Professional Documents
Culture Documents
Avinash C. Balbahadur
APPROVED:
________________________
Dr. R.G. Kirk (Chair)
________________________ ________________________
Dr. M.E.F. Kasarda Dr. R.L. West
________________________ ________________________
Dr. B. Vick Dr. R.K. Kapania
Unlike most instabilities, which are non-synchronous in nature, the Morton Effect
rotors that are supported by fluid film bearings and is caused by differential viscous
shearing within the bearing lubricant. The Morton Effect has also gained much attention
Prior studies of the Morton Effect have used complex analysis in the frequency
domain to model this instability. However, such an approach makes it difficult to develop
a user-friendly design tool for engineers. The current research employs a steady-state
analysis to predict the onset of the Morton Effect, and it uses an instability criterion which
is based on a threshold unbalance caused by a force equal to 15% of the weight of the
rotor. It is hoped that this method will provide a more easily adaptable platform for
The current model has demonstrated good agreement with other theoretical
models and experimental data. This agreement applies to rotors that are supported by
either plain or tilting pad journal bearings and it was found that a worse case scenario for
the Morton Effect would involve centered, circular and large-amplitude bearing orbits.
A test rotor was also designed and built. Initial experimental data revealed an
unusual instability that might have been caused by the Morton Effect.
Dedicated to My Parents:
iii
Acknowledgments
I am profoundly grateful for the advice provided by my advisor Dr. Kirk. Without
his guidance, this project would never have been accomplished. My gratitude also extends
to the members of my committee: Dr. Kasarda, Dr. Vick, Dr. West and Dr. Kapania for
I am deeply indebted to Dr. Swanson for his help with the experimental system.
Also, many thanks to the technicians in the Mechanical Engineering Machine Shop. I am
especially grateful to Mr. James Dowdy for his precision work on various parts of the test
rotor. To Mr. Ben Poe and Mr. Jamie Archual goes my gratitude for rescuing me from a
provided by Ms. Eloise McCoy, Ms. Cathy Hill, Ms. Jackie Buhrdorf, Mr. Mike Harness
I would also like to thank my friends and relatives for all of their support. This
iv
TABLE OF CONTENTS
ACKNOWLEDGMENTS.................................................................................................................... IV
LIST OF FIGURES.............................................................................................................................VII
NOMENCLATURE ...............................................................................................................................X
1. INTRODUCTION ...............................................................................................................................1
1.1. LITERATURE REVIEW .......................................................................................................................2
1.2. MOTIVATION FOR RESEARCH .........................................................................................................10
1.3. RESEARCH OBJECTIVES ..................................................................................................................11
1.4. DEVELOPMENT OF THE CURRENT MODEL .......................................................................................11
2. THE MORTON EFFECT IN ROTORS WITH PLAIN JOURNAL BEARINGS..........................16
2.1. INITIAL MECHANICAL UNBALANCE.................................................................................................16
2.2. FILM THICKNESS FOR PLAIN JOURNAL BEARINGS ............................................................................17
2.3. STATIC EQUILIBRIUM JOURNAL POSITION .......................................................................................19
2.4. SYNCHRONOUS ORBIT ...................................................................................................................20
2.5. TEMPERATURE DISTRIBUTION ........................................................................................................24
2.6. THERMAL UNBALANCE ..................................................................................................................29
2.7. INSTABILITY CRITERION ................................................................................................................33
2.8. COMPUTER PROGRAM FLOWCHART ................................................................................................34
2.9. CASE STUDIES ...............................................................................................................................36
2.9.1. Keogh and Morton ...............................................................................................................36
2.9.2. Faulkner, Strong and Kirk...................................................................................................43
2.10. SOLUTION SENSITIVITY TO THE THRESHOLD UNBALANCE FORCE ..................................................51
3. THE MORTON EFFECT IN ROTORS WITH TILTING PAD JOURNAL BEARINGS ............53
3.1. FILM THICKNESS FOR TILTING PAD JOURNAL BEARINGS ..................................................................54
3.2. STATIC EQUILIBRIUM JOURNAL POSITION .......................................................................................64
3.2.1. Reynolds’ Equation...............................................................................................................65
3.2.2. Pressure Distribution ............................................................................................................67
3.2.3. Forces and Moments.............................................................................................................72
3.3. SYNCHRONOUS ORBIT ...................................................................................................................78
3.4. TEMPERATURE DISTRIBUTION ........................................................................................................82
3.5. CASE STUDIES ...............................................................................................................................88
3.5.1. de Jongh and Morton............................................................................................................88
3.5.2. de Jongh and van der Hoeven...............................................................................................97
3.5.3. Overhung Compressor ........................................................................................................103
4. DESIGN OF TEST ROTOR...........................................................................................................109
4.1. DESIGN PARAMETERS ..................................................................................................................110
4.2. ANALYSIS OF TEST ROTOR DESIGNS .............................................................................................114
4.3. TEST ROTOR COMPONENTS ..........................................................................................................120
4.3.1. Test Shaft............................................................................................................................121
4.3.2. Bearings..............................................................................................................................122
4.3.3. Coupling .............................................................................................................................123
v
4.3.4. Overhang Disk ....................................................................................................................124
4.3.5. Inboard Disk .......................................................................................................................126
4.4. ANALYSIS OF FINAL TEST ROTOR DESIGN .....................................................................................127
5. EXPERIMENTAL RESULTS FROM TEST ROTOR..................................................................140
5.1. ROTOR BALANCING .....................................................................................................................140
5.2. FIXED SPEED DATA......................................................................................................................142
5.3. ACCEL-DECEL DATA ...................................................................................................................145
6. CONCLUSIONS..............................................................................................................................160
6.1. CONCLUSIONS BASED ON THE THEORETICAL WORK.......................................................................160
6.2. CONCLUSIONS BASED ON THE EXPERIMENTAL RESEARCH ..............................................................161
7. SUGGESTED FUTURE RESEARCH ...........................................................................................162
7.1. SUGGESTED FUTURE THEORETICAL WORK....................................................................................162
7.2. SUGGESTED FUTURE EXPERIMENTAL RESEARCH ...........................................................................163
APPENDIX A: MATLAB GUI PROGRAM FOR PLAIN JOURNAL BEARINGS........................164
APPENDIX C: MATLAB GRAPHICS PROGRAM FOR FORTRAN TPJB PROGRAM DATA .207
REFERENCES....................................................................................................................................213
VITA....................................................................................................................................................216
vi
List of Figures
Figure 1. 1 - An Example of the Newkirk Effect........................................................................................2
Figure 1. 2 - An Example of the Morton Effect..........................................................................................6
Figure 1. 3 - Schematic Diagram of The Morton Effect Model used by Other Researchers.......................11
Figure 1. 4 - Schematic Diagram of Current Morton Effect Model ..........................................................13
Figure 3. 1 - Schematic Diagrams of Tilting Pad Journal Bearings Showing Load-on-Pad and Load-
between-Pads Configurations...........................................................................................................54
Figure 3. 2- Schematic Diagram of a Cylindrical Pivot TPJB ..................................................................55
Figure 3. 3 - Triangles from Diagram of The Cylindrical Pivot TPJB ......................................................57
Figure 3. 4 - Comparison of Film Thickness Expressions for a 5-Pad Cylindrical Pivot TPJB..................63
Figure 3. 5 - Coordinate System for Reynolds’ Equation..........................................................................65
Figure 3. 6 - Pad Surface Showing Pressure Boundary Conditions...........................................................67
Figure 3. 7 - A 4-Node Isoparametric Quadrilateral Element...................................................................70
Figure 3. 8 - Comparison between Calculated and Actual Dimensionless Stiffness and Damping
Coefficients from Someya’s 5 -Pad Bearing .....................................................................................81
Figure 3. 9 - Temperature Distribution for Knight’s Bearing...................................................................86
Figure 3. 10 - Film Thickness, Pressure and Temperature at the Centerline of Knight’s Bearing (Table 3.
6).....................................................................................................................................................87
Figure 3. 11 - VT-FAST Model of deJongh and Morton Centrifugal Compressor Rotor...........................89
Figure 3. 12 - Critical Speed Maps for VT-FAST Model and Actual Compressor ....................................91
Figure 3. 13 - Unbalance Curves for the Non-Driven End of the deJongh and Morton Compressor (Before
Modifications) .................................................................................................................................92
Figure 3. 14 - Unbalance Curves for the Driven End of the de Jongh and Morton Compressor (Before
Modifications) .................................................................................................................................93
Figure 3. 15 - Unbalance Curves for the Non-Driven End of the de Jongh and Morton Compressor (After
Modifications) .................................................................................................................................95
Figure 3. 16 - Unbalance Curves for the Driven End of the de Jongh and Morton Compressor (After
Modifications) .................................................................................................................................96
vii
Figure 3. 17 - VT-FAST Model of a Compressor Analyzed by de Jongh and van der Hoeven ..................97
Figure 3. 18 - Critical Speed Maps for VT-FAST Model and an Actual Compressor Analyzed by de Jongh
and van der Hoeven .......................................................................................................................100
Figure 3. 19 - Unbalance Curves for the Non-Driven End of a Compressor Analyzed by de Jongh and van
der Hoeven (Before Modifications) ................................................................................................101
Figure 3. 20 - Unbalance Curves for the Non-Driven End of a Compressor (with Heat Sleeve Barrier)
Analyzed by de Jongh and van der Hoeven ....................................................................................102
Figure 3. 21 - Synchronous Vibration Data obtained from an X-Probe at the NDE of the Overhung
Compressor ...................................................................................................................................104
Figure 3. 22 - Synchronous Vibration Data obtained from an X-Probe at the DE of the Overhung
Compressor ...................................................................................................................................105
Figure 3. 23 - Unbalance Curves for the NDE of the Overhung Compressor..........................................106
Figure 3. 24 - Unbalance Curves for the DE of the Overhung Compressor.............................................107
Figure 5. 1 - Vibration Readings from Probe 5 (Before and After Balancing) ........................................141
Figure 5. 2 - Speed and Probe 3 Response for a Fixed Speed Data Plot ..................................................144
Figure 5. 3 - Probe 3 Response Data for 5000-9000 RPM Acceleration-Deceleration (Oil Inlet
Temperature = 100 oF)...................................................................................................................145
Figure 5. 4 - Probe 5 Response Data for 5000-9000 RPM Acceleration-Deceleration (Oil Inlet
Temperature = 100 oF)...................................................................................................................146
Figure 5. 5 - Cascade Plot for Vibrations Shown in Figure 5. 3 .............................................................147
Figure 5. 6 - Cascade Plot for Vibrations Shown in Figure 5. 4 .............................................................148
Figure 5. 7 - Shaft Centerline Position (from Probes 3 and 4) near NDE Bearing .................................149
Figure 5. 8 - Shaft Centerline Position (from Probes 5 and 6) near DE Bearing .....................................150
Figure 5. 9 - Probe 3 Response Data for 5000-9000 RPM Acceleration-Deceleration (Oil Inlet
Temperature = 78.3 oF)..................................................................................................................152
Figure 5. 10 - Probe 5 Response Data for 5000-9000 RPM Acceleration-Deceleration (Oil Inlet
Temperature = 78.3 oF)..................................................................................................................153
Figure 5. 11 - Probe 3 Response Data for 5000-9500 RPM Acceleration-Deceleration (Oil Inlet
Temperature = 100 oF)...................................................................................................................154
viii
Figure 5. 12 - Probe 5 Response Data for 5000-9500 RPM Acceleration-Deceleration (Oil Inlet
Temperature = 100 oF)...................................................................................................................155
Figure 5. 13 - Synchronous Response at the NDE Bearing of de Jongh and Morton’s LP/IP Compressor
(de Jongh and Morton, 1994).........................................................................................................157
List of Tables
Table 1. 1 - Comparison between the Newkirk Effect and the Morton Effect .............................................7
Table 1. 2 - Comparison between Other and Current Morton Effect Models ............................................13
ix
Nomenclature
Dimensions: M = Mass, L = length, t = time, T = temperature
e eccentricity (L)
m preload factor
x
Ob bearing center
Oj journal center
L-1 t-2)
Re Reynolds’ number
turbulent)
tp Pad thickness
U resultant unbalance (M L)
Um mechanical unbalance (M L)
Ut thermal unbalance (M L)
xi
Wb bearing load (M L t-2)
surface
center
ε eccentricity ratio
θ circumferential angle
xii
unbalances
ya attitude angle
13
1. Introduction
The term ‘rotordynamics’ or ‘rotor dynamics’ consists of two words: ‘rotor’ and
‘dynamics’. ‘Rotor’ refers to objects which spin or rotate while ‘dynamics’ deals with the
Since the invention of the wheel in Mesopotamia circa 3500-3000 B.C., humans
have been fascinated by spinning objects. However, it was not until 1869 that Rankine
wrote the first article on rotordynamics (Kirk, 1998). In this article Rankine concluded
that operation above the critical speed would be unstable. This hypothesis was later
disproved by DeLaval (in 1895) when he demonstrated that a steam turbine was capable
of running above the first critical speed. A theoretical explanation of this phenomenon
was offered by the English dynamicist, H. H. Jeffcott in 1919. The Jeffcott rotor model
consisted of a mass-less elastic shaft on rigid supports with an unbalanced disk at the mid-
span of the shaft. This model showed that the rotor response peaked only at the critical
speed and attenuated to stable values at speeds above and below this critical. Even though
this model is simplistic, and only accounts for a single critical speed, it is still used today in
1
1.1. Literature Review
Just seven years after Jeffcott developed his rotor model, Newkirk published a
paper (Newkirk, 1926) on a thermally-induced problem which was observed in the rotor
of a water-wheel generator. Newkirk concluded that the cause of the problem was a
stator element which rubbed against the shaft to produce a hot spot (Figure 1. 1).
bearing bearing
disk
shaft
* mechanical
unbalance
bearing bearing
2
During this operation, the rotor is executing synchronous whirl, i.e. the shaft or journal is
rotating at the same speed as the rate at which the shaft center is orbiting the bearing
center. This configuration ensures that a particular area of the shaft surface rubs against
the stator piece to create a hot spot, while the diametrically opposite portion of the shaft
never experiences this frictional contact and remains at the ambient temperature. Such a
temperature difference causes a thermal gradient to develop across the shaft and
eventually leads to thermal bending. Newkirk observed that if the shaft was running
below its first critical speed then the mechanical unbalance enhances the thermal bend
caused by the hot spot, as shown in Figure 1. 1(a). This leads to increased shaft vibrations
which promote more contact with the stator. As a result, the system becomes unstable via
a positive feedback mechanism. On the other hand, if the shaft is running above the first
critical speed (Figure 1. 1(b)), then the mechanical unbalance is roughly out of phase with
the hot spot and inhibits the unbalance produced by the thermal bend vector. Hence, the
system shown in Figure 1. 1(b) will be stable which is concordance with Newkirk’s
observations. It should be noted that the Newkirk Effect is not limited to occurring below
the first critical speed. Most of the later researchers, who are referenced in the next
R.P. Kroon and W.A. Williams later gave a mixed qualitative and quantitative
analysis (Kroon and Williams, 1939) of the Newkirk Effect, but it was decades before a
the static thermal bow due to an arbitrary heat input and then combined this bow with the
system dynamics to yield a dynamic thermal bow. The final model consisted of two
3
digital computer to solve this problem may be one reason why it took about 30 years
indicated that the Newkirk Effect consisted of 3 modes: a spiralling mode in which
sinusoidal vibrations, and a constant mode with a fixed time-independent vibration level.
Dimarogonas concluded that the main factor which determines the vibration mode is the
phase difference between the static mechanical bow and the dynamic thermal bow relative
The Newkirk Effect was then studied by Kellenberger (Kellenberger, 1980) who
between the turbogenerator rotor and stator, via a seal ring, produced the hot spot
equations by making some simplifying assumptions, e.g. the shaft thermal bow is linearly
proportional to the shaft temperature. He also found that the interaction between the
thermal and mechanical unbalances was an essential driving force behind the Newkirk
Effect. In addition, he discovered that the ratio of the heat flow into the shaft to the heat
flow out of the shaft determined the system stability, i.e. if this ratio was above some
threshold curve then the system would yield unstable spiral vibrations. Schmied (Schmied,
rotors by using a finite element code. He found that the hot spots could also be caused by
slipring brushes rubbing against the rotor shaft. More recent work concerning the
Newkirk Effect has been done by Goldman and Muzynska (Goldman and Muzynska,
1995) and Goldman, Muzynska and Bently (Goldman et al., 1998). These researchers
4
have noted that the speed of the transient thermal effects is very slow compared to the
rotor vibration velocities. As a result, the thermal problem can be uncoupled from the
rotor vibrations. They have also accounted for inelastic impacts during rotor-to-stator
contact.
Schmied (Schmied, 1987) indicated that unstable spiral vibrations could also arise
from hot spots developing within the fluid-film bearing. Research done by Keogh and
Morton (Keogh and Morton, 1994) confirms the existence of such instabilities which
occur primarily in overhung rotors (Figure 1. 2). These unstable vibrations constitute the
5
hot spot caused by
mechanical
uneven viscous shearing
unbalance
within bearing
*
inboard rotor mass shaft
overhung disk
bearing bearing
mechanical
inboard rotor mass shaft unbalance
*
overhung disk
bearing bearing
The Morton Effect occurs when the journal is executing a synchronous orbit around the
bearing center. This orbit causes one portion of the journal surface to always be at the
minimum film thickness, while a diametrically opposite section of the journal surface is
always at the maximum film thickness. Lower film thickness areas are generally
associated with higher viscous shear stresses which produce higher temperatures. As a
result, a hot spot will develop on the journal surface exposed to the minimum film
6
thickness region and a cold spot will be formed on the surface at maximum film thickness.
This temperature difference leads to a temperature gradient developing across the journal.
If the hot spot is 180 degrees out of phase with the overhung unbalance (Figure 1. 2(b)),
and if the temperature gradient and the magnitude of the unbalance is adequate, thermal
bending will occur. Under these conditions the bent shaft will decrease the bearing
clearance and elevate the thermal gradient. The increased temperature gradient will then
initiate more thermal bending. These actions describe a positive feedback mechanism
Both the Newkirk Effect and the Morton Effect involve the development of a
temperature gradient across the journal which eventually leads to unstable thermal bending
via a positive feedback mechanism. Some of the other similarities and differences between
Table 1. 1 - Comparison between the Newkirk Effect and the Morton Effect
Mechanical Unbalance in phase with hot spot out of phase with hot spot
The first thorough analysis (Keogh and Morton, 1994) of the Morton Effect
7
dependent thermal bend was first calculated by combining the heat transfer equations with
the dynamic equations of the rotor. This thermal bend was then transformed to the
frequency domain where it was incorporated into a positive feedback loop. The stability
characteristics of this loop were then obtained by plotting Nyquist and eigenvalue graphs
as a function of rotor speed. Keogh and Morton thus obtained a range of instability
speeds for which the real part of the eigenvalues were positive.
A more practical insight into the Morton Effect was obtained by deJongh and
Morton (deJongh and Morton, 1994) who investigated a synchronous vibration instability
application. This compressor was mounted on two oil-lubricated, tilting pad journal
bearings and had a mass of about 450 kg (990 lbm). DeJongh and Morton found that the
RPM. They also noticed that when the rotor speed was reduced the unstable vibrations
gradually subsided. This hysteresis in response indicated that the problem could not have
been caused by running too close to a critical speed. In any case, the nearest critical was
around 14500 RPM which was sufficiently separated from the operating speed. Such
behavior implied that the instability could be thermal in nature, so the researchers removed
the mechanical and labyrinth seals and restarted the rotor. Unfortunately, the unstable
vibrations persisted which implied that rotor-to-stator contact (via the seal rings) could
not have been responsible for this anomaly. Hence, the Newkirk Effect could be ruled out
as the potential culprit. The compressor was finally stabilized by reducing the overhang
unbalance which was achieved by replacing some of the steel parts with lighter ones made
8
overhang unbalance, seemed to imply that the Morton Effect could be the source of the
problem. DeJongh and Morton confirmed this hypothesis by building a test rotor with
identical dynamic characteristics as the actual rotor. They measured the temperature
difference across the journal and showed that a rise in this transverse temperature gradient
Faulkner, Strong and Kirk (Faulkner et al., 1997) also encountered the Morton
Effect operating in a large turbocharger. The synchronous orbit and the overhung turbine
wheel caused a hot spot to develop within the turbine-end bearing. This problem was
eccentricity. Such a configuration forced a less-centered orbit which caused the hot spot
to be alternately heated and then cooled. As a result, the average temperature of the hot
spot diminished and the thermal gradient was reduced. The thermal bending was therefore
Another method for decreasing the thermal gradient from the Morton Effect was
given by de Jongh and van der Hoeven (de Jongh and van der Hoeven, 1998). These
researchers installed a heat barrier sleeve around the portion of the shaft within the
bearing. This sleeve had a much lower thermal conductivity than the shaft and partially
insulated the shaft from the heat generated within the bearing. Consequently, the
temperature difference across the shaft was reduced and the influence of the Morton
Recently, there has been great interest in the Morton Effect. Some researchers
(Gomiciaga and Keogh, 1999) have used CFD techniques to obtain a better estimate of
9
the journal temperature distribution. Others, such as Larsson (Larsson I and II, 1999),
have extended the work of Keogh and Morton to include tilting pad journal bearings.
studied and is better understood than the Morton Effect. In addition, the Morton Effect
has ignited an abundance of interest in the past few years. As a result, there is a greater
precedence for current research to be directed at the Morton Effect than at the Newkirk
Effect.
So far, most of the analytical approaches to the Morton Effect have been identical
develop a simpler and more user-friendly model which can easily be adapted to become a
design tool for industrial applications. State-of-the-art research has also failed to make
adequate comparisons between theoretical models and experimental data. Hence, there is
room for research in this particular area. Furthermore, the relative novelty of the Morton
Effect means that there are still unanswered questions concerning its mechanism. For
instance, Why is the overhung configuration necessary? Does the orbit shape affect the
mechanism? These and other questions give further incentive for more research to be
10
1.3. Research Objectives
The objectives of this research are as follows:
(2) Use the model to obtain a better understanding of the Morton Effect.
(3) Compare the model with other theoretical research and experimental data.
analyses (Keogh and Morton, 1994; Larsson I and II, 1999), and the typical outline of
synchronous circumferential
Initial Bend temperature Thermal Bend
orbit
+ distribution
Dynamic Thermal
Σ
Fluid Film
Rotor Model Bearing Model Rotor Model
+
Researchers
In these models, an initial deformed rotor configuration is assumed and is then used as an
input in a dynamic model of the rotor system. This dynamic model produces a
synchronous orbit, i.e. the rate at which the shaft rotates about its own center is equal to
the frequency at which the shaft center rotates about the bearing center. Such an orbit
11
the bearings. A thermal bend, which depends on the temperature distribution and the
rotor thermal characteristics, is subsequently initiated. This thermal bend affects the rotor
dynamics and gives rise to positive feedback behavior. The stability of the resulting
feedback transfer function can then be analyzed by using Nyquist plots or another
equivalent method.
The initial mechanical and thermal bends, which are used in the models by other
researchers, can be related to unbalances. An unbalance arises when some of the rotor
mass lies off the axis of rotation. As a result, a centrifugal force, which is equal to the
product of the unbalance (mass x distance b/w mass and rotation axis) and the square of
the angular speed, is created. This unbalance affects the response of the rotor system and
therefore determines the size of the synchronous orbit. For a perfectly balanced rotor
(unbalance = 0), there is no orbit and the shaft simply rotates at the static equilibrium
position. The main point of this discussion is that the bends in the previous models can be
vibration, or unbalance, threshold levels which are given in standards published by various
organizations. The American Petroleum Institute (API) is one such organization and, in
1979, it provided API standard 617 for centrifugal compressors (American Petroleum
Institute, 1979). This standard said that, after final balancing, the maximum allowable
residual unbalance should be equal to the unbalance caused by a force which is 10% of the
journal static load. In other words, the maximum allowable centrifugal force on the
compressor is 10% of the static load. Such a criterion determines whether the rotor will
be “stable” or “unstable” during operation, i.e. if the unbalance level exceeds the threshold
12
level given by the 10% criterion then the rotor will be “unstable” and will have to be re-
balanced. This notion of practical stability will be utilized in the current model.
The current model for the Morton Effect is shown in Figure 1. 4 and it first
This unbalance is then put into the dynamic rotor model to obtain the synchronous orbit.
A fluid film bearing model is then used to establish the temperature distribution which
Figure 1. 3 and Figure 1. 4 illustrate two different models for the Morton Effect
Number of Calculations several per rotor speed one per rotor speed
13
The most fundamental difference between the two models is that prior representations
transform all equations to the frequency domain, while the current model operates in the
time domain. Hence, the current model avoids the complications encountered when
A time domain model also allows a simpler instability criterion -- the resultant
unbalance has to exceed some threshold value -- to be employed. However, other models
have to use the more complicated Nyquist stability theorem. In addition, the steady-state
nature and the fewer calculations of the current model make it less computationally
intensive than the other models. Previous models have to update the thermally-deformed
rotor shape after each successive iteration. As a result, these models tend to involve more
greater detail. Once a method for estimating the initial mechanical unbalance has been
dynamic orbit on the static equilibrium position of the journal. Obtaining this equilibrium
position requires a knowledge of the lubricant film thickness which depends on the
geometry of the bearings supporting the rotor. The dynamic orbit is then calculated using
a rotordynamics software package that is currently used in both the classroom and
industry. This information can then be used in conjunction with an energy equation to
obtain the circumferential temperature distribution (within the bearings) around the shaft
thermal unbalance. The resultant of the thermal and the initial mechanical unbalances can
14
then be determined and compared with a threshold value. If the resultant unbalance is
greater than the threshold value then the system will be unstable and the Morton Effect
4. Use VT-FAST to get the dynamic orbit. This orbit is then combined with the
Like previous research, the current model will focus on rotors supported by plain journal
bearings (Keogh and Morton, 1994; Faulkner et al., 1997) and tilting pad journal bearings
(deJongh and Morton, 1994; de Jongh and van der Hoeven, 1998). The simpler plain
15
2. The Morton Effect in Rotors with Plain
Journal Bearings
The development of the Morton Effect model for rotors supported by plain journal
deformation. As a result, the amount and the location of the mechanical unbalance is very
unbalance (Um) will be assumed. This mechanical unbalance will be defined as the
unbalance created from a centrifugal force equal to 10% of the total static rotor weight
(W). Um will occur when the rotor is running at its maximum continuous operating speed
01
.W
eq2. 1 Um =
ω 2MCOS
This unbalance will act at an angle of zero degrees with respect to a coordinate
system on the rotor, and will be located at the center-of-gravity of the overhung mass.
16
2.2. Film Thickness for Plain Journal Bearings
The film thickness expression for a plain journal bearing has been derived
extensively in the literature. The following derivation is based on work done by Cameron
(Figure 2. 1) which showed the shaft or journal center (Oj) displaced from the bearing
Lubricant
Line-of-centers B
Rb δ
h
Ob
Rj
e
ξ
Oj
Journal
ψa
17
Applying the sine rule to the triangle ObOjB shown in Figure 2. 1 gives:
R
sin(ξ)
e e
eq2. 2 = b ⇒ sin(δ) =
sin(δ) sin ξ Rb
where Rb = bearing radius and ξ = circumferential angle between the film thickness line
and the line-of-centers. The above equation can be used with a trigonometric identity to
yield:
e2
eq2. 3 cos(δ) = 1− sin 2 (δ) = 1− sin 2 (ξ)
Rb 2
This means that cos(δ) is approximately equal to 1 because e2/Rb2 << 1 (typical values are
e = 2 mils or 50.8 µm and Rb = 2 inches or 5.08 cm, which gives e2/Rb2 = 10-6. The
maximum value of sin2(θ) is unity). The approximation for cos(δ) can be used in
conjunction with Figure 2. 1 to obtain an expression for the film thickness, h. From the
triangle in Figure 2. 1:
Defining the bearing clearance as Cb = Rb-Rj and the eccentricity ratio ε = e/Cb enables the
Therefore,
eq2. 6 [
h = C 1+ ε cos(ξ)
b ]
18
2.3. Static Equilibrium Journal Position
The film thickness expression (eq2. 6) can be used to obtain the static eccentricity
ratio (ε0 - eccentricity ratio before elliptical orbit is superimposed) and the attitude angle
(ψa0 - angle between vertical static bearing load and the line-of-centers) which are needed
to establish the static equilibrium journal position. These parameters can be derived by
solving a reduced form of the Reynolds equation for the lubricant pressure distribution,
and subsequently integrating this pressure distribution over the journal surface area within
the bearing. The resultant force must balance the bearing load, Wb, so that static
These derivations have been done in detail by Cameron (Cameron, 1966) who
Wb C 2b π 2 ε 20 ε 40
eq2. 7 = +
3
µωR j L3 2
161− ε 0
24
1− ε 0
Equation eq2. 7 is valid for a narrow plain journal bearing, which means that the bearing
length, L, is small compared to the journal diameter, 2Rj. Furthermore, the lubricant
angular speed, ω. Unfortunately, the value of the eccentricity ratio is needed to obtain this
An 8th order polynomial can be obtained from the simplification of eq2. 7 and then
solved for the eccentricity ratio, ε0. The solution is presented within the computer code in
19
the appendices. Cameron (Cameron, 1966) also gives an expression for the static attitude
2
π 1− ε 0
eq2. 8 tan ψ =
a0 4 ε
0
By using equations eq2. 7 and eq2. 8, the static equilibrium position of the journal in the
equilibrium position, Oj0 (see Figure 2. 2). The rate of whirl must equal the rate of
rotation in order for the orbit to be described as synchronous. Once this configuration is
achieved, a general point, P, on the surface of the journal will also travel along an elliptical
path. If the orbit was non-synchronous then the path taken by P would be an elliptical
With reference to Figure 2. 2, the path taken by the general point (P(x,y)) on the
( )
(a) x = e cos θ j0 + A cos ωt + φ x + R cos(ωt + λ )
0 x j
eq2. 9
(b) y = e sin θ j0 + A sin ωt + φ y + R sin(ωt + λ )
0 y j
The geometry in Figure 2. 2 can be related to that of Figure 2. 1 by using the following
relationship:
20
eq2. 10 ξ = π − ωt + λ − θ j
where,
e 0 sin θ j0 + A y sin ωt + φ y
eq2. 11 θ = tan − 1
j e cos θ + A cos ωt + φ
0 j0 x ( )
x
and
3π
eq2. 12 θ = +ψ
j0 2 a0
21
Y
P(x,y)
Journal Rj
Oj ωt+λ Bearing Sleeve
Aysin(ωt+φy)
e
ξ
Oj0 γ
e0 Axcos(ωt+φx)
θj
θj0
Ob
X
Elliptic Orbit
Centered at Oj0
From the geometry in Figure 2. 2, the location of the hot and cold spots can be
determined. The hot spot (H) will be defined as the point closest to the bearing sleeve at
time t = 0. In this configuration, it is also assumed that γ = ωt+λ so that Oj0, Oj and P=H
A
y sin φ y
eq2. 13 γ = γ 0 = tan −1 =λ
A x cos φ x
22
The time-dependent locations of the hot spot (H(xH,YH)) and the diametrically-opposite
cold spot (C(xC,yC)) can now be obtained by using the definition of γ0:
(a ) x
H 0 x ( j ) (
= e cos θ j0 + A cos ωt + φ x + R cos ωt + γ 0 )
eq2. 14
(b) y = e sin θ j0 + A sin ωt + φ y + R sin(ωt + γ 0 )
H 0 y j
(a ) x
C 0 x ( j ) (
= e cos θ j0 + A cos ωt + φ x + R cos ωt + γ 0 + π )
eq2. 15
(a ) y = e sin θ j0 + A sin ωt + φ y + R sin(ωt + γ 0 + π )
C 0 y j
However, for plain journal bearings, calculations are usually done in the ξ-coordinate
system and it is necessary to use eq2. 10 to obtain the locations of the hot and cold spots
in this reference:
It should be noted that the parameters Ax, Ay, φx and φy can be obtained directly
from the forced response program in VT-FAST (Virginia Tech - Front-end Automated
currently used in both the classroom and industry. The transfer matrix option in VT-
FAST was chosen to execute the forced response program, and the required stiffness and
damping coefficients for plain journal bearings were obtained from the literature (Vance,
1988):
23
Wb 2
(a ) K = 4 2π + 16 − π 2 ε 2 Q(ε)
xx Cb
2 2 2 4
Wb π − π + 2π ε + 16 − π ε Q( ε)
2
(b) K xy = −
Cb 1/ 2
ε1− ε 2
2 2 2 2 4
Wb π π + 32 + π ε + 216 − π ε Q( ε)
(c) K yx = −
Cb 1/ 2
ε1− ε 2
2 2 2 2 4
Wb 4 π + 32 + π ε + 216 − π ε Q( ε )
(d ) K yy =
Cb 1− ε 2
The complete synchronous dynamic orbit and the location of the hot and cold spots can
now be obtained.
at the journal surface within the bearing. In order to obtain this temperature distribution,
an energy equation must first be formulated and solved. The physical domain, over which
24
E& b
Bearing
E& visc
Lubricant
ωRj
E& j
Journal
Figure 2. 3 represents a simple model for energy flow in a plain journal bearing. It
assumes negligible axial heat flow and steady-state conditions within the control box. It
should be noted that the control box encompasses a portion of the journal surface. Hence,
the temperature within the box will be the same as the circumferential journal surface
temperature. Another key assumption in this model is that the final temperature of the
journal is obtained by averaging the distributions from each of the dynamic positions. As a
result, the average journal temperature distribution can be obtained by considering several
25
Conservation of energy can generally be stated as:
The steady-state assumption means that the accumulation rate is zero. Therefore,
eq2. 19 E& visc = E& lub x + dx − E& lub x + E& j + E& b + 0
ωR j dT ωR j
ωR τdxdz = ρ hdz c T + dx − ρ hdz cT
eq2. 20 j l 2 l dx l 2 l
( )
+ fHdxdz T − Tamb + (1− f ) Hdxdz T − Tamb ( )
The rate of viscous energy dissipation, E& visc , depends on the speed of the journal surface
and the viscous shear stress, τ, which acts over an area dxdz (z = axial dimension). This
energy source term heats up the advecting lubricant which has a density of ρl and a
specific heat capacity of cl. A fraction, f, of the remaining heat is transferred to the journal
while the rest is assumed to be lost to the bearing housing and surroundings. It is also
assumed that the total heat loss is equal to Hdxdz(T-Tamb), where H = heat transfer
coefficient and Tamb = average ambient journal temperature = average ambient bearing
ωR j
eq2. 21 ωR j τ = ρ l h
2
cl
dT
dx
(
+ H T − Tamb )
If the lubricant is Newtonian, then the shear stress will be given by τ = µdu/dy.
Furthermore, a linear velocity profile can be assumed, i.e. the shear strain rate du/dy =
26
ωRj/h. This Petroff type of simplification has been found to give reasonable accuracy
(Cameron, 1966). It should also be noted that since h << Rj (journal radius), x ~ Rjξ. The
dT 2H 2 HT 2µωR 2j
eq2. 22 + T− amb + =0
dξ ρ l c l ωh ρ l c l ωh ρ c h 2
l l
At this point, the temperature rise above the ambient temperature can be defined as
~
( )
T = T − Tamb . Once these substitutions are used, equation eq2. 22 can be rearranged to
give :
2
2 H ~ 2µωR j
~
dT
eq2. 23 + T− =0
dξ ρ l c l ωh ρl c l h 2
result, an effort has been made to derive an effective lubricant viscosity (µ) that can be
used in all equations. The first step was to replace Cb with h, and substitute the known
lubricant supply viscosity (µ0) for µ, in equation eq2. 23. Hence, an approximate 1D
~
dT ~
+ AT − B = 0
dξ
eq2. 24
2H 2µ 0 ωR 2j
A= B=
ρ l c l ωC b ρ l c l C 2b
27
With the boundary condition T( ξ= 0) = T = T − T
~ ~
(T0 = lubricant supply
0 0 amb
B ~ B
eq2. 25 ()
T ξ = + T0 − e − Aξ
~
A A
Now, an average temperature rise (∆Tav) can be derived by assuming that this temperature
B ~
∆Tav = T( π) − T0 ≈ Tmax − T0 = − T0 1− e − Aπ
~ ~
eq2. 26
A
The effective viscosity may be defined according to the exponential Reynolds formulation:
− β∆Tav
eq2. 27 µ = µ 0e
where µ0 is the lubricant viscosity at the supply temperature, T0, and β = thermoviscosity
coefficient. Some researchers have indicated that this expression is not extremely accurate
(Cameron, 1966; Williams, 1994). However, it is simple and provides sufficient accuracy
Equation eq2. 23 can now be solved numerically (see the Matlab program in the
appendices). The required boundary condition sets the temperature at maximum film
thickness (the lowest temperature) to the value of the lubricant supply temperature, i.e.
~
( ~
)
T h max ≡ T ξ = 0 = T0 . This equation (eq2. 23) is solved at several different points in the
~
dynamic synchronous orbit, and the temperature difference between the hot and the cold
spots are extracted from each solution. These temperature differences are then averaged
to give an overall mean temperature difference (∆T) between the hot and cold spots. The
28
resultant ∆T will be used to compute the thermal unbalance generated by the Morton
Effect.
Figure 2. 4. The static bending moment (M) in the journal can be expressed as:
dψ
eq2. 28 M = EI
dz
where E = Young’s Modulus, I = area moment of inertia and y = bend angle as shown in
Figure 2. 4.
29
Ld
L/2 L/2
hot spot
Journal z 2Rj
y
cold spot
Overhung Mass
Bearing
hot spot
yb
Journal
yd1 yd
cold spot ψb
Bearing
Overhung Mass
30
The maximum stress (smax) in the shaft can be written as:
R jM
eq2. 29 σ max =
I
where a = thermal conductivity of the journal and ∆T is the overall mean temperature
difference between the hot and the cold spots. It is assumed that ∆T is independent of
axial position within the bearing. Furthermore, the small deflections allow the thermal and
Equation eq2. 28 can be substituted into eq2. 29, and eq2. 29 and eq2. 30 can be
equated to give:
dψ α∆T
eq2. 31 =
dz R j
For a given journal speed and orbit, the right hand side of equation eq2. 31 is essentially
constant. Therefore, eq2. 31 can be integrated between the start of the bearing -- where it
is assumed that the bend angle is zero -- and some arbitrary axial distance, z (see Figure 2.
4), to give:
dy α∆T
eq2. 32 ψ= = z
dz R j
This equation can be evaluated at the end of the bearing to give the bend angle at this
location:
31
α∆T
eq2. 33 ψb = L
Rj
Also, eq2. 32 can be integrated over the bearing to give the deflection (yb) at the end of
the bearing:
1 α∆T 2
eq2. 34 yb = L
2 Rj
The additional deflection (yd1) at the overhung mass location can then be calculated using
L L
eq2. 35 y
d1 d 2 ( )
= L − sin ψ b ≈ L d − ψ
2 b
By substituting eq2. 33 into eq2. 35 and by adding up eq2. 34 and eq2. 35, the total
α∆T
eq2. 36 yd = LL d
Rj
Equation eq2. 36 gives us the thermal deflection of the overhung mass. This deflection
eq2. 37 U t = md y d
32
2.7. Instability Criterion
The resultant unbalance (U) from Um and Ut can be represented as shown in Figure
2. 5.
Um
U Line joining
ωt cold spot
to hot spot
θCH
X
Ut
With reference to Figure 2. 5, the mechanical and thermal unbalances can be added
vectorially to produce the resultant unbalance (U) which can be represented as follows:
eq2. 38 (
U = U 2t + U 2m − 2U t U m cos ωt − θ CH )
The angle (ωt-θCH) is actually calculated by averaging similar angles at several different
defined. This unbalance was assumed to be caused by a force equal to 15% of the rotor
33
015
. W
eq2. 39 U =
thr
ω2
where W = rotor weight and ω = variable angular journal speed. The value of 15% was
chosen because it gave the best results with the cases studied (see section 2.10).
The rotor will be unstable whenever U exceeds Uthr,. As a result, the threshold
speed for instability (ωthr) occurs when U = Uthr. This instability criterion can be obtained
graphically from the intersection of the U vs. ω and the Uthr vs. ω curves, and it determines
journal bearing cases. At the beginning of the program, the rotor speed is updated and the
equilibrium position. The mechanical unbalance (Um) is next used in VT-FAST to get the
dynamic orbit which is combined with the static position to obtain the total synchronous
orbit. The temperature difference between the hot spot and the cold spot is then
calculated at equal time intervals in the orbit. By averaging these temperature differences,
the overall mean temperature difference (∆T) between the hot spot and the cold spot is
acquired. Next, the unbalances (Ut, U and Uthr) are computed and the process is repeated
for each rotor speed. U and Uthr can then be compared over the speed range to determine
It should be noted that two programs were written. A Matlab Program (Appendix
A) is to be used with rotors supported by plain journal bearings, while a Fortran program
34
(Appendix B) is to be used with rotors supported by tilting pad journal bearings. The
graphics for the Fortran program is handled by the Matlab code in Appendix C.
START
Speed
Effective Viscosity
Thot-Tcold
NO Finished #
of avg. ?
YES
∆T, Unbalances
NO Finished #
of spd. ?
YES
Plot U, Uthr vs. Speed
END
Figure 2. 6 - Flowchart for Plain and Tilting Pad Journal Bearing Programs
35
2.9. Case Studies
This rotor consists of a shaft with a step variation in diameter. The shaft is supported by
two identical plain journal bearings and a disk of constant mass is placed, at an overhung
position, on either shaft end. The pertinent data for this rotor is summarized in Table 2. 1.
36
Table 2. 1 - Data for Keogh and Morton rotor
Bearing Properties
Length L 0.035 m 1.38 in
Radial clearance Cb 1.00E-04 m 3.94 mils
Heat transfer coefficient H 50 W/m2/oC 2.43e-5 hp/in2/oF
Bearing load Wb 2500 N 562 lbf
Journal radius Rj 0.050 m 1.97 in
α
o -1 o -1
Journal C.T.E 1.10E-05 C 6.11E-06 F
Rotor Properties
Rotor weight W 5000 N 1124 lbf
Overhung mass md 32.3 kg 71.3 lbm
Overhang Distance Ld 0.16 m 6.4 in
Max. Cont. Op. Speed ωMCOS 1047 rad/s 10000 RPM
Initial Mech. Unbalance Um 4.54e-4 kg m 0.63 oz in
The Keogh and Morton model examined the stability of a complex rotor thermal bend
angle. From this complex analysis, they were able to calculate the eigenvalues of the rotor
system and determine the corresponding growth factors (σ = real part of eigenvalue).
Since the system response depends on est, a positive growth factor would indicate a
theoretical instability. The growth factor plot for the Keogh and Morton case is shown in
Figure 2. 8.
37
Figure 2. 8 - Growth Factor Plot for Rotor (Keogh and Morton, 1994)
From Figure 2. 8, Keogh and Morton concluded that the critical speed range, in which
rotor thermal bending can occur, is between 1023 rad/s (9769 RPM) and 1086 rad/s
(10371 RPM).
The current analysis used the data in Table 2. 1 to predict a synchronous thermal
instability range from 1047 rad/s (10001 RPM) to 1206 rad/s (11521 RPM) which
overlaps with the Keogh and Morton range. In addition, the current model indicates
another range of instability after 1286 rad/s (12277 RPM). These instability ranges were
obtained from the intersection of the resultant unbalance and threshold unbalance curves
shown in Figure 2. 9.
38
Figure 2. 9 - Unbalance Curves from Current Analysis of Keogh and Morton Rotor
The agreement of the current model with the Keogh and Morton data is not excellent, but
it is still acceptable. One reason for some discrepancy could be the difficulty encountered
in fitting the Keogh and Morton viscosity information with an exponential profile. In
some instances (particularly at higher speeds) the viscosity values from the curve fit were
greater than the given data. This disagreement could have contributed to higher
temperatures and larger thermal unbalances at the elevated speeds. As a result, the
predicted instability range would be more prolific than the actual one.
The driving force behind the instability hump in Figure 2. 9 is the high overall mean
temperature (∆T) between the hot and cold spots. As mentioned above, higher viscosity
values could have contributed to these large ∆T values. However, the rotor dynamics
39
plays the more important role in generating these high ∆Ts which lead to greater thermal
and resultant unbalances. Figure 2. 10 shows that the rotor has a critical speed (4th
Figure 2. 10 - Forced Response Plots for Keogh and Morton Rotor (Unb. = 1 oz in)
This lateral critical speed creates large amplitude orbits in this region. One such orbit (at
40
Figure 2. 11 - Synchronous Orbit of the Keogh and Morton Rotor at 10505 RPM
The lower left and the upper right corners of this orbit show that, in these regions, the
rotor is running very close to the radial bearing clearance, Cb. This configuration implies
that the film thickness at the hot spot (x) would be very low, while the diametrically-
opposite cold spot (o) would experience higher film thickness values. Since a lower film
thickness is associated with higher viscous dissipation and higher temperatures, a strong
thermal gradient will develop across the journal. Such a gradient will cause the ∆T value
On the other hand, lower amplitude orbits (see Figure 2. 12), which are further
away from the critical speed, keep the rotor away from the clearance circle and are less
likely to be associated with high ∆T values. Hence, the orbit shown in Figure 2. 12 would
41
not give rise to a thermal bending instability. The lower speed of the 5730 RPM orbit also
reduces the amount of viscous dissipation and this further stabilizes this orbit relative to
Figure 2. 12 - Synchronous Orbit of the Keogh and Morton Rotor at 5730 RPM
42
2.9.2. Faulkner, Strong and Kirk
Faulkner, Strong and Kirk (Faulkner et al., 1997) did an experimental study of a
Compressor
Turbine
Impeller
Disk
The turbocharger had a centrifugal compressor impeller at one end and a radial inflow
turbine disk at the other end. This machine was supported by two 3 axial-groove journal
bearings which consist of a plain journal bearing with 3 small grooves cut along the
bearing length. For simplicity, this type of bearing will be approximated as a plain journal
bearing. Other relevant information on the turbine end of this turbocharger is shown in
Table 2. 2.
43
Table 2. 2 - Data for Turbine End of Faulkner, Strong and Kirk Turbocharger
Bearing Properties
Length L 0.057 m 2.24 in
Radial clearance Cb 7.11e-5 m 2.8 mils
Heat transfer coefficient H default W/m2/oC default hp/in2/oF
Bearing load Wb 916 N 206 lbf
Journal radius Rj 0.041 m 1.63 in
α
o -1 o -1
Journal C.T.E 1.10E-05 C 6.11E-06 F
Rotor Properties
Rotor weight W 1877 N 422 lbf
Overhung mass md 61.7 kg 136 lbm
Overhang Distance Ld 0.20 m 7.71 in
Max. Cont. Op. Speed ωMCOS 1047 rad/s 10000 RPM
Initial Mech. Unbalance Um 1.73e-4 kg m 0.24 oz in
If the heat transfer coefficient is not known (as is the case with the turbocharger bearings)
− 0.4
eq2. 40 H = (05
. )(255
0.7
. ) ωR j (µ 0 )− 0.2 2πR j
Equation eq2. 40 is based on an expression that was derived for tilting pad journal
During operation, it was observed that the turbocharger became unstable near
around 9900 RPM. Faulkner, Strong and Kirk initially thought that the instability was due
to the turbine wheel becoming loose at the high operation speeds. However, a careful
inspection of the turbine wheel position, before and after operation, indicated that the
44
wheel did not move while the turbocharger was running. Furthermore, a damped critical
speed analysis failed to justify the existence of a lateral critical speed near 9900 RPM. It
was finally concluded that the source of the instability was the thermal bowing of the rotor
After using the current thermal instability model to analyze the turbocharger, it
was found that the thermal instability was predicted to occur around 1009 rad/s (9640
45
Unlike the Keogh and Morton case, the turbocharger does not encounter criticals with low
damping in the given speed range . As a result, there are no large amplitude orbits to
produce a sudden instability hump. The monotonic increase in U can be explained by the
increase in speed which leads to increased viscous dissipation, higher ∆T values and more
incentive for thermal bending. Figure 2. 15 and Figure 2. 16 illustrate this trend of higher
46
Figure 2. 16 - Synchronous Orbit from Turbocharger (Turbine End) at 11000 RPM
The figures above also show that the orbits become more centered at higher speeds, i.e.
the center of the elliptical orbit approaches the bearing center (x/Cb = 0, y/Cb = 0) as the
rotor speed increases. A more centered orbit means that the hot spot and the cold spot
experience almost constant film thickness values, and such a configuration causes a
relatively steady thermal gradient to develop. In fact, this gradient would be the most
steady when the orbit is a circle that is completely centered. The Keogh and Morton case
show that large amplitude orbits tend to produce high thermal gradients which favor the
Morton Effect. If these high gradients are combined with a steady orientation, the worst
case scenario for the Morton Effect is achieved; large-amplitude, circular and centered
47
orbits are the most likely to be associated with the Morton Effect. This scenario can be
Figure 2. 15 shows a 9000 RPM synchronous elliptical orbit at the turbine end of
the turbocharger. The resultant (U) and threshold (Uthr) unbalances for this speed and
configuration are 0.36 oz in and 0.44 oz in respectively. If this orbit is forced to center
(Figure 2. 17) then the resulting ∆T is drastically increased from 0.87 oF to 6.07 oF. The
resultant unbalance is now 1.09 oz in. which is greater than the threshold value of 0.44 oz
in. As a result, the turbocharger is now experiencing the Morton Effect at 9000 RPM.
Figure 2. 17 - Forced 9000 RPM Turbine End Orbit: The Ellipse is Centered
φx, the elliptical orbit can be changed into an equivalent circular one (Figure 2. 18). The
48
radius of the circular orbit is equal to the semi-major axis of the ellipse and the ∆T value is
Figure 2. 18 - Forced 9000 RPM Turbine End Orbit: The Centered Ellipse is
After increasing the diameter of the circular orbit by a factor of 10, Figure 2. 19 is
obtained. The ∆T value has risen to 10.85 oF and U is equal to 1.74 oz in. This entire
process shows that the Morton Effect becomes progressively worse as an orbit is
49
Figure 2. 19 - Forced 9000 RPM Turbine End Orbit: The Centered Circular Orbit is
50
2.10. Solution Sensitivity to the Threshold
Unbalance Force
The threshold unbalance force in the two cases studied above was 15% of the
rotor weight. Table 2. 3 shows the effect on the calculated instability speeds if different
Table 2. 3 - Predicted and Actual Instability Speed Values (RPM) for Different
>12277
It appears that a change of only 5% in the threshold unbalance force can change the
calculated instability speed by several hundred RPM. The solution is therefore sensitive to
the choice threshold unbalance force. From the data shown in Table 2. 3, the optimum
threshold unbalance force seems to be 15% of the rotor weight. This choice of unbalance
force is the one which gives data that most corresponds with the Keogh and Morton
instability range. In addition, this force magnitude predicts an instability speed that lies
51
close to the observed thermal instability speed at the turbine end of the Faulkner, Strong
and Kirk turbocharger. As a result, the 15% unbalance force level was utilized in the
52
3. The Morton Effect in Rotors with Tilting
Pad Journal Bearings
The procedure for developing a Morton Effect model for rotors with tilting pad
journal bearings (TPJB) is similar to that for plain journal bearings. In fact, the steps for
calculating the initial mechanical unbalance (Um), the thermal unbalance (Ut) and the
instability threshold unbalance (Uthr), are identical. In addition, the computer flowchart,
which is given in chapter 2, is also applicable to rotors supported by tilting pad journal
Tilting pad journal bearings are generally more stable than plain journal bearings
because the tilting pads theoretically set the bearing cross-coupled stiffness coefficients to
zero (Kirk, 1998). This property minimizes the tangential forces acting on the shaft and
thus stabilizes the system. The fixed geometry of the plain journal bearings does not allow
the cross-coupled stiffness coefficients to be set to zero and, as a result, plain journal
However, the presence of tilting pads, which are free to move independently of
each other, increases the complexity of the modeling. A numerical solution will therefore
have to be adopted. Furthermore, the film thickness expressions for TPJBs have not been
explicitly derived in any current work of literature. An attempt will be made to develop
53
film thickness expressions for the cylindrical pivot TPJBs and extend the modeling to
Lubricant
Journal
Pad
Pivot
LOP LBP
Regardless of the configuration, a single pad of a TPJB with a cylindrical pivot can be
illustrated as shown in Figure 3. 2. Most current TPJBs carry 4 or 5 pads (Figure 3. 1),
but only one pad is shown in Figure 3. 2 in order to avoid congesting the drawing.
54
Y
bearing
M
tp
δ
N
h
J pad
Rb
Rp
lubricant θc
θ
Ob θp
X
e
Op Oj
journal
Op’
55
This pad is depicted before (dotted line) and after (solid line) tilting it through an angle δ.
The center of the bearing is represented by Ob and the center of the journal is Oj . Points
Op and Op’ respectively refer to the pre-tilt and post-tilt positions of the pad center of
curvature. Rb corresponds to the bearing radius while Rj denotes the journal radius. The
distance between the pivot point, M, and the surface of the pad is tp, and the pad radius of
coordinate system with origin Ob. θc represents the angle to the line-of-centers which
connects Ob to Oj. The distance between Ob and Oj is the journal eccentricity, e. The film
thickness, h, is measured at an angle θ with respect to the given coordinate system. This
circumferential angle, θ, is different from the angle ξ defined for the plain journal bearing.
In the plain journal bearing case, ξ was defined relative to the moving line-of-centers, with
the clockwise direction being positive. On the other hand, the TPJB case defines θ with
being positive. This new definition enables a more conventional coordinate system to be
used which simplifies the numerical analysis. However, it can be shown that if the line-of-
centers position is taken into account, and the pad effects are neglected, the TPJB film
thickness expression can be reduced to give an equivalent plain journal bearing one.
56
M
δ
Y
Rb+tp
Rp+tp
(θp+δ) N
θp
Ob γ X Y ObJ+h λ
J r ξ
Op’
ο Rp
π−(θc−θ) θ
Rj Ob γ X
r κ
Ob Op’
e ι
Oj
Rj e O J
eq3. 1 = = b
[
sin π − (θc − θ) ] sin ο sin ι
eq3. 2 ο = sin − 1
(
e sin θ c − θ )
Rj
57
− 1 e sin θ c − θ
eq3. 3 ι = π − π − θ c − θ + ο = θ c − θ − sin
Rj
Hence,
O J=
R j sin ι
=
Rj
−
sinθ c − θ − sin
1
e sin θ − θ
c ( )
eq3. 4
b
( )
sin θ c − θ sin θ c − θ (
) R j
Since e << Rj, the sin-1 term approximates to its argument. Therefore, eq3. 4 can be
Rj e sin θ − θ
( )
eq3. 5 ObJ =
(
sin θ c − θ − ) c
(
sin θ c − θ
) R j
Rj e sin θ − θ
( ) − cos θ − θ sin e sin(θ c − θ)
eq3. 6 O b J =
( )
sin θ c − θ cos
c
( c ) R
( )
sin θ c − θ
Rj
j
Applying the small angle approximations cos(x) ~ 1 and sin(x) ~ x to eq3. 6 gives the
eq3. 7 O b J = R j − e cos θ c − θ ( )
The cosine rule can be used on Figure 3. 3b to give:
1
2 2 2
eq3. 8 r = R p + t p + R b + t p − 2 R p + t p R b + t p cosδ
58
Typical values for the parameters in eq3. 8 are tp = 0.5 in. (1.27 cm), Rp = 2.006 in. (5.095
cm), Rb = 2.003 in. (5.088 cm) and δ = 0.05 degrees. These values give r = 0.0037 in.
(94.3 µm) or r/Rp = 0.0018 which is much less than 1. From Figure 3. 3c:
Rp r O N
eq3. 9 = = b
sin(θ + γ ) sin λ sin κ
(
r sin θ + γ
)
≈ r sin θ + γ( )
eq3. 10 λ = sin −1
Rp
Rp
Also,
r sin(θ + γ )
eq3. 11 κ = π−θ− γ −λ = π−θ− γ −
Rp
ObN =
R p sin κ
=
Rp
sinπ − θ − γ −
r sin θ + γ ( )
eq3. 12
( )
sin θ + γ sin θ + γ
( ) Rp
Since sin(π-x) = sin(x), this equation can be re-written and expanded to give:
Rp
r sin θ + γ ( )
r sin θ + γ( )
eq3. 13 ObN = (
sin θ + γ cos ) + cos (
θ + γ )
sin
( )
sin θ + γ
Rp
Rp
eq3. 14 O b N = R p + r cos(θ + γ )
59
Figure 3. 3b indicates that γ = π − θ p − δ − ξ . If this expression for γ is substituted into
r Rb +tp
eq3. 16 = ⇒ r sin ξ = R b + t p sin δ
sin δ sin ξ
Therefore,
1
1/ 2 2
eq3. 17 r cos ξ = r 1− sin 2 ξ
( )
= r − Rb + tp 2 sin 2 δ 2
eq3. 19 r cos(δ + ξ) = r cos δ cos ξ − r sin δ sin ξ = R p + t p cos δ − R b + t p
Since δ is very small (~ 0.05 deg or 8.73e-4 rad), sin(δ) ~ δ, and cos(δ) ~ 1. Therefore,
eq3. 21 r cos(δ + ξ) = R p − R b
60
eq3. 22 r sin(δ + ξ) = δ R p + t p
From Figure 3. 3c, it can be deduced that the film thickness is:
eq3. 24 h =O b N −ObJ
Combining eq3. 7, eq3. 23 and eq3. 24 gives an expression for the film thickness of a
eq3. 25 ( )
h = R p − R j + e cos θ c − θ − R p − R b cos θ − θ p − δ R p + t p sin θ − θ p
Other researchers have published expressions for the film thickness of a cylindrical pivot
TPJB and some of these formulae are listed below (the symbols have been changed to
Monmousseau et al., 1997 (the pad deformation effects have been neglected):
eq3. 26 h = C p − C p m cos θ − θ p − R j + t p δ sin θ − θ p + e cos θ c cos θ + e sin θ c sin θ
Ha et al., 1995:
eq3. 28 h = C p + e cos θ c cos θ + e sin θ c sin θ − C p − C b cos θ − θ p − δR j sin θ − θ p
61
where the radial pad clearance, Cp = Rp-Rj, the radial bearing clearance, Cb = Rb-Rj and the
preload factor, m = 1 - Cb/Cp. The above equations can be compared by using some
PARAMETER VALUE
The fractional angular position of the pad pivot, fp = (distance between the leading edge of
the pad and the pivot )/(total angular length of the pad, ∆p). The data in Table 3. 1 was
used to generate Figure 3. 4. This figure shows that h from the current research compares
fairly well with the film thickness equations from the other studies. In fact, along with
Monmouseau et al., this research gives the most conservative estimate of the film
thickness, i.e. the lowest values of h. This is because the current research and the work
62
done by Monmousseau take into account the pad thickness, tp. This additional
consideration causes the film thickness to be lower, especially at pad 4 which is the most
heavily loaded pad. As a result, the current research allows for a safer design which
would be less likely to suffer from the journal destroying the pad babbitt due to an
Current Research
Monmousseau et al.
6 Ha et al.
Kim et al.
5
Film Thickness, h (mils)
0
0 50 100 150 200 250 300 350 400
theta (degrees)
Pivot TPJB
an equivalent expression for the film thickness of a plain journal bearing. In other words,
63
eq3. 25 consists of the plain journal bearing film thickness ( (Rb - Rj) + ecos(θc-θ) = Cb[1+
Some TPJBs contain spherical pivots which have more degrees of freedom than
cylindrical pivots. However, any pad-tilting about the x-axis or y-axis usually leads to
journal misalignment which could create an unstable unbalance response (Vance, 1988).
Therefore, it will be assumed that the spherical pivots only tilt about the z-axis and will
thus have the same film thickness expression as the cylindrical pivot TPJBs.
thickness expression make it necessary for a numerical solution to be used to find the
journal position. At the inception of this project, it was thought that the complicated
geometry of the TPJB domain would be difficult to model accurately using the rectangular
grid of the finite difference method. As a result, the finite element method was chosen to
deal with this complex geometry. In retrospect, however, finite difference techniques
distribution must first be obtained by numerically solving the Reynolds’ equation. This
pressure must then be used to balance the forces on the journal and the moments on each
pad in the TPJB so that the static equilibrium position of the journal can be determined.
64
3.2.1. Reynolds’ Equation
A local coordinate system can be set up in the region between the pad and the
Pad Surface
h
Lubricant
y
In 1886, Osborne Reynolds derived the following equation from the Navier-Stokes
∂ h 3 ∂P ∂ h 3 ∂P ∂h
eq3. 29 + = 6u j
∂x µ ∂x ∂z µ ∂z ∂x
where P = pressure and µ = lubricant viscosity. The derivation of equation eq3. 29 has
also been done by Cameron (Cameron, 1966) and will not be repeated here. Instead, the
65
assumptions made in the derivation of equation eq3. 29, and their implications to this
The following assumptions (Cameron, 1966) were made in order to derive the
Reynold’s equation:
(1) Body forces are neglected. This means that gravitational, magnetic and electrical
forces are assumed not to be present. Such an assumption is true for most of the
(2) Pressure is constant through the film thickness. As a result, the pressure gradient in
the y-direction is zero and no lubricant motion is assumed to occur in this direction. If
the lubricant is highly compressible, e.g. a gas, then this assumption will not be true.
(3) Bearing surface curvatures >> h. Displacement in the x-direction can therefore be
Table 3. 1.
(5) Newtonian lubricant. A Newtonian lubricant has a shear stress (τ) which is equal to
the shear strain rate (e.g. du/dy) multiplied by the lubricant viscosity (µ).
(6) Laminar flow. The Reynolds number (p. 441, Cameron, 1966) for bearings can be
defined as Re = ωRjCbρl/µ and the critical Reynolds’ number which demarcates the
10,000 RPM (1047 rad/s), Rj = 2 in. (5.08 cm), Cb = 3 mils (76.2 µm), ρl = 0.031
lbm/in3 (860 kg/m3), µ = 2.0 µreyn (13.8 cP or 13.8 mPa s) are used then Re = 253 and
66
Rec = 1066. Therefore, most continuous fluid film bearings typically operate under
(7) Neglect fluid inertia. This assumption allows the mass x acceleration term in the
gradients can be equated to provide a starting point for the development of the
Reynolds’ equation.
(8) Constant viscosity across the film thickness. This simplifies cross-film integration
to give the pressure distribution. A typical meshed pad surface with essential pressure
z (axial)
Mesh of
Quadrilateral
θ,x (circumferential) Elements
z = +L/2, P = P0
θ=
θ=
θ e,
θ s,
P=
P=
P0
P0
z = -L/2, P = P0
67
where θs = start angle of the pad, θe = end angle of the pad and L = axial pad length. The
above domain can be meshed using quadrilateral elements which will be discussed in detail
shortly. It should be noted that the h-version (instead of the p-version) of the finite
element method will be chosen for simplicity. In other words, lower order basis or
interpolation functions will be used and convergence will be achieved by refining the mesh.
The p-version uses a given number of elements and achieves convergence by increasing
the order of the polynomials used as basis functions. Generally, the p-version converges
In the finite element method, basis functions are used to approximate the exact
solution to a differential equation. These basis functions are defined over simple elements
which comprise a mesh of the particular domain. Accompanying these basis functions are
unknown coefficients which can be obtained by setting the weighted integral of a residual
to be equal to zero (Reddy, 1993). The residual is obtained from the original governing
3 3
∂ h ∂P ∂ h ∂P ∂h
eq3. 30 ∫∫ w + − 6u j dA = 0
∂
element x µ ∂ x ∂ z µ ∂z ∂ x
∂f ∂w ∂( wf ) ∂w
eq3. 31 ∫∫ w dA = − ∫∫ f dA + ∫∫ dA =− ∫∫ f dA + ∫ wfn x dl
∂x ∂x ∂x ∂x
equation eq3. 30 can be written as:
68
3 3
h ∂P ∂w + h ∂P ∂w dA =
∫∫
µ ∂x ∂x µ ∂z ∂z
element
eq3.32
3 3
h ∂P n + h ∂P n dl − ∂h
∫ w µ ∂x x µ ∂z z ∫∫ w 6u j dA
∂x
boundary element
At this point, the basis function approximations (eq3. 33) and the Galerkin approximation
NPE
eq3. 33 P ≈ ∑ P jφ j
j= 1
eq3. 34 w =φ i
where NPE = number of nodes per element and φk is used to represent the basis function.
The weight function, w, can also be thought of as a variation of the primary variable, P. If
P is specified on the boundary the variation of P, and consequently w, will be zero. From
Figure 3. 6 it can be seen that P is specified as the lubricant supply pressure (P0) on all
four boundaries of the pad surface. Hence, the boundary integral on the right hand side of
eq3.32 evaluates to zero. Noting this observation and substituting eq3. 33 and eq3. 34
NPE h 3 ∂φ i ∂φ j ∂φ i ∂φ j
∂h
eq3. 35 ∑ ∫∫ + dA P j =− ∫∫ φ i 6u J dA
j= 1 element µ ∂x ∂x ∂z ∂z element ∂x
eq3. 36
NPE
∑ KE ijP j = FE i
j =1
h 3 ∂φ i ∂φ j ∂φ i ∂φ j ∂h
where KE = ∫∫ + dA and FE = − ∫∫ φ i 6u J dA
ij µ ∂x ∂x ∂z ∂z i ∂x
element element
69
KE refers to the element “stiffness” matrix and FE is the element load or “force ” vector.
Equation eq3. 36 constitutes the foundation for the finite element method.
Before the integrals contained in KEij and FEi can be evaluated, the basis functions
have to be defined. These linear basis functions will be used to facilitate numerical
integration over elements which are similar to the one shown in Figure 3. 7.
4 3
dz η
1 2
dx = Rj dθ
The linear basis or interpolation functions and their derivatives can be expressed as
follows:
1
(a ) φ = 1+ ξ cjξ 1+ ηcj η
j 4
∂φ j ξ cj
eq3. 37 (b) = 1+ ηcjη
∂ξ 4
∂φ j ηcj
(c) = 1+ ξ ξ
∂η 4 cj
The values for some of the terms in equation eq3. 37 are given in Table 3. 2:
70
Table 3. 2 - Basis Function Parameters
1 -1 -1 1 -1/√3 -1/√3 1
2 +1 -1 2 +1/√3 -1/√3 1
3 +1 +1 3 +1/√3 +1/√3 1
4 -1 +1 4 -1/√3 +1/√3 1
The chain rule can now be used to express the isoparametric derivatives (from eq3. 37) in
∂φ ∂x ∂z ∂φ
∂ξ ∂ξ ∂ξ ∂x
=
∂z ∂φ
eq3. 38
∂φ ∂x
∂η ∂η ∂η ∂z
The Jacobian matrix (JA) can be approximated in terms of the derivatives of the basis
functions:
NPE ∂φ j NPE ∂φ j
∂x ∂z ∑ x ∑ zj
j ∂ξ ∂ξ
∂ξ ∂ξ j=1 j= 1
eq3. 39 JA = =
∂x ∂z NPE ∂φ j NPE ∂φ j
∂η ∂η ∑ x j ∂η ∑ zj
∂η
j=1 j= 1
Hence, the Cartesian derivatives which are required by eq3. 36 can be obtained from eq3.
71
∂φ j −1
NPE NPE ∂φ j
∂φ ∑ x ∑ zj ∂φ
j ∂ξ ∂ξ
∂x j= 1 j =1 ∂ξ
eq3. 40 =
∂φ NPE ∂φ j NPE ∂φ j ∂φ
∂z ∑ x j ∑ zj ∂η
j= 1 ∂η j =1 ∂η
The integration in eq3. 36 is usually done numerically by Gaussian Quadrature which can
be illustrated as follows:
1 1 NG
eq3. 41 ( )
∫∫ f ( x, z)dA = ∫∫ f ( x, z)dxdz = ∫ ∫ f ξ, η JA dξdη = ∑ w m f ξ m , η m JA
element element −1 −1 m =1
where NG = number of Gauss points = 4, and the values for the other parameters are
given in Table 3. 2. In order to avoid the use of two loops in the computer program, the
traditional double summation, which is implied by the integration, has been collapsed into
The integrals in eq3. 36 can now be evaluated using eq3. 41 and the resulting
element matrices can be assembled to give global matrices. The boundary conditions can
now be applied, and these global matrices can then be solved for nodal pressure values
using a banded matrix solver. All of these operations are executed by the program listed
in Appendix B.
force which can be translated into a moment. For static equilibrium to be achieved, these
72
L/ 2 xe
eq3. 42 MOMENTi = ∫ ∫ R j + t p θ − θ p P cos θ − θ p dxdz = 0, i =1,2,..., N pads
− L/ 2 xs
Equation eq3. 42 is an approximation which was derived from Figure 3. 2 and it can be
numerically evaluated using eq3. 41. The symbols xs = Rjθs and xe = Rjθe. Like the pads,
the journal is also free to move. Static equilibrium requires that the X and Y forces on the
N pads L / 2 x e
∑ FX = ∑ ∫ ∫ ( P cos θ)dxdz = 0
i = 1 − L / 2 x s
i
eq3. 43
N pads L / 2 x e
∑ FY = ∑ ∫ ∫ ( P sin θ) dxdz − Wb = 0
i =1 − L / 2 x s
i
where Wb = bearing load. Equations eq3. 42 and eq3. 43 can be solved simultaneously by
using the Newton-Raphson method for systems. This method solves the system of
MOMENT
1 0
M M
⇒ F( X) = 0
MOMENT ~ ~ ~
eq3. 44 N pads = 0
∑ FX 0
∑ FY 0
~
where X consists of the tilt angles (δ) for each pad, the journal eccentricity (e) and the
angle to the line-of-centers (θc). At the foundation of the Newton-Raphson method lies
~
~ ~ ~ ~ ∂ F
X
~ ~ ~
eq3. 45 F X k + 1 ≈ F X k + ~ k +1 − X k = 0
∂X ~
Xk
73
~
~ ∂F
( )
~ ~ ~ ~
Defining J = ~ and Y = X k +1 − X k yields after rearranging eq3. 45:
∂X X
~
k
~ −1 ~ ~
eq3. 46
~ ~
( )
Y =− J F X k
~
Equation eq3. 46 can be solved for each successive kth iteration until Y becomes less than
some pre-defined tolerance. At this point the iterations have converged to give a solution
~
to eq3. 45 which, in turn, solves eq3. 42 and eq3. 43. The final result yields a X vector
which contains the equilibrium tilt angles for the pads and the e and θc which define the
The FEM program was checked for convergence and validated by comparing the
results with those from other studies. The first comparison was done with data presented
74
Table 3. 3 - Input Data for Bearing Number 10 (5-Pad TPJB), Test Number 10
(Someya, 1989)
PARAMETER VALUE
Someya obtained an experimental static eccentricity ratio of 0.63 (p. 224 Someya, 1989),
but the attitude angle (ψa0 = θc-π/2) was not given. He also developed a numerical model
which yielded ε0 = 0.58 and ψa0 = 0.0 deg. The following data is from the program which
75
was developed in the last few sections. NEX = number of elements in the x-direction and
(circ.) 5 10 15
From Table 3. 4, it appears that the numerical solution has converged to values of ε0 =
0.64 and ψa0 = 0.05 deg for NEX = 10 and NEZ = 5 (in order to maintain symmetry, 6
elements will be used in the z-direction along with the 10 elements in the x-direction).
Furthermore, the solution of ε0 = 0.64 is closer to the experimental value than the solution
obtained by Someya. It should be noted that most TPJBs, which are symmetric about the
Y-axis, have a static equilibrium position that lies on the negative Y-axis. This means that
the attitude angle should be close to zero degrees, as is evident from the data in Table 3.
4. The slight deviation from the theoretical zero degrees in Table 3. 4 could be due to a
76
Another TPJB was investigated by Josiah Knight (Knight, 1990). This 5-pad
TPJB was loaded in a load-between-pad or LBW configuration, which means that there is
no pivot at θp = 270 deg as in the Someya case (load-on-pad or LOP configuration). Data
PARAMETER VALUE
The results from the current program yielded ε0 = 0.62 and ψa0 = 0.005 deg, while
Knight’s program predicted ε0 = 0.60 and ψa0 = 0.0 deg. The actual experimental values
were ε0 = 0.64 and ψa0 = -2.2 deg. The negative non-zero value of ψa0 in the
77
experimental data indicates that there was some sort of asymmetry introduced into the
bearing, e.g. friction between the pad and the pivot could cause some of the pads to stick
in a fixed orientation which would force the journal to conform to that pre-defined
position.
the plain journal bearing rotors. In fact, there is no ξ-coordinate system in the TPJB case
and this makes it easier to obtain the locations of the hot and the cold spots. These spots
can be located by using the equations that are given in chapter 2 for xH, yH, xC and yC.
Unfortunately, the stiffness and damping coefficients (which are required by VT-
FAST) for tilting pad journal bearings do not have a convenient closed form solution as do
those for the plain journal bearings. As a result, a numerical computation is required and
(1) Reynolds’ equation is first solved and the resulting data is used to establish an
(2) A perturbation is then applied to the journal. This perturbation can be a displacement
(3) The pads are then allowed to adjust to the new journal position by obtaining tilt angles
(4) The differences between the equilibrium forces and the new X and Y forces on the
journal are then calculated. Dividing these force differences by the perturbation
78
Before a perturbation can be applied to the journal, the Reynolds’ equation must
rewritten as follows:
∂ h 3 ∂P ∂ h 3 ∂P ∂h ∂h
eq3. 47 + = 6u j +12
∂x µ ∂x ∂z µ ∂z ∂x ∂t
The transient term in eq3. 47 is sensitive to the perturbations and can be evaluated by
taking the partial derivative of the film thickness with respect to time. Equation eq3. 25
(the film thickness expression) contains the term ecos(θc-θ) which can be expanded to give
ecos θc cos θ +esin θc sin θ = -Xj cos θ - Yj sin θ. Now, the only terms in eq3. 25 which
are time-dependent are Xj (the X-coordinate of the journal center Oj), Yj (the Y-
coordinate of the journal center Oj) and δ, the tilt angle of the pad. Therefore, the
transient part of the last term in eq3. 47 can now be expressed as:
As a result of this new term, the finite element formulation of the Reynolds’ equation has
to be reformulated as follows:
eq3. 49
NPE
∑ KE ij P j = FE i
j= 1
h 3 ∂φ i ∂φ j ∂φ i ∂φ j ∂h ∂h
where KE = ∫∫ + dA and FE = − ∫∫ φ i 6u J −12 dA
ij µ ∂x ∂x ∂z ∂z i ∂x ∂t
element element
79
The tilt angle velocity term, δ& , is typically zero if the above-mentioned method for
obtaining the stiffness and damping coefficients is used. However, if a pad perturbation
technique is used (Allaire et al., 1981; Nicholas et al, 1977) then this term will be non-
zero. For the purposes of this research, δ& will be set to zero. The stiffness and damping
The cross-coupled coefficients Kxy, Kyx, Cxy and Cyx are usually close to zero and
negligible compared with the other direct coefficients (Shapiro and Colsher, 1977). This
implies that the TPJBs are more stable than the plain journal bearings. Cross-coupled
coefficients tend to produce tangential follower forces (Chapter 1, Vance, 1988) which
augment rotor whirl and eventually drive the system unstable. As a result, bearings with
lower cross-coupling terms, such as TPJBs, are generally more stable than bearings with
higher cross-coupled coefficients, e.g. plain journal bearings. In the current research, it
will be assumed that the TPJB cross-coupled coefficients are effectively zero.
It should be noted that the stiffness and damping coefficients in equation eq3. 50
Cp
K′ = K
Wb
eq3. 51
ωC p
C′ = C
Wb
80
where K represents a generic stiffness coefficient and C represents a generic damping
12
Someya-Exp.
Someya-Theory
10 Current Theory
8
Dimensionless Value
0
K'xx K'yy C'xx C'yy
Coefficient
Figure 3. 8 shows that the current model gives fair agreement with experimental and
81
3.4. Temperature Distribution
The energy equation from chapter 2 can also be adapted for use with the TPJBs.
However, the complexity of the TPJB film thickness expression and the use of Fortran
equation. Euler’s method, which is a finite difference formulation, can be used to solve
2H 2 HT 2µωR 2j
amb
eq3. 52 Ti +1 = 1− ∆θ i
T + ∆θ +
ρ l c l ωh i ρ c ωh
l l ρ c h 2
l l i
where ∆θ = increment in pad angle. Equation eq3. 52 can now be solved by specifying an
inlet temperature at the leading edge of each tilted pad. The inlet temperature (Tin) at one
of the pads is taken to be the sum of the lubricant supply temperature and the estimated
h
eq3. 53 Tin = T0 + min ∆Tav
h in
where hmin = minimum inlet film thickness and hin = inlet film thickness for current pad.
This relationship assumes that higher inlet film thicknesses are associated with lower inlet
temperatures. It should be noted that the mixing of the lubricant in the pockets between
the pads and convective heat transfer at the pad boundaries were neglected. The bearing
heat transfer coefficient will be determined from the following equation (Ettles, 1992):
− 0.4
eq3. 54 . u 0j .7 µ 0− 0.2 R j∆ p
H = 255
82
where H = heat transfer coefficient in Wm-2K-1, uj = journal surface speed in m/s, µ0 =
A pertinent question is: Why wasn’t the finite element method used to solve a
more complex form of the energy equation for the temperature distribution? The answer
is an attempt was made to obtain an FEM model of the temperature field. However,
concordance with the observations of Abdel-Hadi et al. (1985) who did work on
the SUPG (Streamline Upwinding Petrov-Galerkin) model (Abdel-Hadi et al., 1985 and
Kim et al., 1994). This model accounts for the fluid moving upstream, and uses weighting
functions which are chosen to be different from the basis functions as is done in the
SUPG model were unsuccessful and the more simple model summarized in eq3. 52 was
selected. This simpler model gives a reasonable estimate of the temperature distribution
Knight (Knight, 1990) was chosen to check for convergence and accuracy. The input data
83
Table 3. 6 - Input Data for a 5-pad LBP TPJB (Knight, 1990)
PARAMETER VALUE
The temperature at the leading edge of pad #5 was calculated for different mesh sizes
(NEX = number of elements in the x-direction). Since eq3. 52 is one dimensional, only
84
the θ or x direction meshing will be considered. The results from this analysis are shown
in Table 3. 7.
Temperature @
NEX leading edge of
Pad #5 (oF)
5 157
10 158
15 158
Table 3. 7 shows that the solution has converged for NEX = 10. The pressure calculation
also requires NEX to be at least 10 for the journal position to converge. Therefore, the
mesh size used will be NEX = 10 and NEZ = 6 which gives a total of 60 elements per pad.
The resulting temperature of 158 oF in Table 3. 7 is close to the actual experimental value
of about 156 oF, and gives a better approximation than Knight’s theoretical value of 131
o
F. The complete temperature profile for all of the pads is given in Figure 3. 9.
85
Figure 3. 9 - Temperature Distribution for Knight’s Bearing
shown in Figure 3. 9. Nevertheless, this model has its limitations. It does not account for
heat transfer at the pad boundaries, and therefore does not predict the experimental
temperature drops at the trailing (end) edges of the pads. Knight’s theoretical model does
take into account this type of heat transfer and thus mimics these temperature drops as
indicated in Figure 3. 9. These temperature drops are important for a physically consistent
model. The TPJB is a circular system, which means that the trailing edge of pad 2 shares
a lubricant pocket with the leading edge of pad 3 and the temperature within that pocket
has to be some average of the pad 2 outlet and pad 3 inlet temperatures. Therefore, it is
necessary for the pad 2 trailing edge temperature to drop in order to avoid a physically
86
implausible temperature discontinuity. Even though the current model does not depict this
temperature drop, it does give a decent overall estimate for the temperature profile. In
fact, especially for pads 3 to 5, the current model gives a better approximation than
The temperature profile, the pressure distribution and the film thickness can also be
As indicated in Figure 3. 10, the maximum temperatures usually occur on the most highly
loaded pads (greatest pressures) which are associated with the lowest film thickness
87
values. This is in agreement with physical expectations. The graphics program was
Jongh and Morton, 1994) in their analysis of a LP/IP centrifugal compressor. The
compressor was engaged in an off-shore gas-lift operation and its rotor was supported by
two 5-pad LOP TPJBs. It was found that it was impossible to attain the maximum
encountered around 11400 RPM near the non-driven end (NDE) bearing. These fast-
growing vibrations reached a significant level and thus forced the rig operators to reduce
Upon examining the rig in a balancing facility it was discovered that the rotor was
very sensitive to small unbalances. It was also observed that the vibration problem
persisted when the mechanical and labyrinth seals were removed. This precluded the
possibility of the Newkirk Effect. However, the hysteresis in the vibration growth and
decay seemed to suggest a thermal phenomenon. De Jongh and Morton then built a test
rotor with identical dynamic characteristics as the full-sized compressor rotor and
measured the temperature difference across the shaft within the bearings. Resulting data
confirmed that the vibration problem was caused by thermally-induced bending and was
therefore a manifestation of the Morton Effect. The final solution to the vibration problem
88
Using the data in deJongh and Morton paper, a VT-FAST model of the centrifugal
compressor was constructed. The compressor stages and other rotor portions were
simplified and modeled as disks, but the dimensions and gyroscopics were essentially the
same. Figure 3. 11 shows the resulting model and Table 3. 8 gives the corresponding
input data before any modifications were done to solve the thermal instability problem.
NDE DE
Rotor
89
Table 3. 8 - Input Data for deJongh and Morton Compressor Rotor Before
Modifications
Journal/Bearing Parameters
Rj, Radius of journal 1.93 in. (4.9 cm) 1.93 in. (4.9 cm)
Wb, Bearing load 488 lbf (2172 N) 506 lbf (2250 N)
Md, Overhang Mass 59.2 lbm (26.9 kg) 74.0 lbm (33.6 kg)
Ld, Overhang Distance 7.0 in. (0.18 m) 4.7 in. (0.12 m)
ωMCOS, Max. Cont. Op. Spd. 11947 RPM (1251 rad/s) 11947 RPM (1251 rad/s)
Um, Initial Mechanical Unbalance 0.39 oz in. (0.28 kg mm) 0.39 oz in. (0.28 kg mm)
Lubricant Parameters
P0, Lubricant supply pressure 19.1 psi (0.132 MPa) 19.1 psi (0.132 MPa)
T0, Lubricant supply temperature 122 oF (50oC) 122 oF (50oC)
β, Thermoviscosity index 0.017 /oF (0.031/ oC) 0.017 /oF (0.031/ oC)
µ0, Lubricant supply viscosity 2.94 µreyn (20.3 cP) 2.94 µreyn (20.3 cP)
ρl, Lubricant density 0.03 lbm/in3 (860 kg/m3) 0.03 lbm/in3 (860 kg/m3)
cl, Lubricant sp. heat capacity 0.48 Btu/lbm/ oF 0.48 Btu/lbm/ oF
(2000 J/kg/K) (2000 J/kg/K)
In order to ensure that the model was similar to the deJongh and Morton compressor
rotor, the critical speed maps were compared as shown in Figure 3. 12. The undamped
critical speed maps show that there is reasonable agreement between the VT-FAST model
90
Figure 3. 12 - Critical Speed Maps for VT-FAST Model and Actual Compressor
The non-driven end (NDE) of the centrifugal compressor was then analyzed using the
Morton Effect program for the TPJB and the resulting unbalance curves are shown in
Figure 3. 13. These curves predict the onset of the instability at 11508 RPM which is very
close to the 11400 RPM threshold instability value observed by de Jongh and Morton.
91
Figure 3. 13 - Unbalance Curves for the Non-Driven End of the deJongh and
The hump around 10000 RPM in Figure 3. 13 is due to the 3rd critical speed (see
Figure 3. 12) which provides larger amplitude orbits. As mentioned in chapter 2, these
orbits reduce the film thickness near the hot spot and increase the temperature in this
region. More incentive for thermal bending is thus obtained and the unbalance level rises.
The response diminishes and hot spot film thickness increases after the rotor goes through
the critical. As a result, the thermal gradient and the overall unbalance decrease around
11000 RPM. A thermal instability is eventually achieved when the unbalance level rises
due to an increase in viscous dissipation at higher speeds. The 4th critical speed around
92
The driven-end (DE) of the compressor was also analyzed (Figure 3. 14) and this
end seemed to exhibit the Morton Effect at speeds greater than 12488 RPM.
Figure 3. 14 - Unbalance Curves for the Driven End of the de Jongh and Morton
This DE instability was not mentioned in the paper by de Jongh and Morton (de Jongh and
Morton, 1994). However, their solution to the thermal instability at 11400 RPM involved
decreasing the overhang masses at both the NDE and the DE.
In order to stabilize the compressor rotor, de Jongh and Morton reduced the NDE
overhang mass by 27.6 lbm (12.5 kg) and decreased the DE overhang mass by by 17.9 lbm
(8.1 kg). These mass reductions were achieved primarily by reducing the density of the
overhang components -_ they replaced some of the steel components with titanium ones.
93
After these modifications, some of the model input data was changed as indicated in Table
3. 9. The input data not mentioned in Table 3. 9 is the same as that given in Table 3. 8.
Table 3. 9 - Input Data for de Jongh and Morton Compressor Rotor After
Modifications
By using the data in Table 3. 9, unbalance curves were obtained for the NDE (Figure 3.
15). These curves show that the thermal instability is not present in the given speed range
The reduction in the NDE overhang mass stabilizes the rotor system in several
1. The lowered mass forces the 4th critical (at around 145000 RPM) to a higher speed
and thus reduces the influence of this critical on the lower speed orbits. As a result,
2. Decreasing the overhang mass directly reduces md which, in turn, lowers the thermal
3. Since the mass reduction occurs primarily at the ends of the rotor, the center-of-
gravity of the overhung mass would be drawn closer to the NDE bearing and the Ld
value (overhang distance) would diminish (compare Table 3. 8 and Table 3. 9). This
94
decrease in Ld would lower the thermal deflection and thus reduce Ut. Hence, U
would be further reduced and prevented from exceeding the threshold value (Uthr).
4. The unbalance peak due to the critical near 10000 RPM has been significantly
attenuated. This attenuation may be partially due to the increase in phase angle
between the initial mechanical (Um) and thermal (Ut) unbalances. Before
modifications, this angle was 31o near 10000 RPM but after modifications this angle
increased to 61.8o near 10500 RPM (near the new position of the attenuated peak).
As a result, Um and Ut are more out of phase which decreases the overall unbalance,
Figure 3. 15 - Unbalance Curves for the Non-Driven End of the de Jongh and
95
Figure 3. 16 shows that the mass decrease at the DE also seems to have removed the
thermal instability from the operating speed region . Therefore, it appears that there are
no Morton Effect possibilities close to the MCOS of 11947 RPM (1251 rad/s). This is in
Figure 3. 16 - Unbalance Curves for the Driven End of the de Jongh and Morton
96
3.5.2. de Jongh and van der Hoeven
Another example of the Morton Effect operating in rotors supported by TPJBs
was provided by de Jongh and van der Hoeven (de Jongh and van der Hoeven, 1998).
They analyzed a synchronous instability in two identical pipeline compressors which were
being used in a Dutch gas station to transport natural gas. These 584 lbm (265 kg)
compressors operated between 5370 RPM and 9400 RPM and one is schematically shown
in Figure 3. 17.
DE NDE
Hoeven
97
During operation, these compressors exhibited large vibration levels above 7200
RPM and the machines had to be shut down. Analysis confirmed that these vibrations
were synchronous in nature and that there was a hysteresis in the amplitudes. This means
that the magnitude of the deceleration vibrations were up to 4 times the magnitude of the
acceleration vibrations over the same speed range. The phase readings from these
responses were also erratic. These observations pointed to a thermal instability and the
labyrinth seals were checked to determine whether the Newkirk Effect was the cause of
the problem. However, these seals did not seem to be initiating a rub and further
examination of the shaft did not reveal any scuff marks that are normally associated with
shaft rubbing. It was finally concluded that the Morton Effect was the source of the
The researchers found that increasing the bearing clearance slightly would solve
this problem. Increasing the bearing clearance causes a less centered journal orbit which
leads to more overall cooling of the hot spot and attenuates the thermal gradient. As a
result, the Morton Effect is mitigated and the associated unstable vibration levels would be
reduced. A more detailed explanation on the relationship of orbit position and the Morton
Even though increasing the clearance was a solution to this problem, de Jongh and
van der Hoeven decided to adopt a different approach. They found that the minimum
required increase in bearing clearance was only 0.03% which was not a very practical
design specification. As a result, they invented a heat sleeve barrier which reduced the
temperature gradient across the shaft and decreased the amount of thermal bending caused
98
The current model was applied to the de Jongh and van der Hoeven case and the
Table 3. 10 - NDE Data for Compressors Analyzed by de Jongh and van der Hoeven
Journal/Bearing Parameters
Rj, Radius of journal 2.0 in. (5.08 cm)
Wb, Bearing load 490 lbf (2180 N)
Md, Overhang Mass 249 lbm (113 kg)
Ld, Overhang Distance 10.5 in. (0.27 m)
ωMCOS, Max. Cont. Op. Spd. 9400 RPM (984 rad/s)
Um, Initial Mechanical Unbalance 0.37 oz in. (0.27 kg mm)
Lubricant Parameters
P0, Lubricant supply pressure 19.1 psi (0.132 MPa)
T0, Lubricant supply temperature 122 oF (50oC)
β, Thermoviscosity index 0.017 /oF (0.031/ oC)
µ0, Lubricant supply viscosity 2.94 µreyn (20.3 cP)
ρl, Lubricant density 0.03 lbm/in3 (860 kg/m3)
cl, Lubricant sp. heat capacity 0.48 Btu/lbm/ oF
(2000 J/kg/K)
The overhang mass at driven-end (DE) of the compressors was much smaller than that of
the non-driven end (NDE) which indicates that the Morton Effect would be more likely to
occur at the NDE (as observed by de Jongh and van der Hoeven). As a result, only the
99
Before any Morton Effect analysis is done, the current model must be checked to
ensure that it dynamically matches the actual compressor. One way of accomplishing this
task is to compare the critical speed map from the model with values from the compressor
(Figure 3. 18).
Figure 3. 18 - Critical Speed Maps for VT-FAST Model and an Actual Compressor
Figure 3. 18 does not show as good a match between the compressor and the model as
does Figure 3. 12. However, Figure 3. 18 indicates that there is a reasonable match in the
region of the actual support (primarily from bearings) characteristic for the compressor.
This match means that there is good agreement between the model and the compressor
100
within the operating region of the compressor. It is therefore safe to proceed with the
The unbalance curves for the model (Figure 3. 17) were generated by using the
data given in Table 3. 10. These curves are depicted in Figure 3. 19.
The unbalance curves predict the onset of thermal instability at 7070 RPM and this agrees
fairly well with the 7200 RPM value recorded by de Jongh and van der Hoeven.
primarily responsible for the shape of the unbalance curve. Critical speeds also seem to
influence the unbalance curve, e.g. around 5000 RPM and around 9500 RPM.
101
De Jongh and van der Hoeven solved this thermal instability by installing a heat
sleeve barrier around the shaft portion within the bearing. This barrier trapped a layer of
air close to the shaft and insulated it from the hot lubricant. As a result, there was minimal
heat transfer to the shaft and the thermal gradient across the journal was reduced by about
85%. If this reduction in thermal gradient is included in the current model, the following
According to Figure 3. 20, the model predicts that the compressor would be stabilized if
the heat sleeve barrier were used. This statement agrees with the experimental
102
3.5.3. Overhung Compressor
A compressor company graciously provided some experimental data for an
overhung compressor which seemed to exhibit the symptoms of the Morton Effect. This
Journal/Bearing Parameters
Rj, Radius of journal 2.25 in. (5.72 cm) 2.00 in. (5.08 cm)
Wb, Bearing load 302.9 lbf (1347 N) 55.7 lbf (248 N)
Md, Overhang Mass 123.82 lbm (56.2 kg) 43.04 lbm (19.5 kg)
Ld, Overhang Distance 9.68 in. (0.25 m) 7.36 in. (0.19 m)
ωMCOS, Max. Cont. Op. Spd. 12075 RPM (1264 rad/s) 12075 RPM (1264 rad/s)
Um, Initial Mechanical Unbalance 0.1386 oz in. (0.1 kg mm) 0.1386 oz in. (0.1 kg mm)
Lubricant Parameters
P0, Lubricant supply pressure 25.0 psi (0.172 MPa) 25.0 psi (0.172 MPa)
T0, Lubricant supply temperature 140 oF (60oC) 140 oF (60oC)
β, Thermoviscosity index 0.01309 /oF (0.024/ oC) 0.01309 /oF (0.024/ oC)
µ0, Lubricant supply viscosity 1.7 µreyn (11.7 cP) 1.7 µreyn (11.7 cP)
ρl, Lubricant density 0.03414 lbm/in3 0.03414 lbm/in3
(945 kg/m3) (945 kg/m3)
cl, Lubricant sp. heat capacity 0.4685 Btu/lbm/ oF 0.4685 Btu/lbm/ oF
(1962 J/kg/K) (1962 J/kg/K)
103
The experimental data obtained from the compressor is shown in Figure 3. 21 and Figure
accel-decels were done between 10000 RPM and 12000 RPM in attempt to surpass the
instability. Unfortunately, these were unsuccessful and were responsible for the loops seen
104
Figure 3. 22 - Synchronous Vibration Data obtained from an X-Probe at the DE of
The synchronous vibrations, the hystereses in magnitudes and the erratic phase responses
depicted in Figure 3. 21 and Figure 3. 22 indicate that the Morton Effect might be the
cause of the problem. By using forced response data, which was provided by the
compressor company, unbalance curves for the Morton Effect were obtained (Figure 3. 23
105
Figure 3. 23 - Unbalance Curves for the NDE of the Overhung Compressor
106
Figure 3. 24 - Unbalance Curves for the DE of the Overhung Compressor
Figure 3. 23 and Figure 3. 24 show that the NDE is predicted to exhibit the Morton Effect
between 10630 RPM and 12640 RPM while the DE is predicted to go unstable after
12990 RPM. The experimental results indicate that the NDE went unstable near 12000
RPM which agrees fairly well with the predictions. On the other hand, the DE also went
unstable around 12000 RPM even though the analysis does not predict any DE instability
in this region.
According to the data, the DE response decreased in amplitude from about 8000
RPM to 12000 RPM. At this point the vibrations suddenly increased to unstable values.
It is possible that the NDE went unstable first and the DE was sensitive to the resultant
107
unbalances at different locations. As a result of these different sensitivities, the high
unbalances from the Morton Effect can apply a centrifugal force to an overhung region of
the rotor and thus lead to unstable vibrations elsewhere. Such vibrations would explain
108
4. Design of Test Rotor
One of the goals of this research was to initiate experimental work on the Morton
Effect. In order to accomplish this goal, an attempt was made to design and build a test
Since the Morton Effect is a thermal bending phenomenon, the measurement of the
measurement has only been accomplished by deJongh and Morton (deJongh and Morton,
1994) who installed Pt-100 RTDs inside the journal portion within the bearing at a
distance of 1.3 mm from the journal surface. A “special slipringless transmitter” was then
used to transfer data from the rotating shaft to a stationary data acquisition system.
install thermocouples within the shaft. However, even though thermocouple wire is
cheap, the extra machining on the shafts could prove to be costly and this would elevate
the already steep price of the 303 stainless steel shafts. In addition, any asymmetric
machining could further unbalance the rotor and lead to mechanical vibration problems.
109
The second problem would be to transfer the data from the rotating shaft to the
stationary data acquisition system. Traditionally, slipring devices are used for this purpose
but these devices are usually very costly (one quotation for a slipring system from
Michigan Scientific Corporation was in excess of $11,400). Slipring brushes are also
subjected to frictional heating at high shaft speeds and this could lead to discrepancies in
the temperature measurements. Attempts were also made to find temperature sensitive
paints, crystals or crayons that could detect the temperature difference which is typically
around 10oC. Unfortunately, these efforts proved to be unfruitful and the idea to measure
the transverse temperature difference was abandoned due to high cost and lack of
Even though the transverse temperature gradient will not be measured, the design
of a test rotor would still be beneficial. The Morton Effect has other symptoms such as
hysteresis in vibration magnitude plots, erratic phase graphs and synchronous vibration
responses which could still be identified with the available capabilities and budget.
However, before the test rig could be designed, it was necessary to establish several
design parameters which would restrict the design space and hopefully yield a good
design.
to modify an existing rotor. This rotor and its associated data acquisition system was last
110
Fluid film TPJBs
The existing rotor was built upon a table that was 8ft. (2.4 m) long, 2 ft. (0.6 m) wide and
3 ft. (0.9 m) high when the safety cover was in place. At the non-driven end (NDE), a ball
and holder was used to eliminate the axial vibrations of the rotor shaft which was primarily
supported by two 5-pad LOP spherical pivot fluid film bearings. These bearings were
supplied by a oil system that consisted of a 10 gallon (38 liter) reservoir of Teresstic - 32
light turbine oil at a flow rate of about 1.85 GPM (7 LPM). A thermostat was used to
measure and control the oil temperature. The 303 stainless steel rotor shaft or journal was
about 43 in. (1.1 m) long with a 1 in. (2.5 cm) diameter, and a large test disk was mounted
on the shaft at the bearing midspan. A Kop-Flex KD2 disk coupling was used to attach
the test shaft to the drive shaft at the driven-end (DE) of the rotor. Two ball bearings
were used to support the drive shaft which was belted to a 10 HP (7.5 kW) motor. The
111
belt allowed the motor speed to be stepped up by a factor of 3 which translates to an
operating rotor speed range of 450 RPM (7.5 Hz) to about 10,000 RPM (167 Hz). Three
pairs of probes (from Bently Nevada) were used to record the horizontal and the vertical
vibrations of the shaft. These probes were referenced to a synchronous key phasor trigger
which was at the end of the drive shaft that fit into the Kopflex coupling. All of the
transducer probes were attached to a proximeter console that relayed the signals to the
It was decided that most the data acquisition system should be maintained but the
final display should use an ADRE 8-channel data acquisition instrument instead of the 16
channel box. This choice was made because only 7 probes (7 channels) would be used
and the ADRE 108 DAI also allowed a variety of plots and tabular data to be obtained.
The rotor should also be modified to attain the overhung configuration (Figure 4.
Figure 4. 2- Schematic Diagram Showing the Modified Rotor Configuration for The
Morton Effect
112
In order to minimize cost, bearing power loss and unwanted pedestal vibrations, the size
and weight of the modified rotor was kept to a minimum. These criteria, and the given
existing rotor foundation, were the basis for selecting a shaft length of 36 in. (91 cm) and
a 1 in. (2.5 cm) thick, 3 in. (7.6 cm) diameter disk at the NDE overhang position. The
shaft diameter also needed to be small to minimize the weight, but large enough to allow a
thermal gradient to develop across the shaft. As a result, 2 shaft diameters of 1.0 in (2.5
cm) and 1.5 in (3.8 cm) were considered in the design process. The bearings were to be
custom-made by Rotating Machinery Technology and thus the bearing clearance was also
chosen to be an unknown factor that had to be determined from the analysis of the
designs.
The various designs were analyzed using the following design parameters:
(1) Attainable Instability Speed. The average speed at which the Morton Effect would
occur should be within the 450-10000 RPM operating range of the rig.
(2) Operation Away from the Critical Speeds (Ncr). If any of the designs have an
instability speed close to a lateral critical speed of the rotor, then the unstable
vibrations may be blamed on the critical response. Therefore, a good design should
have the Morton Effect occurring at a speed which is separated as far as possible from
the lateral criticals. On the other hand, criticals do provide the large amplitude orbits
(see chapters 2 and 3) which favor the Morton Effect. Hence, if the final design does
encounter the Morton Effect near a critical speed, and if the Morton Effect response
can be clearly separated from that of the critical, then this design would be acceptable.
(3) φ should be close to 0 . An optimum configuration for the Morton Effect is one in
o
which the thermal and the mechanical unbalances are exactly in phase. This
113
arrangement causes the resultant unbalance (U) to be maximized and increases the
(4) Centered Orbit. A more centered synchronous orbit allows one side of the journal to
Hence, the likelihood of a transverse journal temperature gradient is increased and the
(5) Adequate Orbit Amplitude. The orbit must be large enough to generate the viscous
energy required for the Morton Effect. Larger orbits tend to reduce the film thickness
which increases the shear strain rate of the fluid. As a result, the shear stresses are
higher and more viscous dissipation is produced. If the orbit is too large then the
journal could contact the pads and annihilate the babbitt. The optimum orbit,
therefore, should be restricted to a suitable operating range (0.6 - 0.8 Cp was chosen
parameters given in the previous section. As mentioned before, only the shaft diameter
and the diametral bearing clearance (2Cb) were chosen as design variables. An overhang
length of 10 in. (25.4 cm) was arbitrarily chosen (see Figure 4. 2).
114
Table 4. 1- Description of Rotor Designs
The average instability speed (Nmort) for the onset of the Morton Effect was evaluated for
each of the designs along with the second and third lateral undamped critical speeds. The
115
10000
9000
8000
Speed (RPM)
Nmort
7000 Ncr2
Ncr3
6000
5000
4000
S1.0C2.50
S1.0C3.00
S1.0C3.50
S1.0C4.00
S1.0C4.50
S1.5C3.00
S1.5C4.00
S1.5C5.25
S1.5C6.00
S1.5C7.00
Case
Figure 4. 3 - Average Morton Effect Instability Speeds and Critical Speeds for the
Designs
All of the Nmort values appear to be within the 450 - 10000 RPM operating range of the rig
but some of the designs, e.g. S1.0C4.50, appear to have Nmort values that are too close to
critical speeds. This proximity may present problems if the Morton Effect response cannot
be distinguished from the vibrations caused by the excitation of the critical speed.
116
The phase difference between the mechanical and the thermal unbalances for the
120
100
80
| | (deg)
60
40
20
0
S1.0C2.50
S1.0C3.00
S1.0C3.50
S1.0C4.00
S1.0C4.50
S1.5C3.00
S1.5C4.00
S1.5C5.25
S1.5C6.00
S1.5C7.00
Case
Figure 4. 4 - |Angle| between Mechanical and Thermal Unbalances for the Different
Designs
From Figure 4. 4 it can be inferred that S1.5C3.00 would have the most out-of-phase
mechanical and thermal unbalances. Based on this criterion, the S1.5C3.00 case would be
the least likely to exhibit the Morton Effect. On the other hand, the S1.5C6.00 case has
117
the most in-phase mechanical and thermal unbalances and would therefore be a better
In order to find out whether the design orbits are centered, the static eccentricity
ratio was considered (Figure 4. 5). Designs with lower eccentricity ratios have more
centered orbits which favor the Morton Effect. The attitude angles are not considered
because these angles are usually zero for tilting pad journal bearings.
0.400
0.350
0.300
0.250
0.200
0.150
0.100
0.050
0.000
S1.0C2.50
S1.0C3.00
S1.0C3.50
S1.0C4.00
S1.0C4.50
S1.5C3.00
S1.5C4.00
S1.5C5.25
S1.5C6.00
S1.5C7.00
Case
118
Figure 4. 6 shows the orbit size and its required limitations for the designs.
Designs such as S1.0C3.00 have too small orbits and therefore may not produce enough
viscous energy to fuel the Morton Effect. However, designs like S1.5C5.25 fall within the
6.00
5.00
4.00
Radial Displacement (mils)
NDE + ecc
3.00 0.6 Cp
0.8 Cp
2.00
1.00
0.00
S1.0C2.50
S1.0C3.00
S1.0C3.50
S1.0C4.00
S1.0C4.50
S1.5C3.00
S1.5C4.00
S1.5C5.25
S1.5C6.00
S1.5C7.00
Case
All of the data presented so far can be summarized and analyzed using Table 4. 2.
The design with the 1.5” diameter shaft and 5.25 mils diametral clearance is the only one
119
to satisfy all of the design parameters. Hence, this design will be chosen as the basis for
S1.0C3.00 x x x
S1.0C3.50 x x
S1.0C4.00 x
S1.0C4.50 x x
S1.5C3.00 x x
S1.5C4.00 x x x x
S1.5C5.25 x x x x x
S1.5C6.00 x x x x
S1.5C7.00 x x x
necessary for additional existing rig components to be modified or replaced. Some of the
new rig components that were changed or replaced are discussed in the following sections.
120
4.3.1. Test Shaft
The test shaft (Figure 4. 7) was to be made from 303 stainless steel which is
suitable for machining and possesses superior fatigue resistance when compared with
Notes
• All keyway dimensions should have
Test Shaft
tolerances of +0.0020,-0.0000 inches.
• The dimension tolerance on the shaft
length is not critical. 303 stainless steel
• Shaft surface finish ~ 16 micro-inches.
Originally, it was thought that the previous shaft supplier for the Rotor Lab, i.e. Winfred
Berg, would be able to furnish the shaft without problems. Unfortunately, this was not the
case. After several weeks of negotiating with Winfred Berg , it was decided to look for
Several shaft suppliers from this search were then contacted and eventually Staples
Research Corp. (7036 Tech Circle, Manassas, VA 20109, Tel.: (888) 765-7654, Fax:
121
(703) 335-1234) was selected as the shaft supplier. Two shafts were ordered at about
$235 per shaft and shaft 1 was found to have a diameter of 1.4985” whereas 1.4989” was
4.3.2. Bearings
Rotating Machinery Technology, Inc., a member of the Rotordynamics Affiliates
Group, supplied the 5-pad LOP tilting pad bearings. Both of these bearings had a bore of
1.5042” which meant that shaft 1 would have a bearing clearance of 5.7 mils and shaft 2
would have a clearance of 5.3 mils. Since shaft 2 was closer to the recommended 5.25
mils then this shaft was selected for future experimental work.
The bearing fits into the housing as shown in Figure 4. 8. Unfortunately, the
housing was initially designed for a 1.0” diameter shaft, and so the apertures in the end-
shields were too small to accommodate the new 1.5” diameter shaft. New end-shields had
to be fabricated from Lexan which is more ductile and less likely to crack than perspex.
122
Figure 4. 8 - Bearing Housing Components
4.3.3. Coupling
Only one of the Kop-Flex coupling hubs needed to be replaced. The hub adjoining
the 1.0” drive shaft was still going to be used since the drive shaft would be unchanged.
However, the driven test shaft had a diameter of 1.5”, which meant that the other coupling
hub (which was designed for a 1.0” shaft) had to be changed and some balancing holes
had to be drilled. After some tedious negotiating with Motion Industries, Inc., which is a
distributor for Kop-Flex, it was agreed to purchase an entire coupling with the machine
work done as specified in Figure 4. 9. The final cost for the total package was in excess
of $600.
123
Figure 4. 9 - Schematic Diagram of the Kop-Flex Coupling Showing the Required
Modifications
to be secured to the shaft, and one which did not. The disk without a keyway would
torque and the shaft may be damaged in the process. Furthermore, the external diameter
of a $51.56 keyless mount from Trantorque is 2.375”. This large diameter only leaves
about 0.3125” on the 3.0” diameter disk for balancing holes which may become necessary
124
– a large unbalance due to asymmetry in the disk or in the shaft mount could prevent the
If the disk with the keyway was chosen then there would have to be balancing
However, the shaft would not be damaged and the disk could easily be installed and
replaced. In addition, no more funds would have to be spent on a shaft mount and there
would be more room on the disk for balancing holes. The disk with the keyway (Figure 4.
10) was chosen for the overhang disk design in light of these reasons.
Overhang Disk
125
4.3.5. Inboard Disk
During operation of the rig it was found that an inboard disk was required for
additional balancing of the rotor. Since this disk was to be placed in between the bearings
no keyway could be used. Instead, a small screw would be used to secure the disk to the
shaft. This method of attachment prevented the use of a very large disk because high
inertial forces could easily dislodge the disk. As a result, a thin 3” diameter disk was
Inboard Disk
This inboard disk was checked before and after the rotor was run at different speeds. No
shifting in the disk position was observed and the disk did assist in the overall balancing of
the rotor.
126
4.4. Analysis of Final Test Rotor Design
After the rotor components were assembled and all of the modifications were
made, the final test rotor design resembled the drawing in Figure 4. 12.
Figure 4. 12 - The Final Design for the Morton Effect Test Rotor
A comparison of Figure 4. 12 with Figure 4. 2 shows that some changes have been made.
The major alterations were a shortening of the bearing span from 21.5” to 21”, the
inclusion of an inboard disk and the repositioning of some of the probes. These changes
were done primarily to allow the rotor to be better balanced. The actual Morton Effect
127
overhang DE
NDE inboard coupling
disk brg hsg
brg hsg disk
The rest of the systems (lubrication, speed control and data acquisition) were described in
the Design Parameters section. However, a few minor changes had to be made. The oil
flow rate was decreased to 1 GPM in order reduce the leakage of oil from the center hole
in the end-shields. In addition, the oil inlet temperature was increased to 100oF to ensure
that the oil flowed smoothly through the supply pipes, the bearing housing and the
drainage pipes. Figure 4. 14 illustrates the Morton Effect rotor and its accompanying
systems.
128
master
ADRE 108 power
DAI
control rotor
box
computer with
ADRE Software
As a result of the modifications, the dynamics of the rotor were slightly altered
and the analyses (including the Morton Effect analysis) had to be repeated in greater
detail. These analyses were essential because they would give a final prediction for the
The first step in the analysis was to obtain a VT-FAST model by using the
dimensions and information in the given drawings. This model is shown in Figure 4. 15.
129
NDE DE
Input data for the VT-FAST model and the other analyses is listed in Table 4. 3. The
rotor was originally designed to have the Morton Effect at the non-driven end. However,
Table 4. 3 shows that the driven end has a significant overhang and may also be
susceptible to this thermal bending phenomenon. Consequently, both ends of the rotor
130
Table 4. 3 - Data for Morton Effect Rotor
Journal/Bearing Parameters
Rj, Radius of journal 0.75 in. (1.9 cm) 0.75 in. (1.9 cm)
Wb, Bearing load 12.58 lbf (56 N) 13.94 lbf (62 N)
Md, Overhang Mass 6.64 lbm (3.0 kg) 8.57 lbm (3.9 kg)
Ld, Overhang Distance 6.04 in. (0.15 m) 4.35 in. (0.11 m)
Um, Initial Mechanical Unbalance 0.022 oz in. 0.022 oz in.
(0.016 kg mm) (0.016 kg mm)
Lubricant Parameters
P0, Lubricant supply pressure 20 psi (0.138 MPa) 20 psi (0.138 MPa)
T0, Lubricant supply temperature 100 oF (38 oC) 100 oF (38 oC)
β, Thermoviscosity index 0.017 /oF (0.031/ oC) 0.017 /oF (0.031/ oC)
µ0, Lubricant supply viscosity 4.24 µreyn (29.2 cP) 4.24 µreyn (29.2 cP)
ρl, Lubricant density 0.03 lbm/in3 (860 kg/m3) 0.03 lbm/in3 (860 kg/m3)
cl, Lubricant sp. heat capacity 0.48 Btu/lbm/ oF 0.48 Btu/lbm/ oF
(2000 J/kg/K) (2000 J/kg/K)
One of the options in the VT-FAST program is a stability analysis. This feature calculates
the damped critical speeds of the rotor system and thus provides a better estimate of the
critical speeds than the undamped critical speed map. A knowledge of the location of the
critical speeds is essential for both the safe operation of the rig and the Morton Effect
model. By using this stability option, the first four criticals were obtained (see the
following figures.)
131
Figure 4. 16 - Predicted 1st Damped Critical Speed for the Morton Effect Rotor
132
Figure 4. 17 - Predicted 2nd Damped Critical Speed for the Morton Effect Rotor
133
Figure 4. 18 - Predicted 3rd Damped Critical Speed for the Morton Effect Rotor
134
Figure 4. 19 - Predicted 4th Damped Critical Speed for the Morton Effect Rotor
Figure 4. 16, Figure 4. 17, Figure 4. 18 and Figure 4. 19 show the predicted damped
critical speeds for the Morton Effect rotor. These critical speeds are around 4344 RPM,
5443 RPM, 8150 RPM and 24266 RPM respectively. Therefore, within the rotor’s
operating speed range of 450 RPM to 10000 RPM, there are 3 critical speeds. A forced
response analysis was done in order to determine which of these critical speeds would
produce unsafe vibration levels. In this analysis, an unbalance was applied to the rotor (at
the NDE overhang mass c.g.) and the resulting vibration levels were plotted as a function
135
Figure 4. 20 - Response at Probes 3,4 due to a 1oz in. Unbalance at the NDE
From Figure 4. 20, it can be seen that the 3rd mode (near 8150 RPM) produces the most
significant response to the applied unbalance. The 1st and 2nd modes have either been
damped out or have not been sufficiently excited. Nevertheless, Figure 4. 20 provides a
The final step in the analysis is to obtain the unbalance curves for the Morton
Effect. Both the NDE and the DE were analyzed and the resulting curves are shown in
136
Figure 4. 21 - Unbalance Curves for the Non-Driven End of the Morton Effect Test
137
Figure 4. 22 - Unbalance Curves for the Driven End of the Morton Effect Test Rotor
(Final Design)
The figures above indicate that both ends of the rotor would be susceptible to the Morton
Effect. An instability range of 7280 RPM - 8380 RPM is observed for the NDE, whereas
the DE seems to undergo the Morton Effect from 7744 RPM - 8256 RPM. Besides
overlapping each other, these two ranges also encompass the 8150 RPM 3rd critical speed.
As previously-mentioned, critical speeds can produce large responses which can aid the
vibration amplitude and erratic phase angles, which can be used to distinguish the critical
speed behavior from the Morton Effect instability. It was therefore decided to run the
rotor in order to determine: (1) whether the Morton Effect would occur within the
138
predicted speed ranges and (2) if the thermal bending response could be distinguished
139
5. Experimental Results from Test Rotor
After assembly (see chapter 4), the rotor had to be balanced in order to attain the
predicted Morton Effect speed range. Two types of experimental data were then
obtained by running the rotor close to and within the predicted instability region. The first
data set involved holding the rotor at a constant speed, for several minutes, within the
Morton Effect range. An acceleration and a deceleration through the Morton Effect speed
range produced the second set of data. Both of these data sets and the rotor balancing are
displaced from the axis of rotation. This mass produces a centrifugal force which can
drastically increase the rotor vibrations to levels that can damage the bearings. Therefore,
in order to safely attain the rotor operating speeds, it is frequently necessary to reduce
these unbalance levels to more acceptable values. This process is called rotor balancing.
The general purpose of rotor balancing is to reduce the current vibrations (As-Is
Vibrations) in the desired speed range to smaller values which are observed at lower
140
speeds (Run-Out Vibrations). Such a reduction is achieved by applying correction masses,
The Morton Effect rotor was particularly difficult to balance. The primary
obstacle was to reduce the vibration levels near the 3rd critical speed (~ 8000 RPM) to
more acceptable values. It was originally thought that single-plane balancing would have
balancing was then tried with no avail. Finally, three-plane balancing was done and the
rotor was successfully run up to 9500 RPM. A least squares program (written by Dr.
Kirk) was used to calculate the correction masses for the 3 balancing planes. A
comparison of the unbalance levels before and after balancing is shown in Figure 5. 1.
141
Figure 5. 1 shows that the response has been attenuated by about 1 mil and that the 3rd
critical was successfully traversed as the rotor was accelerated up to 9500 RPM. It should
be noted that, as predicted in chapter 4, the 1st and 2nd critical speeds did not present
rotor may have been due to the balancing speed (7900 RPM) being too close to a bending
mode (3rd critical at ~ 8000 RPM). The elastic deformation of the rotor in this mode
redistributes the mass about the rotational axis. As a result, the unbalance configuration
can vary as this mode is approached and such behavior complicates the balancing process.
Another source of unbalance variation could have been the slight misalignment of
the test shaft. It was found that the coupling did not completely uncouple the two shafts
and, as a result, the orientation of the drive shaft skewed the test shaft position. Since the
coupling is flexible (but not enough to adequately uncouple the test and the drive shafts),
this degree of skew could vary as the rotor is brought up to speed. Such a variation
would cause a portion of the test shaft to be non-uniformly displaced from the rotation
axis and thus contribute to a varying mechanical unbalance. This behavior would present
balancing and the rotor was now suitable for experimental work.
thermal gradient to develop within the bearing. In the case of the turbocharger studied by
142
Faulkner, Strong and Kirk (Faulkner et al., 1997), it took about half an hour for this
phenomenon to occur at relatively constant speed. On the other hand, the Morton Effect
occurred after ~ 15 minutes at constant speed in the de Jongh and van der Hoeven
compressor (de Jongh and van der Hoeven, 1998). The current rig has a smaller shaft
diameter than either the turbocharger or the compressor and it may take less time for a
thermal gradient to develop in this shaft. However, smaller shaft diameters create less
viscous dissipation which would decrease the magnitude of the thermal gradient. In other
words, there was no certainty as to when the Morton Effect would occur in the current
rig.
In order to solve this dilemma, it was decided that the rotor should be held at a
fixed speed, for several minutes, within the predicted Morton Effect range. This process
would then be repeated for different speeds within the predicted instability region. A
typical plot from these experiments is shown in Figure 5. 2. This plot shows that the rotor
was taken up to 7800 RPM and held there for about 50 minutes. None of the probes,
including probe 3, recorded the significant increase in vibration levels that are normally
associated with the Morton Effect. The probe 3 response in Figure 5. 2 has some rippling
in its shape but this due to background electrical and mechanical noise. Therefore, the
fixed speed data did not yield any evidence of the Morton Effect.
143
8
0
14:45:36 14:52:48 15:00:00 15:07:12 15:14:24 15:21:36 15:28:48 15:36:00 15:43:12 15:50:24 15:57:36
Time (hh:mm:ss)
Figure 5. 2 - Speed and Probe 3 Response for a Fixed Speed Data Plot
144
5.3. Accel-Decel Data
A second set of data was obtained by accelerating and decelerating the rotor
through a speed range which encompassed the predicted Morton Effect instability region.
Prior to this accel-decel, the oil was heated to 100oF at 500 RPM. The rotor speed was
then increased to 5000 RPM and data was recorded every 50 RPM until a speed of 9000
RPM was attained. At this point, the rotor was decelerated at the same rate (~ 4000
RPM/minute) back to 5000 RPM. Results from this accel-decel are shown in Figure 5. 3
and Figure 5. 4.
145
Figure 5. 4 - Probe 5 Response Data for 5000-9000 RPM Acceleration-Deceleration
The figures above show the responses from a non-driven end (Probe 3) probe and a driven
end (Probe 5) probe. Both of these figures show responses that exhibit hysteresis in
amplitude and erratic phase. As the rotor is accelerated, both the DE and the NDE probes
show a peaking in amplitude around 8000 RPM and a phase drop that is expected when
critical speeds are traversed. However, during the deceleration, the amplitudes increase by
a factor of ~ 4 and the phase responses differ from the responses observed during
acceleration.
146
A cascade plot is a three-dimensional plot that has been collapsed into two
dimensions. It shows the magnitude and the frequency of the responses for different rotor
speeds. The cascade plots corresponding to the vibrations shown in Figure 5. 3 and
Frequency (Hz)
147
Frequency (Hz)
The figures above indicate that the large deceleration vibration occurs at a frequency
which is equal to 1 X shaft rotational speed. Hence, this unstable response is synchronous
in nature.
By using the DC Gap voltage readings from the probes, the shaft centerline
position can be obtained as a function of speed. The NDE probes 3 and 4 were used to
obtain the location of the shaft center near the non-driven end bearing, while the DE
probes 5 and 6 were used to get the same data near the driven end bearing. This
148
Figure 5. 7 - Shaft Centerline Position (from Probes 3 and 4) near NDE Bearing
149
Figure 5. 8 - Shaft Centerline Position (from Probes 5 and 6) near DE Bearing
The NDE bearing (Figure 5. 7) appears to behave as a normal tilting pad bearing.
The attitude angle is close to zero and the shaft position rises with speed as a result of
increased lift. It should be noted that the 8000 RPM decel data point is the most centered
Figure 5. 8 shows that the DE part of the test shaft is locked into a position close
to the bearing center and does not respond much to speed increases. This fixed position
may have been due to insufficient uncoupling between the drive shaft and the test shaft.
Such lack of uncoupling also presented problems during the balancing process.
150
(1) Decrease in Support Damping. An rise in the decel response (see Figure 5. 3 and
Figure 5. 4) could have been caused by a decrease in damping within the bearing.
However, a decrease in damping would have also produced a steeper slope on the phase
curve at the ~ 8000 RPM critical speed. Figure 5. 3 and Figure 5. 4 show that this is not
the case. The slope of the phase curve at 8000 RPM seems to be the same for both the
accel and the decel cases. Hence, a decrease in damping was ruled out as a potential
(2) Increase in Air Content of Lubricant. During the accel-decel runs, it was observed
that there was some foam present in the lubricant. These air bubbles in the oil could lead
to turbulent behavior which may produce the erratic results that were recorded. In order
to test this hypothesis, it was decided to increase the air content in the oil by lowering the
oil reservoir temperature. The colder oil would dissolve more air and should therefore
make the problem worse. Of course, lowering the oil temperature also increases the
viscosity and thus enhances the damping. Nevertheless, an instability of the recorded
magnitude should be able to supersede the increased damping if the driving force is the
increased air content in the lubricant. The results for the colder oil accel-decel runs are
151
Figure 5. 9 - Probe 3 Response Data for 5000-9000 RPM Acceleration-Deceleration
152
Figure 5. 10 - Probe 5 Response Data for 5000-9000 RPM Acceleration-Deceleration
Figure 5. 9 and Figure 5. 10 indicate that the magnitude of instability is around 8 mils p-p
which is less that the ~ 11 mils p-p recorded in Figure 5. 3 and Figure 5. 4. Therefore, it
does not appear that the increased air content in the oil was the cause of the 8000 RPM
decel instability. However, more work should be done to verify this conclusion, e.g.
blowing air bubbles into the oil without significantly changing the temperature.
(3) Usual Accel-Decel Characteristic. Most rotors usually exhibit different response
lower than the acceleration vibrations for both linear and non-linear systems (Ishida et al.,
1987). Figure 5. 3 and Figure 5. 4 show that the current rig displays the opposite trend,
153
i.e. the decel response is higher than the accel response. Therefore, this rotor does not
(4) Influence of Residual Energy from Critical. It was suggested during an interview with
residual energy from the critical speed. Hence, the instability should be eliminated by
decelerating the rotor from a higher speed which is less influenced by the 8000 RPM
critical speed. The results from one such deceleration are shown in Figure 5. 11 and
Figure 5. 12.
154
Figure 5. 12 - Probe 5 Response Data for 5000-9500 RPM Acceleration-Deceleration
The figures above show that the ~ 11 mils responses in Figure 5. 3 and Figure 5. 4 have
been reduced to about 7 mils. As a result, it does appear that the critical exerts some
influence over the decel instability. Operating further away from the critical reduces the
amount of energy put into the system at the start of the decel and thus reduces the overall
excitation of that critical. However, the instability still exists and the erratic phase shifts
are relatively unchanged. Nevertheless, the speed at which the decel begins does influence
the magnitude of the instability. More research needs to be done in this area, e.g. use a
more powerful motor to initiate a decel well above (> 10000 RPM) the 3rd critical speed
155
(5) Increase in Mechanical Unbalance. A sudden increase in the mechanical unbalance
could also cause the high decel responses seen in Figure 5. 3 and Figure 5. 4. However,
such an elevation in the mechanical unbalance, e.g. blade loss in a turbine or loss of a
correction mass in the current rotor, would be present at all synchronous speeds and
would get worse at higher speeds. This type of behavior is not observed in the
experimental results. The rotor stabilizes after the instability and the magnitude decreases
in the lower speed region. The experimental results are also repeatable which is contrary
(6) Newkirk Effect. The hysteresis in the response, the erratic phase angles and the
synchronous vibrations can also indicate that a thermal instability, such as the Newkirk
Effect, could be present. However, there are no labyrinth (or any other) seals in the rotor
and there is a low probability of any contact between the rotor and a solid surface. The
most likely candidate to initiate a thermal rub would be the end shields and these have a
clearance of about 200 mils p-p. This value is much greater than the 11 mils p-p vibration
levels which were observed. Therefore, the Newkirk Effect may not be a potential source
(7) Morton Effect. Another synchronous thermal instability, which could produce the
unstable response, would be the Morton Effect. The 8000 RPM instability peak is within
the predicted range for the Morton Effect (NDE: 7280 RPM - 8380 RPM, DE: 7744
RPM - 8256 RPM) and the experimental results also show that both ends go unstable
156
which is concordance with the analytical predictions. This synchronous instability also
shows a hysteresis in amplitude and yields erratic phase results. Furthermore, the shaft
centerline plots (Figure 5. 7 and Figure 5. 8) indicate that the orbits are fairly centered
near the unstable region. All of these characteristics describe or favor the Morton Effect.
Figure 5. 13 shows the response at the NDE bearing of the de Jongh and Morton
157
This rotor exhibited the Morton Effect and was studied intensively by de Jongh and
Morton (de Jongh and Morton, 1994). Figure 5. 13 indicates that the rotor was
accelerated up to about 11400 RPM when it encountered the Morton Effect. The rotor
was subsequently decelerated. However, since it takes some time for the thermal gradient
to subside, there is a hysteresis in the displacement plot and the decel amplitudes are
higher than the accel ones. In addition, the thermal bending causes the resultant unbalance
to vary in phase, i.e. the decel phase was higher than its accel counterpart. This erratic
phase variation may be due to the feedback nature of the Morton Effect. The initial
mechanical unbalance determines the response (phase and magnitude) at the bearing which
then leads to a thermal unbalance. The resultant unbalance can then cause a new response
at the bearing which can then invoke a new thermal unbalance. As a result of this process,
By comparing Figure 5. 13 with Figure 5. 3 or Figure 5. 4, one can see that there
are some similarities and some differences. Both sets of plots show that the decel
magnitude and phase data are generally higher than the corresponding accel values.
However, the de Jongh and Morton data show that the phase shifts and the amplitude
hystereses occur at the same speed, whereas Figure 5. 3 or Figure 5. 4 have the major
phase shifts occurring at lower speeds. In addition, the current rotor has the instability
occurring at a critical, but the de Jongh and Morton results have the Morton Effect in a
critical-free region. Even though critical speeds can aid the Morton Effect, they can also
make it difficult to distinguish the Morton Effect response from the vibrations due to the
excitation of the critical. Finally, the de Jongh and Morton compressor became
progressively unstable as the rotor was accelerated. The current rotor, on the other hand,
158
only became unstable when the rotor was decelerated. If the instability in the current rotor
was the Morton Effect then a possible explanation could be that the decel motion was
required to kick-start this phenomenon. Within the bearing lubricant, there is a velocity
gradient which causes the fluid closer to the shaft to be moving at higher velocities. As
the rotor is decelerated, some of this fluid retains the momentum from the original higher
speed and thus leads to an inertial effect at the current lower speed. It is possible that
such an inertial effect can alter the dynamics and reduce the phase difference between the
mechanical and thermal unbalances. This decreased phase separation would favor the
Morton Effect.
In summary, it can be said that the unusual instability observed may have been
caused by the Morton Effect or by influence from residual critical energy. More work
therefore needs to be done to investigate this instability. Besides extending the speed
range to check the influence of the critical, the temperature difference across the shaft
could be measured. This measurement would definitely support or refute the existence of
159
6. Conclusions
Several conclusions can be made based on the work presented in this dissertation.
These conclusions will be divided into those concerning the theoretical work and those
overhung rotors that are supported by fluid film bearings and is caused by differential
(2) Most of the prior models of the Morton Effect consisted of frequency-domain
(3) This dissertation proposes a new model to predict the onset of the Morton Effect. The
current steady-state model uses an unbalance threshold (based on 15% of the rotor
160
(4) The current model indicates that a worst-case scenario for the Morton Effect would be
one with a centered, circular and large amplitude orbit. Decreasing the phase difference
between the thermal and mechanical unbalances would also increase the likelihood of the
Morton Effect.
(5) Predictions made by the current model have both slightly underestimated and slightly
overestimated the actual instability speeds. However, there is generally good agreement
between the current model and other theoretical models and experimental data. This good
agreement applies to rotors which are supported by plain or tilting pad journal bearings.
(2) During deceleration, this rotor exhibited an unusual instability near 8000 RPM. The
erratic phase response which are characteristics of the Morton Effect. Furthermore, this
(3) The Morton Effect may have caused this unusual instability, but the influence of
residual energy from the 3rd critical speed (~8000 RPM) could also be the culprit.
161
7. Suggested Future Research
Tech. It is hoped that it may provide some ground work for future studies. A few
suggestions and recommendations for future theoretical and experimental research are as
follows:
(1) More case studies can be done to test the program. Continued assistance from
(2) More investigations on the prediction of the onset of the Morton Effect as a function
(3) Currently, the Morton Effect programs are designed to predict the onset of this
thermal bending phenomenon. These steady-state programs, however, do not give the
transient response that is available in other programs. Such a modification would allow
for better prediction of the magnitude and the phase of the thermal bending response.
162
7.2. Suggested Future Experimental Research
(1) Before any additional experimental work is done, it is suggested that the uncoupling
between the test shaft and the drive shaft should be increased. This task can be
accomplished by increasing the length of the coupling spacer tube. More uncoupling
(2) As mentioned in chapter 5, more work can be done to investigate the effects of the
(3) The influence of the 3rd critical speed on the decel instability can also be checked by
operating at higher speeds. These experiments will require a more powerful motor to
(4) Measurements of the transverse temperature gradient in the shaft portion (within the
bearing) could be done. This data would confirm or refute the existence of a thermal
bending phenomenon. However, such data would be difficult and expensive to obtain and
a new rig may have to be built. As a result, it is suggested that efforts can be made to
modify an existing rig (which exhibits the Morton Effect) from a willing company. The
resulting data can be shared with the company and a better understanding of the Morton
163
Appendix A: Matlab GUI Program for Plain
Journal Bearings
%***********************************************************************
% clbck.m = program to execute when RUN button is pressed
%***********************************************************************
%Initialize program
clear all
164
rhol = eval(get(findobj(gcbf,'Tag','rhol'),'String'))/(2.2046*0.0254^3);
capl = eval(get(findobj(gcbf,'Tag','capl'),'String'))*1055.1*2.2046*1.8;
mu0 = eval(get(findobj(gcbf,'Tag','mu0'),'String'))*6894.8;
beta = eval(get(findobj(gcbf,'Tag','beta'),'String'))*1.8;
T0 = (eval(get(findobj(gcbf,'Tag','T0'),'String'))-32)/1.8;
Tamb = (eval(get(findobj(gcbf,'Tag','Tamb'),'String'))-32)/1.8;
T02 = T0 - Tamb;
L = eval(get(findobj(gcbf,'Tag','L'),'String'))*0.0254;
Rj = eval(get(findobj(gcbf,'Tag','Rj'),'String'))*0.0254;
cb = eval(get(findobj(gcbf,'Tag','cb'),'String'))*1e-3*0.0254;
Wb = eval(get(findobj(gcbf,'Tag','Wb'),'String'))*4.4482;
alphaj = eval(get(findobj(gcbf,'Tag','alphaj'),'String'))*1.8;
HTC = eval(get(findobj(gcbf,'Tag','HTC'),'String'))*745.7*1.8/0.0254^2;
NS = eval(get(findobj(gcbf,'Tag','NS'),'String'));
W = eval(get(findobj(gcbf,'Tag','W'),'String'))*4.4482;
md = eval(get(findobj(gcbf,'Tag','md'),'String'))/2.2046;
Ld = eval(get(findobj(gcbf,'Tag','Ld'),'String'))*0.0254;
Um = eval(get(findobj(gcbf,'Tag','Um'),'String'))*0.0254/(2.2046*16);
speed = eval(get(findobj(gcbf,'Tag','speed'),'String'));
omega = speed*2*pi/60;
xamp = eval(get(findobj(gcbf,'Tag','xamp'),'String'))*1e-3*0.0254;
xph = eval(get(findobj(gcbf,'Tag','xph'),'String'))*pi/180;
yamp = eval(get(findobj(gcbf,'Tag','yamp'),'String'))*1e-3*0.0254;
yph = eval(get(findobj(gcbf,'Tag','yph'),'String'))*pi/180;
CaseText = get(findobj(gcbf,'Tag','case'),'String');
%angular speed
w = omega(i);
%orbit parameters
Ax = xamp(i); Ay = yamp(i);
phx = xph(i); phy = yph(i);
165
A1 = 2*H/(rhol*capl*w*cb); B1 = 2*mu0*w*Rj^2/(rhol*capl*cb^2);
DTeff = (B1/A1-T02)*(1-exp(-A1*pi));
mu = mu0*exp(-beta*DTeff);
LT = length(T); T1 = T';
T2 = [T1(LT:-1:2) T1]; LT2 = length(T2);
if nxih == 0
nxih = 1;
elseif nxih > LT2
nxih = LT2;
end
166
if nxic == 0
nxic = 1;
elseif nxic > LT2
nxic = LT2;
end
%plot orbits
figure(gfnum+i)
orbplt(cb,delT(i2)*1.8,t,w,sf,Rj,Ax,Ay,phx,phy,xje,yje);
end
Ut(i) = md*yd*2.2046*16/0.0254;
Utt = md*yd;
chiav = mean(chi);
phi = pi-chiav;
figure(gfnum+i)
title(sprintf('%s%s %7.0f %s %5.2f %s',CaseText,':',speed(i),...
' RPM, DT = ',1.8*abs(DT),' deg F'))
xlabel('x/Cb')
ylabel('y/Cb')
hold on
%total unbalance
U(i) = sqrt(Um^2 + Utt^2 - 2*Um*Utt*cos(chiav))*2.2046*16/0.0254;
Uthr(i) = (thrfac*W/w^2)*2.2046*16/0.0254;
end
unb = spline(speed,U,sp2);
167
unbthr2 = spline(speed,Uthr,sp2);
j = 1;
for i = 1:length(sp2)-1
diff1 = unb(i)-unbthr2(i);
diff2 = unb(i+1)-unbthr2(i+1);
if diff1*diff2 < 0
Nthr(j) = sp2(i);
Uval(j) = unb(i);
j = j + 1;
end
end
plot(krpm,unb,'b--'), hold on
%Label plot
legend('U','Uthr')
title(CaseText)
xlabel('Speed (krpm)')
ylabel('Unbalance (oz in)')
%***********************************************************************
%***********************************************************************
% epsilon.m = function used by clbck.m. This function gives the static
% equilibrium eccentricity ratio for a narrow plain journal bearing.
%***********************************************************************
168
pi^4*(-16 + pi^2)^3 + ...
4*a* ...
(-1048576 + 131072*pi^2 + ...
14336*pi^4 - 1152*pi^6 + ...
27*pi^8)))^0.3333333333333333)^ ...
2)/ ...
(a*((-16 + pi^2)^3 + ...
72*a*(1024 - 64*pi^2 + 3*pi^4) + ...
12*sqrt(3)*sqrt(a)* ...
sqrt(67108864*a^2 + ...
pi^4*(-16 + pi^2)^3 + ...
4*a* ...
(-1048576 + 131072*pi^2 + ...
14336*pi^4 - 1152*pi^6 + 27*pi^8)) ...
)^0.3333333333333333))/(4.*sqrt(3)) - ...
sqrt(384 - (4*(-16 + 96*a + pi^2))/a + ...
(3072*a - (-16 + pi^2)^2)/ ...
(a*((-16 + pi^2)^3 + ...
72*a*(1024 - 64*pi^2 + 3*pi^4) + ...
12*sqrt(3)*sqrt(a)* ...
sqrt(67108864*a^2 + ...
pi^4*(-16 + pi^2)^3 + ...
4*a* ...
(-1048576 + 131072*pi^2 + ...
14336*pi^4 - 1152*pi^6 + 27*pi^8) ...
))^0.3333333333333333) - ...
((-16 + pi^2)^3 + ...
72*a*(1024 - 64*pi^2 + 3*pi^4) + ...
12*sqrt(3)*sqrt(a)* ...
sqrt(67108864*a^2 + ...
pi^4*(-16 + pi^2)^3 + ...
4*a*(-1048576 + 131072*pi^2 + ...
14336*pi^4 - 1152*pi^6 + 27*pi^8))) ...
^0.3333333333333333/a - ...
(24*sqrt(3)*(-32 + pi^2))/ ...
(a*sqrt((-3072*a + ...
(16 - pi^2 + ...
((-16 + pi^2)^3 + ...
72*a*(1024 - 64*pi^2 + 3*pi^4) + ...
12*sqrt(3)*sqrt(a)* ...
sqrt(67108864*a^2 + ...
pi^4*(-16 + pi^2)^3 + ...
4*a* ...
(-1048576 + 131072*pi^2 + ...
14336*pi^4 - 1152*pi^6 + ...
27*pi^8)))^0.3333333333333333)^ ...
2)/ ...
(a*((-16 + pi^2)^3 + ...
72*a*(1024 - 64*pi^2 + 3*pi^4) + ...
12*sqrt(3)*sqrt(a)* ...
sqrt(67108864*a^2 + ...
pi^4*(-16 + pi^2)^3 + ...
4*a* ...
(-1048576 + 131072*pi^2 + ...
14336*pi^4 - 1152*pi^6 + ...
27*pi^8)))^0.3333333333333333))) ...
)/(4.*sqrt(3)))/sqrt(2));
%***********************************************************************
169
%***********************************************************************
% difeq2.m = function used by clbck.m. This function gives the
% differential equation for the circumferential temperature profile.
% The equation is solved by the Matlab ode45 command.
%***********************************************************************
function dT = difeq2(th,T,flag,A,B,eps)
dT = [-A*T(1)/(1+eps*cos(th)) + B/(1+eps*cos(th))^2];
%***********************************************************************
%***********************************************************************
% orbplt.m = function used by clbck.m. This function plots the total
% synchronous orbit and appropriately scales the journal radius to show
the
% locations of the cold and the hot spots at different points in the
orbit.
%***********************************************************************
function [] = orbplt(cb,dT,t,w,sf,Rj,Ax,Ay,phx,phy,xje,yje)
Rj = sf*Rj/cb;
Ax = Ax/cb; Ay = Ay/cb;
xje = xje/cb; yje = yje/cb;
g0 = atan2(Ay*sin(phy),Ax*cos(phx));
th = 0:0.01*pi:2*pi;
t1 = 0:0.01*2*pi/w:2*pi/w;
xp = xje + Ax*cos(w*t1 + phx);
yp = yje + Ay*sin(w*t1 + phy);
plot(xp,yp,'k-')
axis equal
hold on
xj = xoj + Rj*cos(th);
yj = yoj + Rj*sin(th);
hold on
%***********************************************************************
%***********************************************************************
% mpippi.m = function used by clbck.m. This function places any given
% radian angle in the range -pi to +pi.
%***********************************************************************
170
function [xi2] = mpippi(xi)
if xi < -pi
while xi < -pi
xi = xi + 2*pi;
end
elseif xi > pi
while xi > pi
xi = xi - 2*pi;
end
end
xi2 = xi;
%***********************************************************************
171
Appendix B: Fortran Program for Tilting Pad
Journal Bearings
!*************************************************************************
! FINITE ELEMENT PROGRAM TO SIMULATE THE MORTON EFFECT OPERATING
! IN A CYLNDRICAL PIVOT TILTING PAD JOURNAL BEARING.
!*************************************************************************
!*************************************************************************
! PRE-PROCESSOR
!*************************************************************************
IMPLICIT NONE
INTEGER GDOF, EDOF, NBN, NBN2,NEX, NEZ, NE, NNX, NNZ, NPADS, NDOF,
* HBW, NPE, XXDOF, KCDOF, MAXITER, TGDOF, NSPD, NTAVG
REAL*8 PI, TOL, TOL2, DXX
PARAMETER ( NEX = 10, !# of elements in x direction
* NEZ = 6, !# of elements in z direction
* NE = NEX*NEZ, !total # of elements
* NDOF = 1, !number of nodal dof
* NPE = 4, !# of nodes per element
* EDOF = NDOF*NPE, !element degrees of freedom
* NNX = NEX + 1, !# of nodes in x direction
* NNZ = NEZ + 1, !# of nodes in z direction
* GDOF = NDOF*NNX*NNZ, !global degrees of freedom
* NBN = 2*NNZ + 2*(NNX-2),!# of press boundary nodes(EBCs)
* NBN2 = NNZ, !# of temp boundary nodes (EBCs)
* MAXITER = 20, !maximum # of iterations (Newton)
* NTAVG = 5, !# of trans.temp. averages
* DXX = 1D-6, !increment for XX
* TOL = 1D-6, !convergence tolerances
* TOL2 = 1D-6,
* HBW = (NNZ+2)*NDOF, !system half-bandwidth
* PI = 3.14159265359) !value for pi
172
REAL*8 THS, MINHIN, TH
INTEGER PADNUM, PERTFLAG, SDFLAG, ITH1, ITH2
INTEGER I, I2, I3, IC, NPI, J, J2, J3, J4
CHARACTER:: INF1*20, OUTF1*20, OUTF2*20, TITLE*100
!------------------------------------------------------------------
!GET INPUT DATA FROM FILE
!------------------------------------------------------------------
PRINT *, 'Please enter the input file name:'
READ *, INF1
READ(1,*)
READ(1,*)
READ(1,*) (THETAP(I), I = 1,NPADS)
READ(1,*)
READ(1,*)
READ(1,*)
READ(1,*)
READ(1,*) NSPD, RJ, ALPHAJ, LOAD
READ(1,*)
READ(1,*)
READ(1,*) LD, MD, WROTOR, UNBM
READ(1,*)
READ(1,*)
DO I = 1,NSPD
READ(1,*) SPEED(I), XAMP(I), XPH(I), YAMP(I), YPH(I)
END DO
READ(1,*)
READ(1,*)
READ(1,*)
READ(1,*)
READ(1,*) P0, T0, TAMB, MU0, BETA
173
READ(1,*)
READ(1,*)
READ(1,*) RHOL, CAPL
READ(1,*)
READ(1,*)
READ(1,*)
READ(1,*)
READ(1,*) SDFLAG ! SDFlAG = 0 => no coeff will be calculated
! SDFlAG = 1 => K&C coeff will be calculated
READ(1,*)
READ(1,*)
READ(1,*)
READ(1,*)
READ(1,*) OUTF1, OUTF2
CLOSE(1)
!------------------------------------------------------------------
!CONVERT INPUT DATA TO SI UNITS
!------------------------------------------------------------------
!Calculated Parameters
CB = CP*(1-M) ! Radial bearing clearance (m)
RP = RJ + CP ! Radius of curvature of pad
!Perturbation Parameters
RATIO = 0.005 ! Perturb. increment in ecc. ratio
!*************************************************************************
! PROCESSOR
!*************************************************************************
174
!Open unbalance data file
OPEN(10, FILE = OUTF2, STATUS = 'UNKNOWN')
!------------------------------------------------------------------
! FIND EQUILIBRIUM POSITION
!------------------------------------------------------------------
! SPECIFY NO PERTURBATION
PERTFLAG = 0
VX = 0.0
VY = 0.0
VD = 0.0
175
* NDOF, HBW, NPE, XXDOF,MAXITER,PERTFLAG,TOL,
DXX,L,
* PADANGLE,M, OFFSET, THETAP, TP, RJ, RP, CB,
* P0, VX, VY, VD, LOAD,
* U, MU, SFX, SFY,FX, FY, DEL, MOMENT,ECC, THC)
DO I = 1,TGDOF
TEMP2(I) = T0
END DO
!Get inlet pad film thicknesses. These will help assign TIN
MINHIN = 10000.0
DO PADNUM = 1,NPADS
DELTA = DEL(PADNUM)
THP = THETAP(PADNUM)
HIN(PADNUM) = H
END DO
DELTA = DEL(PADNUM)
THP = THETAP(PADNUM)
176
DO I = 1,GDOF
J = GDOF*(PADNUM-1) + I
TEMP2(J) = TEMP(I)
PRESS2(J) = PRESS(I)
END DO
EQECC = ECC
EQTHC = THC
EQSFX = SFX
EQSFY = SFY
EQEPS(I2) = EQECC/(1.0D0*CB)
EQPSI(I2) = EQTHC*180.0D0/PI - 90.0
EQMU(I2) = MU
CASE(1)
OPEN(1, FILE = 'PAD1')
CASE(2)
OPEN(1, FILE = 'PAD2')
CASE(3)
OPEN(1, FILE = 'PAD3')
CASE(4)
OPEN(1, FILE = 'PAD4')
CASE(5)
OPEN(1, FILE = 'PAD5')
CASE(6)
OPEN(1, FILE = 'PAD6')
CASE(7)
OPEN(1, FILE = 'PAD7')
CASE(8)
OPEN(1, FILE = 'PAD8')
CASE(9)
OPEN(1, FILE = 'PAD9')
CASE(10)
OPEN(1, FILE = 'PAD10')
177
END SELECT
DO I = 1,GDOF
END DO
CLOSE (1)
KXX(I2) = KIJ(1)*0.0254/4.4482D0
KYY(I2) = KIJ(4)*0.0254/4.4482D0
CXX(I2) = CIJ(1)*0.0254/4.4482D0
CYY(I2) = CIJ(4)*0.0254/4.4482D0
END IF
178
!CALCULATE AVERAGE TRANSVERSE TEMPERATURE DIFFERENCE
IF (SDFLAG == 0) THEN
TIME = (J3-1)*2.0D0*PI/(1.0D0*OMEGA*NTAVG)
179
CHI(J3) = OMEGA*TIME-THCH
DO I = 1,TGDOF
TEMP2(I) = T0
END DO
PERTFLAG = 5
!Get inlet pad film thicknesses. These will help assign TIN
MINHIN = 1000.0
DO PADNUM = 1,NPADS
DELTA = DEL(PADNUM)
THP = THETAP(PADNUM)
HIN(PADNUM) = H
END DO
DELTA = DEL(PADNUM)
THP = THETAP(PADNUM)
180
DO I = 1,GDOF
J = GDOF*(PADNUM-1) + I
TEMP2(J) = TEMP(I)
END DO
THP = THETAP(PADNUM)
THS = THP - OFFSET*PADANGLE
DO I = 1,NNX
I3 = IC + (I-1)*NNZ
J = GDOF*(PADNUM-1) + I3
J2 = NNX*(PADNUM-1) + I
TEMPC(J2) = TEMP2(J)
TH = THS + (I-1)*PADANGLE/(NEX*1.0D0)
THTEMPC(J2) = DATAN2(DSIN(TH),DCOS(TH))
END DO
END DO
END DO
181
SUMDTAVG = 0.0D0
DO J3 = 1,NTAVG
SUMDTAVG = SUMDTAVG + DTAVG(J3)
END DO
DT = SUMDTAVG/(1.0D0*NTAVG)
CHIAV = SUMCHI/(1.0D0*NTAVG)
PHI(I2) = PI-CHIAV
!Calculate phase of U
PHIU(I2) = DACOS((UNB**2 + UNBM**2 - UNBT**2)/(2.0D0*UNB*UNBM))
UNBT2(I2) = DABS(UNBT*2.2046*16.0D0/0.0254D0)
UNB2(I2) = DABS(UNB*2.2046*16.0D0/0.0254D0)
UNBTHR2(I2) = DABS(UNBTHR*2.2046*16.0D0/0.0254D0)
DELT(I2) = DABS(DT*1.8)
END IF
182
WRITE(10,230) NTAVG, RJ, CB, WROTOR, 0.0, 0.0
!*************************************************************************
! POST-PROCESSOR
!*************************************************************************
!==================================================================
!WRITE INPUT/OUTPUT FILE
!==================================================================
OPEN (3, FILE = OUTF1, STATUS = 'UNKNOWN')
DO I = 2,NPADS
WRITE(3,20) '=', THETAP(I)*180.0/PI,'deg'
END DO
183
WRITE (3,200) 'RHOL, Density =', RHOL*2.2046*0.0254**3, 'lbm/in^3'
WRITE (3,200) 'CAPL, Specific heat capacity =',
* CAPL/4186.8,'Btu/(lbm F)'
WRITE(3,*) ' '
DO I = 1,NSPD
WRITE(3,170) SPEED(I),SOM(I),DABS(EQEPS(I)),
* EQMU(I)*1.0D6/6894.8,
* 1.0D3*HMIN(I)/0.0254,PMAX(I)/6894.8,1.8*TMAX(I)-459.67
END DO
IF (SDFLAG == 1) THEN
DO I = 1,NSPD
WRITE(3,190) SPEED(I),KXX(I),KYY(I),CXX(I),CYY(I)
END DO
END IF
IF (SDFLAG == 0) THEN
DO I = 1,NSPD
WRITE (3,260) SPEED(I), UNBM/(0.02835*0.0254D0),
* UNBT2(I), UNB2(I), UNBTHR2(I), PHI(I)*180.0D0/PI,
* DELT(I)
END DO
END IF
CLOSE(3)
184
!*************************************************************************
! FORMATS
!*************************************************************************
10 FORMAT (5(E10.3,1X))
20 FORMAT (A40,1X,F8.2,1X,A16)
30 FORMAT (A6,1X,A11,1X,A8,1X,A8,1X,A8,1X,A11,1X,A16)
40 FORMAT (I2,5X,F5.0,7X,F8.0,1X,F8.0,1X,F8.0,1X,F7.5,5X,F7.3)
50 FORMAT (A40,1X,F8.4,1X,A10)
60 FORMAT (A40,1X,F8.0,1X,A10)
70 FORMAT (A40,1X,F8.1,1X,A10)
80 FORMAT (F6.0,1X,F6.2,1X,F6.2,1X,F6.2,1X,F6.2,1X,F6.2)
90 FORMAT (A4,1X,A10,1X,A15,1X,A4,1X,A10,1X,A15)
100 FORMAT (A4,1X,E10.4,1X,F10.2,6X,A4,1X,E10.4,1X,F10.2)
110 FORMAT (A10,1X,F10.0,1X,A15)
120 FORMAT (A10,1X,F10.2,1X,A15)
130 FORMAT (F10.0,2X,4(E10.4,2X))
140 FORMAT (A40,1X,F8.0,1X,A16)
145 FORMAT (A40,1X,I8,1X,A16)
150 FORMAT ('1',A4)
160 FORMAT (7(A8,1X))
170 FORMAT (F8.0,1X,F8.2,1X,F8.2,1X,F8.2,
* 1X,F8.1,1X,F8.0,1X,F8.0)
180 FORMAT (5(A10,1X))
190 FORMAT (F10.0,1X,4(E10.4,1X))
200 FORMAT (A40,1X,F8.3,1X,A16)
210 FORMAT (I6,1X,I6,1X,I6,1X,I6,1X,F6.2,1X,F6.0)
220 FORMAT (F6.2,1X,F6.2,1X,F6.2,1X,F6.2,1X,F6.2,1X,F6.2)
230 FORMAT (I6,1X,E10.4,1X,E10.4,1X,E10.4,1X,F6.2,1X,F6.2)
240 FORMAT (E10.4,1X,E10.4,1X,E10.4,1X,E10.4,1X,E10.4,1X,E10.4)
250 FORMAT (7(A8,1X))
260 FORMAT (F8.0,1X,4(F8.3,1X),2(F8.1,1X))
!*************************************************************************
! SUBROUTINES
!*************************************************************************
!------------------------------------------------------------------
!SUBROUTINE TO CALCULATE THE CONNECTIVITY VECTORS FOR EACH ELEMENT
!------------------------------------------------------------------
SUBROUTINE CONNVEC(EL, NPE, NEX, NEZ, ELNODES)
INTEGER NPE
INTEGER EL, NEX, NEZ, ELNODES(NPE)
INTEGER I, I1, I2
185
ELNODES(3) = ELNODES(2) + 1
ELNODES(4) = ELNODES(1) + 1
END IF
END DO
END
!------------------------------------------------------------------
!SUBROUTINE TO CALCULATE NODAL X AND Z COORDINATES
!------------------------------------------------------------------
SUBROUTINE NODCOORD(RJ, THP, NEX, NEZ, NNX, NNZ, OFFSET,
* PADANGLE, L, ELNODES, XJ, ZJ)
DO II = 1,NNX
END DO
DO JJ = 1,NNZ
!Calculate the z coordinate for the node
IF( MOD(I-JJ,NNZ) == 0 ) THEN
ZJ = ZS + (JJ-1)*L/NEZ
END IF
END DO
END
!------------------------------------------------------------------
!SUBROUTINE FOR THE MAXIMUM TILT ANGLE (DMAX)
!------------------------------------------------------------------
SUBROUTINE MAXDELTA(PADANGLE,M,OFFSET,THP,TP,RJ, ECC,CB,THC,DMAX)
186
REAL*8 M, CB, CP, RJ, RB, RP, ECC
REAL*8 OFFSET, PADANGLE, THC, TP
REAL*8 THP, THE, DMAX
END
!------------------------------------------------------------------
!SUBROUTINE TO CALCULATE FILM THICKNESS EXPRESSIONS FOR A
!TILTING PAD JOURNAL BEARING WHICH HAS A CYLINDRICAL PIVOT.
!------------------------------------------------------------------
SUBROUTINE FILMTH(DELTA, M, THP,TP,RJ, ECC,CB,THC, XJ, DHTHJ, HJ)
TH = XJ/RJ
END
!------------------------------------------------------------------
!SUBROUTINE TO CALCULATE BASIS FUNCTIONS PHI, DPHI/DX, DPHI/DZ
!AND THE DETERMINANT OF THE JACOBIAN
!------------------------------------------------------------------
SUBROUTINE BASFUNC(XI, ETA, XJ, ZJ, PH, DPHX, DPHZ, DETJA)
187
PH(I) = (1 + XIC(I)*XI) * (1 + ETAC(I)*ETA)/4.0
DPHXI(I) = XIC(I) * (1 + ETAC(I)*ETA)/4.0
DPHETA(I) = ETAC(I) * (1 + XIC(I)*XI)/4.0
END DO
END
!------------------------------------------------------------------
!SUBROUTINE TO PERFORM NUMERICAL INTEGRATION AND EVALUATE K AND
!F OVER A GIVEN ELEMENT
!------------------------------------------------------------------
SUBROUTINE ELMAT(VX, VY, VD, TP, THP, NPE, EDOF, XJ, ZJ, HJ,
* DHTHJ, MU, RJ, RP, U, KE, FE)
DO J = 1,EDOF
KE(I,J) = 0.0D0
188
END DO
END DO
DO I = 1,EDOF
H = H + PH(I)*HJ(I)
TH = TH + PH(I)*XJ(I)/RJ
DHTH = DHTH + PH(I)*DHTHJ(I)
END DO
!Calculate coefficients
A1 = H**(3.0D0) / MU
A2 = 6.0D0*U*DHTH/RJ - 12.0D0*(VX*DCOS(TH) + VY*DSIN(TH) +
* VD*(RP+TP)*DSIN(TH-THP))
DO J = 1,EDOF
END DO
END DO
END
!------------------------------------------------------------------
!SUBROUTINE TO ASSEMBLE THE BANDED GLOBAL STIFFNESS MATRIX (KG)
!AND THE GLOBAL LOAD VECTOR FG. CODE ADAPTED FROM J.N. REDDY,
!"AN INTRODUCTION TO THE FINITE ELEMENT METHOD", (1993), P.649
!------------------------------------------------------------------
SUBROUTINE GLOBMATBW(HBW, NPE, EDOF, NDOF, GDOF, ELNODES,
* FE, KE, FG, KG)
189
DO I = 1,NPE
NR = (ELNODES(I)-1)*NDOF
DO II = 1,NDOF
NR = NR + 1
L = (I-1)*NDOF + II
DO J = 1,NPE
NCL = (ELNODES(J)-1)*NDOF
DO JJ = 1,NDOF
M = (J-1)*NDOF + JJ
NC = NCL + JJ + 1 - NR
END DO
END DO
END DO
END
!------------------------------------------------------------------
!SUBROUTINE TO GENERATE THE BOUNDARY NODES
!------------------------------------------------------------------
SUBROUTINE BNODES(NNX, NNZ, NBN, BOUNOD)
DO I = 1,NNZ
S1NOD(I) = I
S3NOD(I) = NNZ*(NNX-1) + I
BOUNOD(I) = S1NOD(I)
BOUNOD(NNZ + I) = S3NOD(I)
END DO
DO I = 1,NNX-2
S2NOD(I) = 2*NNZ + (I-1)*NNZ
S4NOD(I) = (NNZ+1) + (I-1)*NNZ
BOUNOD(2*NNZ + I) = S2NOD(I)
BOUNOD(2*NNZ + (NNX-2) + I) = S4NOD(I)
END DO
END
!------------------------------------------------------------------
!SUBROUTINE TO APPLY BOUNDARY CONDITIONS. CODE ADAPTED FROM J.N.
!REDDY,"AN INTRODUCTION TO THE FINITE ELEMENT METHOD",(1993),P.654
!------------------------------------------------------------------
190
SUBROUTINE APPLYBCSBW(FG, KG, HBW, NDOF, GDOF, NBN, NNX, NNZ,VAL1)
DO NB = 1,NBN
IE = (BOUNOD(NB) - 1)*NDOF + BOUDOF(NB)
VALUE = VAL1
IT = HBW - 1
I = IE - HBW
DO II = 1,IT
I = I + 1
IF (I >= 1 ) THEN
J = IE - I + 1
FG(I) = FG(I) - KG(I,J)*VALUE
KG(I,J) = 0.0
END IF
END DO
KG(IE,1) = 1.0
FG(IE) = VALUE
I = IE
DO II = 2,HBW
I = I + 1
END DO
END
C ________________________________________________________________
C
C The subroutine solves banded symmetric system of equations using
C gauss elimination. The banded matrix is input as BAND(NEQNS,NBW)
C and the right-hand side is input as RHS(NEQNS), where NBW is the
C half band width and NEQNS is the number of equations. If IRES is
C greater than zero, the right hand elimination is skipped.
C CODE FROM J.N. REDDY,"AN INTRODUCTION TO THE FINITE ELEMENT
C METHOD",(1993),P.676
C ________________________________________________________________
191
SUBROUTINE SOLVEBW(NRM,NCM,NEQNS,NBW,BAND,RHS,IRES)
IMPLICIT REAL*8(A-H,O-Z)
DIMENSION BAND(NRM,NCM),RHS(NRM)
COMMON/IO/IN,ITT
MEQNS=NEQNS-1
IF(IRES.LE.0) THEN
DO 30 NPIV=1,MEQNS
NPIVOT=NPIV+1
LSTSUB=NPIV+NBW-1
IF(LSTSUB.GT.NEQNS) THEN
LSTSUB=NEQNS
ENDIF
DO 20 NROW=NPIVOT,LSTSUB
C
C INVERT ROWS AND COLUMNS FOR ROW FACTOR
C
NCOL=NROW-NPIV+1
FACTOR=BAND(NPIV,NCOL)/BAND(NPIV,1)
DO 10 NCOL=NROW,LSTSUB
ICOL=NCOL-NROW+1
JCOL=NCOL-NPIV+1
10 BAND(NROW,ICOL)=BAND(NROW,ICOL)-FACTOR*BAND(NPIV,JCOL)
20 RHS(NROW)=RHS(NROW)-FACTOR*RHS(NPIV)
30 CONTINUE
ELSE
40 DO 60 NPIV=1,MEQNS
NPIVOT=NPIV+1
LSTSUB=NPIV+NBW-1
IF(LSTSUB.GT.NEQNS) THEN
LSTSUB=NEQNS
ENDIF
DO 50 NROW=NPIVOT,LSTSUB
NCOL=NROW-NPIV+1
FACTOR=BAND(NPIV,NCOL)/BAND(NPIV,1)
50 RHS(NROW)=RHS(NROW)-FACTOR*RHS(NPIV)
60 CONTINUE
ENDIF
C
C Back substitution
C
DO 90 IJK=2,NEQNS
NPIV=NEQNS-IJK+2
RHS(NPIV)=RHS(NPIV)/BAND(NPIV,1)
LSTSUB=NPIV-NBW+1
IF(LSTSUB.LT.1) THEN
LSTSUB=1
ENDIF
NPIVOT=NPIV-1
DO 80 JKI=LSTSUB,NPIVOT
NROW=NPIVOT-JKI+LSTSUB
NCOL=NPIV-NROW+1
FACTOR=BAND(NROW,NCOL)
80 RHS(NROW)=RHS(NROW)-FACTOR*RHS(NPIV)
90 CONTINUE
RHS(1)=RHS(1)/BAND(1,1)
RETURN
END
!------------------------------------------------------------------
192
!SUBROUTINE TO SOLVE THE NON-BANADED LINEAR SYSTEM DX = B USING
!GAUSSIAN ELIMINATION WITH PARTIAL PIVOTING - ALGORITHM 6.2 PAGE 340,
!R.L. BURDEN, J.D. FAIRES, NUMERICAL ANALYSIS, FIFTH EDITION, 1993
!------------------------------------------------------------------
SUBROUTINE SOLVE(D, B, N, X)
!Load D into A
DO I = 1,N
DO J = 1,N
A(I,J) = D(I,J)
END DO
END DO
!Find the maximum pivot value in the Ith column and the
!corresponding row number
DO J = I+1,N
IF ( DABS(A(J,I)) > DABS(AP) ) THEN
AP = A(J,I)
P = J
END IF
END DO
DO K = 1,N+1
A(J,K) = A(J,K) - M*A(I,K)
END DO
END DO
END DO
193
!Quit if denominator for calculating xn is zero
IF ( A(N,N) == 0 ) THEN
PRINT *, 'NO UNIQUE SOLUTION - XN DENOMINATOR = 0'
PRINT *, N
PRINT *, (A(I,N), I = 1,N)
GO TO 1000
END IF
!Back substitution
X(N) = A(N,N+1)/(1.0D0*A(N,N))
DO K = 1,N-1
I = N-K
SUM = 0
DO J = I+1,N
SUM = SUM + A(I,J)*X(J)
END DO
1000 END
!------------------------------------------------------------------
!SUBROUTINE TO CALCULATE MOMENT
!------------------------------------------------------------------
SUBROUTINE MOCALC(GDOF, EDOF, NEX, NEZ, NE, NNX, NNZ,
* NPE, L, PADANGLE, OFFSET, THP, TP,
* RJ, PRESS, MOMENT)
INTEGER GDOF, EDOF, NEX, NEZ, NE, NNX, NNZ, NPE, ELNODES(NPE),NG
PARAMETER (NG = 4)
REAL*8 XJ(EDOF), ZJ(EDOF), PJ(EDOF), MOMENT
REAL*8 XI(NG), ETA(NG), W(NG), P, PRESS(GDOF)
REAL*8 L, PADANGLE, OFFSET, THP, TP, RJ
REAL*8 PH(EDOF), DPHX(EDOF), DPHZ(EDOF), DETJA, TH
INTEGER I, J, EL
!Initialize moment
MOMENT = 0.0D0
194
!Calculate the nodal moment arms and pressures
DO I = 1,NPE
PJ(I) = PRESS(ELNODES(I))
END DO
DO I = 1,EDOF
P = P + PH(I)*PJ(I)
TH = TH + PH(I)*XJ(I)/RJ
END DO
!Assemble MOMENT
MOMENT = MOMENT + W(J)*DETJA*(RJ+TP)*P*
* (TH-THP)*DCOS(TH-THP)
END
!------------------------------------------------------------------
!SUBROUTINE TO CALCULATE THE STATIC LOAD FOR A PAD
!------------------------------------------------------------------
SUBROUTINE PLOADCALC(NE, NPE, EDOF, GDOF, NEX, NEZ, NNX, NNZ,
* RJ, THP, OFFSET, PADANGLE, L, PRESS,
* PLOADX, PLOADY)
INTEGER NE, EDOF, GDOF, NPE, NEX, NEZ, NNX, NNZ, ELNODES(NPE)
REAL*8 XJ(EDOF), ZJ(EDOF), PJ(EDOF), PLOADX, PLOADY, RJ
REAL*8 XI(EDOF), ETA(EDOF), W(EDOF), P, TH, THP
REAL*8 OFFSET, PADANGLE, L, PRESS(GDOF)
REAL*8 PH(EDOF), DPHX(EDOF), DPHZ(EDOF), DETJA
INTEGER I, M, EL
DO EL = 1,NE
!CALCULATE THE CONNECTIVITY VECTOR FOR THE ELEMENT
CALL CONNVEC(EL, NPE, NEX, NEZ, ELNODES)
195
!GET THE GLOBAL COORDINATES FOR EACH NODE IN THE CONNECTIVITY
VECTOR
DO I = 1,NPE
CALL NODCOORD(RJ, THP, NEX, NEZ, NNX, NNZ, OFFSET,
* PADANGLE, L, ELNODES(I), XJ(I), ZJ(I))
END DO
DO I = 1,EDOF
P = P + PH(I)*PJ(I)
TH = TH + PH(I)*XJ(I)/RJ
END DO
END DO
END DO
END
!------------------------------------------------------------------
!SUBROUTINE TO PERTURB THE JOURNAL FOR THE CALCULATION OF THE
!STIFFNESS AND DAMPING COEFFICIENTS
!------------------------------------------------------------------
SUBROUTINE PERTURB(PERTFLAG, PI, DECC, DV, EQECC, EQTHC,
* ECC, THC, VX, VY)
REAL*8 PI, ECC, THC, VX, VY, DVX, DVY, EQECC, EQTHC
REAL*8 ECCX, ECCY, DECCX, DECCY, DECC, DV
INTEGER PERTFLAG
196
!Perturb eccy -> Kyy, Kxy
CASE(2)
DECCY = DECC
ECCX = - EQECC*DCOS(EQTHC)
ECCY = DECCY - EQECC*DSIN(EQTHC)
ECC = DSQRT(ECCX**2 + ECCY**2)
THC = PI + DATAN2(ECCY,ECCX)
END SELECT
END
!------------------------------------------------------------------
!SUBROUTINE TO CALCULATE STIFFNESS AND DAMPING COEFFICIENTS
!------------------------------------------------------------------
SUBROUTINE SDCOEFF(MAXITER,GDOF, EDOF, NBN, NEX, NEZ, NE, NNX,
* NNZ, NDOF, HBW, NPE, NPADS, U, MU, P0, L, PADANGLE, M,OFFSET,
* THETAP, TP, TOL2, DXX, VX, LOAD, VY, VD, RJ, RP, CB, OMEGA,
* EQECC, EQTHC, EQSFX, EQSFY, RATIO, PI, KIJ, CIJ)
INTEGER GDOF, EDOF, NBN, NEX, NEZ, NE, NNX, NNZ, NDOF,
* HBW, NPE, NPADS, PERTFLAG, MAXITER
REAL*8 L, PADANGLE, M, OFFSET, THETAP(NPADS),
* TP, TOL2, DXX, LOAD, FX(NPADS), FY(NPADS), DEL(NPADS),
* RJ, EQECC, CB, PI, EQTHC, RP, U, MU, P0, MOMENT(NPADS),
* VX, VY, VD, EQSFX, EQSFY, SFX, SFY, OMEGA, RATIO,
* ECC, THC, DECC, DV, KIJ(4), CIJ(4)
DECC = RATIO*CB
DV = OMEGA*DECC
DO PERTFLAG = 1,4
VX = 0.0D0
VY = 0.0D0
VD = 0.0D0
CALL NEWTON(GDOF, EDOF, NBN, NEX, NEZ, NE, NNX, NNZ, NPADS,
* NDOF, HBW, NPE, NPADS,MAXITER,PERTFLAG,TOL2,DXX,
* L,PADANGLE,M, OFFSET, THETAP, TP, RJ, RP, CB,
* P0, VX, VY, VD, LOAD,
* U, MU, SFX, SFY,FX, FY, DEL, MOMENT,ECC, THC)
197
CASE(1)
KIJ(1) = (SFX-EQSFX)/DECC !Kxx
KIJ(2) = (SFY-EQSFY)/DECC !Kyx
CASE(2)
KIJ(3) = (SFX-EQSFX)/DECC !Kxy
KIJ(4) = (SFY-EQSFY)/DECC !Kyy
CASE(3)
CIJ(1) = (SFX-EQSFX)/DV !Cxx
CIJ(2) = (SFY-EQSFY)/DV !Cyx
CASE(4)
CIJ(3) = (SFX-EQSFX)/DV !Cxy
CIJ(4) = (SFY-EQSFY)/DV !Cyy
END SELECT
END DO
END
!------------------------------------------------------------------
!SUBROUTINE TO CALCULATE THE PAD PRESSURE DISTRIBUTION
!------------------------------------------------------------------
SUBROUTINE PRESSCALC(GDOF, EDOF, NBN, NEX, NEZ, NE, NNX, NNZ,
* NDOF, HBW, NPE, U, MU, P0, L, PADANGLE, M,OFFSET,
* THP, TP, VX, VY, VD, RJ, RP,ECC, CB, DELTA, THC, PRESS)
INTEGER GDOF, EDOF, NBN, NEX, NEZ, NE, NNX, NNZ, NDOF,
* HBW, NPE, ELNODES(NPE), I, J, EL
REAL*8 KE(EDOF,EDOF), FE(EDOF), KG(GDOF,HBW), FG(GDOF),PRESS(GDOF)
REAL*8 L, PADANGLE, M, OFFSET, THP, TP, RJ, ECC, CB, THC, RP
REAL*8 U, MU, P0, DELTA
REAL*8 XJ(NPE), ZJ(NPE), DHTHJ(NPE), HJ(NPE), VX, VY, VD
DO J = 1,HBW
KG(I,J) = 0
END DO
END DO
198
CALL FILMTH(DELTA, M,THP,TP,RJ,
* ECC,CB,THC, XJ(I), DHTHJ(I), HJ(I))
END DO
END DO
END
!------------------------------------------------------------------
!SUBROUTINE TO CALCULATE THE PAD TEMPERATURE DISTRIBUTION
!------------------------------------------------------------------
SUBROUTINE TEMPCALC(GDOF, NEX, NEZ, NNX, NNZ, U, MU, RHOL, CAPL,
* T0, HTC, TIN, TAMB, L, PADANGLE, M,OFFSET,
* THP, TP, RJ, ECC, CB, DELTA, THC, TEMP)
!Initialize
DO I = 1,NNX
TC(I) = TIN
END DO
199
DO I = 1,NNX-1
TH = THS + 1.0D0*I*DTH
X = RJ*TH
A = 2.0D0*HTC/(RHOL*CAPL*OMEGA*H)
B = A*TAMB + 2.0D0*MU*OMEGA*RJ**2.0/(RHOL*CAPL*H**2.0)
DO I = 1, NNX
DO J = I1,I2
END DO
END
!------------------------------------------------------------------
!SUBROUTINE TO CALCULATE XX - VECTOR OF TILT ANGLES AND JOURNAL POS.
!------------------------------------------------------------------
SUBROUTINE XXCALC(NPADS,PADANGLE,M,OFFSET,THETAP,TP,RJ,ECC,
* PERTFLAG, CB,THC, XXDOF, XX)
DO I = 1,NPADS
THP = THETAP(I)
CALL MAXDELTA(PADANGLE,M,OFFSET,THP,TP,RJ,ECC,CB,THC,DMAX(I))
XX(I) = 0.5*DMAX(I)
END DO
IF (PERTFLAG == 0) THEN
XX(XXDOF-1) = ECC
XX(XXDOF) = THC
END IF
200
END
!------------------------------------------------------------------
!SUBROUTINE TO USE NEWTON'S METHOD TO SOLVE A SYSTEM OF EQUATIONS
!------------------------------------------------------------------
SUBROUTINE NEWTON(GDOF, EDOF, NBN, NEX, NEZ, NE, NNX, NNZ, NPADS,
* NDOF, HBW, NPE, XXDOF,MAXITER,PERTFLAG,TOL,
DXX,L,
* PADANGLE,M, OFFSET, THETAP, TP, RJ, RP, CB,
* P0, VX, VY, VD, LOAD,
* U, MU, SFX, SFY,FX, FY, DEL, MOMENT,ECC, THC)
INTEGER GDOF, EDOF, NBN, NEX, NEZ, NE, NNX, NNZ, NPADS, NDOF,
* HBW, NPE, XXDOF, MAXITER
REAL*8 TOL, DXX, PRESS(GDOF)
REAL*8 L, PADANGLE, M, OFFSET, THP, THETAP(NPADS), TP
REAL*8 RJ, RP, ECC, CB, THC, LOAD
REAL*8 U, MU, P0, DELTA, DEL(NPADS)
REAL*8 VX, VY, VD
REAL*8 FX(NPADS), FY(NPADS), SFX,SFY,MOMENT(NPADS)
REAL*8 FXA(NPADS), FYA(NPADS), SFXA,SFYA,MOMENTA(NPADS)
REAL*8 XX(XXDOF), FF(XXDOF), XXA(XXDOF), FFA(XXDOF)
REAL*8 ECCA, THCA
REAL*8 JAC(XXDOF,XXDOF), YY(XXDOF), YYMAX
INTEGER NITER, PADNUM, CONFLAG, PERTFLAG
INTEGER I, J, K
!Initialize XX
CALL XXCALC(NPADS,PADANGLE,M,OFFSET,THETAP,TP,RJ,ECC,
* PERTFLAG, CB,THC, XXDOF, XX)
DO PADNUM = 1,NPADS
201
CALL MOCALC(GDOF, EDOF, NEX, NEZ, NE, NNX, NNZ,
* NPE, L, PADANGLE, OFFSET, THP, TP,
* RJ, PRESS, MOMENT(PADNUM))
END DO
!Calculate FF
DO PADNUM = 1,NPADS
FF(PADNUM) = MOMENT(PADNUM)
END DO
FF(XXDOF-1) = SFX
FF(XXDOF) = DABS(SFY)-LOAD
DO K = 1,XXDOF
XXA(K) = XX(K)
END DO
DO PADNUM = 1,NPADS
202
!Calculate the x,y force components ON each pad
CALL PLOADCALC(NE, NPE, EDOF, GDOF, NEX, NEZ,
* NNX, NNZ,RJ, THP, OFFSET, PADANGLE, L, PRESS,
* FXA(PADNUM), FYA(PADNUM))
END DO
!Calculate FFA
DO PADNUM = 1,NPADS
FFA(PADNUM) = MOMENTA(PADNUM)
END DO
FFA(XXDOF-1) = SFXA
FFA(XXDOF) = DABS(SFYA)-LOAD
203
!Initialize XX
CALL XXCALC(NPADS,PADANGLE,M,OFFSET,THETAP,TP,RJ,ECC,
* PERTFLAG, CB,THC, XXDOF, XX)
DO PADNUM = 1,NPADS
END DO
!Calculate FF
DO PADNUM = 1,NPADS
FF(PADNUM) = MOMENT(PADNUM)
END DO
DO K = 1,XXDOF
XXA(K) = XX(K)
END DO
DO PADNUM = 1,NPADS
204
!Calculate pad moments
CALL MOCALC(GDOF, EDOF, NEX, NEZ, NE, NNX, NNZ,
* NPE, L, PADANGLE, OFFSET, THP, TP,
* RJ, PRESS, MOMENTA(PADNUM))
END DO
!Calculate FFA
DO PADNUM = 1,NPADS
FFA(PADNUM) = MOMENTA(PADNUM)
END DO
ELSE
DO PADNUM = 1,NPADS
205
* NDOF, HBW, NPE, U, MU, P0, L, PADANGLE, M,OFFSET,
* THP, TP, VX, VY, VD, RJ, RP,ECC, CB, DELTA, THC, PRESS)
END DO
END IF
END
!*************************************************************************
206
Appendix C: Matlab Graphics Program for
%***********************************************************************
%THIS PROGRAM IS THE GRAPHICS FILE FOR THE MORTON_TPJB.FOR FORTAN
%PROGRAM.
%***********************************************************************
%***********************************************************************
% EXTRACT DATA FROM INPUT FILE
%***********************************************************************
% Remove all prior definitions from MATLAB's memory
clear all
close all
dat = load(sprintf('%s',fname));
ldat = length(dat);
207
for i = 1:NS
i2 = (i-1)*(NAVG+2) + 1;
i3 = i*(NAVG+2);
%units = m, rad
xamp(i) = dat(i2,1);
xph(i) = dat(i2,2);
yamp(i) = dat(i2,3);
yph(i) = dat(i2,4);
xjeq(i) = dat(i2,5);
yjeq(i) = dat(i2,6);
speed(i) = dat(i3,1);
Uthr(i) = dat(i3,2);
U(i) = dat(i3,3);
DT(i) = dat(i3,4);
phi(i) = dat(i3,5);
eps0(i) = dat(i3,6);
for j = 1:NAVG
j2 = i2 + j;
delT(j,i) = dat(j2,1);
end
end
omega = speed*2*pi/60;
%***********************************************************************
% PROGRAM TO PLOT TEMPERATURE, PRESSURE AND FILM THICKNESS FOR LAST SPD.
%***********************************************************************
% Calculate index for centerline temperatures
if mod(N,2) == 0
ic = N/2;
else
ic = (N+1)/2;
end
switch pn
case 1,
fn = 'PAD1';
case 2,
fn = 'PAD2';
case 3,
fn = 'PAD3';
case 4,
fn = 'PAD4';
case 5,
fn = 'PAD5';
case 6,
fn = 'PAD6';
case 7,
fn = 'PAD7';
case 8,
fn = 'PAD8';
case 9,
fn = 'PAD9';
208
case 10,
fn = 'PAD10';
end
fid = fopen(fn);
A = fscanf( fid, '%g %g %g %g %g', [5 inf]);
A = A';
fclose(fid);
for j = 1:M
r = 1 + (j-1)*N;
r2 = ic + (j-1)*N;
X(j) = A(r,1); % theta (degrees)
TC(j) = A(r2,3); % centerline temperature (deg F)
end
for i = 1:N
Y(i) = A(i,2); % z (inches)
end
for j = 1:M
for i = 1:N
r2 = i + (j-1)*N;
Z1(i,j) = A(r2,3); % temperature (deg F)
Z2(i,j) = A(r2 ,4); % pressure (psi)
Z3(i,j) = A(r2,5); % film thickness (mils)
end
end
figure(1)
subplot(3,1,1)
surf(X, Y, Z1)
view(-10,40)
title(sprintf('%s%s %6.0f %s', ti,': Speed =',...
speed(NS),'RPM'))
text(xt(pn),yt,zt,fn)
zlabel('Temperature (F)')
grid on
hold on
subplot(3,1,2)
surf(X, Y, Z2)
view(-10,40)
zlabel('Pressure (psi)')
grid on
hold on
subplot(3,1,3)
surf(X, Y, Z3)
view(-10,40)
xlabel('Theta (deg.)')
ylabel('Z (in.)')
zlabel('h (mils)')
grid on
hold on
end
209
%***********************************************************************
%PROGRAM TO GENERATE ORBIT PLOTS
%***********************************************************************
%Do calculations in speed loop
for i = 1:NS
%angular speed
w = omega(i);
%orbit parameters
Ax = xamp(i); Ay = yamp(i);
phx = xph(i); phy = yph(i);
%plot orbits
figure(1+i)
orbplt(cb,delT(i2,i),t,w,sf,Rj,Ax,Ay,phx,phy,xje,yje);
end
figure(1+i)
title(sprintf('%s%s %7.0f %s %5.2f %s',ti,':',speed(i),...
' RPM, Avg DT = ',abs(DT(i)),' deg F'))
xlabel('x/Cb')
ylabel('y/Cb')
hold on
end
%***********************************************************************
%PROGRAM TO GENERATE UNBALANCE PLOTS AND CALCULATE THRESHOLD SPEEDS
%***********************************************************************
%Plot unbalance curves
figure(NS+2)
sp2 = min(speed):0.001*(max(speed)-min(speed)):max(speed);
om2 = sp2*2*pi/60;
krpm = 1e-3*sp2;
unb = spline(speed,U,sp2);
unbthr2 = (thrfac*W./(om2.^2))*2.2046*16/0.0254;
j = 1;
for i = 1:length(sp2)-1
diff1 = unb(i)-unbthr2(i);
diff2 = unb(i+1)-unbthr2(i+1);
if diff1*diff2 < 0.00 | diff1 == 0.00
Nthr(j) = sp2(i);
Uval(j) = unbthr2(i);
j = j + 1;
end
end
210
plot(krpm,unb,'b--'), hold on
%Label plot
legend('U','Uthr')
title(ti)
xlabel('Speed (krpm)')
ylabel('Unbalance (oz in)')
for i = 1:NS
disp(sprintf('%7.0f %7.2f %7.2f %7.1f
%7.1f',speed(i),U(i),Uthr(i),phi(i),abs(DT(i))))
end
%***********************************************************************
%***********************************************************************
% orbplt.m = function used by MORTON_TPJB. This function plots the
%total synchronous orbit and appropriately scales the journal radius to
%show the locations of the cold and the hot spots at different points in
%the orbit.
%***********************************************************************
function [] = orbplt(cb,dT,t,w,sf,Rj,Ax,Ay,phx,phy,xje,yje)
Rj = sf*Rj/cb;
211
Ax = Ax/cb; Ay = Ay/cb;
xje = xje/cb; yje = yje/cb;
g0 = atan2(Ay*sin(phy),Ax*cos(phx));
th = 0:0.01*pi:2*pi;
t1 = 0:0.01*2*pi/w:2*pi/w;
xp = xje + Ax*cos(w*t1 + phx);
yp = yje + Ay*sin(w*t1 + phy);
plot(xp,yp,'k-')
axis equal
hold on
xj = xoj + Rj*cos(th);
yj = yoj + Rj*sin(th);
hold on
%***********************************************************************
212
References
Abdel-Hadi, E.A.A., T.R. Hsu and K.S. Bhatia, “Upwind Finite Element Analysis of
Advection-Diffusion Equation”, Numerical Methods in Thermal Problems, Vol. 5, Pine
Ridge Press, Swansea England, 1985, pp. 756-770.
Allaire, P.E., J.K. Parsell and L.E. Barrett, “A Pad Perturbation Method for the Dynamic
Coefficients of Tiliting-Pad Journal Bearings”, Wear, 72, 1981, pp. 29-44.
de Jongh, F.M. and P.G. Morton, “The Synchronous Instability of a Compressor Rotor
due to Bearing Journal Differential Heating”, ASME Paper 94-GT-35.
de Jongh, F.M. and P. van der Hoeven, “Application of a Heat Barrier Sleeve to Prevent
Synchronous Rotor Instability”, Proceedings of the 27th Turbomachinery Symposium,
Turbomachinery Laboratory, Texas A&M University, College Station, Texas, 1998.
Ettles, C.M., “The Analysis of Pivoted Pad Journal Bearing Assemblies Considering
Thermoelastic Deformation and Heat Transfer Effects”, Tribology Transactions, Volume
35, No.1, 1992, pp.156-162.
Faulkner, H.B., W.F. Strong and R.G. Kirk, “Thermally Induced Synchronous Instability
of a Radial Inflow Overhung Turbine, Part II”, Proceedings of ASME Design Engineering
Technical Conferences, September 1997, Sacramento, California, DETC97/VIB-4174.
Goldman, P., A. Muszynska and D.E. Bently, “Thermal Bending of the Rotor due to
Rotor-to-Stator Rub”, ISROMAC-7, Honululu, Hawaii, February 1998.
213
Ha, H.C., H.J. Kim, K.W. Kim, “Inlet Pressure Effects on the Thermohydrodynamic
Performance of a Large Tilting Pad Journal Bearing”, Journal of Tribology, Vol. 117,
January 1995, pp.160-165.
Kellenberger, W., “Spiral Vibrations due to the Seal Rings in Turbogenerators: Thermally-
Induced Interaction between Rotor and Stator”, Journal of Mechanical Design, Vol. 102,
January 1980, pp. 177-184.
Keogh, P.S., and P.G. Morton, “The Dynamic Nature of Rotor Thermal Bending due to
Unsteady Lubricant Shearing within a Bearing”, Proceedings of the Royal Society of
London, Series A, Vol. 445, 1994, pp. 273-290.
Kim, J., A.B. Palazzolo, R.K. Gadangi, “TEHD Analysis for Tilting-Pad Journal Bearings
using Upwind Finite Element Method”, Tribology Transactions, Volume 37, 1994, No. 4,
pp. 771-783.
Kirk, R.G., Course Notes for ME 5504 - Introduction to Rotordynamics Analysis, Fall
1998.
Kroon, R.P. and W.A. Williams, “Spiral Vibration of Rotating Machinery”, Proceedings of
5th International Congress of Applied Mechanics, Wiley, New York, 1939, p. 712.
Larsson, B., “Journal Asymmetric Heating - Part II: Alteration of Rotor Dynamic
Properties”, Journal of Tribology, Vol. 121, January 1999, pp. 164-168.
Newkirk, B. L., “Shaft Rubbing”, Mechanical Engineering, Vol. 48, No. 8, August 1926,
pp. 830-832.
Nicholas, J.C., E.J. Gunter and P.E. Allaire, “Stiffness and Damping Coefficients for the
Five-Pad Tilting-Pad Bearing”, ASLE Transactions, Volume 22, 2, 1977, pp. 113-124.
214
Pinkus, O., Thermal Aspects of Fluid Film Tribology, 1990.
Schmied, J., “Spiral Vibrations of Rotors”, Rotating Machinery Dynamics, Vol. 2, 1987,
pp. 449-456.
Wang, X., “The Influence of a Skewed Disk on a Flexible Rotating Shaft”, M.S. Thesis in
Engineering Mechanics, Virginia Tech, December 1997.
215
Vita
Avinash Chetnand Balbahadur was born on May 29, 1973 in Georgetown, Guyana. He
graduated from Queen’s College Secondary School (Guyana) in 1991 and subsequently
M.S. degree in Mechanical Engineering at the same college, he enrolled in the Mechanical
Engineering Ph.D. program at Virginia Tech in 1997. His doctoral research has lead to
A.C. Balbahadur and R.G. Kirk, “A Model for Thermal Bending in Overhung Rotors”,
Proceedings of the ASME Design Engineering Technical Conference, 17th Biennial
Conference on Mechanical Vibration and Noise, September 12-15, 1999, Las Vegas,
Nevada, DETC99/VIB-8298.
R.G. Kirk and A.C. Balbahadur, “Thermal Distortion Synchronous Rotor Instability”,
IMechE 2000, C576/041/2000.
216