Effect of Shaft Vibration on the Dynamics of Gear and Belt

Drives

Dissertation

Presented in Partial Fulfillment of the Requirements of the Degree Doctor of

Philosophy in the Graduate School of The Ohio State University

By

Sanjib Chowdhury, M.E.

Graduate Program in Mechanical Engineering

The Ohio State University

2010

Dissertation Committee:

Prof. Rama K. Yedavalli, Adviser

Prof. Daniel A. Mendelsohn, Co-adviser

Prof. Ahmet Kahraman

Prof. Gary L. Kinzel


c Copyright by

Sanjib Chowdhury

2010
 ABSTRACT

Gear and belt drives are widely used as effective means of transmitting power includ-

ing, but not limited to automobiles, heavy duty vehicles, helicopters, etc. Noise and wear-

related failures in these systems have their roots in the vibration of different components of

these systems (e.g., the dynamic forces at the gear mesh is a major source of noise in a gear

box). Secondary sources of noise appear due to vibration of various internal components

such as shafts and bearings. This work aims at developing linear mathematical models for

single-mesh spur and helical gears mounted on compliant parallel shafts, with and with-

out considering gyroscopic effects, and single span of a serpentine belt with pulleys also

mounted on parallel compliant shafts. The findings of the research can provide practical

guidance during design or troubleshooting under operating conditions.

Spur gears are simple to design and widely used to transmit power from a drive shaft to

a driven shaft in many applications. In the majority of cases, the rotational speeds are low.

A mathematical model for the closed form approximate eigensolution of a pair of coupled

non-gyroscopic spur gears on parallel shafts is developed. The model is a hybrid discrete-

continuous one where the gears modeled as rigid disks along with the mesh spring form

the discrete elements while the elastic shafts having transverse as well as torsional flexi-

bility constitute the continuous elements. The non-dimensional governing equations along

with the natural boundary conditions are developed using the Hamiltons principle. The

governing equations of the shafts for flexural and torsional vibrations and the equations of

ii
motion of the disks are written in an extended operator form to prove the self-adjointness

of the system. Matching boundary conditions at the disk-shaft interfaces prevent the use of

Galerkins method, a natural extension of the extended operator formulation. An efficient

Rayleighs principle based energy method, the assumed modes method is used to discretize

the system equations where the matching conditions are incorporated with the use of La-

grange multipliers. Orthonormal global basis functions for flexure and torsion are chosen

from separate families. The sensitivities of the natural frequencies of different modes to

mesh stiffness, torsional and flexural rigidities of the shafts, and lengths of the shafts are

examined and the results are correlated with the modal energy distributions. Excitation in

the form of the loaded static transmission error at the gear mesh is identified and converted

to the discretized form and the response for the same is calculated.

Helical gears are employed for quiet operation in drive trains where the rotational

speeds are sufficiently high. Three-dimensional gyroscopic model of a pair of helical gears

mounted at the ends of compliant spinning shafts is developed. The gears are modeled as

rigid disks with the tooth compliance modeled as translational and rotational springs con-

necting the disks with the shafts having transverse and torsional vibrations. The rotational

spring at the gear mesh accounts for the energy stored due to the relative tilting of the gears.

The rotational speed is high and therefore, the gyroscopic effect is non-negligible. Hamil-

tons principle is used to obtain the non-dimensional governing equations and equations of

motion of the disks. Excitation in the form of the loaded static transmission error at the gear

mesh is incorporated in the equations of motion. This excitation appears as a combination

of forces and moments and these are identified from the expression of the virtual work. An

extended operator formulation is employed to simplify the system equations to a compact

analytical form, which is prototypical of a gyroscopic system involving mass, gyroscopic,

iii
stiffness, and rotational stiffness operators. This is then conveniently discretized, along

with the excitation forces and moments using Galerkins method as there are no match-

ing boundary conditions as in the previous model. Basis functions used for the rotating

model are global basis functions similar to that used in the non-gyroscopic model. Natural

frequency sensitivity to various system parameters such as the rotational speed and mesh

stiffness is determined and explained using modal strain energies. Response due to the

loaded static transmission error is obtained by performing modal analysis after reducing

the discretized system to a first order form. The study shows that the gyroscopic effect is

present even for short lengths of the shafts. Natural frequency veering between flexural fre-

quencies is present at low speeds and rare at high speeds where stiffening due to rotational

stress reduces the coupling between the shafts. For coupled frequencies, response of the

system is found to be approximated by modal superposition of a smaller number of modes.

The advantages of belt drives are low cost, easy maintenance, flexible locations of driver

and driven shafts to name a few. Serpentine belt drives are widely used in automobiles and

heavy vehicles for driving accessories such as the alternator, air conditioner, water pump,

etc., by the engine power delivered from the crankshaft. The drive suffers from noise and

belt tension fluctuation, which have their roots in the system vibration. At high rotation

speeds, flexibility of the shafts attached to the pulleys provides additional degrees of free-

dom to the system. The mathematical model consists of the belt modeled as a combination

of extensional and torsional spring, which is justified for flat, wide belts with high initial

tension, having small bending stiffness. The compliant shafts are attached to rigid pul-

leys which are connected by the belt in the reversed wrapped configuration. This hybrid

discrete-continuous system consists of the shafts modeled as continuous elements while

the pulleys with the attached belt modeled as discrete elements. The mathematical model

iv
and the analysis are similar to the helical gear-shaft model, except that the helix angle is

zero and the axis of the torsional spring is different. One of the pulleys is assumed to be

attached to the crankshaft. Periodic load fluctuation from the engine in the form of a force

on the shaft attached to this pulley is considered and the response in the form of tension

fluctuation of the extensional spring is determined.

The study provides practical guidance to the analysis and design engineers who con-

sider the noise and vibration in gear and belt drives. Consideration of shaft flexibility for

specific design parameters makes the system free of secondary sources of vibration induced

noise and failure.

v
 To

My Mother

vi
 ACKNOWLEDGMENTS

I would like to express my most sincere gratitude to my adviser, Prof. Yedavalli who

has been and will be a constant source of inspiration to me. He created an environment con-

ducive of conducting fruitful research. I would like to thank my co-adviser, Prof. Mendel-

sohn for having an eye for the details, starting from the mathematical complexities of the

work to the write up of the thesis. Many thanks to Prof. Kinzel who not only served as

a committee member but also encouraged the research by providing me the much needed

financial support. I am greatly indebted to him. I would like to thank Prof. Kahraman for

serving as a committee member and suggesting me to widen the scope of applicability of

the work.

I would especially like to thank Prof. Parker for his valuable inputs and suggestions. I

have benefited from the courses he offered. Interactions with him, most of the times, are

intellectually stimulating.

I am grateful to ANSYS Inc. for allowing me to work as an intern, which also provided

me with the uninterrupted time for finishing up the thesis. It was also a great opportunity

to see how research problems can be solved by working as a team.

I am thankful to the Prof. Srinivasan and the Mechanical Engineering department at the

Ohio State for having the TA program and a great computing lab. I am indebted to my for-

mer and present lab members, friends in Columbus, and staff members of the department.

vii
 Last, but not the least, I am ever grateful to my family who endured the most for my

education. I am grateful to my upbringing and the values that my family inculcated in me

that helped me face tough situations with steadfastness.

viii
 VITA

January 24, 1976........................................Born-Calcutta, India

1998............................................................B.E., Jadavpur University, India
2001............................................................M.E., Indian Institute of Science, India
2004-present...............................................Graduate Research, The Ohio State University

PUBLICATIONS

1. Sen, D., Chowdhury, S., and Pandey, S. R., 2004, Geometric Design of Interference-
Free Planar Linkages, Mechanism and Machine Theory, 39(7), pp. 737759.

FIELDS OF STUDY

Major Field: Mechanical Engineering

ix
 TABLE OF CONTENTS

Page

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Vita . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Chapters:

1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Objectives and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2.1 Rotor systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.2.2 Geared shaft systems . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2.3 Belt drives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3 Scope of the Present Work . . . . . . . . . . . . . . . . . . . . . . . . . 17

2. EIGENSOLUTION OF NONGYROSCOPIC SPUR GEARED SYSTEM AT-

TACHED TO COMPLIANT SHAFTS . . . . . . . . . . . . . . . . . . . . . . 21

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2 Modeling and Equations . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3 Extended Operator Formulation . . . . . . . . . . . . . . . . . . . . . . 33
2.4 Discretization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.5 Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . 45

x
 2.5.1 Sensitivity of the Natural Frequency to Various System Parameters 47
2.5.2 Response due to Loaded Static Transmission Error . . . . . . . . 54

3. EIGENSOLUTION OF GYROSCOPIC HELICAL GEARED SYSTEM AT-

TACHED TO COMPLIANT SHAFTS . . . . . . . . . . . . . . . . . . . . . . 59

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.2 Kinematics of Helical Gear Pair on Parallel Shafts . . . . . . . . . . . . 62
3.3 Modeling and Equations of Motion . . . . . . . . . . . . . . . . . . . . 65
3.4 Extended Operator Formulation and Discretization . . . . . . . . . . . . 72
3.5 Response Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.6 Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.6.1 Sensitivity of the Natural Frequency to Various System Parameters 82
3.6.2 Response due to Loaded Static Transmission Error . . . . . . . . 89

4. DYNAMICS OF BELT PULLEY SHAFT SYSTEMS . . . . . . . . . . . . . . 93

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.2 Modeling and Equations of Motion . . . . . . . . . . . . . . . . . . . . 97
4.3 Extended Operator Formulation and Discretization . . . . . . . . . . . . 104
4.4 Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.4.1 Sensitivity to the Rotational Speed . . . . . . . . . . . . . . . . 108
4.4.2 Force Response due to Excitation at the Shaft . . . . . . . . . . . 110

5. SUMMARY AND FUTURE WORK . . . . . . . . . . . . . . . . . . . . . . 115

5.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

5.1.1 Eigensolution of nongyroscopic spur geared system attached to
compliant shafts . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.1.2 Eigensolution of gyroscopic helical geared system attached to
compliant shafts . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.1.3 Dynamics of Belt Pulley Shaft System . . . . . . . . . . . . . . 119
5.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.2.1 Eigensolution of the geared shaft systems considering bearing
flexibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.2.2 Eigensolution of the geared shaft systems with flexibility of the
gear bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.2.3 Eigensolution of the geared shaft systems with Timoshenko beam
theory and axial vibration . . . . . . . . . . . . . . . . . . . . . 123
5.2.4 Effect of shaft vibration on the dynamics of continuous belt ser-
pentine belt drive system with and without considering the belt
flexural vibration . . . . . . . . . . . . . . . . . . . . . . . . . . 123

xi
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

xii
 LIST OF TABLES

Table Page

2.1 List of symbols defining spur geared shaft system. Subscripts 1 and 2 de-
note the quantities belonging to the first and the second shafts, respectively.
Overbar indicates dimensional quantities, otherwise it is non-dimensional. . 27

2.2 List of flexural and torsional deformations expressed as a combination of a

series of basis functions. . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

2.3 Dimensional and dimensionless parameters of a coupled disk-shaft system . 46

2.4 Normalized modal strain energies . . . . . . . . . . . . . . . . . . . . . . 48

3.1 List of symbols defining spinning helical geared shaft system. Subscripts
1 and 2 denote the quantities belonging to the first and the second shafts,
respectively. Overbar indicates dimensional quantities, otherwise it is non-
dimensional. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

3.2 Dimensional and dimensionless parameters of the coupled disk-shaft system 80

4.1 List of symbols defining belt-pulley-shaft system. Subscripts 1 and 2 de-

note the quantities belonging to the first and the second shafts, respectively.
Overbar indicates dimensional quantities, otherwise it is non-dimensional. . 100

4.2 Dimensional and dimensionless parameters of a belt-pulley-shaft system . . 109

xiii
 LIST OF FIGURES

Figure Page

1.1 Jeffcott rotor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Gear pair dynamic model. . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Two models of geared rotor system (a) Non-gyroscopic spur (b) Gyro-
scopic helical geared shaft systems. . . . . . . . . . . . . . . . . . . . . . 6

1.4 Accessory drive system of an automobile. . . . . . . . . . . . . . . . . . . 7

1.5 Schematic of a serpentine belt drive. . . . . . . . . . . . . . . . . . . . . . 8

1.6 Belt-pulley-shaft model for single span of the belt. . . . . . . . . . . . . . 9

2.1 Model of the coupled gear-shaft system. . . . . . . . . . . . . . . . . . . . 26

2.2 First six normalized basis functions for flexure of the first shaft; (a) Basis
functions 1i (x1 ) for series expansion of the left segment; (b) Basis func-
tions 2i (x1 ) for series expansion of the right segment. . . . . . . . . . . . 39

2.3 First six normalized basis functions for torsion of the first shaft; (a) Basis
functions 1i (x1 ) for series expansion of the left segment; (b) Basis func-
tions 2i (x1 ) for series expansion of the right segment. . . . . . . . . . . . 40

2.4 Mode shapes for the system shown in Table 1. (a) Mode 20 (b) Mode 22
(c) Mode 26. The amplitude of torsional vibration is proportional to length
of the arrow and are shown at the ends and the disk locations. . . . . . . . . 47

2.5 Natural frequency vs. the non-dimensional mesh stiffness (km ) along with
the normalized modal strain energies at different modes. Modes 20, 22,
and 26 are shown in bold. . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

xiv
2.6 Modal mesh energies vs. the non-dimensional mesh stiffness (km ) for the
modes 20, 22, and 26. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

2.7 Natural frequency vs. K1 along with the normalized modal strain energies
at different modes. Modes 20, 22, and 26 are shown in bold. . . . . . . . . 50

2.8 Modal torsional energies vs. the non-dimensional torsional rigidity (1st
shaft) for the modes 20, 22, and 26. . . . . . . . . . . . . . . . . . . . . . 50

2.9 Natural frequency vs. K8 along with the normalized modal strain energies
at different modes. Modes 20, 22, and 26 are shown in bold. . . . . . . . . 52

2.10 Modal bending energies vs. K8 for the modes 20, 22, and 26. . . . . . . . . 53

2.11 Variation of the 22nd and 26th natural frequencies with K8 for different shaft
lengths; (a) Variation of the mode 22 for the nominal lengths (), 0.95
(), 0.90 ( ), and 0.85 ( ) times the nominal lengths; (b) Variation
of the mode 26 for the nominal lengths (), 0.50 (), and 0.35 ( )
times the nominal lengths. . . . . . . . . . . . . . . . . . . . . . . . . . . 54

2.12 Non-dimensional system responses due to the loaded static transmission

error for three different modal damping ratios (). (a) The 1st shaft (b) The
2nd shaft (c) Mesh deflection (d) Mesh force. Torsional deflection of the
first disk ( ) and flexural deflection at the disk center of mass () are
shown for (a) and (b). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

2.13 Non-dimensional actual and approximate responses due to the loaded static
transmission error for 4 % modal damping along with the one, two, and
three mode approximations. (a) The 1st shaft (b) The 2nd shaft (c) Mesh
deflection (d) Mesh force. Torsional deflection of the first disk ( ) and
flexural deflection at the disk center of mass () are shown for (a) and (b). . 57

2.14 Non-dimensional actual and approximate responses due to the loaded static
transmission error for 0.5 % modal damping along with the one, two, and
three mode approximations. (a) 1st shaft (b) 2nd shaft (c) Mesh deflection
(d) Mesh force. Torsional deflection of the first disk ( ) and flexural
deflection at the disk center of mass () are shown for (a) and (b). . . . . . 58

xv
3.1 (a) Model of the coupled spinning gear-shaft system. The dotted line is
parallel to the line of action. (b) Schematic to visualize the torsional mesh
stiffness. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

3.2 Helical gear translational mesh deflection. (a) Axial deflection a (AB)
due to tilting of the gears about the z1 axis. (b) Components of the total
translational mesh deflection. . . . . . . . . . . . . . . . . . . . . . . . . . 64

3.3 Force on the first gear due to the loaded static transmission error. . . . . . . 68

3.4 Full scale model of the coupled spinning gear-shaft system. . . . . . . . . . 81

3.5 Natural frequency (dimensionless) sensitivity to the rotation speed () for

2 to 12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

3.6 Dimensionless real parts of the eigenvlues vs the rotation speed () for 2
to 12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

3.7 Dimensionless natural frequency and the normalized modal bending and
mesh energy variations of the modes 8 and 9 with the dimensionless rota-
tion speed . (a) Natural frequencies 8 and 9 (zoomed in view from Fig.
3.5) (b) Bending energies of mode 8 (c) Bending energies of mode 9 (d)
Mesh energies of modes 8 and 9. . . . . . . . . . . . . . . . . . . . . . . . 83

3.8 Natural frequency (dimensionless) sensitivity of 6 to the mesh stiffness

(km ) (N/m) in log scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

3.9 Normalized dimensionless modal energy variations of the mode 6 at vari-

ous dimensionless rotational speeds () with the mesh stiffness (km ) (N/m)
in log scale. (a) Bending energies of both the shafts (b) Mesh and torsional
mesh energies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

3.10 Natural frequency (dimensionless) sensitivity of 6 to the torsional mesh

stiffness (KT ) (Nm/rad) in log scale at different K8 . (a) K8 = 1.0 (b) K8 =
10.0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

3.11 Dimensionless normalized modal bending and mesh energies of mode 6 at

various dimensionless rotational speeds () with the torsional mesh stiff-
ness (KT ) (Nm/rad) in log scale. Two different flexural rigidities of the 2nd
shaft are chosen. (a),(b) K8 = 1.0 (c),(d) K8 = 10.0 . . . . . . . . . . . . . 88

xvi
3.12 Dimensionless force response of the shafts (dB) due to the loaded static
transmission error (with unit amplitude) at the gear mesh. The maximum
deflections of the centroid of the disks are shown. The damping ratio () =
4%. (a),(b) Flexural response of the 1st shaft perpendicular to and in plane
of the longitudinal spring. (c),(d) Flexural response of the 2nd shaft in the
same two planes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

3.13 Dimensionless force response at the mesh due to the loaded static transmis-
sion error (with unit amplitude) at the gear mesh. The damping ratio () =
4%. (a) Mesh deflection (b) Torsional mesh deflection. . . . . . . . . . . . 91

3.14 Dimensionless normalized significant modal energy variations of the modes

5 and 6 vs. the dimensionless rotation speed (). (a) Mode 5 (b) Mode 6. . 91

4.1 (a) Model of the coupled belt-pulley-shaft system (b) Pulley tilting causing
torsional belt deflection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

4.2 Model of the coupled belt-pulley-shaft system subject to a sinusoidal force

on the first shaft in the coupled direction. . . . . . . . . . . . . . . . . . . . 108

4.3 Dimensionless natural frequency () vs. the rotational speed (). . . . . . . 110

4.4 Dimensionless normalized significant modal energy variations of the modes

3 to 6 vs. the dimensionless rotational speed (). . . . . . . . . . . . . . . 111

4.5 Dimensionless force response of the shafts (dB) due to a cyclic load of
unity magnitude on the first shaft in the coupled plane. The maximum
deflections of the centroids of the disks are shown. The damping ratio
() = 4%. (a),(b) Flexural response of the 1st shaft in the uncoupled and
coupled planes. (c),(d) Flexural response of the 2nd shaft in the uncoupled
and coupled planes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

4.6 Dimensionless forced response of the shafts (dB) due to a cyclic load of
unity magnitude on the first shaft in the coupled plane. The maximum
torsional deflections at the disk ends are shown. The damping ratio () =
4%. (a) Torsion (1st shaft) (b) Torsion (2nd shaft). . . . . . . . . . . . . . . 113

4.7 Dimensionless forced response of the longitudinal and torsional belt springs
due to a cyclic load of unity magnitude on the first shaft in the coupled
plane. The maximum deflections are shown. The damping ratio () = 4%.
(a) Longitudinal belt spring (b) Torsional belt spring responses. . . . . . . . 113

xvii
4.8 Dimensionless normalized modal energy variations of the modes 2 vs. the
dimensionless rotational speed (). . . . . . . . . . . . . . . . . . . . . . . 114

5.1 Model of the gear-shaft system with bearing flexibility. (a) Spur gear sys-
tem (b) Helical gear system. . . . . . . . . . . . . . . . . . . . . . . . . . 120

5.2 Flexible gear body (modified from [1]). . . . . . . . . . . . . . . . . . . . 122

xviii
 CHAPTER 1

INTRODUCTION

1.1 Objectives and Motivation

A basic rotor system consists of a rotating structure or rotor supported by bearings and

finds application ranging from steam turbines and jet engines to computer disk drives. The

supporting structure is called the stator and consists of bearings and other stationary parts.

An interesting phenomena of this kind of system is that the natural frequency and amplitude

of vibration is often dependent on the rotation speed. As the speed of rotation is increased,

the amplitude of vibration often passes through a maximum value. This speed is known as

the critical speed. Research on rotor dynamics in the early part of the twentieth century was

mainly experimental and concerned with finding the critical speed and stabilizing the rotor

near this speed. The earlier misconception was that a rotor cannot be operated beyond the

critical speed due to shaft failure, unless the evidence of the existence of a second critical

speed appeared. This fueled the need for theoretical research on rotor dynamics by the

Royal Society of London. After WWII, rotor dynamics became an international endeavor

when many variation of the basic model (Fig. 1.1) emerged to satisfy particular applications

with different modeling and solution techniques. For example, as the distinction between

disk and shaft is blurred in a typical aircraft gas turbine, a more general modeling technique

1
known as the Transfer Matrix Method (TMM) developed [24]. It is still a method of choice

for most industrial rotor dynamic analyses.

G
S

Figure 1.1: Jeffcott rotor.

Other than the critical speed, a general rotating machine can develop inertia effects

that can be analyzed to improve design and decrease the possibility of failure. Inertia

effects in rotating structures are usually caused by gyroscopic acceleration introduced by

taking into consideration the precise motions of the vibrating rotor as it spins. As spin

velocity increases, this acceleration of the rotor becomes critical. The system can be stable

at supercritical speeds due to the effect of gyroscopic acceleration. Not accounting for

inertia effects at the design level can lead to bearing and support structure damage during

operation.

Gear drive is one of the most popular power transmission devices for a definite velocity

ratio, making it suitable for being used even in precision equipment such as watches. Driver

and follower are in close proximity in gear drives, making it one of the most compact

drives. Gears can be divided according to the tooth profile (e.g., involute and cycloid gears),

type of cut of the tooth (e.g., spur, helical, bevel, hypoid, worm, etc.), arrangement of the

gears (e.g., internal, external, planetary, etc.). Spur gears are the simplest and the most

2
common types of gears with the tooth aligned to the axis of rotation, easy to manufacture

and fitted on parallel shafts. Impact loading on the tooth and noise is an issue in spur gears.

Helical gear is a refinement over the spur gears with the leading edges of the tooth always

at a constant angle known as the helix angle with the axis of the gear. Helical gears can

be engaged in parallel or crossed orientation. As the contact between the engaging teeth

is gradual, helical gears operate more smoothly and quietly than spur gears. Hence, the

operating speed for the helical gears can be high.

T2
T1 e(t)
cm
~

I1 km I2

Figure 1.2: Gear pair dynamic model.

Modeling of the gears in mesh for theoretical and experimental analyses started in

1950s. Various linear and nonlinear models are developed [58] for considering the ef-

fects of the time varying mesh stiffness and tooth errors contributing to the dynamic tooth

load. Static transmission error in a gear pair is defined as the difference in the angular

positions of the driven gear and the angular position where it would be if the gears had

perfect conjugate action with no error [9]. Most single degree of freedom models [10]

applicable to spur, helical and spiral bevel gears use the loaded static transmission error

as an input rather than individual gear errors. The gear pair here is modeled as two disks

representing the inertia of the gears and the spring and the dashpot representing the mesh

3
stiffness and damping, respectively (Fig. 1.2). The stiffness of the spring is equal to the

mean mesh stiffness over a loading cycle. The static transmission error is represented as

an additional displacement excitation at the gear mesh. This model is revolutionary in the

sense that the analytical solution for more complicated gear systems seem possible with

this simplification.

In the late 1960s and early 1970s, it was realized that dynamic models with shaft and

bearing masses and flexibilities were necessary for more general purpose designs. If the

stiffness of these elements are comparable to the mean mesh stiffness, the coupled vibration

among different elements is non-negligible. These models are either torsional models in

which the torsional flexibilities of the attached shafts are considered, or flexural-torsional

models in which both the flexural and torsional compliances are taken into account. Geared

rotor systems are modeled as two disk-rotor systems connected by a spring representing the

compliance of the gear teeth. The rigid disks represent inertia of the gears. Work is also

done on simplified models assuming that the coupling between the torsional vibration and

flexural vibration is in the power transmitting direction [1114]. Pioneering models are

developed to study the whirling of the gear carrying shafts instead of only the dynamics of

the gears. The constraint imposed by the gear mesh on the pair of rotors result in coupled

vibration. Coupling is present between the rotors as well as the flexural, torsional, and mesh

modes of vibrations and result in significant changes in natural frequency and predicted

responses [1517].

Numerical techniques using lumped parameter models have usually been employed to

solve geared rotor systems. Two common methods - (a) Holzer method for torsional vi-

brations and (b) Myklestad-Prohl method for lateral vibrations are combined to get the

system matrices. The coupling due to mesh deflection is included as an additional coupling

4
term [18, 19] in the system matrices. For reliable dynamic analysis, the type of model that

should be used depends on the objective of the study. Lumped parameter model simpli-

fies the system but the fidelity of the model is lost, especially to predict the higher natural

frequencies. To accurately predict the dynamic behavior of the system at higher frequen-

cies, both the continuous vibrations of the shafts and the discrete vibrations of the spring

and the attached disks need to be considered in a model. Also, calculation of the system

responses to an arbitrary excitation is not straightforward for lumped parameter models.

Finite element models [20, 21] are more accurate than lumped parameter models but are

computationally exhaustive and sometimes, the physical insight into a dynamic problem

is not realized. Thus, a comprehensive dynamic model is preferable that will preserve the

characteristics of different components of the system as it is.

As the rotation speed of helical gears can be high, the gears and the continuous shafts

are subject to coriolis acceleration making the system as a gyroscopic system. Mathe-

matical complexities associated with the hybrid discrete-continuous gyroscopic systems

prevent them from being studied in literature. Response calculation of gyroscopic systems

using finite element methods is cumbersome. Analytical formulation for eigensolutions

and response analysis of such systems demand novel mathematical approaches. Two dif-

ferent models of geared rotor systems are studied - (a) non-gyroscopic spur geared rotor

system (Fig. 1.3(a)) and (b) gyroscopic helical geared rotor system (Fig. 1.3(b)). Both

of these models have individual challenges that need to be overcome in the mathemati-

cal models. The non-gyroscopic model has the gears located on the shaft in between the

bearings. Hence, the matching conditions for flexure and torsion at the gear-shaft inter-

faces pose additional boundary conditions in the formulation. Response calculation for the

gyroscopic helical geared shaft model needs an additional transformation of the system

5
equations. Development of novel mathematical approaches for both non-gyroscopic and

gyroscopic geared rotor systems is an objective of the present research.

Gear1

Shaft1

Gear 1

Shaft2
Shaft 1
Gear 2

Shaft 2
Gear2
(a) (b)

Figure 1.3: Two models of geared rotor system (a) Non-gyroscopic spur (b) Gyroscopic
helical geared shaft systems.

Belt drives are employed as useful means for transmission of power and are found in

various applications such as conveyors, machine tools, stationery or mobile powered ro-

tating equipment. The main advantages of belt drives over other drives are (a) flexible

locations of the driver and driven shafts, (b) lower cost of installation and maintenance, (c)

quieter operation, etc. Until late 1970s, V-belts were exclusively used in automotive acces-

sory drive systems. However, life-spans of V-belts are small and maintaining proper tension

throughout the belt life is difficult. Nowadays, serpentine belt drives have replaced V-belt

drives in automobiles and heavy vehicles for driving the accessories such as alternator, air

conditioner, water pump, etc. by the engine power delivered from the crankshaft (Fig. 1.4).

6
They are advantageous over the conventional V-belt drives in respect of longevity, com-

pactness, simplified assembly, and proper maintenance of the belt tension throughout belt

life by the use of a tensioner. Despite these advantages, serpentine drives suffer from noise

and belt tension fluctuation, which find their roots in the system vibration. Tension fluctu-

ation affects the life expectancy of the belts. The crankshaft of a car typically rotates at v

6000 rpm. For a racing car the speed can go up to v 18000 rpm. Hence, tension fluctuation

of the belt can be a major issue in these applications.

Figure 1.4: Accessory drive system of an automobile.

The shafts attached to the pulleys in the accessory drive are generally considered to

be rigid. But at high speeds, flexibility of the shafts cannot be ignored. Vibration of the

pulleys carrying the belt can be mainly attributed to the excitation coming from the attached

shafts. Pulley vibration ultimately affects the belt vibration and its operation. Serpentine

belts are different in structure compared to the V-belts in the sense that the belt thickness

7
is small. Hence, the bending stiffness in the transverse direction can be neglected and the

belt can be modeled as a spring. In practical applications, belts also experience torsional

vibration about the axis in the direction of belt motion, particularly when they are wide

and thin. The system can be considered as a discrete-continuous system where each shaft

forms continuous part having flexural and torsional vibrations. The pulleys with the belt

span constitute the discrete part, which are connected by a belt, modeled as a combination

of longitudinal and torsional spring.

Alternator

Idler

Tensioner

Water Pump

Power
Steering
Air
Conditioner

Crankshaft

Figure 1.5: Schematic of a serpentine belt drive.

Until now, vibrations of the shafts and their effects on the longitudinal vibration of the

moving belt has not been investigated. The system can be a complicated one if the gyro-

scopic effects of the continuous shafts and the discrete pulleys are taken into consideration.

In automotive accessory drive system, the number of pulleys and the corresponding belt

8
segments is at least seven (Fig. 1.5). Analytical eigensolution of the whole system would

be interesting, but would involve complications in terms of the number of coupled equa-

tions. As a first attempt, effect of vibration of the shafts with the adjacent pulleys on the

vibration of the connecting belt span, held in the reverse wrapped configuration (Fig. 1.6),

is determined. The excitation force is considered as a sinusoidal force acting on one of the

adjacent shafts. The response of the coupled system is considered mainly in the form of

belt stretching.

Pulley 1

Shaft 1

Belt

Shaft 2

Pulley 2

Figure 1.6: Belt-pulley-shaft model for single span of the belt.

1.2 Literature Review

In this section, a brief description of the related work is presented. The review is di-

vided into three parts. The first part focuses on the discussion of spinning or non-spinning

flexible rotor systems. The second part addresses the development in the area of geared

shaft systems. The third part reviews the work on belt drives.

9
1.2.1 Rotor systems

Earlier works on disk rotor systems employ the method of characteristic equation. Sri-

nath and Das [22] is the first to analyze the vibration of a non-rotating, simply supported

beam carrying a mass with rotary inertia placed at an arbitrary location on the span. For

simplification, the beam along with the attached mass is divided into two spans such that the

mass and the moment of inertia of the attached mass are conserved. The governing equation

for each span is written and the matching conditions are employed to solve the eigenvalues

using the method of characteristic equation. Critical speeds for a spinning shaft-disk system

using a similar procedure is solved by Eshleman and Eubanks [23]. Assuming continuous

shaft, separate frequency equations are obtained for forward and backward whirls and the

first four critical speeds are determined for different disk locations. Gyroscopic effect of

the disk is considered while that of the shafts is ignored. In the above works, the disk is

assumed to be rigid. The influence of disk flexibility on the bending natural frequencies

and critical speeds of an axisymmetric rotating shaft-disk system is investigated by Chivens

and Nelson [24]. Even though, the natural frequencies and critical speeds are determined

with respect to the rotating coordinate system, a detailed discussion on natural frequencies

in the stationary and rotating coordinate systems and their relation, whirl ratio, forward and

backward whirls, etc., are presented in this paper. Comparisons of the results are made

with that of [22, 23] in the limiting case of a rigid disk for non-rotating and rotating cases

and a good agreement is found. It is observed that natural frequencies may be significantly

affected by the disk flexibility whereas critical speeds are not.

Characteristic equation method works for simpler systems but fails for complicated

systems. Several different methods exist in literature for vibration and response analysis of

shaft-disk systems. Transfer matrix method is useful for systems involving several disks on

10
a single shaft as the matrices have smaller dimension than the conventional finite element

methods [4, 25, 26]. Coupling between the torsional and bending deformations of a single

shaft is shown by Nataraj [27]. The displacements are expressed as a perturbation series in

the nonlinear analysis. A simply supported uniform circular shaft is chosen as an example.

It is shown that torsional vibration at twice the rotation speed is caused by flexural vibration.

Natural frequencies and mode shapes of a spinning disk-spindle assembly consisting of

multiple elastic circular plates mounted on a short rigid shaft, as applicable to computer

disk drives, is examined by Shen and Ku [28]. As the boundary conditions are not simple,

a new approach is developed where the kinetic and strain energies are discretized and the

discretized equations of motion are obtained using Lagranges equations. Lee and Chun

[29] developed assumed modes method to investigate the effect of multiple flexible disks

on the vibration modes of a flexible rotor of varying annular cross section.

Parker and Mote [30] solve the eigenvalue problem for a non-spinning coupled disk-

spindle-clamp vibration with the use of extended operator formulation. The asymmetries

are introduced at the inner boundary of the disk in the form of shape deviation and, assum-

ing the clamp to be attached to circumferentially varying linear and rotational springs. It

is observed that for the axisymmetric system, only one nodal diameter disk mode couple

with the spindle mode. For the asymmetric system, coupling occurs solely due to disk

asymmetry. Later, Parker et al. extend this work to a spinning asymmetric (due to mass

unbalance) [31] and axisymmetric [32] disk-spindle systems. Galerkins method is used

to discretize the continuous system equations that comes as a natural extension of the ex-

tended operator formulation.

Finite element method is useful for practical rotor systems modeled as higher order

beams or having complex geometries. Nelson and McVaugh [33] develop a finite element

11
model to include the effects of rotary inertia, gyroscopic moments, and axial load of a rotor

bearing system with rigid disks. Equations of motion of the element are presented in both

the fixed and the rotating reference frames. This model is still used as a demonstrative

model in many industrial applications. Finite element simulation of a multi-bearing rotor

system including the effects of shear deformation and internal damping in addition to rotary

inertia, gyroscopic moments and axial force is presented by Ozguven and Ozkan [34]. A

shaft element model adding torsional deformation is developed by Qin and Mao [35] for

accurate estimation of the dynamic behavior of the coupled transverse-torsional motion that

exists in large power systems.

1.2.2 Geared shaft systems

Coupled vibration analysis of geared shaft system is comparatively new. Iida et al.

[11] derive the natural frequencies and modes of a pair of spur geared rotors considering

torsional-flexural vibration. The driving shaft is considered to be rigid in bending, while

the driven shaft is taken as flexible. The flexural deflections are distinguished between

the coupled and uncoupled directions. With the shafts modeled as lumped parameters,

characteristic equation method is applied to the system after omitting the vibration in the

uncoupled direction. Steady state response to geometrical eccentricity and mass unbalance

is calculated using first order perturbation of the parameters. Iida et al. also determine the

dynamic characteristics of a countershaft in a three-shaft gear train system with [14] and

without [13] considering the bearing stiffness. Coupled dynamics is observed when the

spur gears on the countershaft have different power transmitting directions. The natural

frequencies and mode shapes are affected by the angle between the two directions.

12
 In the aforementioned works, only one of the shafts is flexible in bending. Taking both

the shafts as compliant in torsional, flexural, and axial directions, David and Mitchell [15]

show that coupling terms play an important role in the correct response prediction. The

shafts are modeled as massless while only the linear coupling terms are included. The

study indicate that inclusion of the coupling terms corresponding to the gears produce sig-

nificant changes (up to eight orders of magnitude) in the predicted response in a spur geared

rotor system. Effect of shaft inertia in gear dynamics using continuous shaft model is first

investigated by Sener and Ozguven [36]. Dynamic factor in a geared system is defined as

the ratio of maximum dynamic mesh force to the maximum static mesh force. Dynamic

factor is calculated using a simplified linear model where only the torsional deformations

of the shafts are considered and the method of characteristic equation is used to solve the

system equations. Results are compared with an equivalent four degree of freedom discrete

model where shaft masses are ignored and are shown to be in reasonable agreement with

previous nonlinear [10], experimental [37], and finite element [21] models. The approach

is presented as a useful alternative to the finite element model.

Non-linearities in gear systems appear due to backlash, teeth separation in mesh, back

side collision following tooth separation, etc. Ozguven [38] develop a six degree of free-

dom nonlinear geared shaft model including the effects of lateral-torsional compliances

of the shafts, time varying mesh stiffness and damping, material damping of the shafts,

transverse bearing compliances with damping, etc., with the non-linear effects at the gear

mesh. This is an extension of the previous single degree of freedom model [10] when cou-

pling between the mesh and other vibrations are unavoidable. Time varying mesh stiffness

with non-linearities produces Mathieus equation, which is solved by a numerical itera-

tion procedure. A new method of analysis, e.g., the discrete time transfer matrix method

13
combining the advantages of the transfer matrix method and the numerical integration pro-

cedure is developed by Kumar and Sankar [39] for the dynamic response analysis of large

dynamic systems. Application is shown for a single stage spur geared rotor system.

Finite element formulation can be useful for non-gyroscopic geared rotor systems. It

can be cumbersome for response calculation of gyroscopic models. Neriya et al. [20] obtain

a simplified forty-one degrees of freedom finite element model including the degree of free-

dom of the contact point of the mating gear teeth. The coupling action between torsion and

flexure is introduced in the appropriate locations of the global stiffness matrix. Assuming

constant mesh stiffness, the natural frequencies and responses to the mass unbalance and

the geometric eccentricity for the linear system are obtained. Kahraman et al. [21] develop

finite element model of a geared rotor system to study the effect of bearing flexibility. The

general purpose rotor dynamics program developed by Ozguven and Ozkan [34] is used

for the analysis. This model takes rotary inertia, axial loading, flexibility and damping of

the bearings, material damping of the shafts, etc., into account. To simplify the problem,

the gyroscopic moment effects are ignored and internal damping of the shafts is included in

the damping matrix. Excitations in the form of static transmission error, mass unbalance,

and gear runouts are considered and the natural frequencies and responses are shown to

be in good agreement with the previous experimental [11] and theoretical [20] findings.

It is concluded that when bearings are stiff their effects on the natural frequencies can be

ignored. Lee et al. [40] use coupled lateral and torsional vibration FE model of a gear pair

to include gyroscopic and damping effects. It is observed that a given dominant mode may

change from an initial torsional mode to a lateral mode and vice versa.

Transfer matrix methods have been extensively used for the analysis of rotor bearing

systems. Choi and Mau [19] extend this method to a geared rotor system of a previous

14
work [21] by adding a subsystem of gear mesh. Rotating shafts are modeled as Timoshenko

beams with the effects of shear deformation and gyroscopic moment taken into account.

The gear mesh stiffness is considered to be time variant and approximated by a truncated

Fourier series at multiples of tooth passing frequency. Natural frequencies and mode shapes

are shown to be in good agreement with [21]. Campbell diagrams, critical speeds, and

steady state responses due to mass unbalance, geometric eccentricity, and transmission

error are obtained.

1.2.3 Belt drives

In serpentine belt drives, with a single belt, the engine power gets distributed to various

accessories. The system can generally be treated as linear or nonlinear. Considerable re-

search on serpentine belt drives have been done on both the areas of study. The nonlinearity

appears due to the large rotational motion associated with the tensioner arm, which comes

as transcendental terms in the system equations. In addition, research have been done on

models that consider only the pulley rigid body rotational vibration, considering the belt

spans as longitudinal (axial) springs. This results in a discrete system. There are models

that take the transverse vibration of the belt into account as well resulting in a discrete

continuous hybrid system.

In one of the early works, Hawker [41] develop a linearized model of a serpentine belt

drive to consider the rotational vibration of the pulleys along with longitudinal deflection

of the belt for free vibration and forced response analyses. The belt is modeled as an

extensional spring. It is identified that the angular oscillations of an accessory drive result

from the extension and relaxation of the belt strand connecting the pulleys. The onset of belt

slip is predicted by Barker et al. [42] in the analysis of the transient response of the previous

15
model. Barker and Yang [43] approximate the belt in free spans as massless linear springs

with viscous dashpots and assume that the axial deformations are more significant than

transverse or torsional vibrations. The system equations are considered as linear equations

when the tensioner arm is fixed.

Hwang et al. [44] develop a nonlinear model to find the longitudinal response of the belt

and the rotational response of the crankshaft and the accessory pulleys. The key assump-

tions are: (a) the belt does not slip over the pulleys and (b) the lateral and longitudinal belt

responses are decoupled, so that the belt can be modeled as springs. Nonlinearities come

due to finite rotation of the tensioner arm. The equations of motion are linearized about the

equilibrium position of the tensioner to obtain the natural frequencies and mode shapes.

Numerical solutions of the nonlinear equations of motion indicate that under certain engine

operating conditions, the dynamic tension fluctuations may be sufficient to cause the belt

to slip on a particular accessory pulley. Leamy and Perkins [45] make similar assumptions

to obtain the rotational response of the entire serpentine belt drive. Nonlinearities in this

work appear due to coulomb friction at the tensioner hub in addition to the large tensioner

arm motion. Beikmann et al. [46, 47] consider coupling of the transverse and rotational

vibrations in the classical three pulley serpentine drive system. The transverse vibration is

similar to that of a vibrating string. Both linear and nonlinear models are considered in this

undamped system. Nonlinearities arise due to finite stretching of the belt spans. The linear

model identifies the coupling between the rotational pulley vibration and transverse belt

vibration adjacent to the tensioner arm. The nonlinear analysis uses the eigensolutions of

the linear model to obtain the discretized nonlinear model and evaluate the coupled vibra-

tion response. The results indicate that large transverse belt response results from internal

resonance. Damping due to viscous belt span and coulomb damping at the tensioner arm

16
are included in the nonlinear analysis by Kraver et al. [48]. A complex modal method is

used to bypass numerical iteration, which is time consuming.

Kong and Parker [49] use Beikmanns three pulley model and add transverse flexural

vibration of the span away from the tensioner in between the fixed pulleys. The exis-

tence of vibration coupling between span transverse vibration and pulley rotation is shown.

Abrate [50] compiles the existing literature associated with the moving belt and classifies

them according to transverse, torsional, and axial models. Coupling between rotational and

transverse vibrations are neglected in most of the works. It is stated that depending on the

combination of bending rigidity, initial tension, and span length, a belt span can be modeled

as a moving string or a beam.

1.3 Scope of the Present Work

The current research addresses analytical modeling for the linear vibration analysis

of single-mesh geared rotor systems. It also addresses the effect of shaft vibration on

the dynamics of belt-pulley drive by considering a single span of a belt in a serpentine

belt drive system. It aims to establish validated mathematical models for two possible

configurations of a geared rotor system. Two efficient and well known analysis methods are

eventually applied to find the dynamic characteristics of these models. These analyses are

of practical importance in reducing noise level in industrial power transmission operations.

An important aspect of this research is the incorporation of shaft flexural and torsional

vibrations into the existing gear and belt drive models. The thesis is organized as follows:

Chapter 1 provides the necessary background and literature review.

Chapter 2 establishes the non-gyroscopic mathematical model of a spur gear pair on

parallel shafts where matching conditions at the gear-shaft interfaces impose additional

17
geometric constraints. The linearized system equations are derived and their conservative

natures are verified with the use of extended operators, by proving that they are self-adjoint.

An efficient energy-based method known as the assumed modes method is used to incorpo-

rate the matching constraints with Lagrange multipliers, discretize the discrete continuous

system, and determine the system matrices. Discretization is achieved with the use of basis

functions that are global in nature. The coefficients of these bases serve as generalized

coordinates in the assumed modes method. Additional transformation is employed to ob-

tain the classical form of the vibration equation involving only the mass and the stiffness

matrices. The natural frequencies and the mode shapes are obtained thereafter. Interesting

dynamic behavior is discovered for coupled modes. Sensitivities of the natural frequencies

and the normalized modal energies to different system parameters are determined and cor-

related for these modes. The response due to the loaded static transmission error at the gear

mesh is obtained and the resonance at different excitation frequencies are explained with

the mode shapes of the respective modes.

Chapter 3 derives the kinematic model of a helical gear pair in mesh, which are mounted

at the end of parallel compliant shafts. The mesh deflections of the longitudinal and tor-

sional springs at the gear mesh is obtained and used in the strain energy expression. The

kinetic energy expression includes the rotational speeds of the shafts. Hamiltons princi-

ple is employed to obtain the linearized system equations from these energy expressions.

The forcing terms due to the loaded static transmission error are obtained by including an

additional displacement term in the expression for the strain energy. An extended opera-

tor formulation is employed to write the system equations in a compact analytical form,

excluding the geometric boundary conditions. This operator form is helpful in establish-

ing a canonical gyroscopic system structure to these equations involving mass, gyroscopic,

18
stiffness, and rotational stiffness operators. It is also helpful in analytically proving the self-

adjointness or skew self-adjointness of these operators. To discretize the system equations,

Galerkins method is used. This comes as a natural extension of the extended operator

form as the global basis functions are written as an extended variable. These basis func-

tions come from the same family as the previous model and satisfy the geometric boundary

conditions. Modal analysis is performed on the gyroscopic system after transforming the

discretized system into the first order form. The response due to the loaded static transmis-

sion error is obtained using the reduced system from the modal analysis. Natural frequency

sensitivity to different system parameters are also obtained.

The analysis technique developed in Chapter 3 is applied to a single span of belt of

a serpentine belt drive system with the attached pulleys mounted on compliant shafts in

Chapter 4. The dynamic model of the belt is established as a combination of translational

and torsional springs attached to rigid pulleys. The governing equations and the equations

of motion are derived, Galerkins method is employed to discretize system equations, and

the modal analysis is performed, similar to Chapter 3. The excitation to the system is

considered as a periodic force pulsation on one of the shafts in the coupled plane. Response

mainly in the form of belt tension variation is determined.

Contributions of the thesis:

1. The methodology developed for the linear dynamic analysis of the non-gyroscopic

geared rotor system is suitable for the dynamic analysis of similar systems with non-

classical constraints. Knowledge of natural frequency sensitivity and response char-

acteristics is essential for the design and troubleshooting under operating conditions.

Natural frequency veering is useful to trace any dramatic changes in vibration modes

under variations of parameters.

19
2. Study of the gyroscopic effect in a geared shaft system is important for high speed

applications and is a natural extension of the non-gyroscopic model. Inclusion of

helical gears in the model provides larger number of design parameters. The method

developed for analytical eigensolution calculation is suitable for gyroscopic systems

with classical boundary conditions. Modal analysis of the present system can be fol-

lowed in the response calculation of any discretized gyroscopic system. Phenomena

such as mode splitting, critical speeds, forward and backward whirls, etc., explain

many interesting and crucial behaviors of high speed geared rotor systems and hence

can be used as a design and analysis guideline.

3. The work on vibration analysis of belt-pulley-shaft model establishes the influence

of shaft flexibility on the dynamics of high speed serpentine belt drives. Campbell

diagram shows the gyroscopic system characteristics including mode splitting and

critical speeds. Excitation to this system is considered to be a force excitation on

the crankshaft, which is practical for an accessory drive. Correlation of resonance

frequency with the modal frequency serves as a practical analysis guide.

20
 CHAPTER 2

EIGENSOLUTION OF NONGYROSCOPIC SPUR GEARED

SYSTEM ATTACHED TO COMPLIANT SHAFTS

In this chapter, a combination of variational calculus and an energy method is applied

for the linear vibration analysis of a pair of coupled gear-shaft system. The hybrid discrete-

continuous system consists of the gear mesh modeled as the discrete element and the at-

tached shafts modeled as the continuous elements. Location of the gears on the shafts as

opposed to the ends of the shafts imposes additional constraint equations other than the

boundary conditions, which makes the method developed in this chapter suitable for ana-

lyzing systems with similar constraints.

2.1 Introduction

For power generation and transmission, rotating machineries are extensively used. Geared

units are frequently used in these machineries, the purpose being to connect different

branches for transmission of power, variation of speed, and change of direction of work. In

recent years, extensive research has been done on the vibration induced by the gear mesh.

Gears mounted on slowly rotating short shafts are assumed to be rigidly mounted laterally,

21
making the gear mesh the only source of vibration and noise. The same assumption is in-

valid for gears mounted on long, slender shafts, which fall under the class of geared rotor

dynamics problems.

A number of different methods are available in literature to solve the vibration analysis

problems of a flexible and/or rigid shaft carrying flexible or rigid single or multiple disks.

Lumped parameter approximation of the continuous system to obtain the overall transfer

matrix is mainly employed in the previous methods [25, 39]. Lumped parameter approach

is inevitable when dealing with a complicated system, but at the cost is accuracy, as the

model is not a representative model of the true physical system. Solution of continuous

system equations of a non-rotating shaft carrying a rigid disk at an arbitrary location along

the span is first obtained by Srinath and Das [22]. For simplicity of the analysis, both

the span and the disk are divided such that the total mass and the moment of inertia of

the disk are conserved. Gyroscopic coupling effect due to the disk on the critical speed

of a spinning shaft-disk system is analyzed by Eshleman and Eubanks [23]. Characteristic

equation method in some form is used in these works and disk thickness is ignored. Chivens

and Nelson [24] determine the influence of disk flexibility on the natural frequency of

bending and critical speeds of a rotating disk-shaft system. Nonlinear dynamic analysis

of a rotating shaft to show the interaction between the torsional and flexural vibrations is

developed by Nataraj [27]. Extended operator formulation to solve the eigenvalue problem

of a non-spinning asymmetric [30] and spinning [31, 32] disk-spindle system is developed

by Parker et al.. Shen and Ku [28] examine the coupled vibration of multiple flexible disks

and the rocking motion of the attached short, rotating, rigid shaft. Lee and Chun [29] use

assumed modes method to find the effect of disk flexibility on the vibration modes of the

rotating flexible shaft carrying multiple flexible disks.

22
 Study of geared spindle systems is quite recent. This system is usually modeled as

two disk-spindle systems with a gear mesh coupling them. For most of the cases, only

torsional vibrations of the shafts are considered, while the disks are considered to be rigid.

For more compliant shafts, the lateral vibrations due to flexure are also included. For

a general rotor system with length-to-diameter ratio less than 100, coupling between the

flexural and torsional vibrations may be neglected. For a geared disk-spindle system, due

to the gear meshing effect, this coupling phenomenon may arise for a length-to-diameter

ratio well below 100. Analytical solution for the coupled flexural and torsional vibration

characteristics of a geared rotor system is rare. Iida [11] was the first to include both the

flexural and torsional vibrations in a pair of spur geared rotors. Importance of flexure for

the high speed rotating shaft is identified even for lumped parameter approximation of the

shafts where inertias of the shafts are ignored. Application is also sought in geared train

of rotors [1214]. Importance of dynamic coupling terms in the response predictions is

shown by David and Mitchell [15]. Effect of shaft inertia in gear dynamics for a geared

shaft system using linear continuous torsional shaft model is first studied by Sener and

Ozguven [36]. Ozguven [38] conducts dynamic analysis of a six-degree-of-freedom non-

linear model consisting of a spur gear pair with time varying mesh stiffness and backlash.

Transverse vibration of the gears perpendicular to the direction of the line of action is

neglected. Shaft dynamics is neglected while dynamics of the bearings is considered in his

model.

Finite element methods can be useful to obtain the natural modes and calculate the

responses of simplified geared shaft model developed by Neriya et al. [20] or more compli-

cated model developed by Kahraman et al. [21]. Lee et al. [40] examine both the coupled

and uncoupled flexural and torsional vibration characteristics of a geared rotor bearing

23
system by finite element method. They systematically show that some modes yield cou-

pled phenomena when the gear mesh stiffness is higher. While finite element methods are

straightforward, the physical insight into a dynamic problem is not always realized. Trans-

fer matrix method is extended from the domain of rotor dynamics to the scope of geared

rotor systems by Choi and Mau [19].

In all the above works, it is concluded that it is important to consider the flexural and/or

the combined flexural and torsional vibration of the attached shafts for the correct predic-

tion of the natural frequencies of a geared rotor system. For certain parameter values, the

gear mesh vibration can couple with the flexural-torsional vibration of the attached shafts

producing unwanted noise.

Current work aims at analytical eigensolution of a geared rotor system subject to flex-

ural and torsional vibrations. The meshing gears are spur gears. The aim is to find the

eigensensitivity to various system parameters as well as effect of gear mesh on the coupled

flexural and torsional vibration.

Galerkin discretization technique is applied to systems where the basis functions sat-

isfying all the geometric boundary conditions are readily available. An example of such

a system is two shafts with meshing gears attached at their ends. Basis functions satisfy-

ing the geometric boundary conditions at the bearing ends of the shafts are easy to obtain

in this case. When gears having finite thickness are located on the shafts, certain addi-

tional geometric boundary conditions (matching conditions) come into picture, resulting in

difficulties in obtaining suitable basis functions for the various displacements.

The aim of this chapter is to overcome this limitation of the Galerkins method. A

different kind of method known as the assumed modes method is applied to calculate the

24
natural frequencies, vibration modes, and dynamic response due to the loaded static trans-

mission error at the gear mesh. The transverse and torsional displacements of the shafts

are represented by a series expansion of basis functions, which again are used to discretize

the kinetic and strain energies of the system. The sensitivities of the natural frequencies

to various system parameters are discussed qualitatively, which aid in designing and trou-

bleshooting by avoiding the resonant operating conditions. The sensitivities of the natural

frequencies are closely related to the various modal strain energies depending on the type

of the mode. In a typical geared transmission, the main source of excitation is the static

transmission error, which includes the effects of elasticity and shape deviation of the gear

teeth. For the loaded static transmission error, the deflections of the gear teeth contribute to

the total strain energy of the system. Response due to this loaded static transmission error

is determined.

2.2 Modeling and Equations

The coupled gear-shaft system consists of two units, each of which has an axisymmetric

gear mounted on a shaft. In the simplified model (Fig. 2.1), the disks are modeled as rigid

plates with uniform thickness. The flexibility of the gear teeth are taken into account by

considering the two disks connected together with a linear elastic spring along the line of

action. The z1 and z2 axes are parallel to the longitudinal direction of the spring. The

shafts are modeled as uniform Euler-Bernoulli beams having both flexural and torsional

deformations. The bearings are assumed to be rigid and provide fixed-fixed end conditions

for the shafts to transverse vibrations but add no constraints to the torsional vibration.

25
 L1
y1 a
h1

x1
z1 First disk
y2 First shaft

z2 x2 Second disk
b
Second shaft
L2 h2

Figure 2.1: Model of the coupled gear-shaft system.

The following non-dimensional variables are defined (See the Table 2.1 for definitions

of the symbols.)
s
E1 Isy1 L2 x1 x2 u1 u2 v1
t = t 4
, L1 = 1, L2 = , x1 = , x2 = , u1 = , u2 = , v1 = ,
s1 L1 L1 L1 L1 L1 L1 L1

v2 uc1 uc2 vc1 vc2 h1 h2 hc1

v2 = , uc1 = , uc2 = , vc1 = , vc2 = , h1 = , h2 = , hc1 = ,
L1 L1 L1 L1 L1 L1 L1 L1

hc2 R1 R2 a b m1 m2
hc2 = , R1 = , R2 = ,a = ,b = , m1 = , m2 = ,
L1 L1 L1 L1 L1 s1 L1 s1 L1

Icy1 Icy2 Icz1 Icz2 Im1 Im2

Icy1 = , Icy = , Icz = , Icz = , Im = , Im = ,
s1 L31 2
s1 L31 1
s1 L31 2
s1 L31 1
s1 L31 2
s1 L31

km G1 J1 G2 J2 J1 J2 s2
km = 3
, K 1 = , K 2 = , K 3 = 2
, K 4 = 2
, K5 = ,
E1 Isy1 /L1 E1 Isy1 E1 Isy1 s1 L1 s1 L1 s1

Isz1 E2 Isz2 E2 Isy2

K6 = , K7 = , K 8 = .
Isy1 E1 Isy1 E1 Isy1
26 (2.1)
Table 2.1: List of symbols defining spur geared shaft system. Subscripts 1 and 2 denote
the quantities belonging to the first and the second shafts, respectively. Overbar indicates
dimensional quantities, otherwise it is non-dimensional.

t Time
L1 , L2 Lengths of the shafts
s1 , s2 Mass per unit length of the shafts
u1 , u2 Flexural deflections of the shafts in the uncoupled plane
uc1 , uc2 Flexural deflections of the centers of mass of the disks
in the uncoupled plane
v1 , v2 Flexural deflections of the shafts in the coupled plane
vc1 , vc2 Flexural deflections of the centers of mass of the disks
in the coupled plane
1 , 2 Torsional deflections of the shafts about their respective axes
R1 , R2 Radii of the disks
h1 , h2 Thickness of the disks
hc1 , hc2 Distances of the centers of mass of the disks from
the left hand end of the disks
a, b Lengths of the left hand portions of the first and the second shafts
m1 , m2 Masses of the disks
Icy , Icz Mass moments of inertia of the disk with respect to the y and z axes
Im Mass moments of inertia of the disk with respect to the x axis
Isy , Isz Area moments of inertia of the shaft cross section with respect to
the y and z axes
J1 , J2 Polar moments of inertia of the shaft cross-sections
km Mesh stiffness
K1 , K2 Torsional rigidities of the shafts
K7 , K8 Flexural rigidities of the second shaft in the uncoupled and
the coupled planes

27
 The equations of motion as well as the governing equations are developed using Hamil-

tons principle. The non-dimensional kinetic energy

T
T =
E1 Isy1 /L1
a
Z1 Za Z1
1 v1 2 v1 2 u1 2 u1 2
Z
= ( ) dx1 + ( ) dx1 + ( ) dx1 + ( ) dx1
2 t t t t
0 a+h1 0 a+h1
b
ZL2 Zb ZL2
1 v2 2 v2 2 u2 2 u2 2
Z
+ K5 ( ) dx2 + ( ) dx2 + ( ) dx2 + ( ) dx2
2 t t t t
0 b+h2 0 b+h2
a
Z1
1 1 2 1 2
Z
+ K3 ( ) dx1 + ( ) dx1
2 t t
0 a+h1
b
ZL2
1 2 2 2 2
Z
+ K4 ( ) dx2 + ( ) dx2
2 t t
0 b+h2
1 vc1 2 1 uc1 2 1 vc2 2 1 uc2 2 1 1 2
+ m1 ( ) + m1 ( ) + m2 ( ) + m2 ( ) + Icy1 ( )
2 t 2 t 2 t 2 t 2 t
1 1 2 1 2 2 1 2 2 1 1 (a) 2 1 2 (b) 2
+ Icz1( ) + Icy2 ( ) + Icz2( ) + Im1 ( ) + Im2 ( ) .
2 t 2 t 2 t 2 t 2 t
(2.2)

u1 u2 v1 v2
1 (a), 2 (b) and 1 (a), 2 (b) are the slopes of the shaft
x1 x2 x1 x2
elements at the disk locations in the uncoupled and the coupled planes, respectively.

28
 The non-dimensional strain energy

V
V =
E1 Isy1 /L1
a
Z 2 Z1 2
1 v1 v1
= ( 2 )2 dx1 + ( 2 )2 dx1
2 x1 x1
0 a+h1
a
Z 2 Z1 2
1 u1 u1
+ K6 ( 2 )2 dx1 + ( 2 )2 dx1
2 x1 x1
0 a+h1
b
Z 2 ZL2 2
1 v2 2 v2 2
+ K8 ( 2 ) dx2 + ( 2 ) dx2
2 x2 x2
0 b+h2
b
Z 2 ZL2 2
1 u2 u2
+ K7 ( 2 )2 dx2 + ( 2 )2 dx2
2 x2 x2
0 b+h2
a
Z1
1 1 2 1 2
Z
+ K1 ( ) dx1 + ( ) dx1
2 x1 x1
0 a+h1
b
ZL2
1 2 2 2 2 1
Z
+ K2 ( ) dx2 + ( ) dx2 + km {R1 1 (a) + R2 2 (b) vc1 + vc2 }2 .
2 x2 x2 2
0 b+h2
(2.3)

The last term in the equation (2.3) is the strain energy stored in the mesh spring. The

disks are subject to elastic restoring forces and moments due to bending and torsion from

the shafts. In addition, a displacement excitation (approximation for the loaded static trans-

mission error) is applied at the location of the mesh spring. Static transmission error (STE)

in a gear pair is defined as the difference in the angular positions of the driven gear and the

angular position where it would be if the gears were perfect [9]. If static transmission error

occurs solely due to the teeth flexibility, it is defined as the loaded static transmission error.

For the loaded STE, a time dependent displacement excitation e(t) is applied at the mesh

spring having a constant time averaged mesh stiffness km . The modified strain energy in

29
the mesh spring is then

1
Vs = km {R1 1 (a) + R2 2 (b) vc1 + vc2 + e(t)}2 (2.4)
2

Using variational approach, the terms in Vs associated with e(t) are collected and the

coefficients of the variations of different variables are used as forcing terms in the equations

of motion. The forcing terms appear in the translational and rotational equations of motion

of the disks. As Vs = W for a conservative system, the virtual work done by the loaded

STE is given as
W = km e(t) {vc1 vc2 R1 1 (a) R2 2 (b)}
(2.5)
= {F1 vc1 + F2 vc2 + T1 1 (a) + T2 2 (b)}
Rt
Substituting all of the above quantities into the Hamiltons principle, (T V +
0
W )dt = 0, the non-dimensional governing equations for the continuous shafts and the

equations of motion of the discrete disks are obtained. The non-dimensional governing

equations for bending and torsion of the shafts are

4 v1 2 v1
+ 2 =0 (2.6)
x41 t
4 u1 2 u1
K6 + =0 (2.7)
x41 t2
2 1 2 1
K3 K 1 =0 (2.8)
t2 x21
4 v2 2 v2
K8 + K 5 =0 (2.9)
x42 t2
4 u2 2 u2
K7 + K 5 =0 (2.10)
x42 t2
2 2 2 2
K4 K 2 =0 (2.11)
t2 x22
Equations (2.6) (2.8) are for the first shaft, and equations (2.9) (2.11) are for the

second shaft. The two disks are coupled by the mesh spring. The flexural vibrations of

30
the shafts are distinguished for the coupled (in the plane of the spring) and uncoupled

planes with the displacements v and u, respectively. The domain of x1 is 0 < x1 < a and

a + h1 < x1 < 1 and that of x2 is 0 < x2 < b and b + h2 < x2 < L2 . The dimensionless

quantities are defined in the Appendix. Geometric compatibility at the disk-shaft interfaces

imposes the conditions given for the first and second disk-shaft interfaces as

u1 u1 uc1
(a) = (a + h1 ) = (2.12)
x1 x1 x1
v1 v1 vc1
(a) = (a + h1 ) = (2.13)
x1 x1 x1

1 (a) = 1 (a + h1 ) (2.14)

uc1 = u1 (a) + hc1 1 = u1 (a + h1 ) (h1 hc1 )1 (2.15)

vc1 = v1 (a) + hc1 1 = v1 (a + h1 ) (h1 hc1 )1 (2.16)

u2 u2 uc2
(b) = (b + h2 ) = (2.17)
x2 x2 x2
v2 v2 vc2
(b) = (b + h2 ) = (2.18)
x2 x2 x2

2 (b) = 2 (b + h2 ) (2.19)

uc2 = u2 (b) + hc2 2 = u2 (b + h2 ) (h2 hc2 )2 (2.20)

vc2 = v2 (b) + hc2 2 = v2 (b + h2 ) (h2 hc2) 2 (2.21)

The above conditions have to be satisfied for the continuity of displacement and slope for

bending, and continuity of torsional rotation on each side of the disks. The bearings at each

end of the shafts offer the boundary conditions

v1 u1 1
v1 = 0, = 0, u1 = 0, = 0, = 0, for x1 = 0 and 1 (2.22)
x1 x1 x1
v2 u2 2
v2 = 0, = 0, u2 = 0, = 0, = 0, for x2 = 0 and L2 . (2.23)
x2 x2 x2

31
The governing equations for the disk translations and rotations also result from the Hamil-

tons principle. The equations governing the disk translations are

2 uc1 3 u1 3 u1
m1 K 6 (a) + K 6 (a + h1 ) = 0 (2.24)
t2 x31 x31
2 vc1 3 v1 3 v1
m1 (a) + (a + h1 ) Fk = F1 (2.25)
t2 x31 x31
2 uc2 3 u2 3 u2
m2 K 7 (b) + K 7 (b + h2 ) = 0 (2.26)
t2 x32 x32
2 vc2 3 v2 3 v2
m2 K 8 (b) + K 8 (b + h2 ) + Fk = F2 (2.27)
t2 x32 x32
Fk = km [R1 1 (a) + R2 2 (b) vc1 + vc2 ] (2.28)

where Fk is the spring force. The equations governing the disk rotations are
2 1 3 u1 3 u1 2 u1
Icz1 + hc K 6 (a) + (h1 hc )K 6 (a + h1 ) + K 6 (a)
t2 1
x31 1
x31 x21 (2.29)
2 u1
K6 2 (a + h1 ) = 0
x1
2 1 3 v1 3 v1 2 v1
Icy1 + hc1 (a) + (h1 hc1 ) (a + h1 ) + (a)
t2 x31 x31 x21 (2.30)
2 v1
2 (a + h1 ) = 0
x1
2 2 3 u2 3 u2 2 u2
Icz2 + h K
c2 7 (b) + (h2 hc2 )K 7 (b + h2 ) + K 7 (b)
t2 x32 x32 x22 (2.31)
2 u2
K7 2 (b + h2 ) = 0
x2
2 2 3 v2 3 v2 2 v2
Icy2 + h K
c2 8 (b) + (h2 hc2 )K 8 (b + h2 ) + K 8 (b)
t2 x32 x32 x22 (2.32)
2 v2
K8 2 (b + h2 ) = 0
x2
2 1 1 1
Im1 2
(a) + K1 (a) K1 (a + h1 ) + R1 Fk = T1 (2.33)
t x1 x1
2 2 2 2
Im2 2 (b) + K2 (b) K2 (b + h2 ) + R2 Fk = T2 . (2.34)
t x2 x2
These equations are derived using the Hamiltons principle. Note that the translations of

the disks are coupled with the torsional motions through the spring force Fk . The forces

32
F1 , F2 and torques T1 , T2 are associated with the loaded static transmission error, as shown

earler.

2.3 Extended Operator Formulation

The coupled system equations can be cast in a structured form with the use of extended

operators. This kind of form is useful in demonstrating the self-adjointness of the system

and sometimes the natural use of classical analysis methods, e.g., Galerkins method. The

system deformation is represented as the extended variable [30, 51]

w(x, t) = {u1L (x1 , t), u1R (x1 , t), v1L (x1 , t), v1R (x1 , t), u2L (x2 , t), u2R (x2 , t),

v2L (x2 , t), v2R (x2 , t), 1L (x1 , t), 1R (x1 , t), 2L (x2 , t), 2R (x2 , t), uc1(t), vc1 (t),

uc2(t), vc2 (t), 1 (t), 1 (t), 2 (t), 2 (t), 1 (a, t), 2 (b, t)}T ,
(2.35)

which comprises of continuous and discrete degrees of freedom. The subscripts L, R and 1,

2 denote displacements of the left and right segments of the shafts, and the first and second

shafts, respectively. The equations of motion (2.24) (2.34) are written in the compact

form

Mwtt + Lw = f (2.36)

where M and L are the extended mass and stiffness operators, and f is the force vector.

Mw = {u1L , u1R , v1L , v1R , K5 u2L , K5 u2R , K5 v2L , K5 v2R , K3 1L , K3 1R , K4 2L ,

K4 2R , m1 uc1, m1 vc1 , m2 uc2 , m2 vc2 , Icz1 1 (t), Icy1 1 (t), Icz2 2 (t), Icy2 2 (t),

Im1 1 (a, t), Im2 2 (b, t)}T ,

(2.37)

33
 4 u1L 4 u1R 4 v1L 4 v1R 4 u2L 4 u2R 4 v2L
Lw = {K6 , K 6 , , , K 7 , K 7 , K 8 ,
x41 x41 x41 x41 x42 x42 x42

4 v2R 2 1L 2 1R 2 2L 2 2R 3 u1L
K8 , K 1 , K 1 , K 2 , K 2 , K 6 (a, t)
x42 x21 x21 x22 x22 x31

3 u1R 3 v1L 3 v1R

+K6 (a + h1 , t), (a, t) + (a + h1 , t)
x31 x31 x31

3 u2L
km [R1 1L (a, t) + R2 2L (b, t) vc1 (t) + vc2 (t)] , K7 (b, t)
x32

3 u2R 3 v2L 3 v2R

+K7 (b + h2 , t), K 8 (b, t) + K 8 (b + h2 , t)
x32 x32 x32

3 u1L
+km [R1 1L (a, t) + R2 2L (b, t) vc1 (t) + vc2 (t)] , hc1 K6 (a, t)
x31

3 u1R 2 u1L 2 u1R

+(h1 hc1 )K6 (a + h1 , t) + K 6 (a, t) K 6 (a + h1 , t),
x31 x21 x21

3 v1L 3 v1R 2 v1L 2 v1R

hc1 (a, t) + (h1 hc1 ) (a + h1 , t) + (a, t) (a + h1 , t),
x31 x31 x21 x21

3 u2L 3 u2R 2 u2L

hc2 K7 (b, t) + (h2 hc2 )K 7 (b + h2 , t) + K 7 (b, t)
x32 x32 x22

2 u2R 3 v2L 3 v2R

K7 (b + h2 , t), hc K 8 (b, t) + (h2 hc )K 8 (b + h2 , t)
x22 2
x32 2
x32

2 v2L 2 v2R 1L 1R
+K8 2
(b, t) K 8 2
(b + h2 , t), K1 (a, t) K1 (a + h1 , t)
x2 x2 x1 x1

2L
+R1 km [R1 1L (a, t) + R2 2L (b, t) vc1 (t) + vc2 (t)] , K2 (b, t)
x2

2R
K2 (b + h2 , t) + R2 km [R1 1L (a, t) + R2 2L (b, t) vc1 (t) + vc2 (t)]}T ,
x2
(2.38)

f = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, F1, 0, F2, 0, 0, 0, 0, T1, T2 }T . (2.39)

34
The inner product between two extended variables w1 and w2 is defined as

Za Z1 Za
< w1 , w2 >= r1L s1L dx1 + r1R s1R dx1 + g1L h1L dx1 +
0 a+h1 0
Z1 Zb ZL2 Zb
g1R h1R dx1 + r2L s2L dx2 + r2R s2R dx2 + g2L h2L dx2
a+h1 0 b+h2 0
ZL2 Za Z1 Zb
+ g2R h2R dx2 + p1L q1L dx1 + p1R q1R dx1 + p2L q2L dx2
b+h2 0 a+h1 0
(2.40)
ZL2
r1L s1L
+ p2R q2R dx2 + rc1 sc1 + gc1 hc1 + rc2 sc2 + gc2 hc2 + (a, t) (a, t)
x1 x1
b+h2

g1L h1L r2L s2L g2L h2L

+ (a, t) (a, t) + (b, t) (b, t) + (b, t) (b, t)
x1 x1 x2 x2 x2 x2

+p1L (a, t)q1L (a, t) + p2L (b, t)q2L (b, t)

where r1L , r1R , r2L , r2R , g1L , g1R , g2L , g2R , p1L , p1R , p2L , and p2R are the elements of

the extended variable w1 , and s1L , s1R , s2L , s2R , h1L , h1R , h2L , h2R , q1L , q1R , q2L , and

q2R are the elements of the extended variable w2 .

With this inner product and the boundary and matching conditions, the operators M

and L are symmetric, which ensures the orthogonality of the extended eigenfunctions with

respect to these operators. Moreover, M is positive definite and L is positive semi-definite.

Therefore, the eigenvalues are real and non-negative.

Using the separable form w weit to separte the spatial and temporal variables, the

eigenvalue problem is
2 M + L w = 0.


(2.41)

The dimensional natural frequencies are given by

s
E1 Isy1
= (2.42)
s1 L41

35
2.4 Discretization

Closed-form analytical solution of equation (2.41) is prohibitive. Characteristic equa-

tion method works for simpler system with less number of equations [52]. For the present

system, numerical difficulties prohibit the use of the method of characteristic equation.

The extended operator formulation and the associated inner product allow discretization of

many hybrid discrete-continuous systems with the use of the Galerkins method [31]. The

Galerkins method, however, requires the global basis functions to satisfy all the bound-

ary conditions. In other words, the basis functions need to be comparison functions. The

extended operator formulation casts the governing equations and the equations of motion

together in a compact form, leaving only the simpler boundary conditions. Hence, the basis

functions need to satisfy only these boundary conditions, giving some amount of flexibility

in their choice. Additional constraint equations at the disk-shaft interfaces makes finding

the suitable basis functions even for the Galerkin discretization difficult for the present sys-

tem. The assumed modes method ( [4], [53]) is a Rayleighs principle based energy method,

which is useful for the discretization of the current system. This is a global discretization

method as the basis functions are defined over the entire domain, and they need to satisfy

the boundary conditions only at the bearing ends. To enforce the matching conditions at

the disk-shaft interfaces, Lagrange multipliers are employed.

All the deformations are expanded in series of basis functions as given in Table 2.2.

36
Table 2.2: List of flexural and torsional deformations expressed as a combination of a series
of basis functions.
N1
P
u1L (x1 , t) = ai (t)1i (x1 ), 0 < x1 < a
i=1
PN2
u1R (x1 , t) = bi (t)2i (x1 ), a + h1 < x1 < 1
i=1
N3
P
v1L (x1 , t) = ci (t)3i (x1 ), 0 < x1 < a
i=1
PN4
v1R (x1 , t) = di (t)4i (x1 ), a + h1 < x1 < 1
i=1
PN5
u2L (x2 , t) = ei (t)5i (x2 ), 0 < x2 < b
i=1
PN6
u2R (x2 , t) = fi (t)6i (x2 ), b + h2 < x1 < L2
i=1
N7
P
v2L (x2 , t) = gi (t)7i (x2 ), 0 < x2 < b
i=1
PN8
v2R (x2 , t) = ui (t)8i (x2 ), b + h2 < x2 < L2
i=1
M1
P
1L (x1 , t) = mi (t)1i (x1 ), 0 < x1 < a
i=1
M2
P
1R (x1 , t) = ni (t)2i (x1 ), a + h1 < x1 < 1
i=1
M3
P
2L (x2 , t) = pi (t)3i (x2 ), 0 < x2 < b
i=1
M4
P
2R (x2 , t) = qi (t)4i (x2 ), b + h2 < x2 < L2
i=1

37
 Here the ji and ji are the non-dimensional basis functions for bending and torsion,

respectively, and satisfy the geometric boundary conditions (equations (2.22) and (2.23)).

Each of these basis functions are global as each of them define the mode shape of the

structure in their respective domains. The ji and ji form complete sets as well. The

chosen basis functions for bending are polynomials satisfying the orthonormality relations
Ra R1 Ra R1
1r 1s dx1 = rs , 2r 2s dx1 = rs , 3r 3s dx1 = rs , 4r 4s dx1 = rs ,
0 a+h1 0 a+h1
Rb RL2 Rb RL2
5r 5s dx2 = rs , 6r 6s dx2 = rs , 7r 7s dx2 = rs , and 8r 8s dx2 = rs .
0 b+h2 0 b+h2
The basis functions for torsion, ji, are chosen to be the eigenfunctions of each segment of

a single shaft with a concentrated inertia at the disk location for free-free end conditions.
Ra R1
They are also orthonormal functions determined by 1r 1s dx1 = rs , 2r 2s dx1 = rs ,
0 a+h1
Rb RL2
3r 3s dx2 = rs , and 4r 4s dx2 = rs . The Gram-Schmidt method is applied to or-
0 b+h2
thogonalize these basis functions. The first six normalized basis functions for flexure of the

first shaft are shown in Fig. 2.2 and the first four normalized basis functions for torsion of

the first shaft are shown in Fig. 2.3. The basis functions for the second shaft is similar to

that of the first shaft except that they are defined over the normalized length of the second

shaft. Note that 1i =3i , 2i =4i , and 5i =7i , 6i =8i .

The non-dimensional discretized kinetic energy is

1 1 1 1
T = nT MF lex n + nT MT or n + nT MCT rans n + nT MDRot n
2 2 2 2 (2.43)
= TF lex + TT or + TCT rans + TDRot ,

38
 5
7

4
6
11
15 5 21
3 13

Basis Functions, (x )

1
24 23

Basis Functions, (x )
4 26

1
25

1i
2

2i
22
3

1
2

0 1

0
-1
1
-2
2
16
14 12
-3 3
0 0.1 0.2 0.3 0.4 0.5 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Normalized Distance along the shaft (a=0.6) Normalized Distance along the shaft (a=0.6, t1 = 0.05)

(a) (b)

Figure 2.2: First six normalized basis functions for flexure of the first shaft; (a) Basis
functions 1i (x1 ) for series expansion of the left segment; (b) Basis functions 2i (x1 ) for
series expansion of the right segment.

where,
Za Z1
F lex
M = diag[ 1i1 (x1 )1j1 (x1 )dx1 , 2i2 (x1 )2j2 (x1 )dx1 ,
0 a+h1

Za Z1 Zb
3i3 (x1 )3j3 (x1 )dx1 , 4i4 (x1 )4j4 (x1 )dx1 , K5 5i5 (x2 )5j5 (x2 )dx2 ,
0 a+h1 0

ZL2 Zb ZL2
K5 6i6 (x2 )6j6 (x2 )dx2 , K5 7i7 (x2 )7j7 (x2 )dx2 , K5 8i8 (x2 )8j8 (x2 )dx2 ],
b+h2 0 b+h2
(2.44)
Ra R1
MT or = diag[K3 1i1 (x1 )1j1 (x1 )dx1 , K3 2i2 (x1 )2j2 (x1 )dx1 ,
0 a+h1

(2.45)
Zb ZL2
K4 3i3 (x2 )3j3 (x2 )dx2 , K4 4i4 (x2 )4j4 (x2 )dx2 ],
0 b+h2

39
 4 6

3
11
4

24
2

Basis Functions, (x )

Basis Functions, (x )
1

1
1 2

1i

2i
0
0
1

2 2

3 14
4
12
13 22
4 21 23

5 6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Normalized Distance along the shaft (a=0.6) Normalized Distance along the shaft (a=0.6,t 1 =0.05)

(a) (b)

Figure 2.3: First six normalized basis functions for torsion of the first shaft; (a) Basis
functions 1i (x1 ) for series expansion of the left segment; (b) Basis functions 2i (x1 ) for
series expansion of the right segment.

MCT rans = diag[m1 1i1 (a) + hc1 1i

0

0


1
(a) 1j1 (a) + hc1 1j 1
(a) ,

0

0

0


m1 3i3 (a) + hc1 3i (a) 3j3 (a) + hc1 3j (a) , m2 5i5 (b) + hc2 5i5 (b)
3 3
(2.46)
0 0


0


5j5 (b) + hc2 5j 5
(b) , m2 7i7 (b) + hc2 7i7
(b) 7j7 (b) + hc2 7j 7
(b) ],

MDRot = diag[Icy1 3i
0
3
(a)3j
0
3
(a), Icz1 1i
0
1
(a)1j
0
1
(a), Icy2 7i
0
7
(b)7j
0
7
(b),

0 0 (2.47)
Icz2 5i5
(b)5j 5
(b), Im1 1i1 (a)1j1 (a), Im1 3i3 (b)3j3 (b)].

The superscripts CTrans and DRot denote the quantities related to the translations of

the centers of mass of the disks and the rotations of the disks, respectively. The ns are

the sub-vectors of the generalized coordinate qs corresponding to the flexure or torsion as

applicable to the equation (2.43). For the basis function pip , p varies from 1 to Np , whereas

for pip , p varies from 1 to Mp . The non-dimensional discretized strain energy is

40
 1 3 M M
1 T F lex 1 T T or 1 X X
V = n K n + n K n + km [R1 mi (t)1i (a) + R2 pi (t)3i (b)
2 2 2 i=1 i=1
N3
X N3
X N7
X N7
X (2.48)
ci (t)3i (a) hc1 0
ci (t)3i (a) + gi (t)7i (b) + hc2 0
gi (t)7i (b)]2
i=1 i=1 i=1 i=1
= VF lex + VT or + VM esh ,

where,
Za Z1
KF lex = diag[K6 00
1i 1
00
(x1 )1j 1
(x1 )dx1 , K6 00
2i 2
00
(x1 )2j 2
(x1 )dx1 ,
0 a+h1

Za Z1 Zb
00 00 00 00 00 00
3i 3
(x1 )3j 3
(x1 )dx1 , 4i 4
(x1 )4j 4
(x1 )dx1 , K7 5i 5
(x2 )5j 5
(x2 )dx2 ,
0 a+h1 0

ZL2 Zb ZL2
00 00 00 00 00 00
K7 6i 6
(x2 )6j 6
(x2 )dx2 , K8 7i 7
(x2 )7j 7
(x2 )dx2 , K8 8i 8
(x2 )8j 8
(x2 )dx2 ],
b+h2 0 b+h2

(2.49)
Za Z1
KT or = diag[K1 0
1i1
0
(x1 )1j 1
(x1 )dx1 , K1 0
2i2
0
(x1 )2j 2
(x1 )dx1 ,
0 a+h1
(2.50)
Zb ZL2
0 0 0 0
K2 3i3
(x2 )3j 3
(x2 )dx2 , K2 4i4
(x2 )4j 4
(x2 )dx2 ].
0 b+h2

The principal causes of noise in a geared system are the large amplitude gear tooth

forces and bearing loads. If the bearings are assumed to be sufficiently rigid containing

no unbalance or damping, the only excitation to the system comes from the loaded static

transmission error (STE). Several models of mesh stiffness to incorporate the static trans-

mission error exist in literature. In our analysis we assume the loaded static transmission

error to include the effect of the elasticity of gear tooth as proposed by Liu and Parker [54]

in their second group of lumped-parameter models. Although this model includes the mesh

41
stiffness fluctuation indirectly, it simplifies the problem and provides reasonable estimates

for the gear dynamics. Accordingly, in present analysis, the mesh stiffness is assumed to

be constant with the time varying part being included in the static transmission error. To

obtain the discretized force due to the loaded STE, this time varying term is included in the

expression for the strain energy due to the mesh spring. Assuming the time varying part as

e(t), the modified strain energy expression due to the mesh spring can be discretized as
1
Vs = km spring 2
2
1
= km [R1 1 (a) + R2 2 (b) vc1 + vc2 + e(t)]2
2
M1 M3 N3
1 X X X (2.51)
= km [R1 mi (t)1i (a) + R2 pi (t)3i (b) ci (t)3i (a)
2 i=1 i=1 i=1
N3
X N7
X N7
X
hc1 0
ci (t)3i (a) + gi (t)7i (b) + hc2 0
gi (t)7i (b) + e(t)]2
i=1 i=1 i=1

The discretized form of the loaded STE is obtained from this expression of the modified

strain energy (later in this section) and is shown in the Appendix.

42
 The holonomic constraints on the generalized coordinates come from the ten matching

conditions shown earlier (equations (2.12) to (2.21))

1 = u1 (a + h1 , t) u1 (a, t) h1 1 (a, t)
N2
X XN1 N1
X
0
= bi (t)2i (a + h1 ) ai (t)1i (a) h1 ai (t)1i (a) = 0,
i=1 i=1 i=1
2 = v1 (a + h1 , t) v1 (a, t) h1 1 (a, t)
XN4 XN3 N3
X
0
= di (t)4i (a + h1 ) ci (t)3i (a) h1 ci (t)3i (a) = 0,
i=1 i=1 i=1
3 = u2 (b + h2 , t) u2 (b, t) h2 2 (b, t)
N6
X XN5 N5
X
0
= fi (t)6i (b + h2 ) ei (t)5i (b) h2 ei (t)5i (b) = 0,
i=1 i=1 i=1
4 = v2 (b + h2 , t) v2 (b, t) h2 2 (b, t)
XN8 N7
X N7
X
0
= ui (t)8i (b + h2 ) gi (t)7i (b) h2 gi (t)7i (b) = 0,
i=1 i=1 i=1
5 = 1 (a + h1 , t) 1 (a, t)
N2
X X N1
0 0
= bi (t)2i (a + h1 ) ai (t)1i (a) = 0,
i=1 i=1
(2.52)
6 = 1 (a + h1 , t) 1 (a, t)
N4
X X N3
0 0
= di (t)4i (a + h1 ) ci (t)3i (a) = 0,
i=1 i=1
7 = 2 (b + h2 , t) 2 (b, t)
N6
X XN5
0 0
= fi (t)6i (b + h2 ) ei (t)5i (b) = 0,
i=1 i=1
8 = 2 (b + h2 , t) 2 (b, t)
N8
X XN7
0 0
= ui (t)8i (b + h2 ) gi (t)7i (b) = 0,
i=1 i=1
9 = 1 (a + h1 , t) 1 (a, t)
M2
X XM1
= ni (t)2i (a + h1 ) mi (t)1i (a) = 0,
i=1 i=1
10 = 2 (b + h2 , t) 2 (b, t)
M4
X XM3
= qi (t)4i (b + h2 ) pi (t)3i (b) = 0.
i=1 i=1

43
There should not be any confusion between the symbols used for the coefficients of the

basis functions and the various parameters used in the mathematical model.

The non-dimensional Lagranges equations with constraints are

 
d T T V
+ = 1 A1s + 2 A2s + ... + 10 A10s (2.53)
dt qs qs qs

where s = 1, 2, ...., N and N = N1 +N2 +...+N8 +M1 +...+M4. Ais = i /qs are the

coefficients of the sth generalized coordinate in the constraints i , where the generalized co-

ordinates qs are ordered as a1 , ...aN1 , b1 , ...bN2 , c1 , ...cN3 , d1 , ...dN4 , e1 , ...eN5 , f1 , ...fN6 , g1 , ...

gN7 , u1 , ...uN8 , m1 , ...mM1 , n1 , ...nM2 , p1 , ...pM3 , q1 , ...qM4 . The loaded static transmission

error is represented as an additional time varying displacement e(t) = E0 sin(t) at the

gear mesh, which modifies the strain energy at the mesh spring (equation (2.51)) and con-

tribute to the total strain energy, V . Accordingly, a discretized generalized force f1 appears

in the right hand side.

Applying the equation (2.53), the N equations for the generalized coordinates along

with the ten constraints for the forced system are put in matrix form as [53]

       
M1 0 h L1 L2 h f1
+ = . (2.54)
0 0 LT2 0 0

The matrices M1 , L1 , h, L2 , and the vectors and f1 are given in the Appendix. The second

matrix in the equation (2.54) is the augmented stiffness matrix consisting of the stiffness

matrices and the constraint equations. The equation (2.54) is similar to the classical form of

the vibration equation for a self-adjoint system, except that both the mass and the stiffness

matrices are semi-definite matrices and hence cannot be used directly. Note similarity

between f1 and f in the equation (2.36).

44
 From the equation (2.54),

M1 h + L1 h+ L2 = f1 . (2.55)

In order to eliminate , we obtain a matrix D whose columns are the basis for the null

space of LT2 so that DT L2 =0, and premultiply both sides of the equation (2.55) by DT . The

same procedure is followed for a gyroscopic system consisting of a serpentine belt drive

in [53]. D is obtained from the singular value decomposition of L2 . Substituting h = Dy

in the equation (2.55) and premultiplying by DT lead to

Mtot y + Ltot y = DT f1 . (2.56)

Mtot =DT M1 D, Ltot = DT L1 D are the modified mass and stiffness matrices. Note that,

the dimensions of the matrices in the transformed equation (2.56) is reduced from that

of the original system equation (2.55). This is a classic case that demonstrates how the

holonomic constraint equations reduce the accessibility of a system.

Using the separable form y yeit , the eigenvalue problem is

( 2 Mtot + Ltot )y = 0. (2.57)

The system modes h are obtained back by using h = Dy.

2.5 Results and Discussions

The shafts are considered to be uniform shafts with circular cross-section although the

method developed is applicable to shafts with any cross section having an axis of symmetry.

The disks are circular with uniform thickness having the center of mass coinciding with the

axes of the corresponding shafts. The nominal parameter values are given in the Table 2.3.

45
Table 2.3: Dimensional and dimensionless parameters of a coupled disk-shaft system
Masses (kg) m1 = 3.00, m2 = 3.75

Moment of Inertias (kg m2 ) Im1 = 0.03375, Im2 = 0.06592, Icz1 = 0.01687,

Icz2 = 0.03296, Icy1 = 0.01687, Icy2 = 0.03296

Lengths (m) L1 = 0.8, L2 = 1.0, a = 0.38, b = 0.48, R1 = 0.1500,

R2 = 0.1875, h1 = 0.04, h2 = 0.04, hc1 = 0.02,

hc2 = 0.02, d1 = 0.045, d2 = 0.045

Elastic moduli (N/m2 ) E1 = 50 103 , E2 = 70 103 , G1 = 34 103 ,

G2 = 14 103

Stiffness (N/m) km = 47

Density (kg/m3 ) = 7800

Dimensionless rigidities K1 = 1.36, K2 = 0.56, K6 = 1.00, K7 = 1.40,

K8 = 1.40

Dimensionless inertias K3 = 396 106 , K4 = 396 106 , K5 = 1.0

46
 For the system described in the Table 2.3, the effects of various system parameters

on the natural frequencies of the modes are studied. To suitably investigate the coupled

interaction of bending and torsion with the spring force, the non-dimensional parameters

K1 , K8 , and km are chosen. K1 and K8 are the non-dimensional torsional and flexural

rigidities of the first and the second shafts, respectively, while km is the non-dimensional

mesh stiffness. For the range of the parameter values chosen, the modes 20, 22, and 26 are

influenced the most by these parameters, and hence they are taken as the example modes to

study the characteristics of the system.

2.5.1 Sensitivity of the Natural Frequency to Various System Param-

eters

y1
y1 1
x1 = 0
1
x1 = 0 VC1
VC1 1
y2 x1 = 1
1 2 z1 x1
y2 x1 x1 = 1 x2 = 0
2 z1
2
x2 = 0 x2 = b
2 z2 1
x2 = b x2 x1 = a
z2 1
x2 x1 = a

VC2
VC2
2 2
x2 = L 2 x2 = L 2

(a) (b)
y1
1
x1 = 0
VC1
y2
1
z1 x1 = 1
x1
2
x2 = 0 2
x2 = b
z2
x2
1 2
x1 = a
x2 = L 2

VC2

(c)

Figure 2.4: Mode shapes for the system shown in Table 1. (a) Mode 20 (b) Mode 22 (c)
Mode 26. The amplitude of torsional vibration is proportional to length of the arrow and
are shown at the ends and the disk locations.

47
 The mode shapes for the modes 20, 22, and 26 at nominal values are given in the

Fig. 2.4. Variation of natural frequency with the mesh stiffness is shown in the Fig. 2.5.

All other parameter values are kept at their respective nominal values. Modal strain energy

values at different points on the curves are shown as bar charts with the abscissa marked

from 1 to 7 implying the normalized modal energies given in Table 2.4.

Table 2.4: Normalized modal strain energies

1 Flexural energy of the 1st shaft in the uncoupled plane

2 Flexural energy of the 1st shaft in the coupled plane

3 Flexural energy of the 2nd shaft in the uncoupled plane

4 Flexural energy of the 2nd shaft in the coupled plane

5 Torsional energy of the 1st shaft

6 Torsional energy of the 2nd shaft

7 Energy stored in the mesh spring

For the modes 21, 23, 24, and 25, mesh energy is negligible, which implies that the

variation of mesh stiffness has negligible effect on the natural frequencies of these four

modes. Natural frequencies of the modes 20, 22, and 26 show variations with the change

in km . For these modes, the significant energies are from the mesh, torsion (1st shaft), and

bending (2nd shaft). From the Fig. 2.5, the following facts are observed: For the mode 20,

the maximum slope occurs at the earlier part of the curve (km 1933). For the mode 26, it

occurs at a later part of the curve (km 4070), whereas for the mode 22, it occurs at km

48
 300

Non-dimensional natural frequency

Mode 26
280 km = 4070
260
1 2 3 4 5 6 7
240
Mode 25 Mode 24
1 2 3 4 5 6 7
220
Mode 22
200 km = 2544

180 km = 1933
160 Mode 21 Mode 23
Mode 20
140

120
1 2 3 4 5 6 7 1 2 3 4 5 6 7
100
0 1017 2035 3052 4070 5087 6105 7122
Non-Dimensional Mesh Stiffness

Figure 2.5: Natural frequency vs. the non-dimensional mesh stiffness (km ) along with the
normalized modal strain energies at different modes. Modes 20, 22, and 26 are shown in
bold.

3
x 10
25
Non-dimensional Modal Mesh Energy

20

15
Mode 20
Mode 22
Mode 26
10

0
0 1017 2035 3052 4070 5087 6105 7122
Non-Dimensional Mesh Stiffness

Figure 2.6: Modal mesh energies vs. the non-dimensional mesh stiffness (km ) for the
modes 20, 22, and 26.

2544. The modal mesh energies for these three modes are also plotted against the mesh

stiffness in the Fig. 2.6. The mesh energy reaches the maximum values for the modes 20,

49
 Non-dimensional Natural Frequency
260
Mode 26
240

220
Mode 22
200

180
Mode 20
160
1 2 3 4 5 6 7
140
Mode 25
120

100
1 2 3 4 5 6 7 1 2 3 4 5 6 7 1 2 3 4 5 6 7
80
Mode 24 Mode 21 Mode 23
60
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Non-Dimensional Torsional Rigidity of the First Shaft

Figure 2.7: Natural frequency vs. K1 along with the normalized modal strain energies at
different modes. Modes 20, 22, and 26 are shown in bold.

4
x 10
3
Non-dimensional Modal Torsional Energy

2.5 Mode 20
Mode 22
Mode 26
2

1.5

0.5

0
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Non-Dimesional Torsional Rigidity of the First Shaft

Figure 2.8: Modal torsional energies vs. the non-dimensional torsional rigidity (1st shaft)
for the modes 20, 22, and 26.

22, and 26 at the location of the maximum slopes of the Fig. 2.5 (point of contraflexure).

50
 i
A general conjecture about the eigensensitivity can be drawn, which is: UM esh ,
km
where i = eigenvalue, km = mesh stiffness, and UM esh = modal mesh energy.

Variation of the natural frequency with K1 , the non-dimensional torsional rigidity of

the 1st shaft, is shown in the Fig. 2.7. The torsional modal energy plots of the first shaft

are shown in the Fig. 2.8. A relation between the slope of the curves in the Fig. 2.7 and
i
the modal torsional energy (1st shaft) exists here as well, which is: UT or , where
K1
i = eigenvalue and UT or = modal torsional energy (1st shaft). From the modal energy

plot (not shown), the mode 21 is a pure torsional mode where natural frequency increase is

proportional to K1 . The modes 23, 24, and 25 have zero torsional energy as seen in their

respective bar charts in the Fig. 2.7, which further implies that the variation of K1 has a

negligible effect on the natural frequencies of these three modes.

We observe two veering phenomena between the modes 20, 22, and 26. In the first case,

veering is observed between the modes 22 and 26. As K1 is increased from 1.40 to 2.02,

the mode 26 becomes a combined torsional-spring mode from a combined flexural-spring

mode. On the other hand, the mode 22 changes from a combined torsional-spring mode to

flexural modes of the shafts. In the second case, veering is observed between the modes 20

and 22. In this case, as K1 is increased from 0.9 to 1.5, the mode 20 becomes spring mode

from a torsional mode. Mode 22, on the other hand, changes to a torsional mode from the

spring mode. In the veering zone, the modes are strongly coupled and markedly different

from the other zones. Knowledge of veering patterns is essential in selecting the design

range of the respective parameters.

Variation of the natural frequency with K8 for nominal parameter values is shown in

the Fig. 2.9. The same analogy between the maximum slope of the curves in the natural

51
frequency vs. K8 plot and the maximum value of the bending energy (2nd shaft) is ob-

served. The modal energies are plotted against K8 in the Fig. 2.10. It is observed that
i
UBending , where i = eigen value and UBending = modal bending energy (2nd shaft).
K8
Also, the mode 25, being a pure flexural mode is linear in both the plots.

Non-dimensional Natural Frequency 250

Mode 24
1 2 3 4 5 6 7

Mode 26 Mode 22
200

Mode 20

150

1 2 3 4 5 6 7 1 2 3 4 5 6 7 1 2 3 4 5 6 7

Mode 25 Mode 23 Mode 21

100
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Non-Dimensional Flexural Rigidity of the Second Shaft

Figure 2.9: Natural frequency vs. K8 along with the normalized modal strain energies at
different modes. Modes 20, 22, and 26 are shown in bold.

Veering between the modes 22 and 26 is observed as K8 is increased from 1.1 to 1.8

(Fig. 2.10). While the mode 22 changes to a torsional mode from a combined torsional-

flexural mode, the mode 26 changes from a combined torsional-spring mode to a flexural

mode.

Lengths of the shafts have significant effect on the natural frequency of the system as

lengths are proportional to the compliances of the shafts. The Fig. 2.11 shows the variation

of the 22nd and the 26th natural frequencies with K8 , which is the flexural rigidity of the

2nd shaft for the vibration in the coupled plane. The Fig. 2.11(a) shows the variation of the

natural frequencies with K8 for four different shaft lengths (or compliances) of the mode

52
 4
x 10
3

Non-dimensional Modal Bending Energy

Mode 20
2.5 Mode 22
Mode 26
2

1.5

0.5

0
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Non-Dimensional Flexural Rigidity of the Second Shaft

Figure 2.10: Modal bending energies vs. K8 for the modes 20, 22, and 26.

22: shafts having nominal lengths with the dimensions given in Table 2 in the Appendix B,

0.95 times the nominal lengths, 0.90 times the nominal lengths, and 0.85 times the nominal

lengths. The Fig. 2.11(b) shows the same for the mode 26. However, here the lengths are:

nominal lengths, half the nominal lengths, and 0.35 times the nominal lengths. In all the

cases, the gears are at the midspan. If the mode shapes of these two modes are examined

from the case of long shafts to short shafts, the mode 22 becomes more and more torsional,

whereas the lateral vibration of the 2nd shaft becomes more significant for the mode 26.

As shown in the Fig. 2.11(a), in the case of short shafts, the value of the natural frequency

for the mode 22 does not change as the value of K8 exceeds a limiting value (after K8 =

1.3), since the mode considered becomes purely torsional. For the other mode, the natural

frequency increases monotonically with K8 for all the lengths as the lateral vibrations are

dominant as shown in the Fig. 2.11(b).

53
 210 290

Natural Frequency of Mode 26

Natural Frequency of Mode 22
LNominal
205 280

Non-Dimensional
Non-Dimensional
200
270
95% LNominal min
al
LNo
195
260
al
min
190 LNo
250 50%
90% LNominal min
al
185 LNo
35%
240
180

230
175 85% LNominal

170 220
0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 1.4 1.45 1.5 1.55 1.6 1.65 1.7 1.75 1.8 1.85 1.9
Non-Dimensional Flexural Rigidity (Second Shaft) K 8 Non-Dimensional Flexural Rigidity (Second Shaft) K8

(a) (b)

Figure 2.11: Variation of the 22nd and 26th natural frequencies with K8 for different shaft
lengths; (a) Variation of the mode 22 for the nominal lengths (), 0.95 (), 0.90 ( ),
and 0.85 ( ) times the nominal lengths; (b) Variation of the mode 26 for the nominal
lengths (), 0.50 (), and 0.35 ( ) times the nominal lengths.

The sensitivities of the natural frequencies and the vibration modes indicate how the

system parameters affect the free vibration characteristics of the system. On the other

hand, knowledge of veering is helpful to identify any dramatic changes of vibration modes

under the variation of a parameter.

2.5.2 Response due to Loaded Static Transmission Error

The only excitation in the system is the loaded static transmission error as discussed

earlier. The loaded static transmission error is modeled as a variable mesh deflection ap-

proximated by a mean mesh deflection with a sinusoidal variation. For the response anal-

ysis, the modes of interest are the modes 20 to 26, where the maximum variation of the

natural frequencies with different parameter values is observed. The response is calculated

assuming that the amplitude of the excitation as unity. The responses of the system at the

54
 20 20

Response Amplitude (dB)

15

Response Amplitude (dB)

15

10

10
= 0.02 5

= 0.04
5 0 = 0.01
= 0.04
= 0.02
-5
0
= 0.01
-10
Mode 20 Mode 20
-5
Mode 22 Mode 26 -15 Mode 22 Mode 26
-10 -20
160 170 180 190 200 210 220 230 240 250 160 170 180 190 200 210 220 230 240 250
Excitation Frequency Excitation Frequency
(a) (b)
3
x 10
30 70

= 0.01
60 = 0.01
25
Mesh Deflection

50

Mesh Force
20

= 0.02 40 = 0.02
15
30

10 = 0.04 = 0.04
20

5
10

0 0
160 170 180 190 200 210 220 230 240 250 160 170 180 190 200 210 220 230 240 250
Excitation Frequency Excitation Frequency
(c) (d)

Figure 2.12: Non-dimensional system responses due to the loaded static transmission error
for three different modal damping ratios (). (a) The 1st shaft (b) The 2nd shaft (c) Mesh
deflection (d) Mesh force. Torsional deflection of the first disk () and flexural deflection
at the disk center of mass () are shown for (a) and (b).

nominal parameter values are shown in the Fig. 2.12 for the modal damping ratios of 1, 2,

and 4%.

The Fig. 2.12(a) shows the amplitude of the flexural vibration at the center of the 1st

disk as well as the amplitude of the torsional vibration of the disk versus the excitation

frequency. The Fig. 2.12(b) shows the same for the 2nd disk. The maximum absolute

value of the mesh deflection as well as the mesh force is plotted against the excitation

frequency in Figs. 2.12(c) and 2.12(d). The peak responses occur at the modes 20, 22 and

55
26. The highest response amplitude occurs at the frequency corresponding to the mode 20

followed by that of the modes 26 and 22. The reason becomes clear if the individual mode

shapes shown in Figs. 2.4(a) 2.4(c) are observed. For all the three modes, the transverse

and torsional vibrations are in phase at the disk location. For the mode 20, the torsional

deflection of the 1st shaft is large and in phase with the rotation of the disk. Also the flexural

deflections of the center of the disks are large. For the mode 22, even though the torsional

deflection of the 1st shaft is larger, it is in opposite phase to the rotation of the disk. Also,

the flexural deflection of the center of the 2nd disk is small. For the mode 26, torsional

deflection (out of phase with the rotation of the 1st disk) as well as flexural deflection of the

disk centers is moderate.

The responses are verified by the one, two, and three mode approximations as depicted

in the Fig. 2.13 for 4% and the Fig. 2.14 for 0.5% modal damping ratios, respectively. The

three modes selected are the modes 20, 22, and 26. Three mode approximation is closer to

the actual response (calculated by the modal superposition of a large number of the modes

into account). Also, when the modal damping ratio is small, even one mode approximation

can accurately measure the response for the corresponding frequency.

56
 15
15
Response Amplitude (dB)

Response Amplitude (dB)

10
10

5
5

0
Three Mode Approximation 0
Three Mode Approximation
Actual
-5 -5

One Mode Approximation

-10 -10 Actual

Two Mode Approximation
One Mode Approximation
Two Mode Approximation
-15 -15
160 170 180 190 200 210 220 230 240 250 160 170 180 190 200 210 220 230 240 250
Excitation Frequency Excitation Frequency
(a) (b)
3
8 x 10
20
Actual Actual
1 mode 18 1 mode
7 2 mode 2 mode
3 mode 16 3 mode
6
Mesh Deflection

Mesh Force

14
5
12

4 10

3 8

6
2
4
1
2

0 0
160 170 180 190 200 210 220 230 240 250 160 170 180 190 200 210 220 230 240 250
Excitation Frequency Excitation Frequency
(c) (d)

Figure 2.13: Non-dimensional actual and approximate responses due to the loaded static
transmission error for 4 % modal damping along with the one, two, and three mode approx-
imations. (a) The 1st shaft (b) The 2nd shaft (c) Mesh deflection (d) Mesh force. Torsional
deflection of the first disk ( ) and flexural deflection at the disk center of mass () are
shown for (a) and (b).

57
 25

Response Amplitude (dB)

25
Response Amplitude (dB)

20 20

15
15

10
10
5
5 0
One Mode
-5
0 Three Mode Approximation
Approximation -10
-5
-15
Actual Actual
-10 -20 Three Mode Two Mode
Two Mode
One Mode Approximation Approximation Approximation Approximation
-15
-25
160 170 180 190 200 210 220 230 240 250 160 170 180 190 200 210 220 230 240 250
Excitation Frequency Excitation Frequency
(a) (b)
6
60 x 10
0.14
Actual Actual
1 mode 1 mode
50
2 mode 0.12 2 mode
3 mode 3 mode
0.10
Mesh Force
Mesh Deflection

40

0.08
30

0.06

20
0.04

10
0.02

0 0
160 170 180 190 200 210 220 230 240 250 160 170 180 190 200 210 220 230 240 250
Excitation Frequency Excitation Frequency
(c) (d)

Figure 2.14: Non-dimensional actual and approximate responses due to the loaded static
transmission error for 0.5 % modal damping along with the one, two, and three mode
approximations. (a) 1st shaft (b) 2nd shaft (c) Mesh deflection (d) Mesh force. Torsional
deflection of the first disk ( ) and flexural deflection at the disk center of mass () are
shown for (a) and (b).

58
 CHAPTER 3

EIGENSOLUTION OF GYROSCOPIC HELICAL GEARED

SYSTEM ATTACHED TO COMPLIANT SHAFTS

In this chapter, a model of a pair of helical gears mounted on compliant parallel spin-

ning shafts is developed. A well known spatial discretization method is applied for the

linear vibration analysis of this hybrid discrete-continuous system where the gear mesh is

modeled as discrete element and the attached shafts are modeled as continuous elements.

The gear mesh is modeled as a combination of translational and torsional springs, which

is more accurate for high speed applications than the model used in chapter 2. Inclusion

of coriolis and centripetal acceleration terms in the analysis leads to gyroscopic system

characteristics not captured in the previous models.

3.1 Introduction

Research on the noise in geared systems has been centered on gear-specific compo-

nents, e.g., time varying mesh stiffness [8], tooth profile modification [55], gear body vi-

brations [1], etc. This has produced useful information and led to reduction in the vibration

induced failures and noise of geared systems. However, the role of internal components

such as shafts and bearings in the dynamic response and noise has not been thoroughly in-

vestigated. As gears are generally mounted on shafts supported by bearings, the dynamics

59
of these internal components can be important. Their inclusion casts the problem in the

domain of rotor dynamics.

Previous methods to solve the vibration analysis problems of a flexible and/or rigid shaft

carrying flexible or rigid single or multiple disks are mainly based on lumped parameter

approximation using transfer matrix methods [25, 39]. A lumped parameter approach sim-

plifies the problem but is not convenient to accurately predict the higher natural frequencies

and critical speeds of a shaft-disk system.

The problem of a continuous shaft carrying a rigid disk is solved by Srinath and Das

[22] for the non-rotating case and Eshleman and Eubanks [23] for the rotating case. The

characteristic equation method is used to solve for system eigenvalues. Nataraj [27] devel-

ops a mathematical model to obtain the interaction between the torsional and flexural de-

flections of a uniform shaft rotating at a constant speed. Parker and Mote [30] introduce an

extended operator formulation to find the eigensolutions of a stationary disk-clamp-spindle

system with asymmetries due to geometric shape imperfections. Parker [31] extends this

work to a spinning elastic disk-spindle system and obtains the coupled disk-spindle eigen-

solutions using Galerkin discretization.

Geared spindle systems are usually represented as two disk-spindle systems coupled

through a gear mesh. In most of the cases, the disks are considered to be rigid, and the shafts

are subject to torsional vibration. If the shafts are compliant enough, the lateral vibrations

due to flexure may not be neglected. The vibration problem then becomes a combined

flexural-torsional-mesh vibration problem. Iida et al. [11] obtain the natural frequency and

vibration modes for coupled torsional-flexural vibration of geared rotors. In their model,

one of the shafts is assumed to be rigid, and the response to geometric eccentricity and

mass unbalance of the three degree of freedom model is determined. The compliant effect

60
of the gear mesh is ignored in their model. Later, Iida et al. [1214] apply this model to a

gear train. David and Mitchell [15] show the importance of including dynamic coupling of

system elements to produce tractable solutions of a torsional-axial-lateral coupled geared

system. Neriya et al. [20] and Kahraman et al. [21] use finite element methods to study the

modes and the dynamic response of the coupled torsional-flexural model of geared rotors.

Gyroscopic effects are not considered in their study. Choi and Mau [19] show that the

lateral motion along the direction of the gear mesh might induce coupled lateral-torsional

mode for a geared rotor-bearing system. The transfer matrix method is employed in this

study. Lee et al. [40] systematically investigate the uncoupled and coupled lateral and

torsional natural frequencies using finite element model. Gyroscopic effects are considered

as part of the detailed analysis in these studies.

For high speed applications, the coriolis acceleration term acts as an additional coupling

term and is necessary to consider for accurate estimation of the vibrational frequencies and

mode shapes. Some of the above works recognise the importance of including the coriolis

acceleration and the corresponding gyroscopic force term in the equations of motion.

In powertrain and gearing applications (e.g., helicopters, automobiles, turbo-prop en-

gines, etc.) vibration and noise is of major concern. Dynamic excitation from changing

mesh stiffness is the primary source of noise. A secondary source of noise and vibration

is due to the vibration of the attached shafts. Gears used in transmissions generally have

short shafts. Higher order beam theory is generally used for short shafts. However, as a first

attempt toward developing an analytical model of a gyroscopic geared-rotor system, sim-

plified beam theory (Euler-Bernoulli beam theory) is used here. Also, for shorter beams,

the torsional effect may not be significant. Hence, the coupling effect is expected mainly

61
between the flexural and mesh vibrations, even though torsion is included. The objectives

of the current study are:

(a) Build an analytical single-mesh linear gyroscopic model of a helical geared shaft

system with the gears modeled as rigid disks connected by translational and rotational

springs, and the shafts having flexural-torsional compliance.

(b) Develop a methodology to solve the eigenvalue problem of such a system using the

Galerkin discretization. Obtain the sensitivities of these eigenvalues to different system

parameters including the rotational speed and the mesh stiffness. The eigenvalues obtained

by this method are more efficient and accurate than the conventional finite element methods.

(c) Perform modal analysis appropriate for gyroscopic systems and obtain the response

due to the loaded static transmission error.

3.2 Kinematics of Helical Gear Pair on Parallel Shafts

u1 y
e2
1 L1 1
h1
v1 1 Gear 1
u2
z x1
2
km
Gear 2
v2 2
KT
e2
x2
L2

h2 2

(a) (b)

Figure 3.1: (a) Model of the coupled spinning gear-shaft system. The dotted line is parallel
to the line of action. (b) Schematic to visualize the torsional mesh stiffness.

62
 The model (Fig. 3.1(a)) consists of two helical gears in mesh mounted on parallel

shafts. The gear pair is modeled as two rigid cylindrical disks with radii equal to the base

circle radii of the corresponding gears. The x1 and x2 axes are along the axes of the first

and second shafts. The z axis is perpendicular to the x1 axis, lying parallel to the plane of

the gear bodies and is parallel to the tangent to both base circles. The right-handed inertial

orthonormal basis {e1 , e2 , e3 } is chosen such that e3 is parallel to the tangent connecting

these two base circles and lies parallel to the plane of action of the gear mesh (x1 z plane)

in the undeflected configuration. The line of action at the gear mesh is rotated at an angle

from e3 (Fig. 3.1(a)). The origin is at the base of the first shaft at x1 = 0. The y axis is

perpendicular to the plane of action.

The shafts are modeled as uniform Euler-Bernoulli beams for flexural deformation. The

torsional deformations are modeled using first order torsion theory. The flexural vibrations

of the shafts are distinguished between vibrations v1 and v2 along z and vibrations u1 and

u2 in the y direction. (See Table 3.1 for definitions of the symbols.) The tilting of the

disks in these planes are defined by the slopes of the shafts at the disk-shaft interfaces:
v1 v2 u1 u2
1 (1), 2 (L2 ) and 1 (1), 2 (L2 ) for normalized shaft
x1 x2 x1 x2
lengths. The non-dimensional quantities are given in the subsequent section.

The gear mesh in this three-dimensional model is represented by an equivalent stiffness

model, which is the combination of a translational and torsional spring at the tooth mesh

located midway on the active facewidth [56]. The translational spring has the time averaged

equivalent spring stiffness (km ) over a mesh cycle representing the classical gear mesh

stiffness. In the three-dimensional gear pair model, the translational mesh spring acts along

the direction of the line of action in the helical gear pair, which is at an angle with respect

to the z axis (e3 ). It resists deflections in the tangential and axial directions. The tangential

63
 e2

Tangential Direction

R1
1 Axial Direction
Undeflected
Shaft 1 e1 contact point
e1

os
a
A

t c
t
R2

2
Shaft 2 a sin a
Deflected
contact
point e3
(a) (b)

Figure 3.2: Helical gear translational mesh deflection. (a) Axial deflection a (AB) due
to tilting of the gears about the z1 axis. (b) Components of the total translational mesh
deflection.

deflection t is the same as the deflection of the mesh spring of a spur gear pair, which

consists of the relative deflection of the contacting teeth in the (e3 ) direction. The axial

mesh force is generated due to relative motion a of the contact point in the e1 direction.

This force causes tilting of the gears in the x1 y plane (Fig. 3.2(a)), which couples

with the flexural deflections of the shafts in the same plane. For a spur gear pair, flexural

deflections of the shafts in this plane are unaffected by the gear mesh. The total translational

mesh deflection () comprises of the components due to the axial (a ) and the tangential

(t ) deflections and is governed by the helix angle (Fig. 3.2(b)). The corresponding strain

energy stored in the mesh spring is given as km 2 /2.

Relative tilting of the gears about the y axis introduces an additional strain energy term

at the gear mesh. The resistance to relative tilting is more pronounced for large facewidth

64
gears. The torsional mesh spring with stiffness KT is needed to represent this strain energy

KT 2 /2. Here is the relative tilting of the gears (1 2 ) about the y axis (Fig. 3.1(b)).

An average value of KT is chosen over the gear mesh cycle.

3.3 Modeling and Equations of Motion

Coupling between the torsional, flexural, and axial motions is an interesting phenomenon

in helical gear systems [57, 58]. Shaft axial vibrations are neglected even though the axial

vibration of the mesh due to the gear tilting effect remains. The torsional and flexural vi-

brations are considered in modeling the shafts, while the disks remain rigid. The bearings

are assumed to be rigid and provide fixed-fixed end conditions for the shafts to transverse

vibrations but add no constraints to the torsional vibrations.

The governing equations of motion are derived from Hamiltons principle using the

kinetic and strain energies and geometric constraints. The following non-dimensional vari-

ables are defined (See Table 3.1 for definitions of the symbols.):
s
E1 Isy1 L2 x1 x2 u1 u2 v1
t = t 4
, L1 = 1, L2 = , x1 = , x2 = , u1 = , u2 = , v1 = ,
s1 L1 L1 L1 L1 L1 L1 L1

v2 uc1 uc2 vc1 vc2 h1 h2 hc1

v2 = , uc1 = , uc2 = , vc1 = , vc2 = , h1 = , h2 = , hc1 = ,
L1 L1 L1 L1 L1 L1 L1 L1

hc2 R1 R2 m1 m2 Icy1 Icy2

hc2 = , R1 = , R2 = , m1 = , m2 = , Icy1 = , Icy = ,
L1 L1 L1 s1 L1 s1 L1 s1 L31 2
s1 L31

Icz1 Icz2 Im1 Im2 km

Icz1 = , Icz = , Im = , Im = , k m = ,
s1 L31 2
s1 L31 1
s1 L31 2
s1 L31 E1 Isy1 /L31

KT G1 J1 G2 J2 J1 J2 s2
KT = , K 1 = , K 2 = , K 3 = , K 4 = , K 5 = ,
E1 Isy1 /L1 E1 Isy1 E1 Isy1 s1 L21 s1 L21 s1

Isz1 E2 Isz2 E2 Isy2

K6 = , K7 = , K 8 = .
Isy1 E1 Isy1 E1 Isy1
(3.1)

65
Table 3.1: List of symbols defining spinning helical geared shaft system. Subscripts 1 and
2 denote the quantities belonging to the first and the second shafts, respectively. Overbar
indicates dimensional quantities, otherwise it is non-dimensional.

t Time
1 , 2 Rotational speeds of the shafts
L1 , L2 Lengths of the shafts
s1 , s2 Mass per unit length of the shafts
u1 , u2 Flexural deflections of the shafts in the x1 y plane
uc1, uc2 Flexural deflections of the centers of mass of the disks
in the x1 y plane
v1 , v2 Flexural deflections of the shafts in the x1 z plane
vc1 , vc2 Flexural deflections of the centers of mass of the disks
in the x1 z plane
1 , 2 Torsional deflections of the shafts about their respective axes
R1 , R2 Radii of the disks
h1 , h2 Thickness of the disks
hc1 , hc2 Distances of the centers of mass of the disks from
the left hand end of the disks
m1 , m2 Masses of the disks
Icy , Icz Mass moments of inertia of the disk with respect to
the y and z axes
Im Mass moments of inertia of the disk with respect to the x axis
Isy , Isz Area moments of inertia of the shaft cross section with respect to
the y and z axes
J1 , J2 Polar moments of inertia of the shaft cross-sections
km Mesh stiffness
KT Torsional mesh stiffness
Helix angle
K1 , K2 Torsional rigidities of the shafts
K7 , K8 Flexural rigidities of the second shaft in the x1 y
and the x1 z planes
i ith mode
i Natural frequency of i

66
 The non-dimensional kinetic energy is

T
T =
E1 Isy1 /L1
Z1 Z1
1 u1 1 v1
= ( 1 v1 )2 dx1 + ( + 1 u1 )2 dx1
2 t 2 t
0 0
L
1 Z 2 u ZL2
2 1 v2
+ K5 ( + 2 v2 )2 dx2 + ( 2 u2 )2 dx2
2 t 2 t
0 0
Z1  2 ZL2  2
1 1 1 2
+ K3 dx1 + K4 dx2
2 t 2 t
0 0
1 uc1 1 1 vc1 1
+ m1 [ + (1 + hc1 ) 1 vc1 ]2 + m1 [ + (1 + hc1 ) + 1 uc1]2
2 t t 2 t t
1 uc2 2 1 vc2 2
+ m2 [ + (L2 + hc2 ) + 2 vc2 ]2 + m2 [ + (L2 + hc2 ) 2 uc2]2
2 t t 2 t t
1 1 2 1 1 2 1 2 2
+ Icy1 [1 1 + ] + Icz1 [1 1 ] + Icy2 [2 2 ]
2 t 2 t 2 t
1 2 2 1 1 1 2
+ Icz2 [2 2 + ] + Im1 [1 + (1)]2 + Im2 [2 (L2 )]2 .
2 t 2 t 2 t
(3.2)

The non-dimensional strain energy is

V
V =
E1 Isy1 /L1
Z1  2 2 Z1  2 2 ZL2  2 2
1 v1 1 u1 1 v2
= 2
dx1 + K6 2
dx1 + K8 dx2
2 x1 2 x1 2 x22
0 0 0 (3.3)
ZL2  2 Z1  2 ZL2  2
1 2 u2 1 1 1 2
+ K7 dx2 + K1 dx1 + K2 dx2
2 x22 2 x1 2 x2
0 0 0
1 1
+ KT (1 2 )2 + km [a sin + t cos ]2
2 2
where

a = R1 1 + R2 2 (3.4)

t = R1 1 (1) + R2 2 (L2 ) vc1 + vc2 . (3.5)

67
The last two terms in equation (3.3) are the strain energies stored in the mesh springs. The

boundary conditions corresponding to the fixed support for bending and free support for

torsion are

v1 u1 1
v1 = 0, = 0, u1 = 0, = 0, = 0, for x1 = 0 (3.6)
x1 x1 x1
v2 u2 2
v2 = 0, = 0, u2 = 0, = 0, = 0, for x2 = 0. (3.7)
x2 x2 x2

e2

First Gear

Undeformed e1
location
e3
Deformed
location
km
F STE = km e(t)
e(t)
Second Gear

e1

Figure 3.3: Force on the first gear due to the loaded static transmission error.

The static transmission error (STE) in a gear pair is defined as the difference in the

angular position (about (e1 )) of the driven gear and the angular position where it would be

68
if the gears had perfect conjugate action with no elastic deflection [9]. In this model, the

loaded static transmission error is used as an input excitation as recommended in [8,10,59].

This displacement excitation e(t) is applied along the direction of the line of action at the

location of the mesh spring with a constant time averaged mesh stiffness (Fig. 3.3). Fig.

3.3 depicts the force FST E on the first disk due to the loaded STE. This results in the

forces along the e1 (km e(t) sin ) and e3 (km e(t) cos ) axes and moments about the e1

(km e(t)R1 cos ) and e3 (km e(t)R1 sin ) axes on the first disk. The force acting on the

second disk is equal in magnitude but opposite in sign, with the corresponding moments

produced about the e1 and e3 axes. The virtual work done by the loaded STE over a mesh

cycle is given as
W = km e(t)[cos vc1 cos vc2 R1 sin 1 R2 sin 2

R1 cos 1 (1) R2 cos 2 (L2 )] (3.8)

= {F1 vc1 + F2 vc2 + M1 1 + M2 2 + T1 1 (1) + T2 2 (L2 )}

Rt
Substituting all of the above quantities into the Hamiltons principle, (T V +
0
W ) dt = 0, the non-dimensional governing equations for the continuous shafts and the

equations of motion of the discrete disks are obtained. The non-dimensional governing

equations for the coupled bending and torsion of the shafts are

4 u1 2 u1 v1
K6 4
+ 21 21 u1 = 0 (3.9)
x1 t2 t
4 v1 2 v1 u1
4
+ 2 + 21 21 v1 = 0 (3.10)
x1 t t
2 1 2 1
K3 K 1 =0 (3.11)
t2 x21
4 u2
 2 
u2 v2 2
K7 4 + K5 + 22 2 u2 = 0 (3.12)
x2 t2 t
4 v2
 2 
v2 u2 2
K8 + K5 22 2 v2 = 0 (3.13)
x42 t2 t

69
 2 2 2 2
K4 K 2 = 0. (3.14)
t2 x22

Equations (3.9) (3.11) are for the first shaft, and equations (3.12) (3.14) are for the

second shaft. The domain of x1 is 0 < x1 < 1 and that of x2 is 0 < x2 < L2 . Each disk

is subject to elastic forces and moments due to bending and torsion from the shafts. The

equations governing the disk translations are

2 uc1 2 1 3 u1
  
vc1 1
m1 21 21 uc1 + (1 + hc1 ) 1 K6 |x =1 = 0
t 2 t t 2 t x31 1
(3.15)
2 vc1 2 1 3 v1
  
uc1 1
m1 + 21 21 vc1 + (1 + hc1 ) + 1
|x =1 Fk cos = F1
t 2 t t 2 x31 1
t
(3.16)
 2  2
3 u2

uc2 vc2 2 2 2
m2 + 2 2 2 u c2 + (L 2 + h c2 ) + 2 K 7 |x =L = 0
t2 t t2 t x32 2 2
(3.17)
 2  2
3 v2

vc2 uc2 2 2 2
m2 22 2 vc2 + (L2 + hc2 ) 2 K 8 |x =L +Fk cos = F2
t2 t t2 t x32 2 2
(3.18)

Fk = km {(R1 1 + R2 2 ) sin + (R1 1 |x1 =1 + R2 2 |x2 =L2 vc1 + vc2 ) cos } (3.19)

where Fk is the translational spring force. The equations governing the disk rotations are
 2   
1 1 1 2
Icz1 2
1 Icy1 1 + 1 1 + (1 + hc1 ) m1
 2 t t t
2 1 2 u1

uc1 vc1
+ (1 + hc1 ) 1 + R F
1 k sin + K 6 |x =1 (3.20)
t2 t2 t x21 1
3 u1
+hc1 K6 3 |x1 =1 = M1
x1
 2   
1 1 2 1
Icy1 2
+ 1 Icz1 1 1 1 + (1 + hc1 ) m1
 2 t t t
2 1 2 v1 3 v1

vc1 uc1 (3.21)
2
+ (1 + h c1 ) 2
+ 1 + |
2 x1 =1
+ hc1 |
3 x1 =1
t t t x1 x1
+KT (1 2 ) = 0

70
 2 2
   
2 2 2
Icz2 2
+ 2 Icy2 2 2 2 + (L2 + hc2 ) m2
 2 t t t
2 2 2 u2

uc2 vc2
+ (L 2 + hc2 ) + 2 + R F
2 k sin + K 7 |x =L (3.22)
t2 t2 t x22 2 2
3 u2
+hc2 K7 3 |x2 =L2 = M2
x2
 2   
2 2 2 2
Icy2 2
2 Icz2 2 2 + 2 + (L2 + hc2 ) m2
 2 t t t
2 2 2 v2 3 v2

vc2 uc2 (3.23)
+ (L2 + hc2 ) 2 2 + K8 2 |x2 =L2 + hc2 K8 3 |x2 =L2
t2 t t x2 x2
KT (1 2 ) = 0
2 1 1
Im1 2
|x1 =1 + K1 |x =1 + R1 Fk cos = T1 (3.24)
t x1 1
2 2 2
Im2 2
|x2 =L2 + K2 |x =L + R2 Fk cos = T2 . (3.25)
t x2 2 2

The disk translations along the y axis are unaffected by the axial or tangential mesh

force, but the translations along the z direction are influenced by them. Tilting motions of

the disks in the x1 y plane, however, are affected by the moment produced by the mesh

force coming as coupling terms in equations (3.20) and (3.22). This, in turn, influences

the flexural vibration of the shaft in this plane. In this context, the role of the helix angle

cannot be overlooked. When the helix angle is zero (spur gear), the motion of the disks and

shaft bending can be clearly distinguished between the coupled and uncoupled planes. In

that case, both the axial mesh force and the aforementioned coupling terms are zero. The

non-zero helix angle increases the degree of coupling in the system. Tilting motions of the

disks about the y axis are only influenced by the torsional mesh spring as the force of the

translational spring intersects the y axis.

The forces due to the coriolis and centripetal accelerations appear as additional gyro-

scopic (coriolis) and stiffness (centripetal) terms in the classical forms of the flexural beam

vibration equations as well as in the equations for the discrete degrees of freedom. The

gyroscopic acceleration in a given shaft bending plane has magnitude proportional to the

71
out of plane velocity, and this couples vibration between shaft bending motions in different

planes. Increasing the rotation speed increases this coupling. Torsional shaft vibrations are

integral parts of the mesh deflection as they appear as tangential terms in it (equation 3.19).

Hence, the system involves both the torsional and flexural vibrations (in orthogonal planes

coupled by gyroscopic acceleration) of the shafts coupled with torsional and translational

mesh deflections.

3.4 Extended Operator Formulation and Discretization

The equations of motion for the continuum shaft deflections and the discrete disk dis-

placements and rotations can be written in a compact form by the use of an extended oper-

ator formulation [30,51]. The gyroscopic system structure is apparent with the use of these

operators. The extended operator form is useful in the subsequent Galerkin discretization

and modal analysis. The coupled system deformation is represented as the extended vari-

able
w(x, t) = { u1 (x1 , t), v1 (x1 , t), u2 (x2 , t), v2 (x2 , t), 1 (x1 , t), 2 (x2 , t), uc1(t), vc1 (t),

uc2(t), vc2 (t), 1 (t), 1 (t), 2 (t), 2 (t), 1 (1, t), 2 (L2 , t)}T ,
(3.26)

which comprises of continuous and discrete degrees of freedom. The variable x belongs to

the extended space over which w is defined. The equations of motion (3.9) (3.25) are

written in the compact form

 
Mwtt + Gwt + L 2 L w = f (3.27)

where M, G, L, and L are the extended mass, gyroscopic, stiffness, and rotational stiffness

operators, respectively, operating on w. The f is the force vector and = 1 . The rotation

72
 R1
speed of the second shaft is 2 = 1 .
R2

Mw = {u1 , v1 , K5 u2 , K5 v2 , K3 1 , K4 2 , m1 {uc1 + (1 + hc1 )1 } ,

m1 {vc1 + (1 + hc1 )1 } , m2 {uc2 + (L2 + hc2 )2 } , m2 {vc2 + (L2 + hc2 )2 } ,

Icz1 1 + (1 + hc1 )m1 {uc1 + (1 + hc1 )1 } , Icy1 1 + (1 + hc1 )m1

(3.28)
{vc1 + (1 + hc1 )1 } , Icz2 2 + (L2 + hc2 )m2 {uc2 + (L2 + hc2 )2 } ,

Icy2 2 + (L2 + hc2 )m2 {vc2 + (L2 + hc2 )2 } , Im1 1 (1), Im2 2 (L2 )}T ,

Gw = {2v1 , 2u1, 2K5 v2 (R1 /R2 ), 2K5 u2 (R1 /R2 ), 0, 0, m1

{2vc1 + (1 + hc1 )1 } , m1 {2uc1 + (1 + hc1 )1 } , m2 {2vc2 + (L2 + hc2 )2 }

(R1 /R2 ), m2 {2uc2 + (L2 + hc2 )2 } (R1 /R2 ), (Icy1 + Icz1 )1

(3.29)
(1 + hc1 )m1 vc1 , (Icy1 + Icz1 )1 + (1 + hc1 )m1 uc1,

[(Icy2 + Icz2 )2 + (L2 + hc2 )m2 vc2 ] (R1 /R2 ),

[(Icy2 + Icz2 )2 (L2 + hc2 )m2 uc2 ] (R1 /R2 ), 0, 0}T ,

73
 4 u1 4 v1 4 u2 4 v2 2 1 2 2 3 u1
Lw = {K6 , , K 7 , K 8 , K 1 , K 2 , K 6 (1),
x41 x41 x42 x42 x21 x22 x31

3 v1
(1)
x31

km [(R1 1 + R2 2 ) sin() + {R1 1 (1) + R2 2 (L2 ) vc1 + vc2 } cos()]

cos(),

3 u2 3 v2
K7 (L2 ), K 8 (L2 )+
x32 x32

km [(R1 1 + R2 2 ) sin() + {R1 1 (1) + R2 2 (L2 ) vc1 + vc2 } cos()]

2 u1 3 u1
cos(), K6 (1) + hc K 6 (1)+
x21 1
x31

R1 km [(R1 1 + R2 2 ) sin() + {R1 1 (1) + R2 2 (L2 ) vc1 + vc2 } cos()]

(3.30)
2 3 2
v1 v1 u2
sin(), 2
(1) + hc1 3 (1) + KT (1 2 ), K7 2 (L2 )+
x1 x1 x2

3 u2
hc2 K7 (L2 )+
x32

R2 km [(R1 1 + R2 2 ) sin() + {R1 1 (1) + R2 2 (L2 ) vc1 + vc2 } cos()]

2 v2 3 v2 1
sin(), K8 2
(L2 ) + hc 2 K 8 3
(L2 ) KT (1 2 ), K1 (1) + R1 km
x2 x2 x1

[(R1 1 + R2 2 ) sin() + {R1 1 (1) + R2 2 (L2 ) vc1 + vc2 } cos()] cos(),

2
K2 (L2 ) + R2 km
x2

[(R1 1 + R2 2 ) sin() + {R1 1 (1) + R2 2 (L2 ) vc1 + vc2 } cos()]

cos()}T ,
Lw = {u1 , v1 , K5 u2 (R1 /R2 )2 , K5 v2 (R1 /R2 )2 , 0, 0, m1uc1 , m1 vc1 , m2 uc2(R1 /R2 )2 ,

m2 vc2 (R1 /R2 )2 , Icy1 1 , Icz1 1 , Icy2 2 (R1 /R2 )2 , Icz2 2 (R1 /R2 )2 , 0, 0}T ,
(3.31)

74
 f = {0, 0, 0, 0, 0, 0, 0, F1, 0, F2 , M1 , 0, M2 , 0, T1 , T2 }T . (3.32)

The inner product between two extended variables w1 and w2 is defined as

Z1 Z1 ZL2 ZL2
(w1 , w2 ) = r1 s1 dx1 + g1 h1 dx1 + r2 s2 dx2 + g2 h2 dx2
0 0 0 0
Z1 ZL2
+ p1 q1 dx1 + p2 q2 dx2 + rc1 sc1 + gc1hc1 + rc2 sc2 + gc2 hc2
0 0 (3.33)
r1 s1 g1 h1 r2 s2
+ (1) (1) + (1) (1) + (L2 ) (L2 )
x1 x1 x1 x1 x2 x2

g2 h2
+ (L2 ) (L2 ) + p1 (1)q1 (1) + p2 (L2 )q2 (L2 )
x2 x2
where r1 , g1 , r2 , g2 , p1 , and p2 are the elements of the extended variable w1 , and s1 , h1 , s2 ,

h2 , q1 , and q2 are the elements of the extended variable w2 .

With that, the operators M, L, and L are symmetric, and G is skew-symmetric. More-

over, M is positive definite and L is positive semi-definite. The gyroscopic terms captured

in the eigenvalues can even be complex quantities with non-zero real parts. Subcritical

speed eigenvalues appear as purely imaginary complex conjugate pairs n = jn with

the corresponding eigenfunctions being complex conjugates. For some rotation speeds of

the shafts, real parts of the eigenvalues are non-zero and they appear as positive-negative

pairs. Using the separable form, w(x, t) = u(x)eit , the eigenvalue problem is

 
2 M + iG + L 2 L u = 0 (3.34)

Closed-form analytical solution of equation (3.34) is prohibitive. Galerkin discretiza-

tion is a weighted residual technique that can be readily applied to the extended operator

formulation because of the mathematical structure of the operators with the associated in-

ner product (3.26). The trial functions are taken from the space of admissible functions [4].

75
As suggested by Jha and Parker [60], stationary system basis functions for beam vibra-

tions are used as trial functions for this rotating beam analysis. The extended variable w is

approximated with the expansion

N
X
wN (x) = aj (t)j (x),
j=1

N = N1 + N2 + N3 + N4 + M1 + M2 ,

j (x) = {1j (x1 ), 0, 0, 0, 0, 0}T j = 1, 2, ...N1 ,

= {0, 2j (x1 ), 0, 0, 0, 0}T j = 1, 2, ...N2 ,

(3.35)
= {0, 0, 3j (x2 ), 0, 0, 0}T j = 1, 2, ...N3 ,

= {0, 0, 0, 4j (x2 ), 0, 0}T j = 1, 2, ...N4 ,

= {0, 0, 0, 0, 1j (x1 ), 0}T j = 1, 2, ...M1 ,

= {0, 0, 0, 0, 0, 2j (x2 )}T j = 1, 2, ...M2 ,

where j s belong to the space of w and not the actual deflections of the shafts. They are

written in the form of an extended variable. Ni s are the number of basis functions for bend-

ing of each shaft in individual planes and Mi s are the number of basis functions for torsion

of each shaft. The 2i and 4i are the basis functions for bending in the plane of action,

and 1i , 3i are those perpendicular to the plane of action. Note that, 1i = 2i and 3i =

4i and they are chosen as polynomials. The 1i and 2i are the eigenfunctions for torsion

of each shaft with a concentrated inertia at the disk end. Each one of them is a global basis

function as each define the mode shape of the whole structure in their respective domains.

They satisfy the boundary conditions (equations (3.6) and (3.7)). They also form a com-
R1 R1
plete set and satisfy the orthonormality relations 1r 1s dx1 = rs , 2r 2s dx1 = rs ,
0 0
RL2 RL2 R1 RL2
3r 3s dx2 = rs , 4r 4s dx2 = rs , 1r 1s dx1 = rs , 2r 2s dx2 = rs .
0 0 0 0

76
 Substituting (3.35) into equation (3.27), the Galerkin discretization yields

h i
[Mqr ] ar + [Gqr ] ar + Lqr 2 Lqr ar = g, r, q = 1, ..., N (3.36)

where

Mqr = (Mr , q ) = Mrq , Gqr = (Gr , q ) = Grq , Lqr = (Lr , q ) = Lrq ,

 
Lqr = Lr , q = Lrq , gq = (f, q )

The discretized force vector g corresponding to the loaded STE is given as

R1 sin [1i (1)]N1 1

0

0

cos [2i (1) + hc1 2i (1)]N 1
2

0


R 2 sin [3i (L 2 )] N 3 1


g = km e(t)
(3.37)
0
cos [4i (L2 ) + hc2 (L2 )]
4i N4 1



R 1 cos [ 1i (1)] M1 1



R2 cos [2i (L2 )]M2 1

The prime indicates derivative with respect to the the axial coordinate x1 or x2 . The

discretized eigenvalue problem is obtained by setting g = 0 in equation (3.36). The di-

mensionless natural frequencies (imaginary part of the eigenvalues) are related to the

dimensional ones by s
E1 Isy1
= (3.38)
s1 L41

3.5 Response Analysis

Modal analysis precedes response analysis. The extended operator formulation permits

the use of modal analysis for gyroscopic systems [6163]. Modal analysis is performed for

77
the discretized eigenvalue problem corresponding to equation (3.36), which is written in

the discretized state space form [62] as

A + B=0. (3.39)

Here s are imaginary eigenvalues and s are complex state eigenvectors. and appear

in complex conjugate pairs. With n as the eigenvectors of the discretized eigenvalue

problem,
     
Mqr 0 Gqr Kqr r r
A = , B = , = . (3.40)
0 Kqr Kqr 0 r
R
With the inner product h1 , 2 i = T1 2 dD defined over the state space, the orthonor-
D
mality conditions are

hAq , r i = qr , hBq , r i = q qr (3.41)

The normalized eigenvectors are collected from the normalized state eigenvectors as in

(3.40). The inner product between two state variables x and y is defined as

< x, y >= (x1 , y1 ) + (x2 , y2 ) (3.42)

where xi and yi are the elements of x and y, and the parentheses indicate the usual inner

product between the extended variables (equation (3.33)). With this, the operator A is

symmetric and B is skew-symmetric. The equations of motion (3.36) can also be cast in

the discretized first order state space form as

Aut + Bu=F (3.43)

where
   
ar gr
u = F = .
ar 0

78
The response is assumed to be of the form [63, 64]
N
X
u= [k (t)VRk + k (t)VIk ] (3.44)
k=1

where k (t) and k (t) are modal co-ordinates and VRk and VIk are the real and imaginary

parts of the eigenvectors obtained by the modal analysis of the eigenvalue problem (3.39).

Substituting (3.44) into (3.43) and introducing the modal damping ratio k , the decoupled

equations are [63]

k + 2k k k + k2 k = jFI + k FR (3.45)

k + 2k k k + k2 k = jFR k FI . (3.46)

Here, FR (=<F,VRk >) and FI (=<F,VIk >) constitute the discretized modal forces and

is the excitation frequency. When calculating the response due to the loaded STE given by

equation (3.37), the time varying mesh deflection is represented by e(t) = E0 eipt . E0 is the

amplitude of the tooth deflection and p is the mesh frequency at the gear mesh, i.e. = p,

which is a multiple of the tooth passing frequency.

3.6 Results and Discussions

The shafts are considered to be of uniform circular cross-section although the method

developed is applicable to shafts of any cross section having an axis of symmetry. The

disks are circular with uniform thickness, with their centers of mass coinciding with the

axes of the corresponding shafts. The nominal parameter values are given in Table 3.2 and

the full scale model of the system is given in Fig. 3.4.

79
Table 3.2: Dimensional and dimensionless parameters of the coupled disk-shaft system
Masses (kg) m1 = 3.00, m2 = 3.00

Moments of Inertia (kg m2 ) Im1 = 6.34 103 , Im2 = 8.44 103 ,

Icz1 = 3.17 103 , Icz2 = 4.22 103 ,

Icy1 = 3.17 103 , Icy2 = 4.22 103

Lengths (m) L1 = 0.045, L2 = 0.045, R1 = 0.065,

R2 = 0.075, h1 = 0.04, h2 = 0.04, hc1 = 0.02,

hc2 = 0.02, d1 = 0.045, d2 = 0.045

Elastic moduli (N/m2 ) E1 = 200 109 , E2 = 200 109 ,

G1 = 84 109 , G2 = 84 109

Stiffness (N/m or N m/rad) km = 47 106 , KT = 10 103

Density (kg/m3 ) = 7800

Dimensionless rigidities K1 = 0.84, K2 = 0.84, K6 = 1.00,

K7 = 1.00, K8 = 1.00

Dimensionless inertias K3 = 0.125, K4 = 0.125, K5 = 1.000

80
 Gear 1

Gear 2

Figure 3.4: Full scale model of the coupled spinning gear-shaft system.

4.5
Mode 12 V2
B
4 Mode 9
V1 Mode 8
Natural Frequency (Imag ( ))

3.5
Modes 10,11
A
3
C( = 0.2167)
2.5 mode 7

2
Critical Speed
1.5 ( = 0.3578) Mode 2

1 Mode 6
Mode 4
0.5
G
Modes
3,4 0
0 0.5 1 1.5 2
Non-dimensional Rotation Speed ()

Figure 3.5: Natural frequency (dimensionless) sensitivity to the rotation speed () for 2
to 12 .

81
 0.25

0.2

0.15

Real ()
0.1

0.05

0
0 0.5 1 1.5 2
Non-dimensional Rotation Speed ( )

Figure 3.6: Dimensionless real parts of the eigenvlues vs the rotation speed () for 2 to
12 .

3.6.1 Sensitivity of the Natural Frequency to Various System Param-

eters

In equation (3.36), symmetric mass and stiffness matrices but skew symmetric gyro-

scopic matrices signify that the eigenvalues can be complex quantities. The system is

stable when the real parts of the eigenvalues are zero. For the system described in Table

3.2, effects of various system parameters on the natural frequency, (imaginary part of an

eigenvalue) of the modes are studied. Of particular interest is the rotation speed, . In Fig.

3.5, 2 to 12 are plotted against the non-dimensional rotation speed of the first shaft. Fig.

3.6 depicts the real values of against . Hence, it can be conluded that for values of

not extremely high, the system is stable. The 1 is the rigid body mode. Natural frequency

veering and splitting is identified in Fig. 3.5. Veering is apparent in the interaction of the

82
 4 25
BE(1st)
3.98 BE(2nd)

Natural frequencies of 8, 9

bending energies of mode 8

3.96 20

NonDimensional modal
3.94 Mode 9
3.92 15

3.9

3.88 10
Mode 8
3.86

3.84 5

3.82

3.8 0
0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15 0.2
Nondimensional Rotational Speed () Nondimensional Rotational Speed ()

(a) (b)

0.08

NonDimensional modal mesh energies

12 Mode 9
0.07 Mode 8
10
bending energies of mode 9

st 0.06
BE(1 Shaft)
NonDimensional modal

nd
BE(2 Shaft)
8 0.05

of modes 8 and 9
0.04
6

0.03
4
0.02

2
0.01

0 0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0 0.05 0.1 0.15 0.2
Nondimensional Rotational Speed () Nondimensional Rotational Speed ()

(c) (d)

Figure 3.7: Dimensionless natural frequency and the normalized modal bending and mesh
energy variations of the modes 8 and 9 with the dimensionless rotation speed . (a) Natural
frequencies 8 and 9 (zoomed in view from Fig. 3.5) (b) Bending energies of mode 8 (c)
Bending energies of mode 9 (d) Mesh energies of modes 8 and 9.

natural frequencies 8 to 12 (shown within the boxes V1 and V2 in Fig. 3.5). In the range

0.13 0.14, 8 and 9 repel each other (Fig. 3.7(a)), whereas 11 and 12 repel

each other in the range 0.018 0.021 (within the box V2 in Fig. 3.5). Figs. 3.7(b)

to 3.7(d) show the bending and mesh energies for the associated vibration modes of 8

and 9 . These modal energies are calculated from the eigenvectors, which are normalized

83
using modal analysis. Hence, different modal energies are also normalized and compari-

son between them is valid. A dramatic exchange of modal properties between the flexural

modes of the shafts associated with these two modes is observed, as they veer away from

each other. For = 0.136, a strong flexural-mesh modal coupling exists between these

two modes, with both the shafts having significant flexural vibration. Similar phenomena

is observed for the pairs 11 , 12 for lower values of .

Mode splitting is observed at low speeds (locations A, B, and G) as shown in Fig. 3.5.

What separates the degenerate modes 3 and 4 at G is the larger amount of flexural vibration

for 4 in the x1 z plane, which leads to the eigenvalue and eigenfunction splitting for

6= 0. Increase in the natural frequency of 4 with speed can be attributed to the fact

that as is increased, flexural energies of the shafts along with the torsional mesh energy

decrease, while the mesh energy increases. This results in more coupling among the shafts,

thus increasing the overall stiffness. For 3 , flexural energy of the second shaft as well

as the torsional mesh energy increases while the mesh energy decreases rapidly with .

In other words, the second shaft tends to vibrate in an uncoupled manner without much

affecting the vibration of the first shaft. As is increased further, the rotational stiffness

component takes over and reduces the overall stiffness. This continues until the first critical

speed for which = 0. At the location A (Fig. 3.5), two degenerate modes 7 and 8

coincide for = 0 at which point 8 has non-zero mesh deflection and higher flexural

energy. With 6= 0, both the flexural and mesh vibrations decrease due to stiffening

effect (i.e., one mode of vibration acts in a manner so as to decrease the amplitude of

the other) and reach a steady value. Hence, with increasing , increases for 8 . For

7 at A, mesh deflection is zero with a small amount of flexural vibration. With 6= 0,

higher flexural vibration leads to mesh vibration and vice versa, resulting in softening of

84
the overall system. The regions of flutter instability (real parts of the eigenvalues 6= 0) exist

as a result of interaction between flexural and mesh vibrations at high rotational speeds.

Variation of 6 with k m for different

0.5
=0
= 0.25
0.45
= 0.5

Natural frequency of 6 0.4

0.35

0.3

0.25

0.2

2 3 4 5 6 7 8
log10 ( k m )

Figure 3.8: Natural frequency (dimensionless) sensitivity of 6 to the mesh stiffness (km )
(N/m) in log scale.

Natural Frequencies are sensitive to other system parameters as well, such as mesh

stiffness (km ), although this is not as prominent as with . Fig. 3.8 shows the variation of

6 with km , plotted in a semi-log fashion to capture a wide range of the mesh stiffness. All

other parameters are kept at their nominal values. The 6th mode is chosen as it is a coupled

mode and is not influenced by other modes, over a wide range of parameter variations.

From Fig. 3.5, 6 increases monotonically with till over 1.0. Accordingly, values of 6

are increasingly higher for higher speeds as in Fig. 3.8. In all the cases, significant changes

in 6 appear in the range 10 107 (N/m) to 10 108 (N/m). The sharp change in the semi-

log plot specifies a continuous change in the normal plot. Increase in the natural frequency

can be explained by the variation of the strain energies.

85
 Variation of Mesh Energies with k m for 6
Variation of Bending Energy with k m for 6 0.9
8 Mesh Energy = 0

Non-Dimensional mesh energy of 6

0.8
= 0.25
7
0.7 = 0.50

6
bending energies (xz) of
Torsional Mesh Energy = 0

Non-Dimensional modal
6
0.6 = 0.25
5 = 0.50
0.5
st
1 shaft = 0
4
= 0.25 0.4

3
= 0.50
0.3
2nd shaft = 0
2 = 0.25 0.2
= 0.50
1 0.1

0 0
2 3 4 5 6 7 8 2 3 4 5 6 7 8
log10 (km ) log10 (km )

(a) (b)

Figure 3.9: Normalized dimensionless modal energy variations of the mode 6 at various
dimensionless rotational speeds () with the mesh stiffness (km ) (N/m) in log scale. (a)
Bending energies of both the shafts (b) Mesh and torsional mesh energies.

Fig. 3.9 depicts the variations of the normalized modal bending and mesh energies

of 6 with km . The bending or mesh energies do not vary much for lower km . Bending

energy reduces drastically for higher values of mesh stiffness. Mesh energy, on the other
i
hand, increases for higher km . Note that modal mesh energy. Also, from the mode
km
shapes, the torsional vibrations of the shafts increase from lower to higher values of km and

act so as to compress the mesh spring together, resulting in stiffening of the overall system.

Due to these two contributing factors, increases for higher km (Fig. 3.8). Torsional mesh

energy follows the same pattern as the bending energy, as deflection of the torsional spring

results from the relative flexural deflections of the shafts.

Natural frequencies are influenced by the torsional mesh stiffness (KT ) as well. Fig.

3.10 depicts the variation of 6 with KT for two different values of the flexural rigidities

of the second shaft (K8 ). Fig. 3.11 shows the corresponding modal strain energies. For

K8 = 1.0, increase in the natural frequency (Fig. 3.10(a)) in the range of KT from 1e5

86
 Variation of 6 with K T for different for K8 = 1.0 Variation of 6 with K T for different for K8 = 10.0
1.6 1.6

=0 =0
1.4 = 0.25 1.4 = 0.25
= 0.50 = 0.5

6
Natural frequency of 6

Natural frequency of
1.2 1.2

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2
2 3 4 5 6 7 8 2 3 4 5 6 7 8
log10( K T ) log10( K T )

(a) (b)

Figure 3.10: Natural frequency (dimensionless) sensitivity of 6 to the torsional mesh stiff-
ness (KT ) (Nm/rad) in log scale at different K8 . (a) K8 = 1.0 (b) K8 = 10.0

to 1e7 (Nm/rad) can be attributed to the fact that the flexural vibrations of the shafts are

in opposite phase in the x1 z plane and they increase with higher KT . This increases

the deflection of the torsional mesh spring and thereby making the mode shape a coupled

flexural mesh mode, which is evident from Figs. 3.11(a) and 3.11(b). This results in

an increase in the overall stiffness and higher 6 . For KT > 1e7 (Nm/rad), the slopes

of the shafts at the disk ends are small even though the deflections themselves are high,

reducing the deflection of the torsional mesh spring. For 6= 0, for small KT , flexural

deflections in the x1 y plane are of the same order of magnitude as in the x1 z plane

due to the gyroscopic effect. As KT is increased, flexural vibration in the x1 z plane

becomes dominant involving both the shafts. This increases the stiffness of the system

with corresponding increase in for higher .

For K8 = 10, flexural rigidity of the second shaft in the x1 z plane is 10 times higher

than that of the first shaft. The flexural modal energy of 6 for the second shaft is much

87
 Variation of Bending Energy of 6 with KT for K8 = 1.0 Variation of Mesh Energy of 6 with K T for K8 = 1.0
8 25

Non-Dimensional mesh energies of 6

1st shaft = 0 Mesh Energy = 0
7 = 0.25 = 0.25
Non-Dimensional modal bending

= 0.50 20 = 0.50
6 Torsional Mesh Energy = 0
2nd shaft = 0
6

= 0.25 = 0.25
energies (xz) of

5
= 0.50 15 = 0.50

10
3

2
5
1

0 0
2 3 4 5 6 7 8 2 3 4 5 6 7 8
l og10 (K T ) log10(K T )

(a) (b)
Variation of Bending Energy of 6 with K T for K8 = 10.0 Variation of Mesh Energy of 6 with KT for K8 = 10.0
70 45
Mesh Energy = 0
Non-Dimensional mesh energies of 6

40 = 0.25
Non-Dimensional modal bending

60
= 0.50
1st shaft = 0 35
Torsional Mesh Energy = 0
50 = 0.25
= 0.25
energies (xz) of 6

= 0.50 30
nd = 0.50
40 2 shaft = 0 25
= 0.25
30 = 0.50 20

15
20
10

10
5

0 0
2 3 4 5 6 7 8 2 3 4 5 6 7 8
log10 (K T ) log10 ( KT )

(c) (d)

Figure 3.11: Dimensionless normalized modal bending and mesh energies of mode 6 at var-
ious dimensionless rotational speeds () with the torsional mesh stiffness (KT ) (Nm/rad)
in log scale. Two different flexural rigidities of the 2nd shaft are chosen. (a),(b) K8 = 1.0
(c),(d) K8 = 10.0

88
higher than the first shaft (Fig. 3.11(c)), at least for small KT . As KT is increased, energy

is distributed to the first shaft in the form of flexural energy. The torsional mesh energy is

reduced for very high KT (Fig. 3.11(d)) as the shafts tend to become parallel at the disk

ends.

The sensitivity of natural frequencies and vibration modes indicate how system param-

eters affect the free vibration characteristics of the system. On the other hand, knowledge

of the veering properties is helpful to trace any dramatic change of vibration modes with

respect to the variation of system parameters.

3.6.2 Response due to Loaded Static Transmission Error

This section considers excitation due to the loaded static transmission error and its re-

sponse as discussed earlier. The variable mesh deflection is approximated by a mean mesh

deflection with a sinusoidal variation with time. The response is defined as the maximum

vibration amplitudes of flexure and torsion. For the discretized multi degree of freedom

system, response is calculated using the modal superposition method. The amplitude of the

excitation is taken as unity (E0 = 1) and the modal damping ratio is 4%. The flexural and

mesh responses versus the excitation frequency (p) are given in Figs. 3.12 and 3.13. The

excitation frequency (p) is the mesh frequency and is a multiple of the tooth passing fre-

quency. For each of these plots, three peaks are identified. They are closely approximated

by the single mode approximations using modes 2, 4, and 5 (e.g., flexural response of the

1st shaft (Figs. 3.12(a) and 3.12(b))) or matched approximately by a combination of these

modes (e.g., torsional mesh response (Fig. 3.13(b))). The third peak (p 0.287) for all

these plots, however, are matched quite well by the single mode approximation of mode 5.

Resonance occurs when the mesh frequency is close to the natural frequency of a particular

89
 10 5
Total
15 Total 10
Mode 2
Mode 2
15 Mode 4
20 Mode 4

Response amplitude (dB)

Response amplitude (dB)

Mode 5
Mode 5 20
25
25
30
30
35
35
40
40
45
45
50 50

55 55

60 60
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Dimensionless excitation frequency (p) Dimensionless excitation frequency (p)

(a) (b)

5 5
Total Total
10 10
Mode 3 Mode 5
15 Mode 5 15
Response amplitude (dB)

Response amplitude (dB)

20 20

25 25

30 30

35 35

40 40

45 45

50 50

55 55

60 60
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Dimensionless excitation frequency (p) Dimensionless excitation frequency (p)

(c) (d)

Figure 3.12: Dimensionless force response of the shafts (dB) due to the loaded static trans-
mission error (with unit amplitude) at the gear mesh. The maximum deflections of the
centroid of the disks are shown. The damping ratio () = 4%. (a),(b) Flexural response
of the 1st shaft perpendicular to and in plane of the longitudinal spring. (c),(d) Flexural
response of the 2nd shaft in the same two planes.

mode. From the natural frequency plot (Fig. 3.5), natural frequency of mode 5 reaches

0.287 at 0.280. Modal energy variations of the 5th and 6th modes with are shown

in Fig. 3.14. From Fig. 3.14(a), 5 is seen to be a coupled flexural-mesh mode at

0.280 unlike other modes at the same (e.g., Fig. 3.14(b) for 6 ). Hence for 5 , vibration

at the mesh excites the other degrees of freedom (e.g. bending) of the system by similar

proportion. For 2 and 4 , rotational speeds corresponding to the peak frequencies of p

90
 6 0.7
Total Total
Mode 5 Mode 5
0.6 Mode 2
Response amplitude (dB) 5

Response amplitude (dB)

Mode 4
0.5 Mode 2,3,4
4

0.4
3
0.3

2
0.2

1
0.1

0 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Dimensionless excitation frequency (p) Dimensionless excitation frequency (p)

(a) (b)

Figure 3.13: Dimensionless force response at the mesh due to the loaded static transmission
error (with unit amplitude) at the gear mesh. The damping ratio () = 4%. (a) Mesh
deflection (b) Torsional mesh deflection.

0.7 0.6
Dimensionless normalized modal energies

Dimensionless normalized modal energies

Bending (1st) Bending (1st)

Bending (2nd) Bending (2nd)
0.6 Mesh 0.5 Mesh

0.5
0.4

0.4
0.3
0.3

0.2
0.2

0.1
0.1

0 0
0 0.05 0.1 0.15 0.2 0.25 0 0.05 0.1 0.15 0.2 0.25
Nondimensional Rotational Speed () Nondimensional Rotational Speed ()

(a) (b)

Figure 3.14: Dimensionless normalized significant modal energy variations of the modes 5
and 6 vs. the dimensionless rotation speed (). (a) Mode 5 (b) Mode 6.

91
 0.154 and 0.201 are at 0.104 and 0.230 respectively. Both mesh and flexural

deflections at these speeds are small. Hence, a combination of these modes along with

other modes is needed to match the corresponding peaks (Fig. 3.13(b)).

92
 CHAPTER 4

DYNAMICS OF BELT PULLEY SHAFT SYSTEM

In this chapter, the model of a single span of a belt in a serpentine belt drive system

along with the attached pulleys and the shafts is developed. The hybrid discrete-continuous

system consists of the belt modeled as a combination of translational and torsional springs

while the shafts are continuous having flexural and torsional compliances. Galerkins dis-

cretization method is applied to discretize the system equations. Campbell diagram for this

gyroscopic system and response due to the excitation on one the shafts is determined.

4.1 Introduction

Belt drive systems come under the broader context of axially moving media, which can

be power transmission chains, band saws in the wood and metal industries, magnetic tapes,

and even pipes conveying flowing fluids. Vibration of the axially moving elements in these

systems can be critical for the operation, e.g., large transverse vibration of the band results

in poor cutting accuracy and surface quality in the wood industry. High level noise propa-

gating from the supporting structures or directly from vibrating band itself is a major issue.

Alspaugh [65] investigates the torsional vibration of a thin rectangular strip translating at

a constant speed in the longitudinal direction. Torsional buckling of the strip is predicted

by the frequency-load diagram. Mote [66] analyzes small vibrations of a moving band saw

93
theoretically, showing that the flexural natural frequencies always decrease continuously

from a maximum at the zero speed to the zero value with increasing velocity. The rate of

decrease depends on the relative motion between the band and the pulley axes. Fixed pul-

ley axes results in constant band tension and a rapid decrease in the natural frequency with

speed. On the other hand, if the band is allowed to extend under dynamic load, the natural

frequency decreases less rapidly. Torsional vibration of the shaft of a belt drive trans-

mission in a precision machine tool is investigated by Mashinostroeniya [67]. The belt is

represented as a weightless link possessing elastic and damping properties. The equations

of motion consist of the torsional equations of motion of the two pulleys. The parameters

are determined from the amplitude-phase frequency characteristics (APFC) plots obtained

on the test rig. Experiments on precision boring machines show that the amplitude of the

tool vibration are caused by disturbances derived from the drive and amount to 20-30 % of

the total relative vibration amplitude between the tool and the workpiece.

Belt tension fluctuation in an automotive serpentine belt drive is caused by the vibration

of the crankshaft and accessory pulleys. Tension fluctuations of the belts may give rise to

large amplitude belt span vibrations. Three kinds of span vibrations are identified: (a)

extensional, (b) transverse, and (c) torsional. Transverse vibration can be uncoupled from

the other two kinds [44, 45]. Hence, for case (a), it is sufficient to model the belt span

as springs having extensional vibrations only. Hawker [41] studies the effect of angular

vibrations resulting from the extension and relaxation of the belt strands connecting the

accessory pulleys. Three pulleys including the tensioner pulleys and two belt spans are

considered in the analysis. Each belt span is assumed as a linear spring with individual span

stiffness. Experimental validation of the mathematical model is achieved by comparing the

natural frequencies, mode shapes, forced response, and dynamic belt tension. Barker et

94
al. [42] consider two kinds of belt stiffness: (a) open span belt stiffness and (b) belt stiffness

along the arc of contact over the pulleys. The mathematical model captures the pulley

responses and belt tension force. Numerical examples presented show good agreement

with the experimental data obtained by simulating various accessories being turned off or

on. Belt slippage is considered in the mathematical model and the exact value is calculated

experimentally. Barker and Yang [43] analyze a simplified model with no slip with the

belt modeled as massless linear springs with viscous dashpots. The equations of motion

are formulated as a base excitation problem for the torsional vibrations of the pulley inertia

considering the belt and shaft stiffness. The predominant mode of vibration for the belt

is observed to be the one in which axial deformations are more significant than either the

transverse or torsional displacement.

Both the transverse and torsional vibrations are considered by Beikmann et al. [46, 47]

in his three pulley serpentine belt drive system. The transverse vibrations of the two spans

adjacent to the tensioner are identified to capture their coupling with the tensioner rotation.

The span away from the tensioner, in between the fixed pulleys, is modeled as an axially

moving string. It is noted that the torque fluctuations in the crankshaft that excite the

torsional response may also excite large transverse belt response due to belt stretching.

The natural frequencies obtained by considering the torsional and transverse coupling are

slightly higher than those obtained without considering it. Zhang and Zu [68] use the

previous model for the modal analysis and subsequent response calculation. The methods

of Meirovitch [64] and Wickert and Mote [69] are followed for the modal analysis. Instead

of using the iteration method, an explicit characteristic equation for the natural frequencies

is obtained. Kong and Parker [49] use the same model as Beikmann with the additional

feature of including the flexure of the span away from the tensioner to show that some

95
amount of coupling exist between the third span and the pulley rotations. The longitudinal

stiffness is assumed to be much greater than the transverse stiffness and hence the inertia

terms for the longitudinal motion are neglected. Response of the nonlinear belt drive system

is studied by Leamy and Parkins [45] and Hwang et al. [44]. Nonlinearities appear due to

large tensioner arm motions and Coulomb damping.

Abrate [50] discusses the various issues in a single span of a belt. It is shown that the

modeling of the belt as a string depends on several combination of parameters, e.g., bending

rigidity, initial tension, and span length. While increase in the bending rigidity invalidates

the string assumption, increase in initial tension and length of the span validates it.

In the present work, effect of the vibration of the shaft-pulley combination on the ten-

sion fluctuation of the attached span of the belt is studied. The extensional and torsional

vibrations of the belt span are considered. This assumption is valid when the belt is con-

sidered to be thin and wide, with large initial tension and span length. The shafts attached

to the pulleys are considered to be compliant in flexure and torsion. Hence, it is a hybrid

discrete-continuous system where the pulleys with the belt element constitute the discrete

part and the shafts form the continuous part. Extended operator formulation [51] is an ef-

fective means for analyzing such systems. Gyroscopic effects are generally considered for

large rotational speeds of rotor systems [19, 21, 40]. Belt transmission pulleys generally

have short shafts. Higher order beam theories are generally used for short shafts. However,

as a first attempt towards developing an analytical model of a gyroscopic belt-pulley-shaft

system, simplified beam theory (Euler-Bernoulli beam theory) is applied. The objectives

of the present study are:

(a) Build an analytical linear gyroscopic model of the single span of belt of a serpentine

belt drive system held in the reversed wrap configuration [41]. The pulleys are modeled as

96
rigid disks connected by the belt span modeled as translational and rotational springs, and

the shafts have flexural-torsional compliance.

(b) Develop a methodology to solve the eigenvalue problem of such a system using

Galerkin discretization. Obtain the response of the belt in terms of tension fluctuation

when the system is subject to a sinusoidally varying excitation force applied on one of the

shafts in the plane of the belt.

4.2 Modeling and Equations of Motion

u1 y1

1 L1
h1
1
v1 1
u2 y2
z1 x1
2
z1
kb

v2 2
KTz
z2
x2
C

L2
z2
h2 2
(a) (b)

Figure 4.1: (a) Model of the coupled belt-pulley-shaft system (b) Pulley tilting causing
torsional belt deflection.

The model consists of two rotating rigid cylindrical pulleys on parallel shafts connected

by a belt span. The model can be viewed as a fundamental building block of a serpentine

belt drive. To consider the effect of shaft flexibility, the whole belt drive needs to be con-

sidered. This is because in real cases, vibration (transverse) of one of the belt spans might

97
be influenced by the rotational vibration of pulleys not adjacent to the span. Taking small

number of pulleys might give a qualitative idea, but is still insufficient to predict the dy-

namic behavior of the whole system. Even though a single span of a belt between two

pulleys may not physically exist, it is sufficient to demonstrate a methodology that can be

applied to the whole system. In this, the belt is modeled as a combination of longitudinal

and torsional springs (Fig. 4.1(a)) with the axis parallel to the plane of the pulleys. The

z1 and z2 axes are parallel to the direction of belt motion. The torsional spring comes into

picture due to tilting of the pulleys, which causes the belt twist (Fig. 4.1(b)). A further

simplification is made by assuming that the parallel shafts are Euler-Bernoulli beams for

flexural vibration, even though they can be short. The torsional deformations are mod-

eled using first order torsion theory. The bearings are assumed to provide fixed support

for bending but no constraints for torsion. The pulleys are modeled as rigid disks. The

flexural vibrations of the shafts are distinguished for the coupled (in the plane of the belt

motion) and uncoupled planes with the displacements v and u, respectively. The following

98
non-dimensional variables are defined (See Table 4.1 for definitions of the symbols.)
s
E1 Isy1 L2 C x1 x2 u1 u2
t = t 4
, L1 = 1, L2 = ,C = , x1 = , x2 = , u1 = , u2 = ,
s1 L1 L1 L1 L1 L1 L1 L1

v1 v2 uc1 uc2 vc1 vc2 h1

v1 = , v2 = , uc1 = , uc2 = , vc1 = , vc2 = , h1 = ,
L1 L1 L1 L1 L1 L1 L1

h2 hc1 hc2 R1 R2 m1 m2
h2 = , hc1 = , hc2 = , R1 = , R2 = , m1 = , m2 = ,
L1 L1 L1 L1 L1 s1 L1 s1 L1

Icy1 Icy2 Icz1 Icz2 Im1

Icy1 = , Icy = , Icz = , Icz = , Im = ,
s1 L31 2
s1 L31 1
s1 L31 2
s1 L31 1
s1 L31

Im2 kb KT z G1 J1 G2 J2
Im2 = , k b = , K T z = , K 1 = , K 2 = ,
s1 L31 E1 Isy1 /L31 E1 Isy1 /L1 E1 Isy1 E1 Isy1

J1 J2 s2 Isz1 E2 Isz2 E2 Isy2

K3 = , K 4 = , K 5 = , K 6 = , K 7 = , K 8 = .
s1 L21 s1 L21 s1 Isy1 E1 Isy1 E1 Isy1
(4.1)

The bearings impose the following boundary conditions for fixed support for bending

and free support for torsion:

v1 u1 1
v1 = 0, = 0, u1 = 0, = 0, = 0, for x1 = 0 (4.2)
x1 x1 x1
v2 u2 2
v2 = 0, = 0, u2 = 0, = 0, = 0, for x2 = 0. (4.3)
x2 x2 x2
Rt
Hamiltons principle, (T V + W ) dt = 0 is employed to derive the non-
0
dimensional governing equations and the equations of motion of the system. For this,

the kinetic and strain energies are written in the integral form first. Variational calculus and

99
Table 4.1: List of symbols defining belt-pulley-shaft system. Subscripts 1 and 2 denote
the quantities belonging to the first and the second shafts, respectively. Overbar indicates
dimensional quantities, otherwise it is non-dimensional.

t Time
1 , 2 Rotational speeds of the shafts.
L1 , L2 Lengths of the shafts
s1 , s2 Mass per unit length of the shafts
u1 , u2 Flexural deflections of the shafts in the uncoupled plane
uc1, uc2 Flexural deflections of the centers of mass of the disks
in the uncoupled plane
v1 , v2 Flexural deflections of the shafts in the coupled plane
vc1 , vc2 Flexural deflections of the centers of mass of the disks
in the coupled plane
1 , 2 Torsional deflections of the shafts about their respective axes
R1 , R2 Radii of the disks
h1 , h2 Thickness of the disks
hc1 , hc2 Distances of the centers of mass of the disks from the left hand end of the disks
C Center distance between the pulleys.
m1 , m2 Masses of the disks
Icy , Icz Mass moments of inertia of the disk with respect to
the y and z axes
Im Mass moments of inertia of the disk with respect to the x axis
Isy , Isz Area moments of inertia of the shaft cross section with respect to
the y and z axes
J1 , J2 Polar moments of inertia of the shaft cross-sections
kb Longitudinal belt stiffness
KT z Torsional belt stiffness about the direction of the belt travel
K1 , K2 Torsional rigidities of the shafts
K7 , K8 Flexural rigidities of the second shaft in the uncoupled
and the coupled planes
i ith mode
i Natural frequency of i

100
boundary conditions above are employed next. The non-dimensional kinetic energy
T
T =
E1 Isy1 /L1
Z1 Z1
1 u1 2 1 v1
= ( 1 v1 ) dx1 + ( + 1 u1 )2 dx1
2 t 2 t
0 0
L
1 Z 2 u ZL2
2 2 1 v2 2

+ K5 ( + 2 v2 ) dx2 + ( 2 u2 ) dx2
2 t 2 t
0 0
Z1  2 ZL2  2
1 1 1 2
+ K3 dx1 + K4 dx2
2 t 2 t
0 0
1 uc1 1 1 vc1 1
+ m1 [ + (1 + hc1 ) 1 vc1 ]2 + m1 [ + (1 + hc1 ) + 1 uc1]2
2 t t 2 t t
1 uc2 2 1 vc2 2
+ m2 [ + (L2 + hc2 ) + 2 vc2 ]2 + m2 [ + (L2 + hc2 ) 2 uc2]2
2 t t 2 t t
1 1 2 1 1 2 1 2 2
+ Icy1 [1 1 + ] + Icz1 [1 1 ] + Icy2 [2 2 ]
2 t 2 t 2 t
1 2 2 1 1 1 2
+ Icz2 [2 2 + ] + Im1 [1 + (1)]2 + Im2 [2 (L2 )]2 .
2 t 2 t 2 t
(4.4)
v1 v2 u1 u2
1 (1), 2 (L2 ) and 1 (1), 2 (L2 ) are the slopes of the shaft
x1 x2 x1 x2
elements at the pulley locations. The non-dimensional strain energy
V
V =
E1 Isy1 /L1
Z1  2 2 Z1  2 2 ZL2  2 2
1 v1 1 u1 1 v2
= 2
dx1 + K6 2
dx1 + K8 dx2
2 x1 2 x1 2 x22
0 0 0 (4.5)
ZL2  2
2 Z1  2 ZL2  2
1 u2 1 1 1 2
+ K7 dx2 + K1 dx1 + K2 dx2
2 x22 2 x1 2 x2
0 0 0
1 1
+ KT z (1 2 )2 + km {R1 1 (1) + R2 2 (L2 ) vc1 + vc2 }2 .
2 2
The last two terms in the equation (4.5) are the strain energies stored in the springs. In

a serpentine belt drive, excitation comes on the crankshaft in the form of torque fluctuation

or a periodic lateral force. In order to demonstrate the effect of the lateral flexibility of

101
the shaft, external excitation in the form of a periodic force is considered. Assuming this

lateral force f is applied on the first shaft in the coupled plane at a distance of x0 from the

bearing end (Figure 4.2), the virtual work done due to this force is given as

Z1
W = f (x1 x0 )v1 (x1 , t)dx1
(4.6)
0
= f v1 (x0 , t).

The resulting non-dimensional governing equations for coupled bending and torsion of the

shafts are
4 u1 2 u1 v1
K6 4
+ 2
21 21 u1 = 0 (4.7)
x1 t t
4 v1 2 v1 u1
4
+ 2 + 21 21 v1 = 0 (4.8)
x1 t t
2 1 2 1
K3 K 1 =0 (4.9)
t2 x21
4 u2
 2 
u2 v2 2
K7 + K5 + 22 2 u2 = 0 (4.10)
x42 t2 t
4 v2
 2 
v2 u2 2
K8 + K5 22 2 v2 = 0 (4.11)
x42 t2 t
2 2 2 2
K4 K 2 =0 (4.12)
t2 x22

Equations (4.7) (4.9) are for the first shaft, and equations (4.10) (4.12) are for the

second shaft. The two pulleys are coupled by the springs. The domain of x1 is 0 < x1 < 1

and that of x2 is 0 < x2 < L2 . (See Table 4.1 for definitions of the symbols.)

Each pulley is subject to elastic forces and moments due to bending and torsion from

the shafts. The equations governing the pulley translations for the unforced system are

2 uc1 2 1 3 u1
  
vc1 1
m1 21 21 uc1 + (1 + hc1 ) 1 K6 |x =1 = 0
t 2 t t 2 t x31 1
(4.13)

102
 2 vc1 2 1 3 v1
  
uc1 1
m1 + 21 21 vc1 + (1 + hc1 ) + 1 |x =1 Fk = 0

t 2 t t 2 x31 1
t
(4.14)
 2  2
3 u2

uc2 vc2 2 2 2
m2 + 2 2 2 u c2 + (L 2 + h c2 ) + 2 K 7 |x =L = 0
t2 t t2 t x32 2 2
(4.15)
 2  2
3 v2

vc2 uc2 2 2 2
m2 22 2 vc2 + (L2 + hc2 ) 2 K 8 |x =L +Fk = 0
t2 t t2 t x32 2 2
(4.16)

Fk = kb {R1 1 |x1 =1 + R2 2 |x2 =L2 vc1 + vc2 } (4.17)

where Fk is the longitudinal spring force. The equations governing the pulley rotations are
 2   
1 1 1 2
Icz1 2
1 Icy1 1 + 1 1 + (1 + hc1 ) m1
 2 t t t
2 1 2 u1

uc1 vc1
+ (1 + h c1 ) 1 + K 6 |x =1 (4.18)
t2 t2 t x21 1
3 u1
+hc1 K6 3 |x1 =1 + KT z (1 2 ) = 0
x1
 2   
1 1 2 1
Icy1 2
+ 1 Icz1 1 1 1 + (1 + hc1 ) m1
 2 t t t
(4.19)
2 1 2 v1 3 v1

vc1 uc1
+ (1 + h c1 ) + 1 + | x =1 + hc1 | x =1 = 0
t2 t2 t x21 1 x31 1
 2   
2 2 2 2
Icz2 2
+ 2 Icy2 2 2 2 + (L2 + hc2 ) m2
 2 t t t
2 2 2 u2

uc2 vc2
+ (L2 + hc2 ) + 2 + K7 2 |x2 =L2 (4.20)
t2 t2 t x2
3 u2
+hc2 K7 3 |x2 =L2 KT z (1 2 ) = 0
x2
 2   
2 2 2 2
Icy2 2
2 Icz2 2 2 + 2 + (L2 + hc2 ) m2
 2 t t t
2 2 2 v2 3 v2

vc2 uc2
+ (L 2 + h c2 ) 2 + K 8 | x =L + h K
c2 8 |x =L = 0
t2 t2 t x22 2 2 x32 2 2
(4.21)
2 1 1
Im1 2 |x1 =1 + K1 |x =1 + R1 Fk = 0 (4.22)
t x1 1
2 2 2
Im2 |x =L + K 2 |x =L + R2 Fk = 0. (4.23)
t2 2 2
x2 2 2

103
 Note that the translations of the pulleys are coupled with the torsional motions through

the belt spring force Fk , and the rotational speeds, 1 and 2 .

4.3 Extended Operator Formulation and Discretization

The coupled system equations can be cast in a structured form with the use of extended

operators. The system deformation is represented as the extended variable [30, 51]

w(x, t) = { u1(x1 , t), v1 (x1 , t), u2 (x2 , t), v2 (x2 , t), 1 (x1 , t), 2 (x2 , t), uc1 (t),

vc1 (t), uc2(t), vc2 (t), 1 (t), 1 (t), 2 (t), 2 (t), 1 (1, t), 2 (L2 , t)}T ,
(4.24)

which comprises of continuous and discrete degrees of freedom. The equations of motion

are written in the compact form

 
2
Mwtt + Gwt + L L w = 0 (4.25)

where M, G, L, and L are the extended mass, gyroscopic, stiffness, and rotational stiffness

operators, respectively, operating on w and = 1 , the rotation speed of the first shaft.
R1
Rotation speed of the second shaft is 2 = 1 .
R2

Mw = {u1 , v1 , K5 u2 , K5 v2 , K3 1 , K4 2 , m1 {uc1 + (1 + hc1 )1 } ,

m1 {vc1 + (1 + hc1 )1 } , m2 {uc2 + (L2 + hc2 )2 } , m2 {vc2 + (L2 + hc2 )2 } ,

Icz1 1 + (1 + hc1 )m1 {uc1 + (1 + hc1 )1 } , Icy1 1 + (1 + hc1 )m1

(4.26)
{vc1 + (1 + hc1 )1 } , Icz2 2 + (L2 + hc2 )m2 {uc2 + (L2 + hc2 )2 } ,

Icy2 2 + (L2 + hc2 )m2 {vc2 + (L2 + hc2 )2 } , Im1 1 (1), Im2 2 (L2 )}T ,

104
 Gw = {2v1 , 2u1, 2K5 v2 (R1 /R2 ), 2K5 u2 (R1 /R2 ), 0, 0, m1

{2vc1 + (1 + hc1 )1 } , m1 {2uc1 + (1 + hc1 )1 } , m2 {2vc2 + (L2 + hc2 )2 }

(R1 /R2 ), m2 {2uc2 + (L2 + hc2 )2 } (R1 /R2 ), (Icy1 + Icz1 )1

(4.27)
(1 + hc1 )m1 vc1 , (Icy1 + Icz1 )1 + (1 + hc1 )m1 uc1,

[(Icy2 + Icz2 )2 + (L2 + hc2 )m2 vc2 ] (R1 /R2 ),

[(Icy2 + Icz2 )2 (L2 + hc2 )m2 uc2 ] (R1 /R2 ), 0, 0}T ,

4 u1 4 v1 4 u2 4 v2 2 1 2 2 3 u1
Lw = {K6 , , K 7 , K 8 , K 1 , K 2 , K 6 (1),
x41 x41 x42 x42 x21 x22 x31

3 v1 3 u2
(1) k b [R1 1 (1) + R2 2 (L2 ) vc1 + vc2 ] , K 7 (L2 ),
x31 x32

3 v2 2 u1
K8 (L2 ) + k b [R1 1 (1) + R2 2 (L2 ) vc1 + vc2 ] , K 6 (1)
x32 x21

3 u1 2 v1 3 v1 2 u2
+hc1 K6 3 (1) + KT z (1 2 ), (1) + hc1 3 (1), K7 2 (L2 ) (4.28)
x1 x21 x1 x2

3 u2 2 v2 3 v2
+hc2 K7 (L2 ) K Tz (1 2 ), K 8 (L2 ) + h K
c2 8 (L2 ),
x32 x22 x32

1 2
K1 (1) + R1 kb [R1 1 (1) + R2 2 (L2 ) vc1 + vc2 ] , K2 (L2 )+
x1 x2

R2 kb [R1 1 (1) + R2 2 (L2 ) vc1 + vc2 ]}T ,

Lw = {u1 , v1 , K5 u2 (R1 /R2 )2 , K5 v2 (R1 /R2 )2 , 0, 0, m1uc1 , m1 vc1 , m2 uc2(R1 /R2 )2 ,

m2 vc2 (R1 /R2 )2 , Icy1 1 , Icz1 1 , Icy2 2 (R1 /R2 )2 , Icz2 2 (R1 /R2 )2 , 0, 0}T
(4.29)

105
The inner product between two extended variables w1 and w2 is defined as
Z1 Z1 ZL2 ZL2
(w1 , w2 ) = r1 s1 dx1 + g1 h1 dx1 + r2 s2 dx2 + g2 h2 dx2
0 0 0 0
Z1 ZL2
+ p1 q1 dx1 + p2 q2 dx2 + rc1 sc1 + gc1hc1 + rc2 sc2 + gc2 hc2
0 0 (4.30)
r1 s1 g1 h1 r2 s2
+ (1) (1) + (1) (1) + (L2 ) (L2 )
x1 x1 x1 x1 x2 x2

g2 h2
+ (L2 ) (L2 ) + p1 (1)q1 (1) + p2 (L2 )q2 (L2 )
x2 x2
where r1 , g1 , r2 , g2 , p1 , and p2 are the elements of the extended variable w1 , and s1 , h1 , s2 ,

h2 , q1 , and q2 are the elements of the extended variable w2 .

With this definition of the inner product, the operators M, L, and L are symmetric,

while G is skew-symmetric. Moreover, M is positive definite and L is positive semi-

definite. Hence, the eigenvalues can be complex quantities with non-zero real parts at su-

percritical speeds. For those speeds, the real parts of the eigenvalues are non-zero and they

appear in positive-negative pairs. Subcritical speed eigenvalues appear in purely imagi-

nary complex conjugate pairs n = jn . Corresponding eigenfunctions are also complex

conjugates. Using the separable form, w(x, t) = u(x)eit , the eigenvalue problem is
 
2 2
M + iG + L L u = 0 (4.31)

The dimensionless natural frequencies (imaginary part of the eigenvalues) are related to

the dimensional frequencies by

s
E1 Isy1
= (4.32)
s1 L41

The forced problem corresponding to equation (4.25) is given as

 
Mwtt + Gwt + L 2 L w = f, (4.33)

106
where, f is the excitation force on system. The flexural vibration of the shafts in the belt-

pulley-shaft system can affect the system vibration characteristics. In order to demonstrate

this, f is considered to be a force acting on one of the shafts. Closed-form analytical

solution of equation (4.33) is prohibitive. Galerkin discretization technique is a weighted

residual technique, which can be readily applicable to the extended operator formulation

because of the structure and the associated inner product. The required trial functions

are taken from the space of admissible functions [4]. As stated by Jha [60], stationary

system basis functions for beam vibration are used for this rotating beam analysis. With
N
P
the expansion wN (x) = aj (t)j (x) in equation (4.25), where j s are the stationary
j=1
system basis functions, Galerkin discretization yields

h i
[Mqr ] ar + [Gqr ] ar + Lqr 2 Lqr ar = g (4.34)

where

Mqr = (Mr , q ) = Mrq , Gqr = (Gr , q ) = Grq , Lqr = (Lr , q ) = Lrq ,

 
Lqr = Lr , q = Lrq , gq = (f, q )

For the present analysis, f is taken as a sinusoidal force of magnitude E0 acting at a

distance of x0 from the bearing end of the first shaft, directed along the coupled plane (Fig.

4.2). Taking 1i , 3i and 2i , 4i as the basis functions for bending in the uncoupled and

coupled planes, and 1i , 2i as those for torsion, the discretized force vector g correspond-

ing to f is given as
0

N1 1
g = E0 eipt
[2i (x0 )]N2 1 .
(4.35)
0

(N3 +N4 +M1 +M2 )1

107
 u1
1 1

v1 1
u2 x0
2
f = E e ipt
v2

L2
2

L0

Figure 4.2: Model of the coupled belt-pulley-shaft system subject to a sinusoidal force on
the first shaft in the coupled direction.

The response analysis is similar to that of the gyroscopic helical geared shaft system

(section 3.5).

4.4 Results and Discussions

Uniform circular cross-sections of the shafts are chosen although the method developed

is applicable to shafts with any cross section having an axis of symmetry. The circular disks

have uniform thickness with the center of mass coinciding with the axes of the correspond-

ing shafts. Nominal system parameter values are given in Table 4.2.

4.4.1 Sensitivity to the Rotational Speed

The Campbell diagram for this gyroscopic system is shown in Fig. 4.3 for the modes

2 to 6. 1 is a rigid body mode. As in Table 1, the only difference between the individual

shaft-disk system is the disk radii. Accordingly, in the Campbell diagram, two sets of

curves are obtained which are similar. Mode splitting is observed as is increased for

108
 Table 4.2: Dimensional and dimensionless parameters of a belt-pulley-shaft system
Masses (kg) m1 = 3.00, m2 = 3.00

Moments of Inertia (kg m2 ) Im1 = 6.34 103 , Im2 = 8.44 103 ,

Icz1 = 3.17 103 , Icz2 = 4.22 103 ,

Icy1 = 3.17 103 , Icy2 = 4.22 103

Lengths (m) L1 = 0.090, L2 = 0.090, R1 = 0.065,

R2 = 0.075, h1 = 0.15, h2 = 0.15, hc1 = 0.075,

hc2 = 0.075, d1 = 0.045, d2 = 0.045,

C = 0.500

Elastic moduli (N/m2 ) E1 = 200 109 , E2 = 200 109 ,

G1 = 84 109 , G2 = 84 109

Stiffness (N/m or N m/rad) kb = 47 103 , KT z = 10 102

Density (kg/m3 ) = 7800

Dimensionless rigidities K1 = 0.84, K2 = 0.84, K6 = 1.00,

K7 = 1.00, K8 = 1.00

Dimensionless inertias K3 = 31.25 103 , K4 = 31.25 103 , K5 = 1.000

109
the degenerate modes 3, 4 and 5, 6. For this system, not only the eigenvalues but also the

eigenvectors split at > 0. Fig. 4.4 depicts the significant normalized modal energies

of the decoupled modes 3 to 6 for subcritical speeds. The outer split modes 3 and 6 are

flexural modes of the first shaft, whereas the inner split modes 4 and 5 are flexural modes

of the second shaft. Note that, as the rotational speed approaches the critical speeds, 3 and

4 becomes more and more rigid body modes. This is recognised by the drop in the modal

energies of these two modes to zero near the corresponding critical speeds.

0.9

0.8
Natural Frequency ()

0.7

0.6

0.5
Mode 6
0.4
Mode 3
0.3 Mode 5
Mode 4
0.2
Mode 2
0.1

0
0 0.5 1 1.5 2
Rotational Speed ()

Figure 4.3: Dimensionless natural frequency () vs. the rotational speed ().

4.4.2 Force Response due to Excitation at the Shaft

This section discusses the response due to a sinusoidal force on the first shaft along the

coupled plane as discussed earlier. The force is applied on the midspan of the shaft, i.e. x0

= L1 /2 as in Fig. 4.2 and the amplitude of the force is taken as unity (E0 = 1).

110
 6

Dimensionless normalized modal energies

5

st
2 : Bending (1 shaft)
3
nd
4: Bending (2 shaft)
nd
1 : Bending (2 shaft)
5
st
6: Bending (1 shaft)
0
0.05 0.1 0.15 0.2 0.25 0.3 0.35
Nondimensional Rotational Speed ()

Figure 4.4: Dimensionless normalized significant modal energy variations of the modes 3
to 6 vs. the dimensionless rotational speed ().

Figs. 4.5, 4.6, and 4.7 shows the frequency sweep for the maximum values of the shaft

flexural, torsional and belt responses, respectively, obtained by the modal superposition of

a large number of modes. The flexural response of each shaft is divided into uncoupled

and coupled planes, whereas the belt response comprises of longitudinal and torsional belt

responses. Superposed on these plots are the single or two mode response plots of the first

few modes for better understanding of the nature of the peak responses. The peak response

for the longitudinal belt is significant at the excitation frequency p = 0.035 (Fig. 4.7(a)),

which coincides with the peak response for mode 2. From the natural frequency vs.

plot (Fig. 4.3), 2 is constant at 0.035. Also from the modal energy plot (Fig. 4.8), 2 is

primarily a logitudinal belt spring mode, unlike any other mode in the set 2 to 6 (Fig.

4.4). This explains the highest longitudinal belt peak response at p = 0.035. External

flexural excitation does not excite mode 2 flexurally much, but sets in the longitudinal belt

111
 10 10
Total Total
Mode 3 Mode 3
0 Mode 6 0 Mode 6

Response amplitude (dB)

Response amplitude (dB)

10 10

20 20

30 30

40 40

50 50
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Dimensionless excitation frequency (p) Dimensionless excitation frequency (p)

(a) (b)

30 20
Total Total
Mode 2 Mode 2
40 Mode 3
Mode 3 40
Mode 4 Mode 4
Response amplitude (dB)

Response amplitude (dB)

50 Mode 5,6 Mode 5,6
60

60
80
70

100
80

120
90

100 140
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Dimensionless excitation frequency (p) Dimensionless excitation frequency (p)

(c) (d)

Figure 4.5: Dimensionless force response of the shafts (dB) due to a cyclic load of unity
magnitude on the first shaft in the coupled plane. The maximum deflections of the centroids
of the disks are shown. The damping ratio () = 4%. (a),(b) Flexural response of the 1st
shaft in the uncoupled and coupled planes. (c),(d) Flexural response of the 2nd shaft in the
uncoupled and coupled planes.

vibration. This, in turn, initiates the torsional vibrations of the attached pulleys at the same

frequency (Fig. 4.6).

The flexural responses are plotted in Fig. 4.5, which has the peak amplitudes coinciding

with the peaks of the modes 3 to 6 . 3 , 4 , and 6 have peak responses at frequencies

corresponding to the half critical speed ( = 0.2). At this speed, 3 , 4 , and 6 are 0.103,

0.114, and 0.277, respectively (Fig. 4.3). 3 and 6 are primarily the flexural modes of the

112
 20 20
Total Total
Mode 2 Mode 2
0 Mode 3 0 Mode 3
Mode 6 Mode 6
Response amplitude (dB)

Response amplitude (dB)

20 20

40 40

60 60

80 80

100 100
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Dimensionless excitation frequency (p) Dimensionless excitation frequency (p)

(a) (b)

Figure 4.6: Dimensionless forced response of the shafts (dB) due to a cyclic load of unity
magnitude on the first shaft in the coupled plane. The maximum torsional deflections at the
disk ends are shown. The damping ratio () = 4%. (a) Torsion (1st shaft) (b) Torsion (2nd
shaft).

2 0.9
Total Total
1.8 Mode 2 0.8 Mode 3
Mode 3 Mode 6
1.6
Mode 6 0.7
1.4
Response amplitude

Response amplitude

0.6
1.2
0.5
1
0.4
0.8
0.3
0.6
0.2
0.4

0.2 0.1

0 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Dimensionless excitation frequency (p) Dimensionless excitation frequency (p)

(a) (b)

Figure 4.7: Dimensionless forced response of the longitudinal and torsional belt springs
due to a cyclic load of unity magnitude on the first shaft in the coupled plane. The maxi-
mum deflections are shown. The damping ratio () = 4%. (a) Longitudinal belt spring (b)
Torsional belt spring responses.

113
 4 Modal energies of 2 with
x 10
6

Dimensionless normalized modal energies

5
Bending (1st)
Bending (2nd)
4 Torsion (1st)
Torsion (2nd)
Lonitudinal Belt
3
Torsional Belt

0
0 0.02 0.04 0.06 0.08 0.1
Nondimensional Rotational Speed ()

Figure 4.8: Dimensionless normalized modal energy variations of the modes 2 vs. the
dimensionless rotational speed ().

shafts (in both the uncoupled and coupled planes) along with the longitudinal and torsional

belt modes (Fig. 4.4). Hence, excitation on the first shaft produces flexural vibration of

the shafts (Fig. 4.5) and aids in the longitudinal and torsional belt spring vibrations (Figs.

4.7(a), 4.7(b)) at p = 0.103 and 0.277, respectively. Torsional belt spring deflection results

from the relative deflections of the pulleys in the uncoupled plane. 4 is the flexural mode

of the second shaft (Fig. 4.4). Hence, only the second shaft gets excited at 4 (Fig. 4.5(c),

(d)).

114
 CHAPTER 5

SUMMARY AND FUTURE WORK

5.1 Summary

Analytical methods for the linear vibration analysis of geared shaft and belt-pulley-

shaft systems are developed in the present work. Until now, lumped parameter based trans-

fer matrix method or finite element method are used for this purpose, which are either

inaccurate for predicting the higher natural frequencies or non-intuitive. Both the methods

developed in this work are efficient and accurate as the basis functions are global in nature.

Discretization is not at a physical level but at an abstract level. Both the geared shaft and

belt-pulley-shaft models are the fundamental building blocks for automotive transmissions

and accessory drive systems, respectively. The summary on specific topics are stated below.

5.1.1 Eigensolution of nongyroscopic spur geared system attached to

compliant shafts

Analytical calculation of the natural frequencies and modes, modal sensitivities, and dy-

namic response of a non-gyroscopic discrete continuous geared shaft system is presented.

The complete set of governing equations for the system is derived using Hamiltons prin-

ciple. The equations of motion constitute a coupled model involving bending, torsion, and

115
gear mesh deflection. Use of extended operators acting on an extended variable (in sync

with the various system displacements) allows the governing equations to be cast in the

canonical form of the mass (M) and stiffness (L) operators. With a natural inner prod-

uct defined on the extended space, the extended inertia and stiffness matrices are shown

to be self-adjoint. Assumed modes method is used to discretize the system spatially. Ap-

propriately orthonormalized basis functions are chosen for bending and torsion for each

continuous span of the shafts. Trigonometric basis functions (eigenfunctions for pure tor-

sion of individual disk-shaft unit) are chosen for torsional and algebraic basis functions are

chosen for flexural displacements. These basis functions satisfy the boundary conditions

at the bearing ends. With Lagrange multipliers, it possible to incorporate the matching

conditions, which is not possible in other methods such as the Galerkins discretization

method.

Variation of the natural frequencies with the changes in various system parameters is

studied. One of the important parameters is the mesh stiffness. Mesh spring couples the

vibrations of the two shafts. Other important parameters are the torsional and flexural

rigidities of the shafts. From the two dimensional plots of the natural frequency with these

parameters, it is observed that some of the modes are sensitive to the changes of these

parameters while the natural frequencies of the other modes do not change at all. Modal

energy distributions for the first kind of modes are proportional to the sensitivity of the

natural frequency to these parameters.

Lengths of the shafts affect the natural frequencies of the system as lengths are directly

proportional to the compliance of the system. This influence, however, varies from mode

to mode and depends on the modal energy distribution in that mode. For example, if the

116
lengths of the shafts are reduced beyond a certain limit, mode twenty-two becomes insen-

sitive to the changes in the flexural rigidity of the second shaft.

For the loaded static transmission error, responses are calculated for both the shafts for

bending and torsion, and mesh deflection using modal superposition method. The peak

responses are matched with one, two, and three mode approximations of the dominant

modes. These responses are also predicted by observing the mode shapes.

5.1.2 Eigensolution of gyroscopic helical geared system attached to

compliant shafts

In this work, first, the mathematical model of two helical gears in mesh and mounted on

parallel rotating shafts is developed. The shafts are modeled as gyroscopic continua while

the gear pairs are modeled as rigid disks connected by mesh springs. The governing equa-

tions as well as the equations of motion, along with the boundary conditions, are derived

using Hamiltons principle. The equations of motion constitute a coupled model involving

bending, torsion, and mesh vibrations along with the terms due to the gyroscopic accel-

eration. An extended variable is chosen to incorporate both the continuous and discrete

displacements. Extended operators acting on this variable are chosen to cast the equations

in the canonical gyroscopic continua form involving inertia (M), stiffness (L), gyroscopic

(G), and rotational stiffness (L) operators. M, L, and L are shown to be self-adjoint,

whereas G is shown to be skew self-adjoint after defining a natural inner product on the

extended space. Galerkin discretization technique is employed efficiently and accurately to

transform this system into a discretized multi-degree of freedom system. Basis functions

for this global discretization are chosen to be the orthonormalized polynomial bases for

bending and eigenfunctions for torsion. Use of extended operators provides the governing

and the equations of motion with a structured form, which is conducive to the application

117
of the Galerkins method.

Natural frequency plots versus the rotation speed display several interesting traits. Some

degenerate eigenvalues at zero speed split as a result of interactions between the individual

shaft-disk systems (stiffening and softening effects). Critical speed phenomenon is ob-

served at a high rotation speed resulting from the softening due to the rotational stiffness.

Regions of flutter instability are present at sufficiently high speeds and exist as a result of

interactions between the flexural and mesh vibrations. In the operating range below all

the critical and flutter speeds, some of the natural frequencies are coupled flexural-mesh

vibrations for the range of the nominal parameter values chosen. Variation of the rotation

speed causes sharp veerings between the flexural frequencies with the associated changes

in the vibrations of the shafts. At the veering zone, the modes are coupled and significantly

different from either of the approaching modes.

Natural frequency variations with parameters such as mesh stiffness and torsional mesh

stiffness are explained using the mode shapes. If the modal vibration is such that the mesh

springs undergo large deformations, natural frequency increases.

Modal analysis is performed after casting the discretized system in the first order form

with the use of system matrices. The advantage of this approach is that, not only are the

eigenvectors and hence the modal strain energies normalized, but also the response to any

arbitrary excitation is obtained. Response due to the loaded static transmission error is ob-

tained and matched with one, two, or three mode approximations of the associated peaks.

For highly coupled modes (e.g. mode 5), single mode approximation can match the peak

response of the modal superposition quite well. For not so coupled modes, a minimum

number of modes are needed to match a peak response.

118
5.1.3 Dynamics of Belt Pulley Shaft System

The equations of motion display a coupled model involving the continuous shaft vibra-

tions for flexure and torsion, and vibration of the connecting belt. A simplified model of

the belt span is chosen consisting of extensional and torsional vibrations, which is possible

when the belt is thin and wide with large initial tension and span length. The torsional vi-

bration is about an axis along the belt motion. Accordingly, two kinds of belt stiffness are

taken into account: longitudinal and torsional. The system is a discrete continuous system

with the shafts modeled as continuous and the belt with the associated pulleys modeled as

discrete parts. An extended variable is chosen to incorporate the continuous and discrete

motions. Use of extended operators allows the equations to be cast in the canonical form of

a gyroscopic continua. Galerkin discretization technique is employed to get the discretized

form of the system, which is a natural extension of the extended operator formulation. Ba-

sis functions for this global discretization are same as the geared shaft system.

Natural frequency versus the rotation speed plot of the first six modes display interest-

ing characteristics. The degenerate eigenvalues at zero speed split as a result of interactions

due to flexural vibration between the individual shaft-pulley system. At non-zero speeds,

the modes split into flexural modes of the first and second shafts. Critical speed is observed

at a high rotation speed resulting from the softening due to the rotational stiffness.

The normalized eigenvectors after the modal analysis are utilized by calculating and

comparing the normalized modal energies of individual modes. The lower modes are de-

coupled flexural or longitudinal belt modes. Response due to a sinusoidal excitation force

acting on one of the shafts is obtained and matched with one or two mode approxima-

tions of the isolated modal peaks. The participation of different modes in the response are

determined by this method.

119
5.2 Future work

The future work is recommended for the following modified models to take care of the

vibration due to flexibilities of different components other than the shafts.

5.2.1 Eigensolution of the geared shaft systems considering bearing

flexibility

In addition to the shafts, the bearings also have some amount of flexibilities. They can

also be a secondary source of noise and vibration in a geared rotor system, other than the

shafts. Bearings act as interfaces between the stationary and rotating parts and are lubri-

cated in order to reduce friction and power loss. A typical ball bearing consists of rolling

elements, and the inner and outer races. Geometric imperfections in all the three compo-

nents cause vibration related noise at the bearings. Though the bearings can be modeled

with time dependent stiffness with nonlinearity, the simplistic bearing model consists of

linear springs along the two orthogonal directions as shown in the mathematical model of

the geared shaft systems (Fig. 5.1).

km
2
KTy

(a) (b)

Figure 5.1: Model of the gear-shaft system with bearing flexibility. (a) Spur gear system
(b) Helical gear system.

120
 Analytical vibration analyses of geared shaft systems with bearing flexibility would be

somewhat involved as the basis functions used in the original models (with infinite bearing

stiffness) are not applicable here. For example, after casting the governing equations and

the equations of motion of the gyroscopic geared rotor system in the extended operator

form, the boundary conditions that remain are the geometric boundary conditions corre-

sponding to the fixed ends of the shafts. For Galerkins method, the basis functions are

the comparison functions satisfying all the boundary conditions. As the boundary condi-
v1
tions are simple (e.g., v1 = 0, = 0), the basis functions can be chosen from a single
x1
family of functions (e.g., xi+1 for flexure) before orthogonalization. In the case of systems

with bearing flexibility, the boundary conditions at the bearing ends are natural, involving

the spring force. The bearing ends of the shafts are not fixed, but are allowed to displace.

Hence, use of the previous basis functions leads to extremely slow convergence, and for

slightly higher frequencies, no convergence at all. To ensure the convergence of the natural

frequencies, the basis functions are to be chosen as combinations from different families

(e.g., a possible basis function can be a linear combination from families satisfying only the

fixed and only the free boundary conditions). This kind of basis functions are called quasi-

comparison functions [4]. In some cases, the quasi-comparison functions are found to have

better convergence properties than even the comparison functions, if they are available.

5.2.2 Eigensolution of the geared shaft systems with flexibility of the

gear bodies

The gears are considered to be rigid in this dissertation. For practical purposes, if gears

are thin, they can deform elastically. In that case, the gears can be modeled as annular

Kirchhoff plates with bending vibration (Fig. 5.2). The governing equations for plate

vibration are fourth order equations in two dimensions, which can be derived along with

121
the beam governing equations using Hamiltons principle. The vibration modes consist of

circular (m = 0) and diametrical (n = 0) modes, or a combination of both (m, n 6= 0) for

the vibration of a single plate [4]. When these plates are connected by the mesh spring, part

of the boundary has to satisfy the natural boundary condition corresponding to the spring

force. The inner boundary of the plate is clamped and part of the outer boundary is free.

The eigenfunctions consist of Bessels functions of both the first and the second kinds.

The basis function for discretization can be taken as these eigenfunctions. Shaft vibrations

along with the gear flexural vibrations result in a coupled system with interesting dynamics

where different combination of modes are possible, e.g., plate mode, flexural or torsional

shaft mode, and spring mode.

Figure 5.2: Flexible gear body (modified from [1]).

122
5.2.3 Eigensolution of the geared shaft systems with Timoshenko beam
theory and axial vibration

For the gyroscopic geared shaft system (chapter 3), lengths of the shafts are small. For

accurate results, higher order Timoshenko beam theory is preferable. Here, effects of both

the rotary inertia and shear deformation are taken into consideration. Total deformation

of a shaft, w(x, t) is caused by both bending and shear. While bending causes the rota-

tion of a shaft element, shear causes the element to distort. The slope of the deflection
 
w(x, t)
curve consists of two parts: (a) due to bending ((x, t)) and (b) due to shear
x
((x, t)). Hence, the governing equation is a coupled equation of bending and shear, which

can be derived using suitably modified energy expressions in the Hamiltons principle.

For spur geared shaft system, the axial deformation of the shafts are negligible. Axial

deformation in the helical geared shaft system is useful in predicting the bearing forces and

moments, which would be useful in designing the bearings.

5.2.4 Effect of shaft vibration on the dynamics of continuous belt ser-

pentine belt drive system with and without considering the belt
flexural vibration

To accurately predict the dynamic behavior of an entire serpentine belt drive system, it

is essential to incorporate in the model the discrete pulley vibrations along with the con-

tinuous vibrations of the attached shafts as well as the continuous motion of the belt. The

continuous belt can be modeled as an axially moving string or a beam. Consequently, the

gyroscopic acceleration terms due to the moving belt vibration will appear in the analy-

sis. Belt-pulley coupling with the belt modeled as a moving beam has been studied pre-

viously [49]. Addition of shaft flexibility would be an interesting coupling problem for a

practicing engineer.

123
 APPENDIX

Terms in equation (2.54)

M1 = diag[IN1 N1 + M1Lxy , IN2 N2 , IN3 N3 + M1Lxz , IN4 N4 , K5 IN5 N5

+M2Lxy , K5 IN6 N6 , K5 IN7 N7 + M2Lxz , K5 IN8 N8 , K3 IM1 M1 + M1LT or , (0)

K3 IM2 M2 , K4 IM3 M3 + M2LT or , K4 IM4 M4 ],

K6 K1Lxy

0 0 0 0 0
1Rxy

0 K6 K 0 0 0 0

0 0 K1Lxz 0 0 0

0 0 0 K1Rxz 0 0
2Lxy

0 0 0 0 K7 K 0
0 0 0 0 0 K7 K2Rxy
L1 =

0 0 Kt4 0 0 0

0 0 0 0 0 0
t1

0 0 K 0 0 0

0 0 0 0 0 0
0 0 Kt2 0 0 0
0 0 0 0 0 0


0 0 0 0 0 0
0 0 0 0 0 0

Kt4 0 Kt1 0 Kt2 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0
,
2Lxz t3 t5
K8 K 0 K 0 K 0

0 K8 K2Rxz 0 0 0 0

t3 1LT or t7
K 0 K 0 K 0

1RT or
0 0 0 K1 K 0 0

Kt5 0 Kt7 0 K2 K2LT or 0
0 0 0 0 0 K2 K2RT or

124
where

M1Lxy
si = m1 {1s (a) + tc1 1s
0
(a)} {1i (a) + hc1 1i
0
(a)} + Icz1 1s
0
(a)1i
0
(a), (0)

Ra
K1Lxy
si = K6 1s
00
(x1 )1i
00
(x1 )dx1 , (0)
0

R1
K1Lxy
si = K6 2s
00
(x1 )2i
00
(x1 )dx1 , (0)
a+h1

M1Lxz
si = m1 {3s (a) + tc1 3s
0
(a)} {3i (a) + hc1 3i
0
(a)} + Icy1 3s
0
(a)3i
0
(a), (0)

Ra
K1Lxz
si = 3s
00
(x1 )3i
00
(x1 )dx1 + km {3s (a) + hc1 3s
0
(a)} {3i (a) + hc1 3i
0
(a)} , (0)
0

Ktsi1 = km {3s (a) + hc1 3s

0
(a)} R1 1i (a), (0)

Ktsi2 = km {3s (a) + hc1 3s

0
(a)} R2 3i (b), (0)

Ktsi4 = km {3s (a) + hc1 3s

0
(a)} {7i (b) + hc2 7i
0
(b)} , (0)

R1
K1Rxz
si = 4s
00
(x1 )4i
00
(x1 )dx1 , (0)
a+h1

M2Lxy
si = m2 {5s (b) + hc2 5s
0
(b)} {5i (b) + hc2 5i
0
(b)} + Icz2 5s
0
(b)5i
0
(b), (0)

125
 Rb
K2Lxy
si = K7 5s
00
(x2 )5i
00
(x2 )dx2 , (0)
0

RL2
K2Rxy
si = K7 6s
00
(x2 )6i
00
(x2 )dx2 , (0)
b+h2

M2Lxz
si = m2 {7s (b) + hc2 7s
0
(b)} {7i (b) + hc2 7i
0
(b)} + Icy2 7s
0
(b)7i
0
(b), (0)

Rb
K2Lxz
si = K8 7s
00
(x2 )7i
00
(x2 )dx2 + km {7s (b) + hc2 7s
0
(b)} {7i (b) + hc2 7i
0
(b)} , (0)
0

Ktsi3 = km {7s (b) + hc2 7s

0
(b)} R1 1i (a), (0)

Ktsi5 = km {7s (b) + hc2 7s

0
(b)} R2 3i (b), (0)

Ktsi6 = km {7s (b) + hc2 7s

0
(b)} {3i (a) + hc1 3i
0
(a)} = Ktis4 , (0)

RL2
K2Rxz
si = 8s
00
(x2 )8i
00
(x2 )dx2 , (0)
b+h2

M1LT
si
or
= Im1 1s (a)1i (a), (0)

Ra
K1LT
si
or
= K1 1s
0
(x1 )1i
0
(x1 )dx1 + km R12 1s (a)1i (a), (0)
0

126
 Ktsi7 = km R1 R2 1s (a)3i (b), (0)

Ktsi7 = km R1 R2 1s (a)3i (b), (0)

Ktsi8 = km R1 1s (a) {3i (a) + hc1 3i

0
(a)} = Ktis1 , (0)

Ktsi9 = km R1 1s (a) {7i (b) + hc2 7i

0
(b)} = Ktis3 , (0)

R1
K1RT
si
or
= K1 2s
0
(x1 )2i
0
(x1 )dx1 , (0)
a+h 1

M2LT
si
or
= Im2 3s (b)3i (b), (0)

Rb
K2LT
si
or
= K2 3s
0
(x2 )3i
0
(x2 )dx2 + km R22 3s (b)3i (b), (0)
0

Ktsi10 = km R1 R2 3s (b)1i (a) = Ktis7 , (0)

Ktsi10 = km R1 R2 3s (b)1i (a) = Ktis7 , (0)

Ktsi11 = km R2 3s (b) {3i (a) + hc1 3i

0
(a)} = Ktis2 , (0)

Ktsi12 = km R2 3s (b) {7i (b) + hc2 7i

0
(b)} = Ktis5 , (0)

127
 RL2
K2RT
si
or
= K2 4s
0
(x2 )4i
0
(x2 )dx2 . (0)
b+h2

h = {a1 , ...aN1 , b1 , ...bN2 , c1 , ...cN3 , d1 , ...dN4 , e1 , ...eN5 , f1 , ...fN6 ,

(0)
g1 , ...gN7 , u1 , ...uN8 , m1 , ...mM1 , n1 , ...nM2 , p1 , ...pM3 , q1 , ...qM4 }T ,

A(1)1 A(2)1 ... A(10)1


A(1)2 A(2)2 ... A(10)2

L2 =

,
(0)
. . .


A(1)N A(2)N A(10)N

= {1 , 2 , ..., 10 }T , (0)

0(N1 +N2 )1


[km {3i (a) + hc1 3i
0
(a)}]N3 1




0(N4 +N5 +N6 )1



[km {7i (b) + hc2 7i0
(b)}]N7 1



0(N8 )1
f1 =

E0 sin(t)
(0)


[km R1 1i (a)]M1 1




0(M2 )1




[km R2 3i (b)]M3 1



0(M )1 4

128
 BIBLIOGRAPHY

[1] Vinayak, H., and Singh, R., 1998. Multi body dynamics and modal analysis of
compliant gear bodies. Journal of Sound and Vibration, 210(2), Feb., pp. 171214.

[2] Prohl, M. A., 1945. A general method for calculating critical speeds of flexible
rotors. ASME Journal of Applied Mechanics, 12(3), pp. 142148.

[3] Myklestad, N. O., 1944. A new method of calculating natural modes of uncoupled
bending vibration. J. Aeronaut. Sci., 1, pp. 153162.

[4] Meirovich, L., 1997. Principles and Techniques of Vibrations. Prentice-Hall, Inc.

[5] Tuplin, W. A., 1953. Dynamic loads on gear teeth. Machine Design, 25, pp. 203
211.

[6] Tuplin, W. A., 1950. Gear tooth stresses at high speed. In Proceedings of the
Institution of Mechanical Engineers, 16, pp. 162167.

[7] Harris, S. L., 1958. Dynamic loads on the teeth of spur gears. Inst. Mech. Eng.,
172, pp. 87112.

[8] Gregory, R. W., Harris, S. L., and Munro, R. G., 1963. Dynamic behavior of spur
gears. Proceedings of the Conference on Vibrations in Rotating Machinery, Institu-
tion of Mechanical Engineers, 178(8), pp. 207226.

[9] Kahraman, A., and Singh, R., 1990. Non-linear dynamics of a spur gear pair.
Journal of Sound and Vibration, 142(1), Oct., pp. 4975.

[10] Ozguven, H. N., and Houser, D. R., 1988. Dynamic analysis of high speed gears
by using loaded static transmission error. Journal of Sound and Vibration, 125(1),
Aug., pp. 7183.

[11] Iida, H., Tamura, A., Kikuchi, K., and Agata, H., 1980. Coupled torsional-flexural
vibration of a shaft in a geared system of rotors (1st report). Bulletin of the Japanese
Society of Mechanical Engineer, 23(186), pp. 21112117.

129
[12] Iida, H., and Tamura, A., 1984. Coupled torsional-flexural vibration of a shaft in a
geared system. In Proceedings of the Conference on Vibrations in Rotating Machin-
ery, Institution of Mechanical Engineers, pp. 6772.

[13] Iida, H., Tamura, A., and Oonishi, M., 1985. Coupled dynamic characteristics of a
counter shaft in a gear train system. Bulletin of the Japanese Society of Mechanical
Engineer, 28(245), pp. 26942698.

[14] Iida, H., Tamura, A., and Yamamoto, H., 1986. Dynamic characteristics of a gear
train system with softly supported shafts. Bulletin of the Japanese Society of Me-
chanical Engineer, 29(252), pp. 18111816.

[15] David, J. W., and Mitchell, L. D., 1986. Linear dynamic coupling in geared rotor
systems. Transactions of the ASME, Journal of Vibration, Acoustics, Stress, and
Reliability in Design, 108, pp. 171176.

[16] David, J. W., 1984. Analytical investigation of dynamic coupling in nonlinear geared
rotor systems. PhD thesis, Virginia Polytechnic Institute and State University.

[17] Daws, J. W., 1979. An analytical investigation of three dimensional vibration in

gear-coupled rotor systems. PhD thesis, Virginia Polytechnic Institute and State
University.

[18] Lund, J. W., 1978. Critical speeds, stability and response of a geared train of rotors.
ASME Journal of Mechanical Design, 100, pp. 535538.

[19] Choi, S.-T., and Mau, S.-Y., 2001. Dynamic analysis of geared rotor-bearing systems
by the transfer matrix method. Transactions of the ASME, Journal of Mechanical
Design, 123(4), pp. 562568.

[20] Neriya, S. V., Bhat, R. B., and Sankar, T. S., 1985. Coupled torsional-flexural vibra-
tion of a geared shaft system using finite element analysis. The Shock and Vibration
Bulletin, 55, pp. 1325.

[21] Kahraman, A., Ozguven, H. N., Houser, D. R., and Zakrajsek, J. J., 1992. Dynamic
analysis of geared rotors by finite elements. ASME Journal of Mechanical Design,
114, pp. 507514.

[22] Srinath, L. S., and Das, Y. C., 1967. Vibrations of beams carrying mass. Transac-
tions of the ASME, Journal of Applied Mechanics, E 34, Sept., pp. 784785.

[23] Eshleman, R. L., and Eubanks, R. A., 1967. On the critical speeds of a continuous
shaft-disk system. ASME Journal of Engineering for Industry, Nov., pp. 645652.

130
[24] Chivens, D. R., and Nelson, H. D., 1975. The natural frequencies and critical speeds
of a rotating, flexible shaft-disk system. ASME Journal of Engineering for Industry,
97, pp. 881886.

[25] Huang, T. P., 1988. The transfer matrix impedance coupling method for the eigenso-
lutions of multi-spool rotor systems. ASME Journal of Vibration, Acoustics, Stress,
and Reliability in Design, 110, pp. 468472.

[26] Huang, T. P., 1985. The transfer matrix impedance coupling method with its appli-
cations. In ASME, 85-DET-139.

[27] Nataraj, C., 1993. On the interaction of torsion and bending in rotating shafts.
ASME Journal of Applied Mechanics, 60, pp. 239241.

[28] Shen, I. Y., and Ku, C. P. R., 1997. A nonclassical vibration analysis of a multi-
ple rotating disk and spindle assembly. ASME Journal of Applied Mechanics, 64,
pp. 165174.

[29] Lee, C.-W., and Chun, S.-B., 1998. Vibration analysis of a rotor with multiple flex-
ible disks using assumed modes method. Journal of Vibration and Acoustics, 120,
pp. 8794.

[30] Parker, R. G., and Mote Jr., C. D., 1996. Vibration and coupling phenomena in
asymmetric disk-spindle system. ASME Journal of Applied Mechanics, 63, pp. 953
961.

[31] Parker, R. G., 1999. Analytical vibration of spinning, elastic disk-spindle systems.
ASME Journal of Applied Mechanics, 66, pp. 218224.

[32] Sathe, P. J., and Parker, R. G., 1999. Free vibration and stability of a spinning disk-
spindle system. ASME Journal of Vibration and Acoustics, 121, pp. 391396.

[33] Nelson, H. D., and McVaugh, J. M., 1976. The dynamics of rotor-bearing systems
using finite elements. Journal of Engineering for Industry, 98, May, pp. 593600.

[34] Ozguven, H. N., and Ozkan, Z. L., 1984. Whirl speeds and unbalance response of
multibearing rotors using finite elements. ASME Journal of Vibration, Acoustics,
Stress, and Reliability in Design, 106, Jan., pp. 7279.

[35] Qin, Q. H., and Mao, C. X., 1996. Coupled torsional-flexural vibration of shaft sys-
tems in mechanical engineering-1. finite element model. Computers and Structures,
58(4), pp. 835843.

[36] Sener, O. S., and Ozguven, H. N., 1993. Dynamic analysis of geared shaft systems by
using a continuous system model. Journal of Sound and Vibration, 166(3), pp. 539
556.

131
[37] Kubo, A., 1978. Stress condition, vibration exciting force and contact pattern of he-
lical gears with manufacturing and alignment error. Journal of Mechanical Design,
100, Jan., pp. 7784.

[38] Ozguven, H. N., 1991. A non-linear mathematical model for dynamic analysis of
spur gears including shaft and bearing dynamics. Journal of Sound and Vibration,
145(2), pp. 239260.

[39] Kumar, A. S., and Sankar, T. S., 1986. A new transfer matrix method for response
analysis of large dynamic systems. Computers and Structures, 23(4), pp. 545552.

[40] Lee, A. S., Ha, J. W., and Choi, D.-H., 2003. Coupled lateral and torsional vibration
characteristics of a speed increasing geared rotor-bearing system. Journal of Sound
and Vibration, 263, pp. 725742.

[41] Hawker, L. E., 1991. A vibration analysis of automotive serpentine accessory drive
systems. PhD thesis, University of Windsor, Ontario, Canada.

[42] Barker, C. R., Oliver, L. R., and Breig, W. F., 1991. Dynamic analysis of belt drive
tension forces during rapid engine acceleration. In SAE International Congress and
Exposition, 910687.

[43] Barker, C. R., and Yang, Y. L., 1989. Dynamic analysis of automotive belt drive sys-
tem. In Proceedings of the First International Applied Mechanical Systems Design
Conference, 75, pp. 110.

[44] Hwang, S. J., Perkins, N. C., Ulsoy, A. G., and Meckstroth, R. J., 1994. Rotational
response and slip prediction of serpentine belt drive systems. ASME Journal of
Vibration and Acoustics, 116, pp. 7178.

[45] Leamy, M. J., and Perkins, N. C., 1998. Nonlinear periodic response of engine acces-
sory drives with dry friction tensioners. ASME Journal of Vibration and Acoustics,
120, pp. 909916.

[46] Beikmann, R. S., 1992. Static and dynamic behavior of serpentine belt drive systems:
Theory and experiments. PhD thesis, University of Michigan.

[47] Beikmann, R. S., Perkins, N. C., and Ulsoy, A. G., 1996. Free vibration of serpentine
belt drive systems. ASME Journal of Vibration and Acoustics, 118, pp. 406413.

[48] Kraver, T. C., Fan, G. W., and Shah, J. J., 1996. Complex modal analysis of a flat belt
pulley system with belt damping and coulomb-damped tensioner. ASME Journal of
Mechanical Design, 118, pp. 306311.

[49] Kong, L., and Parker, R. G., 2003. Equilibrium and belt-pulley vibration coupling in
serpentine belt drives. ASME Journal of Applied Mechanics, 70(5), pp. 739750.

132
[50] Abrate, S., 1992. Vibrations of belts and belt drives. Mechanism and Machine
Theory, 27(6), pp. 645659.

[51] Mote Jr., C. D., 1977. Moving-load stability of a circular plate on a floating central
collar. Journal of the Acoustical Society of America, 61(2), Feb., pp. 439447.

[52] Banerjee, J. R., 1999. Explicit frequency equation and mode shapes of a cantilever
beam coupled in bending and torsion. Journal of Sound and Vibration, 224(2),
pp. 267281.

[53] Parker, R. G., 2004. Efficient eigensolution, dynamic response and eigensensitivity
of serpentine belt drives. Journal of Sound and Vibration, 270, pp. 1538.

[54] Liu, G., and Parker, R. G., 2008. Dynamic modeling and analysis of tooth profile
modification for multimesh gear vibration. Journal of Mechanical Design, 130, Dec.,
pp. 121402112140213.

[55] Lee, C., Lin, H. H., Oswald, F. B., and Townsend, D. P., 1991. Influence of linear
profile modification and loading conditions on the dynamic tooth load and stress of
high contact ratio spur gears. Transactions of the ASME, Journal of Mechanical
Design, 113, pp. 473480.

[56] Eritenel, T., and Parker, R. G., 2009. Computational nonlinear vibration analysis
of gear pairs using a three-dimensional model. In Proceedings of the ASME 2009
International Design Engineering Conferences IDETC 2009.

[57] Neriya, S. V., Bhat, R. B., and Sankar, T. S., 1988. On the dynamic response of a
helical geared system subjected to a static transmission error in the form of determin-
istic and filtered white noise inputs. ASME Journal of vibration, acoustics, stress,
and reliability in design, 110, pp. 501506.

[58] Eritenel, T., and Parker, R. G., 2009. Modal properties of three-dimensional helical
planetary gears. Journal of Sound and Vibration, 325, pp. 397420.

[59] Ozguven, H. N., and Houser, D. R., 1988. Mathematical models used in gear dy-
namics a review. Journal of Sound and Vibration, 121(3), pp. 383411.

[60] Jha, R. K., and Parker, R. G., 2000. Spatial discretization of axially moving media
vibration problems. Transactions of the ASME, Journal of Vibration and Acoustics,
122, pp. 290294.

[61] Meirovitch, L., 1975. A modal analysis for the response of linear gyroscopic sys-
tems. Transactions of ASME, Journal of Applied Mechanics, 42, June, pp. 446450.

[62] Wickert, J. A., and Mote Jr., C. D., 1991. Response and discretization methods for
axially moving materials. ASME Applied Mechanical Review, 44(11), pp. 279284.

133
[63] Sathe, P. J., and Parker, R. G., 1999. Exact solutions for the free and forced vibration
of a rotating disk-spindle system. Journal of Sound and Vibration, 223(3), June,
pp. 445465.

[64] Meirovitch, L., 1974. A new method of solution of the eigenvalue problem for gyro-
scopic systems. AIAA Journal, 12, pp. 13371342.

[65] Alspaugh, D. W., 1967. Torsional vibration of a moving band. Journal of the
Franklin Institute, 283(4), pp. 328338.

[66] Mote Jr., C. D., 1965. A study of band saw vibrations. Journal of the Franklin
Institute, 279, pp. 430444.

[67] Mashinostroeniya, V., 1975. Torsional vibration of a moving band. Journal of the
Franklin Institute, 55(2), pp. 4547.

[68] Zhang, L., and Zu, J. W., 1999. Modal analysis of serpentine belt drive systems.
Journal of Sound and Vibration, 222(2), pp. 259279.

[69] Wickert, J. A., and Mote Jr., C. D., 1990. Classical vibration analysis of axially
moving continua. ASME Journal of Applied Mechanics, 57, pp. 738744.

134