ARTICLE IN PRESS

Mechanical Systems
and
Signal Processing
Mechanical Systems and Signal Processing 18 (2004) 1369–1380
www.elsevier.com/locate/jnlabr/ymssp

Analysis of whirl speeds for rotor-bearing systems supported

on fluid film bearings
Madhumita Kalitaa, S.K. Kakotyb,*
a
Assam Engineering Institute, Guwahati 781003, India
b
Department of Mechanical Engineering, Indian Institute of Technology, North Guwahati 781 039, India

Abstract

There are several numerical approximations for vibration analysis of rotor bearing systems and the most
popular approach is the Finite Element Method. In the light of the above it is proposed to undertake a
study on the dynamic behaviour of Timoshenko beam supported on hydrodynamic bearings incorporating
internal damping using FEM model. Critical speeds are estimated for synchronous whirl at different
operating conditions using Campbell diagrams. It is observed in the analysis that in addition to the natural
whirl frequencies, for every spin speed another whirling frequency appears in the solution, which is around
half the spin speed. In case of fluid film bearings, half whirl is very common phenomenon. In case of
dynamic coefficients evaluated using short bearing approximation, it is observed that these additional
frequencies are of the same order as that of the synchronous whirling frequencies. The additional
frequencies around half the spin speeds are found using only finite bearing dynamic coefficients; this clearly
indicates the deficiency in the use of short bearing approximation in similar work.
r 2003 Published by Elsevier Ltd.

1. Introduction

Rotating machines are extensively used in diverse engineering applications. The accurate
prediction of dynamic characteristics is important in the design of any type of rotating machinery.
There have been many studies relating to the field of rotor dynamics during the past years. Out of
the published works, the most extensive portion of literature on rotor dynamics is concerned with
determination of critical speeds, natural whirl frequencies, instability thresholds and imbalance
response. Several numerical approximations have been successfully developed to analyse the
dynamic behaviour of rotor systems. However, the most popular approach well suited for
modelling large-scale and complicated rotor systems is the Finite Element Method. Ruhl [1]

*Corresponding author.
E-mail addresses: madhumita [email protected] (M. Kalita), [email protected], [email protected] (S.K. Kakoty).

0888-3270/$ - see front matter r 2003 Published by Elsevier Ltd.

doi:10.1016/j.ymssp.2003.09.002
 ARTICLE IN PRESS

1370 M. Kalita, S.K. Kakoty / Mechanical Systems and Signal Processing 18 (2004) 1369–1380

and Ruhl and Booker [2] are the first to utilise the finite element method to study the stability and
unbalance response of turbo rotor systems. In their analyses, only elastic bending energy and
translational kinetic energy were included. Many effects, such as the rotary inertia, gyroscopic
moments, shear deformations, internal and external damping were neglected. Thorkildsen [3]
included rotary inertia and gyroscopic moments in the finite element model. Nelson and
McVaugh [4] generalised Ruhl’s work by utilising a finite element formulation including the
effects of rotary inertia, gyroscopic moments and axial load to model a flexible rotor system
supported on linear stiffness and viscous damping bearings. The work of Zorzi and Nelson [5] was
the generalisation of the work of Nelson and McVaugh [4] by including both internal viscous and
hysteretic damping in the same finite element model. Nelson [6] utilised Timoshenko beam theory
for establishing shape functions and, thereby included transverse shear effects. Ozguven and
Ozkan [7] presented the combined effects of shear deformation and internal damping to analyse
the natural whirl speeds and unbalance response of rotor-bearing system. Chen and Ku [8]
developed a C0 class Timoshenko beam finite element model and Ku [9] included internal
damping in the same model. Rao [10] provided analysis of dual rotor supported on fluid film
bearings.
In view of the above the present investigation is made to find out the behaviour of rotor-bearing
system supported on fluid film bearings. A finite element formulation is carried out [10] for
Timoshenko beam incorporating translational inertia, rotary inertia, bending deformation, shear
deformation, gyroscopic effect and internal damping. Stiffness and damping coefficients of the
bearings are estimated based on formulation of Lund [11].

2. Theory

2.1. Finite element formulation of rotors

A rotor-bearing system is composed of a uniform shaft of length l rotating at a constant speed

O (Fig. 1) and supported by two bearings. It is assumed that as compared to translational motion,
axial motion is negligible. Fixed reference is X–Y–Z and rotating reference is x–y–z. A typical
cross-section of the shaft, in a deformed state, located at a distance x from the left end can be
described by translations V ðx; tÞ; W ðx; tÞ and rotations Bðx; tÞ; Gðx; tÞ in Y and Z directions. The
relationships can be expressed as
V ðx; tÞ ¼ Vb ðx; tÞ þ Vs ðx; tÞ;
W ðx; tÞ ¼ Wb ðx; tÞ þ Ws ðx; tÞ;
Bðx; tÞ ¼ qWb ðx; tÞ=qx;
Gðx; tÞ ¼ qVb ðx; tÞ=qx; ð1Þ

where Vb; Vs ; and Wb ; Ws are translations due to bending and shear in the Y and Z directions,
respectively.
Timoshenko beam includes all the effects, i.e., bending deformation, rotary inertia, gyroscopic
effect and shear deformation. The strain energy is due to bending and shear. Translation, rotation
and gyroscopic effect contribute towards kinetic energy. In the present model each element has
 ARTICLE IN PRESS

M. Kalita, S.K. Kakoty / Mechanical Systems and Signal Processing 18 (2004) 1369–1380 1371

Fig. 1. Displacement variables and coordinate system.

two nodes and each node has four generalised displacements. The nodal displacement vector is
given by

fqge ¼ fV1 ; W1 ; B1 ; G1 ; V2 ; W2 ; B2 ; G2 gT : ð2Þ

For shaft element, strain energy and dissipation function are given as in [9]. Shear correction
factor k is given by k ¼ 6ð1 þ nÞ=ð7 þ 6nÞ with n as the Poisson’s ratio. Internal damping is
included using the formulations of Ozguven and Ozkan [7],
" #
Za Zb
½Z ¼ ;
Zb Za
1 þ ZH
Za ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffi;
1 þ Z2H
ZH
Za ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ OZv ; ð3Þ
1 þ Z2H

ZH and Zv denote the viscous damping coefficient and hysteretic loss factor of the shaft material.
Following Lagrangian approach, the elemental equation of motion is given by [9]

. e þ ðZv ½K e  O½G e Þfqg

ð½MT e þ ½MR e Þfqg ’ þ ðZa ½K e þ Zb ½KC e Þfqge ¼ fF e g; ð4Þ

where

½K e ¼ ½Kb e þ ½Ks e ;
½G e ¼ ½H e  ½H eT ;
Z l
0 e
½Kb ¼ ½Nb0 T EI½Nb0 dx;
0
 ARTICLE IN PRESS

1372 M. Kalita, S.K. Kakoty / Mechanical Systems and Signal Processing 18 (2004) 1369–1380

Z l
½Ks0 e ¼ ½Ns0 T kGA½Ns dx;
0
Z l
e
½MT ¼ ½Nt T rA½Nt dx;
0
Z l
e
½MR ¼ ½Nb T Id ½Nb dx;
0
Z l " #
e T 0 0
½H ¼ Ip ½Nb ½Nb dx: ð5Þ
0 1 0

The equation of motion for disk is given as [9]

½M d fq. d g þ O½G d fq’ d g ¼ ½0 ; ð6Þ
where ½M d and ½G d are mass and gyroscopic matrices of disk only.

2.2. Finite element formulation for bearings

Eight spring and damping coefficients are employed for the modelling of bearings in the present
work [9]. In this model, the forces at each bearing are assumed to obey the governing equations of
the following form
" # " #
cyy cyz kyy kyz
fq’ b g þ fqb g ¼ fF b g; ð7Þ
czy czz kzy kzz

where fqb g ¼ f v w gT is the bearing displacement vector and cij and kij are the bearing and
damping coefficients; fF b g is the vector of bearing forces.
The resultant system equation of motion then becomes
. e  ð½cb  O½G s Þfqg
½M s fqg ’ þ ð½K s þ ½kb Þfqge ¼ fF g; ð8Þ

4 disks
60.3 kg each

0.09m

Bearing Bearing

0.5 0.5 0.5 0.5 0.5 0.5 0.5 m

Element No.

(1) (2) (3) (4) (5) (6) (7)

1 2 3 4 5 6 7 8
Node No.

Fig. 2. Rotor-bearing system with disks.

 ARTICLE IN PRESS

M. Kalita, S.K. Kakoty / Mechanical Systems and Signal Processing 18 (2004) 1369–1380 1373

where ½M s ; ½G s ; ½K s are the assembled mass, gyroscopic and stiffness matrix of the system
incorporating the disk mass matrix; ½cb and ½kb are the bearing damping and stiffness matrices,
respectively (see Fig. 2). To determine the eigenvalues and eigenvectors the right-hand side of the
equation is set to zero. The element matrices are identical to Rao [10].

3. Results and discussion

3.1. Validation

A steel shaft having diameter 10.16 cm and length 127 cm supported by two identical isotropic
bearings at both the ends is discretised into seven equal finite elements. The present set of results are
compared with those obtained by Ku [9] and Ozguven and Ozkan [7] as shown in Table 1 and are
found to be in good agreement. The whirl speed map of simple rotor systems supported on undamped
isotropic bearings having stiffness coefficients Kvv ¼ Kww ¼ 1:7513 107 N/m and Kvw ¼ Kwv ¼
0:0 are presented in Figs. 3 and 4. In Fig. 3 hysteretic damping is considered, whereas viscous damping
is considered in Fig. 4. It has been observed that for the shaft material with viscous internal damping
Zv ¼ 0:0002 s, critical speeds for the first and second forward modes are found to be 5000 and
10,782 rpm, respectively in the present case, whereas these are reported as 4960 and 10,500 rpm,
respectively by Ku [9]. All other results are found to be in good agreement. The damping coefficients
Cvv ¼ Cww ¼ 1:7513 103 ; Cvw ¼ Cwv ¼ 0:0 are also included along with the stiffness coefficients
for the damped isotropic bearings. The whirl speed maps are presented in Figs. 5 and 6. These plots
are compared with those obtained by Ku [9] and found to be in good agreement.
In view of the above, the model used for the hypothetical bearings is used for analysis of the
rotor-bearings systems supported on fluid film bearings.

3.2. Numerical example

A typical simply supported rotor disk system mounted on two identical fluid film bearings is
analysed using the present finite element model (Fig. 2). The physical properties of the shaft and
bearing geometry are given in Table 2.

Table 1
Whirl speeds in rad/s of a uniform shaft with isotropic undamped flexible bearings at a spin speed of 4000 rpm
ZH ¼ 0:0002 Zv ¼ 0:0002 s
Mode Present work Ref. [9] Ref. [7] Present work Ref. [9] Ref. [7]
1F 519.4 519.78 520.06 519.4 519.75 520.01
1B 520.0 519.23 521.79 520.0 521.48 519.54
2F 1091.6 1094.40 1096.01 1091.6 1095.13 1095.28
2B 1094.8 1090.90 1095.34 1094.8 1094.52 1091.77
3F 2228.0 2238.53 2222.78 2227.9 2216.81 2244.72
3B 2241.9 223.80 2206.94 2241.8 2201.25 2229.82
4F 4954.1 4968.16 4447.40 4933.6 4413.32 5020.12
4B 4987.7 4935.91 4411.81 4987.2 4378.95 4986.74
 ARTICLE IN PRESS

1374 M. Kalita, S.K. Kakoty / Mechanical Systems and Signal Processing 18 (2004) 1369–1380

Fig. 3. Campbell diagram for rotor-bearing system with undamped isotropic bearing.

Fig. 4. See caption of Fig. 3.

The dynamic coefficients for plain journal bearing with two axial grooves 180 apart and of 20
circumferential extent each are evaluated using the code based on the numerical scheme of Lund
[11]. Initially Sommerfeld numbers are calculated for different operating conditions including
speeds; then corresponding eccentricity ratios are found from the code. The code facilitates
estimation of the stiffness and damping coefficients for particular eccentricity ratio. These
dynamic coefficients of the bearings are used in the FEM model to estimate the natural whirl
frequencies of the rotor-bearing system. Results are presented in the form of Campbell diagrams
for finite bearing in Figs. 7–9. Fig. 7 presents critical speeds for first four modes for rotor with
hysteretic damping. Critical speeds for rotors with viscous damping are presented in Fig. 8. Fig. 9
demonstrates the effect of both hysteretic and viscous damping. Not much difference in critical
speeds is observed in all the three cases considered namely hysteretic damping, viscous damping
and both hysteretic and viscous damping.
 ARTICLE IN PRESS

M. Kalita, S.K. Kakoty / Mechanical Systems and Signal Processing 18 (2004) 1369–1380 1375

Fig. 5. Campbell diagram for rotor-bearing system with damped isotropic bearing.

Fig. 6. See caption of Fig. 5.

Table 2
Physical property of the shaft and geometry of the bearings
Density of the shaft material, r 7830 kg/m3
Elastic modulus of the shaft material, E 2.08E+11 N/m2
Diameter of the bearing, D 0.09 m
Bearing load, W 1960.0 N
Radial clearance, C 0.00254 cm

3.2.1. Half-frequency whirling

An interesting observation of the analysis is that in addition to the synchronous natural whirl
frequencies for the first four modes, for every spin speed another whirling frequency appears in
the solution as shown in Figs. 7–9. This particular frequency is around half the spin speed; and
 ARTICLE IN PRESS

1376 M. Kalita, S.K. Kakoty / Mechanical Systems and Signal Processing 18 (2004) 1369–1380

Fig. 7. Campbell diagram for rotor-bearing system with hysteretic damping using finite bearing stiffness and damping
coefficients.

Fig. 8. Campbell diagram for rotor-bearing system with viscous damping using finite bearing stiffness and damping
coefficients.

therefore this value of frequency is identified as the half-frequency whirling due to the use of fluid
film bearings. It may be mentioned that half-frequency whirling or oil whirl is very common in
case of fluid film bearings.

3.2.2. Short bearing approximation

It has been found that many researchers use short bearing approximation to estimate the
stiffness and damping coefficients of fluid film bearings [10]. The results of the system with
hysteretic damping supported on fluid film bearings using short bearing stiffness and damping
coefficients are also presented. Critical whirl frequencies for both finite and short bearings are
estimated as shown in Table 3. For the first mode, critical speeds for both short and finite bearing
are same. But from the second mode, in case of short bearings, critical speeds are more than those
 ARTICLE IN PRESS

M. Kalita, S.K. Kakoty / Mechanical Systems and Signal Processing 18 (2004) 1369–1380 1377

Fig. 9. Campbell diagram for rotor-bearing system with hysteretic and viscous damping using finite bearing stiffness
and damping coefficients.

Table 3
Critical speeds in rad/s for rotor-bearing system with hysteretic damping
Mode no. Viscosity=0.00568 Ns/m2
Finite bearing Short bearing
Backward Forward Backward Forward
I 79.03 79.03 79.03 79.03
II 304.83 310.52 304.83 317.54
III 619.29 654.38 620.96 657.89
IV 933.87 998.24 937.09 994.73

for finite bearing case. However, for the fourth forward mode, short bearing results show
decreasing trend of the frequencies.
In addition to the synchronous whirling frequencies, additional frequencies for the first four
modes are also plotted in Fig. 10 using short bearing stiffness and damping coefficients. It may be
observed that these additional frequencies are of the same order as that of the regular
synchronous whirling frequencies. But in the case of finite bearings these frequencies depicted the
half-frequency whirling condition. This has shown the deficiency in the use of short bearing
approximation in studies relating to estimation of whirling frequencies, both synchronous and
half-frequency whirl conditions.

3.2.3. Effect of bearing damping coefficients

To demonstrate the effect of the bearing damping coefficients, another Campbell diagram is
presented in Fig. 11. The half-frequency whirl is not found in the absence of bearing damping
coefficients as seen in the figure. Comparing this figure with Fig. 9, it has been observed that the
synchronous whirling frequencies in all the four modes are not much different in absence of
bearing damping coefficients.
 ARTICLE IN PRESS

1378 M. Kalita, S.K. Kakoty / Mechanical Systems and Signal Processing 18 (2004) 1369–1380

Fig. 10. Campbell diagram for rotor-bearing system with hysteretic damping using short bearing stiffness and damping
coefficients.

Fig. 11. Campbell diagram for rotor-bearing system without bearing damping coefficients (finite bearing).

3.2.4. Effect of cross-coupled stiffness

To investigate the effect of cross-coupled stiffness coefficients of the bearings, a Campbell
diagram is plotted as shown in Fig. 12. In this case the bearing cross-coupled stiffness coefficients
are not used for estimating whirling frequencies. Comparing Fig. 12 with Fig. 9, it can be observed
that the additional frequencies in absence of cross-coupled stiffness coefficients are of much lower
order. Comparing Fig. 12 with Fig. 11, it is found that in absence of bearing damping coefficients,
the additional frequencies are not at all there. But in absence of bearing cross-coupled stiffness
additional frequencies of lower order are present. Therefore, it can be concluded that the root
cause of half-frequency whirling is bearing damping coefficients and not the cross-coupled
stiffness coefficients.
 ARTICLE IN PRESS

M. Kalita, S.K. Kakoty / Mechanical Systems and Signal Processing 18 (2004) 1369–1380 1379

Fig. 12. Campbell diagram for rotor-bearing system without bearing cross-coupled stiffness coefficients (finite bearing).

4. Conclusion

The study to find out the behaviour of rotor-bearing system mounted on two identical fluid film
bearings results in some interesting observations. From the presented results it can be concluded
that in addition to the natural whirl frequencies for the first four modes, for every spin speed
another whirling frequency appears in the solution, which is around half the spin speed. But in
case of short bearings, these additional frequencies are of the same order as that of the regular
whirling frequencies. The half-frequency whirling phenomenon can be attributed due to bearing
damping; as these frequencies disappear when the bearing damping coefficients are not used.
The presented results may be of interest to the designers and researchers. Nevertheless, the
observations made here from the theoretical analysis should have experimental verification, which
was not possible at the present time due to unavailability of test rig. It would be perhaps better to
carry out more rigorous, particularly experimental, investigations to characterise any rotor-
bearing system supported on fluid film bearings.

References

[1] R.L. Ruhl, Dynamics of distributed parameter rotor systems: transfer matrix and finite element techniques, Ph.D.
dissertation, Cornell University, 1970.
[2] R.L. Ruhl, J.F. Booker, A finite element model for distributed parameter turbo rotor systems, Journal of
Engineering for Industry ASME 94 (1972) 128–132.
[3] T. Thorkildsen, Solution of a distributed mass and unbalanced rotor system using a consistent mass matrix
approach, MSE Engineering Report, Arizona State University, 1972.
[4] H.D. Nelson, J.M. McVaugh, The dynamics of rotor-bearing systems using finite elements, Journal of Engineering
for Industry 98 (1976) 593–600.
[5] E.S. Zorzi, H.D. Nelson, Finite element simulation of rotor-bearing systems with internal damping, Journal of
Engineering Power ASME 99 (1977) 71–76.
[6] H.D. Nelson, A finite rotating shaft element using Timoshenko beam theory, Journal of Mechanical Design
ASME 102 (1980) 793–803.
 ARTICLE IN PRESS

1380 M. Kalita, S.K. Kakoty / Mechanical Systems and Signal Processing 18 (2004) 1369–1380

[7] H. Nevzat Ozguven, Z. Levent Ozkan, Whirl speeds and unbalance response of multibearing rotors using finite
elements, Journal of Vibration, Acoustics, Stress, Reliability in Design 106 (1984) 72–79.
[8] L.W. Chen, D.M. Ku, Finite element analysis of natural whirl speeds of rotating shafts, Computers and Structures
40 (3) (1991) 741–747.
[9] D.M. Ku, Finite element analysis of natural whirl speeds for rotor-bearing systems with internal damping, Journal
of Mechanical Systems and Signal Processing 12 (5) (1998) 599–610.
[10] J.S. Rao, Rotor Dynamics, New Age International Publishers, New Delhi, pp. 146–266.
[11] J.W. Lund, Rotor-Bearing Dynamics, Technical University of Denmark Lecture notes, ISBN 83-04-00267-1, 1979.