Documente Academic
Documente Profesional
Documente Cultură
Drives
Dissertation
By
2010
Dissertation Committee:
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
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
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
of forces and moments and these are identified from the expression of the virtual work. An
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
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
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
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
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
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
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
viii
VITA
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
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
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
3.1 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
4.1 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
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
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
xiii
LIST OF FIGURES
Figure Page
1.3 Two models of geared rotor system (a) Non-gyroscopic spur (b) Gyro-
scopic helical geared shaft systems. . . . . . . . . . . . . . . . . . . . . . 6
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
2.13 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
3.6 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
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
4.1 (a) Model of the coupled belt-pulley-shaft system (b) Pulley tilting causing
torsional belt deflection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.3 Dimensionless natural frequency () vs. the rotational speed (). . . . . . . 110
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
xviii
CHAPTER 1
INTRODUCTION
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
G
S
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
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
T2
T1 e(t)
cm
~
I1 km I2
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
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
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-
prevent them from being studied in literature. Response calculation of gyroscopic systems
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
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
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
Alternator
Idler
Tensioner
Water Pump
Power
Steering
Air
Conditioner
Crankshaft
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
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 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
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.
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
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-
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
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
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
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
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
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
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
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
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
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
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
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
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:
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
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
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-
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
a serpentine belt drive system with the attached pulleys mounted on compliant shafts in
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
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-
acteristics is essential for the design and troubleshooting under operating conditions.
Natural frequency veering is useful to trace any dramatic changes in vibration modes
19
2. Study of the gyroscopic effect in a geared shaft system is important for high speed
helical gears in the model provides larger number of design parameters. The method
with classical boundary conditions. Modal analysis of the present system can be fol-
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
of shaft flexibility on the dynamics of high speed serpentine belt drives. Campbell
diagram shows the gyroscopic system characteristics including mode splitting and
20
CHAPTER 2
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-
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
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
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
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
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
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
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
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
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.
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
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
hc2 R1 R2 a b m1 m2
hc2 = , R1 = , R2 = ,a = ,b = , m1 = , m2 = ,
L1 L1 L1 L1 L1 s1 L1 s1 L1
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
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-
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
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)
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)
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
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-
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.
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
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.
K4 2R , m1 uc1, m1 vc1 , m2 uc2 , m2 vc2 , Icz1 1 (t), Icy1 1 (t), Icz2 2 (t), Icy2 2 (t),
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 u2L
km [R1 1L (a, t) + R2 2L (b, t) vc1 (t) + vc2 (t)] , K7 (b, t)
x32
3 u1L
+km [R1 1L (a, t) + R2 2L (b, t) vc1 (t) + vc2 (t)] , hc1 K6 (a, t)
x31
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)
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
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
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
Using the separable form w weit to separte the spatial and temporal variables, the
eigenvalue problem is
2 M + L w = 0.
(2.41)
35
2.4 Discretization
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
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
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.
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
applicable to the equation (2.43). For the basis function pip , p varies from 1 to Np , whereas
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
42
The holonomic constraints on the generalized coordinates come from the ten matching
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.
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
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
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
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
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
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
G2 = 14 103
Stiffness (N/m) km = 47
K8 = 1.40
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
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
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
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.
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
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,
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
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.
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
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
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
53
210 290
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
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
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
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
56
15
15
Response Amplitude (dB)
5
5
0
Three Mode Approximation 0
Three Mode Approximation
Actual
-5 -5
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
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
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
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
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
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
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-
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. 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
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.
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
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
(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
(b) Develop a methodology to solve the eigenvalue problem of such a system using the
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
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
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.
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
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)).
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
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
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
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)
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)
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
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
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
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
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.
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
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
Icy2 2 + (L2 + hc2 )m2 {vc2 + (L2 + hc2 )2 } , Im1 1 (1), Im2 2 (L2 )}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
cos(),
3 u2 3 v2
K7 (L2 ), K 8 (L2 )+
x32 x32
2 u1 3 u1
cos(), K6 (1) + hc K 6 (1)+
x21 1
x31
3 u2
hc2 K7 (L2 )+
x32
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
2
K2 (L2 ) + R2 km
x2
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)
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 ,
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
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)
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
N = N1 + N2 + N3 + N4 + M1 + 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
Lqr = Lr , q = Lrq , gq = (f, q )
The prime indicates derivative with respect to the the axial coordinate x1 or x2 . The
mensionless natural frequencies (imaginary part of the eigenvalues) are related to the
dimensional ones by s
E1 Isy1
= (3.38)
s1 L41
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
A + B=0. (3.39)
Here s are imaginary eigenvalues and s are complex state eigenvectors. and appear
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
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
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
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
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,
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
79
Table 3.2: Dimensional and dimensionless parameters of the coupled disk-shaft system
Masses (kg) m1 = 3.00, m2 = 3.00
G1 = 84 109 , G2 = 84 109
K7 = 1.00, K8 = 1.00
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 .
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
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
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
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
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.
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
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
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
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
becomes dominant involving both the shafts. This increases the stiffness of the system
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
= 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
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
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)
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
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
90
6 0.7
Total Total
Mode 5 Mode 5
0.6 Mode 2
Response amplitude (dB) 5
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
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
92
CHAPTER 4
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-
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
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
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
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,
system, simplified beam theory (Euler-Bernoulli beam theory) is applied. The objectives
(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
(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
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
h2 hc1 hc2 R1 R2 m1 m2
h2 = , hc1 = , hc2 = , R1 = , R2 = , m1 = , m2 = ,
L1 L1 L1 L1 L1 s1 L1 s1 L1
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
The bearings impose the following boundary conditions for fixed support for bending
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)
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 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
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
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
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
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 ,
With this definition of the inner product, the operators M, L, and L are symmetric,
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
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
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
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
Lqr = Lr , q = Lrq , gq = (f, q )
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).
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.
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
C = 0.500
G1 = 84 109 , G2 = 84 109
K7 = 1.00, K8 = 1.00
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 ().
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
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
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)
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
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)
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
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
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.
Analytical calculation of the natural frequencies and modes, modal sensitivities, and dy-
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
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-
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.
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
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
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
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
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
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
119
5.2 Future work
The future work is recommended for the following modified models to take care of the
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
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.
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
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
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
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
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
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
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)
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
127
RL2
K2RT
si
or
= K2 4s
0
(x2 )4i
0
(x2 )dx2 . (0)
b+h2
= {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.
[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