DIAGNOSTICS OF SUBSYNCHRONOUS VIBRATIONS IN

ROTATING MACHINERY – METHODOLOGIES TO IDENTIFY

POTENTIAL INSTABILITY

A Thesis

by

RAHUL KAR

Submitted to the Office of Graduate Studies of

Texas A&M University
in partial fulfillment of the requirements for the degree of

MASTER OF SCIENCE

August 2005

Major Subject: Mechanical Engineering

 DIAGNOSTICS OF SUBSYNCHRONOUS VIBRATIONS IN

ROTATING MACHINERY – METHODOLOGIES TO IDENTIFY

POTENTIAL INSTABILITY

A Thesis

by

RAHUL KAR

Submitted to the Office of Graduate Studies of

Texas A&M University
in partial fulfillment of the requirements for the degree of

MASTER OF SCIENCE

Approved by:

Chair of Committee, John M.Vance

Committee Members, Alan Palazzolo
Goong Chen
Head of Department, Dennis O’Neal

August 2005

Major Subject: Mechanical Engineering

 iii

ABSTRACT

Diagnostics of Subsynchronous Vibrations in Rotating Machinery – Methodologies to

Identify Potential Instability. (August 2005)

Rahul Kar, B.E., National Institute of Technology, India

Chair of Advisory Committee: Dr. John M.Vance

Rotordynamic instability can be disastrous for the operation of high speed

turbomachinery in the industry. Most ‘instabilities’ are due to de-stabilizing cross

coupled forces from variable fluid dynamic pressure around a rotor component, acting in

the direction of the forward whirl and causing subsynchronous orbiting of the rotor.

However, all subsynchronous whirling is not unstable and methods to diagnose the

potentially unstable kind are critical to the health of the rotor-bearing system.

The objective of this thesis is to explore means of diagnosing whether

subsynchronous vibrations are benign or have the potential to become unstable. Several

methods will be detailed to draw lines of demarcation between the two. Considerable

focus of the research has been on subharmonic vibrations induced from non-linear

bearing stiffness and the study of vibration signals typical to such cases. An analytical

model of a short-rigid rotor with stiffness non-linearity is used for numerical simulations

and the results are verified with actual experiments.

Orbits filtered at the subsynchronous frequency are shown as a diagnostic tool to

indicate benign vibrations as well as ‘frequency tracking’ and agreement of the

 iv

frequency with known eigenvalues. Several test rigs are utilized to practically

demonstrate the above conclusions.

A remarkable finding has been the possibility of diagnosing instability using the

synchronous phase angle. The synchronous phase angle β is the angle by which the

unbalance vector leads the vibration vector. Experiments have proved that β changes

appreciably when there is a de-stabilizing cross coupled force acting on the rotor as

compared to when there is none. A special technique to calculate the change in β with

cross-coupling is outlined along with empirical results to exemplify the case.

Subsequently, a correlation between the synchronous phase angle and the phase angle

measured with most industrial balancing instruments is derived so that the actual

measurement of the true phase angle is not a necessity for diagnosis.

Requirements of advanced signal analysis techniques have led to the

development of an extremely powerful rotordynamic measurement teststand –

‘LVTRC’. The software was developed in tandem with this thesis project. It is a stand-

alone application that can be used for field measurements and analysis by

turbomachinery companies.
 v

DEDICATION

To my Parents and Dr.Vance

“It's a magical world, Hobbes, Ol' Buddy . . . let's go exploring!”

- Bill Watterson
 vi

ACKNOWLEDGMENTS

I am indebted to Dr.Vance for the opportunity to conduct research at the

Turbomachinery Laboratory. Without his deep insights and guidance this thesis would

not have been possible. Mr. Preston Johnson (National Instruments) has been an

inspiration behind the development of LVTRC – the ‘eyes’ and ‘ears’ of the research.

I am grateful to Dr. Palazzolo and Dr. Chen for consenting to oversee my thesis.

Special thanks are due to my friends and colleagues at the Turbomachinery

Laboratory, especially to Mohsin Jafri for his analytical inputs, Bugra Ertas for helping

me set up the test rigs and Kiran Toram for assisting with the experiments. Vivek

Choudhury and Ahmed Gamal added definite spice to what has been two stimulating

years at the laboratory.

The Turbomachinery Research Consortium has been kind enough to sponsor the

research for two consecutive years.

Last but not the least, my parents have been outstanding in their support in every

manner possible and this thesis is the outcome of their sacrifices.

 vii

TABLE OF CONTENTS

Page

ABSTRACT.................................................................................................................... iii

DEDICATION..................................................................................................................v

ACKNOWLEDGMENTS ...............................................................................................vi

TABLE OF CONTENTS................................................................................................vii

LIST OF FIGURES .........................................................................................................ix

CHAPTER

I INTRODUCTION ................................................................................................1

Objective .......................................................................................................3
Literature Review..........................................................................................4
Research Procedure .......................................................................................6

II THE ROTORDYNAMIC MEASUREMENT TESTSTAND – ‘LVTRC’ ...........9

The Teststand ................................................................................................9

Test Setup....................................................................................................11
Functional Blocks........................................................................................13

III SUBHARMONIC VIBRATION OF ROTORS IN BEARING

CLEARANCE ....................................................................................................18

Review of F.F.Ehrich’s 1966 ASME Paper ................................................18

Natural Vibration Waveform from Intermittent Contact ............................21

IV SHORT RIGID ROTOR MODEL TO SIMULATE INTERMITTENT

CONTACT OF ROTOR WITH SINGLE STATOR SURFACE ......................25

Short Rigid Rotor Model.............................................................................25

 viii

CHAPTER Page

Equations of Motion....................................................................................26
Non-Dimensional Equations of Motion ......................................................29
Euler Integration Scheme ............................................................................33
Numerical Simulation .................................................................................35

V EXPERIMENTAL DETERMINATION OF ROTOR BEHAVIOR

INDUCED FROM NON-LINEAR ROTOR-BEARING STIFFNESS ..............42

Experimental Setup – Rig Description and Instrumentation.......................42

Introducing Non-linear Bearing Stiffness in the Rig ..................................45
Experimental Results with Non-linear Bearing Stiffness ...........................47
Experimental Results – Linear System .......................................................52

VI DIAGNOSING UNSTABLE SUBSYNCHRONOUS VIBRATIONS

USING THE SYNCHRONOUS PHASE ANGLE...........................................54

Instability and Cross-coupled Stiffness.......................................................54

Synchronous Phase Angle (β) .....................................................................58
Relationship between the Synchronous and Instrument Phase Angle ........60
Experiments to Diagnose Subsynchronous Vibration.................................62
XLTRC Simulations to Study Phase Shifts from Cross-Coupled Forces ...64

VII DIAGNOSTIC INDICATORS OF BENIGN FREQUENCIES ......................69

Benign Subsynchronous Vibration and Their Indicators ............................69

Experiments with Whirl Orbit Shapes ........................................................71

VIII CONCLUSION ...............................................................................................75

REFERENCES ...............................................................................................................76

APPENDIX I ..................................................................................................................78

APPENDIX II .................................................................................................................81

VITA...............................................................................................................................87
 ix

LIST OF FIGURES

Page

Fig 1: An unstable eigenvalue ..........................................................................................2

Fig 2 : Subharmonic response at the first critical speed ....................................................7

Fig 3: The LVTRC teststand ...........................................................................................10

Fig 4: Three channels of input into the NI 4472 DAQ device ........................................11

Fig 5: Experimental setup illustrating wire connections from the probes to

individual pins of the DAQ boards .....................................................................12

Fig 6: LVTRC channel configuration screen .................................................................12

Fig 7: Time series and resampled-compensated data.....................................................13

Fig 8: Spectrum and orbit plots ......................................................................................14

Fig 9: Cursor driven orbit...............................................................................................15

Fig 10: Bode plot..............................................................................................................16

Fig 11: Waterfall plot obtained from a proximity probe..................................................16

Fig 12: Rotor at bearing center.........................................................................................19

Fig 13: Rotor in constant contact with the stator .............................................................19

Fig 14: Rotor in intermittent contact with the stator ........................................................20

Fig 15: Natural vibration waveform of the rotor from intermittent contact.....................21

Fig 16: Plot of the ratio of the 1st and 2nd harmonic amplitude vs. β .............................23

Fig 17: FFT of the vibration waveform............................................................................23

Fig 18: Analog computer simulation showing response to 2ω excitation for β=0.2........24

Fig 19: Rotordynamic model to analyze the effect of non-linear stiffness ......................25
 x

Page

Fig 20: Free body diagram of the short rigid rotor...........................................................26

Fig 21: Non-dimensional waveform at the critical speed ................................................36

Fig 22: Initial values and system settings. A large value of δ ensures a linear

system..................................................................................................................37

Fig 23: Synchronous response for no contact ..................................................................38

Fig 24: Initial and parametric constants to simulate small rotor-stator

clearance. δ = 0.2..................................................................................................38

Fig 25: Response with small clearance at different rotational speeds..............................39

Fig 26: Higher harmonics on the X –spectrum at a speed lower than the critical............39

Fig 27: Time waveform at a speed lower than the critical ...............................................40

Fig 28: Vibration signatures at twice the critical speed – time trace,

spectra and orbit ..................................................................................................41

Fig 29: Subsynchronous vibration disappears on increasing the speed ...........................41

Fig 30: Experimental rig ..................................................................................................42

Fig 31: Closer view of the rig...........................................................................................43

Fig 32: Swirl inducer sectional view................................................................................44

Fig 33: First forward mode shape ....................................................................................44

Fig 34: Non-rotating bearing support without and with stiffener ....................................45

Fig 35: Stiffness measurements at the bearing.................................................................45

Fig 36: Close-up view of the stiffener..............................................................................46

Fig 37: Non-linear stiffness of the bearing.......................................................................46

 xi

Page

Fig 38: X and Y probe signals at 1300 RPM - below the critical speed ..........................47

Fig 39: Probe signals at 3000 RPM – above the first critical...........................................47

Fig 40: Frequency demultiplication at twice the critical speed........................................48

Fig 41: Probe signals at 4600 RPM..................................................................................48

Fig 42: Spectrum at 1800 RPM indicate the presence of higher harmonics ....................49

Fig 43: Subsynchronous vibration at 4200 RPM at 0.5X ................................................50

Fig 44: Subsynchronous vibration disappears on increasing the speed ...........................51

Fig 45: Waterfall plot from the X-probe showing the onset of subsynchronous

vibration at 0.5X..................................................................................................51

Fig 46: Orbits before, at and after the onset of subsynchronous vibration

at twice the critical speed .....................................................................................52

Fig 47: Frequency spectrum for linear bearing support at different running speeds .......53

Fig 48: Plot of β against ω/ωn for various values of cross coupled

stiffness. Direct stiffness = 90000, direct damping coefficient = 10% ................56

Fig 49: Plot of dβ/dK against K at different speeds ........................................................57

Fig 50: Effect of cross coupled stiffness on stability ......................................................58

Fig 51: Synchronous phase angle β .................................................................................59

Fig 52: Measurement of the synchronous phase angle ....................................................60

Fig 53: Instrumentation for phase measurements ...........................................................61

Fig 54: Virtual instrument to calculate β..........................................................................62

Fig 55: Benign subsynchronous vibration at 0.5X...........................................................63

 xii

Page

Fig 56: Unstable subsynchronous vibration at 0.5X when the swirl inducer

is turned on ..........................................................................................................63

Fig 57: Geo plot of the rotor rig .......................................................................................64

Fig 58: Response curve for the rotor................................................................................65

Fig 59: Rotor model with cross coupled ‘bearing’ at station 17 ......................................65

Fig 60: Details of the cross coupled bearing at station 17 ...............................................65

Fig 61: Instrument phase plots from XLTRC simulations with and without

cross coupling.......................................................................................................66

Fig 62: Response of the rotor with higher damping.........................................................67

Fig 63: Destabilizing bearing ...........................................................................................68

Fig 64: Change of instrument phase angle with larger direct damping............................68

Fig 65: Cross coupled force and velocity vectors in circular and elliptical orbits ...........70

Fig 66: The Shell rotor rig................................................................................................72

Fig 67: Bently rotor kit.....................................................................................................72

Fig 68: Worn-out sleeve bearing......................................................................................73

Fig 69: Almost circular unstable orbit from the Shell rotor rig .......................................73

Fig 70: Benign subsynchronous orbit from the Bently rotor kit. .....................................74
 1

CHAPTER I

INTRODUCTION

A serious problem affecting the reliability of modern day high speed

turbomachinery is “rotordynamic instability” evidenced as subsynchronous whirl. The

cause of instability is never unbalance in a rotor bearing system, but de-stabilizing cross

coupled follower forces around the periphery of some rotor component. In mathematical

terms (Lyapunov), “instability” is when the motion tends to increase without limit

leading to destructive consequences. In most real cases, a “limit cycle” is reached,

because the system parameters (stiffness/damping) do not remain linear with increasing

amplitude. The rotor may then be operated at non-destructive amplitudes for years but

needs rigorous monitoring tools, since any minor perturbation can destabilize the system

and produce a rapid growth in amplitude. In the industry, large subsynchronous

amplitudes are not a common occurrence, but are more destructive and difficult to

remedy than imbalance problems when they do occur. Quite often, they are load or

speed dependant, and build up to catastrophic levels under certain conditions.

Delineation based on definitions of critical speed and “instability” proves

instructive. Vance [1] defines critical speed as the speed at which the response to

unbalance is maximum. Instability is self-excited and is not dependent on the unbalance.

This thesis follows the style and format of Journal of Engineering for Gas Turbines and Power.
 2

It is possible to pass through the critical speed of the machine without destruction,

whereas a “threshold speed of instability” often cannot be exceeded without large-scale

damage.

Mathematically, dynamic instability is the solution to homogenous linear

differential equations of motion characterized by a complex eigenvalue with a positive

real part. The real part is responsible for the exponential growth (or decay, if negative) of

the solution while the imaginary part gives the damped frequency. From a rotordynamic

viewpoint, the solution is the function which determines the time dependant amplitude of

motion. A growing harmonic waveform is thus representative of instability as shown in

Fig 1.

Fig 1: An unstable eigenvalue [1]

 3

The factor in the differential equations that causes the vibration to grow is almost

always cross-coupled stiffness, which models a force (usually from fluid pressure)

driving the whirl orbit.

Lateral vibration signatures (periodic waveforms) captured with adequate

transducers from a rotating machine represents the dynamic whirling motion of the rotor.

Several signal analysis techniques can be used to extract pertinent information from

these signatures – the Fast Fourier Transform (FFT) to obtain all the frequencies present

in the signal, Bode plots for the synchronous rotor response, ‘orbits’ or Lissajous figures

at filtered frequencies, accurate phase and magnitude values for all ‘orders’ etc. Analysis

using these methods will make it possible to predict with a degree of certainty whether a

rotor has the possibility of going unstable. The current research is to explore such

methods and evaluate them with actual experiments at the Turbomachinery laboratory.

It is worth mentioning that existing rotordynamic measurement and signal

analysis tools are not sufficient to extract significant information from subsynchronous

signatures to aid in diagnostics. An advanced teststand with special algorithms is

necessary for purposeful investigation.

Objective

The presence of benign subsynchronous vibrations from a rotor can easily be

mistaken as instability, forcing the machine to be shut down out of fear of destruction. In

a 1977 ASME paper [2], Ferrara at Nuovo Pignone examined whether subsynchronous

vibration is a true indicator of incipient instability in high pressure multi-stage

compressors. He studied two high-pressure compressors and concluded that some

 4

asynchronous vibrations were the result of some “common cause” and was not self-

excited. He recommended that the conventional criteria for determining possible

instabilities should be re-evaluated.

The objective of this research is to study in detail, vibration signatures from

benign and potentially unstable subsynchronous vibrations and find ways of

differentiating between the two.

Literature Review

The occurrence of subsynchronous vibrations in rotating machines has been

widely reported and documented. The first instances began to show up in the early

1920’s when the need arose to operate rotors beyond the first critical speed. The

following section is a brief review of research that has been done in this field.

The first published experimental observation of whirl induced instability was by

Newkirk [3], in which he investigated nonsynchronous whirl in blast furnace

compressors (GE) operating at supercritical speeds. He came to the conclusion that shaft

whipping was because of internal friction from relative motion at the joint interfaces and

not unbalance excitation. Kimball, working with Newkirk, built a test rig that

demonstrated subsynchronous whirl due to internal friction in the rotating assembly.

Newkirk also observed fluid bearing whip caused by the unequal pressure distribution

about the journal. The enclosed fluid in the fluid film bearing clearance circulates with

an average velocity equal to one-half the shaft speed [4]. Alford [5] hypothesized that tip

clearances in axial flow turbines or compressors may induce non-synchronous whirling

of the rotor. Den Hartog [6] described dry friction whip as an instability due to a
 5

tangential Coulomb friction force acting opposite to the direction of shaft rotation

(backward whirl). The tangential force will be proportional to the radial contact force

between the journal and the bearing. Ehrich [7] has done a comprehensive survey of the

possible causes of self-excited instabilities and suggested cures in most cases.

Ehrich [8] reported the occurrence of a large subsynchronous response at twice the

critical speed from the analog computer simulation of a planar rotor model undergoing

intermittent contact with the bearing surface. His 1966 paper has been discussed at

greater length later in this thesis, since it has initiated the premise of research on

subsynchronous rotor response due to bearing non-linearity. A similar phenomenon of

“frequency demultiplication” is also presented by Den Hartog [6]. Bently [9] proposed

and experimentally showed that subsynchronous rotor motion can originate from

asymmetric clearances – “normal-tight’ conditions when the rotor stiffness changes

periodically on contact with a stationary surface or in “normal-loose” conditions when

the radial stiffness changes (hydrodynamic bearings) during synchronous orbiting.

Childs [10] solved non-linear differential equations describing a Jeffcott rotor to prove

that ½ and 1/3 speed whirling occurs in rotors which are subject to periodic normal-loose

or normal-tight radial stiffness variations. Vance [1] stated that rotordynamic instability

is seldom from one particular cause and reviewed the mathematics for analyzing the

stability of rotor-bearing systems through eigenvalue analysis. He also analyzed the case

of supersynchronous instability or the ‘gravity critical’. Wachel [11] presented several

field problems with centrifugal compressors and steam turbines involving instability

along with ‘fixes’ used to overcome the problems. A steam turbine was shown to have a
 6

subsynchronous instability at 1800 rpm when the unit speed was 4800 rpm. The

logarithmic decrement was found to be 0.04. The bearing length was reduced and the

clearance increased so that the log-dec increased to 0.2. The subsynchronous instability

disappeared. Another steam turbine case showed that changing to tilt-pad bearings was

not sufficient to overcome one-half speed instability. The most difficult problem was

instability in a gas re-injection compressor which suffered high vibration trip outs even

before reaching operating speed. The subsynchronous frequency became higher than the

first critical speed after modifying the oil seals. It was found that different seal designs

greatly affected the non-synchronous frequency but did not make the unit stable. The

stability improved greatly when a damper bearing was installed in series with the

inboard bearing. The problem was mitigated by increasing the shaft diameter to raise the

first critical speed and hence the onset speed of instability.

Research Procedure

A robust rotordynamic measurement teststand is required to signal-analyze

subsynchronous vibrations. Currently available applications like ADRE from Bently

Nevada are limited in their applications and will not be useful in the proposed research.

For example, ADRE can only track subsynchronous frequencies in increments of 0.025

orders, which is extremely inflexible for extraction of data for a vibration which occurs

at say, 0.470 orders. Consequently it was found necessary to develop a new

rotordynamic measurement software solution (LVTRC) which exceed the capabilities of

ADRE and help in diagnostics. LabVIEW from National Instruments was chosen as the

software platform with which to develop the software, for its excellent graphical and
 7

data acquisition capabilities. Two add-ons - the Order Analysis Toolset 2.0 and the

Sound and Vibration Toolset, helped with advanced order extraction techniques and

signal processing. Two 8-channel NI-4472 PCI boards were used for data acquisition

and A/D conversion. Once the measurement teststand was in place, several test rigs were

used to study known sources of benign and potentially unstable subsynchronous

vibrations.

The exploration of the rotor response to non-linear bearing supports was of

special interest. The method of investigation was based upon a paper by F.F Ehrich [8]

claiming through analog computer simulations, that such a rotor would have a large

subharmonic response at its first critical frequency when rotating at twice its critical

speed (Fig 2 : Subharmonic response at the first critical speed [8].

Fig 2 : Subharmonic response at the first critical speed [8]

 8

The phenomenon was indeed found to occur in actual experiments and explained the

often noted fact that “the onset of most asynchronous whirl phenomenon is at twice the

induced whirl speed” [8].

A numerical rotordynamic model of a short rigid rotor undergoing intermittent

contact with the bearing housing was developed so that its stiffness varied as a step

function of the displacement along the direction of contact. The simulation results were

compared with empirical data from a rig where non-linearity in bearing stiffness was

artificially introduced.

Vibration signatures typical to instabilities were also studied at length especially

from a rig where a forward acting de-stabilizing air swirl around the rotor could be

turned on at will (a large subsynchronous vibration was induced at the first eigenvalue

above the first critical speed).

It was also discovered that the synchronous phase angle (the angle by which the

unbalance vector leads the vibration vector) was affected by destabilizing cross coupled

forces. The change in phase angle from cross-coupling can be a remarkable tool for the

diagnostics of subsynchronous vibrations in the industry

 9

CHAPTER II

THE ROTORDYNAMIC MEASUREMENT TESTSTAND – ‘LVTRC’

The Teststand

There are a number of differences between rotordynamics and structural

vibrations that make measurement and diagnostics more difficult for rotordynamics.

Examples are the occurrence of subsynchronous vibrations at fractions of the

synchronous speed, the use of Lissajous patterns (orbits) with tachometer marks, the

necessity of ‘runout’ subtraction from shaft vibration measurements and extracting

accurate magnitude and phase information for different ‘orders’ of the running speed.

The difficulties are also compounded by inflexible and limited data analysis systems

currently available like the Bently Nevada ADRE™. Consequently, a new system has

been developed using LabVIEW from National Instruments along with associated state-

of-the-art data acquisition devices, to overcome most shortcomings and to enable

advanced subsynchronous vibration study. A description of the teststand follows.

The teststand (Fig 3) is an Intel Pentium 4, 1.7 GHz personal computer with

Windows®NT operating system, 512 MB RAM and 40 GB of hard drive space. The

system is equipped with two NI 4472 PCI 8-channel boards for data acquisition and

analysis. LabVIEW 7.0 Express with the Order Analysis 2.0 and Sound and Vibration

Toolset is installed for programming and signal processing.

 10

Fig 3: The LVTRC teststand

Proximity probes from Bently Nevada, powered by a -24V power supply are used for all

measurements. The probes are connected to the data acquisition boards via NI

attenuation cables to increase the acceptable voltage range from the input and run the

application DC coupled.
 11

Parallel Port
Security
Key

NI 4472
DAQ boards

Master

Slave

Fig 4: Three channels of input into the NI 4472 DAQ device

Test Setup

All measurements for experiments mentioned in this thesis were carried out using

eddy current proximity probes. The probes were calibrated and set at a voltage gap of

negative 8.0 V DC to ensure linear behavior. The ‘leads’ from the proximity probes were

connected using BNC cables to the DAQ boards as shown in Fig 4. A schematic of the

connection basics is shown in Fig 5.

 12

7 6 5 4 3 2 1 0
Probe Y
DAQ Board: Master

7 6 5 4 3 2 1 0

DAQ Board: Slave

Probe X

Tachometer

Fig 5: Experimental setup illustrating wire connections from the probes to individual pins of the

DAQ boards

The channels have to be configured using the calibration units, probe orientation and

whether to operate in AC or DC coupled mode.

Fig 6: LVTRC channel configuration screen

 13

Functional Blocks

The LVTRC application incorporates most functional blocks, which a standard

rotordynamic data acquisition system has. The capabilities are briefly discussed below.

Time Series Data

Uncompensated

Compensated

Fig 7: Time series and resampled-compensated data

Fig 7 shows the time series data captured from two proximity probes and a tachometer.

The first graph is the raw voltage signal converted to engineering units (in this case mils)

while the second plot is resampled data, compensated and with the DC component

subtracted. Resampling changes the data from the time domain to the angular domain.
 14

Spectrum and Orbit

Malfunctions in rotating machines (loose ball bearings, shaft misalignment, rubs,

instabilities etc) leave tell-tale stamps on the spectrum and orbit plots, in terms of the

fraction of the running speeds they occur, the tendency to track the synchronous

component, the ellipticity of the orbit (Fig 8) etc. LVTRC has advanced tracking filters

which can extract accurate phase and magnitude information for any ‘order’ of the

running speed.

Fig 8: Spectrum and orbit plots

One of the most useful features of the application is the ability to plot orbits exactly at

frequencies where the user places his/her cursor on the Spectrum plot (Fig 9).
 15

Fig 9: Cursor driven orbit

The red cursor is placed at the 40 Hz component on the frequency spectrum. The orbit at

that frequency is plotted. This feature will be used in the study of how orbits filtered at a

particular subsynchronous frequency behave when the rotor is potentially unstable.

Bode Plots

Bode plots (Fig 10) illustrate the rotor response (amplitude and phase) to changing rotor

speed. LVTRC maintains buffered Bode plot data from the raw signals; so that any

channel and any order can be selected at any point of the experiment and the full history

of the rotor response may be obtained.

 16

Fig 10: Bode plot

Waterfall Plot

Fig 11: Waterfall plot obtained from a proximity probe

 17

Several of the features like the waterfall plot (Fig 11) have been used extensively to

obtain very encouraging results. The detailed development of the code is not discussed

here as it is beyond the scope of this thesis.

 18

CHAPTER III

SUBHARMONIC VIBRATION OF ROTORS IN BEARING

CLEARANCE

Review of F.F.Ehrich’s 1966 ASME Paper [8]

Ehrich’s paper describes subsynchronous whirl generated due to non-linearity in

rotor stiffness and uses analog computer simulation to show that “the excitation at the

second harmonic induces a large component of first-harmonic response. It is suggested

that this subharmonic resonance or frequency demultiplication may play a large role in

the often noted fact that the onset of most asynchronous whirl phenomenon is at twice

the induced whirl speed”. The interest of this thesis originates from the idea that any

non-linearity in stiffness of the rotor-bearing system (the source is immaterial) will

produce a large subsynchronous response when running at twice the critical speed and

can be easily mistaken for an instability.

Clearances between the rotor and stator in a high-speed rotating system lead to

non-linear spring stiffness. The static equilibrium position of the rotor may not be at the

bearing center (Fig 12) but may be eccentric and be in contact with one side of the

bearing (Fig 13). Another and perhaps more realistic case is the intermittent contact

between the journal and the bearing due to rotor vibration (Fig 14).
 19

K2

δ
F

K1

K2 Slope = K1

Fig 12: Rotor at bearing center

K2

F
K1

Slope = K1 +K2

K2

Fig 13: Rotor in constant contact with the stator

 20

K2

K1

Slope = K1 +K2
K2 Slope = K1
x
δ

Fig 14: Rotor in intermittent contact with the stator

In Fig 14 the displacement of the stator is exaggerated as K2>>K1.

Case A: Rotor at bearing center

Stiffness of the system = K1

K1
Undamped natural frequency = ω1 =
m

Case B: Rotor in constant contact with the stator

Equivalent stiffness of the system = K1 + K2

K1 + K 2
Undamped natural frequency = ω2 =
m

Case C: Intermittent contact with the stator

The rotor “bounces” on the stator periodically leading to a piecewise linear stiffness of

the system.

Equivalent stiffness of the system = Keq = K1 for x < δ

= K1 + K2 for x ≥ δ
 21

Natural Vibration Waveform from Intermittent Contact

Assuming that the rotor contacts just one side of the stator and the resultant

vibration is not enough for it to contact the other side, the vibration waveform should be

similar to Fig 15.

A
ω1
π
Amplitu

ω2

A
ω2 Time
π
ω1

Fig 15: Natural vibration waveform of the rotor from intermittent contact

π
x (t ) = A1 sin ω1t for 0≤t<
ω1
(a)
⎡ π ⎤ π 2π
x (t ) = − A2 sin ⎢ω2 {t − }⎥ for ≤t<
⎣ ω1 ⎦ ω1 ω

For the waveform to be differentiable at t = π/ω1, x (t) has to be continuous and

dx = dx π −
dt t→ π + dt t→
ω ω
1 1

or, ω1A1 = ω2A2

A1=A/ω1 and A2=A/ω2

 22

The equivalent frequency is computed from the waveform:

2π π π
= + ;
ω ω1 ω2
(b)
2ω1ω2
ω=
ω1 + ω2

The Fourier coefficient Cn from the Fourier series expansion of equation (a) is given as:

⎡ ⎛ ω ⎞2 ⎛ω ⎞
2
⎤
⎢ ⎜ ⎟
ω2 ⎠ ⎜ ⎟ ⎥
Cn =
A
⎢ ⎝ − ⎝ ω1 ⎠ ⎛ nπω ⎞
⎥ × 2 cos ⎜ ⎟
πω ⎢ 2 ⎛ω
2
⎞ 1 − n2 ⎛ ω ⎞ ⎥
2
⎝ 2ω2 ⎠
⎢⎣ 1 − n ⎜ ω ⎟ ⎜ ω ⎟ ⎥
⎝ 2⎠ ⎝ 1⎠ ⎦

⎛ nπω ⎞
with a phase angle φn = ⎜ ⎟
⎝ 2ω2 ⎠

The plot of the ratio of the second harmonic to the first harmonic amplitude indicates the

presence of a large second harmonic component (Fig 16) in the excitation waveform for

β = 0.2 (where β = K1/K2). A similar result is also obtained by the FFT of the waveform

(Fig 15: Natural vibration waveform of the rotor from intermittent contact). It yields that

the waveform has several higher harmonic components with only the second harmonic

being comparable to the first in magnitude (Fig 17).

 23

0 0.2 0.4 0.6 0.8 1

β = K1/K2

Fig 16: Plot of the ratio of the 1st and 2nd harmonic amplitude vs. β

0
0 2 4 6 8 10

Fig 17: FFT of the vibration waveform

If the response of the non-linear system excited at 2ω (refer equation(b)) produces a

subsynchronous vibration at a frequency of ω, then it can be concluded that it is

vibration induced out of the inherent non-linearity in the system stiffness. The analog

computer simulation results from the paper show likewise.

 24

Fig 18: Analog computer simulation showing response to 2ω excitation for β=0.2

The analysis is now extended from considering a plane vibration case to a ‘Short Rigid

Rotor’ model in the following chapter. Numerical methods will replace analog

simulations.
 25

CHAPTER IV

SHORT RIGID ROTOR MODEL TO SIMULATE INTERMITTENT

CONTACT OF ROTOR WITH SINGLE STATOR SURFACE

Short Rigid Rotor Model

Stator
Rotor
Rotor Stiffness
Stiffness
Stiffness M
u
K1 α
C K2
Y

O
∆
X
Im

Stator

K2

Fig 19: Rotordynamic model to analyze the effect of non-linear stiffness

 26

The model illustrated in Fig 19 is used to simulate non-linear rotor-bearing

stiffness due to intermittent contact of the rotor along the X-axis only. For simplicity, the

bearing stiffness is assumed to be symmetric. The bearing stiffness is much greater than

the rotor stiffness so that there is practically no displacement of the surface during

contact. The frame of reference is fixed and the generalized coordinates are X, Y and α.

Equations of Motion

Direction of rotation Y Positive moments

Y
m
m +

u u sin α
C(x,y) Sx+Dx
α C u cos α

r Sy+Dy

X X
O

Fig 20: Free body diagram of the short rigid rotor

Generalized coordinate system: X, Y, α (Fig 20)

 27

X m = X + u cos α

Ym = Y + u sin α

Differentiating with respect to time:

X m = X − (u sin α )α

Ym = Y + (u cos α )α

Xm = X − {(u sin α )α + α 2 (u cos α )}

Ym = Y + {(u cos α )α − α 2 (u sin α )}

From Newton’s Laws of Motion:

∑F x = mX

∑F y = mY

∑M = I mα

where,

Fx = restorative stiffness force (Sx) and ‘damping force’ (Dx) along the x-axis;

Fy = restorative stiffness force (Sy) and ‘damping force’ (Dy) along the y-axis;

M = moments taken about m.

 28

mXm = − K x X − CX

mXm = m[ X − {(u sin α )α + α 2 (u cos α )}] = − K x X − CX

mX + K x X + CX = muα sin α + mα 2u cos α

Similarly,

mY + K yY + CY = − muα cos α + mα 2u sin α

Taking moments about m:-

I mα = K yY (u cos α ) − K x X (u sin α ) − CX (u sin α ) + CY (u cos α )

I mα + K x X (u sin α ) + CX (u sin α ) − K yY (u cos α ) − CY (u cos α ) = 0

The equations of motion are:

mX + K x X + CX = muα sin α + mα 2u cos α

mY + K yY + CY = −muα cos α + mα 2u sin α (1)

I mα + K x X (u sin α ) + CX (u sin α ) − K yY (u cos α ) − CY (u cos α ) = 0

For the particular case under consideration where the static stiffness of the rotor-bearing

system changes as a step function of the rotor clearance (∆) and hence the displacement

along X, the system of equations (1) can be divided into two cases.
 29

Case 1: X< ∆

Kx = Ky= K1

mX + K1 X + CX = muα sin α + mα 2u cos α

mY + K1Y + CY = −muα cos α + mα 2u sin α (2)

I mα + K1 X (u sin α ) + CX (u sin α ) − K1Y (u cos α ) − CY (u cos α ) = 0

Case 2: X ≥ ∆

Kx = K1+ K2, Ky = K1**

mX + ( K1 + K 2 ) X + CX = muα sin α + mα 2u cos α

mY + K1Y + CY = −muα cos α + mα 2u sin α (3)

I mα + ( K1 + K 2 ) X (u sin α ) + CX (u sin α ) − K1Y (u cos α ) − CY (u cos α ) = 0

**Note that the stiffness changes only along the X-axis when contact occurs.
 30

Non-Dimensional Equations of Motion

Case 1: X< ∆

mX + K1 X + CX = muα sin α + mα 2u cos α

(4)
mY + K1Y + CY = −muα cos α + mα 2u sin α

K C
X + 1 X + X = uα sin α + α 2u cos α
m m

X K1
Let, x = and ω1 =
u m

x + ω12 x + 2ξω1 x = α sin α + α 2 cos α



Let τ = ω1t;

dx dx dτ dτ
= × ; = ω1
dt dτ dt dt

dx dx dx
= ω1 ; x = ω1
dt dτ dτ

d 2x
x = ω12

dτ 2

Similarly,

dα d 2α
α = ω1 ; α = ω12 2
dτ dτ

Replacing in (4)
 31

x + 2ξ x + x = α 2 cos α + α sin α



y + 2ξ y + y = α 2 sin α − α cos α



Moment equation:

I mα + K1 X (u sin α ) + CX (u sin α ) − K1Y (u cos α ) − CY (u cos α ) = 0 (5)

Let , I m = mG 2 , where G = radius of gyration

G
Let , g =
u

∴ I m = mu 2 g 2

Replacing in equation(5),

d 2α dx dy
g 2ω12 + ω12 x sin α − ω12 y cos α + 2ξω12 sin α − 2ξω12 cos α = 0
dτ 2
dτ dτ

x y 2ξ 2ξ
or , α + 2
sin α − 2 cos α + 2 x sin α − 2 y cos α = 0
g g g g

Therefore, the non-dimensional set of equations for Case 1 is:

x + 2ξ x + x = α 2 cos α + α sin α



y + 2ξ y + y = α 2 sin α − α cos α

 (6)

x y 2ξ 2ξ
α + 2
sin α − 2 cos α + 2 x sin α − 2 y cos α = 0
g g g g
 32

Case 2: X ≥ ∆

mX + ( K1 + K 2 ) X + CX = muα sin α + mα 2u cos α

(7)
mY + K1Y + CY = − muα cos α + mα u sin α
2

(K + K2 ) C
X + 1 X + X = uα sin α + α 2u cos α
m m

X K1 K
Let, x = , ω1 = ,β = 1
u m K2

( K1 + K 2 ) K1 ( K1 + K 2 ) 1
ω2 = = × = ω1 1 +
m m K1 β

x + ω2 2 x + 2ξω1 x = α sin α + α 2 cos α

or , 

⎛ 1⎞
x + ω12 ⎜1 + ⎟ x + 2ξω1 x = α sin α + α 2 cos α
or , 
⎝ β⎠

Let τ = ω1t;

Non dimensional forms of equation (7) are:

⎛ 1⎞
x + 2ξ x + x ⎜1 + ⎟ = α 2 cos α + α sin α

⎝ β⎠

y + 2ξ y + y = α 2 sin α − α cos α


 33

Moment equation:

I mα + ( K1 + K 2 ) X (u sin α ) + CX (u sin α ) − K1Y (u cos α ) − CY (u cos α ) = 0

(8)
⎛ 1 ⎞  
⇒ I mα + K1 ⎜1 + ⎟ X (u sin α ) + CX (u sin α ) − K1Y (u cos α ) − CY (u cos α ) = 0
⎝ β ⎠

Non-dimensional form of equation (8) is:

x ⎛ 1⎞ y 2ξ 2ξ
α + 2 ⎜
1 + ⎟ sin α − 2 cos α + 2 x sin α − 2 y cos α = 0
g ⎝ β⎠ g g g

Therefore, the non-dimensional set of equations for Case 2 is:

⎛ 1⎞
x + 2ξ x + x ⎜1 + ⎟ = α 2 cos α + α sin α

⎝ β⎠

y + 2ξ y + y = α 2 sin α − α cos α

 (9)

x ⎛ 1⎞ y 2ξ 2ξ
α + 2 ⎜
1 + ⎟ sin α − 2 cos α + 2 x sin α − 2 y cos α = 0
g ⎝ β⎠ g g g

Euler Integration Scheme

Equations (6) and (9) are integrated using Euler’s method to obtain the response

of the system modeled for unbalance excitation. The following substitutions are made:

x = u1; x = u2 ; y = u3 ; y = u4 ;α = u5 ;α = u6
 34

∆
Case 1: For x < δ (where δ = )
u

u2 = −2ξ u2 − u1 + u6 2 cos u5 + u6 sin u5

(10)
u4 = −2ξ u4 − u3 + u6 sin u5 − u6 cos u5
2

u1 u 2ξ 2ξ
u6 = − 2
sin u5 + 32 cos u5 − 2 u2 sin u5 + 2 u4 cos u5
g g g g

Case 2: For x ≥ δ

⎛ 1 ⎞
u2 = −2ξ u2 − u1 ⎜1 + ⎟ + u6 cos u5 + u6 sin u5
2

⎝ β ⎠

(11)
u4 = −2ξ u4 − u3 + u6 2 sin u5 − u6 cos u5

u1 ⎛ 1 ⎞ u3 2ξ 2ξ
u6 = − 2 ⎜
1+ ⎟ sin u5 + 2 cos u5 − 2 u2 sin u5 + 2 u4 cos u5
g ⎝ β ⎠ g g g

Time - marching scheme

u1 (t + δ t ) = u1 (t ) + u2 (t )δ t

u2 (t + δ t ) = u2 (t ) + u2 (t )δ t

u3 (t + δ t ) = u3 (t ) + u4 (t )δ t

u4 (t + δ t ) = u4 (t ) + u4 (t )δ t

u5 (t + δ t ) = u5 (t ) + u6 (t )δ t

u6 (t + δ t ) = u6 (t ) + u6 (t )δ t
 35

Numerical Simulation

Numerical integration of equations (10) and (11) with different initial conditions

is used to simulate an experimental rotor rig with non-linear bearing stiffness. The

system parameters are set to closely represent the actual rotor. The numerical model is

excited by unbalance. Details of the experimental set up have been provided in the next

chapter.

The stiffness ratio K1/K2 = β for the rotor bearing system is 0.23. It is necessary

to find out the non-dimensional critical frequency of the waveform described by Ehrich.

2ω1ω2
ω=
ω1 + ω2

ω2 K1 + K 2 1
= = 1+
ω1 K1 β

1
2ω1 1 +
β
ω=
1
1+ 1+
β
 36

The non-dimensional response waveform for the rotor rotating at the critical speed can

be represented as:

2πω1/ω

X/u

τ =ω1t

Fig 21: Non-dimensional waveform at the critical speed

The frequency of the waveform (Fig 21) is:

1 ω
f = = , where ω is the running speed.
T 2πω1

For a linear system (large clearance) ω = ω1,

1
f = = 0.16 *
2π

The first non-dimensional critical frequency is given as:

1 1
2ω1 1 + 1+
β β
f = = (12)
1 1
2πω1 (1 + 1 + ) π (1 + 1 + )
β β

For the experimental rotor, β = 0.23,

f = 0.222

*This value provides a check to see if the numerical integration has been correctly executed.
 37

Simulation of large rotor-stator clearance with no contact - model verification

A likely proof of the verity of the numerical model will be to obtain the response

of the system to a large clearance (no contact) for different rotational speeds. The

maximum response to unbalance should occur at f = 1 = 0.16 . The initial conditions of

2π

the rotor are shown below (Fig 22). Note that ‘alphadot’ is the rotating speed.

Fig 22: Initial values and system settings. A large value of δ ensures a linear system.

Integration with increasing values of ‘alphadot (0)’ or the rotating speed shows a large

synchronous response at f = 0.16 (Fig 23) as predicted by theory.

 38

Synchronous
Response

Fig 23: Synchronous response for no contact

Simulation of small clearance with contact

The clearance is reduced to δ = 0.2 (Fig 24) and simulations re-run.

Fig 24: Initial and parametric constants to simulate small rotor-stator clearance. δ = 0.2
 39

Fig 25: Response with small clearance at different rotational speeds

With a non-linear stiffness along the x-axis, the non-dimensional critical frequency is

0.225 (Fig 25). At a speed lower than the critical frequency, the spectrum of X shows

higher harmonic components (Fig 26). Y has a clean spectrum.

Synchronous Speed

Higher harmonic
Components

Fig 26: Higher harmonics on the X –spectrum at a speed lower than the critical
 40

The time waveform (Fig 27) also illustrates that collision occurs along the X – axis

clipping the wave.

Fig 27: Time waveform at a speed lower than the critical

“Frequency demultiplication” – Large subsynchronous response at twice the critical

speed

When the rotor speed is twice the critical speed, a large subsynchronous

component (Fig 28) on the X-spectrum at exactly the critical frequency is noted. The Y-

response is largely synchronous. The orbit shows the proverbial inside loop.
 41

Synchronous

Fig 28: Vibration signatures at twice the critical speed – time trace, spectra and orbit

On increasing the speed, the subsynchronous vibration disappears indicating that the

rotor-bearing system is not unstable (Fig 29).

Synch Synch

Fig 29: Subsynchronous vibration disappears on increasing the speed

 42

CHAPTER V

EXPERIMENTAL DETERMINATION OF ROTOR BEHAVIOR

INDUCED FROM NON-LINEAR ROTOR-BEARING STIFFNESS

Numerical simulations from the rotordynamic model in the previous chapter have

yielded typical response to non-linear stiffness which might be used as diagnostic tools

in rotating machinery monitoring. Of particular interest is the peculiar subsynchronous

vibration phenomenon which occurs at twice the critical speed. This chapter shows the

occurrence of “frequency demultiplication” experimentally. A rotor rig was set up to

produce non-linear bearing support stiffness and the signals generated from orthogonally

mounted proximity probes were studied in detail.

Experimental Setup – Rig Description and Instrumentation

Air turbine

Swirl Inducer

100 psi Prox Probes

Line

Proximeter Non-rotating
Power 250 psi
Line bearing
supply support

Fig 30: Experimental rig

 43

Ram Air
Turbine

Prox Probe
Bracket Ball
Bearing

Balance Hollow
Rotor Nozzle
Plane
Steel Rod
Support Swirl Inducer
(non - Housing
rotating) Ball
Bearing

Fig 31: Closer view of the rig

The rotor (Fig 30 and Fig 31) is mounted on ball bearings constrained at the inner

race by a non-rotating cantilevered steel support rod. Pressurized air drives the air-

turbine upto a maximum speed of 6000 rpm. The swirl inducer housing has nozzles

arranged around the periphery of the rotor to induce air swirl when pressurized. The air

swirl is in the direction of rotor rotation and generates whirl instability from

destabilizing cross coupled stiffness above the first critical speed (2100 rpm). A

sectional view of the swirl inducer chamber is shown in Fig 32.

Two calibrated orthogonally mounted eddy current proximity probes were used

to capture vibration data from the rotor. Another proximity probe was used to read

tachometer pulses from a raised notch on the rotor surface. LVTRC was used for all data

acquisition and analysis. The rotor was modeled using XLTRC2 (Appendix II) after the

stiffness of the bearings at the inducer end and the air turbine end were measured in
 44

horizontal and vertical directions. The values were used as bearing inputs and minimal

damping was assumed. The first mode shape (Fig 33) of the rotor showed that the

balance plane should be as close to the inducer end bearing as possible. Fig 31 shows the

balance plane. The single plane orbit method was used to balance the rotor to acceptable

limits of vibration.

Fig 32: Swirl inducer sectional view

Damped Eigenvalue Mode Shape Plot

forward
backward
f=2328.3 cpm
d=.1064 zeta
N=3000 rpm

Fig 33: First forward mode shape

 45

Introducing Non-linear Bearing Stiffness in the Rig

In order to introduce non-linearity in the bearing stiffness as a step function of

the displacement along the horizontal axis, a stiffener was constructed and mounted as

shown in Fig 34.

Stiffener

Non-rotating
bearing support

Fig 34: Non-rotating bearing support without and with stiffener

Stiffener

Chain to apply force

Force
Transducer (lbs)

Fig 35: Stiffness measurements at the bearing

 46

A force transducer and a dial gauge was used (Fig 35 and Fig 36) to estimate the

stiffness at the bearing location and ensure that the force against displacement plot (Fig

37) showed the same step function characteristic as in Ehrich’s 1966 paper.

∆ = Clearance between stiffener

Stiffener and bearing support

Dial gauge

Fig 36: Close-up view of the stiffener

F in lb X in inches
Stiffness
0 0.00
5 0.00050 120
11 0.00110 100
15 0.00150 y = 43068x - 115.78
80
21 0.00210
Force in lbs

60
26 0.00260
31 0.00310 40
y = 8040.1x - 0.4002
41 0.00410 20
50 0.00500 0
70 0.00700 0.0000
-20
0.0010 0.0020 0.0030 0.0040 0.0050 0.0060
90 0.00900 Disp in inches
100 0.01000

Fig 37: Non-linear stiffness of the bearing

 47

Experimental Results with Non-linear Bearing Stiffness

Initial experiments were performed with the stiffener mounted at varying

clearances from the non-rotating bearing support till satisfactory results were obtained.

Fig 37 is such a case and all results reported below (unless otherwise mentioned) pertain

to that case.

Uncompensated
waveforms

Runout-
Compensated
waveforms

Fig 38: X and Y probe signals at 1300 RPM - below the critical speed

Fig 39: Probe signals at 3000 RPM – above the first critical
 48

Uncompensated

Runout-
Compensated
waveforms

Fig 40: Frequency demultiplication at twice the critical speed

Uncompensated
waveforms

Runout-
Compensated
waveforms

Fig 41: Probe signals at 4600 RPM

 49

Fig 38 - Fig 41 demonstrate how the rotor behavior changes as the effect of non-

linear stiffness sets in with increasing speed. At low speeds (below the critical, Fig 38)

the runout-compensated waveforms resemble sinusoids, indicating that the vibration is

mainly 1X from unbalance excitation. At higher speeds (above the first critical, Fig 39)

the X-probe signal is clipped (the stiffener is mounted along the X-axis) and also shows

the presence of higher harmonic components. Fig 40 illustrates the phenomenon of

frequency demultiplication as analyzed by Ehrich [5] with his planar model. Note that

the compensated probe signals at twice the critical speed (2 x 2010 RPM) shown on the

‘Resampled’ plot take two revolutions (x-axis scale shows the number of revolutions) to

complete one cycle. The phenomenon disappears as the rotor speed increases (Fig 41)

and the waveforms return to one cycle per revolution.

Frequency spectrum at various speeds

Fig 42: Spectrum at 1800 RPM indicate the presence of higher harmonics
 50

Fig 42 shows the presence of higher harmonics in the frequency spectrum below the

critical speed and is consonant with the numerical simulation results.

Fig 43: Subsynchronous vibration at 4200 RPM at 0.5X

The order spectrum at 4200 RPM (Fig 43) indicates the presence of a large

subsynchronous vibration which can be easily mistaken for instability. However as

proved in numerical and empirical results above, non-linear stiffness at the bearings

causes the large subsynchronous vibration at twice the first critical speed. A useful

diagnostic feature of such a case is that the subsynchronous component disappears

quickly when the speed is increased slightly as shown in Fig 44 .

 51

Fig 44: Subsynchronous vibration disappears on increasing the speed

A waterfall plot (Fig 45) summarizes the conclusions above.

Subsynchronous vibration
Critical

Fig 45: Waterfall plot from the X-probe showing the onset of subsynchronous vibration at 0.5X
 52

Orbit Plots

The orbit plots (Fig 46) exhibit an inside loop at twice the critical speed but is

transitory. With a slight change in speed, the loop disappears. The orbits bear similarities

with those predicted from the numerical simulations.

Fig 46: Orbits before, at and after the onset of subsynchronous vibration at twice the critical speed
From an analog oscilloscope

Experimental Results – Linear System

The results from the previous section are compared to the case where the rotor-

bearing stiffness was linear throughout its operating speed range. Measurements were

made with the nonlinear stiffener removed. The frequency spectrum does not show any

significant higher harmonic components, nor is there any large subsynchronous

component of vibration (Fig 47).

 53

Fig 47: Frequency spectrum for linear bearing support at different running speeds

It can be concluded from the above experiments and analysis that non-linearity in

stiffness of a rotor-bearing system induces a large subsynchronous vibration when

rotating at twice the critical speed. The vibration disappears on increasing the speed.
 54

CHAPTER VI

DIAGNOSING UNSTABLE SUBSYNCHRONOUS VIBRATIONS

USING THE SYNCHRONOUS PHASE ANGLE

Instability and Cross-coupled Stiffness

Mathematically, dynamic instability is defined as a solution to the linear

differential equations of motion characterized by a complex eigenvalue with a positive

real part. The vibration amplitude associated with instability becomes unbounded

(infinitely large), if linear stability theory holds. In practice, the vibration with growing

amplitude will cause seals or blades to rub or it will reach a ‘limit cycle’ due to non-

linearities in the system e.g. damping increases with amplitude. In rotordynamics,

negative direct damping seldom occurs. Instead cross coupled stiffness, modeled as a

follower force driving the forward going whirl orbit is usually the factor responsible for

instability.

The Jeffcott rotor model with cross coupled stiffness is a modal model for any

real machine operating through its first critical speed. It may be represented

mathematically as:

mx + cx + kx + k xy y = mω 2u cos ωt

(13)
my + cy + ky + k yx x = mω u sin ωt
2

It is found that for the particular case where kxy = - kyx = K ; K > 0; the cross coupled

forces drive the rotor unstable in forward whirl (the common mode in real machines). It
 55

represents a type of force induced by fluid forces around a turbine, impeller, or fluid

seal, or internal friction.

The particular solution to equation (13) is of interest. The amplitude and phase of

synchronous vibration is given as:

mω 2u
| r |=| x + y |=
2 2

(k − mω 2 ) 2 + (ω c − K ) 2
(14)
⎡ ωc − K ⎤
β = tan −1 ⎢
⎣ k − mω ⎥⎦
2

In effect, the equivalent damping due to cross coupled stiffness becomes,

K
ce = c − , (15)
ω

which is a function of the magnitude of cross coupled stiffness and the speed of rotation.

Noticeably, the synchronous phase angle β is also affected by the cross coupled stiffness.

It therefore can be a useful diagnostic value to determine whether the rotating system

can have negative equivalent damping (instability)! As shown by equation (14), both

backward (K < 0) and forward (K > 0) cross coupled follower forces will change the

damping and the value of β - the former increasing and the latter reducing it. However,

backward driving cross-coupled stiffness is rare, a soft continuous rotor rub being one of

the few examples. It is the case of forward driving cross coupling that is the cause of

most instabilities and the need arises to find how ‘β’ changes in this case. A plot of β

against rotating speed for various values of K is educative (Fig 48).

 56

shi := 0.1 wn := 60 K := 90000 w := 0 .. 120

⎡ ⎡ ⎛ 2⋅ shi ⋅ w − k ⎞ ⎤ ⎤
⎢ ⎢⎜ ⎥⎥
beta ( w , k) := ⎢
180
⋅ atan ⎢
⎝ wn K ⎠
⎥⎥
⎢ π ⎢ ⎛ w ⎞
2 ⎥⎥
⎢ ⎢ 1−⎜ ⎥⎥
⎣ ⎣ ⎝ wn ⎠ ⎦⎦

100
87.807

80

60
K= - 60,000
40
beta( w , − 60000)

beta( w , − 30000) K= - 30,000 K=60,000

20
beta( w , − 18000) K= - 18,000 K=30,000
beta( w , 0) 0 K=0
K = 18,000
beta( w , 18000) K=0
K=30,000 K= - 18,000
beta( w , 30000) 20

beta( w , 60000) K=60,000

40

K= - 60,000
60

80

− 87.788 100
0 0.5 1 1.5 2
0 w 2
wn

Fig 48: Plot of β against ω/ωn for various values of cross coupled stiffness. Direct stiffness = 90000,

direct damping coefficient = 10%

 57

A typical chart is plotted in MathCAD, for a rotor with 10% damping. At speeds lower

than the critical speed (for a slightly damped system) a backward driving cross coupled

force increases the value of β whereas a potentially unstable forward driving cross

coupling decreases β. At speeds higher than the first critical, the exact opposite occurs.

Also, the change in β is most sensitive near the critical speed range. This is further

illustrated in Fig 49 where the first derivative of β with respect to the cross coupled

stiffness K is plotted. The maximum change in β occurs near the critical speed range.

.10 5
−5 6
5.29×10

5 ω = 1.1ωn
beta( k , 0.7⋅ wn) 4 .10
d
dk

d 5
beta( k , 0.9wn) 2 .10
dk

d ω = 1.5ωn
beta( k , 1.0wn) 0
dk

d ω = 0.7ωn
beta( k , 1.1wn) 5
dk 2 .10

d
beta( k , 1.5wn) 5
dk 4 .10
ω = 0.9ωn
−5
− 5.847×10 6 .10
5
5 4 4 4 4 4 4 4 4 5
1 .10 8 .10 6 .10 4 .10 2 .10 0 2 .10 4 .10 6 .10 8 .10 1 .10
4 k
− 9×10 9×10

Fig 49: Plot of dβ/dK against K at different speeds

Fig 49 also indicates that there is a definite ratio of the cross coupled stiffness to the

system stiffness for which the cross coupling may have any significant effect on the

system equivalent damping. In this particular case (ξ = 0.1) the ratio is 20-25%. A
 58

similar result is presented in [1] is shown (Fig 50), where λ is the real part of the

eigenvalue.

Fig 50: Effect of cross coupled stiffness on stability [1]

Synchronous Phase Angle (β)

The synchronous phase angle is the angle by which the unbalance vector leads

the vibration vector in synchronous whirl (Fig 51).

 59

Fig 51: Synchronous phase angle β [1]

The synchronous phase angle can be measured (Fig 52) using the influence coefficient

method (actual measurement of the synchronous phase angle is not required for

diagnosis, but can be correlated to the phase angle displayed by most balancing

instruments). The steps for measurement are detailed below:

Step 1:

Run the rotor at a particular speed ‘N rpm’ and save the vibration vector (1X).

V = |V| @ θ1

All phases are measured from a fixed mark ‘P’ on the rotor. The ‘P’ mark is the exact

point below the probe when the tachometer notch is lined up with the tachometer.

Step 2:
Stop the rotor and add a trial known mass at a specified angle.
M = |M| @ γ
 60

Step 3:
Rev the rotor up to N rpm again to find the new vibration vector
V* = |V*| @ θ2

Evaluate ‘β’:
Vector A = V* - V
Vector α = + V/A
Unbalance vector U = α. M = |U| @ δ

‘β’, is calculated as the angle by which the unbalance vector leads the vibration vector.

i.e. β = δ – θ1

Direction of Rotation

U
V
A
P

V*

Fig 52: Measurement of the synchronous phase angle

Relationship between the Synchronous and Instrument Phase Angle

As stated earlier, the measurement of synchronous phase can be circumvented by

using common industrial balancing instruments. The phase displayed by balancing

 61

measurement instruments in the industry is however not the synchronous phase angle. A

simple relationship exists between the two.

Fig 53: Instrumentation for phase measurements [1]

Fig 53 illustrates an experimental setup for phase measurements for balancing. β is the

synchronous phase angle. All phase measurements in balancing machines are from the P

mark on the rotor which is the exact point below the X-probe, when the tachometer

notch is lined up with the tachometer. θ is the constant angular distance of the

unbalance of the rotor from the ‘P’ mark. All angles are measured positive opposite to

the direction of rotation. Following this convention,

L
θ =θ − β
(16)
∆θ = −∆β
 62

Experiments to Diagnose Subsynchronous Vibration

The experiment detailed below is proof of how the synchronous phase angle

conclusively demarcates between instability and a benign subsynchronous vibration. The

experimental rig shown in Fig 30 was used for investigating the variation of β with cross

coupled instability induced from a high pressure air swirl around the rotor. A special

virtual instrument was created with LabVIEWTM (Fig 54) to implement the influence

coefficient method of calculating β and used in tandem with LVTRC.

Fig 54: Virtual instrument to calculate β

Two sets of experiments were performed above the first critical speed – one with the

swirl inducer turned on and one without. The frequency spectrum showed the presence

of a strong subsynchronous component at the first eigenvalue in both the cases (Fig 55

and Fig 56). In the first case (Fig 55), subsynchronous vibration was induced from non-

linear bearing stiffness by mounting a stiffener (as explained in Chapter V) and without

air swirl around the rotor. ‘β’ was almost 180 degrees, as is to be expected for a lightly

damped rotor. In the second case (Fig 56) the swirl inducer was pressurized at 200 psi to
 63

induce de-stabilizing cross coupled forces (the stiffener was still mounted). The value of

β increased to -174 degrees clearly showing that the subsynchronous vibration is an

instability (refer to Fig 48).

Synch

Fig 55: Benign subsynchronous vibration at 0.5X

Synch

Fig 56: Unstable subsynchronous vibration at 0.5X when the swirl inducer is turned on

Note that the instrument phase shows a reverse trend and decreases from 91° to 82°.

Also, ∆β = -7.301 degrees and ∆θ = 9.2 degrees.

 64

The change of phase angle from cross coupling is more pronounced near the

critical speed. The experiments above were carried out at twice the critical speed so that

a benign subsynchronous vibration and an instability can be compared on how they

affect the phase angle. A second reason for the choice of experiment speed is that near

the critical speed, with de-stabilizing cross coupling, damping is minimal. Consequently,

the amplitude of vibration will be beyond safe limits. With greater direct damping (as in

most industrial machines), it will be possible to carry out experiments near the critical

speed. For this thesis, XLTRC simulations are used instead to study phase shifts from

cross coupling in more detail for the rotor described above with higher damping.

XLTRC Simulations to Study Phase Shifts from Cross-Coupled Forces

Complete details of the XLTRC model are included in Appendix (II). Fig 57 and

Fig 58 show the ‘Geo plot’ and the response of the rotor to unbalance excitation. De-

stabilizing cross coupled forces were added to the model (Fig 59 and Fig 60) to find out

how the synchronous phase angle changes.

5
Shaft Radius, inches

16 18
6 8 10
Shaft1 Shaft1
2.5 1 2 4 22 24
14
12 20

-2.5

-5

-7.5

-10
0 5 10 15 20 25

Axial Location, inches

Fig 57: Geo plot of the rotor rig

 65

Rotor dynam ic Re s pons e Plot

35

30
ilspk-pk

25

20
esponse, m

15

10
R

0
0 2000 4000 6000 8000 10000 12000 1400
Rotor Spe e d, r pm

Fig 58: Response curve for the rotor

5
Shaft Radius, inches

16 18
6 8 10
Shaft1 Shaft1
2.5 1 2 4 2224
14
12 20

-2.5

-5 Cross coupled
bearing
-7.5

-10
0 5 10 15 20 25 30

Axial Location, inches

Fig 59: Rotor model with cross coupled ‘bearing’ at station 17

Title: Swirl Inducer cross coupling

Perform a Paste/Special/Link for the Title box within XLTRC to create a link to your rotor model.

Speed Kxx Kxy Kyx Kyy Cxx Cxy Cyx Cyy

rpm lb/in lb/in lb/in lb/in lb-s/in lb-s/in lb-s/in lb-s/in
0 1500 -1500 0 0
2000 1500 -1500 0 0
4000 1500 -1500 0 0

Fig 60: Details of the cross coupled bearing at station 17

 66

Phase change with Swirl

400

350

300

250
Degrees

200
99 degrees
150

100

50
92 degrees
0
0 1000 2000 3000 4000 5000 6000 7000
RPM

Without Swirl With Swirl

Fig 61: Instrument phase plots from XLTRC simulations with and without cross coupling

XLTRC outputs the instrument phase (not β) from the simulations and Fig 61

shows how the instrument phase changes when the cross coupling bearing is

included/excluded from the model. Comparison with Fig 55 and Fig 56 show close

agreement with experimental values at around 4100 RPM

Change of instrument phase from XLTRC = 99° - 92° = 7°

 67

Change of instrument phase from actual experiments = 91° - 82° = 9°

XLTRC Simulations of the Swirl Inducer Rotor with Higher Damping

Industrial turbomachinery have much higher damping than the rotor modeled

above, to decrease the vibration amplitude near the critical speed. They also require

greater values of cross coupled stiffness to drive the rotor unstable. The swirl inducer

model is modified to include larger direct damping at the bearings. The results follow

(Fig 62, Fig 63 and Fig 64).

Rotordynam ic Response Plot

Wit hout Swirl, 40%damping

7
6
Response, mils pk-pk

5
4 Horz Amp
Vert Amp
3
2 Excitat ion = 1x

1
0
0 2000 4000 6000 8000 10000 12000 14000

Rotor Speed, rpm

Fig 62: Response of the rotor with higher damping

 68

Title: Swirl Inducer cross coupling

Perform a Paste/Special/Link for the Title box within XLTRC to create a link to your rotor model.

Speed Kxx Kxy Kyx Kyy Cxx Cxy Cyx Cyy

rpm lb/in lb/in lb/in lb/in lb-s/in lb-s/in lb-s/in lb-s/in
0 4000 -4000 0 0
2000 4000 -4000 0 0
4000 4000 -4000 0 0

Fig 63: Destabilizing bearing

Change of Phase with cross coupling: 40% direct

damping

350

300

250

200
Degrees

150

100

50

0
0 1000 2000 3000 4000 5000 6000 7000
RPM

Without Swirl With Swirl

Fig 64: Change of instrument phase angle with larger direct damping
 69

CHAPTER VII

DIAGNOSTIC INDICATORS OF BENIGN FREQUENCIES

Benign Subsynchronous Vibration and Their Indicators

Ertas, Kar and Vance [12] summarized diagnostic indicators for benign

subsynchronous vibrations. Several amongst them were verified after signal analyzing

vibration data from laboratory test rigs. The diagnostic indicators are discussed below:

Frequency Tracking

If the subsynchronous frequency tracks the running speed at a fixed fraction of

the synchronous, then it cannot be rotordynamic instability. Of course, if the machine

speed cannot be varied then tracking is impossible to determine. An exception to this

rule has been observed in certain pumps where the instability tracks the synchronous at a

constant fraction of around 0.8.

Agreement of Subsynchronous Frequency with known Eigenvalues of the System

Rotordynamic instability is the damped natural frequency of the rotor-bearing

system. The frequency can be determined from an accurate rotordynamic model or from

bump tests.
 70

Orbit Shape – Ellipticity

The sources of most instability are de-stabilizing follower forces pushing on the

rotor in the direction of subsynchronous whirl. Most follower forces are produced by

cross-coupled stiffness, which models a force that is always normal to the instantaneous

rotor deflection vector (orbit radius). As Fig 65 shows, this force can be purely a

follower force only if the orbit is circular. The force normal to the orbit radius becomes

more oblique to the orbital velocity as the orbit becomes more elliptical. The orbit shape

is not an absolute indicator of potential instability. However, it can be stated that highly

elliptical orbits (almost a straight line) cannot become unstable from cross-coupled

stiffness.

Fcc
V V

r Fcc r

Fig 65: Cross coupled force and velocity vectors in circular and elliptical orbits

The rate of energy input to the orbit from the cross coupled force is given as,
G G
J = Fcc .v

J = Fcc v cos θ
 71

where, θ is the angle between the force and velocity vectors.

Since, cos θ ≤ 1, J max = Fcc v for θ = 0 i.e. for a circular orbit.

A rotor going unstable requires energy to be ‘fed’ into the system from the cross coupled

forces in the direction of rotor whirl.

Orbits from benign subsynchronous vibrations are expected to be highly elliptical. The

orbits have to be filtered at exactly the subsynchronous frequency for them to be used as

any diagnostic indicator.

Presence of higher harmonics or multiple frequencies

Subsynchronous frequencies caused by loose bearing clearances, loose bearing

caps, or loose foundations are excited by intermittent impact as the separated surfaces

come together repeatedly as the machine runs. These impacts will excite a number of

natural frequencies, not just the one of concern. Spectral analysis of the complex signal

may expose a rich spectrum.

Experiments with Whirl Orbit Shapes

Two test rigs were used to capture subsynchronous orbits and test the hypothesis

regarding orbit shape. The Shell rig (Fig 66) is mounted on two double row self-aligning

ball bearings with a center disk constrained to the shaft. This test rig has the ability to

operate with different interference fits between the center disk and the shaft. At a critical

shrink fit the rotor goes violently unstable from internal friction. The Bently Rotor kit

(Fig 67) was used to capture a benign subsynchronous vibration induced from the

journal clearance within a worn out ‘Oilite’ bushing (Fig 68). LVTRC with its advanced
 72

filtering capability was required to extract the orbits at the subsynchronous frequency

desired.

B all B earing C enter D isk

H ousing W ith A dju stable
Interference
M otion Probe B all B earing
B racket H ousing

Fig 66: The Shell rotor rig

Vertical
Motion Probe

Old Sleeve
Bearing
Horizontal
Motion Probe

Fig 67: Bently rotor kit

 73

Fig 68: Worn-out sleeve bearing

Synch

Fig 69: Almost circular unstable orbit from the Shell rotor rig
 74

Synch

Fig 70: Benign subsynchronous orbit from the Bently rotor kit.

Fig 69 and Fig 70 show orbits from two different sources. The unstable subsynchronous

orbit from the Shell rig is almost circular whereas the benign orbit is extremely elliptical.
 75

CHAPTER VIII

CONCLUSION

This thesis has elucidated methodologies to ascertain whether a subsynchronous

vibration from a rotor is potentially unstable. It has detailed vibration signatures typical

to a rotating system with non-linear bearing or support stiffness, especially the large 0.5

X subharmonic response that is present when the rotor is running at twice its critical

speed. The experimental results were substantiated with numerical simulations of a short

rigid rotor with stiffness varying as a step function of the rotor displacement along a

particular axis. The subsynchronous vibration in this case is benign, not a true instability.

A new method of diagnosing instabilities by observing changes in the

synchronous phase angle was developed and verified with experiments. XLTRC

simulations were also used to show that de-stabilizing cross coupling changes the

synchronous phase, especially near the critical speed.

Orbit shapes were demonstrated as a potential diagnostic tool (not absolute).

Highly elliptical orbits are less likely to go unstable.

As an upshot of developing advanced rotordynamic measurement and analysis

tools, a state-of-the-art teststand is now available that far exceeds the capability of

current market products. The teststand software – LVTRC is now in use by the

Turbomachinery Research Consortium member industries.

 76

REFERENCES

1. Vance, J.M., 1988, Rotordynamics of Turbomachinery, John Wiley and Sons,

New York.
2. Ferrara, P.L., 1977, “Vibrations in Very High Pressure Compressors”, ASME-
77-DET-15.
3. Newkirk, B.L. and Taylor, H.D., 1925, “Shaft Whipping due to Oil action in

Journal Bearing”, General Electric Review, 28, pp. 559-568.

4. Newkirk, B.L., 1924, “Shaft Whipping”, General Electric Review, 27, pp. 35.

5. Alford, J.S., 1964, “Protecting Turbomachinery from Self Excited Rotor Whirl”,
ASME-64-WA/GTP-4, pp. 333-344.
6. Den Hartog, J.P., 1956, Mechanical Vibrations, 4th Edition, McGraw-Hill, New
York.
7. Ehrich, F.F., 1972, “Identification and Avoidance of Instabilities and Self-
Excited Vibrations in Rotating Machinery”, ASME-72-DE-21.
8. Ehrich, F.F., 1966, “Subharmonic Vibration of Rotors in Bearing Clearance”,
ASME-66-MD-1.
9. Bently, D.E., 1974, “Forced Subrotative Speed Dynamic Action of Rotating
Machinery”, ASME-74-PET-16.
10. Childs, D.W., 1981, “Fractional-Frequency Rotor Motion Due to Non-symmetric
Clearance Effects”, ASME- 81-GT-145.
11. Wachel J.C., 1982, “Rotordynamic Instability Field Problems”, Rotordynamic
Instability Problems in High Performance Turbomachinery -1982, NASA
Conference Publication 2250, Texas A&M University, College Station.
12. Ertas, B., Kar. R, Vance J.M, 2004, “Diagnosing Subsynchronous Vibrations –
Unstable or Benign”, Turbomachinery Consortium Annual Report, pp. 1-55.
 77

Supplemental sources consulted

Agilent Technologies, 2000, The Fundamentals of Signal Analysis - Application Note

243, Agilent Technologies, Palo Alto, California.

Ehrich, F.F., 1991, “Some Observations of Chaotic Vibration Phenomena in High Speed

Rotordynamics”, Journal of Vibration and Acoustics, 113, pp. 50-57.

Ehrich, F.F., 1992, “Observations of Sub critical Super Harmonic and Chaotic Response

in Rotordynamics”, Journal of Vibration and Acoustics, 114, pp. 117-121.

Eisenmann R.C, Sr., Eisenmann Robert C., Jr., 1997, Machinery Malfunction Diagnosis

and Correction, Prentice Hall, New Jersey.

Muszynska, A., 1989 “Rotor-to-stationary Element Rub-Related Vibration Phenomena

in Rotating Machinery”, The Shock and Vibration Digest, 21, pp. 3-11.

National Instruments, 2003, LabVIEW™ Order Analysis Toolset- User Manual, National

Instruments, Austin, Texas.

www.ni.com/developer, Example codes, accessed October 2003- March 2005.

Yao-Qun Lin, 1993, “Rotor Instability Induced by Dead Band Clearance in Bearing

Supports”, Ph.D. Dissertation, Department of Mechanical Engineering, Texas A&M

University, College Station.

 78

APPENDIX I

Details of the LVTRC System

The NI-4472 Data Acquisition Board (courtesy: www.ni.com):

The NI 4472 is a 8-channel data acquisition device which, when used with the Sound

and Vibration Toolset (LabVIEW™), allows high precision vibration measurements.

The test stand uses two such boards, thus possessing the capability of 16 channel

synchronized data acquisition. Input channels incorporate Integrated Electronic

Piezoelectric signal conditioning for accelerometers. The channels simultaneously

digitize signals over an anti-aliasing bandwidth of 0 to 45 kHz.

The NI 4472 PCI 8-Channel DSA board

Hardware

Eight analog inputs with 24 bit resolution analog-to-digital (A/D) converters achieve low

noise and distortion free signals. It is possible to sample at 102.4 kilo samples / second

without distortion.

Antialiasing:

Input signals are passed through fixed analog filters to remove signals with frequency
 79

components greater than those of the ADC’s; then digital antialiasing filters adjust their

cut off frequency to remove any frequency components above half the sampling rate.

Multidevice synchronization:

If more than one DAQ board is used, it is necessary to synchronize their operation.

Synchronization between multiboards are carried out by sharing a digital trigger from

one device and synchronizing all devices to the same clock. Minimal phase mismatch is

important for obtaining phase information in cross channel measurements. The NI 4472

has approximately 0.1 degree phase mismatch for a 1 kHz signal between two channels

across devices.

Triggering:

Several trigger modes are available for signal acquisition. Pre-trigger mode digitizes

signals before and after a trigger condition occurs. Post-trigger mode digitizes signals

after a trigger condition occurs. Delay trigger mode begins signal capture after a

programmable delay from the trigger. The source of the trigger can come from an analog

channel or an external digital trigger input.

Calibration:

The offset voltage and gain accuracy of the analog input are calibrated by National

Instruments. An onboard precision voltage reference is used for internal calibration to

ensure stable, accurate DC specifications.

The SMB-120 NI Attenuation Cables:

The SMB-120 is a specific attenuation cable for the NI-447X Data acquisition boards. It

has a custom attenuation circuit designed to avoid impedance mismatch and distortion.
 80

The attenuation cable ‘weakens’ the signal by 10 dB (a gain of –10dB). The cable is

designed specifically for eddy current proximity probes. By using the SMB-120 cable,

both the AC and the DC components can be measured.

 81

APPENDIX II

XLTRC2 Model of the Experimental Rig

The following section describes in detail the geometrical characteristics and the

XLTRC2 data sheets used to model the response of the experimental rotor.

Bearing input sheets

XLUseKCM™ User Defined Support Stiffness, Damping, and Mass Rotordynamic Coefficients Press Control-F1 for help.
Version 2.0, Copyright 1996 - 1998 by Texas A&M University. All rights reserved.
Title: Swirl Inducer Bearing stiffness inducer end
Perform a Paste/Special/Link for the Title box within XLTRC to create a link to your rotor model.

Speed Kxx Kxy Kyx Kyy Cxx Cxy Cyx Cyy Mxx Mxy Myx Myy
rpm lb/in lb/in lb/in lb/in lb-s/in lb-s/in lb-s/in lb-s/in lb-s**2/in lb-s**2/in lb-s**2/in lb-s**2/in
0 2739.5 0 0 3359.9 3 0 0 3 0 0 0 0
2000 2739.5 0 0 3359.9 3 0 0 3 0 0 0 0
4000 2739.5 0 0 3359.9 3 0 0 3 0 0 0 0
5000 2739.5 0 0 3359.9 3 0 0 3 0 0 0 0

Figure II. 1: Swirl Inducer end bearing input file

XLUseKCM™ User Defined Support Stiffness, Damping, and Mass Rotordynamic Coefficients Press Control-F1 for help.
Version 2.0, Copyright 1996 - 1998 by Texas A&M University. All rights reserved.
Title: Swirl Inducer Rig Bearing at Driver end
Perform a Paste/Special/Link for the Title box within XLTRC to create a link to your rotor model.

Speed Kxx Kxy Kyx Kyy Cxx Cxy Cyx Cyy Mxx Mxy Myx Myy
rpm lb/in lb/in lb/in lb/in lb-s/in lb-s/in lb-s/in lb-s/in lb-s**2/in lb-s**2/in lb-s**2/in lb-s**2/in
0 62335 0 0 116042.3 3 0 0 3 0 0 0 0
2000 62335 0 0 116042.3 3 0 0 3 0 0 0 0
4000 62335 0 0 116042.3 3 0 0 3 0 0 0 0

Figure II. 1: Air Turbine end bearing input file

The values were obtained experimentally using a force transducer and a dial gauge.

Bearing and Seal Definitions

STN 1 STN 2 Bearing Type UCS UCS Curve Type Output
# # Constraint \ Internal Factor Constant Spline \ Poly Loads Link (Paste Special in here)
3 0 Constraint 1 0 Poly no Swirl Inducer Rig Bearing at Driver end
22 0 Constraint 1 0 Poly no Swirl Inducer Bearing stiffness inducer end

Figure II. 2: Bearing stations

 82

Shaft input sheet

Shaft Properties Material Properties

Int. Hyst. Int. Visc. Type of Whirl Axial Elastic Shear
Starting Starting Rotation Shear Density
Shaft Damping Damping Shaft ======= Material Force Modulus Modulus
X0 Y0 Value Const ρ
Coeff Coeff Rotation Spin Const E G
Const or
# inches inches ηH ηV Ratio
rpm or N ω/Ω # (sc or -1) (ax or 0) lbm/in3 lbf/in2 lbf/in2
1 0 0 0 0 Ratio 1 1 1 -1 0 0.283 30.0E+6 12.0E+6
2 -1 0 0.1 11.5E+6 4.0E+6

Rotor Model Data Entry: Multiple Shafts, Elements, Sub-Elements and Layers. Station Numbers
H
Shaft Element Sub- Layer Length Left Right Material Shear Axial Left Right
I
Element OD ID OD ID Interact. Force Station Station
D
# # # # inches inches inches inches inches # ( 0 or 1 ) lbf # #
E
1 1 1 1 0.25 3.75 1.25 3.75 1.25 1 1 0 1 2
1 2 1 1 0.25 4 0 4 0 1 1 0 2 3
1 3 1 1 1.75 4 0 4 0 1 1 0 3 4
1 4 1 1 1.77 1.94 0 1.94 0 1 1 0 4 5
1 5 1 1 0.85 2.97 0 2.97 0 1 1 0 5 6
1 6 1 1 1 1.94 0 1.94 0 1 1 0 6 layer
1 6 1 2 1 6 1.94 6 1.94 2 1 0 layer 7
1 7 1 1 1 1.94 0 1.94 0 1 1 0 7 layer
1 7 1 2 1 6 1.94 6 1.94 2 1 0 layer 8
1 8 1 1 1.07 1.94 0 1.94 0 1 1 0 8 9
1 9 1 1 0.85 2.97 0 2.97 0 1 1 0 9 10
1 10 1 1 0.75 5.93 0 5.93 0 1 1 0 10 11
1 11 1 1 3 1.94 0 1.94 0 1 1 0 11 12
1 12 1 1 3 1.94 0 1.94 0 1 1 0 12 13
1 13 1 1 3.18 1.94 0 1.94 0 1 1 0 13 14
1 14 1 1 0.83 2.97 0 2.97 0 1 1 0 14 15
1 15 1 1 0.534 7.16 6.62 7.16 6.62 2 1 0 15 layer
1 15 1 2 0.534 2.97 0 2.97 0 1 1 0 layer 16
1 16 1 1 1.315 7.16 1.94 7.16 1.94 2 1 0 16 layer
1 16 1 2 1.315 1.94 0 1.94 0 1 1 0 layer 17
1 17 1 1 0.4355 7.16 1.94 7.16 1.94 2 1 0 17 layer
1 17 1 2 0.4355 1.94 0 1.94 0 1 1 0 layer 18
1 18 1 1 0.486 7.16 2.915 7.16 2.915 2 1 0 18 layer
1 18 1 2 0.486 1.94 0 1.94 0 1 1 0 layer 19
1 19 1 1 0.3935 7.16 4.629 7.16 4.629 2 1 0 19 layer
1 19 1 2 0.3935 1.94 0 1.94 0 1 1 0 layer 20
1 20 1 1 2.43 1.94 0 1.94 0 1 1 0 20 21
1 21 1 1 1.75 4 0 4 0 1 1 0 21 22
1 22 1 1 0.25 4 0 4 0 1 1 0 22 23
1 23 1 1 0.25 3.75 1.25 3.75 1.25 1 1 0 23 24

Figure II. 3: Shaft Input

Geo Plot
7.5

5
Shaft Radius, inches

16 18
6 8 10
Shaft1 Shaft1
2.5 1 2 4 22 24
14
12 20

-2.5

-5

-7.5

-10
0 5 10 15 20 25 30

Axial Location, inches

 83

Eigenvalues

Speed zeta1 cpm1 zeta2 cpm2 zeta3 cpm3 zeta4 cpm4 zeta5 cpm5
500. 0.118 2100.7 0.106 2327.3 0.028 9735.5 0.020 11749.6 0.022 22452.9
1000. 0.118 2100.6 0.106 2327.4 0.028 9735.3 0.020 11749.7 0.022 22450.8
1500. 0.118 2100.5 0.106 2327.5 0.028 9734.9 0.020 11750.0 0.022 22447.2
1700. 0.118 2100.4 0.106 2327.6 0.028 9734.8 0.020 11750.1 0.022 22445.4
1900. 0.118 2100.4 0.106 2327.7 0.028 9734.6 0.020 11750.2 0.022 22443.4
2000. 0.118 2100.3 0.106 2327.7 0.028 9734.5 0.020 11750.3 0.022 22442.3
2100. 0.118 2100.3 0.106 2327.8 0.028 9734.3 0.020 11750.4 0.022 22441.1
2200. 0.118 2100.2 0.106 2327.8 0.028 9734.2 0.020 11750.5 0.022 22439.9
2300. 0.118 2100.2 0.106 2327.9 0.028 9734.1 0.020 11750.6 0.022 22438.7
2500. 0.118 2100.1 0.106 2328.0 0.028 9733.8 0.020 11750.7 0.022 22436.0
3000. 0.118 2099.8 0.106 2328.3 0.028 9733.1 0.020 11751.3 0.022 22428.3
3500. 0.118 2099.4 0.106 2328.7 0.028 9732.2 0.020 11751.9 0.022 22419.3
4000. 0.118 2099.0 0.106 2329.1 0.028 9731.2 0.020 11752.6 0.022 22409.0
4500. 0.118 2098.5 0.106 2329.6 0.028 9730.0 0.020 11753.4 0.022 22397.5
6000. 0.118 2096.8 0.107 2331.4 0.028 9725.7 0.020 11756.4 0.022 22355.6
7000. 0.118 2095.4 0.107 2332.9 0.028 9722.2 0.020 11758.9 0.022 22322.2
8000. 0.118 2093.8 0.107 2334.6 0.028 9718.1 0.020 11761.7 0.022 22284.8
9000. 0.118 2092.0 0.107 2336.5 0.028 9713.5 0.020 11764.9 0.022 22243.7
10000. 0.118 2090.1 0.107 2338.5 0.028 9708.4 0.020 11768.4 0.022 22199.4

Figure II. 5: Damped Natural Frequency

Mode shapes

Damped Eigenvalue Mode Shape Plot Damped Eigenvalue Mode Shape Plot

forward forward
backward backward
f=2328.3 cpm f=11761.7 cpm
d=.1064 zeta d=.0203 zeta
N=3000 rpm N=8000 rpm

Figure II. 6: 1st and 2nd Forward Mode shapes

 84

Response

Rotordynam ic Re s pons e Plot

35

30
Response, mils pk-pk

25

20
Horz A mp

15 V ert A mp

10 Excit at ion = 1x

0
0 2000 4000 6000 8000 10000 12000 14000
Rotor Spe e d, rpm

Figure II. 7: Response to 1X excitation

De-stabilizing Cross Coupled Forces to drive the Rotor unstable

De-stabilizing cross coupled forces were added to the XLTRC model to find out

how the synchronous phase angle changes.

Title: Swirl Inducer cross coupling

Perform a Paste/Special/Link for the Title box within XLTRC to create a link to your rotor model.

Speed Kxx Kxy Kyx Kyy Cxx Cxy Cyx Cyy

rpm lb/in lb/in lb/in lb/in lb-s/in lb-s/in lb-s/in lb-s/in
0 1500 -1500 0 0
2000 1500 -1500 0 0
4000 1500 -1500 0 0

Figure II. 8: Bearing model for Destabilizing cross coupled forces

Bearing and Seal Definitions

STN 1 STN 2 Bearing Type UCS UCS Curve Type Output
# # Constraint \ Internal Factor Constant Spline \ Poly Loads Link (Paste Special in here)
3 0 Constraint 1 0 Poly no Swirl Inducer cross coupling
22 0 Constraint 1 0 Poly no Swirl Inducer Bearing stiffness inducer end
17 0 Constraint 1 0 Poly no Swirl Inducer cross coupling

Figure II. 9: Inclusion of the bearing at station 17 (Swirl inducer)

 85

7.5

5
Shaft Radius, inches
16 18
6 8 10
Shaft1 Shaft1
2.5 1 2 4 2224
14
12 20

-2.5

-5

-7.5

-10
0 5 10 15 20 25 30

Axial Location, inches

Figure II. 10: Modified Geo Plot

Speed zeta1 cpm1 zeta2 cpm2 zeta3 cpm3 zeta4 cpm4

500. -0.033 2313.1 0.598 2365.0 0.094 9662.1 0.066 11656.6
1000. -0.033 2315.3 0.598 2361.7 0.094 9661.0 0.066 11657.8
1500. -0.032 2317.5 0.599 2358.4 0.094 9659.8 0.065 11659.0
1700. -0.032 2318.4 0.599 2357.1 0.094 9659.3 0.065 11659.5
1900. -0.032 2319.2 0.599 2355.8 0.094 9658.7 0.065 11660.0
2000. -0.032 2319.7 0.599 2355.1 0.094 9658.4 0.065 11660.3
2100. -0.032 2320.1 0.599 2354.5 0.094 9658.2 0.065 11660.6
2200. -0.032 2320.6 0.600 2353.8 0.094 9657.9 0.065 11660.8
2300. -0.032 2321.0 0.600 2353.2 0.094 9657.6 0.065 11661.1
2500. -0.031 2321.9 0.600 2351.9 0.095 9657.0 0.065 11661.7
3000. -0.031 2324.1 0.600 2348.6 0.095 9655.4 0.065 11663.1
3500. -0.031 2326.3 0.601 2345.3 0.095 9653.6 0.064 11664.7
4000. -0.030 2328.5 0.601 2342.0 0.095 9651.8 0.064 11666.3
4500. -0.030 2330.7 0.602 2338.7 0.095 9649.8 0.064 11668.1
6000. -0.028 2337.4 0.603 2328.9 0.095 9643.1 0.063 11673.8

Figure II. 11: Eigenvalue Table showing that the first forward mode is unstable
 86

Speed Horz Amp Horz Phs Vert Amp Vert Phs Speed Horz Amp Horz Phs Vert Amp Vert Phs
rpm mils pk-pk degrees mils pk-pk degrees rpm mils pk-pk degrees mils pk-pk degrees
200 0.038941 89.12254 0.031744 359.2864 200 0.036677 15.42908 0.03049 289.0259
300 0.088593 88.6693 0.07207 358.9201 300 0.083548 14.96375 0.069438 288.7151
400 0.159992 88.1979 0.12977 358.5419 400 0.151005 14.53617 0.125406 288.4732
500 0.255181 87.70101 0.206167 358.1474 500 0.240927 14.1429 0.199832 288.3012
600 0.377029 87.16996 0.303091 357.7311 600 0.355911 13.78094 0.294685 288.2018
700 0.529463 86.59429 0.422993 357.287 700 0.499445 13.44776 0.412611 288.1794
800 0.717802 85.96088 0.569105 356.8077 800 0.676188 13.14124 0.557142 288.241
900 0.949275 85.25283 0.745677 356.2841 900 0.892354 12.85973 0.733008 288.3959
1000 1.233794 84.44778 0.958317 355.7047 1000 1.15629 12.60219 0.946616 288.6573
1100 1.585177 83.51519 1.214487 355.055 1100 1.479345 12.36852 1.206786 289.0424
1200 2.023106 82.41207 1.52425 354.3156 1200 1.877202 12.1602 1.525939 289.5741
1300 2.576398 81.07569 1.901412 353.4607 1300 2.372 11.98162 1.922067 290.2824
1400 3.288695 79.41075 2.365322 352.4548 1400 2.995814 11.84281 2.422147 291.2065
1500 4.228803 77.26532 2.94381 351.2481 1500 3.796689 11.76489 3.068333 292.3976
1600 5.510398 74.38287 3.678176 349.7687 1600 4.849624 11.79186 3.929785 293.9241
1700 7.331096 70.29928 4.632076 347.9114 1700 6.278337 12.01579 5.126706 295.8825
1800 10.04868 64.10498 5.908813 345.5211 1800 8.30269 12.63431 6.883299 298.4292
1900 14.28696 53.87962 7.691393 342.3664 1900 11.35646 14.09199 9.657573 301.886
2000 20.61637 35.69407 10.37023 337.974 2000 16.42958 17.48213 14.5113 307.1082
2100 26.21211 5.933501 14.88441 330.184 2100 26.23477 26.01073 24.36677 316.9573
2200 24.50698 335.4161 22.13681 313.7107 2200 47.84711 51.85514 47.13383 343.2533
2300 19.99079 316.9228 29.988 285.2 2300 54.77171 112.132 57.28316 42.913
2400 16.76416 305.3876 30.24526 251.2452 2400 32.29723 145.7341 35.59097 74.91663
2500 14.23562 297.1955 24.37293 227.521 2500 21.46615 158.0882 24.59095 85.03119
2600 12.32936 291.8968 19.27838 214.2325 2600 16.19907 164.4869 19.00596 88.96626
2700 10.94585 288.3693 15.85801 206.4886 2700 13.22715 168.6188 15.69316 90.73617
2800 9.922806 285.8687 13.55771 201.5702 2800 11.36396 171.577 13.50932 91.61043
2900 9.143443 284.0011 11.94416 198.2071 2900 10.10418 173.7948 11.96598 92.07722
3000 8.533004 282.5497 10.76332 195.7728 3000 9.202317 175.4961 10.82128 92.34586
3100 8.04341 281.3867 9.867462 193.9318 3100 8.527245 176.8202 9.941543 92.51224
3110 7.99977 281.2829 9.789773 193.7721 3110 8.469003 176.9358 9.86475 92.52529
3120 7.956987 281.1811 9.71395 193.6162 3120 8.4122 177.0487 9.789724 92.53781
3130 7.915037 281.0812 9.639927 193.464 3130 8.35678 177.1589 9.716402 92.54985
3140 7.873895 280.9832 9.567645 193.3153 3140 8.302697 177.2666 9.644734 92.56142
3150 7.833542 280.887 9.497045 193.1701 3150 8.249902 177.3717 9.574666 92.57256
3160 7.793953 280.7926 9.42807 193.0281 3160 8.198351 177.4744 9.506144 92.58328
3170 7.755109 280.7 9.360668 192.8893 3170 8.147998 177.5748 9.439123 92.59361
3180 7.716989 280.609 9.294787 192.7536 3180 8.098805 177.6729 9.373556 92.60356
3190 7.679573 280.5196 9.23038 192.6209 3190 8.05073 177.7687 9.309397 92.61316
3200 7.642845 280.4319 9.167398 192.491 3200 8.003738 177.8623 9.246602 92.62242
3210 7.606782 280.3456 9.105794 192.364 3210 7.957788 177.9538 9.185133 92.63135
3220 7.571372 280.2609 9.045529 192.2396 3220 7.912852 178.0432 9.124947 92.63997
3500 6.787941 278.3647 7.768417 189.5803 3500 6.961244 179.8918 7.838603 92.79678
4100 5.860889 276.0136 6.392496 186.5947 4100 5.921533 181.5954 6.43443 92.88024
4400 5.582558 275.2634 6.008859 185.7106 4400 5.623178 181.947 6.041319 92.86048
4500 5.505434 275.05 5.905271 185.4651 4500 5.541418 182.0264 5.935178 92.84815
4600 5.434475 274.8513 5.811113 185.2389 4600 5.46652 182.0917 5.838716 92.83345
4700 5.368912 274.6655 5.725162 185.0295 4700 5.397589 182.1447 5.750685 92.81665
4800 5.308092 274.4913 5.64639 184.835 4800 5.333865 182.1874 5.670037 92.79798
4900 5.251444 274.3273 5.573933 184.6536 4900 5.274705 182.2212 5.59588 92.77763
5000 5.198482 274.1725 5.507051 184.4838 5000 5.219554 182.2472 5.527456 92.75579
5100 5.148777 274.0259 5.44511 184.3245 5100 5.167935 182.2665 5.464117 92.73261
5200 5.101954 273.8867 5.387565 184.1745 5200 5.119429 182.28 5.405299 92.70823
5300 5.057681 273.7541 5.333945 184.0328 5300 5.073672 182.2883 5.35052 92.68278
5400 5.015662 273.6276 5.283836 183.8987 5400 5.030341 182.2922 5.299356 92.65638
5500 4.975633 273.5066 5.236881 183.7714 5500 4.989146 182.2922 5.251434 92.62911
5600 4.937354 273.3905 5.192759 183.6503 5600 4.949829 182.2887 5.206428 92.60106
5700 4.900608 273.2788 5.151194 183.5348 5700 4.912156 182.2823 5.164053 92.57232
5800 4.865195 273.1713 5.111936 183.4245 5800 4.875915 182.2732 5.124052 92.54296
5900 4.830928 273.0675 5.074763 183.3188 5900 4.840907 182.2618 5.086197 92.51302
6000 4.797633 272.967 5.039479 183.2174 6000 4.806951 182.2484 5.050288 92.48257

Figure II. 12: Response without cross coupling Figure II. 13: Response with cross coupling
 87

VITA

Name: Rahul Kar

Address: Department of Mechanical Engineering, 3123, TAMU, College

Station, Texas, USA.

Email: [email protected]

Education: B.E., Mechanical Engineering, National Institute of Technology,

India, 2002

M.S., Mechanical Engineering, Texas A&M University,

USA, 2005