) orbits
of point C1 (counter-clockwise versus) and of point C (clockwise versus) put in comparison; (b) orbits of point C .

orbits, with a direction that is opposite to that of C , is not exceptional, but it is shared by all
the cases characterised by low values of  in the presence of pure couple unbalance, whatever
the  ratio.
Further, it is worth to remark that under all conditions characterised by low values of  ,
the investigation has shown that the rotor motion, besides the synchronous nature, exhibits an
aspect that is conditional on the type of unbalance, whatever the value of : cylindrical in the
presence of pure static unbalance and conical when the unbalance is only of couple type.
(C) Downstream the bifurcation, the rotor motion turns out to be quasi-periodic or chaotic.
As an illustration, Figures 13a and b show the attractors relative to quasi-periodic motion and
chaotic behaviour, respectively. The values of the parameters were set as follows:
=
Figure 13a:  1 ;  0;  = =
3:75
10 3,  0:077; =
= =
Figure 13b:  2 ;  0:25;  0;  0:075. = =
The quasi-periodic nature of the motion illustrated in Figure 13a can be inferred by the
closed-curve aspect of the attractors. On the other hand, the projection of the Poincaré section
in Figure 13b apparently shows a fractal structure, typical of strange attractors, which indicates
the chaotic character of the solution.
Besides, the attractors of Figure 13 were selected as examples because they make it possible
to point out some general aspects.
In fact, Figure 13a shows that the attractors in the planes of trajectories related to points
C1 and C2 are different from each other, in opposition to the antisymmetrical nature of the
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 77

Figure 13. Attractors relative to the points C1 , C and C2 ; (a)  = 0:857,  = 0,  =

3:75 10 3
, and  = 0:077;
(b)  = 1:200,  = 0:25,  = 0, and  = 0:075.

unbalancing system of forces and, as a consequence, of the whole system, respect to a plane
that is perpendicular to the rotor axis and contains the center C .
The difference between the two attractors can be justified by taking into account the initial
phase of the unbalancing torque. In fact, when the initial shift, unlike the case of Figure 13a,
is given a value that is different from zero, i.e. equal to  , the attractors turn out to be as
in Figure 14. A comparison between Figures 13a and 14 shows that the attractors relative to
points C1 and C2 are exchanged while the attractor of C remains the same. It is worth to
remark that, changing the initial phase from zero to  , makes the sign of Iyz (the only product
of inertia that is different from zero) change.
78 G. Adiletta et al.

Figure 14. Influence of the initial angular phase of the unbalancing torque on the attractors of the points C1 , C


and C2 with  = 0:857,  = 0,  = 3:75 10 3 , and  = 0:077 (compare to Figure 13a).

This behaviour turned out to be shared by all the conditions with  =

1 , pure couple
unbalance and values of  sufficiently high in order to give rise to nonlinear motions.
On the contrary, the motion depicted in Figure 13b is characterised by the highest value
of the conical parameter ( =
2), in the presence of pure static unbalance: under these
conditions, as far as the parameters ,  and  are concerned, the dynamic behaviour is
affected by a reciprocity character respect to the case of Figure 13a. The motion is conical,
as it was in Figure 13a: the three points C1 , C and C2 perform trajectories with motions that
differ in each instant from one point to each other, but here, unlike the case of Figure 13a, the
attractors relative to C1 and C2 are very likely to be equal to each other and both are different
from the attractor relative to the point C .
It is also worth to observe that, according to the above result, the shapes of the attractors
are not influenced by the initial phase of the unbalancing torque, unlike the situations depicted
in Figures 13a and 14.
(D) As it was observed at point (A), when  is given the highest values in the fixed interval,
the rotor motion can still manifest nonperiodic character or return to being periodic again. As
an example, Figure 15a illustrates the attractors corresponding to the orbits of C1 , C and C2 ,
=
for  1 ,  0, = = =
3:75
10 3 , and  0:089. Figure 15b shows the orbits of the same
= = =
points when  1 ,  0:31,  0, and  0:126. =
An examination of these figures makes it possible to infer (in accord to the corresponding
bifurcation diagrams in Figure 7) the nonperiodic nature of the motion characterised by the
closed-curve attractors of Figure 15a, and the synchronous, practically cylindrical character
of the solution depicted through the orbits of Figure 15b.
(E) With  = 1, whatever the assigned value of the static unbalance, the rotor motion
turned out to be cylindrical. In this regard, it must also be pointed out that, when  is given
values that are lower than those selected at the outset, initial conditions that are congruent to
conical motions can make the solution manifest a conical character. Figure 16 illustrates such
= =
a behaviour for the set 3 of the initial conditions and  1 ;  0:12;  0;  0:081. = =
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 79

Figure 15. (a) Attractors relative to the points C1 , C and C2 plotted for  = 0:857,  = 0,  = 3:75 10 3
, and
 = 0:089; (b) orbits of C1 , C and C2 obtained with  = 0:857,  = 0:31,  = 0, and  = 0:126.

The rotor motion is conical and the point C performs quasi-circular, synchronous orbits, while
points C1 and C2 cover their orbits with a period that is twice that of the driving rotation.
Besides, whatever the point, the motion has clockwise versus.
(F) When both static and couple unbalances were present, all the conditions put under
investigation were characterised by conical solutions for the rotor motion.
As an illustration, Figure 17 compares the trajectories of C1 , C and C2 obtained with
=
 0:17,  = 2:65
10 3,  =
0:089 and  =
1 (Figure 17a) with the corresponding
=
orbits that were plotted for  2 and keeping constant the values of all the other parameters
(Figure 17b).
80 G. Adiletta et al.

Figure 16. Orbits of the points C1 , C and C2 obtained with  = 0:857,  = 0:12,  = 0,  = 0:081, and the set 3
of initial conditions.

In both situations, the rotor motion appears to be chaotic and conical, and it is worth to
remark further the influence of the conical parameter that makes the attractors differ from one
case to the other.

5. Conclusions
From the results of the theoretical investigation given in the above section, the following
remarks can be stressed:
1. The operation of rigid rotors supported on lubricated journal bearings can manifest
conditions that are justifiable only if the nonlinearity of the system is taken into account.
2. The nonlinear elements of the system are mainly represented by the lubricated bearings;
consequently, the nonlinear behaviour of the rotor is emphasized with high values of operating
static eccentricity "s .
3. Rotor motion depends on the unbalance, in terms of either its type or value, and the
value of the conical parameter , given that the distance between the rotor supports can be
greater or less than twice the rotor’s transverse radius of inertia.
4. When the static eccentricity values are considerable, rotor motion can exhibit a chaotic,
quasi-periodic or periodic nature; in the latter case the period is the same as the driving rotation
or a multiple of it.
5. The inertia forces that are due to rotor motion determine vibrations in the supports that
can themselves be chaotic, quasi-periodic or periodic with a frequency that is different from
the driving one. Therefore, these vibrations can only be explained by the nonlinear behaviour
of the rotor-supports system.
6. For rotors that are characterised by a conical parameter  whose value is less than 1, the
instability is manifested with a cylindrical whirl.
When the unbalance is static, such rotors, can operate with a synchronous whirl when the
value of  is small; increasing the rotor’s angular speed and consequently raising  , the whirl
remains cylindrical regardless of the nature of the motion.
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 81

Figure 17. Attractors relative to the points C1 , C and C2 plotted for  = 0:17,  =

2:65 10 3
, and  = 0:089:
(a)  = 0:857; (b)  = 1:200.

On the other hand, when the unbalance is only of couple type, the motion at low angular
speeds is practically conical; point C covers its small orbits with a versus that is opposite that
of the driving rotation and a frequency that is double that of the fundamental. An increase in the
rotor’s angular speed makes the motion appear quasi-periodic and affected by a considerable
conical component.
7. For rotors with a conical parameter  that is greater than 1 and in pure static unbalance,
rotor motion is synchronous and cylindrical at low angular speeds; with an increase in the
angular speed the whirl is still cylindrical and periodic, with a frequency that is 1/2 the
fundamental. A further increase in  causes the motion to become chaotic and exhibit a
conical aspect.
When the unbalance is a couple one, the whirl at low angular speeds is synchronous and
practically conical; raising  values causes the synchronous motion to turn into a quasi-
82 G. Adiletta et al.

periodic whirl and successively, with further increase in  , into chaotic motion; in every case
it manifests a considerable conical aspect.
8. From the previous remarks it can be concluded that, regardless of the  values, rotor
motion with small values of  essentially depends on the nature of the unbalance and is syn-
chronous with rotation, cylindrical or prevailingly conical, according to whether the unbalance
is a static or a couple one, respectively.
If  is increased, the motion remains cylindrical only with  < 1 and in static unbalance,
while it maintains its cylindrical aspect under chaotic operating conditions; when  > 1, a
significant conical component characterises the rotor whirl; in this case, the motion can be
synchronous or sub-synchronous, quasi-periodic or chaotic.
Theoretical analysis of the nonlinear dynamic behaviour of a rigid rotor can be carried out
using the method outlined in the present paper.
This starts with the equations of motion written in Section 3 and goes through the deter-
mination of the bifurcation diagrams and frequency analysis up to the plotting of the orbits
described by three points on the rotor axis.
One of the criteria that can be adopted in the investigation in order to assign the unbalance
consists of fixing values that are in agreement with the residual unbalance which is still present
after the balancing procedures.
Two general observations remain to be pointed out. The first concerns the validity of the
proposed method but these issues can be tackled after an experimental check of the theoretical
results. The second observation deals with the unbalance of the rotor that, under more general
conditions, is of dynamic type, i.e. both static and couple unbalances are present.
In this connection, according to the results obtained through the present analysis, it can be
stressed that the motion of a rotor operating with considerable values of the static eccentricity
"s manifests in any case both cylindrical and conical components. This whirl is synchronous
at low values of the  parameter; increasing  , i.e. as a consequence of raising the rotor’s
angular speed, determines the onset of phenomena that should not be disregarded, such as the
presence in the frequency spectra of 1/2 order components, and the quasi-periodic or chaotic
nature of the whirl motion. These behaviours can only be explained once the nonlinear nature
of the system has been taken into account.

Appendix I
A. THE ROTATION MATRIX [R]
The matrix of rotation from Gxyz to Gxyz is [R]:
[R] = [Rz(3 )][Ry (2 )][Rx (1)
2 c c s s c + c s c s c + s s 3
= 4 c2s3 s1s2s3 + c1c3 c1s2s3 + s1c3 75 ;
6 2 3 1 2 3 1 3 1 2 3 1 3

s2 s1c2 c1 c2

 A Rigid Unbalanced Rotor in Journal Bearings. Part I 83

21 0 0 3 2 c 3
with
0 s2
6 0 c s 7 6
[Rx(1 )] = 4 1 1 5 ; [Ry (2)] = 4 0
2
1 0
75 ;
0 s1 c1 s2 0 c2
2 c s 0 3
[Rz(3 )] = 64 s3 c3 0 75 :
3 3

0 0 1
[ ( )]
The symbol Rj  i indicates that a rotation i along its j -axis is made to the frame Gxyz,
starting from its current position; in particular, j  indicates the axis j in its current position.
Under the hypothesis of small angles 1 and 2 , a simplified expression can be given:
2 c   c + s  c +  s 3
[R] = 64 s3 12s3 + c3 2s3 + 1c3 75 :
3 1 2 3 3 2 3 1 3

2 1 1

B. EXPRESSIONS FOR SOME OF THE SYMBOLS IN SECTION 2

=
The vector f! g f!x !y !z gT , is expressed as a function of the speeds 1 , 2 and 3 . The _ _ _
_
contribution of each speed i to the components of f! g along the axes of Gxyz is given by the
=
corresponding vector f! g_i . The expressions of the vectors f! g_i (i 1; 2; 3) can be given
_ _ _
through suitable use of the column vectors f1 ; 0; 0gT , f0; 2 ; 0gT , f0; 0; 3 g and some of the
[ ]
matrices that appear through the definition of the matrix R (see Appendix I-A):
8_ 9
>
< 1 >
=
f!g = [Rz(3)][Ry (2)] > 0 > ;
_
1
:0;
809
< >
> =
f!g = [Rz(3)] > _2 > ;
_
2
:0;
809
>
< > =
f!g = > 0 > :
_
3
: _3 ;
Thus we have
8! 9 8 _ 9
>
< x> = >
< 1> =
f!g = > !y > = f!g + f!g + f!g = [Rv ] > _2 > ; (I.1)
: !z ; _
1
: _ ;
_
2
_
3

2 c c s 0 3 2 c s 0 3
with

[Rv ] = 4 c2s3 c3 0 5 = 4 s3 c3 0 75 :

6 7  6
2 3 3 3 3

s2 0 1 s2 0 1
84 G. Adiletta et al.

Besides, the following relationship holds:

8 0 9
>
< 1> =
f!g = !f!g = ![Rv ] > 20 > :
: 30 ;
The matrix[!] is expressed by means of the vector components f!g:
2 0 ! ! 3
[!] = 64 !z 0 !x 75
z y

!y !x 0
2 0 _1s2 _3 _1c2s3 + _2c3 3
= 64 _1s2 + _3 0 _1c2c3 _2s3 75
_1c2s3 _2c3 _1c2 c3 + _2s3 0
2 0 _12 _3 _1s3 + _2c3 3
= 64 _12 + _3 0 _1c3 _2s3 75 :
_1s3 _2c3 _1c3 + _2s3 0
The expressions (26) and (28) show the inverse of the matrix [Rv ]:
2 c3 s3 0 3 2 c 3
6 c2 7
c2 77  6 s c 0 7 : s 3 0
[Rv ] 1 = 664 s
3

c3 0 5 = 4 3 3 5
3
tan 2 c3 tan 2 s3 1
2c3 2s3 1

The derivative of (I.1) with respect to time makes it possible to obtain:

0 8> _ 9 1
< 1>=
f!_ g = ddt B
@[Rv ] > _2 >C = [R ]fg + [R_ v ]f_g;
: _ ;A v
3

with
2 _2s2c3 _3c2s3 _3c3 0 3 2 _22c3 _3s3 _3c3 0 3
[R_ v ] = 64 _2s2s3 _3c2 c3 3s3 0 75 
= 64 _22 s3 _3c3 _3s3 0 75 :
_2c2 0 0 _2 0 0
Furthermore,
2 0  c 0 s 0 c 0 3
[R_ v ] = ![Rv0 ] = ! 64 20 2s3 30 c3 30 s3 0 75 :
2 2 3 3 3 3 3

20 0 0

Appendix II: Rotor Unbalance and the Frame Gxyz Jointed to the Rotor
In order to assign the motion of an unbalanced rigid rotor, it is necessary to fix the angular
phase shift between the rotating force, which is due to static unbalance, and the rotating torque,
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 85

F
Figure II.1. Static and couple unbalances of the rotor with the masses m1 and m2 : : total unbalancing force; M
F M
total unbalancing torque; i : unbalancing force due to mi ; i : unbalancing torque due to mi .

corresponding to couple unbalance. In this regard, it is worth pointing out that the above phase
shift is generally unknown in practice.
The need to assign the angular phase shift between the unbalance actions is further justified
since an experimental check of the theoretical results has to be carried out and some practical
values need to be fixed.
The easiest way to give a fixed unbalance to an assigned, perfectly balanced rotor consists
of placing a pair of two small unbalancing masses, m1 and m2 , onto the rotor itself. Each of
the two masses is assumed to be concentrated in a point.
After assigning the principal inertia reference frame C of the balanced rotor (Fig-
ure II.1), the rotor’s matrix of inertia in the same frame is:
2I 0 0
3
6 
[I] = 4 0 I I 75 : (II.1)
0 I I
The two unbalancing masses, m1 and m2 , are placed onto the rotor, at a distance rU from
the rotor axis, under the following conditions: m1 = = =
m2 mU ; d distance of the two
planes that are perpendicular to the  -axis and contain m1 and m2 ; and  =2 phase angles =
taken in the plane  (Figure II.1), in order to fix the positions of m1 and m2 .
Because of the rotation around  , at an angular speed ! , m1 and m2 become responsible
for two centrifugal forces whose resultant F has the direction and the versus of the  axis.
We thus have:
(a) Static unbalance:

F = 2mU rU !2 cos( =2)

86 G. Adiletta et al.

e = eccentricity of the center G of mass = CG = 2 mM

U rU
cos( =2): (II.2)

According to the above, the segment CG is along the  -axis and the centrifugal force F
has the direction of this axis. In dimensionless form static unbalance is expressed by the ratio
=
 e=c1 .
(b) Couple unbalance:

M1 = M2 = mU rU !2d
M = resultant of M1 and M2 = 2mU rU !2d sin( =2): (II.3)

Products of inertia with respect to the reference frame C :

I = 0
I = 2mU rU d sin( =2)
I = 0:
Taking into account that the value of e is practically negligible compared to the value of
rU , it can be assumed that:
Ixx = I
Iyy = I
Izz = I
Ixy = I = 0
Iyz = I = 2mU rU d sin( =2)
Ixz = I = 0: (II.4)

[]
This makes it possible to determine the matrix of inertia I in the frame Gxyz which is fixed
to the rotor, and the initial phase shift (!t =0) of the centrifugal force and the centrifugal
torque, that are due to the rotor’s static and couple unbalance.
It is further worth pointing out that, when = 0, the unbalance is only static, while it
is only of couple type when = =2. Under the latter conditions, the unbalancing torque
rotates in phase ( ==2) or in phase opposition ( = =2) with the x-axis of the frame
fixed to the rotor.
According to (II.4), it can be written that:
2I Ixy Ixz
3 2I 0 0
3
[I ] = 64 Ixy Iyy Iyz 75 = 64 0 I I 75 :
xx 

Ixz Iyz Izz 0 I I

By means of

Ixx = Iyy = ID ; Izz = M2z = IP ; Iyz = IC ; IID = i; IIC = ic ;

P P
 A Rigid Unbalanced Rotor in Journal Bearings. Part I 87

[
the matrices I=Izz and ] [I=Izz ] 1 that appear in the relations (24–26) and (28) can finally be
given the expressions:
21 3
2 3 6 77
 I  1 666 i 1
0 0
 I  6i 0 0
iC 77
Izz = 4 0 i iC 75 ; Izz = 6
66 i i2C i i2C 777 :
0
0 iC 1 4 0 iC i 5
i i2C i i2C

References
1. Badgley, R. H. and Booker, J. F., ‘Rigid-body rotor dynamics: Dynamic unbalance and lubricant temperature
changes’, ASME, Journal of Lubrication Technology, July, 1970, 415–424.
2. Yamamoto, T., Ishida, Y., and Kawasumi, J., ‘Oscillations of a rotating shaft with symmetrical nonlinear
spring characteristics’, Bulletin of JSME 18(123), 1975, 965–975.
3. Childs, D. W., ‘Fractional-frequency rotor motion due to nonsymmetric clearance effects’, ASME, Journal
of Engineering for Power 104, 1982, 533–541.
4. Press, W., Flannery, B. P., Teukolsky, S. A., and Vetterling, W. T., Numerical Recipes, Cambridge University
Press, Cambridge, 1986.
5. Choi, Y. S. and Noah, S. T., ‘Nonlinear steady-state response of a rotor-support system’, ASME Journal of
Vibration, Acoustics, Stress, and Reliability in Design 109, 1987, 255–261.
6. Ishida, Y., Ikeda, T., Yamamoto, T., and Murakami, S., ‘Nonstationary vibration of a rotating shaft with
nonlinear spring characteristics during acceleration through a critical speed (A critical speed of a 1/2-order
subharmonic oscillation)’, JSME International Journal, Series III 32(4), 1989, 575–584.
7. Shaw, J. and Shaw, S. W., ‘Instabilities and bifurcations in a rotating shaft’, Journal of Sound and Vibration
132(2), 1989, 227–244.
8. Szemplinska-Stupnicka, W., The Behaviour of Nonlinear Vibrating Systems, Kluwer, Dordrecht, 1990.
9. Ehrich, F. E., ‘Some observations of chaotic vibration phenomena in high-speed rotordynamics’, ASME,
Journal of Vibration and Acoustics 113, 1991, 50–57.
10. Shaw, J. and Shaw, S. W., ‘Non-linear resonance of an unbalanced shaft with internal damping’, Journal of
Sound and Vibration 147(3), 1991, 435–451.
11. Ishida, Y., Yamamoto, T., and Murakami, S., ‘Nonstationary vibration of a rotating shaft with nonlinear
spring characteristics during acceleration through a critical speed (A critical speed of a 1/3-order subharmonic
oscillation)’, JSME International Journal, Series III 35(3), 1992, 360–368.
12. Abu-Mahfouz, I. A., ‘Routes to chaos in rotor dynamics’, Ph.D. Thesis, Case Western Reserve University,
Cleveland, 1993.
13. Chang, C. O. and Cheng, J. W., ‘Non-linear dynamics and instability of a rotating shaft-disk system’, Journal
of Sound and Vibration 160(3), 1993, 433–454.
14. Genta, G., Delprete, C., Tonoli, A., and Vadori, R., ‘Conditions for noncircular whirling of nonlinear isotropic
rotors’, Nonlinear Dynamics 4, 1993, 153–181.
15. Ishida, Y., ‘Nonlinear vibrations and chaos in rotordynamics’, JSME International Journal, Series C 37(2),
1994, 237–245.
16. Adiletta, G., Guido, A. R., and Rossi, C., ‘Chaotic motions of a rigid rotor in short journal bearings’,
Nonlinear Dynamics 10, 1996, 251–269.
17. Adiletta, G., Guido, A. R., and Rossi, C., ‘Non-periodic motions of a Jeffcott rotor with non-linear elastic
restoring forces’, Nonlinear Dynamics 11, 1996, 37–59.
18. Meirovitch, L., Methods of Analytical Dynamics, McGraw-Hill, New York, 1970.
19. Lund, J. W., ‘Self-excited, stationary whirl orbits of a journal in a sleeve bearing’, Ph.D. Thesis, Rensselaer
Polytechnic Institute, New York, 1966.