Lateral-Torsional Vibration of A Side-Loaded Rotor With Asymmetri
Lateral-Torsional Vibration of A Side-Loaded Rotor With Asymmetri
Lateral-Torsional Vibration of A Side-Loaded Rotor With Asymmetri
A Thesis
presented to
In Partial Fulfillment
by
Clinton Judd
Clinton Judd
ii
Committee Membership
iii
Abstract
By Clinton Judd
Using energy equations a four degrees of freedom analytical model is developed for a two-disk rotor with
shaft stiffness asymmetry. A radial constant force is applied to the outboard disk to emphasize the effects of
gravity or aerodynamic side loading. Special emphasis is placed on characterizing the lateral and torsional
vibration trends associated with shaft asymmetry which may be used to identify failing shafts in operational
rotor systems. Simulation reveals distinct patterns in lateral and torsional response, with strong
dependencies on the magnitude of the side load, magnitude of the asymmetry and proximity of the lateral
and torsional natural frequencies. Notable interaction is also observed between the lateral and torsional
response. Lateral response peaks are found to correlate to torsional response peaks under some conditions.
An experiment is performed to compare the response of a real system with the simulated model.
iv
Acknowledgements
The author wishes to thank the faculty and staff of the Mechanical Engineering Department for their strong
support and encouragement throughout his undergraduate and graduate education. He is especially grateful
for the moral support provided during his diagnosis and successive treatment for Hodgkin’s Lymphoma
v
Table of Contents
Bibliography................................................................................................................................................ 46
vi
B.1 Lateral stiffness of axis-symmetric shaft ........................................................................................... 49
PY
C.3 Special Case: Torsional Vibration Matlab Model For Parametric Study Of ε ........................ 79
k
vii
List of Tables
Table 3: Parameters used for torsional vibration model for gravitational loading........................................ 19
Table 5: Parameters used in vibration model of rotor with increasing shaft asymmetry.............................. 22
viii
List of Figures
1
Figure 5: Orbit response at half the lateral natural frequency ( Ω = ωn = 1750rpm ) .................................. 15
2
PY
Figure 8: Torsional vibration amplitude vs. speed for varying ratios of ε ........................................... 17
k
1
Figure 9: Torsional response at half the torsional natural frequency ( Ω = ωt = 1750rpm )........................ 18
2
Figure 10: Torsional response at the torsional natural frequency ( Ω = ωt = 3500rpm )................................ 18
Figure 11: Torsional vibration amplitude vs. speed for gravitationally loaded rotor ................................... 19
Figure 13: Vertical vibration amplitudes for rotor with increasing shaft asymmetry................................... 23
Figure 14: Torsional vibration amplitudes for rotor with increasing shaft asymmetry ................................ 23
Figure 21: Torsional response for varying ratios of ωt ωn (3D plot) ......................................................... 28
Figure 22: Torsional response for varying ratios of ωt ωn (2D plot) ......................................................... 28
Figure 23: Lateral response for increasing side load (3D plot)..................................................................... 29
ix
Figure 24: Torsional response for increasing side load (3D plot)................................................................. 30
Figure 25: Bently Nevada RK-4 Rotor Kit configured for a side loaded rotor............................................. 31
Figure 26: Rotor kit configured to minimize torsional natural frequency .................................................... 32
Figure 27: Side load provided by gravity through a pulley system .............................................................. 32
Figure 31: Torsional transducer consisting of a gear wheel and two proximity probes ............................... 35
Figure 39: Overall rotor torsional response (difference between inboard & outboard), 1X & 2X ............... 42
x
Chapter I Introduction
I.1 Purpose
Modern high speed rotating machinery is constantly driven to the limits of design, materials, and service
life. Equipment failure can be costly and hazardous. Diagnostic tools are increasingly important for early
detection of component failure. One area of potential failure is in the shaft of a rotating machine which may
crack as a result of fatigue. This study aims to model and study the effects of a compromised shaft on
rotordynamic behavior in an effort to identify warning signs that could be used to identify a developing
failure.
I.2 Background
A comprehensive literature survey of various crack modeling techniques and system behavior of cracked
rotor was given by Wauer (1).This paper contains the modeling of the cracked components of the structures
and searches for different detection strategies to diagnose fracture damage. Dimarogonas (2) provided a
comprehensive literature review of the vibration of cracked structures and cites more than 300 papers. The
review is divided into several sections; methods that describe local flexibility due to cracks, nonlinearities
introduced into the system, and local stiffness matrix descriptions of the cracked section. The crack leads to
a coupled system that can be recognized from additional harmonics in the frequency spectrum. The sub-
harmonic resonances at approximately half and one third of the bending critical speed of the rotor are
reported to be the prominent crack indicators by Gasch (3, 4) and Chan (5). By utilizing a single parameter
“hinge” crack model, Gasch, provided an overview of the dynamic behavior of a simple rotor with
transverse crack. He assumed weight dominance and employed a perturbation method into his analysis.
Cross-coupling stiffness and dynamic response terms were not included in his analysis. Mayes model (6) is
more practical for deep cracks than a hinged model. Based on Mayes modified model, Sawicki and Wu et
al. (7, 8) studied the transient vibration response of a cracked Jeffcott rotor under constant acceleration
ratios and under constant external torque. The angle between the crack centerline and the rotor whirl vector
is employed to determine the closing and opening of the crack. This allows one study of the rotor dynamic
1
response with or without the rotor weight dominance assumption by taking non-synchronous whirl into
account. Sawicki and Wu et al.(9) investigated the nonlinear dynamic response of a cracked one-mass
Jeffcott rotor by means of bifurcation plots. When a rotor with the crack depth of 0.4 spins at some speed
ranges, both the lateral and torsional vibration responses sustain periodic, quasi-periodic or chaotic
behavior. Muszynska et al. (10) and Bently et al.(11) discuss rotor coupled lateral and torsional vibrations
due to unbalance, as well as due to shaft anisotropy under a constant radial preload force. Their
experimental results exhibited the existence of significant torsional vibrations due to coupling with the
lateral modes. In Bently and Muszynska’s experiments, an asymmetric shaft was used to simulate the
behavior of a crack.
The current paper extends the research investigation of Bently et al. (11) by using the simpler anisotropic
shaft model to characterize a cracked shaft and employing the methods used by Wu et al. (12) to develop a
vibration model. Using energy equations a four degrees of freedom analytical model is developed for a
two-disk rotor with shaft asymmetry. A radial constant force is applied to the outboard disk to emphasize
the effects of gravity or aerodynamic side loading. As load and stiffness asymmetry increase, the vibration
amplitudes in both lateral and torsional directions increase, resulting in torsional vibration with amplitudes
that may be measured in practical applications. This paper places special emphasis on characterizing the
torsional vibration trends associated with shaft asymmetry which may be used to identify failing shafts in
2
Chapter II Rotor Model
The system shown in Figure 1 represents a practical rotor system. An electric motor drives a shaft through a
flexible coupling. The shaft is supported by relatively rigid bearings that in turn support two rotors. The
inboard rotor is spanned closely by bearings while the outboard rotor is supported by a wide bearing span.
This system is analogous to a modern rotating machine consisting of a turbine driving a compressor. In this
case the inboard rotor and motor represent the driving torque and rotational inertia of a turbine. The
outboard rotor represents a compressor. Side loading at the outboard rotor might represent the side loading
3
Figure 2: Rotor system model (damping not shown)
Figure 2 shows a vibration model for the physical system depicted by Figure 1. The motor is modeled as a
source of rotational energy providing a constant angular velocity, Ω . The attached coupling acts as both a
torsional damper and torsional spring. The shaft is modeled by multiple elements: a torsional spring
connecting the two rotors, two lateral springs aligned with a fixed X-Y coordinate system, and torsional
and lateral dampers (not shown). Because the inboard rotor is held laterally by close-mounted rigid
bearings, it is modeled with only one rotational degree of freedom, Θ , and no lateral flexibility. The
outboard rotor is allowed to move both rotationally and laterally due to the flexible nature of the shaft, so it
For an axis-symmetric shaft, lateral stiffness is conveniently described in the rotating ξ - η coordinate
system shown in Figure 3. The η -direction is always aligned with the strong direction of the shaft while
the ξ -direction is always aligned with the weak direction of the shaft. This unique coordinate system
exhibits uncoupled stiffness; loads in the η -direction produce displacement only in the η -direction and
4
Figure 3: Outboard rotor parameters
In this special coordinate system, forces exerted by the shaft are described by Eq. II-1. Stiffness is defined
in the strong and weak directions as kξ and kη respectively. The displacements η and ξ are measured
from the geometric bearing centerline to the geometric center of the rotor.
Fξ kξ 0 ξ
F = 0 kη η
η
Eq. II-1
The stiffness matrix in the rotating coordinate system ( K R ) is defined by Eq. II-2.
kξ 0
KR =
0 kη
Eq. II-2
frame. Shaft loads and displacements in the inertial coordinate system are related by the stiffness matrix,
FX k11 k12 X
=
FY k21 k22 Y
Eq. II-3
5
k11 k12
KI =
k21 k22
Eq. II-4
To determine values that appropriately populate the inertial stiffness matrix the following coordinate
K I = TK RT −1
Eq. II-5
where
cos Φ − sin Φ
T =
sin Φ cos Φ
Eq. II-6
kξ + kη kη − kξ
Defining the average stiffness k = and the shaft asymmetry factor q = , Eq. II-5produces
2 2k
1 − q cos 2Φ −q sin 2Φ
KI = k
−q sin 2Φ 1 + q cos 2Φ
Eq. II-7
Given any rotational position Φ and lateral displacement X and Y of the rotor centerline, the inertial
stiffness matrix may be used calculate loads acting on the rotor by the flexible shaft.
The equations of motion governing the system shown in Figure 2 can be derived using Lagrange’s
equations. The general form of Lagrange’s equations is shown in Eq. II-8, where qi is a generalized
coordinate, Qi is the generalized load(s) in the direction of qi that produce work, T and U are the kinetic
and potential energies of the system, D is the dissipation function of the system, and n is the number of
degrees of freedom.
6
∂ ∂T ∂T ∂U ∂D
− + + = Qi i = 1,..., n
∂t ∂q&i ∂qi ∂qi ∂q&i
Eq. II-8
T= 1
{
2
IΦ
&2 + 1
2 2
1
&2
I0Θ
3 { & Y& cos ( Φ + δ ) − X& sin ( Φ + δ ) + ε 2 Φ
+ 12 M X& 2 + Y& 2 + 2εΦ
14444444444424444444444 43
&2 }
Outboard Rotor Inboard Rotor
Rotational Kinetic Energy Outboard Rotor Lateral Kinetic Energy
Rotational Kinetic Energy
Eq. II-9
For a non-specific lateral shaft stiffness having k12 = k21 , the potential energy of the system is
U= ( k X + k Y ) + k XY
1
2 11
14444244443
2
22
2
12 + 2 t (
1
14
k Φ − Θ)
4244 3
2
Energy Stored In Shaft Through Lateral Deflection Energy Stored in Shaft Through Torsional Deflection
Eq. II-10
For the axis-symmetric shaft, Eq. II-7 and Eq. II-10 produce Eq. II-11.
Eq. II-11
Eq. II-12
∂ ∂T ∂T ∂U ∂D
− + + = − PX
∂t ∂X& ∂X ∂X ∂X&
∂ ∂T ∂T ∂U ∂D
− + + = − PY
∂t ∂Y& ∂Y ∂Y ∂Y&
∂ ∂T ∂T ∂U ∂D
& − + + =0
∂t ∂Φ ∂Φ ∂Φ ∂Φ
&
∂ ∂T ∂T ∂U ∂D
& − + +
&
= −TDR
∂t ∂Θ ∂Θ ∂Θ ∂Θ
Eq. II-13
The solutions for Eq. II-12 using Eq. II-9, Eq. II-11, and Eq. II-12 are as follows.
7
X&& + 2ςωn X& + ωn2 (1 − q cos 2Φ ) X = εΦ & 2 cos ( Φ + δ ) + ω 2 q sin 2ΦY − PX
&& sin ( Φ + δ ) + εΦ
n
M
&& + ω 2 ( Φ − Θ ) + 2ς ω ( Φ
Φ & ) = ε 2ςωn cos ( Φ + δ ) Y& − sin ( Φ + δ ) X&
& −Θ
t t t
ρ2
εωn2 εω 2 q
+ cos ( Φ + δ ) Y − sin ( Φ + δ ) X + n2 cos ( Φ − δ ) Y − sin ( Φ − δ ) X
2
ρ ρ
ε PY PX ωn2 q
+ cos ( Φ + δ ) − M ( Φ + δ ) + 2 2 cos 2ΦXY − sin 2Φ ( X − Y )
2 2
M sin
ρ2 ρ
&& − kt ( Φ − Θ ) − Ct ( Φ
Θ & ) = − TDR
& −Θ
I0 I0 I0
Eq. II-14
Assuming that the motor will run at a constant speed and the outboard rotor will vibrate about a static
PX
lateral offset, it is convenient to express Eq. II-14 in terms of vibration coordinates X M = X + ,
k
PY
YM = Y + , ϕ = Φ − Ωt , and θ = Θ − Ωt . Additionally, the motor drive torque can be expressed as a
k
Cc kc
function of the inboard rotor rotational vibration θ and coupling properties CR = and K R = .
Ct kt
Substitution of these terms into Eq. II-14 produces Eq. II-15 below.
( )
X&&M + 2ςωn X& M + ωn2 1 − q cos ( 2 ( Ωt + ϕ ) ) X M − ωn2 q sin ( 2 ( Ωt + ϕ ) ) YM = εϕ&& sin ( Ωt + ϕ + δ )
P P
+ε ( Ω + ϕ& ) cos ( Ωt + ϕ + δ ) − Y q sin ( 2 ( Ωt + ϕ ) ) − X q cos ( 2 ( Ωt + ϕ ) )
2
M M
( )
Y&&M + 2ςωnY&M + ωn2 1 + q cos ( 2 ( Ωt + ϕ ) ) YM − ωn2 q sin ( 2 ( Ωt + ϕ ) ) X M = −εϕ&& cos ( Ωt + ϕ + δ )
P PY
+ε ( Ω + ϕ& ) sin ( Ωt + ϕ + δ ) − X q sin ( 2 ( Ωt + ϕ ) ) + M q cos ( 2 ( Ωt + ϕ ) )
2
M
ε q PY P
ϕ&& + ωt 2 (ϕ − θ ) + 2ς tωt (ϕ& − θ& ) = − 2 cos ( Ωt + ϕ − δ ) − X sin ( Ωt + ϕ − δ )
ρ M M
q PY PX PX
2
PY
2
+
ωn2 ρ 2
2cos ( 2 ( Ωt + ϕ ) ) M M − sin ( 2 ( Ωt + ϕ ) ) + sin ( 2 ( Ωt + ϕ ) ) +Γ
M M
8
ε 2ςωn & − sin ( Ωt + ϕ + δ ) X& + εωn cos ( Ωt + ϕ + δ ) Y − sin ( Ωt + ϕ + δ ) X
2
Γ= cos ( Ωt + ϕ + δ ) Y
ρ2
M
ρ2
M M M
εωn2 q
+ cos ( Ωt + ϕ − δ ) YM − sin ( Ωt + ϕ − δ ) X M
ρ2
2q P P P P
+ 2
cos ( 2 ( Ωt + ϕ ) ) − Y X M − X YM − sin ( 2 ( Ωt + ϕ ) ) Y YM − X X M
ρ M M M M
ωn2 q
+ 2 cos ( 2 ( Ωt + ϕ ) ) X M YM − sin ( 2 ( Ωt + ϕ ) ) X M2 + sin ( 2 ( Ωt + ϕ ) ) YM2
ρ2
Eq. II-15
Non-dimensional time τ = ωn t may be introduced for modeling purposes. The chain rule allows simple
Ω Ω Ω
X M′′ + 2ς X M′ + 1 − q cos 2 τ + ϕ X M − q sin 2 τ + ϕ YM = εϕ ′′ sin τ + ϕ + δ
ωn ωn ωn
ε Ω P q Ω P q Ω
2 (
Ω + ωnϕ ′) cos τ + ϕ + δ − Y 2 sin 2 τ + ϕ − X 2 cos 2 τ + ϕ
2
+
ωn
ωn M ωn ωn M ωn ωn
Ω Ω Ω
YM′′ + 2ς YM′ + 1 + q cos 2 τ + ϕ YM − q sin 2 τ + ϕ X M = −εϕ ′′ cos τ + ϕ + δ
ωn ωn ωn
ε Ω P q Ω P q Ω
2 (
Ω + ωnϕ ′ ) sin τ + ϕ + δ − X 2 sin 2 τ + ϕ + Y 2 cos 2 τ + ϕ
2
+
ωn ωn M ωn ωn M ωn ωn
2
ωt 2ς t ωt εq PY Ω PX Ω
ϕ ′′ + (ϕ − θ ) + (ϕ ′ − θ ′ ) = − 2 2 cos ω τ + ϕ − δ − M sin ω τ + ϕ − δ
ω
n ω n ω nρ M n n
q Ω P P Ω PX
2
Ω PY
2
+ 4 2
2 cos 2 τ + ϕ Y X − τ + ϕ + τ + ϕ + Γn
ω ω
sin 2 sin 2
ωn ρ ωn M M n M n M
d( )
1
For notation purposes, non-dimensional time derivatives are expressed as hatch marks (i.e. ( )′ =
dτ
d2 ( ) ).
and ( )′′ =
dτ 2
9
Ω
ε 2ς Ω ε Ω Ω
Γn = cos τ + ϕ + δ YM′ − sin τ + ϕ + δ X M′ + 2 cos τ + ϕ + δ YM − sin τ + ϕ + δ X M
ρ2
ωn ωn ρ ωn ωn
εq Ω Ω
+ 2 cos τ + ϕ − δ YM − sin τ + ϕ − δ X M
ρ ωn ωn
2q Ω P P Ω P P
+ 2 2 cos 2 τ + ϕ − Y X M − X YM − sin 2 τ + ϕ Y YM − X X M
ωn ρ ωn M M ωn M M
q Ω Ω Ω
+ 2
2 cos 2 τ + ϕ X M YM − sin 2 τ + ϕ X M2 + sin 2 τ + ϕ YM2
ρ ω ω ω
n n n
2 2
ωt ω ω ω
θ ′′ + (1 + CR ) 2ς t θ ′ + (1 + K R ) t θ = t ϕ + 2ς t t ϕ ′
ωn ω
n ω
n ω n
Eq. II-16
While Eq. II-16 provides a complete mathematical description for the system described in section II.1, the
physical interpretation is unclear. In sections II.4 and II.5, special cases are explored to develop a better
For negligible levels of torsional activity, some simplifications can be made to the general equations of
motion. This might be the case when operating at a speed that is not near a torsional resonance. Assuming a
torsionally rigid system, all elements of the rotor move at a constant rotational velocity Ω such that ϕ = 0
and θ = 0 . Note that the inboard rotor has no vibration influence in this simplified model. For a vertical
P
X&&M + 2ςωn X& M + ωn2 (1 − q cos ( 2Ωt ) ) X M = εΩ2 cos ( Ωt ) − Y q sin ( 2Ωt ) + ωn2 q sin ( 2Ωt ) YM
M
P
Y&&M + 2ςωnY&M + ωn2 (1 + q cos ( 2Ωt ) ) YM = εΩ2 sin ( Ωt ) + Y q cos ( 2Ωt ) + ωn q sin ( 2Ωt ) X M
2
M
Eq. II-17
1X (once per motor revolution) and 2X (twice per motor revolution) lateral forcing functions are clearly
shown as functions of the unbalance and side loading respectively. 2X variations in stiffness are also
present as well as 2X coupling terms that describe the interaction between the two lateral directions. It
10
should be noted that all 2X components are related to the shaft asymmetry, q. If no asymmetry were
present, q would become zero and Eq. II-17 would reduce to the description of a Jeffcott rotor with a 1X
When the rotor is operating at a speed that does not excite significant lateral activity, some simplifications
can be made to the general equations of motion. Side loading produces a static deflection, however the
outboard disk remains in a constant lateral position such that X M = 0 and YM = 0 . For rigid coupling
2
ε q PY q P
ϕ&& + 2ς t ωtϕ& + ωt 2ϕ = − 2 cos ( Ωt + ϕ ) + 2 2 Y sin ( 2 ( Ωt + ϕ ) )
ρ M424444
1444 3
ωn ρ M
14444
4244444
3
1X Forcing Term 2X Forcing Term
Eq. II-18
Eq. II-18 resembles a linear oscillator with two forcing terms. The 1X forcing term is linearly proportional
to the side load and unbalance eccentricity. The 2X forcing term is not dependent on the unbalance
eccentricity, but has a quadratic dependency on the side load. Manipulation of Eq. II-18 reveals a
Eq. II-19
Eq. II-19 shows that the 2X torque amplitude is proportional to the product of the side load ( PY ) and the
static displacement due to the side load ( Y = − PY k ). This highlights an interesting phenomena: a torque
that is proportional to Y results from the load PY acting in the direction of Y . Furthermore, the amplitudes
of the 1X and 2X torsional excitations are proportional to the ratio of unbalance eccentricity to the static
To predict the conditions under which the 1X and 2X torsional excitations would be equal in magnitude,
11
2
ε q PY q PY
ρ 2 M = ω2ρ 2 M
n
PY
ε= 2
ωn M
PY
ε=
k
Eq. II-20
Eq. II-20 predicts that the magnitudes of the 1X and 2X torsional excitations would be equal when the
PY
magnitudes of the eccentricity and average displacement from side loading ( ) are equal.
k
12
Chapter III Simulation
For the special case where torsional vibration is considered negligible, ϕ = 0 and θ = 0 . If we consider
only a vertical load on the outboard rotor ( PX = 0 ), assume no initial displacements ( Φ 0 = 0 , Θ0 = 0 ), and
consider the case where unbalance is aligned with the weak axis ( δ = 0 ), the equations of motion reduce to
Eq. II-17. Non-dimensional time τ = ωn t may be introduced for modeling purposes. The chain rule allows
III-1 below.
Ω Ω ε Ω P q Ω
X M′′ + 2ς X M′ + 1 − q cos 2 τ X M − q sin 2 τ YM = 2 Ω2 cos τ − Y 2 sin 2 τ
ωn ωn ωn ωn M ωn ωn
Ω Ω ε Ω P q Ω
YM′′ + 2ς YM′ + 1 + q cos 2 τ YM − q sin 2 τ X M = 2 Ω2 sin τ + Y 2 cos 2 τ
ωn ωn ωn ωn M ωn ωn
Eq. III-1
For the purpose of numerical modeling, let the states u3 = X M , u4 = X M′ , u5 = YM , u6 = YM′ . This
u3′ = u4
Ω Ω ε Ω P q Ω
u4′ = −2ς u4 − 1 − q cos 2 τ u3 + q sin 2 τ u5 + 2 Ω2 cos τ − Y 2 sin 2 τ
ωn ωn ωn ωn M ωn ωn
u5′ = u6
Ω Ω ε Ω P q Ω
u6′ = −2ς u6 − 1 + q cos 2 τ u5 + q sin 2 τ u3 + 2 Ω 2 sin τ + Y 2 cos 2 τ
ωn ωn ωn ωn M ωn ωn
Eq. III-2
The solution to this system of equations can be found using computer software, in this case Matlab version
7.1. The function “ode45” is used to solve this system of equations. This solver integrates the series of
13
differential equations and is based on an explicit version of the Runge-Kutta formula, the Dormand-Prince
pair. It is a one-step method that uses initial conditions and solves for conditions at the next time step.
Matlab programming for this system can be found in the C.1. Using the parameters listed in Table 1,
solutions to Eq. III-2 are generated for a range of rotor speeds. The resulting steady state vibration XM and
ωn ς ςt ε ρ Py/M q
2
3500rpm 0.2 0.05 0.005in 1.0in 2700in/sec 0.15
0.035
Horizontal Displacement X
0.03
Vertical Displacement Y
Peak-to-Peak Amplitude (in)
0.025
0.02
0.015
0.01
0.005
0
0 1000 2000 3000 4000 5000 6000 7000
Rotor Speed (rpm)
Figure 4 shows amplitude peaks at the natural frequency ( ωn ) and half the natural frequency ( 12 ωn ) as
expected. These peaks are associated with the 1X unbalance excitation and 2X asymmetry and side load
excitations respectively.
In Figure 5 and Figure 6 the orbit response of the rotor is plotted at the 12 ωn and ωn rotor speeds.
14
0.012 0.02
0.016
0.008
0.012
0.008
0.004
0.004
0
0
-0.004
-0.004 -0.008
-0.012
-0.008
-0.016
-0.02
-0.012 -0.02 -0.01 -0.01 -0.00 -0.00 0 0.004 0.008 0.012 0.016 0.02
-0.012 -0.008 -0.004 0 0.004 0.008 0.012 6 2 8 4
Horizontal Displacem ent (in) Horizontal Displacem ent (in)
Figure 5: Orbit response at half the lateral Figure 6: Orbit response at the lateral natural
1 frequency ( Ω = ωn = 3500rpm )
natural frequency ( Ω = ωn = 1750rpm )
2
When the rotor operates at a speed that does not excite significant lateral activity, some assumptions and
simplifications can be made to the general equations of motion. It is assumed that side loading produces a
static deflection, however the outboard disk remains in a relatively constant lateral position such that
X M = 0 and YM = 0 . For rigid coupling θ = 0 . Assuming only a vertical side load ( PX = 0 ) and δ = 0 ,
2
ω 2ς ω εq P Ω q Ω PY
2
ϕ ′′ + t ϕ + t t ϕ ′ = − 2 2 Y cos τ + ϕ + sin 2
ω τ + ϕ
ωn ωn ωn ρ M ωn ωn ρ
4 2
n M
Eq. III-3
For the purpose of numerical simulation, let τ = ωn t and the states u1 = ϕ and u2 = ϕ ′ . The equations of
motion become:
u1′ = u2
15
2
ω 2ς ω εq P Ω q Ω PY
2
u2′ = − t u1 − t t u2 − 2 2 Y τ + + τ +
1 ω 1
cos u sin 2 u
ω
n ω n ωnρ M ωn ωn ρ
4 2
n M
Eq. III-4
Matlab is used to find a solution to Eq. III-4 (see Appendix C.2). Peak-to-peak torsional vibration
PY
ε 2
ωn ωt ς ςt ε ρ Py/M q k
2
10000rpm 3500rpm 0.2 0.05 0.005in 1.0in 2700in/sec 0.15 0.49
0.02
0.018
Peak-to-Peak Amplitude (deg)
0.014
0.012
0.01
0.008
0.006
0.004
0.002
0
0 1000 2000 3000 4000 5000 6000 7000
Rotor Speed (rpm)
Two peaks are observed corresponding to ωt and ½ωt respectively. It is observed that the ratio of the peak
PY
amplitudes at ωt and ½ωt approximates the ratio ε . This phenomenon is explored further through
k
PY
parametric study by varying the value of and plotting the amplitude and speed against the ratio
M
PY P P
2
Ratio ε is calculated from parameters ωn, ee, and Py/M using the relationship Y ε = 2Y ee
k k ω M
16
PY P
ε . The resulting 3D plot using the parameters from Table 2 and increasing values of Y is shown in
k M
Figure 8 below.
PY
Figure 8: Torsional vibration amplitude vs. speed for varying ratios of ε
k
PY
It can be seen in Figure 8 that the ratio ε approximates the ratio of torsional vibration amplitudes at ωt
k
PY
and ½ωt (i.e. where ε = 1 the torsional vibration amplitudes at ωt and ½ωt are approximately equal).
k
This suggests that torsional vibration associated with a shaft stiffness asymmetry (such as a transverse
PY
crack) might be characterized by this ratio. For ε < 1 more torsional vibration activity would be
k
PY
expected at speeds near ωt and for ε > 1 more torsional vibration activity would be expected near ½ωt.
k
Torsional response predicted by the model for the parameters in Table 2 is plotted on a time scale for
½ωt=1750rpm and ωt=3500rpm in Figure 9 and Figure 10 respectively. Near ½ωt, peaks are predicted
every 90 degrees of motor rotation, when the side load is aligned with the strong and weak axes of the
shaft. Near ωt peaks are predicted only when the side load is aligned with the weak axis.
17
0.01
0.008
0.004
0.002
-0.002
-0.004
-0.006
-0.008
-0.01
3π 2π 5π
0 π /2 π
2 2
1
Figure 9: Torsional response at half the torsional natural frequency ( Ω = ωt = 1750rpm )
2
0.01
0.008
0.006
Torsional Displacement (deg)
0.004
0.002
-0.002
-0.004
-0.006
-0.008
-0.01
3π 2π 5π
0 π /2 π
2 2
18
The case where gravity is the only source of side loading may be of special interest, as might be the case in
a horizontal turbine with balanced aerodynamic loading. For the case of a gravitationally side loaded rotor
Py/M would be the acceleration due to gravity, generally accepted as 32.17 ft/sec2 or 386.1 in/sec2. Using
the rotor parameters in Table 3, the Matlab model from Appendix C.2 predicts the response shown in
Figure 11. Note the scale of the vertical axis in Figure 11 is the same as in Figure 7 for relative comparison.
Response at ½ωt is insignificant compared with the response at ωt, as might be expected from the low ratio
PY
of ε . Systems with lower lateral natural frequencies may exhibit additional response at ½ωt since the
k
PY
ratio ε would increase, but a general model that considers both lateral and torsional vibration should
k
be used to investigate such a case; the assumption of negligible lateral activity would be invalidated as ωn
approached ωt.
Table 3: Parameters used for torsional vibration model for gravitational loading
PY
ε
ωn ωt ς ςt ε ρ Py/M q k
2
10000rpm 3500rpm 0.2 0.05 0.005in 1.0in 386.1in/sec 0.15 0.07
0.02
0.018
Peak-to-Peak Amplitude (deg)
0.014
0.012
0.01
0.008
0.006
0.004
0.002
0
0 1000 2000 3000 4000 5000 6000 7000
Rotor Speed (rpm)
Figure 11: Torsional vibration amplitude vs. speed for gravitationally loaded rotor
19
III.3 General Vibration
A general vibration model is required to consider torsional and lateral vibration interaction. To simplify
simulation, rigid coupling ( θ = 0 ), no horizontal side load ( PX = 0 ), and δ = 0 are assumed. For the
u1′ = u2
2
ω 2ς ω εq P Ω q Ω PY
2
u2′ = − t u1 − t t u2 − 2 2 Y τ + 1 + τ +
ω 1
cos u sin 2 u
ω
n ω n ωnρ M ωn ωn ρ
4 2
n M
ε 2ς Ω Ω ε (1 + q ) Ω Ω
+ cos τ + u1 u6 − sin τ + u1 u4 + cos τ + u1 u5 − sin τ + u1 u3
ρ2 ωn ωn ρ ωn ωn
2
2q PY Ω Ω
−
ωn2 ρ 2 M cos 2 ω τ + u1 u3 + sin 2 ω τ + u1 u5
n n
q Ω Ω 2 Ω 2
+ 2 cos 2 τ + u1 u3u5 − sin 2 τ + u1 u3 + sin 2 τ + u1 u5
ρ 2 nω nω nω
u3′ = u4
Ω Ω Ω
u4′ = −2ς u4 − 1 − q cos 2 τ + u1 u3 + q sin 2 τ + u1 u5 + ε u2′ sin τ + u1
ωn ωn ωn
ε Ω P q Ω
2 (
Ω + ωnu2 ) cos τ + u1 − Y 2 sin 2 τ + u1
2
+
ωn ωn M ωn ωn
u5′ = u6
Ω Ω Ω
u6′ = −2ς u6 − 1 + q cos 2 τ + u1 u5 + q sin 2 τ + u1 u3 − ε u2′ cos τ + u1
ωn ωn ωn
ε Ω P q Ω
+ 2 ( Ω + ωn u2 ) sin τ + u1 + Y 2 cos 2 τ + u1
2
ωn ωn M ωn ωn
Eq. III-5
The Matlab programming found in Appendix C.4 is used to find a solution to Eq. III-5. The predicted
response amplitudes for the parameters in Table 4 are plotted in Figure 12.
20
Table 4: Parameters used for general vibration model
PY
ε
ωn ωt ς ςt ε ρ Py/M q k
2
1700rpm 2500rpm 0.2 0.04 0.001in 1.0in 1000in/sec 0.2 31.5
ωn ωt
2 2 ωn ωt
0.04 0.14
0.035 0.12
0.03
0.1
0.025
0.08
0.02
0.06
0.015
0.04
0.01
0.005 0.02
0 0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Motor Speed (RPM)
Lateral response amplitude peaks around 1/ 2ωn and ωn as predicted by the simplified model in II.4.
Torsional response amplitude peaks at1/ 2ωt and ωt as predicted by the simplified model in II.5. However
new peaks appear in the torsional response at 1/ 2ωn and ωn that were not predicted by the simplified
model. These peaks are associated with the interaction between the lateral and torsional vibration
The following chapters will attempt to identify trends in the torsional and lateral vibration.
21
III.4 Trends of Increasing Shaft Asymmetry
The general vibration model developed in III.3 may be useful to understand how a developing shaft
asymmetry affects the vibration of a rotor, for example in the case of a propagating crack. As defined in
II.2, the asymmetry factor, q represents the relative difference in stiffness between the strong and weak
axes of the shaft. q = 0 represents a perfectly symmetric shaft while increasing values of q represent
increasing asymmetry. If the stiffness in the strong direction is assumed constant ( kη = constant ), the
asymmetry factor q would increase as the stiffness in the weak direction ( kξ ) decreases. This would
reduce the average stiffness ( k ) and therefore reduce the lateral natural frequency, ωn .
1
ωn = ωn,original
1+ q
Eq. III-6
For the purpose of this study, it is assumed that the source of the lateral stiffness asymmetry does not affect
the torsional stiffness and therefore ωt = constant . The Matlab model show in Appendix C.5 is used to
model a rotor with the parameters in Table 5 and increasing shaft asymmetry factor, q . The resulting
Table 5: Parameters used in vibration model of rotor with increasing shaft asymmetry
ωn,original ωt ς ςt ε ρ Py/M
2
1850rpm 2500rpm 0.2 0.04 0.005in 1.0in 386.1in/sec
22
Figure 13: Vertical vibration amplitudes for rotor with increasing shaft asymmetry
Figure 14: Torsional vibration amplitudes for rotor with increasing shaft asymmetry
23
Figure 13 shows that lateral response at 1/ 2ωn grows with increasing asymmetry, as does the response at
ωn . However there is no apparent lateral response associated with torsional resonances at 1/ 2ωt and ωt .
As indicated by Eq. III-6, increase in q reduces ωn - an effect more apparent in the 2D plot.
It can be seen in Figure 14 that torsional response peaks at 1/ 2ωn , ωn , 1/ 2ωt and ωt grow with
increasing asymmetry. For low values of q , the peak lateral responses at ωn and 1/ 2ωn do not correlate
with significant torsional response, despite proximity in motor speed to ωt and 1/ 2ωt . However Torsional
response at ωt and 1/ 2ωt associated with the unbalance and side load respectively are significant even at
low values of q . Torsional response at 1/ 2ωt appears to grow faster with increasing q than the response
at ωt . For higher values of q , lateral vibration at 1/ 2ωn and ωn is reflected in peaks in torsional response.
The orbit response at 1/ 2ωn is plotted on the same scale for increasing values of q in Figure 15 through
Figure 19. Overall response grows with a characteristic two-loop pattern developing with increasing
asymmetry. This is consistent with the predicted response of the simplified model presented in Figure 5
0.012
0.008
Vertical Dispalcement (in)
0.004
-0.004
-0.008
-0.012
-0.012 -0.008 -0.004 0 0.004 0.008 0.012
Horizontal Displacem ent (in)
24
0.012
0.008
-0.004
-0.008
-0.012
-0.012 -0.008 -0.004 0 0.004 0.008 0.012
Horizontal Displacem ent (in)
0.012
0.008
Vertical Dispalcement (in)
0.004
-0.004
-0.008
-0.012
-0.012 -0.008 -0.004 0 0.004 0.008 0.012
Horizontal Displacem ent (in)
25
0.012
0.008
-0.004
-0.008
-0.012
-0.012 -0.008 -0.004 0 0.004 0.008 0.012
Horizontal Displacem ent (in)
0.012
0.008
Vertical Dispalcement (in)
0.004
-0.004
-0.008
-0.012
-0.012 -0.008 -0.004 0 0.004 0.008 0.012
Horizontal Displacem ent (in)
26
III.5 Effect on Response of the Proximity of ωt to ωn
To understand how torsional and lateral vibration interact, the model developed in III.3 is employed for
varying ratios of ωt ωn . Parameters from Table 5 are used, ωt is varied and the asymmetry factor is held
constant at q = 0.2 . Matlab programming found in Appendix C.6 is used to predict response, which is
plotted in Figure 20, Figure 21 and Figure 22. Lateral response remains completely unchanged, as indicated
by the overlapping plots in Figure 20. Torsional response at ωt and 1/ 2ωt both peak when ωt ωn = 0.5 ,
when ωt corresponds with 1/ 2ωn . Torsional response is less pronounced for ωt ωn = 1 and decreases as
ωt ωn increases beyond ωt ωn = 1 .
ωn
2 ωn
0.045
0.04
0.035
Lateral Amplitude (in pk-pk)
0.03
0.025
0.02
0.015
0.01
0.005
0
0 500 1000 1500 2000 2500 3000 3500 4000
Motor Speed (RPM)
ωt
ωn 0.25 0.5 0.75 1 1.25 1.5 1.75 2
27
Figure 21: Torsional response for varying ratios of ωt ωn (3D plot)
ωn
2 ωn
0.25
Torsional Amplitude (deg pk-pk)
0.2
0.15
0.1
0.05
0
0 500 1000 1500 2000 2500 3000 3500 4000
Motor Speed (RPM)
ωt
ωn 0.25 0.5 0.75 1 1.25 1.5 1.75 2
28
III.6 Effect of Increasing Side Load
The general vibration model developed in III.3 may be used to understand how a side load affects the
vibration of an asymmetric shaft rotor. In real systems, side load may vary with aerodynamic or gearbox
loads. Determining how changing loads are reflected in rotor response could help with diagnosis of an
asymmetry. The Matlab model in Appendix C.7 is used to predict rotor response for varying loads and the
Figure 23: Lateral response for increasing side load (3D plot)
Lateral response is primarily affected at 1/ 2ωn where overall response amplitude increases linearly with
side load. Torsional response increases with side load at 1/ 2ωn , 1/ 2ωt and ωt , with the most notable
sensitivity at 1/ 2ωt . Torsional resonances are not reflected in the lateral response.
29
Figure 24: Torsional response for increasing side load (3D plot)
30
Chapter IV Experiment
The objectives of the experiment are to provide information for verifying and improving the mathematical
model, and to investigate the practical implications of lateral and torsional vibration measurement as
diagnostic tools. To this aim, the experimental foundation is the Bently Nevada RK-4 Rotor Kit shown in
Figure 25. This assembly allows for a wide range of configurations and is adaptable for the specific needs
of this study. Two disks fixed to a common shaft and supported with plain rigid bearings are driven by an
electric motor through a flexible coupling. A shaft-mounted bearing allows the application of a side load.
The inboard disk is constrained by adjacent bearings to prevent lateral vibration. The outboard disk is
supported by an adjustable bearing span to allow lateral flexibility. Torsional flexibility arises from the
section of shaft connecting the two disks. Proximity probes are used to measure vibration and allow
Figure 25: Bently Nevada RK-4 Rotor Kit configured for a side loaded rotor
Based on estimates from the sizes of disks and shafts used by the RK-4 rotor kit, the bearing span of the
outboard mass can be configured to provide a lateral natural frequency as low as 2,500rpm or a torsional
natural frequency as low as 5,700rpm. Additional disks can be added to further reduce both natural
frequencies. Figure 25 shows the rotor configured for minimum lateral natural frequency while Figure 26
31
Figure 26: Rotor kit configured to minimize torsional natural frequency
The pulley system shown in Figure 27 is employed to provide an adjustable side load to the outboard rotor.
Calibrated masses are hung from a string which draws over a pulley and loads a shaft-mounted bearing near
the disk. Additional weights can be added to change loading conditions. Multiple pulleys can be used to
load the shaft either the horizontal or vertical (gravitational) direction. The advantage of this system is
flexibility to run a variety of side loads. However the masses hung have a potentially significant inertial
influence on the rotor that is not modeled by this study. Use of a spring or magnetic load may be more
32
To generate asymmetric lateral stiffness, four flats are carefully ground into the rotor shaft using a precision
machine tool. These flats are oriented 180 degrees apart in two locations along the shaft. This provides
symmetry about the center of the rotor’s bearing span as well as axis-symmetric lateral stiffness. Varying
flat depths result in varied magnitudes of stiffness asymmetry, as measured by load-deflection testing.
A direct current motor provides rotational power to the inboard end of the rotor system. A Bently Nevada
RK4 Speed Controller uses feedback from a coupling-mounted reluctor wheel (shown in Figure 29) to
control average motor rotational speed. Steady and ramped motor speeds can be achieved with this system.
33
Figure 29: Reluctor wheel for motor speed control
Lateral vibration is measured directly at the shaft using eddy current proximity probes as shown in Figure
30. The probes are oriented mutually perpendicular to measure vertical and horizontal displacement
respectively. Due to physical constraints and limitations in linear range, these probes may not always be
positioned near a rotor. Instead they are placed along sections of shaft exhibiting acceptable levels of
vibration. This vibration is assumed to be representative of the nearby rotor’s vibration. Measurements
made by these probes are assumed to be proportional to what would be measured at the rotor itself.
34
In addition to lateral probes, each disk is equipped with a transducer for measuring torsional vibration. The
primary components are a custom machined gear wheel attached to the shaft and two proximity probes that
are fixed 180 degrees apart in the horizontal plane, as shown in Figure 31.
Figure 31: Torsional transducer consisting of a gear wheel and two proximity probes
The principle of operation of the torsional transducer can be described as follows. As the gear wheel
rotates, proximity probes detect the passage of teeth. The time it takes for teeth to pass can be attributed to
2π R 2π R
∆t L = and ∆t R =
(
N RΩ i + V y ) (
N RΩ i − V y )
Eq. IV-1
where Ωi is the instantaneous rotative speed, Vy is the vertical velocity, R is the radius of the gear wheel,
N is the number of teeth, and ∆t L and ∆t R are the times between pulses generated by the left and right
probes respectively. The lateral component of the vibration signal ( Vy ) can be eliminated from the above
π ∆t L + ∆t R
Ωi =
N ∆t L ∆t R
Eq. IV-2
35
In general, the vibration of the disk can be express as
θ = (Ω motor − Ωi )t
Eq. IV-3
where θ is the general vibration angle with respect to the motor and Ω motor is the rotative speed of the
motor.
Bently Nevada equipment is widely used in industry for vibration monitoring and diagnostics. For this
project, the RK-4 Rotor Kit is equipped with a motor speed controller to provide steady rotation at the
Two Bently Nevada TK17 Torsional Vibration Signal Conditions are employed to simplify measurement of
torsional vibration and reduce post processing. Based on the relationships described in Eq. IV-1, Eq. IV-2
and Eq. IV-3, each signal conditioner processes the two proximity probe outputs of a torsional transducer
For data acquisition, a Bently Nevada ADRE 208-P works in unison with a personal computer to store and
process vibration signals in real time, allowing for a wide range of vibration monitoring and data storage.
The 208-P performs many of the same functions of oscilloscopes, spectrum analyzers, filters, and recording
instruments.
Seven signals are fed to the 208-P. Three signals come from both the inboard rotor and outboard rotor
respectively. These include feedback from a vertical proximity probe, horizontal proximity probe, and
torsional transducer (pre-processed by a TK17). A seventh signal from the rotor kit’s Keyphasor provides a
36
IV.3 Experimental Results
In an attempt to excite torsional activity and lateral-torsional vibration interaction, the rotor kit is
configured as shown in Figure 32 (additional notes regarding this trial can be found in Appendix D.1). Two
disks are used at both the inboard and outboard positions to reduce ωt and ωn . The calculated values
assuming rigid bearings and the dimensions of the rotor kit are ωt = 4,100rpm and ωn = 5,800rpm . A very
Py
heavy side load is applied to achieve = 2,600 in/sec 2 .
M
The resulting vibration amplitudes as recorded by the ADRE 208-P are shown in the following figures. The
direct and 1X lateral response of the outboard rotor is shown in Figure 33 and Figure 34. A broad 1X
response peak is noted in the horizontal direction near 4,200rpm, but resonant peaks are not obvious in the
vertical response.
37
Figure 33: Vertical response, direct & 1X
38
The 2X filtered lateral response shown in Figure 35 and Figure 36 shows two or more distinctive peaks,
potentially related to the 1X and 2X excitations associated with the unbalance and side-load asymmetry
interaction respectively. 2X response peaks appear to occur at different speeds for the horizontal and
vertical directions. This may be an effect of the mass used for side-loading. It should be noted that this
39
Figure 36: Horizontal response, 2X only
Torsional responses for the inboard and outboard disks separately are shown in Figure 37 and Figure 38.
The difference in the outboard and inboard response is shown in Figure 39. A 1X resonant peak is clearly
shown at 5,400rpm, but no peak is obvious at half this speed, as might be expected from the simulation
Py
considering the high ratio of . Observing the outboard rotor, some broad 1X response peaks also appear
M
at about 900rpm and 3,600rpm, but these do not appear to correlate to either the natural frequency at
40
Figure 37: Torsional response of inboard disk, direct & 1X
41
Figure 39: Overall rotor torsional response (difference between inboard & outboard), 1X & 2X
Additional trials were performed, but due to difficulties obtaining data from this apparatus, meaningful
results were not obtained. Future experimentation should be performed to verify or refute the results of the
42
Chapter V Conclusion
The research presented in this paper aims to identify characteristics and trends associated with an
asymmetric shaft rotor system. Understanding the behavior of a compromised rotor system is important for
the early detection of failure. If trends can be identified through carefully correlated simulation, vibration
experts might use those trends to prevent catastrophic, hazardous or costly failures of real machines.
This paper examines a rotor consisting of two disks connected by a torsionally flexible shaft. One disk is
assumed to have rigid lateral support while the other is allowed lateral flexibility. A side load is applied to
the lateral flexible disc and the shaft is considered to have asymmetric stiffness. The model mimics a real
system that might consist of a turbine driving a compressor where gravitational or aerodynamic loading
provides a side load and a transverse crack or similar defect results in shaft stiffness asymmetry.
Energy equations are used to derive the mathematical model proposed to describe the system. The model
accounts for four degrees of freedom: rotation for both disks and translation for one. Parameters used in the
mathematical model such as side loads, stiffness, damping ratios, eccentricity and natural frequencies are
The shaft stiffness model for this system assumes loads in orthogonal directions along specific axes of the
shaft produce deflection only in those directions, without interaction. This model is a special case that
allows exploration of the effects of reducing stiffness in one of these directions. The asymmetry factor
defines the magnitude of this unidirectional reduction of stiffness. Real systems may exhibit a similar
reduction of stiffness along one direction in the case of a transverse crack, for example. The asymmetry in
stiffness effectively couples the torsional and lateral vibration of the rotor system.
Special cases are identified that allow simplified evaluation of the system. First, torsional vibration is
assumed negligible and lateral vibration is evaluated independently. The equations of motions describing
this case resemble a Jeffcott rotor with additional forcing terms related to the asymmetry. A 1X forcing
43
term associated with imbalance is present along with 2X forcing terms associated with the asymmetry
factor and side load. Computer simulation of this case reveals lateral response peaks at the lateral natural
frequency and half the lateral natural frequency. Orbit response at half the natural frequency displays a
Also of interest is the case where lateral vibration is negligible and torsional vibration dominates. The
equations of motion describing this case resemble a simple linear oscillator with two forcing terms. A 1X
forcing term is associated with the unbalance, side load and asymmetry while a 2X forcing term is
associated with the side load and asymmetry only. Computer simulation shows torsional response peaks at
the torsional natural frequency and half the torsional natural frequency. An important characteristic
identified in both the mathematical model and computer simulation is the response at half the torsional
natural frequency grows approximately as the square of the side load while the response at the natural
frequency grows approximately linearly with side load. The response amplitudes are approximately equal
when the imbalance eccentricity and the displacement due to side load are equal.
When both lateral and torsional vibration and their interactions are considered, the equations of motion are
more complex and the system must be characterized through parametric computer simulation. Specific
studies are conducted to determine the effects of shaft asymmetry, side load, and proximity of the torsional
and lateral natural frequencies. By observing changes in the response of the system resulting from changes
in these parameters, conclusions can be made about their effects which may later be compared with real
Shaft asymmetry in a side loaded rotor results in torsional response. As the asymmetry factor increases,
torsional amplitude increases around half and one times the torsional natural frequency. Lateral vibration
also begins to influence torsional vibration as the asymmetry factor increases. Simulation predicts
additional torsional response at the lateral and half the lateral natural frequency as asymmetry increases.
Torsional vibration should be monitored at these frequencies to alert the possibility of developing shaft
asymmetry
44
The effect of asymmetry on lateral response is primarily seen at half the lateral natural frequency. Here the
orbit response changes distinctly from a circular to a loop-in-loop shape as asymmetry increases. Overall
lateral vibration amplitude also increases with asymmetry at a higher rate at half the natural frequency than
at the natural frequency. Without shaft asymmetry, no resonance is present at half the natural frequency.
Torsional vibration does not appear to affect lateral vibration as in the opposite case, with no significant
The proximity of the torsional and lateral natural frequencies has a strong influence on the amplitude of
torsional response at the torsional natural frequency. The amplitude of the torsional response at torsional
resonance peaks when it corresponds to the lateral half resonance. Response at half the torsional natural
Side loading impacts both torsional and lateral response. Lateral vibration is primarily affected by the 2X
forcing term associated with side load, which causes increased lateral amplitude near half the resonant
frequency. Side loading produces very little change in response at the resonant speed. Torsional response is
strongly influenced by side load, especially at half the torsional natural frequency. As side load increases,
lateral response at half the lateral natural frequency begins to cause torsional response at that speed as well.
In an attempt to correlate simulation with a real system, an experiment is conducted in which lateral and
torsional vibration is measured for a two disc rotor with an asymmetric shaft and side load. Lateral
vibration is measured with conventional proximity probes while torsional vibration is measured through
gear tooth transducers and special signal processing. Some vibration characteristics are observed to
correlate with simulation, notably the existence of torsional resonance and two peaks in the lateral
amplitude plot. However, verification of trending is not obtained. Further experimentation should be
performed.
45
Bibliography
1. Wauer, J., 1990, “On the Dynamics of Cracked Rotors: A Literature Survey”, Applied Mechanics
2. Dimarogonas, A. D., 1996, “Vibration of Cracked Structures: A State of the Art Review,”
3. Gasch, R., 1976, “Dynamic behavior of a simple rotor with a cross-sectional crack”, Paper
4. Gasch, R. A, 1993, “Survey of the Dynamic Behavior of a Simple Rotating Shaft with a
5. Chan, R. K. C. and Lai, T. C., 1995, “Digital Simulation of a rotating shaft with a transverse
System Containing a Transverse Crack in a Rotor”, ASME Journal of Vibration, Acoustics, Stress,
7. Sawicki, J. T., Wu, X., Baaklini, G.Y. and Gyekenyesi, A., 2003, “Vibration-Based Crack
Vol. 5046.
8. Sawicki, J. T., Bently, D. E., Wu, X., Baaklini, G.Y. and Friswell, M. I., 2003, “Dynamic
9. Sawicki, J.T., Wu, X., Gyekenyesi, A. L. and Baaklini, G. Y., 2005 “Application of Nonlinear
Dynamic Analysis for Diagnosis of Cracked Rotor Vibration Signatures”, submitted to SPIE
10. Muszynska, A., Goldman, P. and Bently, D. E., 1992, “Torsional/Lateral Cross-Coupled
Responses Due to Shaft Anisotropy: A New Tool in Shaft Crack Detection”, I. Mech. E., C432-
46
11. Bently, D. E., Goldman, P. and Muszynska, A., 1997, ““Snapping” Torsional Response of an
Anisotropic Radially Loaded Rotor,” Journal of Engineering for Gas Turbines and Power, Vol.
12. Wu, X., 2005, Doctoral dissertation, “Vibration-based Crack-induced Damage Detection of Shaft-
13. Wu, X.; Meagher, J.; 2007, “Characterization of Shaft Cracks and Rotating Asymmetries using
Full Spectrum and Orbit Plots”, ISCORMA-4, Calgary, Alberta, Canada, 27-31 August 2007.
14. Wu, X., Meagher, J., Judd, C.; 2007, “Investigation of coupled lateral and torsional vibrations of a
cracked rotor under radial load”, IMAC XXV, Society for Experimental Mechanics, Orlando,
15. Wu, X., Meagher, J., Volume 2008, “A Two-Disk Extended Jeffcott Rotor Model Distinguishing a
Shaft Crack from Other Rotating Asymmetries”, International Journal of Rotating Machinery,
Article ID 846365.
47
Appendix A Nomenclature
48
Appendix B Detailed Calculations
Fξ kξ 0 ξ
F = 0 k
η η η
kξ 0
KR =
0 kη
in an inertial coordinate system, the stiffness matrix would be
FX k11 k12 X
F = k
Y 21 k22 Y
k k
K I = 11 12
k21 k22
perform a coordinate transformation to determine the values of the inertial stiffness matrix
K I = TK RT −1
cos Φ − sin Φ kξ 0 cos Φ sin Φ
=
sin Φ cos Φ 0 kη − sin Φ cos Φ
cos Φ − sin Φ kξ cos Φ kξ sin Φ
=
sin Φ cos Φ − kη sin Φ kη cos Φ
k k12
= 11
k21 k22
49
k11 = kξ cos 2 Φ + kη sin 2 Φ
1 + cos 2Φ 1 − cos 2Φ
= kξ + kη
2 2
kξ + kη kη − kξ
= − cos 2Φ
2 2
= k − kq cos 2Φ
= k (1 − q cos 2Φ )
k12 = k21
k 22 = kξ sin 2 Φ + kη cos 2 Φ
1 − cos 2Φ 1 + cos 2Φ
= kξ + kη
2 2
kξ + kη kη − kξ
= + cos 2Φ
2 2
= k + kq cos 2Φ
= k (1 + q cos 2Φ )
1 − q cos 2Φ −q sin 2Φ
KI = k
−q sin 2Φ 1 + q cos 2Φ
50
B.2 Geometric Relationships
Φ = Ωt + ϕ − Φ 0
& = Ω + ϕ&
Φ
&& = ϕ&&
Φ
Θ = Ωt + θ − Θ 0
& = Ω + θ&
Θ
&& = θ&&
Θ
xcm = X + ε cos ( Φ + δ )
& sin ( Φ + δ )
x&cm = X& − εΦ
ycm = Y + ε sin ( Φ + δ )
y& = Y& + εΦ& cos ( Φ + δ )
cm
assuming lateral vibration is about the static offset due to side loading
PX
X = XM −
k
1 P
X = XM − 2 X
ωn M
X& = X& M
X&& = X&& M
PY
Y = YM −
k
1 P
Y = YM − 2 Y
ωn M
Y& = Y&M
Y&& = Y&&
M
51
kξ + kη kη − kξ k kt C Ct
k= , q= , ωn = , ωt = ,ς = , ςt = , I = ρ M , τ = ωnt
2
2 2k M I kM kt I
Fξ kξ 0 ξ
F = 0 k
η η η
kξ 0
KR =
0 kη
FX k11 k12 X
F = k
Y 21 k 22 Y
k k
K I = 11 12
k 21 k 22
perform a coordinate transformation to determine the values of the inertial stiffness matrix
KI = TKRT −1
cos Φ − sin Φ kξ 0 cos Φ sin Φ
=
sin Φ cos Φ 0 kη − sin Φ cos Φ
cos Φ − sin Φ kξ cos Φ kξ sin Φ
=
sin Φ cos Φ −kη sin Φ kη cos Φ
k k
= 11 12
k21 k22
52
k11 = kξ cos2 Φ + kη sin2 Φ
1 + cos2Φ 1 − cos2Φ
= kξ + kη
2 2
k +k k −k
= ξ η − η ξ cos2Φ
2 2
= k − kq cos 2Φ
= k (1 − q cos2Φ)
k12 = k21
1 − q cos 2Φ − q sin 2Φ
KI = k
− q sin 2Φ 1 + q cos 2Φ
53
B.5 Detailed Derivation
T = 12 I Φ
&2+1I Θ& 2 + 1 Mx& 2 + 1 My& 2
2 0 2 cm 2 cm
= 12 I Φ
& 2 + 1 I 0Θ
2 2 { & sin ( Φ + δ ) + Y& + εΦ
& 2 + 1 M X& − εΦ 2
& cos ( Φ + δ )
2
}
= 12 I Φ
&2+1I Θ
2 0 2 { & Y& cos ( Φ + δ ) − X& sin ( Φ + δ ) + ε 2 Φ
& 2 + 1 M X& 2 + Y& 2 + 2εΦ
&2 }
∂T
=0
∂X
∂T & sin ( Φ + δ )
= MX& − ε M Φ
∂X&
∂ ∂T
= MX − ε M Φ sin ( Φ + δ ) − ε M Φ cos ( Φ + δ )
&& && &2
∂t ∂X&
∂T
=0
∂Y
∂T & cos ( Φ + δ )
= MY& + ε M Φ
∂Y&
∂ ∂T
& = MY + ε M Φ cos ( Φ + δ ) − ε M Φ sin ( Φ + δ )
&& && &2
∂t ∂Y
∂T
=0
∂Θ
∂T
= I 0Θ&
∂Θ&
∂ ∂T
& = I0Θ
&&
∂t ∂Θ
54
∂T & X& cos ( Φ + δ ) + Y& sin ( Φ + δ )
= −ε M Φ
∂Φ
∂T
= ( I + ε 2M ) Φ & + ε M Y& cos ( Φ + δ ) − X& sin ( Φ + δ )
∂Φ&
∂ ∂T
& = ( I + ε M ) Φ + ε M (Y − X Φ ) cos ( Φ + δ ) − ( X + Y Φ ) sin ( Φ + δ )
2 && && & & && & &
∂t ∂Φ
k k12 X 1
[X Y ] 11 + k (Φ − Θ)
2
U= 1
2
k21 k22 Y 2 t
k12 = k21
∴U = 1
2 (k
11 X 2 + k22Y 2 ) + k12 XY + 12 kt ( Φ − Θ )
2
∂U
= k (1 − q cos 2Φ ) X − kq sin 2ΦY
∂X
∂U
= k (1 + q cos 2Φ ) Y − kq sin 2ΦX
∂Y
∂U
= kq sin 2Φ ( X 2 − Y 2 ) − 2kq cos 2ΦXY + kt ( Φ − Θ )
∂Φ
∂U
= −kt ( Φ − Θ )
∂Θ
55
Dissipation Function of the System
∂D
= CX&
∂X&
∂D
= CY&
&
∂Y
∂D
= Ct ( Φ &)
& −Θ
∂Φ
&
∂D
= −Ct ( Φ &)
& −Θ
&
∂Θ
56
∂ ∂T ∂T ∂U ∂D
− + + = − PX
∂t ∂X& ∂X ∂X ∂X&
&& sin ( Φ + δ ) − ε M Φ
MX&& − ε M Φ & 2 cos ( Φ + δ ) + k (1 − q cos 2Φ ) X − kq sin 2ΦY + CX& = − P
X
substituting
1 PX 1 P
X = XM − Y = YM − 2 Y
ωn2 M X& = X&
, && &&
M , X = XM , ωn M
, Y& = Y&M , Y&& = Y&&M
1 P
X&&M + 2ςωn X& M + ωn2 (1 − q cos 2Φ ) X M − 2 X
ωn M
substituting
Φ = Ωt + ϕ − Φ 0 , Φ
& = Ω + ϕ& Φ
,
&& = ϕ&&
( )
X&&M + 2ςωn X& M + ωn2 1 − q cos ( 2 ( Ωt + ϕ − Φ0 ) ) X M − ωn2 q sin ( 2 ( Ωt + ϕ − Φ0 ) ) YM = εϕ&& sin ( Ωt + ϕ − Φ0 + δ )
P P
+ε ( Ω + ϕ& ) cos ( Ωt + ϕ − Φ0 + δ ) − Y q sin ( 2 ( Ωt + ϕ − Φ0 ) ) − X q cos ( 2 ( Ωt + ϕ − Φ0 ) )
2
M M
d( ) d( ) ′ d ( ) = ω 2 d ( ) = ω 2 ( )′′
2 2
τ = ωnt , dt = ω = ω ( )
dτ dτ 2
n n 2 n n
let , dt
57
Ω Ω Ω
X M′′ + 2ς X M′ + 1 − q cos 2 τ + ϕ − Φ0 X M − q sin 2 τ + ϕ − Φ0 YM = εϕ ′′ sin τ + ϕ − Φ0 + δ
ωn ωn ωn
ε Ω P q Ω P q Ω
2 (
Ω + ωnϕ ′) cos τ + ϕ − Φ0 + δ − Y 2 sin 2 τ + ϕ − Φ0 − X
2
+ ω 2 cos 2 ω τ + ϕ − Φ0
ωn ωn M ωn ωn M n n
∂ ∂T ∂T ∂U ∂D
− + + = − PY
∂t ∂Y& ∂Y ∂Y ∂Y&
&& cos ( Φ + δ ) − ε M Φ
MY&& + ε M Φ & 2 sin ( Φ + δ ) + k (1 + q cos 2Φ ) Y − kq sin 2ΦX + CY& = − P
Y
substituting
1 PX 1 P
X = XM − Y = YM − 2 Y
ωn2 M X& = X&
, && &&
M , X = XM , ωn M
, Y& = Y&M , Y&& = Y&&M
1 P
Y&&M + 2ςωnY&M + ωn2 (1 + q cos 2Φ ) YM − 2 Y
ωn M
& 2 sin ( Φ + δ ) − PX
&& cos ( Φ + δ ) + εΦ
= −εΦ
PY
q sin 2Φ + q cos 2Φ
M M
substituting
Φ = Ωt + ϕ − Φ 0 , Φ
& = Ω + ϕ& Φ
,
&& = ϕ&&
58
( )
Y&&M + 2ςωnY&M + ωn2 1 + q cos ( 2 ( Ωt + ϕ − Φ 0 ) ) YM − ωn2 q sin ( 2 ( Ωt + ϕ − Φ 0 ) ) X M = −εϕ&& cos ( Ωt + ϕ − Φ 0 + δ )
P P
+ε ( Ω + ϕ& ) sin ( Ωt + ϕ − Φ 0 + δ ) − X q sin ( 2 ( Ωt + ϕ − Φ 0 ) ) + Y q cos ( 2 ( Ωt + ϕ − Φ 0 ) )
2
M M
d( ) d( ) ′ d ( ) = ω 2 d ( ) = ω 2 ( )′′
2 2
τ = ωnt dt = ω = ω ( )
dτ dτ 2
n n n n
let , dt 2
,
Ω Ω Ω
YM′′ + 2ς YM′ + 1 + q cos 2 τ + ϕ − Φ 0 YM − q sin 2 τ + ϕ − Φ 0 X M = −εϕ ′′ cos τ + ϕ − Φ 0 + δ
ωn ωn ωn
ε Ω P q Ω P q Ω
2 (
Ω + ωnϕ ′ ) sin τ + ϕ − Φ 0 + δ − X 2 sin 2 τ + ϕ − Φ 0 + Y 2 cos 2 τ + ϕ − Φ 0
2
+
ωn ωn M ωn ωn M ωn ωn
59
Lagrange Equation for Φ
∂ ∂T ∂T ∂U ∂D
& − + + =0
∂t ∂Φ ∂Φ ∂Φ ∂Φ
&
I 2 &&
+ ε Φ + ε Y cos ( Φ + δ ) − X sin ( Φ + δ ) + ωn q sin 2Φ ( X − Y )
&& && 2 2 2
M
−2ωn2 q cos 2ΦXY + t ( Φ − Θ ) + t ( Φ − Θ) = 0
k C & &
M M
60
from Lagrange equation for Y,
it follows that
61
Y&& cos ( Φ + δ ) − X&& sin ( Φ + δ ) = 2ςωn sin ( Φ + δ ) X& − cos ( Φ + δ ) Y&
+ωn2 (1 − q cos 2Φ ) sin ( Φ + δ ) X − ωn2 (1 + q cos 2Φ ) cos ( Φ + δ ) Y
&& + ω 2 q sin 2Φ cos ( Φ + δ ) X − sin ( Φ + δ ) Y + PX sin ( Φ + δ ) − PY cos ( Φ + δ )
−εΦ n M M
62
substituting into Lagrange equation for Φ
(ρ 2
+ε 2 )Φ
&& + ε 2ςω sin ( Φ + δ ) X& − cos ( Φ + δ ) Y&
n
+εωn2 sin ( Φ + δ ) X − cos ( Φ + δ ) Y + εωn2 q sin ( Φ − δ ) X − cos ( Φ − δ ) Y
+ε
PX P && + ω 2 q sin 2Φ ( X 2 − Y 2 )
sin ( Φ + δ ) − ε Y cos ( Φ + δ ) − ε 2Φ n
M M
−2ωn2 q cos 2ΦXY + ρ 2ωt 2 ( Φ − Θ ) + 2 ρ 2ς tωt ( Φ
& −Θ &)=0
&& + ω 2 ( Φ − Θ ) + 2ς ω ( Φ
Φ & ) = ε 2ςωn cos ( Φ + δ ) Y& − sin ( Φ + δ ) X&
& −Θ
t t t
ρ2
εωn2 εωn2 q
+ 2 cos ( Φ + δ ) Y − sin ( Φ + δ ) X + 2 cos ( Φ − δ ) Y − sin ( Φ − δ ) X
ρ ρ
ε P PX ωn2 q
+ 2 Y cos ( Φ + δ ) − sin ( Φ + δ ) + 2 2 cos 2ΦXY − sin 2Φ ( X − Y )
2 2
ρ M M ρ
63
substituting
1 PX 1 PY
X = XM − Y = Y −
ωn2 M X& = X&
, && &&
M , X = XM ,
M
ωn2 M
, Y& = Y&M , Y&& = Y&&M
&& + ω 2 ( Φ − Θ ) + 2ς ω ( Φ
Φ & ) = ε 2ςωn cos ( Φ + δ ) Y& − sin ( Φ + δ ) X&
& −Θ
t t t
ρ2 M M
εωn2 εωn2 q
+ cos ( Φ + δ ) Y − sin ( Φ + δ ) X + cos ( Φ − δ ) YM − sin ( Φ − δ ) X M
ρ2 M M
ρ2
ε PY PX ε q PY PX
− 2 cos ( Φ + δ ) − sin ( Φ + δ ) − 2 cos ( Φ − δ ) − sin ( Φ − δ )
ρ M M ρ M M
ε P P
+ 2 Y cos ( Φ + δ ) − X sin ( Φ + δ )
ρ M M
ω 2q 1 P
2 2
1 P 1 P 1 P
+ n2 2cos 2Φ X M − 2 X YM − 2 Y − sin 2Φ X M − 2 X − YM − 2 Y
ρ ωn M ωn M ωn M ωn M
&& + ω 2 ( Φ − Θ ) + 2ς ω ( Φ
Φ & ) = ε 2ςωn cos ( Φ + δ ) Y& − sin ( Φ + δ ) X&
& −Θ
t t t
ρ2 M M
εωn2 εωn2 q
+ cos ( Φ + δ ) M
Y − sin ( Φ + δ ) M
X + cos ( Φ − δ ) YM − sin ( Φ − δ ) X M
ρ2 ρ2
ε q PY PX
− 2 cos ( Φ − δ ) − sin ( Φ − δ )
ρ M M
ω 2q 1 P
2 2
1 P 1 P 1 P
+ n2 2cos 2Φ X M − 2 X YM − 2 Y − sin 2Φ X M − 2 X − YM − 2 Y
ρ ωn M ωn M ωn M ωn M
64
aside
1 P 1 PY 1 PX
2
1 PY
2
2cos 2Φ X M − 2 X YM − 2 − sin 2Φ X M − 2 − YM − 2
ωn M ωn M ωn M ωn M
1 P 1 P 1 P P
= 2cos 2Φ X M YM − 2 X YM − 2 Y X M + 4 Y X
ωn M ωn M ωn M M
2 2 PX 1 PX
2
2 2 PY 1 PY
2
− sin 2Φ X M − 2 X M + 4 + sin 2Φ YM − 2 YM + 4
ωn M ωn M ωn M ωn M
1 P 1 PY 1 PX
2
1 PY
2
2cos 2Φ X M − 2 X YM − 2 − sin 2Φ X M − 2 − YM − 2
ωn M ωn M ωn M ωn M
2 P PX PY PX
= 2
cos 2Φ − Y X M − YM − sin 2Φ YM − X M
ωn M M M M
1 PY
2 2
PY PX PX
+ 4 2 cos 2Φ − sin 2Φ + sin 2Φ
ωn M M M M
+2 cos 2ΦX M YM − sin 2ΦX M2 + sin 2ΦYM2
65
combining
&& + ω 2 ( Φ − Θ ) + 2ς ω ( Φ & ) = − ε q PY PX
Φ cos ( Φ − δ ) − sin ( Φ − δ )
& −Θ
t t t
ρ 2 M M
q P P PX
2
PY
2
+ 2 2
2 cos 2Φ Y X − sin 2Φ + sin 2Φ +Γ
ωn ρ M M M M
where
ε 2ςωn
Γ= cos ( Φ + δ ) Y&M − sin ( Φ + δ ) X& M
ρ2
εωn2 εωn2 q
+ cos ( Φ + δ ) Y − sin ( Φ + δ ) X + cos ( Φ − δ ) YM − sin ( Φ − δ ) X M
ρ2 M M
ρ2
2q P P P PX
+ 2
cos 2Φ − Y X M − X YM − sin 2Φ Y YM − XM
ρ M M M M
ω 2q
+ n2 2 cos 2ΦX M YM − sin 2ΦX M2 + sin 2ΦYM2
ρ
substituting
Φ = Ωt + ϕ − Φ 0 , Φ
& = Ω + ϕ& Φ
,
&& = ϕ&& Θ = Ωt + θ − Θ0 Θ
, ,
& = Ω + θ& Θ
,
&& = θ&&
ε q PY PX
ϕ&& + ωt 2 (ϕ − θ − Φ0 + Θ0 ) + 2ς tωt (ϕ& − θ& ) = − 2 cos ( Ωt + ϕ − Φ 0 − δ ) − sin ( Ωt + ϕ − Φ0 − δ )
ρ M M
q PY PX PX
2
PY
2
where
66
ε 2ςωn
Γ= cos ( Ωt + ϕ − Φ 0 + δ ) Y&M − sin ( Ωt + ϕ − Φ 0 + δ ) X& M
ρ2
εωn2
+ cos ( Ωt + ϕ − Φ 0 + δ ) YM − sin ( Ωt + ϕ − Φ 0 + δ ) X M
ρ2
εωn2 q
+ 2 cos ( Ωt + ϕ − Φ 0 − δ ) YM − sin ( Ωt + ϕ − Φ 0 − δ ) X M
ρ
2q P P P P
+ 2
cos ( 2 ( Ωt + ϕ − Φ 0 ) ) − Y X M − X YM − sin ( 2 ( Ωt + ϕ − Φ 0 ) ) Y YM − X X M
ρ M M M M
ω 2q
+ n2 2 cos ( 2 ( Ωt + ϕ − Φ 0 ) ) X M YM − sin ( 2 ( Ωt + ϕ − Φ 0 ) ) X M2 + sin ( 2 ( Ωt + ϕ − Φ 0 ) ) YM2
ρ
67
d( ) d( ) ′ d2( ) 2 d ( )
2
= ωn = ωn ( ) = ωn = ωn2 ( )′′
let
τ = ω n t , dt dτ , dt 2
d τ 2
2
ωt 2ς t ωt εq PY Ω PX Ω
ϕ ′′ + (ϕ − θ − Φ 0 + Θ 0 ) + (ϕ ′ − θ ′ ) = − 2 2 cos ω τ + ϕ − Φ0 − δ − M sin ω τ + ϕ − Φ 0 − δ
ω
n ω n ωnρ M n n
q Ω P P Ω PX
2
Ω PY
2
+ 4 2
2cos 2 τ + ϕ − Φ 0 Y X − τ + ϕ − Φ + τ + ϕ − Φ + Γn
0 0
sin 2 sin 2
ωn ρ
ωn M M ωn M ωn M
ε 2ς Ω Ω
Γn = 2
cos τ + ϕ − Φ 0 + δ YM′ − sin τ + ϕ − Φ 0 + δ X M′
ρ ωn ωn
ε Ω Ω
+ cos τ + ϕ − Φ 0 + δ YM − sin τ + ϕ − Φ 0 + δ X M
ρ2 ωn ωn
εq Ω Ω
+ 2
cos τ + ϕ − Φ 0 − δ YM − sin τ + ϕ − Φ 0 − δ X M
ρ ωn ωn
2q Ω P P Ω P P
+ 2 2
cos 2 τ + ϕ − Φ 0 − Y X M − X YM − sin 2 τ + ϕ − Φ 0 Y YM − X X M
ωn ρ ωn
M M
ωn
M M
q Ω Ω Ω
+ 2 2 cos 2 τ + ϕ − Φ 0 X M YM − sin 2 τ + ϕ − Φ 0 X M2 + sin 2 τ + ϕ − Φ 0 YM2
ω ω
ρ n n ωn
68
Lagrange Equation for Θ
∂ ∂T ∂T ∂U ∂D
& − + + = −TDR
∂t ∂Θ ∂Θ ∂Θ ∂Θ &
IΘ&& − k ( Φ − Θ ) − C ( Φ
& −Θ& ) = −T
0 t t DR
&& − kt ( Φ − Θ ) − Ct ( Φ
Θ & ) = − TDR
& −Θ
I0 I0 I0
&& − kt ( Φ − Θ ) − Ct ( Φ
Θ & ) = − kc θ − Cc θ&
& −Θ
I0 I0 I0 I0
substituting
Φ = Ωt + ϕ − Φ 0 , Φ
& = Ω + ϕ& Φ
,
&& = ϕ&& Θ = Ωt + θ − Θ0 Θ
, ,
& = Ω + θ& Θ
,
&& = θ&&
( )
I 0θ&& − kt (ϕ − Φ 0 − θ + Θ0 ) − Ct ϕ& − θ& = − kcθ − Ccθ&
Cc + Ct & kc + kt k C
θ&& + θ+ θ = t (ϕ + Θ0 − Φ 0 ) + t ϕ&
I0 I0 I0 I0
Ct & k k C
θ&& + (1 + CR ) θ + (1 + K R ) t θ = t (ϕ + Θ0 − Φ 0 ) + t ϕ&
I I I I
θ + (1 + CR ) 2ς t ωtθ + (1 + K R ) ωt θ = ωt (ϕ + Θ0 − Φ 0 ) + 2ς t ωtϕ&
&& & 2 2
d( ) d( ) ′ d2( ) 2 d ( )
2
= ω = ω n( ) = ω = ωn2 ( )′′
let
τ = ω t
n , dt
n
dτ , dt
2 n
dτ 2
2 2
ω ω ω ω
θ ′′ + (1 + CR ) 2ς t t θ ′ + (1 + K R ) t θ = t ϕ + 2ς t t ϕ ′
ωn ωn ωn ωn
69
Special Case, Torsional Vibration Only
Assume
X M = 0 , YM = 0 , θ = 0
Let
PX = 0 , Φ 0 = 0 , Θ0 = 0 , δ = 0
2
ωt 2ς t ωt εq P Ω q Ω PY
2
ϕ ′′ + ϕ+ ϕ′ = − 2 2 Y cos τ + ϕ + sin
ω
2 τ + ϕ
ω
n ω n ω nρ M ωn ωn ρ
4 2
n M
u1 = ϕ u2 = ϕ ′
Let the states and
u1′ = u2
2
ω 2ς ω εq P Ω q Ω PY
2
u2′ = − t u1 − t t u2 − 2 2 Y τ + + τ +
1 ω 1
cos u sin 2 u
ω
n ω n ωnρ M ωn ωn ρ
4 2
n M
70
Special Case, Lateral Vibration Only
Assume
ϕ =0, θ =0
Let
PX = 0 , Φ 0 = 0 , Θ0 = 0 , δ = 0
Ω Ω ε Ω P q Ω
X M′′ + 2ς X M′ + 1 − q cos 2 τ X M − q sin 2 τ YM = 2 Ω2 cos τ − Y 2 sin 2 τ
ωn ωn ωn ωn M ωn ωn
Ω Ω ε Ω P q Ω
YM′′ + 2ς YM′ + 1 + q cos 2 τ YM − q sin 2 τ X M = 2 Ω2 sin τ + Y 2 cos 2 τ
ωn ωn ωn ωn M ωn ωn
u3′ = u4
Ω Ω ε Ω P q Ω
u4′ = −2ς u4 − 1 − q cos 2 τ u3 + q sin 2 τ u5 + 2 Ω2 cos τ − Y 2 sin 2 τ
ωn ωn ωn ωn M ωn ωn
u5′ = u6
Ω Ω ε Ω P q Ω
u6′ = −2ς u6 − 1 + q cos 2 τ u5 + q sin 2 τ u3 + 2 Ω2 sin τ + Y 2 cos 2 τ
ωn ωn ωn ωn M ωn ωn
71
General Vibration
Let
PX = 0 , Φ 0 = 0 , Θ0 = 0
2
ωt 2ς tωt εq P Ω
ϕ ′′ = (θ − ϕ ) + (θ ′ − ϕ ′) − 2 2 Y
cos τ + ϕ
ωn ωn ωn ρ M ωn
q Ω P
2
+ 4 2 sin 2 τ + ϕ Y + Γn
ωn ρ
ωn M
ε 2ς Ω Ω
Γn = cos τ + ϕ YM′ − sin τ + ϕ X M′
ρ2 ωn ωn
ε Ω Ω
+ cos τ + ϕ Y − sin τ + ϕ X
ρ 2 ωn ωn
M M
εq Ω Ω
+ 2
cos τ + ϕ YM − sin τ + ϕ X M
ρ ωn ωn
2q PY Ω Ω
+ − cos 2 τ + ϕ X M − sin 2 τ + ϕ YM
ωn2 ρ 2 M ωn ωn
q Ω Ω 2 Ω 2
+ 2 cos
2 τ + ϕ
M M
X Y − sin
2 τ + ϕ
M
X + sin
2 τ + ϕ YM
ρ 2 ωn ωn ωn
2 2
ωt ω ω ω
θ ′′ = − (1 + CR ) 2ς t θ ′ − (1 + K R ) t θ + t ϕ + 2ς t t ϕ ′
ωn ω
n ω
n ω n
Ω Ω Ω
X M′′ = −2ς X M′ − 1 − q cos 2 τ + ϕ X M + q sin 2 τ + ϕ YM + εϕ ′′ sin τ + ϕ
ω ω ω
n n n
ε Ω P q Ω
2 (
Ω + ωnϕ ′) cos τ + ϕ − Y 2 sin 2 τ + ϕ
2
+
ωn ωn M ωn ωn
72
Ω Ω Ω
YM′′ = −2ς YM′ − 1 + q cos 2 τ + ϕ YM + q sin 2 τ + ϕ X M − εϕ ′′ cos τ + ϕ
ω ω ω
n n n
ε Ω P q Ω
+ 2 ( Ω + ωnϕ ′ ) sin τ + ϕ + Y 2 cos 2 τ + ϕ − Φ 0
2
ωn ωn M ωn ωn
u1′ = u2
2
ω 2ς ω εq P Ω
u2′ = t ( u7 − u1 ) + t t ( u8 − u2 ) − 2 2 Y cos τ + u1
ωn ωn ωn ρ M ωn
q Ω P
2
+ 4 2 sin 2 τ + u1 Y + Γn
ωn ρ ω
n M
Ω
ε 2ς Ω
Γn = cos τ + u1 u6 − sin τ + u1 u4
ρ2
ωn ωn
ε Ω Ω
+ 2 cos τ + u1 u5 − sin τ + u1 u3
ρ ωn ωn
εq Ω Ω
+ 2 cos τ + u1 u5 − sin τ + u1 u3
ρ ωn ωn
2q P Ω Ω
+ 2 2 Y − cos 2 τ + u1 u3 − sin 2 τ + u1 u5
ωn ρ M ωn ωn
q Ω Ω Ω
+ 2 2 cos 2 τ + u1 u3u5 − sin 2 τ + u1 u32 + sin 2 τ + u1 u52
ρ ωn ωn ωn
2 2
ωt ω ω ω
u8′ = − (1 + CR ) 2ς t u8 − (1 + K R ) t u7 + t u1 + 2ς t t u2
ωn ωn ωn ωn
Ω Ω Ω
u4′ = −2ς u4 − 1 − q cos 2 τ + u1 u3 + q sin 2 τ + u1 u5 + ε u2′ sin τ + u1
ωn ωn ωn
ε Ω P q Ω
2 (
Ω + ωnu2 ) cos τ + u1 − Y 2 sin 2 τ + u1
2
+
ωn ωn M ωn ωn
73
Ω Ω Ω
YM′′ = −2ς YM′ − 1 + q cos 2 τ + ϕ YM + q sin 2 τ + ϕ X M − εϕ ′′ cos τ + ϕ
ω ω ω
n n n
ε Ω P q Ω
+ 2 ( Ω + ωnϕ ′ ) sin τ + ϕ + Y 2 cos 2 τ + ϕ − Φ 0
2
ωn ωn M ωn ωn
74
Appendix C Matlab Models
clear all;
close all;
options=odeset('RelTol',1e-6,'AbsTol',1e-6);
[t,x]=ode45(@lateral_fun,tspan,x0,options,omega,wn,wt,si,sit,ee,rho,Py,
q);
x0=x(end,:)'; %use results from last iteration for new
initial cond.
if omega==0.5*wn
xlswrite('result', t(:,1), [int2str(N) ' RPM'], 'A1');
xlswrite('result', x(:,:), [int2str(N) ' RPM'], 'B1');
elseif omega==wn
xlswrite('result', t(:,1), [int2str(N) ' RPM'], 'A1');
xlswrite('result', x(:,:), [int2str(N) ' RPM'], 'B1');
end
75
%Output Data for Amplitude Plot
y(n,1)=omega*60/(2*pi()); %Speed
(rpm)
y(n,2)=max(x(40:end,3))-min(x(40:end,3)); %Amplitude X (in)
y(n,3)=max(x(40:end,5))-min(x(40:end,5)); %Amplitude Y (in)
%Plot Response
figure(1);
hold on;
plot(y(:,1),y(:,2),'k.');
plot(y(:,1),y(:,3),'g.');
axis([0 6500 0 1.10*max(y(:,3))]);
title('Asymetric Shaft Lateral Vibration');
xlabel('Motor Speed(rpm)');
ylabel('Amplitude (in)');
grid;
end
function dx=lateral_fun(t,x,omega,wn,wt,si,sit,ee,rho,Py,q);
dx(1)=0;
dx(2)=0;
dx(3)=x(4);
dx(4)=-2*si*x(4)-(1-
q*cos(2*omega*t/wn))*x(3)+q*sin(2*omega*t/wn)*x(5)+(ee/wn^2)*omega^2*co
s(omega*t/wn)-Py*(q/wn^2)*sin(2*omega*t/wn);;
dx(5)=x(6);
dx(6)=-2*si*x(6)-
(1+q*cos(2*omega*t/wn))*x(5)+q*sin(2*omega*t/wn)*x(3)+(ee/wn^2)*omega^2
*sin(omega*t/wn)+Py*(q/wn^2)*cos(2*omega*t/wn);
76
C.2 Special Case: Torsional Vibration Matlab Model
clear all;
close all;
options=odeset('RelTol',1e-7,'AbsTol',1e-7);
[t,x]=ode45(@torsional_fun,tspan,x0,options,omega,wn,wt,si,sit,ee,rho,P
y,q);
x0=x(end,:)'; %use results from last iteration for new
initial cond.
%Output
y(n,1)=omega*60/(2*pi()); %Speed (rpm)
y(n,2)=(360/(2*pi()))*(max(x(end-500:end,1))-min(x(end-
500:end,1))); %Amplitude (deg pk-pk)
77
n=n+1; %increase index for organizing output matrix
%Plot Response
figure(1);
plot(y(:,1),y(:,2),'k.');
axis([0 5000 0 1.10*max(y(:,2))]);
title('Asymetric Shaft Torsional Vibration');
xlabel('Motor Speed(rpm)');
ylabel('Torsional Amplitude (deg pk-pk)');
grid;
end
function dx=torsional_fun(t,x,omega,wn,wt,si,sit,ee,rho,Py,q);
dx(1)=x(2);
dx(2)=-x(1)*(wt/wn)^2-(2*sit*wt/wn)*x(2)-
(ee*q/(wn*rho)^2)*Py*cos((omega/wn)*t+x(1))+(q/(wn^4*rho^2))*sin(2*((om
ega/wn)*t+x(1)))*Py^2;
dx=[dx(1) dx(2)]';
78
PY
C.3 Special Case: Torsional Vibration Matlab Model For Parametric Study Of ε
k
clear all;
close all;
options=odeset('RelTol',1e-7,'AbsTol',1e-7);
[t,x]=ode45(@torsional_fun,tspan,x0,options,omega,wn,wt,si,sit,ee,rho,P
y,q);
x0=x(end,:)'; %use results from last iteration for new
initial cond.
%Output
y(n,1)=omega*60/(2*pi()); %Speed (rpm)
y(n,2)=(360/(2*pi()))*(max(x(end-500:end,1))-min(x(end-
500:end,1))); %Amplitude (deg pk-pk)
79
y(n,3)=Pyke; %Ratio Py/M
n=n+1; %increase index for organizing output matrix
%Plot Response
figure(1);
plot3(y(:,1),y(:,3),y(:,2),'b.');
title('Asymetric Shaft Torsional Vibration');
xlabel('Motor Speed(rpm)');
zlabel('Torsional Amplitude (deg pk-pk)');
ylabel('Ratio (Py/k)/(ee)');
grid;
end
end
function dx=torsional_fun(t,x,omega,wn,wt,si,sit,ee,rho,Py,q);
dx(1)=x(2);
dx(2)=-x(1)*(wt/wn)^2-(2*sit*wt/wn)*x(2)-
(ee*q/(wn*rho)^2)*Py*cos((omega/wn)*t+x(1))+(q/(wn^4*rho^2))*sin(2*((om
ega/wn)*t+x(1)))*Py^2;
dx=[dx(1) dx(2)]';
80
C.4 General Vibration Model
clear all;
close all;
options=odeset('RelTol',1e-7,'AbsTol',1e-7);
[t,x]=ode45(@torsional_fun,tspan,x0,options,omega,wn,wt,si,sit,ee,rho,P
y,q);
x0=x(end,:)'; %use results from last iteration for new
initial cond.
%Output
y(n,1)=omega*60/(2*pi()); %Speed (rpm)
y(n,2)=(360/(2*pi()))*(max(x(end-500:end,1))-min(x(end-
500:end,1))); %Amplitude (deg pk-pk)
y(n,3)=Pyke; %Ratio Py/M
n=n+1; %increase index for organizing output matrix
%Plot Response
81
figure(1);
plot3(y(:,1),y(:,3),y(:,2),'b.');
title('Asymetric Shaft Torsional Vibration');
xlabel('Motor Speed(rpm)');
zlabel('Torsional Amplitude (deg pk-pk)');
ylabel('Ratio (Py/k)/(ee)');
grid;
end
end
function dx=general_fun(t,x,omega,wn,wt,si,sit,ee,rho,Py,q);
dx(1)=x(2);
dx(2)=-x(1)*(wt/wn)^2-(2*sit*wt/wn)*x(2)-
(ee*q/(wn*rho)^2)*Py*cos((omega/wn)*t+x(1))+(q/(wn^4*rho^2))*sin(2*((om
ega/wn)*t+x(1)))*Py^2+(ee*2*si/rho^2)*(x(6)*cos((omega/wn)*t+x(1))-
x(4)*sin((omega/wn)*t+x(1)))+(ee*(1+q)/rho^2)*(x(5)*cos((omega/wn)*t+x(
1))-x(3)*sin((omega/wn)*t+x(1)))-
(2*q*Py/(wn*rho)^2)*(x(3)*cos(2*((omega/wn)*t+x(1)))+x(5)*sin(2*((omega
/wn)*t+x(1))))+(q/rho^2)*(2*x(3)*x(5)*cos(2*((omega/wn)*t+x(1)))-
x(3)^2*sin(2*((omega/wn)*t+x(1)))+x(5)^2*sin(2*((omega/wn)*t+x(1))));
dx(3)=x(4);
dx(4)=-2*si*x(4)-(1-
q*cos(2*((omega/wn)*t+x(1))))*x(3)+q*sin(2*((omega/wn)*t+x(1)))*x(5)+ee
*dx(2)*sin((omega/wn)*t+x(1))+(ee/wn^2)*(omega+wn*x(2))^2*cos((omega/wn
)*t+x(1))-Py*(q/wn^2)*sin(2*((omega/wn)*t+x(1)));
dx(5)=x(6);
dx(6)=-2*si*x(6)-
(1+q*cos(2*((omega/wn)*t+x(1))))*x(5)+q*sin(2*((omega/wn)*t+x(1)))*x(3)
-
ee*dx(2)*cos((omega/wn)*t+x(1))+(ee/wn^2)*(omega+wn*x(2))^2*sin((omega/
wn)*t+x(1))+Py*(q/wn^2)*cos(2*((omega/wn)*t+x(1)));
82
C.5 Vibration Model for Increasing Shaft Asymmetry
clear all;
close all;
for q=[0:0.02:0.2];
parameters = {'wn (rad/s)', 'wt (rad/s)', 'si', 'sit', 'ee (in)', 'rho
(in)', 'Py (in/sec^2)', 'q'; wn wt si sit ee rho Py q};
xlswrite('result', parameters, 'Parameters', 'A1');
options=odeset('RelTol',1e-7,'AbsTol',1e-7);
[t,x]=ode45(@general_fun,tspan,x0,options,omega,wn,wt,si,sit,ee,rho,Py,
q);
x0=x(end,:)'; %use results from last iteration for new
initial cond.
83
y(n,1)=omega*60/(2*pi()); %Speed
(rpm)
y(n,2)=max(x(2*wn:end,3))-min(x(2*wn:end,3)); %Pk-Pk Amplitude X
(in)
y(n,3)=max(x(2*wn:end,5))-min(x(2*wn:end,5)); %Pk-Pk Amplitude Y
(in)
y(n,4)=(360/(2*pi()))*(max(x(2*wn:end,1))-min(x(2*wn:end,1)));
%Pk-Pk Amplitude torsional (deg)
y(n,5)=q;
%Plot Response
figure(3);
plot3(y(:,1),y(:,5),y(:,4),'b.');
title('Asymmetric Shaft Torsional Vibration');
xlabel('Motor Speed(rpm)');
zlabel('Torsional Amplitude (deg pk-pk)');
ylabel('Asymmetry Factor, q');
grid;
hold on;
%Plot Response
figure(4);
plot3(y(:,1),y(:,5),y(:,3),'b.');
title('Asymetric Shaft Lateral Vibration');
xlabel('Motor Speed(rpm)');
zlabel('Vertical Amplitude (in pk-pk)');
ylabel('Asymmetry Factor, q');
grid;
hold on;
84
xlswrite('result', y(:,:), ['Bode Plot q=' int2str(q)], 'A1');
end
end
85
C.6 Vibration Model for Varying Ratios of ωt ωn
clear all;
close all;
for wtwn=[0.25:0.25:2];
parameters = {'wn (rad/s)', 'wt (rad/s)', 'si', 'sit', 'ee (in)', 'rho
(in)', 'Py (in/sec^2)', 'q'; wn wt si sit ee rho Py q};
xlswrite('result', parameters, 'Parameters', 'A1');
options=odeset('RelTol',1e-7,'AbsTol',1e-7);
[t,x]=ode45(@general_fun,tspan,x0,options,omega,wn,wt,si,sit,ee,rho,Py,
q);
x0=x(end,:)'; %use results from last iteration for new
initial cond.
86
%Output Data for Amplitude Plot
y(n,1)=omega*60/(2*pi()); %Speed
(rpm)
y(n,2)=max(x(2*wn:end,3))-min(x(2*wn:end,3)); %Pk-Pk Amplitude X
(in)
y(n,3)=max(x(2*wn:end,5))-min(x(2*wn:end,5)); %Pk-Pk Amplitude Y
(in)
y(n,4)=(360/(2*pi()))*(max(x(2*wn:end,1))-min(x(2*wn:end,1)));
%Pk-Pk Amplitude torsional (deg)
y(n,5)=wtwn;
%Plot Response
figure(3);
plot3(y(:,1),y(:,5),y(:,4),'b.');
title('Asymmetric Shaft Torsional Vibration');
xlabel('Motor Speed(rpm)');
zlabel('Torsional Amplitude (deg pk-pk)');
ylabel('wt/wn');
grid;
hold on;
end
end
87
C.7 Vibration Model for Increasing Side Load
clear all;
close all;
q=0.2;
wnorig=1850*2*pi()/60; %original lateral natural frequency (rad/s)
wn=(1/(1+q))^0.5*wnorig; %lateral natural frequency after asymetry
(rad/s)
parameters = {'wn (rad/s)', 'wt (rad/s)', 'si', 'sit', 'ee (in)', 'rho
(in)', 'Py (in/sec^2)', 'q'; wn wt si sit ee rho Py q};
xlswrite('result', parameters, 'Parameters', 'A1');
options=odeset('RelTol',1e-7,'AbsTol',1e-7);
[t,x]=ode45(@general_fun,tspan,x0,options,omega,wn,wt,si,sit,ee,rho,Py,
q);
x0=x(end,:)'; %use results from last iteration for new
initial cond.
88
%Output Data for Amplitude Plot
y(n,1)=omega*60/(2*pi()); %Speed
(rpm)
y(n,2)=max(x(2*wn:end,3))-min(x(2*wn:end,3)); %Pk-Pk Amplitude X
(in)
y(n,3)=max(x(2*wn:end,5))-min(x(2*wn:end,5)); %Pk-Pk Amplitude Y
(in)
y(n,4)=(360/(2*pi()))*(max(x(2*wn:end,1))-min(x(2*wn:end,1)));
%Pk-Pk Amplitude torsional (deg)
y(n,5)=Pyke;
%Plot Response
figure(3);
plot3(y(:,1),y(:,5),y(:,4),'b.');
title('Asymmetric Shaft Torsional Vibration');
xlabel('Motor Speed(rpm)');
zlabel('Torsional Amplitude (deg pk-pk)');
ylabel('Ratio (Py/k)/(ee)');
grid;
hold on;
%Plot Response
figure(4);
plot3(y(:,1),y(:,5),y(:,3),'b.');
title('Asymetric Shaft Lateral Vibration');
xlabel('Motor Speed(rpm)');
zlabel('Vertical Amplitude (in pk-pk)');
ylabel('Ratio (Py/k)/(ee)');
grid;
hold on;
89
if omega==0.5*wn
xlswrite('result', y(:,:), ['Pyke' int2str(Pyke) ' '
int2str(N)], 'A1');
elseif omega==wn
xlswrite('result', y(:,:), ['Pyke' int2str(Pyke) ' '
int2str(N)], 'A1');
end
end
end
90
Appendix D Experiment Trial Notes
System consists of two 800g disks as an inboard rotor and two 800g disks as an outboard rotor.
Outboard rotor has asymmetry with 1.5" long flats ground on either side of rotor.
Outboard rotor has vertical loading system (pulleys) with masses hung to produce vertical side load.
Gear wheels at each end for torsional measurement (TK17 inboard / outboard)
91